Modelling
of Welding
SECOND EDITION
0YSTEIN GRONG
Norwegian University of
Science and Technology,
Department of Metallurgy,
N7034 Trondheim, Norway
Editor: H. K. D. H. Bhadeshia
The University of Cambridge
Department of Materials Science
and Metallurgy
T H E INSTITUTE OF MATERIALS
Book 677
First published in 1997 by
The Institute of Materials
1 Carlton House Terrace
London SWlY 5DB
First edition (Book 557)
Published in 1994
The Institute of Materials 1997
All rights reserved
ISBNl 86125 036 3
Originally typeset by
PicA Publishing Services
Additional typesetting and corrections by
Fakenham Photosetting Ltd
Printed and bound in the UK at
The University Press, Cambridge
Besides correcting some minor linguistic and print errors, I have in the second edition included a collection of different exercise problems which have been used in the training of students at NTNU. They illustrate how the models described in the previous chapters can be used
to solve practical problems of more interdisciplinary nature. Each of them contains a 'problem description' and some background information on materials and welding conditions. The
exercises are designed to illuminate the microstructural connections throughout the weld
thermal cycle and show how the properties achieved depend on the operating conditions applied. Solutions to the problems are also presented. These are not complete or exhaustive, but
are just meant as an aid to the reader to develop the ideas further.
Trondheim, 28 October, 1996
0ystein Grong
The purpose of this textbook is to present a broad overview on the fundamentals of welding
metallurgy to graduate students, investigators and engineers who already have a good background in physical metallurgy and materials science. However, in contrast to previous textbooks covering the same field, the present book takes a more direct theoretical approach to
welding metallurgy based on a synthesis of knowledge from diverse disciplines. The motivation for this work has largely been provided by the need for improved physical models for
process optimalisation and microstructure control in the light of the recent advances that have
taken place within the field of materials processing and alloy design.
The present textbook describes a novel approach to the modelling of dynamic processes in
welding metallurgy, not previously dealt with. In particular, attempts have been made to rationalise chemical, structural and mechanical changes in weldments in terms of models based
on well established concepts from ladle refining, casting, rolling and heat treatment of steels
and aluminium alloys. The judicious construction of the constitutive equations makes full use
of both dimensionless parameters and calibration techniques to eliminate poorly known kinetic constants. Many of the models presented are thus generic in the sense that they can be
generalised to a wide range of materials and processing. To help the reader understand and
apply the subjects and models treated, numerous example problems, exercise problems and
case studies have been worked out and integrated in the text. These are meant to illustrate the
basic physical principles that underline the experimental observations and to provide a way of
developing the ideas further.
Over the years, I have benefited from interaction and collaboration with numerous people
within the scientific community. In particular, I would like to acknowledge the contribution
from my father Professor Tor Grong who is partly responsible for my professional upbringing
and development as a metallurgist through his positive influence on and interest in my research work. Secondly, I am very grateful to the late Professor Nils Christensen who first
introduced me to the fascinating field of welding metallurgy and later taught me the basic
principles of scientific work and reasoning. I will also take this opportunity to thank all my
friends and colleagues at the Norwegian Institute of Technology (Norway), The Colorado School
of Mines (USA), the University of Cambridge (England), and the Universitat der Bundeswehr
Hamburg (Germany) whom I have worked with over the past decade. Of this group of people,
I would particularly like to mention two names, i.e. our department secretary Mrs. Reidun
0stbye who has helped me to convert my original manuscript into a readable text and Mr.
Roald Skjaerv0 who is responsible for all linedrawings in this textbook. Their contributions
are gratefully acknowledged.
Trondheim, 1 December, 1993
0ystein Grong
Contents
xiii
xiv
1.1
Introduction ...............................................................................
1.2
1.3
1.4
1.5
1.6
10
1.7
14
1.8
18
1.9
24
24
26
26
28
31
41
45
48
49
vi
Contents
vii
56
59
77
77
80
91
95
61
72
75
96
99
100
References .........................................................................................
103
105
Appendix 1.2: Refined Heat Flow Model for Spot Welding ..............
110
111
112
96
Introduction ...............................................................................
116
2.2
116
2.3
117
2.4
120
2.4.1
120
2.4.2
121
viii
Contents
2.5
122
2.6
123
2.6.1
123
2.6.2
124
124
128
2.8.1
128
2.8.2
128
2.8.3
130
2.8.4
131
131
134
138
140
140
141
141
2.9.1
142
2.9.2
142
2.9.3
143
2.9.4
146
148
148
2.7
2.8
2.9
149
149
152
156
160
Contents
ix
166
166
169
170
171
172
173
174
176
177
179
180
182
184
185
186
188
190
190
192
192
193
195
196
196
201
202
202
202
203
204
Contents
2.12.4 Prediction of Inclusion Composition .............................
2.12.4.1 CMn Steel Weld Metals ..............................
2.12.4.2 Lowalloy Steel Weld Metals ........................
204
204
206
References .........................................................................................
212
215
219
Introduction ...............................................................................
221
3.2
221
3.3
222
3.3.1
225
3.3.2
226
228
3.4.1
228
3.4.2
229
3.4.3
230
230
234
3.4.4
239
240
242
251
3.5.1
251
3.5.2
254
254
256
3.5.3
260
260
261
264
3.4
3.5
Contents
3.6
3.7
3.8
xi
268
3.6.1
268
3.6.2
272
272
3.7.1
Microsegregation .........................................................
272
3.7.2
Macrosegregation ........................................................
278
3.7.3
279
279
281
284
3.7.4
286
286
286
290
3.8.1
290
3.8.2
290
290
292
292
References .........................................................................................
293
296
Introduction ...............................................................................
301
4.2
301
4.2.1
301
4.2.2
303
4.2.3
304
304
308
xii
Contents
4.3
314
4.3.1
314
4.3.2
314
315
315
316
4.4.1
316
319
325
326
328
References .........................................................................................
334
334
4.4
4.4.2
322
329
332
Introduction ...............................................................................
337
5.2
337
5.2.1
337
5.2.2
339
5.2.3
340
5.2.4
341
5.2.5
342
343
5.3.1
343
5.3.2
345
5.3
Contents
xiii
345
345
360
5.4.1
360
360
361
361
363
5.4.2
364
364
367
370
372
375
380
5.3.3
5.4
5.5
5.5.1
347
348
351
356
380
382
References .........................................................................................
382
384
5.5.2
Introduction ...............................................................................
387
6.2
387
6.2.1
387
xiv
Contents
6.3
6.2.2
389
389
390
392
6.2.3
396
396
397
6.2.4
400
400
402
6.2.5
402
403
404
406
6.3.1
406
6.3.2
406
6.3.3
408
408
408
422
6.3.4
427
6.3.5
428
428
429
430
444
6.3.6
404
405
432
6.4
6.5
Contents
xv
6.3.7
Bainite ..........................................................................
6.3.7.1 Upper Bainite ...............................................
6.3.7.2 Lower Bainite ...............................................
444
444
447
6.3.8
Martensite ....................................................................
6.3.8.1 Lath Martensite ............................................
6.3.8.2 Plate (Twinned) Martensite ..........................
448
448
448
453
6.4.1
456
6.4.2
456
458
6.5.1
459
6.5.2
459
461
464
References .........................................................................................
467
471
475
Introduction ...............................................................................
477
7.2
477
7.2.1
477
478
480
485
486
xvi
Contents
7.2.1.5
491
7.2.2
494
495
500
502
7.2.3
509
509
513
514
514
524
524
525
527
7.3.1
527
7.3.2
529
7.3.3
530
7.3.4
532
536
7.4.1
536
7.4.2
540
7.2.4
7.3
7.4
7.4.3
518
541
542
544
547
Contents
7.4.3.1
7.4.3.2
7.4.3.3
7.4.3.4
7.4.4
xvii
547
548
548
550
556
556
557
558
558
560
References .........................................................................................
564
567
Introduction ...............................................................................
571
8.2
571
8.3
583
8.4
587
1
Heat Flow and Temperature Distribution in
Welding
1.1 Introduction
Welding metallurgy is concerned with the application of wellknown metallurgical principles
for assessment of chemical and physical reactions occurring during welding. On purely practical grounds it is nevertheless convenient to consider welding metallurgy as a profession of its
own because of the characteristic nonisothermal nature of the process. In welding the reactions are forced to take place within seconds in a small volume of metal where the thermal
conditions are highly different from those prevailing in production, refining and fabrication of
metals and alloys. For example, steel welding is characterised by:
High peak temperatures, up to several thousand 0 C.
High temperature gradients, locally of the order of 103 0C mm"1.
Rapid temperature fluctuations, locally of the order of 103 0C s 1 .
It follows that a quantitative analysis of metallurgical reactions in welding requires detailed
information about the weld thermal history. From a practical point of view the analytical
approach to the solution of heat flow problems in welding is preferable, since this makes it
possible to derive relatively simple equations which provide the required background for an
understanding of the temperaturetime pattern. However, because of the complexity of the
heat flow phenomena, it is always necessary to check the validity of such predictions against
more reliable data obtained from numerical calculations and in situ thermocouple measurements. Although the analytical models suffer from a number of simplifying assumptions, it is
obvious that these solutions in many cases are sufficiently accurate to provide at least a qualitative description of the weld thermal programme.
An important aspect of the present treatment is the use of different dimensionless groups
for a general outline of the temperature distribution in welding. Although this practice involves several problems, it is a convenient way to reduce the total number of variables to an
acceptable level and hence, condense general information about the weld thermal programme
into twodimensional (2D) maps or diagrams. Consequently, readers who are unfamiliar with
the concept should accept the challenge and try to overcome the barrier associated with the use
of such dimensionless groups in heat flow analyses.
Since heat losses from free surfaces by radiation and convection are usually negligible in
welding, the temperature distribution can generally be obtained from the fundamental differential equations for heat conduction in solids. For uniaxial heat conduction, the governing
equation can be written as:1
(iD
where T is the temperature, t is the time, x is the heat flow direction, and a is the thermal
diffusivity. The thermal diffusivity is related to the thermal conductivity X and the volume
heat capacity pc through the following equation:
(12)
For biaxial and triaxial heat conduction we may write by analogy:1
d3)
and
(14)
The above equations must clearly be satisfied by all solutions of heat conduction problems,
but for a given set of initial and boundary conditions there will be one and only one solution.
Carbon steel
Temperature, 0C
Fig. 1.1. Enthalpy increment H7H0
24.
Table 1.1 Physical properties for some metals and alloys. Data from Refs 2  6 .
Material
(WrTIm10C1)
(mm2 s"1)
(Jmnr 3 0C"1)
(0C)
(J mnr 3 )
(J mnr 3 )
Carbon
Steels
0.040
0.005
1520
7.50
2.0
Low Alloy
Steels
0.025
0.005
1520
7.50
2.0
High Alloy
Steels
0.020
0.005
1500
7.40
2.0
Titanium
Alloys
0.030
10
0.003
1650
4.89
1.4
Aluminium
(> 99% Al)
0.230
85
0.0027
660
1.73
0.8
AlMgSi
Alloys
0.167
62
0.0027
652
1.71
0.8
AlMg
Alloys
0.149
55
0.0027
650
1.70
0.8
X9 W/mm 0C
(a)
Temperature, 0C
(b)
X, W/mm 0C
Temperature, 0C
Fig. 1.2. Factors affecting the thermal conductivity X of steels; (a) Temperature level and chemical
composition, (b) Heat treatment procedure. Data from Refs. 24.
where Q is the net heat input (energy) released at time t = O, and A is the cross section of the
rod.
(ii)
Equations (15), (16) and (17) provide the required basis for a comprehensive theoretical
treatment of heat flow phenomena in welding. These solutions can either be applied directly
or be used in an integral or differential form. In the next sections a few examples will be given
to illustrate the direct application of the instantaneous heat source concept to problems related
to welding.
(a)
Fig. 1.3. Schematic representation of instantaneous heat source models; (a) Plane source in a long rod.
(b)
R
(C)
Fig. 1.3.Schematic representation of instantaneous heat source models (continued); (b) Line source in a
wide plate, (c) Point source in a heavy slab.
The model considers a point source on a heavy slab as illustrated in Fig. 1.4. The heat is
assumed to be released instantaneously at time t = 0 on the surface of the slab. This causes a
temperature rise in the material which is exactly twice as large as that calculated from equation
(17):
(18)
In order to obtain a general survey of the thermal programme, it is convenient to write
equation (18) in a dimensionless form. The following parameters are defined for this purpose:
Dimensionless temperature:
(19)
where Tc is the chosen reference temperature.
Dimensionless time:
d10)
where tt is the arc ignition time.
Dimensionless operating parameter:
(111)
where qo is the net arc power (equal to Qlt(), and (HcHo) is the heat content per unit volume at
the reference temperature.
Dimensionless radius vector:
(112)
By substituting these parameters into equation (18), we obtain:
(113)
Heat source
Isotherms
0Zn1
e/n
T1
^i
Fig. 1.5. Calculated temperatures in arc strikes.
Equation (113) has been solved numerically for different values ofCT1and T1. The results
are presented graphically in Fig. 1.5. Due to the inherent assumption of instantaneous release
of heat in a point, it is not possible to use equation (113) down to very small values OfCT1 and
T1. However, at some distance from the heat source and after a time not much shorter than the
real (assumed) time of heating, the calculated temperaturetime pattern will be reasonably
correct. Note that the heavy broken line in Fig. 1.5 represents the locus of the peak temperatures. This locus is obtained by setting 3In(OAi1VdT1 = 0:
from which
(114)
where Qp is the peak temperature, and e is the natural logarithm base number.
Example (1.1)
Consider a small weld crater formed in an arc strike on a thick plate of low alloy steel. Calculate the cooling time from 800 to 5000C (Af875), and the total width of the fully transformed
region adjacent to the fusion boundary. The operational conditions are as follows:
where r is the arc efficiency factor. Relevant thermal data for low alloy steel are given in
Table 1.1.
Solution
In the present case it is convenient to use the melting point of the steel as a reference temperature (i.e. 0 = 0m = 1 when Tc = TJ. The corresponding values OfZi1 and 9 (at 800 and 5000C,
respectively) are:
Since the cooling curves in Fig. 1.5 are virtually parallel at temperatures below 800 0 C, Af875
will be independent of Cr1 and similar to that calculated for the centreline ((J1 = 0). By rearranging equation (113) we get:
and
and
Alternatively, the same information could have been read from Fig. 1.5. Although it is
difficult to check the accuracy of these predictions, the calculated values for Ats/5 and ARlm are
considered reasonably correct. Thus, because the cooling rate is very large, in arc strikes a
hard martensitic microstructure would be expected to form within the transformed parts of the
HAZ, in agreement with general experience.
The model considers a line source which penetrates two overlapping plates of similar thermal
properties, as illustrated in Fig. 1.7. The heat is assumed to be released instantaneously at time
Heat source
Electrode
Heat source
Fig. 1.7. Idealised heat flow model for spot welding of plates.
t = 0. If transfer of heat into the electrodes is neglected, the temperature distribution is given
by equation (16).
This equation can be written in a dimensionless form by introducing the following group of
parameters:
Dimensionless time:
(115)
where th is the heating time (i.e. the duration of the pulse).
Dimensionless operating parameter:
(116)
where dt is the total thickness of the joint.
Dimensionless radius vector:
(117)
By substituting these parameters into equation (16), we get:
(118)
where 6 denotes the dimensionless temperature (previously defined in equation (19)).
6/n2
e/n2
T2
Fig. 1.8. Calculated temperaturetime pattern in spot welding.
Figure 1.8 shows a graphical representation of equation (118) for a limited range of a 2 and
T2. A closer inspection of the graph reveals that the temperaturetime pattern in spot welding
is similar to that observed during arc ignition (see Fig. 1.5). The locus of the peak temperatures in Fig. 1.8 is obtained by setting d\n{^ln7}ldx2  0.
which gives
and
(119)
Example (1.2)
Consider spot welding of 2 mm plates of low alloy steel under the following operational conditions:
Calculate the cooling time from 800 to 5000C (Af8/5) in the centre of the weld, and the cooling
rate (CR.) at the onset of the austenite to ferrite transformation. Assume in these calculations
that the total voltage drop between the electrodes is 1.6 V. The M^temperature of the steel is
taken equal to 475C.
Solution
If we use the melting point of the steel as a reference temperature, the parameters n2 and 6 (at
800 and 5000C, respectively) become:
The parameter A%5 can be calculated from equation (118). For the weld centreline (CT2 = 0),
we get:
and
The cooling rate at a specific temperature is obtained by differentiation of equation (118) with
respect to time. When (J2 = 0 the cooling rate at 9 = 0.3 (475C) becomes:
and
Since the cooling curves in Fig. 1.8 are virtually parallel at temperatures below 8000C (i.e.
for QZn2 < 0.15), the computed values of Ar8/5 and CR. are also valid for positions outside the
weld centreline. In the present example the centreline solutions can be applied down to
("2m)2 ~ 2. According to equation (119), this corresponds to a lower peak temperature of:
It should be emphasised that the present heat flow model represents a crude oversimplification of the spot welding process. In a real welding situation, most of the heat is generated at the
interface between the two plates because of the large contact resistance. This gives rise to the
development of an elliptical weld nugget inside the joint as shown in Fig. 1.9. Moreover, since
the model neglects transfer of heat into the electrodes, the mode of heat flow will be mixed and
not truly twodimensional as assumed above. Consequently, equation (118) cannot be applied for reliable predictions of isothermal contours and zone widths. Nevertheless, the model
may provide useful information about the cooling conditions during spot welding if the efficiency factor if] and the voltage drop between the electrodes can be estimated with a reasonable
degree of accuracy.
A more refined heat flow model for spot welding is presented in Appendix 1.2.
In thermit welding the time interval between the ignition of the powder mixture and the completion of the reduction process will be short because of the high reaction rates involved.
Assume that a groove of width 2L1 is filled instantaneously at time t = 0 by liquid metal of an
initial temperature Tt (see Fig. 1.10). The metal temperature outside the fusion zone is T0. If
heat losses to the surroundings are neglected, the problem can be treated as uniaxial conduction where the heat source (extending from L 1 to +L1) is represented by a series of elementary
sources, each with a heat content of:
(120)
At time t this source produces a small rise of temperature at position JC, given by equation (1 5):
(121)
The final temperature distribution is obtained by substituting u = (xxy(4at)m (i.e. dx' du(4at)m) into equation (121) and integrating between the limits JC'= L 1 and x' +L1. This
gives (after some manipulation):
(122)
Isl'srau*'*'=]
Fusion
zone
Fig. 1.9. Calculated peak temperature contours in spot welding of steel plates (numerical solution). Operational conditions: / = 23kA, 64 cycles. Data from Bently et al1
Fusion
zone
Fig. 1.10. Idealised heat flow model for thermit welding of rails.
where erf(u) is the Gaussian error function. The error function is defined in Appendix 1.3*.
Because of the complex nature of equation (122), it is convenient to present the different
solutions in a dimensionless form by introducing the following groups of parameters:
*The error function is available in tables. However, in numerical calculations it is more convenient to use the
Fortran subroutine given in Appendix 1.3.
Dimensionless temperature:
(123)
Dimensionless time:
(124)
Dimensionless jccoordinate:
(125)
Substituting these parameters into equation (122) gives:
(126)
Equation (126) has been solved numerically for different values of Q and T3. The results
are presented graphically in Fig. 1.11. As would be expected, the fusion zone itself (Q < 1)
cools in a monotonic manner, while the temperature in positions outside the fusion boundary
(Q > 1) will pass through a maximum before cooling. The locus of the HAZ peak temperatures
in Fig. 1.11 is defined by 3673T3 = 0. Referring to Appendix 1.3, we may write:
which gives
(127)
The peak temperature distribution is obtained by solving equation (127) for different combinations of Qm and T3m and inserting the roots into equation (126).
Example (1.3)
Consider thermit welding of steel rails (i.e. reduction of Fe2O3 with Al powder) under the
following operational conditions:
Calculate the cooling time from 800 to 5000C in the centre of the weld, and the total width
of the fully transformed region adjacent to the fusion boundary. The Ac3temperature of the
steel is taken equal to 8900C.
91
Definition of parameters:
For positions along the weld centreline (Q. = 0) equation (126) reduces to:
From the above relation it is possible to calculate the cooling time from Tt = 22000C to 800 and
5000C, respectively:
and
The computed value for A/8/5 is also valid for positions outside the weld centreline, since
the cooling curves at such low temperatures are reasonably parallel within the fusion zone.
Total width of fully transformed region
The fusion boundary is defined by:
The locus of the 8900C isotherm in temperaturetime space can be read from Fig. 1.11.
Taking the ordinate equal to 0.40, we get:
By inserting this value into equation (127), we obtain the corresponding coordinate of the
isotherm:
Unfortunately, measurements are not available to check the accuracy of these predictions.
Systematic errors would be expected, however, because of the assumption of instantaneous
release of heat immediately after powder ignition and the neglect of heat losses to the surroundings. Nevertheless, the present example is a good illustration of the versatility of the
concept of instantaneous heat sources, since these solutions can easily be added in space as
shown here or in time for continuous heat sources (to be discussed below).
The model considers a continuous (plane) heat source in a long rod as shown in Fig. 1.12(a).
The heat is liberated at a constant rate q'o in the plane x = 0 starting at time / = 0. If we
subdivide the time t during which the source operates into a series of infinitesimal elements dt/
(Fig. 1.12b), each element will have a heat content of:
(128)
(a)
Continuous heat source
(b) q
t
Fig. 1.12. Idealised heatflowmodel for friction welding of rods; (a) Sketch of model, (b) Subdivision of
time into a series of infinitesimal elements dt'.
At time / this heat will cause a small rise of temperature in the material, in correspondance
with equation (15):
(129)
If we substitute t"=t1'into equation (1 29), the total temperature rise at time t is obtained
by integrating from t"= t (t'= 0) to /"= 0 (t'= t):
(130)
In order to evaluate this integral, we will make use of the following mathematical transformation:
where
and
The latter integral can be expressed in terms of the complementary error function* erfc{u)
by substituting:
(131)
If the temperature of the contact section at the end of the heating period is taken equal to Th,
equation (131) can be rewritten as:
(132)
where t'h denotes the duration of the heating period (t < t'h). Measured contact section temperatures for different metal/alloy combinations are given in Table 1.2.
Equation (132) may be presented in a dimensionless form by the use of the following
groups of parameters:
Dimensionless temperature:
(133)
Dimensionless time:
(134)
The complementary error function is defined in Appendix 1.3.
Table 1.2 Measured contact section temperatures during friction welding of some metals and alloys.
Data from Tensi et al.10
Metal/Alloy
Combination
Measuring
Method
Temperature Level
[0C]
Partial
Melting
Steel
Thermocouples
10801340
No
12601400
No/Yes
1080
No
Direct readings
548
Yes
CopperNickel
Direct readings
1083
Yes
AlCu2Mg
Thermocouples
506
Yes
Al4.3Cu
Thermocouples
562
Yes
Al12Si
Thermocouples
575
Yes
Al5Mg
Thermocouples
582
Yes
SteelNickel
Direct readings
SteelTitanium
Direct readings1
CopperAl
Based on direct readings of the voltage drop between the two workpieces.
Dimensionless .^coordinate:
(135)
Equation (136) describes the temperature in different positions from the weld contact section during the heating period. However, when the rotation stops, the weld will be subjected to
free cooling, since there is no generation of heat at the interface. As shown in Fig. 1.13(a) this
can be accounted for by introducing an imaginary heat source of power +qo at time t = t'h
which acts simultaneously with an imaginary heat sink of negative power q o. It follows from
the principles of superposition (see Fig. 1.13b) that the temperature during the cooling period
is given by:9
(137)
where 6"(x4) and 6"(T 4  1) are the temperatures calculated for the heat source and the heat
sink, respectively, using equation (136).
Equations (136) and (137) have been solved numerically for different values of Q'and T4.
The results are presented graphically in Fig. 1.14. Considering the contact section (Q'= 0), the
temperature increases monotonically with time during the heating period, in correspondance
with the relationship:
(138)
q
(a)
Imaginary heat source
Real heat
source
t
Imaginary heat sink
e"
(b)
Heating
period
$ffl9
\
Fig. 1.13. Method for calculation of transient temperatures during friction welding; (a) Sketch of
imaginary heat source/heat sink model, (b) Principles of superposition.
Heating
e"
Cooling
\
Fig. 1.14. Calculated temperaturetime pattern in friction welding.
Example (1.4)
Consider friction welding of 026mm aluminium rods (AlCu2Mg) under the following conditions:
Calculate the peak temperature distribution across the joint. Assume in these calculations
that the thermal diffusivity of the AlCu2Mg alloy is 70mm2 s"1.
Solution
Readings from Fig. 1.14 give:
Next Page
In this particular case, it is possible to check the accuracy of the calculations against in situ
thermocouple measurements carried out on friction welded components made under similar
conditions. A comparison with the data in Fig. 1.15 shows that the model is quite successful in
predicting the HAZ peak temperature distribution. In contrast, the weld heating and cooling
cycles cannot be reproduced with the same degree of precision. This has to do with the fact
that the present analytical solution omits a consideration of the plastic straining occurring
during friction welding, which displaces the coordinates and alters the heat balance for the
system.
Previous Page
In this particular case, it is possible to check the accuracy of the calculations against in situ
thermocouple measurements carried out on friction welded components made under similar
conditions. A comparison with the data in Fig. 1.15 shows that the model is quite successful in
predicting the HAZ peak temperature distribution. In contrast, the weld heating and cooling
cycles cannot be reproduced with the same degree of precision. This has to do with the fact
that the present analytical solution omits a consideration of the plastic straining occurring
during friction welding, which displaces the coordinates and alters the heat balance for the
system.
Temperature, 0C
Heating period
Time, s
Peak temperature, 0C
(b)
Observed relationship
Predicted relationship
by Rosenthal,1314 but the theory has later been extended and refined by a number of other
inve stigators .9'n*15"20
1.10.1 Arc efficiency factors
In arc welding heat losses by convection and radiation are taken into account by the efficiency
factor r\, defined as:
(140)
where qo is the net power received by the weldment (e.g. measured by calorimetry), / is the
welding current (amperage), and U is the arc voltage.
For submerged arc (SA) welding the efficiency factor (r\) has been reported in the range
from 90 to 98%, for SMA and GMA welding from 65 to 85%, and for GTA welding from 22
to 75%, depending on polarity and materials.11
A summary of ranges is given in Table 1.3.
7.70.2 Thick plate solutions
Model (after Rykalin9)
According to Fig. 1.17, the general thick plate model consists of an isotropic semiinfinite
body at initial temperature T0 limited in one direction by a plane that is impermeable to heat.
At time t = 0 a. point source of constant power qo starts on the surface at position O moving in
the positive xdirection at a constant speed U The rise of temperature T T0 in point P at time
t is sought.
During a very short time interval from ^'to t'+ dt'the amount of heat released at the surface
is dQ = qodt'. According to equation (17) this will produce an infinitesimal rise of temperature in P at time t:
(141)
where
to point P.
For a convenient presentation of the pseudosteady state solution, the position P should be
referred to that of the moving heat source. This is achieved by changing the coordinate system
from O to O'(see Fig. 1.17):
and
Solidified
weld metal
Root of arc
Weld pool
X=Vt
Relative position along weld centreline
Fig. 1.16. Schematic diagram showing weld centreline temperature in different positions from the heat
source during steel welding at pseudosteady state.
Table 1.3 Recommended arc efficiency factors for different welding processes. Data from Refs 11,21.
Arc efficiency factor j]
Welding Process
Range
Mean
SA welding (steel)
0.91O.99
0.95
0.660.85
0.80
0.750.93
0.85
0.660.70
0.70
0.250.75
0.40
0.550.80
0.60
0.220.46
0.40
(142)
where
The total rise of temperature at P is obtained by substituting:
and
into equation (142), and integrating between the limits u = (R2l4af)m and u = o.This gives
(after some manipulation):
(143)
It is wellknown that:
(144)
If u is sufficiently small (i.e. when welding has been performed over a sufficient period), we
obtain the pseudosteady state temperature distribution:
(145)
This equation is often referred to as the Rosenthal thick plate solution,1314 in honour of
D. Rosenthal who first derived the relation by solving the differential heat flow equation directly for the appropriate boundary conditions.
1.10.2.1 Transient heating period
It follows from the above analysis that the pseudosteady state temperature distribution is
where
and
Equation (149) has been solved numerically for a limited range of cr3 and x. The results are
presented graphically in Fig. 1.18. A closer inspection of Fig. 1.18 reveals that the duration of
the transient heating period depends on the dimensionless radius vector a 3 . In practice this
means that the Rosenthal equation is not valid during the initial period of welding unless the
distance from the heat source to the observation point is very small. It should be noted, however, that a dimensionless distance o~3 may be 'short' for one combination of welding speed
and thermal diffusivity, while the same position may represent a 'long' distance for another
combination of V and a. Similarly, the dimensionless time T may be 'short' or 'long' at a
chosen number of seconds, depending on the ratio v/2a.
Example (1.5)
Consider stringer bead deposition on a thick plate of aluminium at a constant welding speed of
5 mm s"1. Calculate the duration of the heating period when the distance from the heat source
to the point of observation is 17 mm.
Solution
Ki = (TT0)/(TT0)p.s.
T = v2t/2a
Fig. 1.18. Ratio between real and pseudosteady state temperature in thick plate welding for
different combinations ofCT3and T.
It is seen from Fig. 1.18 that the pseudosteady state temperature distribution is approached
when T ~ 3, which gives:
The above calculations show that the Rosenthal equation is not valid if the ratio between R
and L2 exceeds a certain critical value (typically 0.15 to 0.30 for aluminium welds and 0.4 to
0.6 for steel welds). This important point is often overlooked when discussing the relevance of
the thick plate solution in arc welding.
1.10.2.2 Pseudosteady state temperature distribution
The Rosenthal equation gives, with the limitations inherent in the assumptions, full information on the thermal conditions for point sources on heavy slabs. Accordingly, in order to obtain
a general survey of the pseudosteady state temperature distribution, it is convenient to present
the different solutions in a dimensionless form. The following parameters are defined for this
purpose:11
Dimensionless operating parameter:
(150)
Dimensionless jccoordinate:
(151)
Dimensionless ^coordinate:
(152)
Dimensionless zcoordinate:
(153)
By substituting these parameters into equation (145), we obtain:
(154)
where 6 and a 3 are the dimensionless temperature and radius vector, respectively (previously
defined in equations (19) and (147)).
Equation (154) has been solved numerically for chosen values of a 3 and . A graphical
presentation of the different solutions is shown in Fig. 1.19. These maps provide a good
e/n3
(a)
%
Fig, /./P.Dimensionless temperature maps for point sources on heavy slabs; (a) Vertical sections parallel to the ^axis.
overall indication of the thermal conditions during thick plate welding, but are not suitable for
precise readings. Consequently, for quantitative analyses, the following set of equations can
be used:11
Isothermal zone widths
The maximum width of an isothermal enclosure is obtained by setting 3ln(0M3)/9a3 = 0:
(b)
V
Fig. /.iP.Dimensionless temperature maps for point sources on heavy slabs (continued): (b) Isothermal
contours in the \}/plane for different ranges of 0/n3.
Partial differentiation of the Rosenthal equation gives:
and
(155)
Equation (155) can be used for calculations of isothermal zone widths V/w and cross sectional areas A1. From Fig. 1.20 we have:
(156)
and
(157)
A graphical presentation of equations (155), (156), and (157) is shown in Fig. 1.21.
Length of isothermal enclosures
Referring to Fig. 1.20, the total length of an isothermal enclosure r is given by:
(158)
where ' and "are the distance from the heat source to the front and the rear of the enclosure,
respectively.
x,S
Heat source
y.v
z, C
Vm,O3m,Al
V9P
Fig. 1.21. Dimensionless distance a3m, half width \/m and cross sectional area A1 vs n3 /Qp.
The coordinates ' and " are found by setting a 3 = ^ in equation (154). This gives:
(159)
and
(160)
(162)
d65)
Equation (165) provides a basis for calculating the cooling time within a specific temperature interval (e.g. from O1 to B2):
(166)
For welding of low alloy steel the cooling time from 800 to 5000C is widely accepted as an
adequate index for the thermal conditions under which the austenite to ferrite transformation
takes place. From equation (166), we have:
(167)
(169)
Similarly, the cooling rate at a specific temperature is obtained by differentiating equation
(165) with respect to time:
(170)
By multiplying equation (170) with the appropriate conversion factor, we get:
(171)
Example (1.6)
Consider stringer bead deposition (GMAW) on a thick plate of low alloy steel under the following conditions:
Sketch the contours of the fusion boundary and the Ac3isotherm (9100C) in the \/ (xy)
plane at pseudosteady state.
Solution
As shown in Fig. 1.22(a) it is sufficient to calculate the coordinates in four different (characteristic) positions to draw a sketch of the isothermal contours. If we neglect the latent heat of
melting, the d/n3 ratio at the melting point becomes, according to equations (19) and (150):
Endpoints
The endpoints are readily obtained from equations (159) and (160):
and
Maximum width
The maximum width of an isothermal enclosure can generally be calculated from equations
(155) and (156), or read from Fig. 1.21. Taking O^ /n3 = QIn3 = 0.088, we obtain:
(a)
Pos. (3)
Pos. (2)
Pos. (4)
Pos.(1)
%
(b)
\
x(mm)
y(mm)
Fig. 1.22. Calculation of isothermal contours from Rosenthal thick plate solution; (a) Calculation procedure, (b) Sketch of fusion boundary and Acjisotherm in position = 0 (Example 1.6).
and
which gives
Similarly, the contour of the Ac3isotherm can be determined by inserting BZn3= 0.052 into
the same set of equations. Figure 1.22(b) shows a graphical representation of the computed
isothermal contours.
Example (1.7)
Consider GTA welding on a thick plate of low alloy steel under the following conditions:
Calculate the weld pool volume, the weld bead cross section, the width of the fully transformed HAZ, the cooling time from 800 to 5000C, and the cooling rate at the onset of the
austenite to ferrite transformation (e.g. at 6500C). The arc efficiency factor is taken equal to
0.5.
Solution
If we neglect the latent heat of melting, the GM3 ratio at the melting point becomes:
We can now use equations (163) and (164) to calculate the weld pool volume. This gives:
and
and
which gives
This value is also valid for positions outside the weld centreline, since the cooling curves
at such low temperatures are reasonably parallel down to Tp ~ 9100C (see Fig. 1.19(a)).
Cooling rate at 6500C
The cooling rate at a specific temperature is given by equation (171). Taking the transformation start temperature equal to 6500C, we get:
(172)
According to the assumptions this amount of heat will remain in a slice of thickness dx due
to the lack of a temperature gradient in the welding direction. Since symmetry demands that
the isotherms in the yz plane are semicircles*, the situation becomes identical to the temperature field around a linear instantaneous heat source in a thin plate, provided that the space
above the slab is replaced by solid material and the strength of the source is doubled (see Fig.
1.23(b)). The solution is then given by equation (16) if we replace QId by Iq0 Iv:
d73)
where r* is the twodimensional radius vector in the yz plane.
Equation (173) represents the simplified solution for a fast moving high power source on a
semiinfinite slab, and is valid within a limited range of the more general Rosenthal equation
for threedimensional heat flow (equation (145)). By introducing the dimensionless radius
vector Cr4 = Dr*12a, equation (173) transforms to:
(174)
A graphical presentation of equation (174) gives a family of curves which resembles the
thermal cycles shown in Fig. 1.19(a). Although the cooling conditions close to the weld centreline are similar to those calculated from the Rosenthal equation, the predicted width/depth
of the isotherms will always be greater than that inferred from the general thick plate solution
as illustrated in Fig. 1.24 due to the assumption of 2D heat flow. The parameter o~4m in Fig.
1.24 is obtained by differentiating equation (174) with respect to time:
*The isotherms must meet the plate surface at right angles in the absence of a temperature gradient across the boundary.
(a)
(b)
Symmetry
plane
Fig. 1.23. Fast moving high power source on a semiinfinite slab; (a) Sketch of model, (b) Analogy between a fast moving high power source and an instantaneous line source.
4m m
Asymptote
0 /n
P 3
Fig. 1.24. Theoretical width of isotherms under 2D and 3D heat flow conditions, respectively at pseudosteady state (thick plate welding).
which gives
(175)
It is interesting to note that the dimensionless width a4m will approach \j/ m when the dp /n3
ratio becomes sufficiently small (i.e. less than about 0.1). Under such conditions the isotherms
will be strongly elongated in the welding direction (see Fig. 1.19(b)), which forces the heat to
flow primarily in directions normal to the xaxis.
A general graphical representation of the weld thermal programme can be obtained by
combining equations (174) and (175):
(176)
Equation (176) has been plotted in Fig. 1.25. This graph provides a basis for calculating
the retention time within specific temperature intervals under various welding conditions.
e/ep
W J
Fig. 1.25. Temperaturetime pattern in thick plate welding at high arc power and high welding speed.
Example (1.8)
Consider SA welding on a thick plate of low alloy steel under the following conditions:
Calculate the retention time within the austenite regime (T > 9100C) for points located lmm
outside the fusion boundary.
Solution
If we use the melting point of the steel as a reference temperature, the parameter n3 becomes:
A comparison with Fig. 1.24 shows that the assumption of 2D heat flow is justified when
dp < 1. Hence, the width of the fusion zone (8p = 1) can be calculated from equation (175):
which gives
The total time spent in the thermal cycle from 0 = 0.59 (T= 9100C) to G^ = 0.78 and down
again to 0 = 0.59 can now be read from Fig. 1.25. Taking the ordinate 0/0^ equal to 0.76, we
obtain:
which gives
and
(177)
where
If we refer the position P to that of the heat source at time t, we shall expect a solution
independent of time at pseudosteady state. This is achieved by changing the coordinate system from O to O' (see Fig. 1.26):
K0(U)1K1(U)
U
Fig. 1.27. Graphical representation of the Bessel functions Ko(u) and Kx{u).
d78)
where
to
, we introduce:
from which
d79)
It is wellknown that:
where K0(G5) is the modified Bessel function of the second kind and zero order. A graphical
representation of A^(W) is shown in Fig. 1.27.
Hence, the general thin plate solution can be written as:
(180)
When w is sufficiently large (i.e. when welding has been performed over a sufficient period), we obtain the pseudosteady state temperature distribution:
(181)
Equation (181) is often referred to as the Rosenthal thin plate solution, in memory of its
originator D. Rosenthal.1314 It follows that this model is applicable to all types of welding
processes (including electron beam, plasma arc and laser welding), provided that a full throughthickness penetration is achieved in one pass.
1.10.3.1 Transient heating period
In thin plate welding the duration of the transient heating period is determined by the integral in equation (180). Taking the ratio between the real and pseudosteady state temperature equal to K2, we obtain:
(182)
Consider butt welding of a thin aluminium plate at a constant travel speed of 5mm s"1. Calculate the duration of the transient heating period when the distance from the heat source to the
point of observation is 17mm.
Solution
It follows from Fig. 1.28 that the pseudosteady state temperature distribution is approached
when T ~ 5, which gives:
and
This minimum bead length is nearly twice as large as that calculated for 3D heat flow for
the same combination of welding speed, thermal diffusivity and radius vector (see Example
K2=(TTo)/(TTo)p.s.
t = v2t/2a
Fig. 1.28. Ratio between real and pseudosteady state temperature in thin plate welding for different
combinations of a 5 and T.
(1.5)). Consequently, the duration of the transient heating period is significantly longer in thin
plate welding than in thick plate welding due the pertinent differences in the heat flow conditions.
1.10.3.2 Pseudosteady state temperature distribution
A graphical presentation of the pseudosteady state temperature distribution similar to that
shown in Fig. 1.19 for threedimensional heat flow may be obtained by introducing the
dimensionless plate thickness 8 = vdlla, which is a measure of the relative speeds of the arc
and the heat flow in the material. By rearranging equation (181), we get:
(183)
Plots of this equation are shown in Fig. 1.29. It follows that the pseudosteady state temperature distribution in thin plate welding depends on the parameter 68//?3. In practice, this
means that the shape of the isotherms is not influenced by the welding speed, since both 8 and
n3 are proportional to v.
Isothermal zone widths
The maximum width of an isothermal enclosure is obtained by setting 3(68 /n3) /die, = 0. Noting that (d/du)Ko(u)  Kx(u), where K1 (u) is the modified Bessel function of the second kind
and first order (shown in Fig. 1.27), we get:
(184)
%
\
V
Fig. 1.29. Dimensionless temperature maps for line sources in thin plates for different ranges of 98/n3.
Heat source
Equation (186) can be used for calculations of isothermal zone widths (v/m) and cross
sectional areas (A2) in thin plate welding. Referring to Fig. 1.30, we have:
d87)
and
(188)
Figure 1.31 shows a graphical presentation of equations (186), (187), and (188).
Length of isothermal enclosures
The distance from the heat source to the front (') and the rear (^') of an isothermal enclosure is obtained by substituting o~5 = into equation (183).
This gives:
(189)
and
(190)
^5m.VmA 2 /8
n3/6p5
s\?.*t
Fig. l.Ji.Dimensionless distance a5m, half width \/m and cross sectional area A2/S vs n3/Qph.
n 3 /98
Fig. 1.32. Dimensionless distance from heat source to front (') and rear (") of isothermal enclosure vs
3/68 (thin plate welding).
A graphical presentation of equations (189) and (190) is shown in Fig. 1.32. Included is
also a plot of the total length of the enclosure t vs the parameter n3 /08.
Cooling conditions close to weld centreline
For points located on the weld centreline behind the heat source a 5 =  = X. When x is larger
than about 1, (i.e. t > IaIv1), it is a fair approximation to set K0 (x) exp(x)Vrc/2x (see Fig.
1.27). Hence, equation (183) reduces to:
(191)
Equation (191) provides a basis for calculating the cooling time within a specific temperature interval (e.g. from O1 to 02):
(192)
The dimensionless cooling time from 800 to 5000C is thus given by:
(193)
from which the real cooling time is obtained:
(194)
Taking as average values X = 0.025 W mm"1 0C"1, pc = 0.005 J mm"3 0C"1, and T0 = 200C
for welding of low alloy steels, we have:
d95)
Similarly, the cooling rate at a specific temperature is obtained by differentiating equation
(191) with respect to time:
(196)
(197)
(198)
Example (LlO)
Consider GTA butt welding of a 2mm thick sheet of coldrolled aluminium (AlMg alloy)
under the following conditions:
Sketch the contours of the fusion boundary and the Arisotherm in the \/ (xy) plane at
pseudosteady state. The recrystallisation temperature Ar of the base material is taken equal to
2750 C. Calculate also the cross sectional area of the fully recrystallised HAZ and the cooling
rate at 2750C for points located within this region.
Solution
Referring to Fig. 1.22(a) (Example (1.6)) it is sufficient to calculate the coordinates in four
different (characteristic) positions to sketch the contour of the fusion boundary. If we neglect
the latent heat of melting, the n3/68 ratio at the melting point becomes:
Endpoints
The endpoints can be read from Fig. 1.32:
and
Maximum widths
The maximum width of an isothermal enclosure can generally be calculated from equations
(186) and (187) or read from Fig. 1.31. When n3/QpS = 0.84, we obtain:
and
which gives
Similarly, the contour of the Arisotherm can be determined by inserting n3/db = 2.08 into
the same set of equations. Figure 1.33 shows a graphical representation of the calculated
isothermal contours.
x(mm)
y(mm)
Fig. 1.33. Calculated contours of fusion boundary and Arisotherm in GTA butt welding of a 2mm thick
aluminium plate (Example 1.10).
which gives
(199)
According to the assumptions this amount of heat will remain in a rod of constant cross
sectional area due to the lack of a temperature gradient in the welding direction. Under such
conditions the mode of heat flow becomes essentially onedimensional, and the temperature
distribution is given by equation (15):
(1100)
Equation (1100) represents the simplified solution for a fast moving high power source* in
a thin sheet, and is valid within a limited range of the more general Rosenthal equation for twodimensional heat flow (equation (181)).
By substituting the appropriate dimensionless parameters into equation (1100), we obtain:
* Since the shape of a given isotherm in the xy plane is determined by the qjd ratio, the minimum welding speed
which is required to maintain 1D heat flow increases with decreasing qjVd ratios. Hence, the term 'fast moving high
power source' is also appropriate in the case of the thin plate welding.
(1101)
The locus of the peak temperatures is readily evaluated from equation (1101) by setting
3ln(e8M3)/3T = 0:
which gives
and
(1102)
It is evident from the plot in Fig. 1.35 that the predicted width of the isotherms is always
greater than that inferred from the general thin plate solution (equation (183)) due to the
assumption of onedimensional heat flow. However, such deviations become negligible at
very small Qpb/n3 ratios because of a small temperature gradient in the welding x direction
compared to the transverse y direction of the plate.
A general graphical representation of the weld thermal programme (similar to that shown in
Fig. 1.25 for a fast moving high power source on a heavy slab) can be obtained by combining
equations (1101) and (1102):
m(1D)/m(aD)
Asymptote
0 p 8/n 3
e/ep
Fig. 1.35. Theoretical width of isotherms under 1D and 2D heat flow conditions, respectively at pseudosteady state (thin plate welding).
2t/(vm)2
Fig. 1.36. Temperaturetime pattern in thin plate welding at high arc power and high welding speed.
Next Page
(1103)
Equation (1103) has been plotted in Fig. 1.36.
Example (LU)
Consider butt welding of a 2mm thin plate of austenitic stainless steel with covered electrodes
(SMAW) under the following conditions:
Calculate the retention time within the critical temperature range for chromium carbide
precipitation (i.e. from 650 to 8500C) for points located at the 8500C isotherm.
Solution
If we use the melting point of the steel as a reference temperature, the parameter n3/5 becomes:
A comparison with Fig. 1.35 shows that the assumption of 1D heat flow is justified when
Qp< 1. Hence, the total time spent in the thermal cycle from 6 = 0.43 (T = 6500C) to 0p = 0.56
(Tp = 8500C) and down again to 0 = 0.43 can be read from Fig. 1.36. Taking the ordinate 0/0p
equal to 0.76, we obtain:
which gives
and
The general medium thick plate model considers a point heat source moving at constant speed
across a wide plate of finite thickness d. With the exception of certain special cases (e.g.
watercooling of the back side of the plate), it is a reasonable approximation to assume that the
Previous Page
(1103)
Equation (1103) has been plotted in Fig. 1.36.
Example (LU)
Consider butt welding of a 2mm thin plate of austenitic stainless steel with covered electrodes
(SMAW) under the following conditions:
Calculate the retention time within the critical temperature range for chromium carbide
precipitation (i.e. from 650 to 8500C) for points located at the 8500C isotherm.
Solution
If we use the melting point of the steel as a reference temperature, the parameter n3/5 becomes:
A comparison with Fig. 1.35 shows that the assumption of 1D heat flow is justified when
Qp< 1. Hence, the total time spent in the thermal cycle from 6 = 0.43 (T = 6500C) to 0p = 0.56
(Tp = 8500C) and down again to 0 = 0.43 can be read from Fig. 1.36. Taking the ordinate 0/0p
equal to 0.76, we obtain:
which gives
and
The general medium thick plate model considers a point heat source moving at constant speed
across a wide plate of finite thickness d. With the exception of certain special cases (e.g.
watercooling of the back side of the plate), it is a reasonable approximation to assume that the
plate surfaces are impermeable to heat. Thus, in order to maintain the net flux of heat through
both boundaries equal to zero, it is necessary to account for mirror reflections of the source
with respect to the planes of z = 0 and z = d. This can be done on the basis of the 'method of
images' as illustrated in Fig. 1.37. By including all contributions from the imaginary sources
...2q__2 , 2g_i , 2q\ , 2q2 ,...located symmetrically at distances 2id below and above the upper
surface of the plate, the pseudosteady state temperature distribution is obtained in the form of
a convergent series*:
(1104)
where
Note that equation (1104) is simply the general Rosenthal thick plate solution (equation
(145)) summed for each source.
Fig. 1.37. Real and imaginary point sources on a medium thick plate.
*The number of imaginary heat sources necessary to achieve the required accuracy depends on the chosen values of
R0 and vd/2a.
By substituting the dimensionless parameters defined above into equation (1104), we obtain:
(1105)
where
It follows from equation (1104) that the thermal conditions will be similar to those in a
thick plate close to the centre of the weld. Moreover, Rosenthal1314has shown on the basis of
a Fourier series expansion that equation (1104) converges to the general thin plate solution
(equation (181) for points located sufficiently far away from the source. However, at intermediate distances from the heat source, the pseudosteady state temperature distribution will deviate significantly from that observed in thick plate or thin plate welding because of variable
temperature gradients in the throughthickness direction of the plate. Within this 'transition
region', the thermal programme is only defined by the medium thick plate solution (equation
(1104)).
1.10.4.1 Dimensionless maps for heat flow analyses
Based on the models described in the previous sections, it is possible to construct a series of
dimensionless maps which provide a general outline of the pseudosteady state temperature
distribution during arc welding.20
Construction of the maps
The construction of the maps is done on the basis of the medium thick plate solution (equation
(1105)). This model is generally applicable and allows for the plate thickness effect in a
quantitative manner. Since the other solutions are only valid within specific ranges of this
equation, they will have their own characteristic fields in the temperaturedistance or the temperaturetime space. The extension of the different fields can be determined from numerical
calculations of the temperature distribution by comparing each of these models with the medium thick plate solution, using a conformity of 95% as a criterion.
Similarly, when the 95% conformity is not met between the respective solutions, the fields
are marked 'transition region'. Since any combination of dimensionless temperature, operating parameter, and plate thickness locates a point in a field, it means that the dominating heat
flow mechanism can readily be read off from the maps.
Peak temperature distribution
The variation of peak temperature with distance in the \j/(^jdirection has been numerically
evaluated from equation (1105) for different values of the dimensionless plate thickness (8 =
vdlla). The results are shown graphically in Fig. 1.38(a) and (b) for the two extreme cases of
= 0 (z = 0) and = 8 (z = d), respectively.
An inspection of the maps reveals that the temperaturetime pattern in stringer bead
weldments can be classified into three main categories:
(a)
VV
(b)
Thin plate solution
1D heat flow
V"3
Fig. 1.38. Peak temperature distribution in transverse direction (\/ = \\fm) of plate; (a) Upper plate surface
(^ = 0), (b) Lower plate surface (J = 8).
1. Close to the heat source, the thermal programme will be similar to that in a thick plate
(Fig. 1.38(a)), which means that the temperature distribution is determined by equation (154). For large values of the dimensionless plate thickness, the mode of heat flow may
become essentially twodimensional. This corresponds to the limiting case of a fast moving high power source in a thick plate (equation (174)). Under such conditions the slope of
the Qp/n3\ym curves in Fig. 1.38(a) attains a constant value of2.
2. With increasing distance from the heat source, a transition from threedimensional to
twodimensional heat flow may occur, depending on the dimensionless plate thickness and
the operational conditions applied. Considering the upper surface of the plate (Fig. 1.38(a)),
the extension of the transition region is seen to decrease with increasing values of 8 as the
conditions for thick plate welding are approached. The opposite trend is observed for the
bottom plate surface (Fig. 1.38(b)), since a small dimensionless plate thickness generally
results in a more rapid equalisation of the temperature gradients in the t,(z) direction. When
the curves in Fig. 1.38(b) become parallel with the jcaxis, the temperature at the bottom of
the plate reaches its maximum value. Note that within the transition region, reliable predictions of the pseudosteady temperature distribution can only be made from the medium
thick plate solution (equation (1105)).
3. For points located sufficiently far away from the heat source, the temperature gradients in the throughthickness direction of the plate become negligible. This implies that the
temperature distribution at the upper and lower surface of the plate is similar, and can be
computed from the thin plate solution (equation (183)). When the conditions for onedimensional heat flow are approached (equation (1101)), the slope of the dp/n3\\fm curves
in Fig. 1.38(a) and (b) attains a constant value o f  1 .
Cooling conditions close to weld centreline
Figure 1.39 contains a plot of the cooling programme for points located on the weld centreline
(\j/ = = 0), as calculated from equation (1105). A closer inspection of Fig. 1.39 reveals that
the slope of the cooling curves increases gradually from 1 to 0.5 with increasing distance
from the heat source. This corresponds to a change from threedimensional to onedimensional heat flow.
From Fig. 1.39 it is possible to readoff the cooling time within specific temperature intervals for a wide range of operational conditions. These results are also valid for positions
outside the weld centreline, since the cooling curves are virtually parallel in the transverse \/
direction of the plate. A requirement is, however, that the peak temperature of the thermal
cycle is significantly higher than the actual temperature interval under consideration.
Retention times at elevated temperatures
The retention time, Axn is defined as the total time spent in a thermal cycle from a chosen
reference temperature 0 to the peak temperature dp and down again to 9. This parameter can
readily be computed from equation (1105) by means of numerical methods. The results of
such calculations (carried out in position = 0) are shown graphically in Fig. 1.40 for 9 = 0.5
An inspection of Fig. 1.40 reveals a complex temperaturetime pattern. In this case it is not
possible to determine the exact field boundaries between the respective solutions, since the
e/n3
e/n3
T= ^
e/n 3
X
Fig. 1.39. Cooling programme for points located on the weld centreline (\\f = =0).
AXf
Fig. 1.40. Total time spent in a thermal cycle from 9 through 9p to 9 for a chosen reference temperature of
9 = 0.59p.
mode of heat flow may vary within a single thermal cycle. Hence, the extension of the different fields is not indicated in the graph. The results in Fig. 1.40 provide a systematic basis for
calculating the retention time within specific temperature intervals under various welding conditions.
Isothermal contours
Because of the number of variables involved, it is not possible to present a twodimensional
plot of the isotherms without first specifying the dimensionless plate thickness. Examples of
calculated isotherms in different planes are shown in Figs. 1.41 and 1.42 for 8 equal to 0.5 and
5, respectively. It is evident that an increase in the dimensionless plate thickness from 0.5 to 5
has a dramatic effect on the shape and position of the isothermal contours. However, in order
to explain these observations in an adequate manner, it is necessary to condense the results into
a twodimensional diagram. As shown in Fig. 1.43, this can be done by plotting the calculated
field boundaries in Fig. 1.38(a) at maximum width of the isotherms vs the parameters Qp/n3
and vdlla.
It is seen from Fig. 1.43 that a large plate thickness generally will favour threedimensional
heat flow. With decreasing values of Qp/n3, the conditions for a fast moving high power source
are approached before the transition from the thick plate to the thin plate solution occurs. In
such cases the isotherms at the bottom of the plate will be strongly elongated in the welding
direction and shifted to positions far behind the heat source. The opposite trend is observed at
small values of vdlla, since a rapid equalisation of the temperature gradients in the throughthickness direction of the plate will result in elliptical isotherms at both plate surfaces, located
in an approximately equal distance from the heat source. In either case the temperature at
which the crosssectional isotherms approach a semicircle or become parallel with the XXz)axis can be obtained from Fig. 1.43 by readingoff the intercept between the line for the
dimensionless plate thickness and the respective field boundaries.
Limitations of the maps
Since the maps have been constructed on the basis of the analytical heat flow equations, it is
obvious that they will apply only under conditions for which these equations are valid. The
simplifying assumptions inherent in the models can be summarised as follows:
(a)
(b)
The thermal conductivity, density, and specific heat are constant and
independent of temperature.
(c)
The plate is completely insulated from its surroundings, i.e. there are no
heat losses by convection or radiation from the boundaries.
(d)
(e)
The electrode is not consumed during welding, and all heat is concentrated in a zerovolume point or line.
(a)
(b)
(C)
C
Fig. 1.41. Computed isothermal contours in different sections for 8 = 0.5;
(a) Front view (\j/ = \/m),
(b) Side view (\/ = 0),
(c) Top view ( = 0) and bottom view ( = 8).
(a)
(b)
(C)
ep/n3
Thin plate
solution
5 = vd/2a
Fig. 1.43. Heat flow mechanism map showing calculated field boundaries in transverse direction (i/
= i/m) of plate vs Qp/n3 and 8 = vdlla.
(f)
Pseudosteady state, i.e. the temperature does not vary with time when
observed from a point located in the heat source.
In general, the justification of these assumptions relies on a good correlation between theory
and experiments. However, since the analytical solutions ignore the important role of arc
energy distribution and directed metal currents in the weld pool, predictions of the weld thermal programme should be restricted to positions well outside the fusion zone where such
effects are of less importance (to be discussed below).
Example (1.12)
Consider stringer bead welding (GMAW) on a 12mm thick plate of aluminium (> 99% Al)
under the following conditions:
Based on Fig. 1.43, sketch the peak temperature contours in the transverse section of the
weld at pseudosteady state.
Solution
If we neglect the latent heat of melting, the parameter n3 at the chosen reference temperature
(Tc = Tm) becomes:
Similarly, when v = 3mm s l and a = 85mm2 s"1 we obtain the following value for the dimensionless plate thickness:
^, (0C)
Model System
Comments
0.50 1.0
340 660
Medium thick
plate solution
< 0.17
< 130
A sketch of the HAZ peak temperature contours in the transverse section of the weld is
shown in Fig. 1.44.
Case Study (Ll)
Consider stringer bead GMA welding on 12.5mm thick plates of low alloy steel and aluminium
(i.e. a AlMgSi alloy), respectively under the conditions E = 1.5 kJ mm"1 and r\ = 0.8. Details
of welding parameters for the four series involved are given in Table 1.4. Computed peak
temperature contours in the transverse section of the welds are given in Figs. 1.45 and 1.46.
Fusion
zone
Fig. 1.44. Schematic diagram showing specific peak temperature contours within the HAZ of an aluminium weld at pseudosteady state (Example 1.12).
WELD A1
y(mm)
(b)
WELD A2
z(mm)
(a)
z(mm)
y(mm)
Fig. 1.45. Computed peak temperature contours in aluminium welding at pseudosteady state (Case study
1.1); (a) Weld Al, (b) Weld A2. Black regions indicate fusion zone.
Aluminium welding
In general, the maximum width of the isotherms at the upper and lower surface of the plate can
be obtained from Fig. 1.38(a) and (b), although these maps are not suitable for precise readings. A comparison with the computed peak temperature contours in Fig. 1.45(a) and (b)
reveals a strong influence of the welding speed on the shape and position of the crosssectional
isotherms at a constant gross heat input of 1.5 kJ mm"1. It is evident that the extension of the
fusion zone and the neighbouring isotherms becomes considerably larger when the welding
speed is increased from 2.5 to 5 mm s"1. This effect can be attributed to an associated shift
from elliptical to more elongated isotherms at the plate surfaces (e.g. see Fig. 1.43), which
reduces heat conduction in the welding direction. It follows from Fig. 1.43 that the upper left
corner of the map represents the typical operating region for aluminium welding.
Because of the pertinent differences in the heat flow conditions, the temperaturetime pattern will also vary significantly between the respective series as indicated by the maps in Figs.
1.39 and 1.40. Hence, in the case of aluminium welding the usual procedure of reporting arc
power and travel speed in terms of an equivalent heat input per unit length of the bead is highly
questionable, since this parameter does not define the weld thermal programme. In general,
the correct course would be to specify both qo, v and d, and compare the weld thermal history
on the basis of the dimensionless parameters n3 and 8.
Steel welding
If welding is performed on a steel plate of similar thickness, the operating region will be
shifted to the lower right corner of Fig. 1.43. Under such conditions, the isotherms adjacent to
the fusion boundary will be strongly elongated in the xdirection even at very low welding
speeds (see Fig. 1.42). This implies that the thermal programme approaches a state where the
temperature distribution is uniquely defined by the net heat input r\E, corresponding to the
(a)
W E L D S1
z(mm)
y(mm)
(b)
WELD S2
z(mm)
y(mm)
Fig. 7.46.Computed peak temperature contours in steel welding at pseudosteady state (Case study 1.1);
(a) Weld Sl, (b) Weld S2. Black regions indicate fusion zone.
Series
(W)
(mras"1)
(mm)
(kJmrrr 1 )
AlMgSi
alloy
Al
A2
6000
3000
5
2.5
12.5
12.5
Low alloy
steel
Sl
S2
9600
4800
8
4
12.5
12.5
Material
n3
1.5
1.5
0.36
0.09
0.50
0.25
1.5
1.5
32.6
8.2
10
5
limiting case of a fast moving high power source. As a result, the calculated shape and width
of the fusion boundary and neighbouring isotherms are seen to be virtually independent of
choice of qo and v as illustrated in Fig. 1.46(a) and (b).
1.10.4.2 Experimental verification of the medium thick plate solution
It is clear from the above discussion that the medium thick plate solution provides a systematic
basis for calculating the temperature distribution within the HAZ of stringer bead weldments
under various welding conditions. In the following, the accuracy of the model will be checked
against extensive experimental data, as obtained from in situ thermocouple measurements and
numerical analyses of a large number of beadonplate welds.
Weld thermal cycles
Examples of measured and predicted weld thermal cycles in aluminium welding are presented
in Fig. 1.47. It is evident that the medium thick plate solution predicts adequately the HAZ
temperaturetime pattern under different heat flow conditions for fixed values of the peak
temperature. This, in turn, implies that the model is also capable of predicting the total time
spent in a thermal cycle within a specific temperature interval as shown in Fig. 1.48.
Weld cooling programme
At temperatures representative of the austenite to ferrite transformation in mild and low alloy
steel weldments, the conditions for a fast moving high power source are normally approached
before the transition from thick plate to thin plate welding occurs (see Fig. 1.39). In such
cases, it is possible to present the different solutions for Ax875 (at \/ = = 0) in a single graph by
introducing the following groups of variables*:
Ordinate:
Abscissa:
Relevant literature data for the cooling time between 800 and 5000C are given in Fig. 1.49.
A closer inspection of the figure reveals a reasonable agreement between observed and predicted values in all cases. For welding of thick plates, the ordinate attains a constant value of
1. Similarly, in thin sheet welding, the slope of the curve becomes equal to 1.
In aluminium welding, the thermal conditions will be much more complex because of the
resulting higher base plate thermal diffusivity (see Fig. 1.39). Hence, it is not possible to
describe the weld cooling programme in terms of equations (174) and (1101), which apply to
fast moving high power sources. The plot in Fig. 1.50 confirms, however, that the medium
thick plate solution is also capable of predicting the cooling time within specific temperature
intervals (e.g. from 300 to 1000C) in aluminium weldments.
These groups of variables can be obtained from equations (174) and (1101).
Measured
Predicted
Temperature (0C)
Time (sec)
A tr(s), observed
Fig. 1.47. Comparison between measured and predicted weld thermal cycles in aluminium welding for
fixed values of Tp. Data from Myhr and Grong.20
P
t
A tr(s), predicted
Fig. /.4#. Comparison between measured and predicted retention times in aluminium welding for fixed
values of Tp. Data from Myhr and Grong.20
8/5
AT
n 3 [i4
SMAW
SMAW
SAW
SAW
THICK PLATE
SOLUTION
THIN PLATE
SOLUTION
2Hr + Ir 1
5 6SOO 6SOO
Fig. 1.49. Comparison between observed and predicted cooling times from 800 to 5000C in steel welding
(solid lines represent theoretical calculations). Data from Myhr and Grong.20
At
(sec), observed
T(0C)
t(sec)
At
(sec), predicted
Fig. 1.50. Comparison between observed and predicted cooling times from 300 to 1000C in aluminium
welding. Data from Myhr and Grong.20
Tp, 0C , observed
Tp, 0C , predicted
Fig. 1.51. Comparison between observed and predicted HAZ peak temperatures in aluminium welding.
Data from Myhr and Grong.20
Based on the finite difference method (FDM).
Z (mm)
y (mm)
Fig. 1.52. Comparison between numerically and analytically computed peak temperature contours in
GTA welding of a 3.2mm thin aluminium sheet. (Broken lines: numerical model; solid lines: analytical
model.) Data for welding parameters and material properties are given in Ref. 21.
1. Considering heat flow and temperature distribution in fusion welding, there exists no
defined plate thickness which can be regarded as 'thick' or 'thin'. Accordingly, in a real
welding situation, the mode of heat flow will vary continuously with increasing distance
from the heat source.
2. Close to the centre of the weld, the thermal conditions will be similar to those in a
heavy slab. This means that the temperature distribution is approximately described by the
Rosenthal thick plate solution.
3. At intermediate distances from the heat source, the temperature distribution will deviate significantly from that observed in thick plate welding because of variable temperature gradients in the throughthickness direction of the plate. Within this transition region,
reliable predictions of the pseudosteady state temperature distribution can only be made
from the medium thick plate solution.
4. For points located sufficiently far away from the heat source, the temperature gradients in the throughthickness direction of the plate become negligible. Under such conditions, the weld thermal programme is approximately defined by the Rosenthal thin plate
solution.
5. In general, a full description of the weld thermal history requires that both the arc
power qo, the travel speed V, and the plate thickness d are explicitly specified. Hence, the
usual procedure of reporting welding variables in terms of an equivalent heat input per unit
length of the bead, (kJ mm"1), is highly questionable, since this parameter does not define
the weld thermal programme. An exception is welding of thick steel plates, where the
temperature distribution approaches that of a fast moving high power source because of a
low thermal diffusivity of the base metal.
6. A comparison between theory and experiments shows that the medium thick plate
solution predicts adequately both the peak temperature distribution and the temperaturetime pattern within the HAZ of stringer bead weldments for a wide range of operational
conditions (including aluminium and steel welding). A requirement is, however, that the
equation is calibrated against a known isotherm (e.g. the fusion boundary) due to the simplifying assumptions inherent in the model.
(1106)
where
y
Crater
HAZ"
Finger
Fusion line
Z
Fig. 1.53. Schematic diagram showing the weld crater/weld finger formation during stringer bead welding.
(a)
(b)
Z
Fig. 1.54. General heat flow model for welding on medium thick plates; (a) Physical representation of the
heat distribution by elementary point sources, (b) Method for calculating the temperature field around an
elementary point source displaced along the yaxis.
(C)
Z
Fig. 1.54. General heat flow model for welding on medium thick plates(continued): (c) Method for calculating the temperature field around an elementary (submerged) point source displaced along the zaxis.
Similarly, for a submerged point source located along the zaxis (Fig. 1.54(c))> we obtain:
(1107)
where
and
Note that equation (1107) correctly reduces to equation (1106) when A approaches zero.
The total temperature rise in point P is then obtained by superposition of the temperature
fields from the different elementary heat sources, i.e.:
(1108)
where
In practice, we can subdivide the heat distributions into a relatively small number of elementary point sources, and usually a total number of 8 to 10 sources is sufficient to obtain
good results (i.e. smooth curves). However, the relative strength of each heat source and their
distribution along the y and zaxes must be determined individually by trial and error by
comparing the calculated shape of the fusion boundary with the real (measured) one.
Figure 1.55 shows the results from such calculations, carried out for a single pass (beadingroove) GMA steel weld. It is evident that the important effect of the weld crater/weld finger
formation on the HAZ peak temperature distribution is adequately accounted for by the present
model. A weakness of the model is, of course, that the shape and location of the fusion boundary must be determined experimentally before a prediction can be made.
1.10.5.2 Simplified solution
Similar to the situation described above, the point heat source will clearly not be a good
model when the heat is supplied over a large area. Welding with a weaving technique and surfacing with strip electrodes are prime examples of this kind.
Model (after Grong and Christensen19)
As a first simplification, the Rosenthal thick plate solution is considered for the limiting case
of a high arc power qo and a high welding speed D, maintaining the ratio qo Iv within a range
applicable to arc welding. Consider next a distributed heat source of net power density qo I2L
extending from L to +L on either side of the weld centreline in the ydirection*, as shown
schematically in Fig. 1.56. It follows from equation (173) that an infinitesimal source dqy
located between y and y + dy will cause a small rise of temperature in point P at time f, as:
(1109)
where
and
z (mm)
y (mm)
Fig. 1.55. Calculated peak temperature contours in the transverse section of a GMA steel weldment (Operational conditions: / = 450A, U  30V, v = 2.6mm s"1, d = 50mm).
Fig. 1.56. Distributed heat source of net power density qJ2L on a semiinfinite body (2D heat flow).
(1110)
where erf(u) is the Gaussian error function (previously defined in Appendix 1.3).
Peak temperature distribution
Because of the complex nature of equation (1110), the variation of peak temperature with
distance in the yz plane can only be obtained by numerical or graphical methods. Accordingly, it is convenient to present the different solutions in a dimensionless form. The following
parameters are defined for this purpose:
Dimensionless operating parameter:
(1111)
Dimensionless time:
(1112)
Dimensionless ycoordinate:
(1113)
Dimensionless zcoordinate:
(1114)
By substituting these parameters into equation (1110), we obtain:
(1115)
The variation of peak temperature Qp with distance in the yz plane has been numerically
evaluated from equation (1115) for chosen values of P and 7 (i.e. P = 0, (3 = 3/4, P = I , and
7 = 0). The results are presented graphically in Figs. 1.57 and 1.58 for the through thickness
(z = zm) and the transverse (y = ym) directions, respectively. These figures provide a systematic
basis for calculating the shape of the weld pool and neighbouring isotherms under various
welding conditions.
In Fig. 1.59 the weld width to depth ratio has been computed and plotted for different
combinations of nw and 9p. It is evident that the predicted width of the isotherms generally is
much greater, and the depth correspondingly smaller than that inferred from the point source
model. Such deviations tend to become less pronounced with decreasing peak temperatures
(i.e. increasing distance from the heat source). At very large value of nw, the theoretical shape
of the isotherms approaches that of a semicircle, which is characteristic of a point heat source.
Example (1.13)
Consider GMA welding with a weaving technique on a thick plate of low alloy steel under the
following conditions:
Sketch the contours of the fusion boundary and the Ac r isotherm (71O0C) in the yz plane.
Compare the shape of these isotherms with that obtained from the point heat source model.
Solution
If we neglect the latent heat of melting, the operating parameter at the chosen reference temperature (Tc = Tm) becomes:
Fusion boundary
Here we have:
Ac jtemperature
In this case the peak temperature should be referred to 7200C, i.e.:
ep/nw
V 
ep/nw
Fig. /.57. Calculated peak temperature distribution in the throughthickness direction of the plate at different positions along the weld surface.
y m /L
Fig. 1.58. Calculated peak temperature distribution in the transverse direction of the plate at position y
(Z) = 0.
y m /z m
nw
Fig. 1.59. Effects of nw and dp on the weld width to depth ratio.
from which
Similarly, equation (175) provides a basis for calculating the width of the isotherms in the
limiting case where all heat is concentrated in a zerovolume point. By rearranging this equation, we obtain:
which gives
6.3 mm when Bp = 1(Tp = 15200C)
and
when
Figure 1.60 shows a graphical presentation of the calculated peak temperature contours.
Implications of model
It is evident from Fig. 1.60 that the predicted shape of the isotherms, as evaluated from equation (1110), departs quite strongly from the semicircular contours required by a point heat
source. Moreover, a closer inspection of the figure shows that inclusion of the heat distribution
also gives rise to systematic variations in the weld thermal programme along a specific
isotherm, as evidenced by the steeper temperature gradient in the vdirection compared with
the zdirection of the plate. This point is more clearly illustrated in Fig. 1.61, which compares
the HAZ temperaturetime programme for the two extreme cases of z = 0 and y = 0, respectively. It is obvious from Fig. 1.61 that the retention time within the austenite regime is
considerably longer in the latter case, although the cooling time from 800 to 5000C, Af8/5, is
reasonably similar. These results clearly underline the important difference between a point
heat source and a distributed heat source as far as the weld thermal programme is concerned.
Model limitations
In the present model, we have used the simplified solution for a fast moving high power
source (equation 173)) as a starting point for predicting the temperaturetime pattern. Since
the equations derived later are obtained by integrating equation (173), they will, of course,
apply only under conditions for which this solution is valid.
Moreover, a salient assumption in the model is that the heat distribution during weaving
can be represented by a linear heat source orientated perpendicular to the welding direction.
Although this is a rather crude approximation, experience shows that the assumed heat dis
z, mm
Fig. /.60. Predicted shape of fusion boundary and Acpisotherm during GMA welding of steel with an
oscillating electrode (Example 1.13). Solid lines: Distributed heat source; Broken lines: Point heat source.
Temperature, 0C
Time, s
Fig. L61. Calculated HAZ thermal cycles in positions y = 0 and z = 0 (Example 1.13).
tribution is not critical unless the rate of weaving is kept close to the travel speed. However,
for most practical applications weaving at such low rates would be undesirable owing to an
unfavourable bead morphology.
Case Study (1.2)
Surfacing with strip electrodes makes a good case for application of equation (1110). Specifically, we shall consider SA welding of low alloy steel with 60mm X 0.5mm stainless steel
electrodes. The operational conditions employed are listed in Table 1.5.
It is evident from the metallographic data presented in Fig. 1.62 that neither the bead penetration nor the HAZ depth (referred to the plate surface) can be predicted readily on the basis
of the present heat flow model when welding is carried out with a consumable electrode,
owing to the formation of a reinforcement. This situation arises from the simplifications made
in deriving equation (1110). The problem, however, may be eliminated by calculating the
depth of the Ac3 and Ac1 regions relative to the fusion boundary, i.e. Azm = zm(Qp)  zm (Qp = 1),
or Aym = ym (8p)  ym (6p = 1), for specific positions along the weld fusion line, as shown by the
solid curves in Fig. 1.62 for Qp = 0.54 and 0.45, respectively. An inspection of the graphs
reveals satisfactory agreement between theory and experiments in all three cases, which implies that the model is quite adequate for predicting the HAZ thermal programme as far as strip
electrode welding is concerned. This result is to be expected, since the assumption of twodimensional heat flow is a realistic one under the prevailing circumstances.
Case Study (1.3)
As a second example we shall consider GTA welding (without filler wire additions) at various
heat inputs and amplitudes of weaving within the range from 1 to 2.5 kJ mm"1 and 0 to 15mm,
respectively. Data for welding parameters are given in Table 1.6.
Stainless steel!
y, mm
Weld S3
Fusion line
z, mm
Stainless steel
y, mm
Weld S4
z, mm
Stainless steel!
y, mm
Weld S5
z, mm
Fig. 1.62. Comparison between observed and predicted Ac3 and Ac1 contours during strip electrode welding (Case study 1.2). Data from Grong and Christensen.19
Table 1.5 Operational conditions used in strip electrode welding experiments (Case study 1.2).
Base metal/
filler metal
combination
Weld
No.
S3
Low alloy steel/
stainless steel
/
(A)
730
S4
73
S5
730
U
(V)
27
27
27
v
(mms"1)
2L
(mm)
nw
Cn = 0.7)
1.8
60
0.34
2 2
2.5
60
60
0.28
0.24
Table 1,6 Operational conditions used in GTA welding experiments (Case study 1.3).
I
2L
1
nw
"V=O
12
WeIdNo.
(A)
(V)
(mms" )
(mm)
T] = 0.23
Bl
200
13.5
2.6
9.5
0.20
0.40
B2
200
14.0
1.1
15.0
0.20
0.40
B3
200
13.5
2.5
B4
200
12.5
1.0
Calibration procedure
In general, a comparison between theory and experiments requires that the arc efficiency factor can be established with a reasonable degree of accuracy. Unfortunately, the arc efficiency
factor for GTA welding has not yet been firmly settled, where values from 0.25 up to 0.75 have
been reported in the literature (see Table 1.3). Additional problems result from the fact that
only a certain fraction of the total amount of heat transferred from the arc to the base plate is
sufficiently intense to cause melting. This has led to the introduction of the melting efficiency
factor T]m, which normally is found to be 3070% lower than the total arc efficiency of the
process, depending on the latent heat of melting, the applied amperage, voltage, shielding gas
composition, or electrode vertex angle.23 Consequently, since these parameters cannot readily
be obtained from the literature, the following reasonable values for r\m and r\ have been assumed to calculate nw in Table 1.6, based on a preevaluation of the experimental data: j \ m =
0.12 (fusion zone), TI = 0.23 (HAZ).
It should be noted that the above values also include a correction for threedimensional heat
flow, since the assumption of a fast moving high power source during low heat input GTA
welding is not valid. Hence, both the arc efficiency factor and the melting efficiency factor
used in the present case study are seen to be lower than those commonly employed in the
literature.
Full weaving (welds Bl and B2)
The results from the metallographic examination of the two GTA welds deposited under full
weaving conditions are presented graphically in Fig. 1.63. Note that the shape of the fusion
boundary as well as the Ac3 and the Ac1 isotherms can be predicted adequately from the present
model for both combinations of E and L (an exception is the HAZ end points in position z = 0),
provided that proper adjustments of i\m and Tj are made. The good correlation obtained in Fig.
1.63 between the observed and the calculated peak temperature contours justifies the adaptation of the model to low heat input processes such as GTA welding, despite the fact that the
assumption of twodimensional heat flow is not valid under the prevailing circumstances.
No weaving (welds B3 and B4)
For the limiting case of no weaving (Fig. 1.64), the concept of an equivalent amplitude of
weaving has been used in order to calculate the peak temperature contours from the model.
This parameter (designated Leq) takes into account the effects of convectional heat flow in the
weld pool on the resulting bead geometry, and is evaluated empirically from measurements of
the actual weld samples. At low heat inputs (Fig. 1.64(a)), the agreement between theory and
y, mm
Weld B1
z, mm
Fusion line
y, mm
Weld B2
z, mm
Fig. 1.63. Comparison between observed and predicted fusion line, Ac3 and Ac1 contours during GTA
welding under full weaving conditions (Case study 1.3). Data from Grong and Christensen.19
experiments is largely improved by inserting 2Leq = 7.5mm into equation (1110), when comparison is made on the basis of the point source model. In contrast, at a heat input of 2.5 kJ
mm"1 (Fig. 1.64(b)), the measured shape of the HAZ isotherms is seen to approach that of a
semicircle, and hence the deviation between the present model and the simplified solution for
a fast moving high power point source is less apparent.
Intermediate weaving
At intermediate amplitudes of weaving (2L = 5 and 7.5mm, respectively), convectional heat
flow in the weld pool will also tend to increase the bead width to depth ratio beyond the
theoretical value predicted from the present model, as shown in Fig. 1.65. The plot in Fig. 1.65
Weld B3
y, mm
z, mm
Fusion line
y, mm
Weld B4
z, mm
Fig. 1.64. Comparison between observed and predicted fusion line, Ac3 and Ac1 contours during GTA
welding with a stationary arc (Case study 1.3). Solid lines: Distributed heat source, Broken lines: Point
heat source. Data from Grong and Christensen.19
includes all data obtained in the GTA welding experiments with an oscillating arc, as reported
by Grong and Christensen.19 These results suggest that the applied amplitude of weaving must
be quite large before such effects become negligible. Consequently, adaptation of the model to
the weld series considered above would require an empirical calibration of the weaving amplitude similar to that performed in Fig. 1.64 for stringer bead weldments to ensure satisfactory
agreement between theory and experiments.
1.10.6 Thermal conditions during interrupted welding
Rapid variations of temperatures as a result of interruption of the welding operation can have
an adversely effect on the microstructure and consequently the mechanical properties of the
weldment.
Width
Depth
Theoretical
curve
Fig. 1.65.Comparison between observed and predicted weld width to depth ratios during GTA welding
with an oscillating arc (Case study 1.3). Data from Grong and Christensen.19
T,e
t,t
Fig. 1.66. Idealised heat flow model for prediction of transient temperatures during interrupted welding.
The situation existing after arc extinction may be described as shown in Fig. 1.66. From time
t = t* there is no net heat supply to the weldment. This condition is satisfied if the real source
q0 is considered maintained by adding an imaginary source +qo and sink qo of the same
strength at t*. The temperature at some later time t** in a given position R0 (measured from the
origin 0") is then equal to the difference of temperatures due to the positive heat sources qo
and +qo and the negative heat sink qo. Each of these temperature contributions will be a
product of a pseudosteady state temperature Tps, and a correction factor K1 or K2 (given by
equations (149) and (182), respectively). Hence, for 3D heat flow, we have:
(1116)
where
and
Similarly, for 2D heat flow, we get:
(1117)
where
(ro is the position of the weld with respect to the imaginary heat source at time f* in the x y
plane).
Example (1.14)
Consider repair welding of a heavy steel casting with covered electrodes under the following
conditions:
Suppose that a 50mm long bead is deposited on the top of the casting. Calculate the temperature in the centre of the weld 5 s after arc extinction.
Solution
The pseudosteady state temperature for points located on the weld centreline (i/ = = 0) can
be obtained from equation (165). When t**  t* = 5 s, we get:
Referring to Fig. 1.67, the position of the weld with respect to the imaginary heat source at
time f* is 10mm, which gives:
(a)
3D heat flow
(b)
3D heat flow
(C)
3D heat flow
Fig. /.6#. Recommended correction factor/for some joint configurations; (a) Single Vgroove, (b) Double Vgroove, (c) Tjoint.
and
At these coordinates, the correction factor K1 is seen to be 1 and 0.62, respectively (Fig.
1.18). The temperature in the centre of the weld 5 s after arc extinction is thus:
which is equivalent to
LlOJ
During conventional beadonplate welding the angle of heat conduction is equal to 180 due
to symmetry effects (e.g. see Fig. 1.23). In order to apply the same heat flow equations during
root pass welding, it is necessary to introduce a correction factor,/, which takes into account
variations in the effective heat diffusion area due to differences in the joint geometry. Taking
/equal to 1 for ordinary beadonplate welding (b.o.p.), we can define the net heat input of a
groove weld as:9
(1118)
Recommended values of the correction factor/for some joint configurations are given in
Fig. 1.68.
Example (Ll5)
Consider deposition of a root pass steel weld in a doubleVgroove with covered electrodes
(SMAW) under the following conditions:
Calculate the cooling time from 800 to 5000C (Ar875), and the cooling rate (CR.) at 6500C
in the centre of the weld when the groove angle is 60.
Solution
The cooling time, Ar875, and the cooling rate, CR., can be obtained from equation (168) and
(171), respectively:
Cooling time, Atm
The above calculations show that the thermal conditions existing in root pass welding may
deviate significantly from those prevailing during ordinary stringer bead deposition due to
differences in the effective heat diffusion area. These results are in agreement with general
experience (see Fig. 1.69).
1.10.8 Semiempirical methods for assessment of bead morphology
In fusion welding fluid flow phenomena will have a strong effect on the shape of the weld
pool. Since flow in the weld pool is generally driven by a combination of buoyancy, electromagnetic, and surface tension forces (e.g. see Fig. 3.10 in Chapter 3), prediction of bead morphology from first principles would require detailed consideration of the current and heat flux
distribution in the arc, the interaction of the arc with the weld pool free surface, convective
heat transfer due to fluid flow in the liquid pool, heat of fusion, convective and radiative losses
from the surface, as well as heat and mass loss due to evaporation.
Over the years, a number of successful studies have been directed towards numerical weld
pool modelling, based on the finite difference, the finite element, or the control volume approach.24"31 Although these studies provide valuable insight into the mechanisms of weld pool
development, the solutions are far too complex to give a good overall indication of the heatand fluidflow pattern. The present treatment is therefore confined to a discussion of factors
affecting the nominal composition of singlebead fusion welds. This composition can be obtained from an analysis of the amount of deposit D and the fused part of the base material B,
from which we can calculate the mixing ratio BI(B + D) or DI(B + D). Methods have been
outlined in the preceding sections for handling such problems by means of point or line source
models. The following section gives a brief description of procedures which can be used for
predictions of the desired quantities in cases where the classic models break down, or where
the calculation will be too tedious.
1.10.8.1 Amounts of deposit and fused parent metal
The heat conduction theory does not allow for the presence of deposited metal. The rate of
deposition, dMwldt, is roughly proportional to the welding current /, and is often reported as a
coefficient of deposition, defined as:
(1119)
Since the area of deposited metal D is frequently wanted, we may write:
(1120)
where p is the density, and v is the welding speed.
Plate thickness:
Consider stringer bead deposition (S AW) on a thick plate of low alloy steel under the following conditions:
Table 1.7 Average rates of volume deposition in arc welding. Data from Christensen.32
Welding Process
SMAW
0.30.5
GMAW, steel
0.60.7
GMAW, aluminium
0.9
SAW, steel
0.7
The amount of fused parent material can be obtained from equation (175). If we include an
empirical correction for the latent heat of melting, the dimensionless radius vector a4m becomes:
This gives:
Similarly, the amount of deposited metal can be calculated from equation (1120). Taking
ifc'/p equal to 0.7mm3 A"1 s"1 for SAW (Table 1.7), we get:
Example (L 17)
Consider stringer bead deposition with covered electrodes (SMAW) on a thick plate of low
alloy steel under the following conditions:
In this particular case the conditions for a fast moving high power source are not met. Thus, in
order to eliminate the risk of systematic errors, the amount of fused parent metal should be
calculated from the general Rosenthal thick plate solution (equation (145)) or read from Fig.
1.21. When Tc = Tm (i.e. 8^ = 1), we obtain:
and
Moreover, the amount of deposited metal can be calculated from equation (1120). Taking
k'/p equal to 0.4mm3 A"1 s~l for SMAW (Table 1.7), we get:
and
The above calculations indicate a small difference in the mixing ratio between SA and
SMA welding, but the data are not conclusive. In practice, a value of BI(B + D) between 1/3
and 1/2 is frequently observed for SMAW, while the mixing ratio for SAW is typically 2/3 or
higher. The observed discrepancy between theory and experiments arises probably from difficulties in estimating the amount of fused parent metal from the point heat source model.
1.10.8.2 Bead penetration
It is a general experience in arc welding that the shape of the fusion boundary will depart quite
strongly from that of a semicircle due to the existence of highvelocity fluid flow fields in the
weld pool.24"31 For combinations of operational parameters within the normal range of arc
welding, a fair prediction of bead penetration h can be made from the empirical equation
derived by Jackson et al.:33
(1121)
A summary of Jackson's data is shown in Table 1.8. It is seen that the constant C in equation (1121) has a value close to 0.024 for SAW and SMAW with E6015 type electrodes, and
about 0.050 for GMAW with CO2 shielding gas. Penetration measurements of GMA/Ar + O2,
GMA/Ar, and GMA/He welds, on the other hand, show a strong dependence of polarity, and
shielding gas composition, to an extent which makes the equation useless for a general prediction. Such data have therefore not been included in Table 1.8.
Example (1.18)
Based on the Jackson equation (equation (1121)), calculate the bead penetration for the two
specific welds considered in Examples (1.16) and (1.17). Use these results to evaluate the
applicability of the point heat source model under the prevailing circumstances.
Solution
Table 1.8 Recommended bead penetration coefficients for some arc welding processes. Data from
Jackson.33
Welding Process
Comments
SAW, steel
0.024
SMAW, steel
(E6015)
0.024
GMAW, steel
(CO2  shielding)
0.050
Electrode positive
(h from 6.5 to 8mm)
and
The corresponding values predicted from the point heat source model are:
and
Provided that the Jackson equation gives the correct numbers, it is obvious from the above
calculations that the point heat source model is not suitable for reliable predictions of the bead
penetration during arc welding. This observation is not surprising.
1.10.9 Local preheating
So far, we have assumed that the ambient temperature T0 remains constant during the welding
operation (i.e. is independent of time). The use of a constant value of T0 is a reasonable approximation if the workpiece as a whole is subjected to preheating. In many cases, however,
the dimensions of the weldment allow only preheating of a narrow zone close to the weld.
This, in turn, will have a significant influence on the predicted weld cooling programme, particularly in the low temperature regime where the classic models eventually break down when
T approaches T0.
Model (after Christensen n)
The idealised preheating model is shown in Fig. 1.70. Here it is assumed that the weld centreline temperature is equal to the sum of the contributions from the arc and from the field of
preheating. The former contribution is given by equation (145) for R = x = Vt, provided that
the plate thickness is sufficiently large to maintain 3D heat flow. Similarly, the temperature
field due to preheating can be calculated as shown in Section 1.7 for uniaxial heat conduction
from extended sources (thermit welding). By combining equations (145) and (122), we
obtain the following relation for the weld centreline:
(1122)
Temperature profile
att = 0
Weld
Preheated
zone
z
Fig. 1.70. Sketch of preheating model.
where T* is the local preheating temperature, and L* is the half width of the preheated zone.
Equation (1122) can be written in a general form by introducing the following groups of parameters:
Dimensionless temperature:
(1123)
Time constant:
(1124)
Dimensionless time:
(1125)
It is evident from the graphical representation of equation (1127) in Fig. 1.71 that the
predicted weld cooling programme falls within the limits calculated for Q"^ 0 (no preheating)
and Q /7 > oo (global preheating). The controlling parameter is seen to be the dimensionless
half width of the preheated zone Q", which depends both on the actual width L*, the base plate
thermal properties a, X, and the net heat input qo Iv.
Example (1.19)
Consider stringer bead deposition with covered electrodes (SMAW) on a thick plate of low
alloy steel under the following conditions:
Calculate the cooling time from 800 to 50O0C (A%5), and the cooling time 1Oo measured
from the moment of arc passage to the temperature in the centre of the weld reaches 10O0C.
Solution
First we calculate the time constant to from equation (1124):
e*
^6
Next Page
Cooling time, At8/5
from which
This cooling time is only slightly longer than that calculated from equation (168) for T0 =
200C (6.9s), showing that moderate preheating up to 1000C is not an effective method of
controlling Ar875.
Cooling time, t]00
When T=T0* = 1000C, the dimensionless temperature 9* = 1. Reading from Fig. 1.71 gives
T 6  10, from which:
The above value should be compared with that evaluated from the numerical data of Yurioka
et al.,35 replotted in Fig. 1.72 (see p.104). It follows from Fig. 1.72 that the weld cooling
programme in practice is also a function of the plate thickness d, an effect which cannot readily be accounted for in a simple analytical treatment of the heat diffusion process. For the
specific case considered above the parameter ^100 varies typically from 500 to 900s, depending
on the chosen value of d. This cooling time is significantly shorter than that calculated from
equation (1127), indicating that the analytical model is only suitable for qualitative predictions.
References
1.
2.
3.
4.
5.
6.
H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids; 1959, Oxford, Oxford University
Press.
British Iron and Steels Research Association: Physical Constants of some Commercial Steels
at Selected Temperatures; 1953, London, Butterworths.
R. Hultgren, R.L. Orr, RD. Anderson and K.K. Kelly: Selected Values of Thermodynamic
Properties of Metals and Alloys; 1963, New York, J. Wiley & Sons.
E. Griffiths (ed.): J. Iron and Steel Inst., 1946,154, 83121.
J.E. Hatch (ed.): Aluminium Properties and Physical Metallurgy; 1984, Metals Park (Ohio),
American Society for Metals.
Metals Handbook, 9th edn., Vol. 2, 1979, Metals Park (Ohio), American Society for Metals.
Previous Page
Cooling time, At8/5
from which
This cooling time is only slightly longer than that calculated from equation (168) for T0 =
200C (6.9s), showing that moderate preheating up to 1000C is not an effective method of
controlling Ar875.
Cooling time, t]00
When T=T0* = 1000C, the dimensionless temperature 9* = 1. Reading from Fig. 1.71 gives
T 6  10, from which:
The above value should be compared with that evaluated from the numerical data of Yurioka
et al.,35 replotted in Fig. 1.72 (see p.104). It follows from Fig. 1.72 that the weld cooling
programme in practice is also a function of the plate thickness d, an effect which cannot readily be accounted for in a simple analytical treatment of the heat diffusion process. For the
specific case considered above the parameter ^100 varies typically from 500 to 900s, depending
on the chosen value of d. This cooling time is significantly shorter than that calculated from
equation (1127), indicating that the analytical model is only suitable for qualitative predictions.
References
1.
2.
3.
4.
5.
6.
H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids; 1959, Oxford, Oxford University
Press.
British Iron and Steels Research Association: Physical Constants of some Commercial Steels
at Selected Temperatures; 1953, London, Butterworths.
R. Hultgren, R.L. Orr, RD. Anderson and K.K. Kelly: Selected Values of Thermodynamic
Properties of Metals and Alloys; 1963, New York, J. Wiley & Sons.
E. Griffiths (ed.): J. Iron and Steel Inst., 1946,154, 83121.
J.E. Hatch (ed.): Aluminium Properties and Physical Metallurgy; 1984, Metals Park (Ohio),
American Society for Metals.
Metals Handbook, 9th edn., Vol. 2, 1979, Metals Park (Ohio), American Society for Metals.
Heat input:
E=1.7 kJ/mm
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
K.P. Bentley, J.A. Greenwood, R McKnowlson and R.G. Bakes: Brit. Weld. J., 1963,10, 613619.
N.N. Rykalin, A.I. Pugin and V.A. Vasil'eva: Weld. Prod., 1959, 6, 4252.
N.N. Rykalin: Berechnung der Warmevorgdnge beim Schweissen; 1953, Berlin, VEB Verlag
Technik.
H.M. Tensi, W. Welz and M. Schwalm: Aluminium, 1981, 58, 515518.
N. Christensen, V. de L. Davis and K. Gjermundsen: Brit. Weld. J., 1965,12, 5475.
Welding Handbook, 8th edn., Vol. 2, 1991, Miami (Florida), American Welding Society.
D. Rosenthal: Weld. / , 1 9 4 1 , 20, 220s234s.
D. Rosenthal: Trans. ASME, 1946, 68, 849866.
CM. Adams: Weld. J., 1958, 37, 210s215s.
RS. Myers, O.A. Uyehara and G.L. Borman: Weld. Res. Bull., 1967, 123, 146.
T.W Eagar and N.S. Tsai: Weld. J., 1983, 62, 346s355s.
M.F. Ashby and K.E. Easterling: Ada Metall, 1984, 32, 19351948.
0. Grong and N. Christensen: Mater. ScL Tech., 1986, 2, 967973.
O.R. Myhr and 0 . Grong: Acta Metall. Mater., 1990, 38, 449460.
S. Kou and Y. Le: Metall. Trans., 1983,14A, 22452253.
22.
23.
24.
25.
26.
27.
28.
29.
30.
32.
33.
34.
35.
O.R. Myhr and 0. Grong: Unpublished work, 1990, University of Trondheim, The Norwegian
Institute of Technology.
R.W. Niles and CE. Jackson: Weld. J., 1975, 54, 25s32s.
G.M. Oreper, T.W. Eagar and J. Szekely: Weld J., 1983, 62, 307s312s.
Y.H. Wang and S. Kou: Proc. Int. Conf. on Trends in Welding Research, Gatlinburg, TN, May,
1986, pp. 6569, Publ. ASM International.
S.A. David and J.M Vitek: Int. Mater. Rev., 1989, 34, 213245.
K.C. Mills and BJ. Keene: Int. Mater. Rev., 1990, 35, 185216.
R.L. UIe, Y. Joshi and E.B. Sedy: Metall. Trans., 1990, 21B, 10331047.
T. Zacharia, S.A. David, J.M. Vitek and H.G. Kraus: Metall. Trans., 1991, 22B, 243257.
A. Matsunawa: Proc. 3rd Int. Conf. on Trends in Welding Research, Gatlinburg, TN, 1992,
pp.316, Publ. ASM International.
N. Christensen: Welding Metallurgy Compendium, 1985, University of Trondheim, The Norwegian Institute of Technology.
CE. Jackson: Weld. J., 1960, 39, 226s230s.
O.M. Akselsen and G. Sagmo: Technical Report STF34 A89147, 1989, Trondheim (Norway),
SintefDivision of Metallurgy.
N. Yurioka, M. Okumura, S. Ohshita and S. Saito: HW Doc. XIIE1081, 1981.
Appendix 1.1
Nomenclature
General symbols
thermal diffusivity (mm2 s"1)
integration parameter
integration parameter
integration parameter
dimensionless temperature
temperature (0C)
thermal conductivity
(W mm"1 0C1)
dimensionless time
time (s)
Specific symbols
0
dimensionless time
dimensionless time
locus of peak temperature in 0T1
space
Spot Welding
thickness of overlapping plates
(mm)
dimensionless operating parameter
Friction Welding
integral in equation (130)
dimensionless rvector
dimensionless time
dimensionless temperature
Thermit Welding
dimensionless jccoordinate
locus of peak temperature in 6"Q.' space
dimensionless time
Arc Welding
amount of fused parent metal
(mm2)
constant in Jackson equation
amount of deposited metal (mm2)
gross heat input per unit length of
weld (kJ mm"1)
correction factor for the net heat
input during root pass welding
FCAW
SAW
SMAW
(S)
groove angle
dimensionless jcaxis
dimensionless zo axis
dimensionless xo axis
dimensionless zcoordinate at
maximum depth of isotherm
dimensionless jccoordinate at
maximum width of isotherm
dimensionless yaxis
dimensionless yo axis
dimensionless ^coordinate at
maximum width of isotherm
dimensionless isothermal zone
width
dimensionless zaxis
Appendix 1.2
Refined Heat Flow Model for Spot Welding
The refined model is based on the assumption that all heat is released instantaneously at time
t = 0 in a point located at the interface between the two overlapping plates, which implies that
equation (17) is valid. However, in order to maintain the net flux of heat through both plate
surfaces equal to zero, it is necessary to account for mirror reflections of the source with
respect to the planes z = dt/2 and z =  dt/2. This can be done on the basis of the method of
images, as illustrated in Fig. A 1.1. By including all contributions from the imaginary sources
"Q2 >Qi ,Q\,Qi > located symmetrically at distances idt below and above the centreaxis
of the joint, the temperature distribution is obtained in the form of a convergent series:
(AlI)
where
Z
Fig. Al.l.Refined heat flow model for spot welding of plates.
and i is an integer variable (...1, 0, 1...)A numerical solution of equation (Al.l) gives a peak temperature distribution similar to
that shown in Fig. 1.9.
Appendix 1.3
The Gaussian Error Function
The eiTor function erf(u) and the complementary error function erfc(u) are special cases of the
incomplete gamma function. Their definitions are:
and
and
The following Fortran subroutine can be used for calculations of the error functions with a
fractional error less than 1.2 X 10~7:
FUNCTION ERFC(U)
Z=ABS(U)
T=l./(l.+0.5*Z)
ERFC=T*EXP(Z*Z1.26551223+T*(1.00002368+T*(.37409196+
*
T%09678418+T*(.18628806+T%27886807+T*(l.13520398+
T*(1.48851587+T*(.82215223+T*.17O87277)))))))))
IF (U.LT.O.) ERFC=2.ERFC
RETURN
END
Appendix 1.4
Gaussian Heat Distribution
Following the treatment of Rykalin,9 the situation may be described as shown in Fig. Al.2.
Here we consider a distributed heat source of net power density (in W mm"1):
(Al2)
The total power of the source qo is obtained by integration of equation (Al2). Substituting
and integrating from u =  to u = +<*>, gives:
from which
(Al3)
It follows from equation (173) that an infinitesimal source dqy> located between j ' a n d
will cause a small rise of temperature dTy> in point P at time t, as:
(Al4)
where
and
Integration of equation (Al4) between the limits y'= o and >>'= +00 gives:
(Al5)
q(y)
z
Fig, Al.2. Distributed heat source of net power density q(y) on a semiinfinite body.
where
from which
(Al6)
If we replace the Gaussian heat distribution by a linear source of the same strength, which
extends from L0 to +L0 on either side of the weld centreline in the ydirection (see Fig. A1.3),
we may write:
(Al7)
In practice, the parameter L0 has the same physical significance as the weaving amplitude
L in equation (1110). Consequently, these solutions are equivalent in the sense that they
predict a similar temperaturetime pattern.
q(y)
Fig. Al.3. Physical representation of a Gaussian heat distribution by a linear source of width 2LO.
2
Chemical Reactions in Arc Welding
2.1 Introduction
The weld metal composition is controlled by chemical reactions occurring in the weld pool at
elevated temperatures, and is therefore influenced by the choice of welding consumables (i.e.
combination of filler metal, flux, and/or shielding gas), the base metal chemistry, as well as the
operational conditions applied. In contrast to ladle refining of metals and alloys where the
reactions occur under approximately isothermal conditions, a characteristic feature of the arc
welding process is that the chemical interactions between the liquid metal and its surroundings
(arc atmosphere, slag) take place within seconds in a small volume where the metal temperature gradients are of the order of 1000C mm"1 with corresponding cooling rates up to
1000C s"1. The complex thermal cycle experienced by the liquid metal during transfer from
the electrode tip to the weld pool in GMA welding of steel is shown schematically in Fig. 2.1.
As a result of this strong nonisothermal behaviour, it is very difficult to elucidate the
reaction sequences during all stages of the process. Consequently, a complete understanding
of the major controlling factors is still missing, which implies that fundamentally based predictions of the final weld metal chemical composition are limited. Additional problems result
from the lack of adequate thermodynamic data for the complex slagmetal reaction systems
involved. However, within these restrictions, the development of weld metal compositions
can be treated with the basic principles of thermodynamics and kinetic theory considered in
the following sections.
(ii)
A cooling stage, where the concentrations established during the initial stage
tend to readjust by rejection of dissolved elements from the liquid.
Gas nozzle
Shielding gas
Filler wire
Contact tube
Arc plasma
temperature~10000C
Cold part of
weld pool (< 19000C)
Weld pool
retention
time 21Os
Base plate
Fig. 2.1. Schematic diagram showing the main process stages in GMA welding. Characteristic average
temperature ranges at each stage are indicated by values in parenthesis.
As indicated in Fig. 2.2 the high temperature stage comprises both gas/metal and slag/metal
interactions occurring at the electrode tip, in the arc plasma, or in the hot part of the weld pool,
and is characterised by extensive absorption of elements into the liquid metal. During the
subsequent stage of cooling following the passage of the arc, a supersaturation rapidly increases because of the decrease in the element solubility with decreasing temperatures. The
system will respond to this supersaturation by rejection of dissolved elements from the liquid,
either through a gas/metal reaction (desorption) or by precipitation of new phases. In the latter
case the extent of mass transfer is determined by the separation rate of the reaction products in
the weld pool. It should be noted that the boundary between the two stages is not sharp, which
means that phase separation may proceed simultaneously with absorption in the hot part of the
weld pool.
In the following sections, the chemistry of arc welding will be discussed in the light of this
twostage reaction model.
Peak temperature
Grey j
zonei
Rejection of
dissolved elements
Peak concentration
Concentration
Absorption
of elements
Temperature
'Hot1 part of
weld pool
Time
Fig. 2.2. Idealised twostage reaction model for arc welding (schematic).
Table 2.1 Temperature for 90% dissociation of some gases in the arc column. Data from Lancaster.2
Gas
CO 2
3800
H2
4575
O2
5100
N2
8300
Next, consider a shielding gas which consists of two components, i.e. one inert component
(argon or helium) and one active component X2. When the fraction dissociated is close to
unity, the partial pressure of species X in the gas phase px is equal to:
(22)
where H1 and nx are the total number of moles of components / (inert gas) and X, respectively
in the shielding gas, andptot is the total pressure (in atm).
It follows from equation (21) that two moles of X form from each mole of X2 that dissociates. Hence, equation (22) can be rewritten as:
(23)
where nXl is the total number of moles of component X2 which originally was present in the
shielding gas.
If nXl and H1 are proportional to the volume concentrations of the respective gas components in the shielding gas, equation (23) becomes:
(24)
Taking vol% / = (100  vol% X2) andp,ot = 1 atm, we obtain the following expression for
Px(25)
It is evident from the graphical representations of equations (25) and (27) in Fig. 2.3 that
the partial pressure of the dissociated component X increases monotonically with increasing
concentrations of X2 and YX2 in the shielding gas. The observed nonlinear variation of px
arises from the associated change in the total number of moles of constituent species in the gas
phase due to the dissociation reaction. Moreover, it is interesting to note that the partial pressure px is also dependent on the nature of the active gas component in the arc column (i.e. the
stoichiometry of the reaction). This means that the oxidation capacity of for instance CO2 is
only half that of O2 when comparison is made on the basis of equal concentrations in the
shielding gas (to be discussed later).
Px
Vol%)^ f VoRGYX2
Fig. 2.3. Graphical representation of equations (25) and (27).
(ii)
(iii)
Transport of dissolved elements from the interface to the bulk of the metal.
(28)
where Dx is the diffusion coefficient of the transferring species X (in mm2 s~*).
Although the validity of equation (28) may be questioned, the thin film model provides a
simple physical picture of the resistance to mass transfer during gas absorption.
Partial pressure
Distance
Fig. 2.4. Film model for mass transfer (schematic).
(210)
It follows from equation (210) that the transient concentration of element X in the hot part
of the weld pool is proportional to the partial pressure of the dissociated component X in the
plasma gas. Since this partial pressure is related to the initial content of the molecular species
X2 or YX2 in the shielding gas through equations (25) and (27), we may write:
Arc column
Bulk gas
phase
Stagnant gaseous
boundary layer
Gas/metal interface
Metal phase
Hot part of
weld pool
Fig. 2.5. Idealised kinetic model for gas absorption in arc welding (schematic).
(211)
and
(212)
where C1 and C2 are kinetic constants which are characteristic of the reaction systems under
consideration.
X (dissolved)
(213)
By introducing the equilibrium constant K{ for the reaction and setting the activity coefficient to unity, we get:
(214)
This equation should be compared with equation (210) which predicts a linear relationship
between wt% X and px. If the above analysis is correct, one would expect that the partial
pressure px at the gas/metal interface is directly proportional to the partial pressure of the
dissociated component in the bulk phase. Unfortunately, the proportionality constant is difficult to establish in practice.
(215)
where k'd is the mass transfer coefficient (in mm s 1X and px is the equilibrium partial pressure of component X2 at the gas/metal interface (in atm).
Bulk gas
phase
Stagnant gaseous
boundary layer
Gas/metal interface
Metal phase
Cold part of
weld pool
Fig. 2.6. Idealised kinetic model for gas desorption in arc welding (schematic).
The partial pressure pX2 can be calculated from chemical thermodynamics by considering
the following reaction:
2X(dissolved) = X2 (gas)
(216)
from which
(217)
where K2 is the equilibrium constant, and [wt% X] is the concentration of element X in the
liquid metal (in weight percent). Note that the activity coefficient has been set to unity in the
derivation of equation (217).
The equilibrium constant K2 may be expressed in terms of the solubility of element X in the
liquid metal at 1 atm total pressure Sx. Hence, equation (217) transforms to:
(218)
(219)
Data for the solubility of hydrogen and nitrogen in some metals up to about 22000C are
given in Figs. 2.7 and 2.8, respectively. It is evident that the element solubility decreases
steadily with decreasing metal temperatures down to the melting point. This implies that the
desorption reaction is thermodynamically favoured by the thermal conditions existing in the
cold part of the weld pool.
2.6.2. Sievert's law
It follows from equation (219) that desorption becomes kinetically unfeasible when
Px2 ~ Px2' corresponding to:
(220)
Equation (220) is known as the Sievert's law. This relation provides a basis for calculating
the final weld metal composition in cases where the resistance to mass transfer is sufficiently
small to maintain full chemical equilibrium between the liquid metal and the ambient (bulk)
gas phase.
2.7 Overall Kinetic Model for Mass Transfer during Cooling in the Weld Pool
Because of the complexity of the rate phenomena involved, it would be a formidable task to
derive a complete kinetic model for mass transfer in arc welding from first principles. How
(b)
Aluminium
(a)
Temperature, 0C
Solid Cu
Temperature, 0C
Iron
Temperature, 0C
(d)
(C)
Copper
Nickel
Temperature, 0C
Fig. 2.7. Solubility of hydrogen in some metals; (a) Aluminium, (b) Copper, (c) Iron, (d) Nickel. Data
compiled by Christensen.4
ever, for the idealised system considered in Fig. 2.9, it is possible to develop a simple mathematical relation which provides quantitative information about the extent of element transfer
occurring during cooling in the weld pool. Let [%X]eq denote the equilibrium concentration of
element X in the melt. If we assume that the net flux of element X passing through the phase
boundary A per unit time is proportional to the difference ([%X]  [%X]eqX the following
balance is obtained:3
(221)
where V is the volume of the melt (in mm 3 ), kd is the overall mass transfer coefficient (in
mm s"1), and A is the contact area between the two phases (in mm 2 ).
Temperature, 0C
log (wt% N)
Iron
104AT1 K
Fig. 2.8. Solubility of nitrogen in iron. Data from Turkdogan.5
Phase I i
Phase i
Distance
Net flux of X
Volume (V)
Concentration
Fig. 2.9. Idealised kinetic model for mass transfer in arc welding (schematic).
By rearranging equation (221) and integrating between the limife [%X]( (att = O) and [%X]
(at an arbitrary time t\ we get:
(222)
It is evident from the graphical representation of equation (222) in Fig. 2.10 that the rate of
mass transfer depends on the ratio Vl kji, i.e. the time required to reduce the concentration of
element X to a certain level is inversely proportional to the mass transfer coefficient kd. This
type of response is typical of a first order kinetic reaction.
Although the above model refers to mass transfer under isothermal conditions, it is also
applicable to welding if we assume that the weld cooling cycle can be replaced by an equivalent isothermal holdup at a chosen reference temperature. Thus, by rearranging equation (222), we get:
(223)
It follows that the final concentration of element X in the weld metal depends both on the
cooling conditions and on the intrinsic resistance to mass transfer, combined in the ratio t/to.
When [%X]eq is sufficiently small, equation (223) predicts a direct proportionality between
[%X] and [%X\t (i.e. the initial concentration of element X in the weld pool). This will be the
case during deoxidation of steel weld metals where separation of oxide inclusions from the
weld pool is the rate controlling step. Moreover, when t/t0 1 (small resistance to mass
transfer), equation (223) reduces to:
(224)
(XX^)Z(X1Xeq)
Under such conditions the final weld metal composition can be calculated from simple
chemical thermodynamics.
Because of this flexibility, equation (223) is applicable to a wide range of metallurgical
problems at the same time as it provides a simple physical picture of the resistance to mass
transfer during cooling in the weld pool.
t,s
Fig. 2.10. Graphical representation of equation (222).
FCAW
Very Low
Medium
low
Weld hydrogen level
High
Fig. 2.11. Ranking of different welding processes in terms of hydrogen level (schematic). The diagram is
based on the ideas of Coe.6
(i)
The Japanese method (JIS Z 3131975), which has been adopted with important adjustments from the former ASTM designation A31648T. This method involves collection of
released hydrogen from a single pass weld above glycerine for 48h at 45 0 C. The total volume
of hydrogen is reported in ml per 10Og deposit. Only 5 s of delay are allowed from extinction
of the arc to quenching.
(ii) The French method (N.F.A. 813051975) where two beads are deposited onto core
wires placed in a copper mould. Hydrogen released from this bead is collected above mercury,
and the volume is reported in ml per 10Og fused metal (including the fused core wire metal).
(iii) The International Institute of Welding (HW) method (ISO 36901977), where a single
bead is deposited on previously degassed and weighed mild steel blocks clamped in a quickrelease copper fixture. The weldment is quenched and refrigerated according to a rigorously
specified time schedule. Hydrogen released from the specimens is collected above mercury
for 72 h at 25C, and the results are reported in ml per 10Og deposit, or in g per ton fused metal.
To avoid confusion, it is recommended to use the symbol HDM for the content reported in terms
of deposited metal (ml per 10Og deposit), and HFM for the content referred to fused metal (ml
per 100 g or g per ton fused metal). The relationship between HDM and HFM is shown in Fig.
2.12.
As would be expected, these three methods do not give identical results when applied to a
given electrode. Approximate correlations have been established between the HW criteria
HDM and HFM and the numbers obtained by the Japanese and the French methods (designated
HJIS and HFR, respectively). For covered electrodes tested at various hydrogen levels, we
have:7
Fig. 2.12. The relation between HDM and HFM (0.9 is the conversion factor from ml per 10Og to g per ton).
(225)
(226)
The conversion factor from HFR to HFM applies to a ratio of deposited to fused metal,
DI(B + D), equal to 0.6, which is a reasonable average for basic electrodes.
The use of HFM in preference of HDM is normally recommended, because it is a more rational criterion of concentration. Moreover, HDM values would be grossly unfair, if applied to
high penetration processes like submerged arc welding. In GTA welds made without filler
wire HDM cannot be used at all, since there is no deposit.
It should be noted that the present HW procedure gives the amount of 'diffusible hydrogen'.
For certain purposes the total hydrogen content may be wanted. It is obtained by adding the
content of 'residual hydrogen' determined on the same samples by vacuum or carrier gas
extraction at 6500C. A very small additional amount may be observed on vacuum fusion of the
sample, tentatively labelled 'fixed hydrogen'. There is no clear line of demarcation between
these categories of hydrogen. As will be discussed later, the extent of hydrogen trapping
depends both on the weld metal constitution and the thermal history of the metal. In singlebead basic electrode deposits the diffusible fraction is usually well above 90%.
2.8.3 Reaction model
Normally, measurements of hydrogen in weld metals are carried out on samples from solidified beads. Due to the rapid migration of hydrogen at elevated temperatures, such data do not
represent the conditions in the hot part of the weld pool. Quenched end crater samples would
be better in this respect, but they are not representative of normal welding. Further complications arise from the presence of hydrogen in different states (e.g. diffusible or residual hydrogen) and the lack of consistent sampling methods.
Nevertheless, experience has shown that pickup of hydrogen in arc welding can be interpreted on the basis of the simple model outlined in Fig. 2.13. According to this model, two
zones are considered:
(i) An inner zone of very high temperatures which is characterised by absorption of atomic
hydrogen from the surrounding arc atmosphere.
Electrode
Hot part of weld pool
Absorption of atomic hydrogen
(controlled by pH in the arc column)
Hydrogen
trapped in
weld metal
Weld pool
Fig. 2.13. Idealised reaction model for hydrogen pickup in arc welding.
(ii) An outer zone of lower temperatures where the resistance to hydrogen desorption is
sufficiently small to maintain full chemical equilibrium between the liquid weld metal and the
ambient (bulk) gas phase.
Under such conditions, the final weld metal hydrogen content should be proportional to the
square root of the initial partial pressure of diatomic hydrogen in the shielding gas, in agreement with Sievert's law (equation (220)).
2.8.4 Comparison between measured and predicted hydrogen contents
It is evident from the data in Table 2.2 that the reported ranges for hydrogen contents in steel
weld metals are quite wide, and therefore not suitable for a direct comparison of prediction
with measurement. For such purposes, the welding conditions and consumables must be more
precisely defined.
2.8.4.1 Gasshielded welding
In GTA and GMA welds the hydrogen content is usually too low to make a direct comparison
between theory and experiments. An exception is welding under controlled laboratory conditions where the hydrogen content in the shielding gas can be varied within relatively wide
limits. The results from such experiments are summarised in Fig. 2.14, from which it is seen
that Sievert's law indeed is valid. A closer inspection of the data reveals that the weld metal
hydrogen content falls within the range calculated for chemical equilibrium at 1550 and 20000C,
depending on the applied welding current. This shows that the effective reaction temperature
is sensitive to variations in the operational conditions.
An interesting effect of oxygen on the weld metal hydrogen content has been reported by
Matsuda et al.9 Their data are reproduced in Fig. 2.15. It is evident that the hydrogen level is
significantly higher in the presence of oxygen. This is probably due to the formation of a thin
(protective) layer of slag on the top of the bead, which kinetically suppresses the desorption of
hydrogen during cooling.
Table 2.2 Measured arc atmosphere compositions in steel welding. Also included are typical ranges for
the weld metal hydrogen content. Data compiled by Christensen.4
Arc Atmosphere Composition
(vol%)
Method
Primary Source
of Hydrogen
Weld Metal
Hydrogen
Content (ppm)
CO2
CO
H 2 +H 2 O
Range
Average
9880
220
<0.02
15
GMAW*
(CO2)
Moisture introduced
through the shielding gas
SMAW
(acid)
4
34
62
1030
25
SMAW
(rutile)
~4
42
54
1030
25
SMAW
(basic)
19
77
4
210
35
FCAW
(rutile)
1020
FCAW
(basic)
25
SAW
(basic)
210
*The arc atmosphere composition can vary within wide limits, depending on the operational conditions applied.
Consider GTA welding (Arshielding) on a thick plate of lowalloy steel under the following
conditions:
/ = 200A, U = 15V, v = 3 mm s"1, TI = 0.5, T0 = 20C
The shielding gas contains 0.1 vol% moisture (H2O) and is supplied at a rate of 15Nl mhr 1 .
Calculate the 'potential' hydrogen level, assuming that all hydrogen introduced through the
shielding gas is absorbed in the weld metal.
Solution
The resulting bead cross section and total mass of weld metal per mm can be estimated
from the Rosenthal equation by considering the dimensionless operating parameter at the melting
point (equation (150)):
It is evident from the above calculations that the 'potential' hydrogen level is at least one
order of magnitude higher than the expected weld metal hydrogen content (1 to 3 ppm). This
shows that the hydrogen pickup in GTA welding is not determined by the total amount of
hydrogen which is introduced through the shielding gas, but is mainly controlled by the resulting partial pressure of hydrogen in the ambient (bulk) gas phase.
2.8.4.2 Covered electrodes
In SMA welding the partial pressure of hydrogen is more difficult to assess due to the presence
of trapped moisture and hydrogencontaining compounds in the electrode coating. Such compounds will loose their identity at the stage of introduction into the arc atmosphere. Since very
little information is available on the species present in the arc column, we shall base our estimate on a simple thermodynamic approach, including only the molecular species H2 and H2O
which can be determined by analysis (see data in Table 2.2). It follows that the combined
partial pressure of H2 and H2O in the gas phase is given by:
(227)
The parameter pw can be estimated on the basis of combustion measurements of the electrode coating, assuming that no carbon is picked up or lost from the system in excess of the
amount calculated from an analysis of the base plate and the electrode wire. For a recorded
content of mw g H2O and mc g CO2 per 100 g of electrode coating, we obtain:
(228)
From a thermodynamic standpoint, replacement of pHl b y / ^ in the expression for Sievert's
law requires the use of a modified solubility of hydrogen, defined as:
(229)
where K3 is the equilibrium constant for the H 2 OH reaction, and [%O] is the weld metal
oxygen content. In practice, the correction term ^ K3/(K3+[%0]) does not depart significantly from unity, which means that Sw ~ SH.
During welding with basic covered electrodes considerable amounts of CO2 may form as a
result of decomposition of calcium carbonate, according to the reaction:
(230)
Modern basic electrodes contain between 20 to 40 weight percent CaCO3, which is equivalent with a CO2 content of 9 to 18 percent. Taking as an average mc equal to 15 g CO2 per 100 g
electrode coating, we obtain:
(231)
In Fig. 2.16 the validity of equation (231) has been checked against relevant literature data
(compiled by Chew10). A closer inspection of the data reveals that the weld metal hydrogen
content falls within the range calculated for chemical equilibrium at 1520 to 2000C, taking Sw
equal to the solubility of hydrogen in pure iron at the indicated temperatures (i.e. 27 and 40 ml
H2 per 100 g fused metal, respectively). Although the observed scatter in the effective reaction
temperature is admittedly large, equation (231) points out a very interesting effect, namely
that the hydrogen content of SMA steel weld metals is controlled by the combined partial
pressure of H2 and H2O in the ambient gas phase. For this reason it is frequently recommended that calcium carbonate is added to the electrode coating, which on decomposition
produces considerable amounts of shielding gas in the form of CO2. Hydrogen shielding can
also be achieved by additions of volatile alkalifluorides, which on heating will evaporate and
dilute the atmosphere with respect to hydrogen.
Example (2.2)
Consider SMA welding on mild steel with basic covered electrodes. The electrode coating
contains 35 wt% CaCO3 and 0.5 wt% H2O in the asreceived condition. After drying at 3500C
for 1 h the water content is reduced to 0.2 wt% H2O. Estimate the weld metal hydrogen
content (in ppm) both before and after drying of the electrode. Assume in these calculations an
effective reaction temperature of 18000C.
Solution
First we calculate the CO2 content per 100 g of electrode coating. Taking the atomic weight of
CaCO3 and CO2 equal to 100.1 and 40.0, respectively, we obtain:
The combined partial pressure pw can now be estimated from equation (228). Before
drying we have:
From Fig. 2.7(c) it is evident that the solubility of hydrogen in liquid iron at 18000C is
about 37 ml H2 per 100 g fused metal. This corresponds to a modified solubility Sw (in ppm)
of:
Substituting this value into the expression for Sievert's law gives:
(before drying)
(after drying)
It follows from the above calculations that a low weld metal hydrogen level requires the use
of 'dry' basic electrodes. In practice, this can be achieved by protecting the electrodes against
moisture pickup during storage (see Fig. 2.17). However, in certain cases it is necessary to
differentiate between strongly bound and loosely adsorbed moisture in the coating of basic
electrodes. This point is more clearly illustrated in Fig. 2.18, which shows the HDM content of
hydrogen in basic electrode deposits at various levels of coating moisture. It is seen that water
remaining from an insufficient baking treatment is more dangerous than moisture picked up by
exposure of a properly dried coating. This has to do with the fact that loosely adsorbed mois
very low
low
medium
high
Fig. 2.17. Moisture content in basic electrode coating as a function of exposure time and relative humidity (R.H.) in ambient gas phase. Data from Evans.11
ture will tend to evaporate during the welding operation (before it enters the arc column)
because of resistance heating of the electrode, a process which is not feasible when the water is
bound in rust on the surface of the electrode wire or the iron powder.
2.8.4.3 Submerged arc welding
This method is usually classified as a pure slagshielded process, because carbonates or other
gasproducing compounds are not present in large quantities. A closed arc cavity does exist,
however, as indicated by the falling voltampere curve characteristic of open arcs, and by
observations made by probes inserted through the flux cover.
It is reasonable to assume that the gas contained within this enclosure consists of metal
vapour, volatile constituents originating from the flux, and relatively small fractions of carbon
monoxide and water vapour. Acid fluxes of the calcium silicate type will probably generate
silicon monoxide, while agglomerated fluxes bonded with alkali silicate will produce volatile
alkali fluorides. In addition, carbon monoxide may be present as a result of oxidation of carbon, or decomposition of carbonates.
A small but important contribution to the cavity atmosphere is the trace of moisture remaining in the flux even after careful drying. No direct measurements of partial pressures are available, and the gas composition must therefore be inferred from observations of hydrogen absorption in the weld metal. Hydrogen pickup during SA welding has been examined by Evans
and Bach.12 Their data are replotted in Fig. 2.19. The shape of the observed curve of hydrogen
vs residual water content would seem to indicate a relationship similar to that predicted by
Sievert's law. In fact, a very close fit can be obtained through empirical calibration of the
dilution term in equation (228). This, however, implies unreasonable amounts of CaCO3.
Carbon monoxide in addition to that delivered by carbonates could be formed by oxidation of
carbon. Again, an unreasonable amount of carbon loss would be required. Therefore, it must
be concluded that further research is needed for a proper interpretation of the factors controlling hydrogen pickup in SA welding.
Example (2.3)
Consider SA welding on a thick plate of lowalloy steel under the following conditions:
The flux contains 0.04 wt% H2O and is consumed at a rate of 0.6 g per g weld deposit.
Estimate both 'potential' and 'equilibrium' hydrogen levels when the total oxidation loss of
carbon in the weld pool is 0.03 wt%.
Solution
First we calculate the total amount of fused parent metal and weld deposit formed on welding.
From equations (175) and (1120), we have:
and
When the dilution ratio DI(B + D) is known, it is possible to calculate the total flux consumption per gram fused weld metal:
If we assume that all CO produced by reactions between dissolved carbon and oxygen is
infiltrated in the arc column, the following balance is obtained:
Total number of moles of CO per g fused weld metal:
This gives:
Since the effective reaction temperature of hydrogen absorption in SA welding is not known,
the maximum solubility of hydrogen at 1 atm total pressure is taken equal to 33 ppm, similar to
that in the previous example. By inserting this value in the expression for Sievert's law, we
obtain:
In practice, the 'potential' hydrogen level represents an upper limit for the hydrogen concentration which cannot be exceeded. Thus, the contradictory results obtained in the present
example clearly illustrate the difficulties involved in estimating the effective partial pressure
of hydrogen in SA welding.
2.8.4.4 Implications of Sievert's law
An important implication of Sievert's law is that the fraction of hydrogen picked up from the
arc atmosphere is very high at low hydrogen pressures:
(232)
As seen from equation (232), the first traces of hydrogen added to the atmosphere are
completely absorbed in the metal. At increasing partial pressures the fraction of hydrogen
picked up in the metal will gradually decrease, finally attaining a threshold of (SH/2) in the
case of pure H2. This shows that the concept of 'potential' hydrogen content frequently used to
characterise filler materials (see Fig. 2.11) is a dangerous one, since the rates of absorption are
so different in the high and low ranges of the hydrogen potential.
2.8.4.5 Hydrogen in multirun weldments
So far, no standardised method is available for the determination of hydrogen in multilayer
welds. Early measurements by Roux,13 using an arrangement similar to that subsequently
adopted in French standards, indicate a constant ratio of extracted hydrogen to the mass of
fused metal, regardless of the number of passes. If hydrogen is reported on the basis of deposited metal, this ratio may vary by a factor of 2.5 when comparing a deposit made in five passes
to a single bead.
Exploratory measurements of local hydrogen contents in largesize joints have been made
by Skjolberg,14 who butt welded a 40 mm plate with a selfshielding flux cored wire at an
interpass temperature of 2000C. Samples were cut from a refrigerated part of the weldment at
midthickness, including positions in the weld metal close to the fusion line and samples in the
HAZ. His results are summarised in Table 2.3.
Normal testing of the filler wire according to ISO 3690 gave fused metal hydrogen contents
of 3.3 ppm (diffusible) and 1.7 ppm (residual). A comparison with Table 2.3 shows that the
multirun content of diffusible hydrogen is much lower than the corresponding ISO value,
probably as a result of a high interpass temperature which facilitates loss of hydrogen to the
surroundings through diffusion.
Table 2.3 Measured hydrogen contents in multirun FCA steel weldment. Data from Skjolberg.14
HAZ
Distance from fusion line (mm)
Condition
Weld Metal
Aswelded
PWHT*
(4h/150C)
Oto 5
5 to 10
0.25 ppm
diffusible
0.15 ppm
diffusible
0.15 ppm
diffusible
0.15 ppm
diffusible
10 to 15
0.10 ppm
diffusible
(233)
(234)
where SN is the maximum solubility of nitrogen at 1 atm total pressure,/^ is the activity coefficient, and pNl is the resulting partial pressure of diatomic nitrogen in the gas phase.
The solubility of nitrogen in liquid iron is approximately given by:
(235)
where T is the temperature in K.
At 1600 and 20000C, this equation gives equilibrium concentrations of 446 and 465 ppm,
respectively. In alloyed steel containing large amounts of nitrideforming elements (e.g.
austenitic stainless steel), the activity coefficient of nitrogen fN is about 1/4 and hence, the
solubility will be about 4 times higher than that calculated from equation (235).
From a primitive model of pseudoequilibrium between gaseous N2 and dissolved N a maximum solubility of about 465 ppm would be expected in welding under 1 atm total pressure.
Thus, the maximum pickup of nitrogen in deposition of bare wire in air would be of the order
of 465A/OT8 ppm or 416 ppm. If a tentative estimate of air infiltration in the arc column is
made at 1 vol% N2, the expected pickup of nitrogen would be 465 VoToT or about 47 ppm.
A comparison with the data in Table 2.4 shows that the measured weld metal nitrogen
contents are much higher than predicted from Sievert's law. This implies that the mechanism
of nitrogen desorption is different from that of hydrogen.
2.9.2 Gasshielded welding
Information on the factors controlling nitrogen pickup may be obtained from the work of
Kobayashi et a/.,16 who examined the GMA welding process in a systematic manner. Some of
their results are shown in Fig. 2.20.
Figure 2.20(a), for lowalloy steel, reveals that the square root relationship is a fair approximation only for welding in mixtures of N 2 and H 2 (curve No. 5). Mixtures of N2 + Ar (curve
No. 3), N 2 + CO2 (curve No. 2) and N2 + O 2 (curve No. 1) show increasing deviation from the
predicted behaviour. Pure N2 under reduced pressure gives a curve (No. 4) of an entirely different shape including a maximum at pN ~ 0.05.
Table 2.4 Summary of measured weld metal nitrogen contents. Data compiled by Christensen.4
Welding Method
Material
Lowalloy steel
60180
Stainless steel
550650
Lowalloy steel
200350
Stainless steel
600750
SAW
Lowalloy steel
40140
FCAW
Lowalloy steel
125275
GMAW
Lowalloy steel
50200
Similar features are seen from Fig. 2.20(b) for welding of stainless steel. Again, the deviation becomes more pronounced as the oxidation potential of the gas mixture is increased in the
sequence H 2 ArCO 2 O 2 . Moreover, a comparison with Fig. 2.20(a) reveals that the displacement of the nitrogen concentrations in the presence of chromium is larger than expected from
the calculated reduction of the nitrogen activity coefficient.
The trends observed in Fig. 2.20 have been confirmed by O'Brien and Jordan17 who studied
nitrogen pickup during CO2shielded welding of lowalloy steel. As can be seen from Fig.
2.21 (a) their curves are similar to those of Kobayashi et al.16 for short circuiting metal transfer,
while a mixed spray/globular transfer gives a sharp rise of nitrogen absorption up to pNi = 0.3
followed by a constant or slightly decreasing concentration (Fig. 2.21(b)). Both patterns are
clearly not in accordance with predictions based on Sievert's law (equation (234)).
An interpretation of the observed trends should be made with a view to absorption of hydrogen, where the concept of pseudoequilibrium has proved useful for a semiquantitative
prediction. In both cases the molecular species H2 and N 2 are known to dissociate in the arc
column (see Table 2.1), and would therefore dissolve in the metal to an extent far beyond the
solubility controlled by pH or PN . The excess of dissolved hydrogen is probably released as
gas at weld pool temperatures. This will also be the case with nitrogen in the absence of oxygen, as shown previously in Fig. 2.20(a) and (b). However, under oxidising conditions the
desorption of gaseous nitrogen becomes suppressed by the presence of oxygen at the gas/metal
interface and hence, nitrogen is retained at a level which by far exceeds the solubility limit at
1 atm total pressure of N2. This has been confirmed experimentally by Uda and Ohno18 in their
classic work on surface active elements (i.e. oxygen, sulphur and selenium) in liquid steel. A
similar phenomenon was quoted in Section 2.8.4.1 from the work of Matsuda et al9 even in
the case of hydrogen, where increased entrapment of hydrogen was observed in the presence
of oxygen (see Fig. 2.15).
It appears thus that excessive absorption of nitrogen (and in some cases also hydrogen)
should be interpreted as a state of incomplete release of solute, as described previously in
Sections 2.6 and 2.7. As a consequence, Sievert's law cannot be used for an estimate of nitrogen pickup in steel welding, unless the weld metal oxygen content is extremely low.
2.9.3 Covered electrodes
The nitrogen content of SMA weld deposits is known to be sensitive to variations in the arc
(a)
(b)
(a)
Lowalloy steel
Experiment
Lowalloy steel
Experiment
length (voltage) because of the risk of air infiltration in the arc column. This point is more
clearly illustrated in Fig. 2.22, which shows that the resulting weld metal nitrogen level may
vary significantly from one weld to another, depending on the operational conditions applied.
Consequently, the use of long arcs in SMAW should be avoided in order to prevent excessive
pickup of nitrogen from the surrounding atmosphere.
2.9.4 Submerged arc welding
In submerged arc welding the risk of air infiltration in the arc column is less imminent, since
welding is performed under the shield of a flux. Hence, in multipass welds the filler wire itself
will be the main source of nitrogen (see Fig. 2.23), while in single pass weldments the base
plate nitrogen content is more important because of the high dilution involved. The latter point
is illustrated by the following numerical example.
Example (2.4)
Consider SA (single pass) welding on a thick plate of lowalloy steel under the following
conditions:
Based on the 'rule of mixtures', calculate the weld metal nitrogen content. Assume in these
calculations that the nitrogen content of the base plate and the filler wire is 0.005 and 0.012
wt%, respectively.
Welder No.
Fig. 2.22. Natural fluctuations in nitrogen pickup during SMA welding due to variations in the arc
length. Data from Morigaki et al.19
Next Page
gain
loss
First we calculate the total amount of fused parent metal and weld deposit formed on welding.
From equations (175) and (1120), we have:
and
The 'rule of mixtures' gives us the nominal weld metal nitrogen content, which is defined
as:
The above calculations show that the nitrogen content of single pass SA steel welds is close
to that of the base plate because of the high dilution involved. This is in agreement with general experience.
Previous Page
Material
Lowalloy steel
250^40
Lowalloy steel
5001300
Lowalloy steel
7001400
Lowalloy steel
50100
Lowalloy steel
2001300*
GMAW (CO2)
Lowalloy steel
500800
Lowalloy steel
130200
Lowalloy steel
4502600
Lowalloy steel
400750
Lowalloy steel
180350
(a)
Wt% C
Electrode tip
Chilled metal (falling droplet)
Multilayer weld metal
Electrode wire
Vol% O 2 in Ar
(b)
wt%c
Electrode tip
Chilled metal (failing droplet)
Multilayer weld metal
Electrode wire
V o l % C O 2 in Ar
Fig. 2.24. Measured carbon contents in electrode tip droplets, chilled metal and multilayer weld deposit
vs the oxygen potential of the shielding gas; (a) ArC>2 gas mixtures, (b) ArCC>2 gas mixtures. Data from
Grong and Christensen.1
Carbon control
Silicon control
Pco>
atm
the former case. Normally, homogeneous nucleation of CO gas within the liquid metal is considered impossible, which means that the CO nucleation in practice must take place heterogeneously. However, the most probable site for CO evolution during droplet formation at the
electrode tip will be the gas/metal interface itself, which allows carbon to be oxidised simultaneously with silicon and manganese. It is reasonable to assume that most of the observed
carbon oxidation is located to the hot layers facing the arc, where the reaction is thermodynamically favoured. At other surface positions Si and Mn are expected to prevent carbon from
reacting due to a rather low metal temperature, stated to be only slightly above the melting
point at the time of detachment.21
It is evident from the data in Fig. 2.24(a) and (b) that the carbon losses increase with increasing O2 or CO2 contents in the shielding gas up to a certain critical level. Hence, supply of
oxygen to the tip droplet surface is the rate controlling step for oxidation of carbon at low
oxygen potentials. This conclusion is also consistent with calculations made by Corderoy et
al.,22 who found that transport of atomic oxygen through a stagnant gaseous boundary layer
close to the metal surface controls the oxidation rate of alloying elements at this stage of the
process.
The carbon oxidation will gradually decline with increasing oxygen concentrations in the
shielding gas, probably as a result of buildup of carbon monoxide in the surrounding atmosphere. When the critical CO gas pressure is reached, the carbon reaction is blocked, silicon
(and manganese) now exercising control of the oxygen level, as indicated in Fig. 2.25. For ArO 2 gas mixtures this critical pco pressure is attained at about 10 to 15 vol% O 2 in the shielding
gas, corresponding roughly to 0.05 wt% C oxidised in chilled metal. When welding is performed in ArCO 2 mixtures the reaction is blocked at a much earlier stage of carbon oxidation
(equal to about 0.02 wt% C lost in chilled metal), since dissociation of CO2 in this case will
produce an additional amount of CO to concentrate in the surrounding gas phase.
Temperature, 0C
Fig. 2.25. Break even equilibrium partial pressure of CO vs temperature for silicon control of oxygen
level at 0.8 wt% Si and silica saturation. Data from Elliott et al. 23
The mechanism indicated above is supported by the data presented in Fig. 2.26, which
show that a CO2rich atmosphere even may act carburising if the initial carbon content of the
electrode wire is sufficiently low. Moreover, it is interesting to note that the carbonoxygen
reaction is also influenced by the rate of droplet detachment. Since the highest carbon oxidation losses are normally associated with a coarse globular droplet transfer mode, this suggests
that the reaction time is more important than the effective contact area available for interaction
which depends on the droplet size.
2.10.1.3 Oxidation of silicon
It is evident from the data in Fig. 2.27 that loss of silicon mainly take place in the weld pool, as
indicated by the difference between the measured silicon content in chilled and multilayer
weld metal. In ArO 2 mixtures the Si loss increases steadily with increasing oxygen potential
of the shielding gas. Thus, at 30 vol% O2 in Ar it amounts to 0.59 wt% Si (or 0.67 wt% O)
removed from the weld pool as a result of deoxidation reactions. A similar situation exists in
the case of CO2shielded welding up to about 20 vol% CO2 in Ar. At higher CO2 contents, the
Si loss tends to drop off, finally attaining an upper limit of approximately 0.30 wt% Si corresponding to 0.34 wt% O removed from the weld pool.
In comparison, the amount of Si lost in the two preceding stages (i.e. electrode tip and arc
column) is much smaller, as shown by the data for the chilled metal Si content. Since no slag is
formed under the conditions of rapid cooling, silicon must escape in the form of a gaseous
(Gain)
(Loss)
(a)
Chilled metal (falling droplet)
Multilayer weld metal
Wt% Si
Electrode wire
Vol% O 2 in Ar
(b)
Chilled metal (falling droplet)
Multilayer weld metal
Wt% Si
Electrode wire
Vol% CO 2 in Ar
Fig. 2.27. Measured silicon contents in chilled and multilayer weld metals vs the oxygen potential of the
shielding gas; (a) Ar02 gas mixtures, (b) ArCC>2 gas mixtures. Data from Grong and Christensen.1
product. Evaporation losses can in this case be excluded due to a very low vapour pressure of
silicon at the prevailing temperatures. It is therefore reasonable to assume that SiO(g) forms as
a result of chemical reactions occurring at the electrode tip. The observed decrease in chilled
metal Si content with increasing O2 and CO2 contents in the shielding gas is probably caused
by the presence of CO at the gas/metal interface, which facilitates SiO formation according to
the reaction:
(slag)
(gas)
(gas)
(gas)
(236)
PSiO,103atm
Figure 2.28 shows a plot of the equilibrium partial pressure of SiO vs temperature at silica
saturation for three different CO levels. Note that stoichiometric amounts of CO2 have been
assumed to form in order to calculate psi0. It is evident from the thermodynamical data presented in Fig. 2.28 that the formation of SiO is strongly dependent on the CO partial pressure
at the slag/metal interface. Thus, from a thermodynamic standpoint the silicon loss at the electrode tip should be most pronounced during CO2shielded welding due to the resulting higher
CO pressure. The data shown in Fig. 2.27(a) and (b) support this assumption. It is seen that the
silicon loss is increased by a factor of 3 to 5 in presence of CO2 when comparison is made on
the basis of equal oxygen potential of the shielding gas (i.e. equal loss of deoxidants in the
weld pool). Similar observations have also been made by Heile and Hill27 from determination
of silicon in collected GMA welding fumes. The recorded chilled metal Si loss in Fig. 2.27(a)
and (b) is in good agreement with the reported fume formation rates of silicon. Since all CO
consumed in reaction (2.36) is expected to be regenerated immediately by decomposition of
CO2 at the metal surface, the high CO partial pressure required for SiO formation at the electrode tip is maintained even in the case of extensive silicon losses. Consequently, the net reaction can be written as:
Temperature, 0C
Fig. 2.28. Equilibrium partial pressure of SiO vs temperature at different CO levels. Data from Refs. 25
and 26.
(slag)
(dissolved)
(gas)
(237)
When welding is performed in ArO 2 gas mixtures, pSiO andpco may be taken proportional
to the recorded loss of silicon and carbon in chilled metal (see data in Fig. 2.29). It is evident
from this plot that the silicon loss is directly proportional to the corresponding loss of carbon
up to a certain critical level. Thus, during the initial period of carbon oxidation at the electrode
tip the SiO formation is probably controlled by the resulting partial pressure of CO at the slag/
metal interface, according to reaction (236). When the carbon reaction is blocked, the chilled
metal silicon loss becomes independent of the CO partial pressure (see reaction (237)), since
all CO consumed in the SiO formation will immediately be recirculated within the system.
However, at the break even point for silicon control of the oxygen level at the electrode tip, the
CO pressure in the surrounding gas phase will be the same for both ArO 2 and ArCO 2 gas
mixtures. Hence, the recorded loss of silicon in chilled metal at 20 vol% CO2 in Ar is seen to be
similar to that in Ar + 10 vol% O2, as indicated by the heavy broken line in Fig. 2.29.
At the temperatures where liquid steel is normally deoxidised, silicon and manganese have
a strong affinity to oxygen. Their ability to form stable oxides decreases rapidly with increasing temperature, and above approximately 180O0C silicon and manganese do no longer act as
efficient deoxidation agents. Precipitation of manganese silicate slags is therefore favoured by
the lower metal temperatures prevailing at the electrode tip and in the cold part of the weld
pool. At higher temperatures, these oxides become unstable. Consequently, as a result of the
metal superheating occurring during droplet transfer through the arc column, the macroscopic
slag phase formed earlier at the electrode tip surface (as reported by Corderoy et ah12) will
redissolve in the metal. This gives rise to a relatively high chilled metal oxygen content (to be
discussed below).
Ar+10vol%O2
Example (2.5)
Consider CO2shielded welding on a thin sheet of lowalloy steel with a 0.8mm dia. electrode
wire under the following conditions:
Based on the data presented in Fig. 2.27(b), calculate the fume formation rate (FFR) of
silicon (in mg per min) due to SiO formation at the electrode tip. The wire feed rate is 125mm s"1.
Solution
The total loss of silicon due to SiO formation may be taken equal to the observed difference
between the filler wire and the chilled metal silicon contents. For welding in pure CO2, we get:
The corresponding fume formation rate of silicon (in mg min"1) can readily be calculated
when the wire feed rate (WFR) is known. Taking the density of the steel equal to 7.85mg mm"3,
we obtain:
A comparison with the measured FFR of silicon in Table 2.6 (at / = 13OA) shows that the
calculated value is reasonable correct. Moreover, these data support our previous conclusion
that the SiO formation is favoured by a high CO2 content in the shielding gas due to the
dissociation reaction. In fact, more detailed studies of the reaction kinetics have confirmed that
the rate of SiO formation is proportional to the resulting partial pressure of CO at the gas/metal
interface,28 in agreement with equation (236).
2.10.1.4 Evaporation of manganese
It is seen from the data in Fig. 2.30(a) and (b) that the amount of manganese lost in chilled
metal is virtually independent of the oxygen potential of the shielding gas, as indicated by the
constant difference of about 0.35 wt% between the filler wire and the chilled metal Mn contents in both graphs. This implies that significant amounts of manganese are lost during
droplet transfer through the arc column as a result of evaporation.
At the prevailing temperatures, the vapour pressure of iron is also high due to the almost
unity activity of Fe. If the average arc metal temperature is taken equal to about 24000C,29 the
data in Fig. 2.31 indicate that the vapour pressures of iron and manganese (at 1.27 wt% Mn)
are nearly identical and close to 0.05 atm. In the hot surface layers of liquid metal facing the
arc the temperature will be even higher, which means that iron vapour will dominate.
Measurements of collected GMA welding fume reported by Heile and Hill27 (see data in
Table 2.6) show a substantial higher loss of iron than that derived from simple thermodynamical calculations taking the rate of element loss proportional to the vapour pressure. From their
results a reasonable value of the average mass ratio Fe to Mn in dust is about 5. Consequently,
as a preliminary estimate the loss of iron may be taken 5 times the amount of manganese lost
Table 2.6 Measured fume formation rates in GMA welding of ferrous materials. Data from Heile and
Hill.27
Shielding Gas
Argon
Ar+ 2 % O 2
Ar+ 5% O 2
Ar + 25% CO2
Pure CO2
Current
Voltage
(A)
(V)
Si
Fe
250
29
22
300
350
31
35
1
5
0
4
12
51
150
200
300
400
28
28
29
34
15
12
5
18
8
10
4
16
134
75
35
86
100
200
300
28
28
28
29
16
10
33
22
23
273
129
76
100
23
11
20
75
150
300
27
35
13
29
35
62
105
191
130
150
200
250
300
27
30
30
30
30
13
14
19
31
36
59
63
73
112
125
86
120
126
216
214
in chilled metal, corresponding to a total loss of 1.7 wt% (Fe + Mn) or 500mg (Fe + Mn) per
min. The reported fume formation rate of Fe and Mn under nearly similar conditions is only
one half of that stated above, which indicates that the calculated flux of metal vapour may be
somewhat overestimated. However, if these two values are considered to represent borderline
cases, the volume of metal vapour is in the range from 300 to 500 times that of the droplets
and will therefore be more than sufficient to protect the metal from arc atmosphere oxidation
at this stage of the process. This conclusion is consistent with statements made by Distin et al?1
who claim that iron vapour acts as an effective oxygen getter already at about 19000C.
Very large amounts of manganese are also lost in the weld pool stage, as shown by the
difference between the measured Mn contents in chilled and multilayer weld metals. This
situation appears to be quite similar to that of silicon. For ArO 2 mixtures the manganese loss
increases steadily with increasing oxygen contents in the shielding gas, and reaches a value of
about 1.08 wt% Mn (or 0.31 wt% O) removed from the weld pool at 30 vol% O2 in Ar. In the
case OfArCO2 shielded welding the Mn oxidation loss starts to drop off at a CO2 content of
about 20 vol%, finally attaining an upper limit of about 0.50 wt% Mn corresponding to 0.15
wt% O removed from the weld pool.
A more detailed discussion of oxidation reactions in GMA welding is given in Section
2.10.1.5.
Example (2.6)
Consider GMA welding lowalloy of steel under conditions similar to those in Example 2.5.
(a)
Wt% Mn
Chilled metal
(falling droplet)
Vol% O 2 in Ar
Wt% Mn
(b)
Vol%CO 2 inAr
Fig. 2.30. Measured manganese contents in chilled and multilayer weld metals vs the oxygen potential
of the shielding gas; (a) Ar02 gas mixtures, (b) ArCO2 gas mixtures. Data from Grong and Christensen.1
Based on the data presented in Fig. 2.30(b), calculate the fume formation rate (FFR) of manganese due to evaporation losses occurring during droplet transfer through the arc column. Estimate also the effective mass transfer coefficient for manganese evaporation under the prevailing circumstances by utilising the vapour pressure data in Fig. 2.31. The surface temperature
of the falling droplets is assumed constant and equal to 26000C.
PMr/PFe
LogpMn,atm
Temperature, 0C
Fig. 2.31. Equilibrium manganese vapour pressure and corresponding vapour pressure ratio pMn to PFe vs
temperature at 1.27 wt% Mn in iron. Data from Kubaschewski and Alcock. 30
Solution
The total loss of manganese due to evaporation may be taken equal to the observed difference
between the filler wire and the chilled metal manganese contents. For welding in pure CO2, we
have:
The corresponding fume formation rate of manganese (in mg min"1) can readily be calculated when the wire feed rate (WFR) is known. Taking the density of the steel equal to 7.85mg
mm~3, we obtain:
A comparison with the data in Table 2.6 reveals that the reported fume formation rate of
manganese (at / = 13OA) is much lower than computed in the present example. This discrepancy can probably be attributed to differences in the filler wire manganese content. Assuming
that the evaporation loss of manganese is controlled by a transport mechanism in the gas phase,
we can estimate the effective mass transfer coefficient from equation (215) by inserting a
reasonable average value for the manganese vapour pressure at the gas/metal interface (pMn in
the bulk phase is taken equal to zero). Reading from Fig. 2.31 (at T= 26000C) gives:
If the diameter of the falling droplets is taken equal to the diameter of the filler wire, it is
possible to calculate the total loss of manganese associated with one droplet (in mol):
Since the average flight time of large, globular droplets through the arc column is of the
order of one second, the corresponding flux of manganese vapour per unit time is close to 9.97
X 10~8 mol s"1. Thus, by rearranging equation (215), we obtain the following value for the
effective mass transfer coefficient:
Although the above value is rather uncertain, the calculated mass transfer coefficient is of
the expected order of magnitude. This supports our previous conclusion that the evaporation
kinetics are controlled by a transport mechanism in the gas phase.
2.10.1.5 Transient concentrations of oxygen
It is evident from the data summarised in Fig. 2.32 that the oxygen content in both chilled and
multilayer weld metal increases with increasing oxygen potential of the shielding gas. The
chilled metal analysis is representative of the oxygen absorption occurring at the electrode tip
due to the lack of gas/metal interaction in the arc column. Measurements of manganese oxidation losses during droplet formation have been performed by Corderoy et al.22 From the curve
presented for 20ms tip melting cycles, a reasonable estimate for the manganese oxidation loss
in ArO 2 shielding gas mixtures may be:
(238)
Taking the Si to Mn mass ratio in precipitated slag equal to 0.66,22 the corresponding loss of
silicon is equal to:
(239)
Based on equations (238) and (239) it is possible to calculate the oxygen absorption at the
electrode tip which is associated with the MnO and the SiO2 slag formation. If corrections also
are made for the amount of oxygen simultaneously removed as iron oxide*, we obtain:
(240)
*An approximate correction can be made from an analysis of the Fe to Mn mass ratio in precipitated slag, which under
the prevailing circumstances is close to 0.16.22
(a)
Wt% O
Vol%O2inAr
(b)
Wt% O
Vol% CO2 in Ar
Fig. 2.32. Measured oxygen contents in chilled and multilayer weld metals vs the oxygen potential of
the shielding gas; (a) ArC>2 gas mixtures, (b) ArCO2 gas mixtures. Data from Grong and Christensen.1
At 30 vol% O2 in Ar the absorption of oxygen is roughly 0.27 wt% O, which is reasonably
close to that recorded in the chilled metal analysis (about 0.23 wt% O). The rate controlling
step for metal oxidation at this stage of the process is believed to be transport of atomic oxygen
through a stagnant gaseous boundary layer of Ar and/or CO adjacent to the metal surface.22
On the other hand, the measured chilled metal oxygen contents are much too low to account
for the heavy oxidation losses of deoxidants observed in multilayer welds. This indicates that
considerable amounts of oxygen are introduced in the hot part of the weld pool immediately
beneath the root of the arc. Although the gas/metal interfacial contact area available for reaction is much smaller than that of the electrode tip (or falling) droplets, the strong turbulence
existing in the hot part of the pool will provide an effective circulation of the liquid metal
through the reaction zone.32 Moreover, absorption of oxygen at this stage of the process will be
favoured by the increased time available for gas/metal interaction. If the diameter of the reaction zone is taken equal to about 3mm, the time available for oxygen absorption in the weld
pool is of the order of one second for a typical welding speed of 3mm s"1. In comparison, the
corresponding reaction time during droplet formation at the electrode tip is only 20 to 50 ms.
Calculations of the total oxygen absorption can be done on the basis of the measured difference between silicon and manganese contents in chilled and multilayer weld metals, designated A[%S7] and A[%Mn], respectively*. Moreover, corrections should be made for the amount
of oxygen simultaneously removed as iron oxide from the weld pool. For deoxidation with
silicon and manganese the mass ratio wt% Fe to wt% Mn in precipitated slag is equal to about
O.I.1 This leads to the following balance:
(241)
At 30 vol% O2 in Ar the total oxygen absorption amounts to:
It is seen from the graphical representation of equation (241) in Fig. 2.33(a) and (b) that
most of the oxygen pickup takes place in the weld pool. However, at present it is not clear
whether the calculated values represent a real transient concentration in the hot part of the
weld pool or a number in concentration units representing precipitation of manganese silicate
slags. According to Fischer and Schumacher33 the solubility of oxygen in liquid iron is 0.94
wt% at 19000C and 1.48 wt% at 20000C This temperature range is probably relevant with respect to the hot part of the weld pool, although the surface metal temperature will be even
higher.34 However, in steel weld metals the oxygen concentrations should be well below this
solubility limit in the presence of silicon and manganese. Taking the average weld pool silicon
content equal to 0.50 wt%, the corresponding oxygen content in equilibrium with a silicasaturated slag is roughly 0.10 and 0.20 wt% at 1900 and 20000C, respectively.23 On the other
hand, under the prevailing circumstances there will be no solid/liquid interface available for
nucleation of oxide particles, and hence homogeneous nucleation is the only possibility.
According to Sigworth and Elliott35 this requires a relatively high degree of supersaturation,
which means that the oxygen concentration most likely will exceed the SiMnO equilibrium
value before slag precipitation occurs.
The data in Fig. 2.33(a) and (b) indicate that the weld pool oxygen absorption is controlled
by a complex transport mechanism in the gas phase. However, since mass transfer in gasjet/
liquid systems is not fully understood,3 we shall only consider the limiting case where the
resistance to mass transfer is confined to a stagnant gaseous boundary layer adjacent to the
metal surface. Under such conditions equation (210) predicts that the transient oxygen concentration is determined by the partial pressure of atomic oxygen in the plasma gas, as shown
*Oxygen consumed in CO and SiO formation is not included.
(a)
Wt% O
Vol% O2 in Ar
(b)
Wt% O
Vol% CO2 in Ar
Fig. 2.33. Calculated total oxygen absorption in GMA welding at different oxygen potentials of the
shielding gas; (a) ArO 2 gas mixtures, (b) ArCO 2 gas mixtures. The analytical weld metal oxygen
content is indicated by the broken lines in the graphs. Data from Grong and Christensen.1
by the plots in Fig. 2.34. If we as a borderline case assume that all oxygen present in the plasma
gas is immediately absorbed in the liquid metal, the slope of the curve in Fig. 2.34 for ArO 2
gas mixtures (equal to about 2.67 wt% dissolved oxygen at one atmosphere total pressure of
atomic oxygen) is representative of the total amount of gas passing through the arc column.
Taking the mass of liquid metal leaving/entering the reaction zone equal to 3Og min"1 under
the prevailing circumstances,1 the following gas flow rate is obtained (in Nl min"1).
This value corresponds to about 6 per cent of the total shielding gas flow rate.
In contrast to the situation described in Fig. 2.34 for ArO 2 gas mixtures, a large deviation
from the expected relationship is observed for welding in CO2rich atmospheres. At present,
the reason for this shift in the reaction kinetics is not known. However, it is evident from the
data presented in Fig. 2.34 that the effective mass transfer coefficient decreases by a factor of
about 2.5 when the CO2 content in the shielding gas increases from 10 to 100 vol%. This
implies that the oxidation capacity of pure CO2 is comparable with that of Ar + 13 vol% O2,
although the partial pressure of atomic oxygen in the plasma gas is equivalent with an oxygen
content of about 33 vol%.
Example (2.7)
[%O]t0t,wt%
Consider GMA welding of lowalloy steel in Ar + 20 vol% CO2 and pure CO2, respectively.
Based on the results in Figs. 2.27(b) and 2.30(b), calculate the total weight of top bead slag (in
gram per 100 gram weld deposit) which forms as a result of deoxidation reactions. Assume in
these calculations that the iron to manganese mass ratio in precipitated slag is equal to 0.1.
Pure CO2
P0, atm
Fig. 2.34. Calculated total oxygen absorption in GMA welding at different partial pressures of atomic
oxygen in the plasma gas. Data from Grong and Christensen.1
Solution
The amount of silicon and manganese lost as a result of deoxidation reactions is equal to the
observed difference in the chilled and multilayer weld metal Si and Mn contents. Assuming
that these elements are removed from the weld pool as SiO2 and MnO, respectively, the following balance is obtained:
Thus, for welding in Ar + 20 vol% CO2, the total weight of slag amounts to:
A comparison with the experimental data in Fig. 2.35 shows that there is a fair agreement
between the amount of slag recorded by weighing and that calculated from a simple mass
balance of silicon, manganese and iron.
g slag/10Og deposit
Vol% CO2 in Ar
Fig. 2.35. Comparison between measured and calculated weight of top bead slag in CO2shielded welding. Data from Grong and Christensen.1
(242)
Similarly, for welding with CO2rich shielding gases (i.e. between 10 and 100 vol% CO2),
we obtain:
(243)
Equal oxidation capacity means that the total weld metal oxygen absorption is the same for
both shielding gas mixtures. Hence, we may write:
(244)
when the CO2 content in the shielding gas is less than 10 vol%, and
(245)
when welding is performed with CO2rich shielding gases (more than 10 vol% CO2 in Ar).
Based on equations (244) and (245) it is possible to compare the oxidation capacity of
various shielding gases (see Table 2.7). Included in Table 2.7 is also a slightly modified version of the International Institute of Welding (HW) classification system,36 which is based on
an evaluation of retained (analytical) oxygen in the weld deposit. It is evident from these data
that both systems are applicable and mutual consistent, although the former one utilises a more
rational criterion for the shielding gas oxidation capacity.
2.10.1.7 Overall oxygen balance
In GMA welding with solid wires, the CO content in the exhaust gas provides a direct measure
of the extent of gas/metal interaction. This CO content should be compatible with that calculated from an overall oxygen balance for the reaction system.37
Example (2.8)
Table 2.7 Proposed shielding gas classification scheme for GMA welding of lowalloy steel according to
equations (244) and (245). Included is also a modified version of the corresponding IIW's classification
system.36
Shielding Gas Composition
[%O] tot
Vol%CO2
Vol%O 2
Vol%Ar
(wt%)
IIW's Terminology*
04
02
02
01
balance
balance
balance
<0.10
Lightly oxidising
(<0.02)
410
25
25
12V2
balance
balance
balance
0.10.25
Oxidising
(0.0250.035)
1025
25100
58

balance
balance
0.250.40
Oxidising
(0.0350.045)
813

balance
balance
0.400.60
Strongly oxidising
(0.0450.07)
> 13
balance
>0.60
Extremely oxidising
(> 0.07)
The analytical weld metal oxygen content (in wt%) is given by the values in brackets.
The shielding gas is pure CO2 and is supplied at a constant rate of 15Nl min *. Based on the
composition data in Table 2.8 calculate the resulting CO content in the welding exhaust gas.
Solution
First we calculate the nominal weld metal chemical composition by neglecting oxidation loss
of alloying elements due to chemical reactions:
The dilution ratios BI(B + D) and DI(B + D) can be estimated from the classic heat flow
theory presented in Chapter 1. From equations (175) and (1120), we have:
Table 2.8 Chemical composition of filler wire, base plate and weld metal used in Example 2.8.
C
(wt%)
O
(wt%)
Si
(wt%)
Mn
(wt%)
Filler wire
0.10
0.01
0.93
1.52
Baseplate
0.14
0.007
0.40
1.30
Weld metal
0.09
0.065
0.35
0.81
Element
and
This gives:
The extent of gas/metal interaction can then be evaluated from the observed concentration
displacements:
Calculated values for the concentration displacements of carbon, oxygen, silicon and manganese utilising the composition data in Table 2.8 are given below.
Element
\%X\nom.
[A%X]
C
0
1 2
0.03
O
0
0 0 8
0.057
'
Si
Mn
6 3
L 3 9
0.28
0.58
The total CO evolution (in mol min l) can now be computed from an overall oxygen balance for the reaction system. In these calculations we shall assume that Si and Mn lost as
SiO(g) and Mn(vap.) immediately react with CO2 to form SiO2 and MnO, respectively*. Taking the density of steel equal to 7.85 X 10~3 g mm"3, the total mass of weld metal produced per
unit time amounts to:
Oxidation of silicon:
Oxidation of manganese:
Total CO evolution (sum):(13.4 + 53.5 + 28.3 + 9.6) X 10~3 mol CO min  1 = 104.8 X 10~3 mol
CO min' 1
Based on this information it is possible to calculate the resulting CO content in the welding
exhaust gas:
A comparison with the data in Table 2.2 shows that a CO content of about 15 vol% is
reasonably close to that determined by analysis.
2.10.1.8 Effects of welding parameters
So far, gas/metal interactions in GMA welding has mainly been discussed in terms of the
oxygen potential of the shielding gas. In the following, some consideration will be given to the
effects of welding parameters on the weld metal chemistry.
Amperage
When the welding current is raised, the time available for interaction decreases due to the more
rapid detachment of the electrode tip droplets. At the same time the interfacial contact area
increases as the average droplet size becomes smaller. From measurements of fume formation
rates in GMA welding,27 it has been shown that these two counteracting effects will almost
cancel, i.e. the total amount of emitted dust (in mg per g deposit) is found to be constant and
nearly independent of the applied amperage. On the other hand, the total fume formation rate
is probably not a reliable index for the burnoff of Si and Mn, since the evolution of iron
vapour during droplet transfer will tend to conceal the corresponding loss of alloying elements.
The effect of amperage (or more correct the droplet detachment frequency) on the burnoff
of carbon, silicon, and manganese in CO2shielded welding has been investigated by Smith et
al.38 They found that the recovery of alloying elements in the weld deposit increased with
increasing welding current (i.e. droplet detachment frequency). In view of the previous discussion, it is reasonable to assume that the higher weld metal carbon and silicon contents reported
by Smith et al3S are a result of a reduced CO and SiO gas evolution at the electrode tip due to
the shorter time available for chemical interaction. In the case of manganese reduced evaporation losses because of a more rapid transfer of the droplets through the arc column offers a
reasonable explanation to the increased element recovery. This shows that the weld metal
chemistry is sensitive to variations in the welding current.
Arc voltage
Since the arc voltage neither affects the melting rate nor the droplet size to any great extent,39
variations in the arc voltage should only have a minor effect on the weld metal chemistry. This
conclusion is apparently in conflict with observations made by Lindborg,40 who found that the
oxidation reactions in GMA welding were strongly voltage dependent and at the same time
independent of the welding current, the droplet detachment frequency, and the mode of metal
transfer (spray or shortcircuiting). Consequently, further investigations are required to explain these discrepancies.
Welding speed
It can be inferred from the data of Grong and Christensen1 that the analytical weld metal
carbon and oxygen contents are virtually independent of the welding speed v within the normal range of GMAW (i.e. from 0.4 to 6 mm s"1). However, the intensified losses of silicon and
manganese observed at low welding speeds indicate that more oxygen is absorbed in the weld
pool under such conditions. This point is more clearly illustrated in Fig. 2.36 which shows a
plot of [%O]tot vs v for a series of multipass GMA welds deposited under the shield of Ar +
10 vol% O2. It is evident that the total oxygen absorption increases nearly by a factor of two
when the welding speed decreases from 6 to 0.4mm s"1. This shows that the welding speed has
a marked effect on the transient oxygen pickup in the hot part of the weld pool during GMA
welding, since it controls the time available for element absorption.
2.10.2 Submerged arc welding
In fluxshielded processes the reaction pattern is much more difficult to assess because of the
[%O]tot,wt%
complicating presence of the slag. For this reason most investigators have chosen to analyse
empirically slag/metal reactions in SA welding. Nevertheless, some authors have been able to
interpret their results on more theoretical grounds in spite of the complex reaction systems
involved.41"*5 Unfortunately, these thermodynamic approaches give, at best, only a qualitative
description of the compositional changes occurring during the welding operation.
Recently, a kinetic model has been developed by Mitra and Eagar46 to account for variations in the element recovery in both singlepass and multipass SA steel weldments. From
their work it is evident that the transfer of alloying elements between the slag phase and the
weld metal cannot be adequately described by means of a primitive model of pseudoequilibrium without including a more detailed analysis of the reaction kinetics. This shows that the
conditions existing in SA welding are quite similar to those prevailing during GMA welding,
although the experimental and theoretical challenges are much greater in the former case due
to the complicating presence of a macroscopic slag phase.
2.10.2.1 Flux basicity index
During SA welding of steel, oxygen may be transferred from the slag to the weld metal due to
decomposition of easily reduced oxides at elevated temperatures according to the overall reaction:
MxOy = xM (dissolved) + y O (dissolved)
(246)
where MxOy denotes any oxide component in the slag phase (e.g. SiO2, MnO or FeO).
The basicity index (B.I.), originally adopted from steel ladle refining practice, is most frequently employed for assessment of oxygen pickup in SA welding, since it gives an approximate measure of the flux oxidation capacity. A number of different expressions exists in the
literature, but for the purpose of convenience the basicity index defined by Eagar47 has been
adopted here:
basic oxides"
non basic oxides
(247)
where the concentration of each flux component is given in weight percent.
It is evident from Fig. 2.37, which shows a typical correlation between the weld metal
oxygen content and B.L, that the oxygen level of welds produced under acid fluxes (i.e. low
B.I.) is strongly dependent on the basicity index. In contrast, the oxygen concentrations of
welds deposited under basic fluxes are seen to be essentially independent of B.I., as indicated
by the horizontal part of the curve in Fig. 2.37. It should be noted that this analysis gives no
information about the extent of slag/metal interaction, since it is based on data for retained
oxygen in the weld deposit. Consequently, because of the empirical nature and limited applicability of the basicity index, its role in the choice of welding fluxes for SA welding is a keenly
debated question.
Basic fluxes
Acid fluxes
(251)
The value of k is difficult to evaluate in practice, but in view of the reported mass transfer
coefficients for Mn and Si a reasonable estimate would be about 0.5 in the case of manganese
silicate fluxes.46 Under such conditions the total oxygen absorption, [%O]abs is given by:
(252)
where [A%O] is the observed concentration displacement of oxygen in the weld metal, and
[%O]rej. is the amount of oxygen rejected from the weld pool as a result of deoxidation reactions.
Based on equation (252) it is possible to estimate the total oxygen absorption during SA
welding of CMn steels from an analysis of measured concentration displacements of oxygen,
silicon, and manganese in the weld metal. The results of such calculations are shown graphically in Fig. 2.38, using data from Indacochea et a/.44 It is evident from this plot that the total
oxygen absorption during SA welding is much larger than that inferred from an analysis of
retained oxygen in the weld deposit. The situation is thus quite similar to that observed experimentally in GMA welding (see Fig. 2.33).
It should be noted that the calculated values for [%O]abs. in Fig. 2.38 may be encumbered
by systematic errors due to the number of simplifying assumptions inherent in equation (252).
However, this does not affect our main conclusion regarding the significance of the oxygen
absorption, since more refined calculations give a pattern similar to that observed above (see
Fig. 2.39).
2.10.3 Covered electrodes
Chemical reactions during SMA welding have been studied by several investigators in the
Retained oxygen
Absorbed
oxygen
Retained
oxygen
(253)
* Although gases such as CO and CO2 are widely dissociated and ionised in the arc column, from a thermodynamic
standpoint, there is no objection to the choice of molecular species as components for the system, provided that
equilibrium is maintained down to temperatures where such species are stable.
Carbon boil in the weld pool has been detected experimentally during welding with covered electrodes,49 which implies that heterogeneous nucleation of CO is kinetically feasible
under the prevailing circumstances. Possible nucleation sites for CO are gas bubbles present in
the macroscopic slag layer covering the metal, created by the vigorous stirring action of the arc
plasma jet. It should be noted that this behaviour is in sharp contrast to experience with GMA
welding, where little or no oxidation of carbon takes place in the weld pool, as shown previously in Section 2.10.1.2. It is tentatively suggested that the apparent difference between SMA
and GMA welding regarding the possibilities for CO nucleation in the weld pool arises from
the lack of a macroscopic slag layer in the latter case.
Unlike carbon, the deoxidation capacity of silicon (and manganese) increases rapidly with
decreasing metal temperatures (se Fig. 2.40), which means that carbon oxidation becomes
gradually suppressed during cooling in the weld pool. Upon reaching the critical temperature
indicated in Fig. 2.40, the carbon reaction is blocked, silicon and manganese now control the
oxygen level. An unknown but significant fraction of the manganese silicate inclusions precipitated in the hot part of the weld pool beneath the root of the arc are brought by convection
currents to the interface between the macroslag and the metal, where they are readily absorbed.
The remaining fraction formed in the cold and unstirred part of the weld pool is trapped in the
metal solidification front in the form of finely dispersed oxide particles. This results in a high
and rather unpredictable weld metal oxygen content.
The above reaction model has been tested experimentally against data obtained from a
series of hyperbaric welding experiments carried out in a remotely controlled pressure chamber with basic covered electrodes containing various levels of ferrosilicon in the electrode
coating (see Table 2.9). Welding under hyperbaric conditions offer the special advantage of
assessing the reactions through variations in the ambient pressure without changing the composition of the electrode coating or the core wire. Consequently, if the proposed reaction model
0.1 wt%Cat10bar.
Temperature,C
Fig. 2.40. The break even equilibrium temperature for silicon control of oxygen level at 0.1 wt% C and
0.3 wt% Si. Data from Elliott et al.23
Table 2.9 Contents of ferrosilicon and iron powder in the electrode coating of experimental
consumables used in the hyperbaric welding experiments.
Electrode
Iron Powder
R*
4.5 wt%
31wt%
5.5 wt%
30 wt%
6.5 wt%
29 wt%
7.5 wt%
28 wt%
Electrode R
A
B
C
Fig. 2.41. Carbon absorption in hyperbaric SMA welding. Data from Grong et al.51
is at least qualitatively correct would expect a correlation between the weld metal carbon
content and the concentrations of oxygen, silicon and manganese, both under atmospheric and
hyperbaric welding conditions. The main effect of pressure on weld metal chemistry is thus to
suppress the carbonoxygen reaction in the weld pool at the expense of intensified oxidation
losses of silicon and manganese, as indicated by the thermodynamic data in Fig. 2.40.
2.103.2 Absorption of carbon and oxygen
It is evident from the data presented Fig. 2.41 that the weld metal carbon content increases
monotonically with pressure from 1 to 31 bar for all four electrodes involved. This indicates
that the carbon oxidation in the weld pool is systematically suppressed under hyperbaric welding conditions. Moreover, Fig. 2.41 reveals a small but important effect of electrode deoxidation
capacity on the weld metal carbon content. Since ferrosilicon itself is an insignificant source of
carbon, the observed increase in the carbon concentrations with increasing additions of
ferrosilicon to the electrode coating is an indication that carbon oxidation in the weld pool is
blocked at an earlier stage of the process at high silicon levels, according to the reaction:
(254)
This interpretation is further supported by the results from the oxygen determination contained in Fig. 2.42. Although there is considerable scatter in the data in this figure, it is evident
that the recorded enhancement of the weld metal carbon content at high ferrosilicon levels in
the electrode coating is accompanied by a corresponding reduction in the oxygen concentrations.
2.10.3.3 Losses of silicon and manganese
Suppression of carbon oxidation in the weld pool at elevated pressures gives rise to intensified
oxidation losses of silicon and manganese, as shown in Figs. 2.43 and 2.44. Moreover, it is
apparent that increased additions of ferrosilicon to the electrode coating result in a corresponding increase in both the silicon and the manganese concentrations. This finding suggests that
the final weld metal content of the deoxidants is controlled by the reaction:
Si (dissolved) + 2MnO (slag) = 2 Mn (dissolved) + SiO2 (slag)
(255)
Assuming the activity ratio (tf Mn0 ) / (aSio2 )in precipitated slag to be constant and independent of pressure, equation (255) may be rewritten as:
(256)
In Fig. 2.45 the weld metal manganese content has been plotted versus the square root of
the silicon content by inserting data from Figs. 2.43 and 2.44. As it appears from Fig. 2.45, the
Electrode R
A
B
C
Electrode R
A
B
C
Fig. 2.43. Silicon oxidation in hyperbaric SMA welding. Data from Grong et al.51
Electrode R
A
B
C
Electrode R
A
B
C
1/2
[Silicon content, wt%]
Fig. 2.45. Correlation between weld metal manganese and silicon contents. Data from Grong et a/.51
(258)
Next Page
Here [%0]eq. is the oxygen concentration in an assumed equilibrium with carbon at a given
pressure, and [%O]anaL is the analytical weld metal oxygen content. For this correlation, minor
vaporisation losses of manganese as well as possible reactions between oxygen and liquid iron
have been neglected.
In Fig. 2.46 the product m = [%C] [%O] is plotted vs the total ambient pressure. Calculations of m have been done both on the basis of [%O]anaL and [%O]eq. It can be seen from Fig.
2.46 that the former set of data (i.e. open symbols in the graph) cannot be represented by a
straight line passing through the origin, which should apply to a true equilibrium reaction.
However, when proper corrections are made for the amount of oxygen removed from the weld
pool after the completion of the carbon oxidation, such a correlation may be obtained as shown
by the solid line in Fig. 2.46. No clear effect of the electrode deoxidation capacity (i.e. ferrosilicon
content) on the product m = [%C] [%O] can be observed within the precision of measurements. This result is to be expected if the weld metal carbon content is controlled by a local
equilibrium with oxygen established at elevated temperatures in the weld pool. Also, inspection of the slope of the curve (i.e. heavy solid line in Fig. 2.46) indicates that the product
K5Nco is about 1.14 X 10~3 under the prevailing circumstances. If a reasonable average value
for the equilibrium constant K5 of 2.3 X 10~3 is assumed within the specific temperature range
of the reaction, we get:
Nco~0.5
(259)
The above calculations suggest that the controlling partial pressure of CO in the reaction
product is significantly lower than the ambient pressure under hyperbaric welding conditions.
This probably arises from an extensive infiltration of helium in the nucleating bubbles at the
slag/metal interface which, thermodynamically, will enhance the deoxidation capacity of carbon according to Le Chatelier's Principle. The conditions existing in hyperbaric SMA welding
thus appear to be similar to OBM/QBOP steelmaking, where simultaneous injection of oxygen and inert gas from the bottom of the convenor during the decarburisation stage results in a
steel carbon content which is typically below the value calculated for equilibrium between
oxygen and carbon at 1 atm partial pressure of CO.52
Previous Page
Here [%0]eq. is the oxygen concentration in an assumed equilibrium with carbon at a given
pressure, and [%O]anaL is the analytical weld metal oxygen content. For this correlation, minor
vaporisation losses of manganese as well as possible reactions between oxygen and liquid iron
have been neglected.
In Fig. 2.46 the product m = [%C] [%O] is plotted vs the total ambient pressure. Calculations of m have been done both on the basis of [%O]anaL and [%O]eq. It can be seen from Fig.
2.46 that the former set of data (i.e. open symbols in the graph) cannot be represented by a
straight line passing through the origin, which should apply to a true equilibrium reaction.
However, when proper corrections are made for the amount of oxygen removed from the weld
pool after the completion of the carbon oxidation, such a correlation may be obtained as shown
by the solid line in Fig. 2.46. No clear effect of the electrode deoxidation capacity (i.e. ferrosilicon
content) on the product m = [%C] [%O] can be observed within the precision of measurements. This result is to be expected if the weld metal carbon content is controlled by a local
equilibrium with oxygen established at elevated temperatures in the weld pool. Also, inspection of the slope of the curve (i.e. heavy solid line in Fig. 2.46) indicates that the product
K5Nco is about 1.14 X 10~3 under the prevailing circumstances. If a reasonable average value
for the equilibrium constant K5 of 2.3 X 10~3 is assumed within the specific temperature range
of the reaction, we get:
Nco~0.5
(259)
The above calculations suggest that the controlling partial pressure of CO in the reaction
product is significantly lower than the ambient pressure under hyperbaric welding conditions.
This probably arises from an extensive infiltration of helium in the nucleating bubbles at the
slag/metal interface which, thermodynamically, will enhance the deoxidation capacity of carbon according to Le Chatelier's Principle. The conditions existing in hyperbaric SMA welding
thus appear to be similar to OBM/QBOP steelmaking, where simultaneous injection of oxygen and inert gas from the bottom of the convenor during the decarburisation stage results in a
steel carbon content which is typically below the value calculated for equilibrium between
oxygen and carbon at 1 atm partial pressure of CO.52
[%C][%O]x102
Electrode R
A
B
C
Separation Stage
Buoyancy
Surface tension
Stirring
Growth Stage
Diffusion of reactants in the
melt to the oxide nuclei
Particle coalescence
Nucleation Stage
Homogeneous nucleation
Heterogeneous nucleation
Fig. 2.46. The product [%C][%0] in hyperbaric SMA welding. Solid symbols: calculations based on
[%O]eq.. Open symbols: calculations based on [%O]anai. Data from Grong et al. 51
Reaction time
Fig. 2.47. The three major consecutive steps in steel deoxidation (schematic).
Although rate phenomena in ladle refining of liquid steel are extensively investigated and
reported in the literature,53"55 only recently attempts have been made to include such effects in
an analysis of deoxidation reactions in arc welding. 1 ' 56 ' 57
2.11.1 Nucleation of oxide inclusions
During ladlerefining of liquid steel, it is well established that homogeneous nucleation of
oxide inclusions may occur in certain regions of the melt where the supersaturation is sufficiently high.55'58 Over the composition range normally applicable to deoxidation of steel, the
number of nuclei formed at the time of addition of deoxidisers is approximately 105 mirr 3 . 55
However, in steel weld metals, the number of oxide nuclei formed during the initial stage of
deoxidation must be considerably higher to account for the observed inclusion number density
of about 107 mm~3 to 108 mm"3.56'57 This implies that the supersaturation established in the
weld pool on cooling as a result of rapid temperature fluctuations (~103 0C s"1) exceeds by far
that obtained by additions of deoxidisers to a liquid steel melt under approximately isothermal
conditions.
There are several theories available for treating nucleation phenomena, but for the purpose
of convenience, a simplified version of the model of Turpin and Elliott58 has been adopted
here. Consider a steel melt which is brought to a state of supersaturation by first equilibrating
it with pure MxOy at a high temperature T\ and then rapidly cooling it to a lower temperature
T2, as shown schematically in Fig. 2.48. It follows from the classical theory of homogeneous
nucleation that the required temperature difference T1T2 necessary to achieve spontaneous
precipitation of MxOy is approximately given by the following relationship:
(260)
where AH is the standard enthalpy of reaction, G is the oxidesteel interfacial energy (assumed to be constant and independent of temperature), and Vm is the molar volume of the
nucleus.
The derivation of equation (260) is shown in Appendix 2.2.
Example (2.9)
Assume that precipitation of manganese silicates in the weld pool occurs according to the
following reaction:
Si (dissolved) + Mn (dissolved) + 3 O (dissolved) = MnSiO3 (slag)
(261)
where
AG(J) = 858620 + 3457.
Based on equation (260), calculate the critical temperature interval of subcooling for homogeneous nucleation OfMnSiO3. Typical physical data for liquid steel and manganese silicate slags are given in Table 2.10.
Time
Fig. 2.48. Idealised model for homogeneous nucleation of oxide inclusions in steel weld metals (schematic).
Table 2.10 Physical data for liquid steel and manganese silicate slags at 16000C. Data from Refs. 3 and
53.
Property
Density
(kgnr 3 )
Viscosity
(kgm^s 1 )
Steel
6900
4.3 X 10~3
Silicate slag
2300
Interfacial Energyt
(Jm" 2 )
0.8
Solution
By inserting the appropriate values for Vm, AH, and a in equation (260), we obtain:
from which
If we assume that the supersaturation is released at T2 = 16000C (1873K), the initial temperature of the liquid T1 becomes:
Similar calculations can also be carried out for other types of oxide inclusions, e.g. FeO(I),
SiO2(s), and Al2O3(S). The results of such computations are presented graphically in Fig. 2.49,
using data from Refs 55 and 58. It is evident from these plots that the critical temperature
interval of subcooling depends on the interfacial energy, G. Although data for oxidesteel
interfacial energies are scarce, the following average values are frequently used in the literature, 5558 i.e. a(FeOFe) = 0.3 J m"2; G(SiO2Fe) = 0.9 J m~2; and G(Al2O3Fe)  1.5 J mr2. If
these values are accepted, the results in Fig. 2.49 indicate that the critical temperature interval
of subcooling for homogeneous nucleation of FeO (1), SiO2(s), and Al2O3(S) is of the order of
200 to 30O0C. Considering the fact that the liquid weld metal spans a temperature range of
about 2200 to 15000C,34 it is not surprising to find that nucleation of oxide inclusions occurs
readily in the weld pool during cooling. It should be noted that the quoted data for G are
representative of ladlerefined steel deoxidised at 16000C. At higher metal temperatures, in the
presence of large amounts of dissolved oxygen, the oxidesteel interfacial energies would be
expected to be significantly lower.59 Hence, it is reasonable to assume that the actual temperature interval of subcooling required for spontaneous oxide precipitation in a weld pool is well
below 2000C.
Interfacial energy,o(J/m2)
Fig. 2.49. Critical temperature interval of subcooling for homogeneous nucleation of oxide inclusions in
steel weld metals at 16000C as a function of the interfacial energy a.
(i) Collision
(ii) Diffusion
(iii) Ostwald ripening.
From deoxidation and ladle refining of liquid steel it is well established that the flotation
rate of the oxides generally depends on their growth rates, since large inclusions separate
much more rapidly than small ones, in agreement with Stokes law.55 Growth of the oxides can
proceed either through diffusion of reactants in the melt to the oxide nuclei or by collision and
coalescence of ascending inclusions, and is therefore influenced by factors such as the number
density of the nuclei, interfacial tensions, and the extent of melt stirring.53"55 The last factor is
of particular importance in welding, because the stirring action will increase the possibilities
for collision and coalescence of inclusions and, depending on the direction of flow, can give
rise to circulation of inclusionladen metal to the surface. As a result, the separation of small
oxide particles, i.e. microslag, is strongly favoured by the turbulent conditions existing in the
hot part of the weld pool immediately beneath the root of the arc.
2.11.2.1 Buoyancy (Stokes flotation)
Assessment of the role of buoyancy (Stokes flotation) in the separation of deoxidation products from the weld pool can be done on the basis of the experimental data of Grong et ai56
reproduced in Fig. 2.50. It is evident that a change in the welding position (i.e. from flat to
overhead position) has no significant effect on the weld metal oxygen content. This shows that
the buoyancy effect does not play an important role in the separation process of oxide inclusions during welding.
Upward welding
Downward welding
Horizontal
The above conclusion is also in agreement with predictions based on the Stokes law, which
gives the terminal velocity of the ascending particles (u) relative to the liquid:55
(262)
where g is the gravity constant, dv is the particle diameter, Ap is the difference in densities
between the liquid steel and the inclusions, and x is the steel viscosity.
Taking the Stokes parameter, gAp/18ji, equal to 0.6 jjinr1 s"1 for manganese silicate slags in
steel (Table 2.10), equation (262) becomes:
(263)
Generally, the majority of nonmetallic inclusions in steel weld metals are of a diameter
below 2 jxm.56'57 According to equation (263), such particles have a relative velocity less than
2.4 xm s"1. This implies that the buoyancy effect alone is far too insignificant to promote
flotation of the inclusions out of the weld pool before solidification, when it is recognised that
the average fluid flow velocity in the weld pool is four to five orders of magnitude higher (to
be discussed later).
Stokes law is based on the assumption that the inclusions are completely wetted by the
liquid steel, i.e. there is no slip at the oxide/metal interface.55 Normally, interfacial tension
effects promote slip at the particle/metal interface, which, in turn, enhances the flotation rate
of the ascending oxides (often referred to as the Plockinger effect).60 The concept of a wetting
angle has been used in this context. But, the slip phenomenon is probably perceived better in
terms of a secondary flow in the interfacial region between the liquid metal and the solid
precipitate, which is produced by gradients in interfacial tension. In the rare case of no wetting
(0 = 180), it has been shown that the average terminal velocity of the inclusions is approximately 50% higher than that given by equation (262).55 For silica slags, the wetting angle (in
the presence of air) is close to 115,54 indicating that the correction in the particle terminal
velocity due to the slip at the oxide/metal interface in less than about 30%. Consequently,
interfacial tension effects between the slag and the steel do not significantly affect the flotation
rate of the particles, and, therefore, can be ignored.
2.11.2.2 Fluid flow pattern
It is evident from the above discussion that the separation of deoxidation products during arc
welding is controlled by the fluid flow fields set up within the weld pool. The typical flow
pattern in a submerged arc weld pool is shown in Fig. 2.51. It can be seen from the figure that
a depression is formed at the forward edge of the pool, which forces the melted metal to flow
underneath and on either side of the depression, following the arrows in Fig. 2.51. At the rear
of the weld pool the flow direction is reversed, and the metal streams back along the pool
surface. A similar flow pattern has also been observed in GMA weld pools.61 As a result, the
inclusions precipitated in the turbulent part of the weld pool are rapidly brought to the upper
surface behind the arc due to the highvelocity flow field created within the liquid metal, and
disengaged by the surface tension effects in the pool.
The drag force exerted on the particles because of the liquid flowfield velocity can be
estimated from published data for weld pool flow velocities following electromagnetic fluid
mechanics theory. Normally, for weld pools the Peclet number for heat transfer lies within the
range from 10 to 5000, which indicates that the heat flow is predominately convectional.62 The
limits quoted above for the Peclet number correspond to a range in the average weld pool fluid
velocity from about 0.025 to 0.4 m s"1 in the case of SA welding.6364 Unfortunately, specific
information about the fluid flow velocity gradients in the weld pool are lacking, which prevents a more complete analysis of variations in the flow pattern with increasing distance from
the root of the arc.
Generally, the drag force, Fd, acting on a spherical particle in relative motion to a fluid can
be expressed as:3
(264)
where Cd is the drag coefficient, p is the fluid density, u is the bulk velocity of the fluid
relative to the particle, and dv is the particle diameter. The drag coefficient is, in turn, a function of the particle Reynolds number, A^,:3
(265)
where x is the fluid viscosity. Typical physicalproperty data for liquid steel are given in Table
2.10. By inserting the values for p and JJL from Table 2.10, then using the relative velocity
calculated in equation (263) (2.4 xm s"1) as one extreme and the fluid flow velocity (0.4 m s"1) as
the other (i.e. assuming a stationary particle), one can demonstrate that the particle Reynolds
numbers obtained never exceed 1.3 for most weld metal inclusions. The limit calculated above
for the Reynolds number is within the socalled creeping flow region where Stokes law is
indeed valid. Under such conditions, the ratio between the particle drag force Fd and the corresponding gravity force Fg is given by the expression:3
(266)
Note that equation (266) is the basis for obtaining Stokes law (equation (262)).
For steady motion of the particle, Fd and Fg are, of course, equal. However, the significance
of Fd can also be interpreted in a transient sense, by considering the limiting case where the
particle is stationary and its instantaneous motion is governed by the forces acting on it. In this
case, the ratio of Fd (max) acting on the particle relative to the net gravity force (given by
Electrode
Pulsating cavity
Base plate
Fig. 2.51. Typical flow pattern in SA weld pools. For clarity, the arc, slag and flux have been omitted.
The sketch is based on the ideas of Lancaster.32
equation (266)) can determine the dominant force responsible for the particle's trajectory. In
Fig. 2.52 the above ratio has been calculated and plotted against the particle diameter, dv, for
values of u' equal to 0.025 and 0.4 m s"1, which represent the typical velocities reported for
SA steel weld pools. 63 ' 64 It can be seen from the figure that the drag force is always several
orders of magnitude greater than the gravity force for particles within the typical size range of
weld metal inclusions (i.e. less than 2 xm in diameter). This result is to be expected, since the
relative velocity, based on Stokes law and calculated in equation (263), is negligible compared with the liquid velocities in the weld pool (2.4 \xm s"1 vs 0.02 to 0.4 m s"1). This calculation thus supports our previous conclusion that the separation of the oxide inclusions is controlled solely by the fluid flow behaviour in the weld pool. The fact that the phase separation
proceeds under strongly turbulent conditions is also evidenced by the large number of iron
droplets being mechanically dispersed in the top bead slag of GMA steel welds, as shown by
the optical micrographs in Fig. 2.53.
In the case of GMA welding, the nonmetallic inclusions that are brought to the upper
surface behind the arc coalesce rapidly to form large slag clusters that float on the top of the
bead. Generally, reentrapment of the slag does not occur owing to the decrease in the total
surface free energy of the system, which is caused by the emergence of the inclusions from the
weld metal.54 Consequently, the slag will remain floating on the top of the bead even when
welding is performed in the overhead position, as shown previously in Fig. 2.50.
*ci(max/Fg
(a)
(b)
Fig. 2.53. Optical micrographs showing characteristic "comet's tails" of trapped iron droplets (light
areas) in collected top bead slags of GMA steel welds; (a) Ar + 5vol%O2, (b) Ar + 5vol%CO2.
Electrode
(a)
Arc
S e c t i o n AA
Temperature
(b)
Oxygen content
Fig. 2.54. Schematic diagrams showing the sequence of reactions occurring during weld metal deoxidation;
(a) Longitudinal section of weld pool, (b) Cross section of weld pool along AA. The diagrams are based
on the ideas of Grong and Christensen.1
(i) The hot part of the weld pool, characterised by simultaneous oxidation and deoxidation
of the metal, where the separation of microslag takes place continuously as a result of highly
turbulent flow conditions.
(ii) The cold part of the weld pool, where precipitated slag will largely remain in the metal
as finely dispersed particles as a result of inadequate melt stirring.
Under such conditions equation (223) predicts a direct correlation between absorbed and
retained (analytical) oxygen in the weld metal, i.e.:
(267)
in agreement with experimental observations (see plots in Figs. 2.33, 2.38, and 2.39). Typically, the proportionality constant C4 varies between 0.1 to 0.2, which corresponds to a range
in the t/to ratio from 2.3 to 1.6. This shows that the boundary between 'hot' and 'cold' parts of
the weld pool is not well defined, but depends on the welding system under consideration as
well as on the operational conditions applied.
2.11.3 Predictions of retained oxygen in the weld metal
Although the weld metal oxygen content is controlled by a transport mechanism in the weld
pool, the concept of pseudoequilibrium can still be used for an assessment of slag/metal reactions in arc welding.
2.11.3.1 Thermodynamic model
In the case of silicon deoxidation of steel weld metals, we may write:
Si (dissolved) + 2 O (dissolved) = SiO2 (slag)
(268)
unity for a given temperature. A check of this equation against more recent data for silica
activities in MnOSiO2 slags reported by Turkdogan67 supports the findings of Walsh and
Ramachandran65 that the activity of silica is approximately given by equation (270) for a wide
range in the steel manganesetosilicon ratio (i.e. from 0.1 to 50). By combining equations (269) and (270), it is possible to obtain an expression for the equilibrium oxygen content, solely
in terms of the silicon and manganese concentrations:
(271)
where K^ is a temperaturedependent parameter equal to (K1I K6)05.
The temperature dependence of the SiO reaction (equation (268)) is well established and
is approximately given by the relationship:23
(272)
when pure SiO2 is used as the standard state for the silica activity.
For a rough estimate of the temperature dependence of equation (270), the results of
Turkdogan55 can be used. It should be noted that Walsh and Ramachandran65 did calculate the
temperature dependence of K1 within the range from 1550 to 16500C. However, because equation (270) is empirical, the function cannot be extrapolated beyond these temperature limits.
The data quoted in Ref. 55 are derived directly from the SiMn reaction (equation (255)) and
activity data for MnO at silica saturation. On introduction of the equilibrium constant for
equation (255), we obtain:
(273)
By using data from Ref.55, the initial [%Mn]2/[%Si] ratio for precipitation of silica saturated slags (equal to K9(aMnO)2) at 1500,1550,1600 and 16500C has been recorded and replotted
against temperature, as shown in Fig. 2.55. The figure shows that the critical [%Mn]2/[%Si]
ratio for precipitation of silica saturated slags is temperature dependent and decreases from
about 5 at 15000C to below 1.5 at 16500C. On the basis of a crude extrapolation of the data to
higher metal temperatures, it can, however, be argued that the ratio would approach a constant
value of approximately 0.75 at temperatures beyond about 17500C (indicated by the broken
horizontal line in Fig. 2.55). The above observation reflects the fact that the SiMnO reaction
equilibrium (equation (255)) is not very sensitive to a change in temperature (i.e. the enthalpy
for the reaction is small). Over the composition range normally applicable to SiMn deoxidation
of steel weld metals the observed threshold for the critical [%Mn]2/[%Si] ratio for precipitation of silicasaturated slags in Fig. 2.55 corresponds to a manganesetosilicon ratio closely
equal to unity. Consequently, at temperatures higher than about 17500C, K1 can, as a first
approximation, be taken constant and independent of temperature (i.e. K1 Y). The temperature dependence of equation (271) is thus simply (1/^ 6 ) 05 or:
(274)
[%Mn]2/[%Si]
Temperature, 0C
Fig. 2.55. The critical [%Mn]2/ [%Si] ratio for precipitation of silicasaturated slags as a function of
temperature. Data from Turkdogan.55
GMA Welding
SMA Welding
([%SI] [%Mn]f"25
Fig. 2.56. Examples of pseudoequilibrium in GMA and SMA welding of CMn steels. The solid lines in
the graph represent thermodynamical calculations at indicated temperatures. Data from Refs. 56 and 68.
cles), shape (angular or spherical particles), and crystallographic properties as a result of the
complex alloying systems involved69 (see Fig. 2.57). An exception may be CMn steel welds,
where the oxide inclusions will be predominately glassy, spherical, manganese silicates.1
A survey of important weld metal inclusion characteristics is given in Table 2.11.
2.12.1 Volume fraction of inclusions
It is evident from Table 2.11 that the volume fraction of nonmetallic inclusions in steel weld
metals normally falls within the range from 2 X 10~3 to 8 X 10~3, depending on the type of
weld under consideration. Based on simple stoichiometric calculations it is possible to convert
the analytical weld metal oxygen and sulphur contents to an equivalent inclusion volume fraction when the chemical composition of the reaction products is known. This is shown below.
Table 2.11 Summary of weld metal inclusion characteristics. The data are compiled from miscellaneous sources.
Size Distribution*
Chemical Composition
Type of Weld
CMn steel
weld metals
Lowalloy
steel weld
metals
Constituent
elements
Reported
phases
1050
Si, Mn, O, S
(traces of Al,
Ti, and Cu)
SiO2,
MnOSiO2,
MnS, (CuxS)
1040
MnOAl2O3,
7Al2O35TiN,
MnOSiO2, SiO2,
aMnS, (3MnS,
(Cu^S)
Vv X 10"3
dv
NvX 107
Sv
38
0.30.6
110
26
0.30.7
0.55
*Vy: volume fraction; dv: arithmetic mean (3D) particle diameter (jxm); Nv: number of particles per unit volume
(mm"3); Sv\ total particle surface area per unit volume (mm2 per mm3).
Fig. 2.57. Digital STEM brightfield image and Si, Al, S, Mn and Ti Xray images of a multiphase weld
metal inclusion. After Kluken and Grong.57
Solution
First we calculate the total weight of retained MnOSiO2 per 100 g weld metal:
Similarly, we can calculate the weight and volume fraction of MnS in the weld metal:
and
In practice, the stoichiometric conversion factors for oxygen and sulphur are virtually constant for a wide spectrum of inclusions70 and hence, they can be regarded as independent of
composition. Taking the solubility of sulphur in solid steel equal to 0.003 wt%, the following
relationship is obtained for steel weld metals:57'70
(275)
The validity of equation (275) has been confirmed experimentally by comparison with
microscopic assessment methods.5771
In steel weld metals the majority of the inclusions will be in the submicroscopic range
owing to the limited time available for growth of the oxides. From the histogram in Fig. 2.58 it
is seen that particles with diameters between 0.3 to 0.8 xm contribute to nearly 50 percent of
the total inclusion volume fraction. This trend is not significantly changed by additions of
strong deoxidisers, such as aluminium and titanium, or by a moderate increase/decrease in the
heat input.57
2.12.2 Size distribution of inclusions
As shown in Fig. 2.59 the majority of the threedimensional (3D) inclusion diameters fall
within the range of 0.05 to 1.5 xm, with a characteristic peak in the particle frequency at about
0.4 to 0.5 jim. These data obey the lognormal law, since a plot of the frequencies against the
logarithms of the diameters approximately gives a symmetrical curve. Considering specific
inclusion size classes, deoxidation with aluminium generally results in a higher fraction of
coarse particles (> 1 xm) due to incipient clustering of Al2O3.57' 69 However, the observed
particle clustering has no significant influence on the arithmetic mean 3D inclusion diameter,
as shown by the data in Fig. 2.59.
(a)
Frequency (%)
(b)
High Al (0.053 wt%)
Frequency (%)
(276)
Here do is the initial particle diameter, a is the oxidesteel interfacial energy, Dm is the
element diffusivity, Cm is the element bulk concentration, V'm is the molar volume of the oxide
per mole of the diffusate, and t is the retention time.
Frequency (%)
Fig. 2.60. Effect of heat input on the 3D inclusion size distribution in lowalloy steel weld metals. Data
from Kluken and Grong.57
Fig. 2.61. Variation of arithmetic mean 3D inclusion diameter with heat input during SA welding. Data
from Kluken and Grong.57
For welding of thick plates, the time available for growth of particles in the 'cold' part of
the weld pool can be estimated from the Rosenthal equation, i.e. equation (145) in Section
1.10.2 (Chapter 1). If the characteristic length of the cooling zone is taken equal to the weld
ripple lag (defined in Fig. 2.63), the retention time t is approximately given by the following
relationship:
Temperature, C
Time, seconds
Fig. 2.62. Measured temperature level in the trailing edge of the weld pool during GMA welding. Data
from Kluken and Grong.57
Max. width
Retention time
Weld ripple lag
Welding speed
Heat source
Fusion boundary
Fig. 2.63. Definition of weld ripple lag xm, and retention time t.
(277)
where J\ is the arc efficiency (equal to about 0.95 for SAW and 0.80 for GMAW /SMAW), and
E is the gross heat input (kJ mm"1). Note that equation (277) assumes constant values for the
steel thermal diffusivity and volume heat capacity (5 mm2 s"1 and 0.0063 J mm"3 0C"1, respectively), and no preheating (T0* = 200C).
In Fig. 2.64 the Ostwald ripening theory has been tested against relevant literature data,
which may be considered representative of the 3D particle size distribution. Although there is
some scatter in the data, the observed inclusion growth rates fall within the range calculated
for oxygen diffusioncontrolled coarsening of SiO2 and Al2O3 at 15500C, using the Wagner
equation. In these calculations, a reasonable average value for the bulk diffusivity of oxygen
has been assumed (i.e. 10~2 mm2 s"1).55 If the effective inclusion growth rate constant for lowalloy steel weld metals is taken equal to the slope of the curve in Fig. 2.64, the following
relationship is achieved:
GMAW
SAW
dv,jLim
SMAW
1/3
i i sJ/3
Fig. 2.64. Relation between arithmetic mean (3D) inclusion diameter dv and retention time t for different
arc welding processes. Data compiled by Kluken and Grong.57
CoId1 part of
weld pool
Deoxidation
Phase separation
Deoxidation
Incomplete phase;
separation
Homogeneous
nucleation
Growth of
inclusions
Oxidation
Temperature
Gas/metalslag/metal
reactions
'Hot' part of
weld pool
Ostwald ripening
Time
Fig. 2.65. Proposed deoxidation model for steel weld metals (schematic). The diagram is based on the
ideas of Kluken and Grong.57
(278)
By substituting equation (277) into equation (278), we obtain a direct correlation between
the arithmetic mean 3D inclusion diameter dv and the net heat input T\E:
(279)
Equation (279) predicts that dv is a simple cube root function of E, in agreement with the
experimental data in Fig. 2.61.
It should be noted that the measured shape of the particle distributions (see Figs. 2.59 and
2.60) deviates somewhat from that required by the Wagner equation, which assumes a quasistationary distribution curve, and that the maximum stable particle diameter is about 1.5 times
the mean diameter of the system.73 Although the origin of this discrepancy remains to be
resolved, this suggests that particle clustering is also a significant process in steel weld metals
as it is in other metallurgical systems. In fact, such effects would be expected to be most
pronounced at high aluminium levels because of a large interfacial energy between Al2O3 and
liquid steel, in agreement with experimental observations.57'69
2.12.2.3 Proposed deoxidation model
Referring to Fig. 2.65, the sequence of reactions occurring during weld metal deoxidation can
be summarised as follows. In general, nucleation of oxide inclusions occurs homogeneously
as a result of the supersaturation established during cooling in the weld pool. The diffusioncontrolled deoxidation reactions (i.e. diffusion of reactants to the oxide nuclei) will be essentially complete when the liquid metal temperature attains a constant level of about 15500C at
some distance behind the root of the arc. Growth of the particles may then proceed under
approximately isothermal conditions at a rate controlled by the Wagner equation until the
temperature reaches the melting point of the steel. Since retention times in welding generally
depend on the heat input, it follows that choice of operating parameters will finally determine
the degree of particle coarsening to be achieved.
Example (2.11)
Consider SA welding with a basic flux on a thick plate of lowalloy steel under the following
conditions:
Previous experience has shown that this steel/flux combination gives a weld metal oxygen
and sulphur content of 0.035 and 0.008 weight percent, respectively. Based on the stereometric
relationships given below, calculate the total number of particles per unit volume Nv, the total
number of particles per unit area Na, the total particle surface area per unit volume Sv, and the
mean particle centre to centre volume spacing Xv in the weld deposit:74'75
(280)
(281)
(282)
(283)
Solution
First we calculate the total volume fraction of oxide and sulphide inclusions from equation
(275):
Vv = 102 [5.0(0.035) + 5.4(0.008 0.003)] = 2.0 X 10~3
The arithmetic mean (3D) inclusion diameter can then be evaluated from equation (279):
This gives:
particles per mm3
particles per mm 2
mm 2 per mm3
particles per mm 2
mm 2 per mm3
Silicon
Manganese
Aluminium
Frequency (%)
26 wt%
20 wt%
28 wt%
implies that the 7TiO constituent is not an equilibrium reaction product in steel deoxidation.55'76
Hence, it is reasonable to assume that the titaniumrich crystalline patches observed toward
the edges of weld metal inclusions are titanium nitride. This conclusion has later been confirmed experimentally by Kluken and Grong.57
2.12.3.4 Constituent phases
Based on the above discussion, it is possible to rationalise the formation of primary and secondary reaction products (i.e. oxides, sulphides and nitrides) during cooling in the weld pool,
as shown in Fig. 2.67. In general, the inclusions will consist of an oxide core which is formed
during the primary deoxidation stage. The chemical composition of the deoxidation product
can vary within wide limits, depending on the activities of Al, Ti, Si, Mn, and O in the weld
metal. The surface of the oxides will partly be covered by MnS and TiN (see also STEM
micrograph in Fig. 2.57). Precipitation of these phases occurs after the completion of the weld
metal deoxidation, probably during solidification, where the reactions are favoured by solute
enrichment in the interdendritic liquid. Additional precipitation of TiN may occur in the solid
state as a result of diffusion of titanium and nitrogen to the surface of the inclusions.
2.12.4 Prediction of inclusion composition
Since the diffusioncontrolled deoxidation reactions are completed within a fraction of a second when the number of nuclei is of the order of 107 mm"3 or higher,55 the average chemical
composition of the inclusions should be compatible with that calculated for chemical equilibrium at temperatures close to the melting point (e.g. 15500C).
2.12.4.1 CMn steel weld metals
Over the composition range normally applicable to siliconmanganese deoxidation of steel
weld metals (i.e. between 0.4 to 0.7 wt% Si and 0.8 to 1.5 wt% Mn) the equilibrium reaction
product at 15500C should be silicasaturated slags with a mole fraction of SiO2 close to 0.55. 5565
Since the two other slag components are MnO and FeO, we may write:
(284)
The activity coefficients for MnO and FeO in the ternary system SiO 2 MnOFeO can be
computed from the equations presented by Sommerville et al.:S2
(285)
and
(286)
For the specific case of silicasaturated slags, we obtain:
(287)
and
(288)
(289)
The activity ratio aMnO/aFeO can also be estimated from the equilibrium constant for the FeMn reaction at 15500C,23 i.e.:
(290)
The corresponding mole fractions of MnO and FeO in the slag phase are then obtained by
combining equations (289) and (290):
(291)
and
(292)
Equations (291) and (292) provide a basis for calculating the chemical composition of the
inclusions under different deoxidation conditions. A requirement is, however, that the weld
metal Si to Mn ratio is sufficiently high to promote precipitation of silicasaturated slags at
15500C.
and
This gives the following chemical composition of the inclusions (in wt%):
The above calculations should be compared with the composition data presented in Fig.
2.68. It is evident from this plot that the agreement between predictions and experiments is
reasonably good both at high and low weld metal manganese levels. This justifies the
simplifications made in deriving equations (291) and (292).
2.12.4.2 Lowalloy steel weld metals
In principle, a procedure similar to that described above could also be used to establish a
theoretical basis for predicting the chemical composition of the inclusions in the case lowalloy steel weld metals. Unfortunately, adequate activity data for the FeAlTiSiMnO system are not available in the literature.
An alternative approach would be to calculate the average inclusion composition from simple mass balances, assuming that all oxygen combines stoichiometrically with the various
deoxidation elements to form stable oxides in the order Al2O3, Ti2O3, SiO2, and MnO, according to their oxygen affinity in liquid steel (see Fig. 2.69). If reasonable average values for the
inclusion and steel densities are used (i.e. 4.2 and 7.8 g cm"3, respectively), the following set of
equations* can be derived from a balance of O, Al, Ti, S, Si, and Mn and data for acid soluble
aluminium and titanium in the weld metal:57
Aluminium
The average aluminium content of the inclusions, [%Al]inch can be estimated from the measured difference between total and acid soluble aluminium in the weld metal, (k%Al)we[d. This
difference is, in turn, equal to the total mass of aluminium in the inclusions:
(293)
where mind and Vv (cal) are the total mass and volume fraction of nonmetallic inclusions in
the weld deposit, respectively.
*The numerical constants in the constitutive equations given below could alternatively be expressed in terms of
atomic weights etc. to bring out more clearly their physical significance (e.g. see the treatment of Bhadeshia and
Svensson83).
Mole fraction
Ar  O2 gas mixtures
Ar  CO2 gas mixtures
Si
Ti
Al
(294)
However, since data for acid soluble aluminium (and titanium), in practice, may contain
large inherent errors, the following restriction applies:
(295)
where [%O]anaL is the analytical weld metal oxygen content.
Titanium
Similarly, the average titanium content of the inclusions, [%Ti]incl, can be estimated from the
measured difference between total and acid soluble titanium, (A%Ti)weld. However, since TiN
dissolves readily in strong acid, it is necessary to include an empirical correction for the amount
of titanium nitride which simultaneously forms at the surface of the inclusions during solidification. This can be done on the basis of published data for the solubility product of TiN in
liquid steel.84 If we assume that the nitrogen content of the inclusions is proportional to the
calculated difference between total and dissolved nitrogen at the melting point (15200C), the
following relationship is obtained:
(296)
where [%N]anai is the analytical weld metal nitrogen content, and [%Ti]soL is acid soluble
titanium. Note that the correction term for TiN, in practice, neither can be negative nor exceed
[%Ti]soh
In this case, the maximum amount of titanium which can be bound as Ti2O3 is determined
by the overall oxygen balance:
(297)
Sulphur
If the solubility of sulphur in solid steel is taken equal to 0.003 wt%,70 the average sulphur
content of the inclusions, [%S]ind is given by:
(298)
where [%S]anai is the analytical weld metal sulphur content.
Silicon and manganese
From the experimental data of Saggese et al.S5 reproduced in Fig. 2.70, it is evident that the
mass ratio between SiO2 and MnO in the oxide phase may be considered constant and virtually
independent of composition (i.e. equal to about 0.94). This implies that the average silicon
content of the inclusions, [%Si]ind, can be calculated from a balance of oxygen:
SiO2
Al2O3
MnO
wt% AI2O3
Fig. 2.70. Measured inclusion compositions in lowalloy steel weld metals. Data from Saggese et alP
(299)
Considering manganese, proper adjustments should also be made for the amount of MnS
formed at the surface of the inclusions during solidification. Hence, the average manganese
content of the inclusions, [%Mn]inci, is given by the sum of the oxygen and the sulphur contributions:
(2100)
In Fig. 2.71, the accuracy of the model has been tested against the experimental SA/GMA
inclusion data reported by Kluken and Grong,57 taking the sum (%A1 + %Ti + %S + %Si +
%Mn) equal to 100%. A closer inspection of the graphs reveals a reasonable agreement between calculated and measured average inclusion compositions in all cases, which confirms
that the model is sound. Included in Fig. 2.71 is also a collection of data for SMA lowalloy
steel weld metals (3 kJ mm"1 basic electrodes). Since these results follow the same pattern,
it implies that the model is generally applicable and, therefore, can be adopted to all relevant
arc welding processes.
Aluminium
Sulphur
Titanium
Manganese
Silicon
It can be inferred from equations (294) and (295) that the chemical composition of the inclusion oxide core is directly related to the aluminium to oxygen ratio in the weld metal.
Referring to Fig. 2.72, the fraction of MnOAl2O3 and 7Al2O3 in the inclusions is seen to increase steadily with increasing (A%Al)wdJ[%O]anal ratios until the stoichiometric composition for precipitation of aluminium oxide is reached at 1.13. At higher ratios, the deoxidation product will be pure Al2O3, since aluminium is present in an overstoichiometric amount
with respect to oxygen. When titanium is added to the weld metal, titanium oxide (in the form
of Ti2O3) may also enter the reaction product. At the same time both TiN and aMnS form
epitaxially on the surface of the inclusions during solidification. Consequently, in AlTiSiMn
deoxidised steel weld metals the total number of constituent phases within the inclusions may
approach six. The kinetics of inclusion formation are further discussed in Ref. 87.
SiO2
MnO
Fig. 2.72. Coexisting phases in inclusions at different weld metal aluminiumtooxygen ratios. Shaded
region indicates the approximate composition range for the oxide phase. The diagram is constructed on
the basis of the model of Kluken and Grong57 and relevant literature data.
Example (2.13)
Consider SA welding of lowalloy steel with two different basic fluxes. Data for the weld
metal chemical compositions are given in Table 2.12. Based on Fig. 2.72, estimate the total
number of constituent phases in the inclusions in each case.
Solution
It follows from Fig. 2.72 that the chemical composition of the deoxidation product is determined by the aluminium to oxygen ratio in the weld metal. For weld A, we have:
Since this ratio is higher than the stoichiometric factor of 1.13, all oxygen is probably tied up
as aluminium oxide. In addition, weld A contains small amounts of titanium and sulphur,
which may give rise to precipitation of TiN and MnS at the surface of the inclusions during
solidification. Hence, the three major constituent phases in the weld metal inclusions are
7Al2O3, TiN, and otMnS.
Table 2.12 Chemical composition of SA steel weld metals considered in Example 2.13.
Weld
No.
C
(wt%)
O
(wt%)
Si
(wt%)
Mn
(wt%)
S
(wt%)
N
(wt%)
Al*
(wt%)
Ti*
(wt%)
0.09
0.021
0.45
1.52
0.01
0.006
0.028
(0.003)
0.010
(0.009)
0.09
0.045
0.45
1.52
0.01
0.006
0.028
(0.003)
0.028
(0.010)
*Data for acid soluble Al and Ti are given by the values in brackets.
In the case of weld B the situation is much more complex due to a higher content of oxygen
and titanium. From Table 2.12, we have:
and
These calculations show that Al and Ti are not present in sufficient amounts to tieup all oxygen. Under such conditions Fig. 2.72 predicts that the total number of constituent phases in the
inclusions is six, i.e.: SiO2, MnOAl2O3, 7Al2O3, Ti2O3, aMnS and TiN.
References
1
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
P. Nilles, P. Dauby and J. Claes: Proc. Int. Conf. Basic Oxygen Steelmaking A New Technology Emerges, London, 1978, 6072, The Metals Society (England).
U. Lindborg and K. Torsell: Trans. TMSAIME, 1968, 242, 94102.
N.F. Grevillius: Jernkont. Ann., 1969,153, 547572.
E.T. Turkdogan: Proc. Int. Conf. on Chem. Metall. of Iron and Steel, Sheffield, July 1971,153170, Publ. The Iron and Steel Institute (England).
0 . Grong. T.A. Siewert, G.P. Martins and D.L. Olson: Metall. Trans. A, 1986,17A, 17971807.
A.O. Kluken and 0 . Grong: Metall. Trans. A, 1989, 2OA, 13351349.
M.L. Turpin and J.F. Elliott: J. Iron Steel Inst., 1966, 204, 217225.
E.T. Turkdogan: Physicochemical Properties of Molten Slags and Glasses, 1983, London, The
Metals Society.
E. Plockinger and M. Wahlster: Stahl und Eisen, 1960, 80, 659669.
BJ. Bradstreet: HW Doc. 21213868, 1968.
J.F. Lancaster: Phys. Technol, 1984,15, 7379
F. Eickhorn and A. Engel: HW Doc. 21220170, 1970.
N. Mori and Y. Horii: HW Doc. 21218870, 1970.
R.A. Walsh and S. Ramachandran: Trans. TMSAIME, 1963, 227, 560562.
D.C. Hilty and W. Crafts: Trans.AIME, 1950,188, 425436.
E.T. Turkdogan: Trans. TMSAIME, 1965, 233, 21002112.
G.M. Evans: HW Doc. IIA63084, 1984.
0 . Grong and D.K Matlock: Int. Met. Rev., 1986, 31, 2748.
A.G. Franklin: J. Iron Steel Inst., 1969, 207, 181186.
A.O. Kluken, 0. Grong and J. Hjelen: Mat. ScL Technol., 1988, 4, 649654.
L.M. Hocking: Quart. J. Royal Meterol. Soc, 1959, 85, 4450.
C. Wagner: Z Electrochemie, 1961, 65, 581591.
E.E. Underwood: Quantitative Stereology, 1970, London, AddisonWesley Publ. Co.
M.F. Ashby and R. Ebeling: Trans. TMSAIME, 1966, 236, 13961404.
R. Kiessling: NonMetallic Inclusions in Steel, 1978, London, The Metals Society (TMS).
H. Homma, S. Ohkita, S. Matsuda and K. Yamamoto: Weld. J., 1987, 66, 301s309s.
G. Thewlis: HW Doc. IIA73688, 1988.
J.M. Dowling, J.M. Corbett and H.W. Kerr: Metall. Trans. A, 1986,17A, 16111623.
G.M. Evans: Metal Constr., 1986,18, 631R636R.
J.L. Murray and H.A. Wriedt: Bull. Alloy Phase Diagr., 1987, 8, 148165.
LD. Sommerville, I. Ivanchev and H.B. Bell: Proc. Int. Conf Chem. Metall of Iron and Steel,
Sheffield, July 1971, 2325, Publ. The Iron and Steel Inst. (1973).
H.K.D.H. Bhadeshia and L.E. Svensson: Mathematical Modelling of Weld Phenomena (Eds
H. Cerjak and K.E. Easterling), 1993, London, The Institute of Materials, 109180.
S. Matsuda and N. Okumura: Trans. ISIJ, 1978,18, 198205.
M.E. Saggese, A.R. Bhatti, D.N. Hawkins and J.A. Whiteman: Proc. Int. Conf on the Effect of
Residual, Impurity and MicroAlloy ing Elements on Weldability and Weld Properties, London,
Nov. 1983, Paper 15, Publ. The Welding Institute (England).
A.O. Kluken and 0 . Grong: Report No. STF34 F87125, 1987, Sintef, Trondheim, Norway.
S.S. Babu, S.A. David, J.M. Vitek, K. Mundra and T. DebRoy: Mater. ScL Technol., 1995,11,186199.
Appendix 2.1
Nomenclature
contact area (mm2)
activity coefficient of
element X
difference between
total and acid soluble Al in
weld metal (wt%)
gravity constant
(9.81ms 2 )
thermal diffusivity
(mm2 s 1 )
GTAW
drag coefficient
molar concentration of
element X in the liquid
(mol m~3)
cross section of deposited
metal (mm2)
element diffusivity in liquid
phase (m2 s"1 or mm2 s"1)
diffusivity of element X in
gas phase (mm2 s"1)
gross heat input (kJ mm"1)
FCAW
standard enthalpy of
reaction (J mol1 or kJ
mol1)
hydrogen content related
to deposited metal (ml per
10Og deposit)
hydrogen content related
to fused metal (ml per 10Og
or g per ton)
amperage (A)
dimensionless operating
parameter
absorption/desorption rate
of element X in the weld
pool (mol s"1)
concentration displacement
of silicon referred to
different standard states
(wt%)
SAW
SMAW
STEM
scanning transmission
electron microscope
volume fraction of
inclusions
concentration of component / (vol%)
concentration of component X2 (vol%)
time (s)
time constant (s)
acid soluble titanium in
weld metal (wt%)
difference between total
and acid soluble Ti in weld
metal (wt%)
temperature (0C or K)
concentration of
component YX2 (vol%)
WDX
WFR
concentration of element X
in the weld metal (wt%)
concentration displacement
of element X referred to
nominal weld metal
composition (wt%)
equilibrium concentration
of element X in the weld
pool (wt%)
voltage (V)
initial concentration of
element X in the weld pool
(wt%)
nominal concentration of
element X in weld metal
(wt%)
concentration of element X
in base plate (wt%)
concentration of element X
in filler wire (wt%)
arbitrary element or
gaseous species
density (kg irr 3 or
g mm"3)
thickness of stagnant
gaseous boundary layer
(mm)
wetting angle
mean particle centre to
centre volume spacing
(jxm)
difference in density
between liquid steel and
inclusion (kg irr 3 or
g mm"3)
dimensionless ycoordinate
at maximum width of
isotherm
slag/metal interfacial
energy (J nr 2 )
Appendix 2.2
Derivation of equation (260)
The nucleation rate / as a function of temperature can be expressed as:
(A21)
where J0 is a constant (with the unit nuclei per m3 and s) and AG* is the energy barrier for
nucleation.
By rearranging equation (A21) and inserting reasonable values for J and J0 for the specific
case of homogeneous nucleation of oxide inclusions in liquid steel,55 we obtain:
(A22)
(A23)
where NA is the Avogadro constant, o is the interfacial energy between the nucleus and the
liquid (in J nr 2 ) and AGV is the driving force for the precipitation reaction (in J nr 3 ).
(A24)
where AH and AS are the standard entalpy and entropy of the precipitation reaction, respectively and Vm is the molar volume of the nucleus (in m3 mol"1).
It is evident from Fig. 2.48 that AGV = 0 when T=Tu which gives \S = A//7 T\. Hence,
equation (A24) may be rewritten as:
(A25)
By combining equations (A22), (A23) and (A25), we obtain the following relationship
between T\ and T^
(A2 6)
3
Solidification Behaviour of Fusion Welds
3.1 Introduction
Inherent to the welding process is the formation of a pool of molten metal directly below the
heat source. The shape of this molten pool is influenced by the flow of both heat and metal,
with melting occurring ahead of the heat source and solidification behind it. The heat input
determines the volume of molten metal and, hence, dilution and weld metal composition, as
well as the thermal conditions under which solidification takes place. Also important to solidification is the crystal growth rate, which is geometrically related to weld travel speed and weld
pool shape. Hence, weld pool shape, weld metal composition, cooling rate, and growth rate
are all factors interrelated to heat input which will affect the solidification microstructure.
Some important points regarding interpretation of weld metal microstructure in terms of these
four factors will be discussed below.
Since the properties and integrity of the weld metal depend on the solidification microstructure, a verified quantitative understanding of the weld pool solidification behaviour is essential. At present, our knowledge of the chemical and physical reactions occurring during solidification of fusion welds is limited. This situation arises mainly from a complex sequence of
reactions caused by the interplay between a number of variables which cannot readily be accounted for in a mathematical simulation of the process. Nevertheless, the present treatment
will show that it is possible to rationalise the development of the weld metal solidification
microstructure with models based on well established concepts from casting and homogenising
treatment of metals and alloys.
Shrinkage pipe
Chill zone
Columnar zone
Equiaxed zone
Mould
Fig. 3.1. Transverse section of an ingot showing the chill zone, the columnar zone and the equiaxed zone
(schematic).
grains will change continuously from the fusion line towards the centre of the weld due to a
corresponding shift in the direction of the maximum temperature gradient in the weld pool.
This change in orientation may result in a curvature of the columnar grains (Fig. 3.2(a)). Alternatively, new grains can nucleate and grow in a columnar manner, producing a socalled 'stray'
structure as shown schematically in Fig. 3.2(b). Finally, if the conditions for nucleation of new
grains are favourable, an equiaxed zone will form near the weld centreline similar to that
observed in ingots or castings (see Fig. 3.2(b)).
Although the process of weld pool solidification is frequently compared with that of an
ingot in 'miniature', a number of basic differences, already mentioned, exist which strongly
influence the microstructure and properties of the weld metal. Of particular importance is also
the disparity in cooling rate between a fusion weld and an ingot (see Fig. 3.3). For conventional processes such as shielded metal arc (SMA), gas metal arc (GMA), submerged arc (SA)
or gas tungsten arc (GTA) welding the cooling rate may vary from 10 to 103 0 C s"1, while for
modern high energy beam processes such as electron beam (EB) and laser welding the cooling
rate is typically of the order of 103 to 106 0 C s"1.4 Consequently, to appreciate fully the implications of these differences in general solidification behaviour between a weld pool and an
ingot, it is necessary to consider in detail the sequence of events taking place in the solidifying
weld metal beginning with the initiation of crystal growth at the fusion boundary.
(a)
Welding direction
(b)
Columnar zone
Welding direction
Equiaxed zone
Columnar zone
SMAW, SAW,
GMAW, GTAW
Fig. 3.2. Examples of structural zones in fusion welds (schematic); (a) Curved columnar grains,
(b) Stray grain structure.
Process
Fig. 3.3. Disparity in cooling conditions between casting, welding and rapid solidification.
Fusion
boundary
HAZ
Weld metal
Fig. 3.4. Schematic illustration showing epitaxial growth of columnar grains from partly melted base
metal grains at fusion boundary.
Liquid (L)
Substrate (S)
where VE is the volume of the solid embryo, AGV is the free energy change associated with the
embryo formation, AEL and AES are the areas of the embryoliquid and embryosubstrate interfaces, respectively, and/(P) is the socalled shape factor, defined as:
(33)
The critical radius of the stable nucleus, r / , is found by differentiating equation (32) with
respect to rs and equating to zero:
(34)
By substituting equation (34) into equation (32), we obtain the following expression for
the energy barrier to heterogeneous nucleation (AG^ r ):
(35)
where AHm is the latent heat of melting, Tm is the melting point, and AJT is the undercooling.
It is easy to verify that the first term in equation (35) is equal to the energy barrier to
homogeneous nucleation, AG^om. Hence, we may write:
(36)
Equation (36) shows that AG^ is a simple function of the wetting angle O). Since the
chemical composition and the crystal structure of the two solid phases are usually very similar,
we have:6
Under such conditions equation (31) predicts that the wetting angle 3 ~ 0 (cos(3 ~ 1),
which implies that there is a negligible energy barrier to solidification of the weld metal (}*het
~ 0), i.e. no undercooling of the melt is needed, and solidification occurs uniformly over the
whole grain of the base metal. This is in sharp contrast to conventional casting of metals and
alloys where some undercooling of the melt is always required to overcome the inherent energy barrier to solidification (see Fig. 3.6).
3.3.2 Implications of epitaxial solidification
Since the initial size of the weld metal columnar grains is inherited directly from the grain
growth zone adjacent to the fusion boundary, the solidification microstructure depends on the
grain coarsening behaviour of the base material. This is particularly a problem in high energy
processes such as submerged arc and gas metal arc welding, where grain growth of the base
metal can be considerable. In such cases the size of the columnar grains at the fusion boundary
will be correspondingly coarse, as indicated by the data in Fig. 3.7.
Moreover, during multipass welding the columnar grains can renucleate at the boundary
between for instance the first and the second weld pass and subsequently grow across the
entire fusion zone, as illustrated in Fig. 3.8. This type of behaviour is usually observed in
weldments which do not undergo transformations in the solid state (e.g. aluminium, certain
titanium alloys, stainless steel etc.). In practice, the problem can be eliminated by additions of
inoculants via the filler wire, which facilitates a refinement of the columnar grain structure
through heterogeneous nucleation of new (equiaxed) grains ahead of the advancing interface
(to be discussed later).
AG
Welding
Casting
Homogeneous
nucleation
Fig. 3.6. The free energy change associated with heterogeneous nucleation during casting and weld
metal solidification, respectively (schematic). The corresponding free energy change associated with
homogeneous nucleation is indicated by the broken curve in the graph.
Weld metal
prior austenite grain size (jim)
Fusion
line
HAZ
Weld
metal
2. pass
1. pass
HAZ
Base metal
Fig. 3.8. Optical micrograph showing renucleation of columnar grains during multipass GMA welding
of a Ptitanium alloy.
HAZ isotherms
Fusion boundary
Weld
pool
(a)
HAZ isotherms
Fusion boundary
Weld
pool
(b)
Fig. 3.9. Theoretical shape of fusion boundary and neighbouring isotherms under different operational
conditions; (a) High n3 values, (b) Low ^values.
merited for ordinary GTA austenitic stainless steel welds. 1 2 1 3 The indications are that such
effects become even more important under hyperbaric welding conditions. 14
Electrode
Weld pool
Arc
(a)
Electrode
Weld pool
Arc
(b)
Fig. 3.10. Schematic diagrams illustrating the major fluid flow mechanisms operating in a weld pool;
(a) Buoyancy force (b) Electromagnetic force.
columnar grains as shown in Fig. 3.11(b), since the direction of the maximum temperature
gradient in the melt does not change significantly during the solidification process. The latter condition is known to promote formation of centreline cracking because of mechanical
entrapment of inclusions and enrichment of eutectic liquid at the trailing edge of the weld
pool.
3.4.3 Growth rate of columnar grains
The growth rate of the columnar grains is geometrically related to the weld travel speed and
the weld pool shape.
3.4.3.1 Nominal crystal growth rate
Since the shape of the weld pool remains constant during steady state welding, the growth rate
of the columnar grains must vary with position along the fusion boundary. This point is more
clearly illustrated in Fig. 3.12 which shows a sketch of a single columnar grain growing parallel with the steepest temperature gradient in the weld pool. Taking the angle between the
Surface tension
Temperature
Electrode
Weld pool
Arc
Surface tension
(C)
Temperature
Electrode
Arc
Weld pool
(d)
Fig. 3.10. Schematic diagrams illustrating the major fluid flow mechanisms operating in a weld pool
(continued); (c) Surface tension gradient force (negative gradient); (d) Surface tension gradient force
(positive gradient).
Heat source
Heat source
(a)
(b)
Fig. 3.11. Schematic comparison of columnar grain structures obtained under different welding conditions; (a) Elliptical weld pool (low n3 values), (b) Tearshaped weld pool (high n3 values). Open arrows
indicate the direction of the maximum temperature gradient in the weld pool.
Fusion boundary
Heat source
Crystal
Fig. 3.12. Definition of the nominal crystal growth rate RN.
growth direction and the welding direction equal to a, the steady state growth rate, R N , becomes:
(38)
where v is the welding speed.
Considering spherical or elliptical weld pools, the nominal crystal growth rate is lowest at
the edge of the weld pool (a>90, cosa^0) and highest at the weld centreline where R N
approaches v (a>0, c o s a ^ l ) . In contrast, columnar grains trailing behind a tearshaped weld
pool will grow at an approximately constant rate which is significantly lower than the actual
welding speed (a 0), since the direction of the maximum temperature gradient in the weld
pool does not change during the solidification process. This is also in agreement with practical
experience (see Fig. 3.13).
(a)
Nominal growth rate (RN), mm/s
(b)
Equiaxed zone
Example (3.1)
Consider electron beam (EB) welding of a lmm thin sheet of austenitic stainless steel under
the following conditions:
Estimate on the basis of the Rosenthal thin plate solution (equation 183) the steady state
growth rate of the columnar grains trailing the weld pool.
Solution
The contour of the fusion boundary can be calculated from the Rosenthal thin plate solution
according to the procedure shown in Example (1.10). If we include a correction for the latent
heat of melting, the QbZn3 ratio at the melting point becomes:
Substitution of the above value into equation (183) gives the fusion boundary contour
shown in Fig. 3.14. It is evident from Fig. 3.14 that the weld pool is very elongated under the
prevailing circumstances due to a constrained heat flow in the ^direction. This implies that the
angle a will not change significantly during the solidification process. Taking a as an average,
equal to about 70, the steadystate crystal growth rate R N becomes:
This value is in reasonable agreement with the measured crystal growth rates in Fig. 3.13(b).
3.4.3.2 Local crystal growth rate
Equation (38) does not take into account the inherent anisotropy of crystal growth. For faceted materials the dendrite growth directions are always those that are 'capped' by relatively
slowgrowing (usually lowindex) crystallographic planes.1 Figure 3.15 shows examples of
faceted cubic crystals delimited by {100} and {111} planes, respectively. If the {111} planes
are the slowest growing ones, the {100} planes will grow out, leaving the {111} facets and a
new crystal growing in the <100> directions as shown schematically in Fig. 3.15(b).
Although most metals and alloys do not form faceted dendrites, the anisotropy of crystal
growth is still maintained during solidification.2 In fact, experience has shown that the major
dendrite growth direction is normally the axis of a pyramid whose sides are the most closely
packed planes with which a pyramid can be formed.1 These directions are thus <100> for
body and facecentred cubic structures, < 1010 > for hexagonal closepacked structures, and
<110> for bodycentred tetragonal structures.
Because of the existence of preferred growth directions, the local growth rate of the crystals
RL will always be higher than the nominal growth rate R N defined in equation (38). Consider
now a cubic crystal which grows along the steepest temperature gradient in the weld pool, as
shown schematically in Fig. 3.16. If denotes the angle between the interface normal and the
<100> direction, the following relationship exists between RN and RL:
y(mm)
Columnar zone
Heat source
Equiaxed zone
+x(mm)
Columnar zone
+yjmm)
Fusion boundary
Fig. 3.14. Predicted shape of fusion boundary during electron beam welding of austenitic stainless steel
(Example (3.1)).
(a)
(b)
Fig. 3.25.Examples of faceted cubic crystals; (a) Crystal delimited by {100} planes, (b) Crystal delimited by {111} planes.
Columnar grain
Tip temperature, 0C
Liquidus temperature
Tip velocity, mm/sFig. 3.17. Calculated dendrite tip temperature vs dendrite growth velocity for an Fe15Ni15Cr alloy.
The undercooling of the dendrite tip is given by the difference between the liquidus temperature and the
solid curve in the graph. Data from Rappaz et alP
(39)
which gives:
(310)
Equation (310) shows that the local growth rate increases with increasing misalignment of
the crystal with respect to the direction of the maximum temperature gradient in the weld pool.
Since such crystals cannot advance without a corresponding increase in the undercooling ahead
of the solid/liquid interface (see Fig. 3.17), they will soon be outgrowed by other grains which
have a more favourable orientation. Fusion welds of the fee and bcc type will therefore develop a sharp <100> solidification texture in the columnar grain region, similar to that documented for ingots and castings. The weld metal columnar grains may nevertheless be separated by 'highangle' boundaries, as shown in Fig. 3.1&, due to a possible rotation of the grains
in the plane perpendicular to their <100> length axes.
Example (3.2)
Consider electron beam welding of a 2mm thick single crystal disk of Fe15Ni15Cr under
the following conditions:
The orientation of the disk with respect to the beam travel direction is shown in Fig. 3.19.
Calculate on the basis of the minimum velocity (undercooling) criterion the growth rate of
the dendrites trailing the weld pool under steady state welding conditions (assume 2D
heat flow). Make also schematic drawings of the solidification microstructure in different
sections of the weld. Relevant thermal properties for the Fe15Ni15Cr single crystal are
given below:
Solution
Since the base metal is a single crystal, separate columnar grains will not develop. Nevertheless, under 2D heat flow conditions growth of the dendrites can occur both in the [100] and
the [010] (alternatively the [010]) direction. Referring to Fig. 3.20 the growth rate of the
[100] and the [010] deridrites is given by:
and
Fig. 3.18. Spatial misorientation between two columnar grains growing in the <100> direction (schematic).
Heat source
Weld
Fig. 3.19. Orientation of the single crystal Fe15Ni15Cr disk with respect to beam travel direction (Example (3.2)).
Welding direction
Fig. 3.20. Schematic diagram showing the pertinent orientation relations between the fusion boundary
interface normal and the dendrite growth directions (Example (3.2)).
From this it is seen that the velocity of the [100] dendrites is always equal to that of the heat
source v. In contrast, the growth rate of the [010] dendrites depends both on v and a, and will
therefore vary with position along the fusion boundary. It follows from minimum velocity
criterion that the [100] dendrites will be selected when the interface normal angle a is less than
45, while the [010] dendrites will develop at larger angles. This is shown graphically in Fig.
3.21.
At pseudosteady state the fusion boundary can be calculated from the Rosenthal thin plate
solution (equation (183)) according to the procedure shown in Example (1.10). If we include
hk ,
/V
dendrites
a, degrees
Fig. 3.21.Normalised minimum dendrite tip velocity vs interface normal angle a (Example 3.2)).
a correction for the latent heat of melting, the QbIn3 ratio at the melting point becomes:
Substitution of this value into equation (183) gives the fusion boundary contour shown in
Fig. 3.22(a). Included in Fig. 3.22 are also schematic drawings of the predicted solidification
microstructure in different sections of the weld.
The results in Fig. 3.22 should be compared with the reconstructed 3D image of the solidification microstructure in Fig. 3.23, taken form Rappaz et al.17 Due to partial heat flow in the
zdirection, [001] dendrite trunks will also develop. Nevertheless, these data confirm the general validity of equations (38) and (310) relating crystal growth rate to welding speed and
weld pool shape.
3.4.4 Reorientation of columnar grains
In principle, there are two different ways a columnar grain can adjust its orientation during
solidification in order to accommodate a shift in the direction of the maximum temperature
gradient in the weld pool, i.e.:
(i)
(ii)
Through bowing
Through renucleation.
y (mm)
(a)
dendrites
Heat source
dendrites
+x (mm)
dendrites
+y (mm)
Fusion boundary
(b)
Base plate
dendrites
dendrites
dendrites
Fig. 3.22. Schematic representation of the weld metal solidification micro structure (Example 3.2));
(a) Top view of fusion zone, (b) Transverse section of fusion zone.
Consider a curved columnar grain of iridium which grows from the fusion boundary towards
the weld centreline along a circle segment of length L, as shown schematically in Fig. 3.24.
Based on the assumption that the bowing is accommodated solely by branching of [010] dendrites
in the [100] direction, calculate the maximum local growth rate of the crystal during solidification.
y
X
2
Fig. 3.23. Reconstructed 3D image of solidification microstructure in an electron beam welded Fe15Ni15Cr single crystal. The letters (a), (b) and (c) refer to [100], [010] and [001] type of dendrites, respectively. After Rappaz et al.17
Weld centreline
Fusion line
and
It follows from Fig. 3.21 that growth will occur preferentially in the [010] direction as long
as the interface normal angle a is larger than 45, while the [100] direction is selected at
smaller angles. This means that the local growth rate of the dendrites, in practice, never will
exceed the welding speed v.
3.4.4.2 Renucleation of crystals
In ingots and castings, three different mechanisms for nucleation of new grains ahead of the
advancing interface are operative:12
(i)
(ii)
(iii)
Heterogeneous nucleation
Dendrite fragmentation
Grain detachment.
The former mechanism is of particular importance in welding, since the weld metal often
contains a high number of second phase particles which form in the liquid state. These particles can either be primary products of the weld metal deoxidation or stem from reactions
between specific alloying elements which are deliberately introduced into the weld pool through
the filler wire. The latter process is also known as inoculation.
Nucleation potency of second phase particles
In general, the effectiveness of individual particles to act as heterogeneous nucleation sites can
be evaluated from a balance of interfacial energies, analogous to that described in Section
3.3.1 for epitaxial nucleation. It follows from the definition of the wetting angle (3 in Fig. 3.5
that the energy barrier to heterogeneous nucleation is a function of both the substrate/liquid
interfacial energy ySL, the substrate/embryo interfacial energy yES, and the embryo/liquid interfacial energy yEL. Complete wetting is achieved when:
(311)
Under such conditions, the nucleus will readily grow from the liquid on the substrate. Unfortunately, data for interfacial energies are scarce and unreliable, which makes predictions
based on equation (311) rather fortuitous.18
In pure metals, experience has shown that the solid/liquid interfacial energies are roughly
proportional to the melting point, as shown by the data in Fig. 3.25. On this basis, it can be
expected that the higher melting point phases will reveal the highest ySL values, and thus be
nucleants for lower melting phases. A similar situation also exists in the case of nonmetallic
inclusions in liquid steel, where the highmelting point phases are seen to exhibit the highest
solid/liquid interfacial energies (see Fig. 3.26).
In contrast, very little information is available on the substrate/embryo interfacial energy
yES. For fully incoherent interfaces, yES would be expected to be of the order of 0.5 to 1 J m~2.5
However, this value will be greatly reduced if there is epitaxy between the inclusions and the
nucleus, which results in a low lattice disregistry between the two phases. In general, assessment of the degree of atomic misfit between the nucleus n and the substrate s can be done on
Melting point, K
Interfacial energy, J / m 2
Fig. 3.25. Values of solid/liquid interfacial energy ySL of various metals as function of their melting points.
Data from Mondolfo.18
Melting point, 0 C
Fig. 3.26. Values of interfacial energy 7 5L for different types of nonmetallic inclusions in liquid steel at
16000C as function of their melting points. Data compiled from miscellaneous sources.
(312)
a lowindex
a lowindex
a lowindex
a lowindex
where
Undercooling, 0C
>
Fig. 3.27. Relationship between planar lattice disregistry and undercooling for different nucleants in
steel. Data compiled from miscellaneous sources.
Example (3.4)
In lowalloy steel weld metals, titanium nitride can form in the melt due to interactions between dissolved titanium and nitrogen. Assume that the TiN particles are faceted and delimited by {100} planes. Calculate on the basis of equation (312) the minimum planar lattice
disregistry between TiN and the nucleating deltaferrite phase under the prevailing circumstances. Indicate also the plausible orientation relationship between the two phases. The
lattice parameters of delta ferrite and TiN at 15200C may be taken equal to 0.293 and 0.43 lnm,
respectively.
Solution
Titanium nitride has the NaCl crystal structure, while delta ferrite is bodycentred cubic, as
shown in Fig. 3.28(a) and (b). It is evident from Fig. 3.29(a) that a straight cubetocube orientation relationship between TiN and 8Fe will not result in a small lattice disregistry.
However, the situation is largely improved if the two phases are rotated 45 with respect to
each other (see Fig. 3.29(b)), conforming to the following orientation relationship:
A comparison with the data in Fig. 3.27 shows that the calculated lattice disregistry conforms to an undercooling of about 1 to 2C. This value is sufficiently small to facilitate heterogeneous nucleation of new grains ahead of the advancing interface during solidification.
Considering other inclusions with more complex crystal structures, the chances of obtaining a small planar lattice disregistry between the substrate and the delta ferrite nucleus are
Fe atoms
Natoms
Ti atoms
(a)
(b)
Fig. 3.28. Crystal structures of phases considered in Example (3.4); (a) Titanium nitride, (b) Delta ferrite.
TiN
(a)
TiN
(C)
(b)
Ti atoms
N atoms
8Fe atoms
Fig. 3.29. Possible crystallographic relationships between titanium nitride and delta ferrite (Example (3.4));
(a) Straight cubetocube orientation, (b) Twisted cubetocube orientation, (c) Details of lattice arrangement along coherent TiN/dFe interface.
rather poor (see Fig. 3.27). Nevertheless, such particles can act as favourable sites for heterogeneous nucleation if 7 ^ is sufficiently large compared with yEL and 7 ^ . This is illustrated by
the following example:
Example (3.5)
In lowalloy steel weld metals 7Al2O3 inclusions can form during the primary deoxidation
stage as discussed in Section 2.12.4.2 (Chapter 2). Based on the classic theory of heterogeneous nucleation, evaluate the nucleation potency of such inclusions with respect delta ferrite.
Solution
It is readily seen from Fig. 3.27 that the planar lattice disregistry between delta ferrite and
Al2O3 is very large, which indicates of a fully incoherent interface (i.e. yES 0.75 J m" 2 ).
Moreover, readings from Figs. 3.25 and 3.26 give the following average values for the delta ferrite/liquid and the inclusion/liquid interfacial energies:
and
According to equation (311) complete wetting is achieved when ySL > yES + yEL. This
requirement is clearly met under the prevailing circumstances.
Similar calculations can also be performed for other types of nonmetallic inclusions in
steel weld metals. The results are presented graphically in Fig. 3.30. It is evident that the
nucleation potency of the inclusions increases in the order SiO2MnO, Al 2 O 3 Ti 2 O 3 SiO 2 MnO, Al2O3, reflecting a corresponding increase in the inclusion/liquid interfacial energy ySL.
The resulting change in the weld metal solidification microstructure is shown in Fig. 3.31,
from which it is seen that both the average width and length of the columnar grains decrease
with increasing Al2O3contents in the inclusions. This observation is not surprising, considering the characteristic high solid/liquid interfacial energy between aluminium oxide and steel
(see Fig. 3.26). The important effect of deoxidation practice on the weld metal solidification
microstructure is well documented in the literature.320"22
Rate of heterogeneous nucleation
It can be inferred from the classic theory of heterogeneous nucleation that the nucleation rate
Complete wetting
No wetting
Embryo'
G*he/AGhom
p (degrees)
Inclusion
(Y sf Y ES )/7 EL
Fig. 3.30. Nucleation potency of different weld metal oxide inclusions with respect to delta ferrite.
(a)
Pure AI2O3
(A%AI)weld/[%O]anaL
AI2O3 content (wt%)
(b)
<A%AIW[%0W
Fig. 3.31.Effect of deoxidation practice on the columnar grain structure in lowalloy steel weld metals;
(a) Average width of columnar grains, (b) Average length of columnar grains. Data from Kluken et al.22
Nhet(i.e. number of nuclei which form per unit time and unit volume of the melt) is interrelated to the energy barrier AG^ through the following equation:5
(313)
whereZ1 is a frequency factor, Nv is the density of nucleation sites per unit volume of the melt,
AGD is the activation energy for diffusion of atoms across the interface, and k is the Boltzmann
constant.
Since AGD is often negligible compared with AG^et in liquids, equation (313) reduces to:
(314)
Equation (314) shows that the nucleation rate Nhet depends both on Nv and &Ghe{. Hence,
under full wetting conditions (AGhet ~ 0), the number of nuclei which form per second and mm3
ahead of the advancing interface is directly proportional to the instantaneous concentration of
catalyst particles in the melt. Examples of such particles are TiAl3 in aluminium and TiN/Al2O3
in steel. The important effect of controlled titanium additions and subsequent TiAl3 precipitation on the columnar grain structure in 1100 aluminium welds is illustrated in Fig. 3.32.
where co is the total grain rotation angle, and / is the average length of the columnar grains (in
mm).
Weld metal
Fig. 3.33. Characteristic growth pattern of columnar grains in beadonplate welds (schematic).
By introducing reasonable average values for co and / in the case of SA welding of lowalloy steel,22 we obtain:
(316)
Calculated values for 4>* in steel weld metals are presented in Fig. 3.34, using data from
Kluken et al.22 An expected, the critical cell/dendrite alignment angle in fully aluminium
deoxidised steel welds is seen to be very small (of the order of 2), reflecting the fact that
nucleation of delta ferrite occurs readily at Al2O3 inclusions. The value of 4>* increases gradually with decreasing Al2O3 contents in the inclusions and reaches a maximum of about 4 for
SiMn deoxidised steel weld metals. This situation can be attributed to less favourable nucleating opportunities for delta ferrite at silicacontaining inclusions, which reduces the possibilities of obtaining a change in the crystal orientation during solidification through a nucleation
and growth process.
Dendrite fragmentation
In principle, nucleation of new grains ahead of the advancing interface can also occur from
random solid dendrite fragments contained in the weld pool. Although the source of these
solid fragments has yet to be investigated, it is reasonable to assume that they are generated by
some process of interface fragmentation due to thermal fluctuations in the melt or mechanical
disturbances at the solid/liquid interface.3 At present, it cannot be stated with certainty whether
grain refinement by dendrite fragmentation is a significant process in fusion welding.26
Grain detachment
Since the partially melted base metal grains at the fusion boundary are loosely held together by
liquid films between them, there is also a possibility that some of these grains may detach
themselves from the base metal and be trapped in the solidification front.26 Like dendrite
fragments, such partially melted grains can act as seed crystals for the formation of new grains
in the weld metal during solidification if they are able to survive sufficiently long in the melt.
R L / v cos a
Next Page
Calculated from
equation (310)
R L / v cos a
Previous Page
Calculated from
equation (310)
Cell wall'
Fig. 3.36. Optical micrograph showing the characteristic cellulardendritic substructure in a lowalloy
steel weld. The metallographic section is normal to the columnar grain growth direction. After Kluken
etal.22
A characteristic feature of cellular and cellulardendritic growth is also that the boundary
between two adjacent columnar grains will closely follow the contours of the original cell
boundaries. An illustration of this point is contained in Fig. 3.38. Consequently, since all cell
walls are preferential sites for segregation during solidification (see ion micrograph in Fig.
3.39), the presence of solute at the columnar grain boundaries can strongly alter the kinetics of
subsequent solid state transformation reactions. The indications are that for instance phosphorus segregations at prior austenite grain boundaries will promote the formation of grain boundary ferrite in lowalloy steel weld metals during the 7 to a transformation because of the
associated increase in the A<?3temperature.22'25 We will return to this question in Chapter 6.
(a)
(b)
Fig. 3.39. Ion (SIMS) micrograph showing evidence of phosphorus segregations at primary solidification (cell) boundaries in a lowalloy steel weld. After Kluken et al22
(317)
Provided that equilibrium exists at the solid/liquid interface, i.e. (Cs)t = IcJC1J1, equation
(317) can be rewritten as:
(318)
where DL is the diffusivity of the solute in the liquid, and ko is the equilibrium partition coefficient (Jco < 1).
Temperature
(a)
Concentration
Liquid composition
(b)
Distance (x)
Temperature
(C)
Distance (x)
Fig. 3.40. Constitutional undercooling in alloy solidification; (a) Schematic representation of binary phase
diagram, (b) Buildup of soluteenriched layer in front of solid/liquid interface, (c) Undercooled region
ahead of solid/liquid interface.
With the aid of the phase diagram in Fig. 3.40(a) it is easy to verify that the equilibrium
liquidus temperature increases with distance from the interface because of the lower solute
content. Referring to Fig. 3.40(c) the latent heat of melting AHm will diffuse away from the
interface (thereby stabilising possible interface protuberances) if the actual temperature gradient in the liquid (dTA/dx)t is less than the equilibrium liquidus temperature gradient (3T1ZdX)1.
The latter gradient can be expressed as:
(319)
where (dTL/dCL)i denotes the slope of the liquidus curve in the phase diagram (designated mL).
By combining equations (318) and (319), we obtain the following criterion for the interface stability:
(320)
Taking (C1). = C0Ik0 and Rj = RL this equation can alternatively be written as:
(321)
In aluminium welding, binary AlSi and AlMg alloys are frequently used as filler metals.
Consider stringer bead deposition with an Al4.5wt%Si filler wire at a constant welding speed
of 1.5mm s"1. Based on equation (321) calculate the critical (minimum) temperature gradient
Alloying level (C )
Planar
GL/RL
Fig. 3.41. Schematic representation of the combined effect of crystal growth rate RL and melt thermal
gradient GL on the weld metal solidification microstructure.
Fusion zone
v
Larger R L,
Small RL, Small GL
Large G1
Weld centreline
Equiaxeddendritic
Cellulardendritic
Cellular
HAZ grains
Fig. 3.42. Schematic diagram illustrating structural variations in the weld metal solidification microstructure across the fusion zone.
in the weld pool which gives a planar solidification front. Relevant physical data for the AlSi
system are given below:2
Solution
When mL, ko and D1 are known the critical temperature gradient (G1)cr can readily be calculated from equation (321). Taking RL ~ v, we obtain:
The above calculations show that the temperature gradient in the weld pool must be extremely large in order to promote a planar solidification front. The calculated value for (GL)cr
corresponds to a cooling rate of about 92 4000C s"1 (CR. = G1R1). Besides lowheat input
electron beam and laser welding, such high cooling rates are rarely observed in fusion welding
(see Fig. 3.3). On this basis it is not surprising to find that the weld metal solidification microstructure is normally of the cellular or the dendritic type.
For a rough evaluation of the weld metal solidification microstructure the diagram in Fig.
3.43 can be used. This diagram summarises the various microstructures which can be obtained
(using a typical alloy with a melting range of 500C) when the imposed temperature gradient GL
and crystal growth rate RL are varied. Moving from the lower left to the upper right along the
lines at 45 leads to a refinement of the structure without changing the morphology (constant
GL/RL ratio). Crossing these lines by passing from the upper left to the lower right leads to
changes in the morphology from planar to cellular or dendritic growth, while the scale of the
microstructure remains essentially the same. The gray bands define the regions over which
one structure changes into another. An example of the application of this diagram is given
below.
Example (3.7)
Consider GTA butt welding of a 2mm thin sheet of aluminium (AlMg alloy) under the following conditions:
Based on the diagram in Fig. 3.43, estimate the weld metal solidification microstructure in
different positions from the fusion boundary. In these calculations we shall assume that the
actual thermal gradient in the weld pool is equal to the average thermal gradient within the
temperature interval from 680 to 5500C, as evaluated from the Rosenthal thin plate solution
(equation (183)). Relevant thermal data for the AlMg alloy are given in Table 1.1 (Chapter
1).
Solution
The contours of the fusion boundary and the 680 and the 5500C isotherms can be calculated
according to the procedure shown in Example (1.10). If we neglect the latent heat of melting,
the corresponding 68//?3 ratios at these temperatures become:
GL,C/mm
RL, mm/s
Fig. 3.43. Variation of weld metal solidification microstructure with GL and RL. The diagram is based on
the ideas of Kurz and Fisher.2
Substitution of the above values into equation (183) gives the isothermal contours shown
in Fig. 3.44. It follows from Fig. 3.44 that the thermal gradient decreases from about 25C
mnr 1 at the fusion boundary to approximately 100C mm"1 close to the weld centreline. This
occurs parallel with an increase in the nominal crystal growth rate from 0.8 to 4mm s"1.
According to Fig. 3.43 the calculated values of GL and RL conform to a cellulardendritic
solidification microstructure within the central regions of the weld and an equiaxeddendritic
microstructure close to the weld centreline. Both types of substructures are commonly observed in aluminium weldments.2728
Weld pool (680 0C)
Fusion boundary (650 0C)
HAZ (550 0C)
+x (mm)
+Y (mm)
Fig. 3.44. Predicted shape of fusion boundary and neighbouring isotherms during thin plate aluminium
welding (Example (3.6)).
Tip radius,\irr\
(323)
This gives the following relationship between the primary dendrite arm spacing X1, the tip
radius rd and the thermal gradient GL:
(324)
rd
b
Fig. 3.46. Definition of primary X1 and secondary X2 dendrite
arm spacings.
Ttip
Troot
9
Heat flow
Cell/dendrite
Fig. 3.48. Definition of the dendrite tip Ttip and root Troot temperatures.
If we also take into account that the dendrite tip radius rd is inversely proportional to the
square root of the crystal growth rate RL within the central range of RL (see Fig. 3.45), equation
(324)reduces to:
(325)
where c{ is a kinetic constant which is characteristic of the alloy system under consideration.
It is evident from the above analysis that the primary dendrite arm spacing cannot readily
be characterised by one single parameter (e.g. the cooling rate), since its dependence on GL
and R1 have different exponents.
Example (3.8)
Based on equation (325), show that the following relationships exist between the primary
dendrite arm spacing X1, the net arc power qo, and the weld travel speed v during thick plate
and thin plate welding, respectively:
Thick plate welding:
(326)
Thin plate welding:
(327)
where c2 and c3 are proportionality constants.
Solution
Under 3D heat flow conditions, the pseudosteady state temperature distribution is given by
the Rosenthal thick plate solution (equation (1 45)). For points located on the weld centreline
behind the heat source y = z = 0, and R* = x. Hence, equation (145) reduces to:
The thermal gradient GL in the mushy zone close to the weld centreline can be obtained
by differentiating of the above equation with respect to the xcoordinate:
(328)
By inserting the appropriate expressions for GLandRL into equation (325) (noting that RL=v
at the weld centreline), we arrive at an expression for X1 which is identical with the one
presented in equation (326).
Similarly, under 2D heat flow conditions, the pseudosteady state temperature distribution
is given by the Rosenthal thin plate solution (equation 181). For points located on the weld
centreline behind the heat source y = 0 and r = x = vt. If IxI is sufficiently large, it is a fair
approximation to set Ko(u) ~ exp(u)^n!2u
. Hence, equation (181) reduces to:
(329)
From this we see that the primary dendrite arm spacmg X1 during thin plate welding is
interrelated to qo and v through a relationship of the type shown in equation (327).
In Fig. 3.49 the validity of equation (326) has been checked against the experimental data
of Jordan and Coleman,27 who measured the primary dendrite arm spacing in different GMA
AlMgMn welds. It is evident from this plot that their datapoints can approximately be
represented by straight lines passing through the origin, as required by the theory. Moreover,
a closer inspection of the figure shows that the dendrite arm spacing close to the centreline
varies systematically from the bottom to the top of the weld. This observation is not surprising, considering the fact that the rate of solidification increases progressively from the toe to
the surface of the plate.
Example (3.9)
Consider GTA butt welding of a 2mm thin aluminium sheet (AlMg alloy) under conditions
similar to those employed in Example (3.7). Based on equation (325) calculate the relative
change in the primary dendrite arm spacing X1 from the fusion boundary to the weld centreline during solidification.
Parallel to plate
surface
Just above toe.
of weld
[(qo)1/2/(v)1/4],(WS1/2/mm1/2)1/2
Fig. 3.49. Experimental verification of equation (326). Data from Jordan and Coleman.27
Solution
As shown in Fig. 3.44, the thermal gradient, GL, decreases from about 25C mm"1 at the
fusion boundary to approximately 10C mm"1 close to the weld centreline. This occurs
parallel with an increase in the nominal crystal growth rate from 0.8 to 4mm s"1. Taking the
primary dendrite arm spacing at the fusion boundary equal to X1*, the ratio X1 /X1* in different
positions of the weld becomes:
Central region of columnar zone:
Weld centreline:
In contrast to that predicted above, the smallest dendrite arm spacing is normally observed at
the weld centreline.26 This has to do with the fact that the constant C1 in equation (325), in
practice, decreases with increasing distance from the fusion boundary due to solute segregation, which gradually reduces the coarsening rate of the dendrites.
3.5.3.3 Secondary dendrite arm spacing
It follows from the above analysis that the primary dendrite arm spacing, once it has been
established, will remain constant during subsequent cooling of the weld. This is not true of the
secondary arms (see definition in Fig. 3.46), which undergo a continuous ripening process.
Hence, their size and morphology will change with time as the thicker branches grow larger at
the expense of the smaller ones.2
The driving force for the ripening process is the disparity in chemical potential (interfacial
energy) between branches with different curvature. Since this process is analogous to the
Ostwald ripening of precipitates (see equation (276)), the spacing of the branches, X2, will be
a simple cube root function of the local solidification time fo:1'2
(330)
where M is a mobility term defined as:2
(331)
The parameter Q in equation (331) refers to the GibbsThomson coefficient and is given as:
(332)
where a is the solid/liquid interfacial energy, and AS, is the entropy of fusion.
In practice, the value of M can easily vary by an order of magnitude. Nevertheless, its effect
on the secondary dendrite arm spacing is rather weak, since X2 is proportional to the cube root
of M. Hence, a plot of the secondary dendrite arm spacing vs the local solidification time in a
logarithmic diagram will normally reveal a straightline relationship between X2 and to with a
slope close to 0.33, as shown in Fig. 3.50.
X2^m
Slopes 0.33
to'5
Fig. 3.50. Relation between secondary dendrite arm spacing X2 and local solidification time to in type
AISI 310 stainless steel welds. Data from Kou and Lee.31
Example (3.10)
Based on equation (330), derive a relationship between the secondary dendrite arm spacing
X2, the net arc power qo, and the weld travel speed v during thick plate welding. Calculate then
the secondary dendrite arm spacing in the centre of a thick GTA AlSi weld deposited under
the following conditions:
Relevant physical data for the AlSi system are given below:2
Solution
The local solidification time to is the time for the dendrite array to pass an arbitrary point in the
weld and is therefore a measure of the solidification rate. Referring to Fig. 3.48, the local
solidification time is defined as:
(333)
(334)
(335)
and
Although reliable experimental data are not available for a direct comparison, the calculated value for X2 is considered reasonably correct. In AlCu castings a secondary dendrite
arm spacing of about 4(im corresponds to a cooling rate of the order of 103 C s"1, as shown in
Fig. 3.51. In the present example, the cooling rate at the solid/liquid interface is close to:
The operating parameters given in Example (3.10) are also applicable to single pass butt welding of thin aluminium plates. Based on equation (330), estimate the secondary dendrite arm
spacing in the centre of a 2mm thick AlSi butt weld deposited under such conditions.
Solution
In thin plate welding the local solidification time to is obtained by combining equations (329)
and (333). Noting that RL = v at the weld centreline, we get:
(336)
X 2 ,nm
AI4.5 wt% Cu
(337)
It is evident from the above calculation that the secondary dendrite arm spacing is significantly coarser in thin plate welding compared with thick plate welding. This observation is not
surprising, considering the pertinent difference in the heat transfer mode and thus the cooling rate between these two types of weldments. In the former case, we have:
A comparison with the data in Fig. 3.51 shows that a cooling rate of about 78C s"1 is
compatible with a secondary dendrite arm spacing of 13 jim.
CL
FL
FL
CL
Fig. 3.52.Example of grain refinement in an AlMgSi plasma arc weld due to TiAl3precipitation (FL:
fusion line, CL: centreline). Courtesy of M.I. Onsoien, SINTEF, Trondheim, Norway.
transition occurs when the GJR1 ratio drops below a certain critical value fcn. Hence, we may
write:
(338)
During thick plate welding (3D heat flow) the temperature gradient at the weld centreline
is given by equation (328). Taking RL = v, a combination of equations (328) and (338) gives:
(339)
where c4 is a kinetic constant which is characteristic of the alloy system under consideration.
Similarly, for welding of medium thick plates (mixed heat flow), we get:
(340)
where c5 is a new kinetic constant, and n is an exponent which varies between zero and unity
depending on the mode of heat flow (i.e. thin, medium thick or thick plate welding).
(a)
(b)
Heat flow
Fig. 3.53. Schematic diagrams showing different dendrite growth morphologies in castings; (a) Columnar dendritic growth, (b) Equiaxed dendritic growth. The diagrams are based on the ideas of Kurz and
Fisher.2
Equation (340) predicts that the columnar to equiaxed transition in fusion welds occurs at
critical combinations of the net arc power qo and the welding speed v. Thus, a decrease in qo
must always be compensated by a corresponding increase in v in order to maintain equiaxed
dendritic growth at the weld centreline. This is also in agreement with general experience
(see Fig. 3.54).
Example (3.12)
Based on the experimental data in Fig. 3.54, calculate the critical G1IR1 ratio which provides
equiaxed dendritic growth during thick plate welding of AlMg alloys (5083 series). Relevant
thermal data for the AlMg system are given in Table 1.1 (Chapter 1).
Solution
By multiplying fcr with the welding speed we see that the calculated value corresponds to
a critical temperature gradient of about 928C mm"1. This means that the thermal conditions
existing in welding will favour growth of equiaxed dendrites close to the weld centreline. A
requirement is, of course, that the melt contains a sufficient number of seed crystals to facilitate heterogeneous nucleation of new grains ahead of the advancing solid/liquid interface.
Next Page
Heterogeneous nucleation
Dendrite fragmentation
Grain detachment.
The former mechanism is particularly relevant to welding, since the weld metal often contains a high number of second phase particles which form in the liquid state. As already
mentioned in Section 3.4.4.2, these particles can either be primary products of the weld metal
deoxidation or stem from reactions between specific alloying elements which are deliberately
introduced into the weld pool through the filler wire. The important effect of deoxidation
practice (inclusions) on the columnar to equiaxed transition in ferritic stainless steel GTA
welds is shown in Figs. 3.55 and 3.56.
Under such conditions it is fairly simple to derive an expression for the solute concentration
in the metal as a function of the fraction solidified. Referring to Fig. 3.57, a mass balance
gives:
(341)
Since dz z and Cs = ko C1, equation (341) can be rewritten as:
Previous Page
Heterogeneous nucleation
Dendrite fragmentation
Grain detachment.
The former mechanism is particularly relevant to welding, since the weld metal often contains a high number of second phase particles which form in the liquid state. As already
mentioned in Section 3.4.4.2, these particles can either be primary products of the weld metal
deoxidation or stem from reactions between specific alloying elements which are deliberately
introduced into the weld pool through the filler wire. The important effect of deoxidation
practice (inclusions) on the columnar to equiaxed transition in ferritic stainless steel GTA
welds is shown in Figs. 3.55 and 3.56.
Under such conditions it is fairly simple to derive an expression for the solute concentration
in the metal as a function of the fraction solidified. Referring to Fig. 3.57, a mass balance
gives:
(341)
Since dz z and Cs = ko C1, equation (341) can be rewritten as:
(b)
Aluminium content, wt%
Fig. 3.55. Effect of deoxidation practice (aluminium additions) on the columnar to equiaxed transition in
ferritic stainless steel GTA welds; (a) Average fraction of equiaxed grains observed at the surface of the
welds vs aluminium content, (b) Surface equiaxed grain size vs aluminium content. Data from Villafuerte
etal21
(342)
Integration of this equation gives:
(343)
(a)
(b)
Liquid
Concentration, C
Solid
Distance, z
Fig. 3.57. Solute redistribution during nonequilibrium freezing according to the Scheil equation.
Equation (344) is valid up to CL = Ceut where the remaining melt solidifies in the form of
intercellular or interdendritic eutectics. The eutectic fraction feut is, in turn, given as:
(346)
Figure 3.58 shows how the Scheil equation can be used for an evaluation of the
microsegregation pattern in binary alloy systems by considering a small volume element of
length L* which solidifies perpendicular to the cell/dendrite growth direction.
Since the Scheil equation does not allow for solid state diffusion during solidification, a
slightly refined version of this equation also exists in the literature:1
(347)
(348)
Solid
Liquid
Example (3.13)
Consider GTA welding of an Al2wt%Cu alloy under the following solidification conditions:
Welding speed
Cooling rate
Primary dendrite arm spacing
Estimate on the basis of the Scheil equation the degree of microsegregation occurring during weld metal solidification. Relevant physical data for the AlCu system are given below:
Solution
The local solidification time can be calculated from equation (333). Taking RL = v, we get:
Since the solidification conditions in this case facilitate the formation of a cellulardendritic
type of substructure close to the weld centreline (see Fig. 3.43), the characteristic diffusion
length L* is determined by the primary dendrite arm spacing X1. The product a* ko is then
given as:
Because the numerical value of a* ko is very small, the contribution from diffusion in the
solid state can be neglected. Hence, the extent of microsegregation occurring during solidification can be evaluated from equation (345). Taking C0 = 2wt% and ko = 0.17, we obtain:
These results should be compared with the experimental data of Brooks and Baskes35
replotted in Fig. 3.60. It is evident that the measured copper concentration profile for the Al2wt%Cu GTA weld is similar to that inferred from the Scheil equation, although the observed
dendrite core concentration lies significantly above the predicted one. Consequently, the Scheil
equation gives a reasonable description of the segregation pattern during weld metal solidification, in spite of the simplifying assumptions inherent in the model.
Cu concentration, wt%
z.^m
Cu concentration, wt%
Position, ]im
Fig. 3.60. Electron microprobe analysis of Cu across primary solidification (cell) boundaries in an Al2wt% Cu GTA weld. Data from Brooks and Baskes.35
Fig. 3.6/. Example of solute banding (dark stripes) in a phosphoruscontaining copper weld. After Garland and Davis.36
3.61). Although the origin of the phenomenon is not yet fully understood, it is reasonable to
assume that the pertinent fluctuations in the solidification rate occurs as a result of frequent
variations in the heat flux during welding. Direct experimental evidence for such a correlation
can be obtained from the data of Garland and Davis.36
Moreover, there is a pronounced tendency for alloying and impurity elements to segregate
to the weld centreline where the columnar grains growing from each side impinge. This, in
turn, may produce hot tearing as a result of the formation of lowmelting eutectics between the
dendrite arms. In general, the risk of hot tearing decreases with increasing width to depth ratio
of the weld because of a more favourable crystal growth mode, as illustrated in Fig. 3.62.
3.7.3 Gas porosity
Troublesome impurities in fusion welds are those which precipitate, alone or in combination
with other elements, to form various gaseous reaction products.37 These reaction products may
be simple diatomic gases such as H2 and N2 or more complex gaseous compounds like CO, or
H2O. An illustration of gas porosity due to hydrogen evolution is contained in Fig. 3.63.
Following the discussion in Chapter 2, absorption of gases occurs readily in the hot part of
the weld pool during welding because of interactions with the surrounding arc atmosphere.
During the subsequent stage of cooling a supersaturation rapidly increases due to the associated decrease in the element solubility with decreasing temperatures (see Figs. 2.7 and 2.8).
The system will respond to this supersaturation by rejection of dissolved elements from the
liquid, either through a gas/metal reaction (desorption) or by precipitation of new phases. The
latter incident may result in porosity if the gas bubbles become trapped in the weld metal
solidification front.
3.7.3.1 Nucleation of gas bubbles
In liquid metals a bubble will be stable if the gas pressure inside the bubble is sufficiently high
to balance the external forces. These external forces are the liquid/vapour interfacial energy
Width
Depth
Depth
Width
Fig. 3.62. Effect of weld width to depth ratio on the tendency to centreline cracking; (a) Correct width to
depth ratio, (b) Incorrect width to depth ratio.
Fig. 3.63. Example of gas porosity in a GTA 7106 aluminium weld. After D'annessa.39
Transport of dissolved
gaseous species
Liquid
Gas bubble
Solid
Fig. 3.64. Growth of a gas bubble due to diffusion in the liquid phase (schematic).
a*, the metallostatic pressure head/?m, and the ambient pressure pa. Thus, for a stable bubble,
we may write:1
(351)
where pg is the total gas pressure inside the bubble, and rg is the radius of the gas bubble.
For shallow welds, the contribution from the metallostatic pressure head pm can be ignored. Hence, equation (351) reduces to:
(352)
Since a* is typically of the order of 1 J m 2 (9.87 atm im) for most gasmetal systems, we
may write:
(353)
It is evident from equation (353) that the interfacial energy term is negligible at large
values of r . However, if the radius of curvature becomes sufficiently small, extremely large
pressures are required to maintain a stable bubble. Thus, there is a bubble nucleation problem, which is formally similar to that of nucleation of a solid from a liquid (discussed in
Section 3.3.1). In fact, it can be shown on the basis of classic nucleation theory that the driving
force normally associated with rejection of dissolved gases in liquid metals is by far too small
to allow for homogeneous nucleation of gas bubbles in the weld pool during cooling. This, in
turn, implies that solid particles (e.g. inclusions) entrained in the liquid metal will be the most
probable sites for gas bubble formation in fusion welds.
3.7.3.2 Growth and detachment of gas bubbles
As shown in Fig. 3.64, growth of gas bubbles in liquids is a diffusioncontrolled process where
the time dependence of the mean bubble radius is given by:38
(354)
Here Q.* is the growth constant, defined as:
(355)
where CL is the molar concentration of solute in the supersaturated liquid, C6 is the equilibrium
molar concentration of the solute at the gas/liquid interface, and p^ is the gas density (in the
same units as CL and C6).
Equations (354) and (355) may be used to estimate the growth rate of a bubble while it is
still attached to the solid/liquid interface. The bubble becomes detached when the buoyancy
force, which is pushing it upwards, exceeds the surface tension force, which tends to keep it
attached to the solid surface. The bubble radius at which detachment occurs is given by the socalled Fritz equation:38
(356)
where gc is the gravity constant, and P is the wetting angle (in degrees).
Based on equations (354) and (356) it is possible to evaluate the conditions for growth
and detachment of gas bubbles during weld metal solidification. This is a subject of considerable importance in welding, since the pore formation will inevitably affect the mechanical
integrity of the weldment.
Example (3.14)
Consider GTA butt welding of a 3mm thin AlMg sheet under the following conditions:
Suppose that gas bubbles form at the solid/liquid interface during solidification due to rejection of dissolved hydrogen from a supersaturated liquid. Based on equations (354) and
(356) estimate the maximum theoretical radius of the gas bubbles and the critical radius at
which the bubbles detach themselves from the solid/liquid interface during welding. Relevant
physical data for the AlMg system are given below:
Solution
Since particles located at the solid/liquid interface are the most probable sites for hydrogen
gas evolution, the local solidification time t0 provides a conservative estimate of the growth
time t in equation (354). From equation (336), we have:
In order to calculate the growth constant from equation (355), it is necessary to convert the
concentration driving force to molar units:
The molar density of the gas p is obtained from the ideal gas law:
The maximum theoretical radius of the gas bubbles can now be evaluated from equation
(354) by inserting the appropriate values for Q*, DH, and to:
Similarly, the critical radius at which the hydrogen bubbles become detached may be estimated from equation (356). Since PM > > PH 2 , we obtain:
By inserting this value into equation (354) it is also possible to estimate the average bubble
detachment frequency under the prevailing circumstances:
Since r (crit.) r (max), the maximum pore radius will probably be closer to 0.7mm than
30mm in a real welding situation. This is also in agreement with practical experience (see Fig.
3.65).
GTAW (Aluminium)
Diameter of pores, mm
Fig. 3.65. Measured distribution of pore diameters in some GTA aluminium welds deposited with different hydrogencontaining shielding gases. Data from Tomii et al.40
Based on Stokes law (equation (357)) calculate the rising velocity of a 0.2mm large hydrogen
bubble ascending in liquid aluminium. Relevant physical data for liquid aluminium are given
below:
Solution
It is evident from the above calculations that the flotation rate of such gas bubbles is quite
high and of the same order of magnitude as the weld pool fluid flow velocity (discussed in
Section 2.11.2). Hence, the buoyancy force would be expected to play a significant role in the
separation process of detached gas bubbles in the weld pool. On this basis it is not surprising to
find that a change in the welding position (e.g. from flat to overhead) results in a dramatic
increase in the volume of porosity during GTAW of aluminium alloys (see Fig. 3.66).
Although a great deal has been reported on the causes and effects of porosity in weld metals
(see Ref. 37 for an excellent discussion), little is known about the mechanism of pore formation relative to solidification mechanics, nucleation, growth and transport of gas bubbles in the
weld pool. Consequently, a more fundamental approach to the porosity problem in fusion
welding (along the lines indicated above) is necessary in order to obtain a verified, quantitative
understanding of the phenomenon.
GTAW
Overhead position
Flat position
where C and Cmax are as indicated in Fig. 3.67, lavg is the average distance between adjacent maxima and minima, and Ds is the diffusivity of the solute in the solid.
Equation (358) states that the concentration remains constant and equal to Cavg at positions
JC = 0, JC = lavg, x = 2/ etc., while the peak of the sine wave is attained at distances x = lavg /2,
x = 5lavg /2 etc. during the decay. If only the peak concentration is considered, the sine term
becomes equal to unity and equation (358) reduces to:
(359)
from which the homogenisation time during isothermal heat treatment thom can be obtained:
(360)
Since the diffusion length I , in practice, is equal to the half dendrite arm spacing (X1 /2 or
X2/2), equation (360) predicts that the homogenisation time is proportional to the square OfX1
or X2. The latter parameters are, in turn, determined by the thermal conditions existing within
the mushy zone during weld metal solidification, and are therefore sensitive to variations in
welding variables such as the net arc power qo and the travel speed v.
3.7.4.2 Application to continuous cooling
During cooling of the weld in the solid state, some equalisation of microsegregations will
occur through diffusion. The extent of this diffusion can be reported in terms of an equivalent
isothermal homogenisation time at a chosen reference temperature Tr. If Qs denotes the activation energy for diffusion of the solute in the solid, the equivalent isothermal homogenisation
time tr at Tr is given by:
(361)
Concentration
x
Fig. 3.67. Relaxation of sinusoidal distribution of solute during isothermal annealing.
The integral on the righthand side of equation (361) represents the kinetic strength of the
weld thermal cycle with respect to homogenisation (solid state diffusion), and can be determined by means of numerical methods when the cooling programme is known. The extent of
solute diffusion may then be evaluated from equation (360) by inserting representative
values for t and / :
T
UY'g.
(362)
Example (3.16)
[n lowalloy steel, both carbon, phosphorus, and manganese are known to segregate to the
interdendritic spaces during solidification. Consider SA welding on a thick plate of steel
under the following conditions:
Based on equations (361) and (362) calculate the extent of homogenisation occurring
within the solid weld metal during cooling in the austenite regime (i.e. from 1520 to 7000C).
Relevant data for the diffusivity of carbon, phosphorus, and manganese in austenite are given
below:
Solution
During thick plate welding, the cooling programme can be calculated from equation (145).
For points located on the weld centreline behind the heat source y = z = 0, and R* = x = vt.
Hence, equation (145) reduces to:
From this it is seen that the actual transformation temperatures (i.e. 1520 and 7000C) are
reached after 20.9 and 46.2s, respectively. If 13500C is used as a reference temperature, the
kinetic strength of the cooling cycle with respect to solute diffusion can be expressed as:
Numerical integration of this equation over the weld cooling cycle gives:
It is evident from the above calculations that interdendritic segregations of carbon are
readily removed during cooling of the weld from the solidification temperature due to a high
diffusivity of C in austenite. In contrast, segregations of phosphorus and manganese are much
more persistent, since the diffusion cannot keep pace with the falling temperature. In the latter case the initial concentration gradients will largely be maintained down to temperatures
where the austenite to ferrite transformation occurs (see ion micrograph of phosphorus segregations in Fig. 3.39). This, in turn, can promote the formation of different types of transformation products (ranging from ferrite to martensite depending on the nature of the segregants) along the primary solidification boundaries due to local variations in the steel
hardenability. An example of martensite banding in an AISI 4340 SA steel weld is shown in
Fig. 3.68.
It should be noted that the conditions for equalisation of interdendritic segregations (e.g.
phosphorus) are not dramatically altered by a change in the heat input as long as the tr I l\vg
ratio in equation (362) remains fairly constant. For instance, if the net arc power qQ in the
above example is reduced by a factor of four (conforming to welding with covered electrodes),
the corresponding change in / and J1350 (P) will be:
and
from which
This value is reasonably close to that obtained during highheat input SA welding
(0.036 s jirrr2), as shown in the above example.
Fig. 3.68. Banding of martensite (M) along primary solidification boundaries in an AISI 4340 SA steel
weld. After Burck.42
<3"63)
Because of the nature of this reaction, there is a strong tendency for the secondary phase 7
to grow along the 8^/liquid interface and, thus, to isolate the primary phase from contact with
the liquid.43 Depending on the growth mode, the decomposition of the primary 8^phase is
said to occur either through a peritectic reaction or by a peritectic transformation,44 as
shown schematically in Fig. 3.70 and 3.71, respectively.
3.8.1 Primary precipitation of the y phase
Primary precipitation of the 7^phase from the liquid is also possible if the temperature is
gradually decreasing during solidification.43 As shown by Kerr et al.44^6 this type of precipitation is favoured by the presence of solid particles within the liquid metal which can act as
seed crystals for the 7^phase in a selective manner. Particularly at high cooling rates, the formation of the primary 8^phase can be completely suppressed, thereby allowing the secondary
7 phase to nucleate and grow directly from the supercooled liquid.
3.8.2 Transformation behaviour of lowalloy steel weld metals
In general, lowalloy steels undergo multiple phase transformations during solidification and
subsequent cooling. Depending on the cooling rate, carbon or the substitutional alloy content,
the primary solidification product will either be delta ferrite 8 Fe , austenite 7 Fe or a mixture of
both. 47 ^ 9
3.8.2.1 Primary precipitation of delta ferrite
When a lowalloy steel is slowly cooled below the crystallisation temperature, delta ferrite
will be the first phase to form.50 In fusion welds, the delta ferrite grains will reveal an anisotropic
columnar morphology, with their major axes aligned in the direction of the steepest temperature gradient in the weld pool. On further cooling below the peritectic temperature, austenite
nucleates epitaxially at the primary delta ferrite grain boundaries, as shown schematically in
Fig. 3.72, since these sites provide the lowest energy barrier against heterogeneous nucleation
(requires a KurdjumowSachs or a NishiyamaWasserman type of orientation relationship
between 8Fe and 7Fe).51 Subsequent growth of the austenite into the delta ferrite may then
proceed by a peritectic transformation at a rate which is controlled by diffusion of carbon in
the austenite. This reaction pattern is analogous to that documented for growth of 7 F e Widmanstatten sideplates in duplex stainless steel.5253
Since the austenite is bound by an orientation relationship with the delta ferrite,51 the
austenite cannot grow across the primary delta ferrite solidification boundaries. Hence, after
the peritectic transformation the columnar 7Fe grains will adopt the original 8Fe morphology,
as shown in Fig. 3.72. Direct experimental evidence for such a correspondence can be obtained from the optical micrographs in Fig. 3.38 and the texture data presented in Section 6.3.5
(Chapter 6).
Temperature
Liquid (L)
Growth direction
Liquid
Diffusion of
solute atoms
Fig. 3.70.Peritectic reaction by which the secondary 7pphase grows along the surface of the primary
8pphase (schematic).
Liquid
Fig. 3.7L Peritectic transformation involving longrange diffusion of solute atoms through the secondary
7^phase (schematic).
Delta ferrite
Liquid
Solute segregations
Austenite
Austenite grain
boundary
Fig. 3.72. Primary delta ferrite solidification with subsequent growth of austenite along the boundaries of
the primary 8Fephase (schematic).
Austenite
Columnar
grain
Liquid
Fig. 3.74. Optical micrograph showing austenite grain boundaries crossing primary delta ferrite solidification boundaries in a lowalloy SA steel weld (indicated by arrows). After Kluken et al.22
at prior solidification (cell) and austenite grain boundaries*, respectively in a SA steel weld
metal deposited under the shield of a basic flux. It is obvious from Fig. 3.75 that there is no
matching between the two types of boundaries in this particular case.
It follows from the analysis of Kluken et al?2 that the observed shift in the mechanism of
the peritectic transformation can probably be attributed to heterogeneous nucleation of
austenite at inclusions (e.g. Al2O3), which is energetically more favourable than nucleation at
8Fe/8Fe grain boundaries. Under such conditions, the austenite is not bound by an orientation
relationship with the delta ferrite and is thus free to grow across the original delta ferrite
columnar grain boundaries, as shown schematically in Fig. 3.76. The austenite grains will
therefore adopt a morphology which is different from the columnar one.245556
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
M.C. Flemings: Solidification Processing, 1974, New York, McGrawHill Book Company.
W. Kurz and DJ. Fisher: Fundamentals of Solidification, 3rd Edn, 1989, Aedermannsdorf
(Switzerland), Trans. Tech. Publications.
GJ. Davis and J.G. Garland: Int. MetalL Rev., 1975, 20, 83106.
S.A. David and J.M. Vitek: Int. Mater. Rev., 1989, 34, 213245.
D. A. Porter and K.E. Easterling: Phase Transformations in Metals and Alloys, 1981, Wokingham
(England), Van Nostrand Reinhold Co. Ltd.
K.E. Easterling: Introduction to the Physical Metallurgy of Welding, 1983, London, Butterworths
& Co (Publisher) Ltd.
0. Grong, TA. Siewert and G.R. Edwards: Weld. J., 1986, 65, 279s288s.
G.M. Oreper, TW. Eagar and J. Szekely: Weld. J., 1983, 62, 307s312s.
Y.H. Wang and S. Kou: Proc. Int. Conf. on Advances in Welding Science, Gatlinburg, TN,
May, 1986, 6569, Publ. ASM International.
*Because of a similar atomic weight of BO2~ and AlO~, both types of ions will be sampled in the SIMS analysis.
Consequently, the white spots observed within the interior of the grains are probably traces of aluminium oxide
inclusions.
(a)
(b)
Fig. 3.75. Ion (SIMS) micrographs showing austenite grain boundaries crossing primary delta ferrite
solidification boundaries in a lowalloy SA steel weld; (a) Phosphorus segregations at primary solidification (cell) boundaries, (b) Boron segregations at prior austenite grain boundaries (same area as in Fig.
3.75(a)). After Kluken et alP
Delta ferrite.
Solute segregations
Liquid
Austenite
Austenite grain
boundary
Fig. 3.76. Primary delta ferrite solidification with subsequent nucleation of austenite at inclusions (schematic).
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
K.C. Mills and BJ. Keene: Int. Mater. Rev., 1990, 35, 185216.
T. Zacharia, S.A. David, J.M. Vitek and H.G. Kraus: Metall. Trans., 1991, 22B, 243257.
CR. Heiple and J.R. Roper: Weld. J., 1982, 61, 97s102s.
CR. Heiple and P. Burgardt: Weld. J., 1985, 64, 159s162s.
T. Habrekke, H.O. Knagenhjem and J.O. Berge: Proc. 9th. Int. Conf. on Offshore Mech. and
Arctic Engin., Houston, TX, Feb., 1990, 511515, Publ. ASME.
H. Biloni: Physical Metallurgy, 3rd Edn (Eds R.W. Chan and P. Haasen), 1983, Amsterdam,
NorthHolland Physics Publ., 478579.
T. Senda, F. Matsuda and M. Kato: Techn. Reports of the Osaka University (Japan), 1970,20,
527558.
M. Rappaz, S.A. David, J.M. Vitek and L.A. Boatner: Metall. Trans. A, 1989,2OA, 11251138.
L.F. Mondolfo: Mat. ScL and TechnoL, 1989, 5, 118122.
B.L. Bramfitt: Metall. Trans., 1970, 1, 19871995.
G.N. Heintze and R. McPherson: Weld. /., 1986, 65, 71s81s.
J.C. Villafuerte, E. Pardo and H.W. Kerr: Metall. Trans. A, 1990, 21A, 20092019.
A.O. Kluken, 0. Grong and G. R0rvik: Metall. Trans. A, 1990, 21 A, 20472058.
H. Yunjia, R.H. Frost, D.L. Olson and G.R. Edwards: Weld. J., 1989, 68, 280s289s.
CE. Cross, 0. Grong, S. Liu and J.F. Capes: Applied Metallography, (Ed. G.F. Vander Voort),
1986, New York, Van Nostrand Reinhold Inc., 197210.
A.O. Kluken and 0 . Grong: Proc. Int. Conf. on Recent Trends in Welding ScL and TechnoL
(TWR 89), Gatlinburg, TN, May, 1989, 781786, Publ. ASM International.
S. Kou: Welding Metallurgy, 1987, New York, John Wiley & Sons.
M.F. Jordan and M.C. Coleman: Brit. Weld. J., 1968,15, 553558.
CE. Cross and G.R. Edwards: Treatise on Mat. ScL TechnoL, 1989, 31, 171187.
W. Kurz, B. Giovanola and R. Trivedi: Acta Metall., 1986, 34, 823830.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
Appendix 3.1
Nomenclature
thermal diffusivity
(mm2 s1)
(Lim)
bcc
area of embryoliquid interface (m2)
area of embryosubstrate interface (m2)
eutectic fraction
fraction solidified
shape factor
facecentred cubic structure
equilibrium concentration of
solute at gas/liquid interphase
(wt% or mol cirr3)
GMAW
GTAW
[uvw\
EBW
threedimensional radius
vector (mm)
exponent
SAW
SIMS
secondary
spectrometry
SMAW
ion
mass
temperature (C or K)
ambient temperature (C or
K)
actual temperature in the liquid (C or K)
melting point (C or K)
peritectic temperature (C or
K)
reference temperature (C or
K)
dendrite root temperature (C
or K)
dendrite tip temperature (C
or K)
efficiency factor
undercooling (C or K)
temperature difference between tip and root of dendrite
stem (C or K)
a lowindex direction in
(UcQn
GibbsThomson coefficient
(mm C or mK)
a lowindex direction in
W),
voltage (V)
dimensionless temperature
rising velocity of gas bubbles
(mm s"1)
dimensionless temperature at
the melting point
yaxis/transverse direction
(mm)
[vw]n (degrees)
austenite
delta ferrite
embryo/substrate interfacial
energy (J nr 2 )
substrate/liquid interfacial
energy (J rrr2)
thermal conductivity (W
mm10C"1)
4
Precipitate Stability in Welds
4.1 Introduction
Precipitate stability is an important aspect of welding metallurgy. Normally, modern structural
steels and aluminium alloys derive their balanced package of high strength, ductility and toughness via optimised thermomechanical processing to produce a finegrained, precipitation
strengthened matrix. This delicate balance of microalloy precipitation and microstructure,
however, is significantly disturbed by the heat of welding processes, which, in turn, affects the
mechanical integrity of the weldment.
When a commercial alloy is subjected to welding or heat treatment several competitive
processes are operative which may contribute to a change in the volume fraction and size
distribution of the base metal precipitates. The two most important are:1
(i)
(ii)
Referring to Fig. 4.1, particle coarsening occurs typically at temperatures well below the
equilibrium solvus Te of the precipitates, while particle dissolution is the dominating mechanism at higher temperatures. On the other hand, there exists no clear line of demarcation
between these two processes, which means that particle coarsening can take place simultaneously with reversion in certain regions of the weld where the peak temperature of the thermal
cycle falls within the 'gray zone' in Fig. 4.1. Nevertheless, it is important to regard them as
separate processes, since the reaction kinetics are so different (coarsening is driven by the
surface energy alone, whereas dissolution, which involves a change in the total volume fraction, is driven by the free energy change of transformation).
'Grey zone1
Particle coarsening 
Temperature
Particle dissolution
%B
Fig. 4.1. Schematic diagram showing the characteristic temperature ranges where specific physical
reactions occur during reheating of grain sizecontrolled and dispersionhardened materials.
(41)
At equilibrium, we have:
(42)
where AH and AS are the standard enthalpy and entropy of reaction, respectively. The other
symbols have their usual meaning (see Appendix 4.1).
When pure An Bm is used as a standard state, the activity of the precipitate {aAn Bm) is equal
to unity. In addition, for dilute solutions it is a fair approximation to set aA~[%A] and aB~[%B],
where the matrix concentrations of elements A and B are either in wt% or at%*. Hence, the
solubility product can be written as:
(43)
*For the solute, the standard state is usually a hypothetical 1 % solution. This implies that the activity coefficient is
equal to unity as long as Henry's law is obeyed.
Table 4.1 gives a summary of equilibrium solubility products for a wide range of precipitates in lowalloy steels and aluminium alloys.
In addition to the compounds listed in Table 4.1, different types of mixed precipitates may
form within systems which contain more than two alloying elements.3"6 However, since the
presence of such multiphase particles largely increases the complexity of the analysis, only
pure binary intermetallics will be considered below.
4.2.2 Equilibrium dissolution temperature
Based on equation (43) it is possible to calculate the equilibrium dissolution temperature Td of
the precipitates. By rearranging this equation, we get:
(44)
where [%A]o and [%B]o refer to the analytical content of elements A and B in the base metal,
respectively.
Equation (44) shows that the equilibrium dissolution temperature increases with increasing concentrations of solute in the matrix. This is in agreement with the Le Chatelier's principle.
Table 4.1 Equilibrium solubility products for different types of precipitates in lowalloy steels and
aluminium alloys. Data compiled from miscellaneous sources.
Material/ phase
Type of
Precipitate
C* = AS/R'
D* = Mi0IR'
TiN
0.32
8000
TiC
5.33
10475
NbN
4.04
10230
Lowalloy steel
NbC
2.26
6770
(austenite)t
VN
3.02
7840
VC
6.72
9500
AIN
1.79
7184
Mo2C
5.0
7375
AlMgSi^
Mg2Si
5.85
5010
AlCuMg$
CuMg
6.64
4005
MgZn
5.33
2985
Zn2Mg
7.72
4255
AlZnMgij:
Example (4.1)
Calculate on the basis of the reported solubility products in Table 4.1 the equilibrium dissolution temperature of each of the following three nitride precipitates, i.e. NbN, AlN, and TiN.
Solution
The equilibrium dissolution temperature of the precipitates can be computed from equation (44) by inserting the correct values for C* and D* from Table 4.1:
It is evident from these calculations that precipitates of the NbN and the AlN type will
dissolve readily at temperatures above 1050 to 11000C, while TiN is thermodynamically stable up to about 14500C. In practice, however, a certain degree of superheating is always
required to overcome the inherent kinetic barrier against dissolution, particularly if the heating
rate is high. Consequently, in a real welding situation the actual dissolution temperature of the
precipitates may be considerably higher than that inferred from simple thermodynamic calculations based on the solubility product (to be discussed later).
4.2.3 Stable and metastable solvus boundaries
Due to the lack of adequate phase diagrams for the complex alloy systems involved, thermodynamic calculations based on the solubility product represent in many cases the only practical means of estimating the solid solubility of alloying elements in commercial lowalloy steels
and aluminium alloys.
4.2.3.1 Equilibrium precipitates
In the case of large, incoherent precipitates (where the GibbsThomson effect can be neglected),
the concentration of element A in equilibrium with pure An Bm at different temperatures can be
inferred directly from equation (43). If we replace / ^ by R (i.e. switch from common to
natural logarithms), this equation yields:
(45)
Equation (45) describes the solvus surface within the solventrich corner of the phase diagram. However, when a pure binary compound dissolves the concentration of elements A and
B in solid solution is fixed by the stoichiometry of the reaction. The following relationship
exists between [%B] and [%A]:
(46)
or
[%B]
[%A]0
[%B]o
Excess B
[%A]
Fig. 4.2. Concentration displacements during dissolution of binary intermetallics (equilibrium conditions).
Increased additions
_of element B
Temperature
Concentration of element A
Fig. 4.3. Factors affecting the solid solubility of a binary intermetallic compound in a multicomponent
alloy system (schematic).
crease with increasing temperature when AH is positive. This type of behaviour is characteristic of intermetallics in metals and alloys, since the dissolution process in such systems is
endothermic.7 As a result, increased additions of a second alloying element B will also reduce
the solubility of the first alloying element A by shifting the 'solvus boundary' towards higher
temperatures when an intermetallic compound between A and B is formed.
With the aid of Fig. 4.3 it is easy to verify that the equilibrium volume fraction of the
precipitates/^ at a fixed temperature is given by:
(47)
where fmax is the maximum possible volume fraction precipitated at absolute zero.
Equation (47) provides a basis for estimating the equilibrium volume fraction of binary
intermetallics in complex alloy systems at different temperatures in cases where the concentration of element B is sufficiently high to tieup all A in the form of precipitates. Similarly, if A
is present in an overstoichiometric amount with respect to B, we may write:
(48)
Example (4.2)
In AlMgSi alloys the equilibrium Mg2Si phase may form during prolonged high temperature annealing. Consider a pure ternary alloy which contains 0.75 wt% (0.83 at.%) Mg and 1.0
wt% (0.96 at.%) Si. Estimate on the basis of the solubility product the equilibrium volume
fraction of Mg2Si at 4000C. Make also a sketch of the Mg2Si solvus in a vertical section
through the ternary AlMgSi phase diagram. Relevant physical data for the AlMgSi system are given below:
Solution
The maximum possible volume fraction of Mg2Si precipitated at absolute zero (fmax) can be
estimated from a simple mass balance by considering the stoichiometry of the reaction:
Moreover, the solubility product [at.% Mg]2 [at.% Si] at 4000C (673K) can be obtained
from equation (43) by utilising data from Table 4.1:
from which
If we also take into account the stoichiometry of the reaction, the solubility product can be
expressed solely in terms of the Mgconcentration. Substituting
into the above equation gives [at.% Mg] ~ 0.20. The equilibrium volume fraction OfMg2Si at
4000C is thus:
Similarly, the equilibrium Mg2Si solvus can be calculated from the solubility product by
substituting
into equation (43). By inserting data from Table 4.1 and rearranging this equation, we get:
It is seen from the graphical representation of the above equation in Fig. 4.4 that the Mg2Si
compound is thermodynamically stable up to about 3000C. At higher temperatures the phase
will start to dissolve until the process is completed at 5600C.
It is obvious from these calculations that the microstructure of overaged AlMgSi alloys
should be very persistent to the heat of welding processes. In practice, only a narrow solutionised
zone forms adjacent to the fusion boundary. However, within this zone significant strength
recovery may occur after welding due to reprecipitation of hardening phases from the
supersaturated solid solution. Consequently, in such weldments the ultimate HAZ strength
level is usually higher than that of the base metal, as illustrated in Fig. 4.5.
4.2.3.2 Metastable precipitates
In practice, the solid solubility is also affected by the size of the particles. If, for instance, a
Temperature, 0C
at% Mg
at%Mg Si
Fig. 4.4. Solubility of Mg2Si in aluminium (Example (4.2)).
Strength level
HAZ
Fig. 4.5. Response of overaged AlMgSi alloys to welding and subsequent heat treatment (schematic).
spherical precipitate is acted on by an external pressure of say 1 atm, the same precipitate is
also subjected to an extra pressure AP due to the curvature of the particle/matrix interface, just
as a soap bubble exerts an extra pressure on its content (see Fig. 4.6(a)). The pressure AP is
given as:8
(49)
where 7 is the particle/matrix interfacial energy, and r is the radius of the precipitate.
Because of this extra pressure, the Gibbs energy of a small precipitate will be higher than
that of a large one, which, in turn, increases its solubility (see Fig. 4.6(b)). The important
influence of particle curvature on the solid solubility has been extensively investigated and
reported in the literature.18 Usually, the phenomenon is referred to as the capillary or the
GibbsThompson effect.
In the following we shall assume that the thermodynamic and crystallographic properties
of the metastable precipitates are similar to those of the equilibrium phase and that the reduced thermal stability is only associated with capillary effects. For single phase precipitates
in binary alloy systems, it is fairly simple to show that the concentration of solute across
a curved interface, [%A]r, is interrelated to the equilibrium concentration of solute across a
planar interface, [%A], through the following equation:8
(410)
where Vn is the molar volume of the precipitate (in m3 mol"1), and Q is the contribution of the
interface curvature to the reaction enthalpy (equal to IyVJr).
(a)
Atmospheric
pressure
Matrix
(b)
Gibbs energy
Small precipitate
Matrix
Large precipitate
[%A]
[%A] r
Concentration
Fig. 4.6. Effect of interfacial energy on the solubility of small particles; (a) Schematic representation of
spherical particles embedded in a metal matrix, (b) Integral molar Gibbs energy of matrix and precipitates at a constant temperature.
Assuming that this relationship also holds in the case of binary intermetallics, a combination of equations (45) and (410) gives:
(411)
where
(412)
or
Alternatively, we can express T as a function of the product [%A]rn [%B]rm. This gives the
following expression for the solvus temperature of metastable precipitates T'eq:
(413)
It is evident from the graphical representation of equation (413) in Fig. 4.7 that the solid
solubility at a given temperature is significantly increased at small particle radii. Taking as an
example 7 = 0.5 J n r 2 , Vm = 10~5 m3 moH, R = 8.314 J Kr1 moH, T = 500 K, we obtain from
equation (410):
or
Temperature
Concentration
Fig. 4.7. Graphical representation of equation (413) (schematic).
Example (4.3)
Solution
The metastable [3"(Mg2Si) solvus can now be calculated from the solubility product by
substituting
into equation (413). By inserting data from Table 4.1 and rearranging this equation, we get:
It is evident from the graphical representation of the above equation in Fig. 4.8 that the
particle curvature has a dramatic effect on the solid solubility. A comparison with Fig. 4.4
shows that the dissolution temperature drops from about 5600C in the case of the equilibrium
Mg2Si phase to approximately 225C for the metastable 3"(Mg2Si)phase. On this basis it is
not surprising to find that artificially aged (T6 heat treated) AlMgSi alloys suffer from severe softening in the HAZ after welding, as shown schematically in Fig. 4.9. Moreover, it is
Nominal alloy
composition A
Temperature, 0C
Metastable solvus
boundary
at% Mg
at% Mg 2 Si
Strength level
HAZ
Distance from fusion line
Fig. 4.9. Response of artificially aged AlMgSi alloys to welding and subsequent heat treatment (schematic).
evident that the characteristic low dissolution temperature of the precipitates also gives rise to
the formation of a heat affected zone which is significantly wider than that observed during
welding of overaged AlMgSi alloys.9 This shows that the response of agehardenable aluminium alloys to welding and thermal processing depends strongly on the initial base metal
temper condition.
With the aid of equation (411) it is also possible to calculate an average (apparent) metastable
solvus boundary enthalpy for hardening 3"(Mg2Si)precipitates in AlMgSi alloys. A closer
evaluation of the exponent gives:
This value is in close agreement with the reported solvus boundary enthalpy for (3"(Mg2Si)precipitates in 6082T6 aluminium alloys. 910
where c{ is a kinetic constant, and Qs is the activation energy for the coarsening process (for
binary intermetallics Qs may be taken equal to the activation energy for diffusion of the less
mobile constituent atom of the precipitates in the matrix).
4.3.2.1 Kinetic strength of thermal cycle
It follows that the extent of particle coarsening occurring during a weld thermal cycle can be
calculated by integration of equation (415) between the limits t = t{ and t = t2:
(416)
The integral on the righthand side of equation (416) represents the kinetic strength of the
thermal cycle with respect to particle coarsening, and can be determined by means of numerical methods when the weld thermal (Tt) programme is known. The resulting radius of the
precipitates may then be evaluated from equation (416) by inserting representative values
for the constants ro and C1 (e.g. obtained from quantitative particle measurements).
4.3.2.2 Model limitations
A salient assumption in the classic LifshitzWagner theory is that the particles coarsen at
almost constant volume fraction, i.e. no solute is lost to the surrounding matrix during the
coarsening process. Consequently, equation (416) should only be applied in cases where the
peak temperature of the thermal cycle is well below the equilibrium solvus of the precipitates.
Example (4.4)
Consider stringer bead deposition (GMAW) on a thick plate of a Timicroalloyed steel under
the following conditions:
Assume that the base metal contains a fine dispersion of TiN precipitates in the asreceived
condition. Calculate on the basis of equation (416) and the Rosenthal thick plate solution
(equation (145)) the extent of particle coarsening occurring within the fully transformed heat
affected zone during welding. Relevant physical data for titaniummicroalloyed steels are
given below:
In the present example the problem is to calculate the size of the TiN precipitates in different
positions from the fusion boundary. This requires detailed information about the weld thermal
programme, as shown in Fig. 4.10(a). By substituting the appropriate values for qo, X, a and v
into the Rosenthal thick plate solution, the governing heat flow equation becomes:
where /?* refers to the threedimensional radius vector in the moving coordinate system (designated R in equation (145)), while x is the welding direction (equal to vt at pseudosteady
state).
Since titanium nitride is thermodynamically stable up to the melting point of the steel,
equation (416) can be used to calculate the extent of particle coarsening occurring within the
transformed parts of the HAZ. In the present example, we may write:
(a)
Weld metal
Temperature
HAZ
Y~ regime
Time
(i/r)exp(Qs/RT)
(b)
Time
Fig. 4.10. Kinetic strength of weld thermal cycle with respect to particle coarsening (Example (4.4)); (a)
HAZ temperaturetime programme (schematic), (b) Numerical integration procedure (schematic).
Weld metal
Particle radius, nm
Peak temperature, 0C
Fig. 4.11. Coarsening of TiN during steel welding (Example (4.4)).
Frequency, %
Particle radius, nm
Fig. 4.12. Measured size distribution of TiN before (broken lines) and after (full lines) weld thermal
simulation. Operational conditions as in Example (4.4). Data from Ion et al.14
ing precipitates. Nevertheless, it will be shown below that at least some of them are sufficiently accurate to capture the essential physics of the problem and to give valuable quantitative information on the extent of particle dissolution occurring during the weld thermal cycle.
Particle radius, nm
Steel A
Steel B
Steel C
Annealing time, s
Fig. 4.13. Effects of annealing time and steel chemical composition on the mean particle size of TiN.
Data from Matsuda and Okumura.15
Concentration
Distance
Fig. 4J4. Schematic representation of the concentration profile around a dissolving spherical particle in
an infinite matrix.
ently by Aaron et al.16 and is valid after a certain period of time, provided that there is no
impingement of diffusion fields from neighbouring precipitates. As shown in Fig. 4.15, this
simplified solution gives a reasonable description of the dissolution kinetics of small spherical
precipitates in steel during reheating above the AC1 temperature.
Following the treatment of Agren,18 the time required for complete dissolution of a spherical precipitate td can be obtained from equation (418) by setting r = 0:
(419)
Moreover, the volume fraction of the precipitates/as a function of time is given by:
(420)
where fo is the initial particle volume fraction.
The former equation shows that the dissolution time td depends strongly on the initial particle size rQ.
Example (4.5)
The following example illustrates the direct application of equations (418) and (419). Consider a niobiummicroalloyed steel which contains a fine dispersion of NbC precipitates. Provided that impingement of diffusion fields from neighbouring particles can be neglected, calculate the total time required for complete dissolution of a 100 nm large NbC precipitate at
Simplified analytical
solution (Whelan)
Time, s
Fig. 4.15. Dissolution kinetics of spherical cementite particles in austenite at 8500C. Data from Agren.18
135OC. Data for the steel chemical composition and the diffusivity of Nb in austenite at
13500C are given below:
Steel chemical composition:
In the present example it is reasonable to assume that the dissolution rate of the precipitate is
controlled by diffusion of Nb in austenite. For a single NbC precipitate embedded in a Nbdepleted matrix, the dimensionless supersaturation becomes:
from the solubility product by utilising data from Table 4.1. If we assume that the carbon
concentration at the interface is constant and equal to the nominal value of 0.12 wt% (i.e. the
stoichiometry of the dissolution reaction is neglected), equation (45) reduces to:
This gives:
The dissolution time td can now be calculated from equation (419) by inserting the appropriate values for ro, aNt)J and Dm\
A comparison with Fig. 4.16 shows that the predicted value is off by a factor of about 4
compared with that obtained from more sophisticated numerical calculations. This degree of
accuracy is acceptable and justifies the use of equation (418) for prediction of the dissolution
rate of spherical precipitates under different thermal conditions provided that the model is
calibrated against experimental data points.
4.4.1.2 Application to continuous heating and cooling
Application of the model to continuous heating and cooling requires numerical integration of
equation (418) over the weld thermal cycle:
(421)
Under such conditions the volume fraction of the precipitates is given by:
(422)
Equations (421) and (422) provide a basis for predicting the extent of particle dissolution
occurring within the HAZ during welding in the absence of impingement of diffusion fields
from neighbouring precipitates.
Dissolution time, s
Particle diameter, nm
Fig. 4.16. The dissolution time of NbC in austenite at 13500C as function of initial particle diameter lro
for different Nb and C levels (numerical solution). Data from Suzuki et al.6
Example (4.6)
Consider stringer bead deposition (SAW) on a thick plate of a Nbmicroalloyed steel (0.10
wt% C  0.03 wt% Nb) under the following conditions:
Assume that the base metal contains a fine dispersion of NbC precipitates in the asreceived
condition. Calculate on the basis of equation (422) and the Rosenthal thick plate solution
(equation (145)) the extent of particle dissolution occurring within the fully transformed HAZ
during welding. Relevant physical data for Nbmicroalloyed steels are given below:
Solution
In the present example the problem is to calculate the variation in the//fo ratio across the fully
transformed HAZ. By substituting the appropriate values for qo, X, a, and v into the Rosenthal
thick plate solution, the governing heat flow equation becomes:
Since it is reasonable to assume that the dissolution rate of the precipitates is controlled by
diffusion of Nb in austenite, the dimensionless supersaturation reduces to:
This gives:
By substituting the appropriate expressions for aNb and DNb into equation (422), we obtain:
Here the lower and upper integration limits refer to the total time spent in the thermal cycle
from Ac3 to T and down again to Ac3.
The extent of particle dissolution occurring within the HAZ during welding can now be
calculated in an iterative manner by numerical integration of the above equation over the weld
thermal cycle. The results from such computations are presented graphically in Fig. 4.17.
It is evident from these data that NbC starts to dissolve when the peak temperature of the
thermal cycle T exceeds the equilibrium dissolution temperature Td of the precipitate. The
process is completed when T approaches 13300C, conforming to a temperature interval of
1900C. This shows that considerable superheating is required in order to overcome the inherent kinetic barrier against particle dissolution under the prevailing circumstances.
4.4.2 Numerical solution
In the previous treatment, no consideration is given to impingement of diffusion fields from
neighbouring precipitates or the position of the particle/matrix interface during the dissolution
process. In certain cases, however, such phenomena will have a marked effect on the dissolution kinetics.18"22 A good example is AlMgSi alloys where the hardening P"(Mg2Si)phase
forms a very fine distribution of needleshaped precipitates along <100> directions in the
aluminium matrix. These precipitates are closely spaced and will therefore interact strongly
with each other during dissolution (coupled reversion).
f/fo
No dissolution
Complete dissolution
Peak temperature, 0C
Fig. 4.17. Dissolution of NbC during steel welding (Example (4.6)).
Concentration
(b)
Distance
Fig. 4.18. Numerical model for dissolution of rodshaped particles in a finite, depleted matrix; (a) Dissolution cell geometry, (b) Particle/matrix concentration profile (moving boundary).
For a specific alloy, the ratio between ro and L (the mean interparticle spacing) can be
calculated from a simple mass balance, assuming that all solute is tiedup in precipitates. Taking this ratio equal to 0.06 for rodshaped precipitates in diluted alloys,9 the kinetics of particle
dissolution during isothermal heat treatment have been examined for a wide range of operational conditions. These results are presented in a general form in Fig. 4.19 by the use of the
following groups of dimensionless parameters:
Dimensionless time
(424)
Dimensionless supersaturation
(425)
where c2 is a kinetic constant, and n{ is a time exponent (assumed constant and equal to 0.5
under the prevailing circumstances).
The rate of particle dissolution will gradually decline with increasing values of T as a result
of impingement of diffusion fields from neighbouring precipitates which reduces a. In practice, this is seen as a continuous decrease in the slope of the flfox curves in Fig. 4.19 (nx < 0.5).
In such cases equation (426) will only be valid within small increments of X.
log T
Fig. 4.19. Dissolution kinetics of rodshaped particles in a finite, depleted matrix. Data from Myhr and
Grong.9
(427)
where n{ < 0.5.
This equation can further be simplified if we assume that
and
(428)
For isothermal heat treatment at a chosen reference temperature (Tn), the rate of particle
dissolution is determined by the retention time tr{. Let f* denote the maximum hold time
required for complete dissolution of the precipitates. It follows that equation (428) can be
written in a general form by normalising tn with respect to t*x. The parameter t*n is obtained
by setting flfo = 0:
(429)
where C3 is a new kinetic constant.
If heat treatment is carried out at a different temperature (T * Tn) , the maximum hold time
t* is simply given by:
(430)
By inserting the approximate expressions for C1 and Dm into equation (430) (see previous
examples), and rearranging equation (428), we obtain:
(431)
and
(432)
where Q'apP. is the apparent (metastable) solvus boundary enthalpy (defined in Section 4.2.3.2),
and Qs is the activation energy for diffusion of the less mobile constitutive atom of the precipitates.
Equations (431) and (432) exploit some good modelling techniques. For example, the use
of a dimensionless time eliminates an unknown kinetic constant which premultiplies t and ^1*
in the derivation of equation (432). Moreover, by raising the dimensionless time to a power
H1 means that the premultiplying constant, here unity, is independent of the value of nv and is
itself also dimensionless. Finally, the form of equation (431) eliminates further unknown
kinetic constants, and may readily be calibrated using an experimental time r* at a reference
temperature.
Figure 4.20 shows the variation inflfo with time (on log axes), from a range of isothermal
experiments carried out on 6082T6 aluminium alloys, using hardness (or electrical conductivity) measurements to evaluate/7/\ The curve (equation (432)) extrapolates back to a slope
of 0.5 (the exponent n{) for the case of the early stages of dissolution before impingement of
the diffusion fields. The exponent nx is seen to fall to lower values when the proportion dissolved is higher, in agreement with the theoretical curves in Fig. 4.19.
log(if/f 0 )
log (t/t*)
Fig. 4.20. 'Mastercurve' for dissolution of hardening p"(Mg2Si)precipitates in 6082T6 aluminium
alloys. Data from Myhr and Grong.9
(433)
where G(X) and H{T) are arbitrary functions of X and T, respectively.
For a given thermal history T(t), this essentially means that the differential equation contains separable variables of X and T. The same criterion may also be applied to the models
derived above. In the case of coupled reversion, we may write:
(434)
and
(435)
Since dfldlx and t\ are unique functions of/and T, respectively, the additivity condition is
satisfied. Consequently, when the temperature varies with time, we replace the term 11 t*x in
equation (434) by dt I t*x and integrate over the thermal cycle, giving:
(436)
This integral is called the kinetic strength of the thermal cycle with respect to reversion.
The resulting volume fraction of the precipitates following a heating cycle is then found by
evaluating the integral Z1 numerically (e.g. by utilising input data from Table 4.2) and replacing 111* with Z1 in equation (432), yielding a value for flfQ from the master curve of Fig. 4.20.
Case Study (4.1)
By utilising equation (436) and the general heat flow model for welding on medium thick
plates (i.e. equation (1104)), it is possible to calculate the variation in the f/fo ratio (i.e. the
solute distribution) across the HAZ of single pass 6082T6 aluminium weldments for a wide
range of operational conditions (see Table 4.3). The results from such computations are presented graphically in Fig. 4.21 4.23.
When stringer bead deposition is carried out on a plate of medium thickness, the solute
distribution in the transverse y direction is expected to vary with distance from the plate surface due to a continuous change in the heat flow conditions. A closer inspection of Figs. 4.21
and 4.22 shows that this is correctly accounted for in the present model. In contrast, a full
penetration butt weld will always reveal a similar solute distribution in the transverse direction
of the weld, as shown in Fig. 4.23. This situation arises from the lack of a temperature gradient
in the throughthickness z direction of the plate.
Table 4.2 Basic input data in dissolution model for hardening p" (Mg2Si) precipitates in 6082T6
aluminium alloys. Data from Myhr and Grong.9
Parameter
Q'app
Qs
nx
(starting value)
t\
(375C)
Value
3OkJm0I1
13OkJm 0 I 1
0.5
600 s
Wo
Peak temperature, 0C
(a)
Comolete dissolution
Aym ,mm
f/fo
Peak temperature, 0C
(b)
Complete dissolution
ym,mm
Fig. 4.21. Dissolution of p"(Mg2Si)precipitates during aluminium welding (Weld 1); (a) Upper plate
surface, (b) Lower plate surface. Operational conditions as in Table 4.3.
Table 4.3 Operational conditions used in aluminium welding experiments (Case study 4.1).
Weld
Material
No.*
Plate thickness
Welding speed
(mm)
(kW)
(mms"1)
AA6082
15
9.1
4.2
AA6082
15
5.7
9.1
3
AA6082
13
14.0
5.8
* Calibration of heat flow model is done by including an empirical correction for heat consumed in melting of the
parent material (thermal data for the AA6082 alloy are given in Table 1.1, Chapter 1).
Complete dissolution
Peak temperature, 0C
Mo
(a)
Aym,mm
Complete dissolution
Peak temperature, 0C
f/fo
(b)
ym,mm
Fig. 4.22. Dissolution of (3"(Mg2Si)PrCCiPiIaIeS during aluminium welding (Weld 2); (a) Upper plate
surface, (b) Centre of plate. Operational conditions as in Table 4.3.
Peak temperature, 0C
Wo
Complete dissolution
Ay ,mm
m
Fig. 4.23. Dissolution of p"(Mg2Si)precipitates during aluminium welding (Weld 3). Operational conditions as in Table 4.3.
f/fo
Figure 4.24 shows plots of the variation in the flfo ratio across the HAZ of 6082T6 aluminium weldments for different values of qo Ivd. It follows that a narrow width of the dissolution zone requires the use of a low energy per mm2 of the weld. In practice, this can be
achieved by the use of an efficient welding process (e.g. electron beam or laser welding) which
facilitates deposition off a full penetration butt weld without employing a groove preparation
(i.e. eliminates the need for filler metals).
Scale:
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
R.D. Doherty: Physical Metallurgy, 3rd Edn (Eds R.W. Chan and R Haasen), 1983, Amsterdam, NorthHolland Physics Publ., 9341030.
K.E. Easterling: Introduction to the Physical Metallurgy of Welding, 1983, London, Butterworths
& Co., Ltd.
H. Adrian and RB. Pickering: Mater. ScL TechnoL, 1991, 7, 176182.
B. Loberg, A. Nordgren, J. Strid and K.E. Easterling: MetalL Trans., 1984,15A, 3341.
J. Strid and K.E. Easterling: Acta MetalL, 1985, 33, 20572074.
S. Suzuki, G.C. Weatherly and D.C. Houghton: Acta. MetalL, 1987, 35, 341352.
J.L. PettyGalis and R.D. Goolsby: J. Mater. ScL, 1989, 24, 14391446.
D. A. Porter and K.E. Easterling: Phase Transformations in Metals and Alloys, 1981, Wokingham
(England), Van Nostrand Reinhold Co. Ltd.
O.R. Myhr and 0. Grong: Acta MetalL Mater., 1991, 39, 26932702; ibid., 27032708.
H.R. Shercliff and M.F. Ashby: Acta MetalL Mater., 1990, 38, 17891802; ibid., 18031812.
J.M. Lifshitz and V.V. Slyozov: /. Phys. Chem. Solids, 1961,19, 3550.
C. Wagner: Z. Electrochem., 1961, 65, 581591.
L.C. Brown: Acta MetalL Mater., 1992, 40, 12931303.
J.C. Ion, K.E. Easterling and M.F. Ashby: Acta MetalL, 1984, 32, 19491962.
S. Matsuda and N. Okumura: Trans. ISIJ, 1978,18, 198205.
H.B. Aaron, D. Fainstein and G.R. Kotler: J. Appl. Phys., 1970, 41, 44044410.
MJ. Whelan: Metal ScL J., 1969, 3, 9597.
J. Agren: Scand. J. MetalL, 1990,19, 28.
R.A. Tanzilli and R.W. Heckel: Trans. Met. Soc. AIME, 1968, 242, 23132321.
H.B. Aaron and R. Kotler: MetalL Trans., 1971, 2, 393408.
R. Asthana and S.K. Pabi: Mat. ScL Eng., 1990, A128, 253258.
U.H. Tundal and N. Ryum: MetalL Trans., 1992, 23A, 433444; ibid., 445449.
H.R. Shercliff, 0. Grong, O.R. Myhr and M.F. Ashby: Proc. 3rd Int. Conf. on Aluminium
Alloys Their Physical and Mechanical Properties, Trondheim, Norway, June 1992, Vol. Ill,
pp. 357369, The University of Trondheim, The Norwegian Inst. of Technol.
J.W. Christian: Phase Transformations in Metals and Alloys, 1975, Oxford, Pergamon Press.
Appendix 4.1
Nomenclature
thermal diffusivity (mm2 s l)
equilibrium concentration of
element A in matrix (wt% or
at.%)
analytical content of element
A in alloy (wt% or at.%)
equilibrium concentration of
element A across a curved
particle/matrix interface
(wt% or at.%)
GMAW
H(T)
arbitrary function of T
standard enthalpy of reaction (J mol"1 or kJ mol"1)
amperage (A)
kinetic strength of thermal
cycle with respect to reversion
equilibrium constant
mean interparticle spacing in
dissolution model (m, Jim or
nm)
integer
atomic weight of element A
(g mol"1)
atomic weight of binary precipitate (g mol"1)
atomic weight of element B
(g mol1)
integer
time exponent in dissolution
model
pressure caused by curvature
effects (J nr 3 )
net arc power (W)
apparent (metastable) solvus
boundary enthalpy (kJ mor 1 )
activation energy for diffusion (kJ moH)
arbitrary function of X
threedimensional radius
vector in Rosenthal equation
(mm)
state variable
time (s)
time necessary for complete
dissolution of precipitate(s)
integration limits (s)
time necessary for complete
reversion at T(s)
density of precipitate An Bm
(g cm"3 or kg rrr3)
thermal conductivity (W
mm1 0C1)
temperature (0C or K)
density (g cm"3 or kg nr 3 )
arc efficiency factor
particle/matrix interfacial
energy (J m~2)
5
Grain Growth in Welds
5.1 Introduction
Grain growth is an important aspect of welding metallurgy. Normal grain growth in metals
and alloys is a thermally activated process driven by the reduction in the grain boundary energy. Physically, it occurs by growth of the larger grains at the expense of the smaller ones
which tend to shrink.
Under isothermal heat treatment conditions, normal grain growth is well described by the
following empirical equation:1
(51)
where D is the mean grain size (diameter), D0 is the initial grain size, n is the time exponent,
/ is the isothermal annealing time, Qapp. is the apparent activation energy for grain growth, and
C1 is a kinetic constant. The other symbols have their usual meaning.
For most metals and alloys the time exponent n in equation (51) varies typically in the
range from 0.1 to 0.4, as shown in Fig. 5.1. Only in the case of ultrapure metals annealed at
very high temperatures the time exponent may approach a constant value of 0.5. This corresponds to the limiting case where the grain boundary migration rate is directly proportional to
the driving pressure 7 /D (7 denotes the grain boundary interfacial energy per unit area, while
D is the grain size).
It is well recognised that alloying and impurity elements both in the dissolved state and in
the form of inclusions or second phase particles will retard grain growth.2"* Consequently, a
comprehensive theoretical treatment of grain growth in welds must include a consideration of
such effects. The present analysis will therefore start with a closer examination of factors
affecting the grain boundary mobility in metals and alloys under conditions applicable to welding.
Z.R.AI
Brass
(53)
from which
(54)
It follows that the mean linear grain intercept D* is always smaller than the actual grain size
D.
A common observation in metals and alloys is that the size distributions of grain aggregates
at different annealing times become equivalent when the measured grain size parameter, D, is
normalised (scaled) by the timedependent average of this metric, D (see Fig. 5.2). This
means that grain structures can be completely characterised, in a statistical sense, by simple
probability functions of the standard deviation of the distribution together with the time dependence of the average size scale D.
5.2.2 Driving pressure for grain growth
The thermodynamic driving pressure for motion of a spherically curved element of grain boundary is given by:7
(55)
where y is the grain boundary interfacial energy, and p* is the radius of curvature.
It is conventional practice to replace p* in equation (55) with some measure of the average
grain size, such as the mean linear grain intercept D*, or with the diameter of the equivalent
spherical volume of some geometrically modelled average grain size. Experimental measurements performed on high purity aluminium indicate that p* ~ 3.23 D*.7 This observation is
consistent with the model of Hellman and Hillert,8 which predicts that the curvature of the
most critical element of grain boundary that must be stabilised is p* = 3 D . Under such conditions, the driving pressure becomes:
(56)
F (arbitrary units)
Rayleigh distribution
Lognormal
distribution
D/D
Fig. 5.2. Comparison of measured grain size data in iron with the Rayleigh and lognormal distribution
functions F. The similarity of the size distributions at different annealing times illustrates the selfsimilar
scaling behaviour of normal grain growth. After Pande.6
In practice, the numerical constant in equation (56) can vary by, at least, a factor of three,
depending on the assumptions of the models. Consequently, in the general case the average
driving pressure is given by:
(57)
where c2 is a constant which is characteristic of the system under consideration.
5.2.3 Drag from impurity elements in solid solution
In the dissolved state impurity and alloying elements will retard grain growth through elastic
attraction of the atoms towards the open structure of the grain boundary. The boundary must
then either drag the impurity atoms along (so that its speed is limited by the diffusion rate of
these atoms) or break away if the impurity concentration is sufficiently small or the driving
pressure or temperature is high enough.
An analysis of such effects can be done on the basis of the classic impurity drag theories of
Chan2 and Lucke and Stiiwe,3 which deal with the following two extreme cases:
(i) A low velocity limit, where the rate of grain boundary migration is controlled by diffusion of impurity atoms perpendicular to the boundary.
(ii) A high velocity limit, where the grain boundary migration process is mainly governed
by the diffusion of solvent atoms across the boundary (i.e. controlled by the rate of boundary
self diffusion).
The low velocity limit is associated with either a low driving pressure PG or a high impurity
level C0, and is characterised by a linear type of relationship between the grain boundary
migration rate Va and PG:2
(58)
Here e denotes the intrinsic drag coefficient, while 1F is a parameter depending on the
diffusivity and the interaction energy between the grain boundary and the impurity atoms.
For the other extreme (i.e. the high velocity limit), the grain boundary migration rate Vb is
described by the relationship:2
(59)
where 4Vp2 denotes another complex function of the impurity diffusivity and the interaction
energy between the grain boundary and the impurity atoms.
In the case of spherical grains, the classic grain growth equation predicts that the grain
boundary migration rate V is a power function of the driving pressure (y /D). This is readily
seen by differentiating equation (51) and inserting 7 (the grain boundary interfacial energy)
into the resulting expression:
(510)
where c3 is a new kinetic constant.
Note that at very high driving pressures or low impurity levels, equation (59) approaches
the limiting case where the migration rate Vb becomes directly proportional to PG, corresponding to a time exponent n = 0.5 in the grain growth equation. At this point the grain boundary
will break away from its surrounding impurity atmosphere and migrate at a rate close to the
rate of boundary self diffusion. In most cases, however, the relationship between V and PG
derived from equation (510) will be different from the theoretical one due to the empirical
nature of the grain growth equation.
For high driving pressures and intermediate impurity concentrations, the classic impurity
drag theories predict a discontinuous transition from the high to the low velocity limit. It has,
however, been argued by Vandermeer9 that the observed transition may be considered as continuous. Thus, in a log V vs log PG plot it would appear as a steep curve connecting the lines
for the high and low velocity extremes together, as illustrated schematically in Fig. 5.3.
5.2.4 Drag from a random particle distribution
The retardation of grain growth by second phase particles was first theoretically investigated
by Zener.10 There seems to be general agreement that the maximum pinnig force Fp exerted by
a single particle of radius r on a grain boundary is given by: 410
(511)
log V
Transition region
log PG
Fig. 5.3. Schematic variation of grain boundary migration rate V with driving pressure PQ according to
the classic impurity drag theories of Cahn2 and Liicke and Stiiwe.3 The diagram is based on the ideas of
Vandermeer.9
If the number of interacting particles per unit area of the grain boundary is taken equal to na,
the resulting retardation pressure becomes:
(512)
Assuming that only one half of the particles which touches the grain boundary will interact
with a maximum force, na is related to Nv (the number of particles per unit volume) through the
following equation:410
(513)
Given that the particles are spherical and of uniform size, Nv can be expressed as:
(514)
where/is the particle volume fraction.
A combination of equations (512), (513), and (514) leads to the wellknown expression
for the socalled Zener drag (or Zener retardation pressure):410
(515)
In practice, the numerical constant in equation (515) can vary by, at least, a factor of five,
depending on the assumptions of the models.11 Consequently, in the general case the Zener
retardation pressure is given by:
(516)
where C4 is a constant which is characteristic of the system under consideration.
5.2.5 Combined effect of impurities and particles
As already stated in the introduction of this chapter, the time exponent n is a measure of the
resistance to grain boundary motion in the presence of impurity and alloying elements in solid
solution. Based on equation (51) Hu and Rath 112 have shown that the grain boundary migration rate V is related to the effective driving pressure APG and the time exponent n through the
following equation:
(517)
where M is the grain boundary mobility.
In alloys containing grain boundary pinning precipitates, the effective driving pressure APG
is defined as the numerical difference between PG and Pz. By inserting the correct expressions
for PG and Pz into equation (517), we obtain:
(518)
It follows from equation (518) that the grain boundary migration rate V becomes inversely
proportional to the average grain size D when n = 0.5 a n d / = 0. This corresponds to the
limiting case where the grain boundary will break away from its surrounding impurity atmosphere and migrate at a rate which is controlled by the diffusion of solvent atoms across the
boundary. In most cases, however, the observed relationship between V and APG will be
different from the theoretical one due to drag from second phase particles (f> 0) or impurity
elements in solid solution (n < 0.5).
and
(520)
From this it is seen that the parameters M0 and k are true physical constants which are
related to the grain boundary mobility and the pinning efficiency of the precipitates, respectively.
5.3.1 Limiting grain size
Equation (520) shows that the grain structure is stabilised when (d D ldi)  0. The stable
(limiting) grain size is given by:
(521)
The parameter k (which in the following is referred to as the Zener coefficient) is defined as
the ratio between the numerical constants in equations (57) and (516), respectively. In the
original Zener's model k = 4/3, while other investigators have arrived at different results.811"14
As shown in Fig. 5.4, the limiting grain size may vary by over one order of magnitude, depending upon the assumptions of the models. This makes it difficult to apply equations (520) and
(521) for quantitative grain size analyses without further background information on the Zener
coefficient.
Diim.nm
Gladman
r/f, um
Fig. 5.4. Relation between limiting grain size Dum., particle radius r, and volume fraction/predicted by
different models.
Example (5. J)
Consider multipass GMA welding on a thick steel plate under the following conditions:
Based on the models of Zener,10 Hellman and Hillert,8 and Gladman13 estimate the limiting
austenite grain size Dlim in the transformed parts of the weld HAZ when the oxygen and
sulphur contents of the asdeposited weld metal are 0.04 and 0.01 wt%, respectively.
Solution
As shown in Chapter 2 of this textbook the volume fraction of oxide and sulphide inclusions
can be calculated from equation (275):
Similarly, the average radius of the grain boundary pinning inclusions can be obtained from
equation (279):
This gives the following values for the limiting austenite grain size:
Zener:
Gladman:
and
As expected, the limiting austenite grain size is seen to vary by more than one order of
magnitude, depending on the assumptions of the models. In practice, the Zener coefficient in
lowalloy steel weld metals falls within the range from 0.32 to 0.93, as shown in Fig. 5.5. The
average value of A: is close to 0.52, which is the same as that inferred from the Gladman model
(upper limit). When it comes to intermetallic compounds such as titanium nitride, the Zener
coefficient varies typically between 0.75 and 0.25 during grain growth in the austenite regime. 1617 This suggests that k ~ 0.50 is a reasonable estimate of the grain boundary pinning
efficiency of oxides and nitrides in steel.
5.3.2 Grain boundary mobility
Direct application of equation (520) requires also reliable information on the time exponent n
and the grain boundary mobility M. When n = 0.5 and/= 0, the classic impurity drag theories
predict that the activation energy Qapp, should be close to the value for boundary self diffusion
in the matrix material.2'3 This borderline case is approximately attained in steel welding, as
shown in Fig. 5.6(a) and (b), since the driving pressure for austenite grain growth immediately
following the dissolution of the pinning precipitates is usually so large that the grain boundary
migration rate approaches the higher velocity limit defined in equation (59).18 On this basis it
is not surprising to find that Qapp falls within the range reported for lattice self diffusion (284
kJ mol"1) and boundary self diffusion (170 kJ mol~l) in pure 7iron19 during welding.18 In
most cases, however, the activation energy will be different from the theoretical one due to
complex interactions between impurity atoms and grain boundaries (characterised by a time
exponent n < 0.5). Under such conditions, the value of Qapp has no physical meaning.1
5.3.3 Grain growth mechanisms
Equation (520) provides a basis for evaluating the grain growth inhibiting effect of impurity
elements and second phase particles under different thermal conditions. This also includes
situations where the grain boundary pinning precipitates either coarsen or dissolve during the
heat treatment process.
5.3.3.1 Generic grain growth model
Equation (520) can readily be integrated to give the average grain size D as a function of
time. In the general case we may write:
(522)
(a)
Austenite grain size, Jim
Annealing temperature, 0 C
(b)
Dm. Hm
r/f, urn
Fig. 5.5. Evaluation of the Zener coefficient in steel weld metals containing stable oxide and sulphide
inclusions; (a) Determination of Dum. from isothermal grain growth data (holding time: 30 min),
(b) Variation in Dum. with the inclusion rlf ratio. Data from Skaland and Grong.15
where D{im is the limiting grain size (defined in equation (521)).
The integral I1 on the righthand side of equation (522) represents the kinetic strength of the
thermal cycle with respect to grain growth and can be determined by numerical methods when
the temperaturetime programme is known. In practice, however, it is not necessary to solve this
integral to evaluate the grain growth mechanisms. Consequently, the lefthand side of equation
(a)
Steel A
Log (DxZD1)
Slope: n = 0.4
(b)
LogD7
Steel A
Steel B
Fig. 5.6. Evaluation of the time exponent n and the activation energy Q for austenite grain growth in
steel under thermal conditions applicable to welding; (a) Time exponent n, (b) Activation energy Qapp.
Data from Akselsen et ah18
(522) can be solved explicitly for different values of DUm, n, and Z1. The results may then be
presented in the form of novel diagrams which show the competition between the various
processes that lead to grain growth during heat treatment of metals and alloys.
A more thorough documentation of the predictive power of the model and its applicability
to welding is given in Section 5.4.
5.3.3.2 Grain growth in the absence ofpinning precipitates
In the absence of grain boundary pinning precipitates, we have:/= 0, Dlim > ~ , and (1/ DlitrL) = 0.
Under such conditions, equation (522) reduces to:
(523)
If grain growth occurs in the presence of stable precipitates, the limiting grain size {Dlim) in
equation (522) becomes constant and independent of the thermal cycle. In the specific case
when n = 0.5 the integral on the lefthand side of equation (522) has the following analytical
solution:
D, ^m
(525)
I1W"
Fig. 5.7. Predicted variation in average grain size D with /, and n f o r / = 0 and D0 = 0 ('free' grain
growth).
(a)
D Y ,fim
Steel A
Number of cycles
(b)
D y ,um
Steel B
Number of cycles Fig. 5.8. Illustration of the austenite grain growth inhibiting effect of phosphorus and free nitrogen in
lowalloy steel during reheating above the Ac^ temperature (multicycle weld thermal simulation);
(a) Steel A (50ppm P, 20ppm N), (b) Steel B (180ppm P, 80ppm N). Data from Akselsen et a/.18
from which the average grain size D is readily obtained. In other cases, numerical methods
must be employed to evaluate D.
It is evident from the graphical representation of equation (525) in Fig. 5.9 that the grain
growth inhibiting effect of the precipitates is very small during the initial stage of the process
when D D lim. Under such conditions the grains will coarsen at a rate which is comparable
with that observed for free grain growth (n = 0.5,/= 0). The grain coarsening process becomes
gradually retarded as the average grain size increases because of the associated reduction in
the effective driving pressure APG until it comes to a complete stop when AP0 = 0 (i.e. D =
D Hm)
D, jim
I 1 ^m 2
D, (im
Fig. 5.9. Predicted variation in average grain size D with Z1 and Dnm. for n = 0.5 and D0 = 0 (stable
precipitates).
I 1 , [nm]1/n
Fig. 5.10. Predicted variation in average grain size D with Z1 and n for Dum. = 250Jin and D0 = 0
(stable precipitates). Dotted curves correspond to grain growth in the absence of pinning precipitates.
If grain growth at the same time occurs under the action of a constant drag from impurity
elements in solid solution, the situation becomes more complex. As shown in Fig. 5.10, a
decrease in the time exponent from say 0.5 to 0.2 gives rise to a marked reduction in the slope
of the DI 1 curves, similar to that observed in Fig. 5.7 for particlefree systems (/= 0). However, the predicted grain coarsening rate is lower than that evaluated from equation (524) due
to the extra drag exerted by the grain boundary pinning precipitates. This leads ultimately to a
stabilisation of the microstructure when D = DUm.
5.3.3.4 Grain growth in the presence of growing precipitates
Very little information is available in the literature on the matrix grain growth behaviour of
metals and alloys in the presence of growing second phase particles. So far, virtually all
modelling work has been carried out on two phase a(3 titanium alloys.14 Unfortunately, none
of these models can be extended to more complex alloy systems such as steels or aluminium
alloys.
When grain growth occurs in the presence of growing second phase particles, Dum. will no
longer be constant due to the associated increase in the particle rlf ratio with time. As shown in
Chapter 4 of this textbook, the LifshitzWagner theory2021 provides a basis for modelling
particle growth during welding and heat treatment of metals and alloys in cases where the peak
temperature of the thermal cycle is kept well below the equilibrium solvus of the precipitates.
Under such conditions, the particles will coarsen at almost constant volume fraction (f=fo), in
accordance with equation (416):
(526)
where Qs is the activation energy for the coarsening process, C5 is a kinetic constant, and I2 is
the kinetic strength of the thermal cycle with respect to particle coarsening. The other symbols
have their usual meaning.
If the base metal contains particles of an initial radius ro and volume fraction/^, the limiting
grain size at I2 = 0 (D lim) can be defined as:
(527)
from which
(528)
Similarly, when I2 > 0, we may write:
(529)
By combining equations (526), (528), and (529), we arrive at the following relationship
between (D Um ) and I2:
(530)
It is seen from equation (530) that the limiting grain size in the presence of growing
particles depends on the product (k/fo)3I1. In practice, the grain boundary pinning effect of the
precipitates is determined by the relative rates of particle coarsening and grain growth in the
material, i.e. whether the grain boundary mobility is sufficiently high to keep pace with the
increase in DUm during heat treatment. Generally, the pinning conditions are defined by the
(k/fo)3 I1IIx ratio, which after substitution and rearranging yields:
(531)
In cases where the parameters c5 ,Qs, M0*, and Qapp are known, the average grain size D
can readily be evaluated from equations (522), (530), and (531) by utilising an appropriate
integration procedure. However, since Qs normally differs from Qapp^ the (klfo)3I1I Ix ratio
will depend on the thermal path during continuous heating and cooling. Consequently, solution of these coupled equations generally requires stepwise integration in temperaturetime
space via a fourth heat flow equation. This problem will be dealt with in Section 5.4.
The situation becomes much simpler if heat treatment is carried out isothermally. Under
such conditions the product (k/fo)311 will only differ from Ix by a proportionality constant m,
which is characteristic of the system under consideration. Accordingly, equation (530) can be
rewritten as:
(532)
From this we see that the two coupled equations (522) and (532) can be solved explicitly
for different values of Dnm., n, m, and Z1. Hence, it is possible to present the results in the form
of novel 'mechanism maps' which show the competition between particle coarsening and
grain growth during isothermal heat treatment for a wide range of operational conditions.
Examples of such diagrams are given in Figs. 5.11 and 5.12.
It is evident from these figures that the grain coarsening behaviour during isothermal heat
treatment is very sensitive to variations in the proportionality constant m. For large values of
m, the matrix grains will coarsen at a rate which is comparable with that observed in Fig. 5.7
for particlefree systems (f = 0). This corresponds to a situation where the grain boundary
pinning precipitates will completely outgrow the matrix grains. It is interesting to note that
particle outgrowing is more likely to occur if the time exponent n is small, as shown in Fig.
5.12, because of the associated reduction in the grain boundary mobility in the presence of
impurity elements in solid solution. In other systems, where the proportionality constant m is
closer to unity, the reduced coarsening rate of the precipitates gives rise to a higher Zener
retardation pressure and ultimately to a stagnation in the matrix grain growth. In the limiting
case, when m = 0, the grain growth behaviour becomes idential to that observed in Figs. 5.9
and 5.10 for stable precipitates.
D.jim
I 1 ,^m 2
Fig. 5.11. Predicted variation in average grain size D with Ix and m for Dum. = 50jLim, n = 0.5, and
D0 =0 (growing precipitates).
D,fxm
I1^m1'"
Fig. 5.12. Predicted variation in average grain size D with I1 and m for Dum.  50im, n = 0.3, and
D0 = 0 (growing precipitates).
Example (5.2)
Solution
The initial volume fraction of TiN in the material can be estimated from simple stoichiometric
calculations by considering the difference between total and soluble titanium. Taking the
atomic weight of Ti and N equal to 47.9 and 14.0 g mol"1, respectively, we obtain:
From this we see that the initial radius of the TiN precipitates in the base metal is close to:
Since heat treatment is carried out under isothermal conditions, the parameters m and Z1 can
be obtained directly from equations (531) and (522) without performing a numerical integration:
The average austenite grain size can now be read from Fig. 5.11 by linear interpolation
between the curves for m = 10 and 100 Jim. This gives:
Although experimental data are not available for a direct comparison, the predicted grain
size is of the expected order of magnitude. From this it is obvious that considerable austenite
grain growth may occur in titaniummicroalloyed steels because of particle coarsening, in
spite of the fact that TiN, from a thermodynamic standpoint, is stable up to the melting point of
the steel. The process can, to some extent, be counteracted by the use of a finer dispersion of
TiN precipitates in the material. For example, if the initial particle radius is reduced by a factor
of five (conforming to a change in I W from 50 to 10 Jim), the austenite grain size of the
annealed material decreases from 75 to 65 jLim, as shown in Fig. 5.13. Nevertheless, since
particle coarsening is a physical phenomenon occurring during high temperature heat treatment of metals and alloys, austenite grain growth cannot be avoided. This explains why, for
instance, conventional titaniummicroalloyed steels are not suitable for high heat input welding due to their tendency to form brittle zones of Widmanstatten ferrite and upper bainite in the
coarse grained HAZ region adjacent to the fusion boundary.22
D,jim
Stable particles
I1Hm2
Fig. 5.13. Predicted variation in average grain size D with Z1 and m for DHm. = 10 um, n = 0.5, and
Do = 0 (growing precipitates).
where I3 is the kinetic strength of the thermal cycle with respect to particle dissolution.
From this relation the following expression for the particle volume fraction can be derived
(see equation (422), Chapter 4):
(535)
where fo is the initial particle volume fraction.
By substituting Dlim  k(rlf) and Dnm. = Kr0If0) into equations (534) and (535), it is
possible to obtain a simple mathematical relation which describes the variation in the limiting
grain size with I3 during particle dissolution. After some manipulation, we obtain:
(536)
It is seen from equation (536) that the limiting grain size increases from Dum. at I3 = 0 to
infinite when I3 = (fo Ik)2 (Dum. )2 Since the magnitude of the Zener drag, in practice, depends
on the relative rates of grain growth and particle dissolution in the material, the pinning conditions are defined by the (klfo)2131 Ix ratio:
(537)
Equation (537) shows that the (k/fo)2I3111 ratio is contingent upon the thermal path during
continuous heating and cooling. Consequently, application of the model to welding generally
requires numerical integration of the coupled equations (522), (536), and (537) over the
weld thermal cycle.
However, the integration procedure is largely simplified if heat treatment is carried out
isothermally. In such cases the product (k/fo)213 will only differ from Ix by a proportionality
constant m*, which is characteristic of the system under consideration. By substituting m*Ix
into equation (536), we obtain:
(538)
From this we see thaUhe two coupled equations (522) and (538) can be solved explicitly
for different values of Dun., n,m*, and I1. Hence, it is possible to present the results in the
form of novel 'mechanism maps' which show the competition between particle dissolution
and grain growth during isothermal heat treatment for a wide range of operational conditions.
Examples of such diagrams are given in Figs. 5.14 and 5.15.
As expected, the stability of the second phase particles is sensitive to variations in the
proportionality constant m*. Normally, the precipitates will exert a drag on the grain boundaries as long as they are present in the metal matrix. However, when the dissolution process is
completed, the matrix grains are free to grow without any interference from precipitates. This
D,jim
Complete particle
dissolution
Stable particles
I 1 ,nm 2
Fig. 5.14. Predicted variation in average grain size D with Z1 and ra* for Dum. = 50 um, n = 0.5, and
D0 =0 (dissolving precipitates).
D,jim
I 1 ^m 1 ' 0
Fig. 5.15. Predicted variation in average grain size D with I1 and m* for Dum.  50 Jim, n  0.3, and
Do =0 (dissolving precipitates).
means that the grains, after prolonged high temperature annealing, will coarsen at a rate which
is comparable with that observed in Fig. 5.7 for particlefree systems. In the limiting case,
when m* = 0, the grain growth behaviour becomes identical to that shown in Figs. 5.9 and 5.10
for stable precipitates.
Example (5.3)
Solution
The initial volume fraction of NbC in the material can be estimated from simple stoichiometric
calculations by considering the difference between total and soluble niobium. Taking the
atomic weight of Nb and C equal to 92.9 and 12.0 g mol"1, respectively, we obtain:
From this we see that the radius of the NbC precipitates in the base metal is close to:
By substituting this value into the expression for the proportionality constant m*, we obtain:
The average austenite grain size can now be read from Fig. 5.14 by interpolation between
the curves for m* = 1 and/= 0 (free grain growth). This gives:
Since the calculated value of D1 is reasonably close to that observed for a particlefree
system, it means that the presence of a fine dispersion of NbC in the base metal has no significant effect on the resulting austenite grain size under the prevailing circumstances. Other
Next Page
types of niobium microalloyed steels may reveal a different grain coarsening behaviour, depending on the chemical composition, size distribution, and initial volume fraction of the base
metal precipitates. However, the pattern remains essentially the same, i.e. the growth inhibition is always succeeded by grain coarsening as long as the precipitates are thermally unstable.
Previous Page
types of niobium microalloyed steels may reveal a different grain coarsening behaviour, depending on the chemical composition, size distribution, and initial volume fraction of the base
metal precipitates. However, the pattern remains essentially the same, i.e. the growth inhibition is always succeeded by grain coarsening as long as the precipitates are thermally unstable.
(541)
where the limits tx and t2 refer to the total time spent in the thermal cycle from the chosen
reference temperature Tc to the peak temperature Tp and down again to Tc.
Tuning of coarsening model
When the parameters n, Qapp., Qs, D0, and Dlim are known, it is possible to evaluate the
remaining (unknown) kinetic constants from equations (522) and (530) by an iterative procedure. The following parameter is defined for this purpose:
(542)
The next step is to calculate the integral
(543)
for the same pairs of values of Tp and Ar875 as above. The differential grain growth equation
can now be solved by selecting an appropriate starting value for M0* and evaluating the corresponding Qvalue which conforms to a mean grain size of D\ and D2, respectively. The
computations are repeated by adjusting M0* until a contour in Mo*Q space is built up for each
grain size. The accepted values of M* and Q are then found by considering the intersection
point between the two curves, as shown schematically in Fig. 5.16(a).
(a)
MS.
Accepted,
"" value
Accepted^
" ~vaTue~
a
(b)
MJ
Accepted.
~ value" "
Acce3ted_
value" "
Q*
Fig. 5.16. Method for calibrating unknown kinetic constants to experimental grain growth data;
(a) M*oQ. (coarsening model), (b) M*oQ* (dissolution model).
(546)
(547)
(548)
(549)
where Tp < Tn,.
Welding diagrams of this kind are very useful, since they summarise the effect of the important process variables in a systematic and illustrative manner at the same time as they provide
a good overall indication of the grain growth behaviour of materials during welding. In addition, the diagrams can be used for quantitative predictions of the austenite grain size across the
HAZ of steel welds for a wide range of operational conditions. This will be illustrated below
in various numerical examples.
5.4.2 Case studies
The following section describes grain growth diagrams for different types of steels. The parameters used to construct the maps are either tabulated or included in the text. Some of these
are taken from the literature, while others are arrived at by fitting the theory of the previous
sections to data from real or simulated welds according to the procedure shown in Fig. 5.16.
To obtain a consistent basis, all grain sizes reported here conform to the mean linear intercept.28 Conversion to threedimensional grain sizes may then be done through equation (54)
or by the use of other appropriate conversion factors (e.g. 1.776 as recommended by Ashby et
a/.24'25).
5.4.2.1 Titaniummicroalloyed steels
Titaniummicroalloyed steels are widely used in welded structures. From a thermodynamic
standpoint, additions of small amounts of titanium to steel would be expected to impede austenite
grain coarsening during welding by virtue of its ability to form stable nitrides even at high
temperatures (see discussion in Chapter 4). However, certain restrictions must be adhered
during casting and subsequent thermomechanical processing of the steel so that the number
density of TiN particles is sufficiently high to retard grain growth.22 This currently limits the
use of titanium for austenite grain size control to continuously cast and controlled rolled steels.
Typical HAZ grain growth diagrams for Timicroalloyed steels can be constructed on the
basis the experimental data reported by Ion et al.25 Tables 5.1 and 5.2 contain information
about the steel chemical composition and the parameters used in the computations, while Fig.
5.17 gives an overall indication of the accuracy of the predictions after calibration of the model
against two experimental data points.
Table 5.1 Chemical composition of Timicroalloyed steel used by Ion et al.25 (in wt%).
C
Si
Mn
Al
Ti
0.12
0.23
1.53
0.005
0.006
0.02
0.011
0.009
Table 5.2 Data used to construct welding maps for Timicroalloyed steel (compiled from miscellaneous sources).
"
0.5
I ~PP I Q? I Hm.
I K I ^* I
(kJmol1)
(kJmol1)
(xm)
(jun)
(IJLm2S1)
(1JLm3Ks1)
224
240
20
7.7 X 109
9.9 X 1014
Timicroalioyed steel
E
=i
s
(a
I
Oi
I
i
Calibration point
Evaluate on the basis of the nomograms in Fig. 5.18(a) the variation in the austenite grain
size across the fully transformed HAZ after welding. Estimate also the total width of the HAZ
(referred to the fusion boundary) under the prevailing circumstances.
Solution
First we calculate the net heat input per unit length of the weld:
Readings from Fig. 5.18(a) give the HAZ austenite grain size profile shown in Fig. 5.19.
(a)
Peak temperature, 0C
Thin plate welding (1D heat flow)
Prior austenite grain size dim)
Relative size of pinning precipitates (r/ro)
(b)
Peak temperature, 0C
FZg. 5.18. HAZ grain growth diagrams for titaniummicroalloyed steel; (a) Thick plate welding (2D heat
flow), (b) Thin plate welding (1D heat flow). No preheating (T0 = 200C).
Peak temperature, 0C
Fig. 5.19. Predicted variation in austenite grain size across the fully transformed HAZ of a Timicroalloyed
steel weld (Example 5.4).
The total width of the fully transformed HAZ can now be estimated from equation (547)
by using data from Table 1.1 (Chapter 1). Taking the Ac3 temperature equal to 9100C, we
obtain:
Based on the quoted value of 2 in Table 5.2, it is also possible to estimate the initial volume
fraction of TiN in the base metal. Taking C5 6.67 X 104 im3 K s"1 and k 0.5 for titaniummicroalloyed steels, we obtain from equation (542):
A comparison with the experimental data of Ringer et aill shows that a volume fraction of
2 X 10"4 is reasonably close to that measured by microscopic assessment methods.
Typical HAZ grain growth diagrams for Nbmicroalloyed steels (with chemical composition as in Table 5.3) can be constructed on the basis of the experimental data of Ion et al. 25 The
parameters used to compute the diagrams are listed in Table 5.4. Figure 5.20 shows a correlation between predicted and observed grain sizes after calibration of the model against two
experimental data points.
It is evident from the nomograms in Fig. 5.21 (a) and (b) that considerable austenite grain
growth occurs during welding of niobiummicroalloyed steels because of dissolution of the
grain boundary pinning precipitates. On the average, the HAZ austenite grain size adjacent to
the fusion boundary is four to six times larger than that observed for titaniummicroalloyed
steels. This gives rise to a high HAZ hardenability, which facilitates formation of low temperature transformation products such as martensite and bainite during welding ?2
Example (5.5)
In Table 5.4 the quoted value for the effective grain boundary pinning constant Q* is 5.0 X
1018 im 2 s"1. Based on equations (527) and (544), estimate the initial volume fraction/, and
radius ro of NbC in the parent material under the prevailing circumstances.
Table 5.3 Chemical composition of Nbmicroalloyed steel used by Ion et al.25 (in wt%).
C
Si
Mn
Al
Nb
0.12
0.16
0.91
0.002
0.005
0.04
0.021
0.011
Nbmicroalloyed steel
SAW
GMAW
Laser welding
Weld simulation
Calibration point
(a)
Peak temperature, 0C
Thin plate welding (1D heat flow)
Prior austenite grain size (urn)
Relative volume fraction of pinning precipitates (f/fo)
(b)
Peak temperature, 0C
Fig. 5.21. HAZ grain growth diagrams for niobiummicroalloyed steel; (a) Thick plate welding (2D heat
flow), (b) Thin plate welding (1D heat flow). No preheating (T0 = 200C).
Table 5.4 Data used to construct welding maps for Nbmicroalloyed steel (compiled from miscellaneous sources).
n
0.5
(kJmol1)
(kJ mol1)
((Jim)
224
130
343
11
I ^
(xm)
6
I K* I
"^
(1Xm2S1)
(1Xm2S1)
37 X 109
5.0 X 1018
Reasonable average values for a 0 , Dm, k, and Dnm. are given below:
20.5 (estimated from the solubility product of NbC in austenite, Example
4.6)
5.9 X 1010 Jim2 s"1 (diffusion constant for Nb in austenite)
0.5 (Zener coefficient)
11 Lim (from Table 5.4)
Solution
The initial volume fraction of NbC can be estimated from equation (544). After rearranging
this equation, we obtain:
The corresponding radius of the precipitates can now be obtained from equation (527):
Although experimental data are not available for a direct comparison, the calculated values
for/o and ro are reasonable and of the expected order of magnitude (see Example 5.3).
5.4.2.3 CMn steel weld metals
In CMn steel weld metals the volume fraction of nonmetallic inclusions is considerably
higher than for normal cast steel products because of the limited time available for growth and
separation of the deoxidation products (see discussion in Chapter T). For the same reason, the
weld metal inclusions are also significantly smaller in dimension and more finely dispersed.
Since oxides and sulphides will neither coarsen nor dissolve during the weld thermal cycle,
they will ultimately lead to a stabilisation of the austenite grain structure within the reheated
regions of multipass steel weld metals. This is illustrated schematically in Fig. 5.22.
Grain growth diagrams for CMn steel weld metals (with chemical compostion as in Table
5.5) can be constructed on the basis of the theory described in the previous sections by selecting reasonable average values for the kinetic constants in equation (522). The combination of
parameters listed in Table 5.6 gives a fair agreement between predictions and measurements,
as shown in Fig. 5.23.
The response of the material to reheating above the Ac3temperature under 2D and 1D
heat flow conditions is illustrated in Fig. 5.24(a) and (b). As expected, the presence of finely
Weld metal
Base plate
Peak temperature
Fig. 5.22. Schematic diagram illustrating the effect of nonmetallic inclusions on the weld metal grain
coarsening behaviour.
Table 5.5 Chemical composition of CMn steel weld metal used by Kluken et al.29 (in wt%).
C
Si
Mn
Nb
Al
Ti
0.09
0.034
0.48
1.86
0.01
0.01
0.005
0.004
0.02
0.018
0.005
Table 5.6 Data used to construct welding maps for CMn steel weld metal (compiled from
miscellaneous sources).
"
I
0.5
~pp. I K
1
I bHm. I K?
(kJmol" )
(xm)
(xm)
(JJLm2S1)
224
10
95
20X109
dispersed inclusions within the weld metal gives rise to a strong austenite grain boundary
pinning effect, similar to that documented for TiN in microalloyed steels. However, since no
particle coarsening occurs in the present case, it means that a small austenite grain size
(< 95jnm) is preserved at all relevant peak temperatures, irrespectively of the applied heat
input. Accordingly, the weld metal grain growth behaviour is seen to be quite different from
that of the base metal, even when the nominal chemical composition has not been significantly
changed by the welding process.
Cooling time, A t . . s
8/5
Fig. 5.23. Comparison between measured and predicted austenite grain sizes after calibration of model
to data reported by Kluken et al.29 for CMn steel weld metal (simulated thick plate heat cycles).
Example (5.6)
Consider deposition of a cap layer (GMAW) on the top of a thick multipass Nbmicroalloyed
steel weld under the following conditions:
Estimate on the basis of the nomograms in Figs. 5.21(a) and 5.24(a) the variation in the
prior austenite grain size across the fully transformed HAZ at different locations along the
periphery of the weld after arc extinction. The situation is illustrated in Fig. 5.25.
Solution
First we calculate the net heat input per unit length of the weld:
Readings from Figs. 5.21(a) and 5.24(a) give the HAZ austenite grain size profiles shown
in Fig. 5.26. From this we see that the prior austenite grain size adjacent to the fusion boundary of a cap layer will vary significantly with position along the periphery of the weld, depending on the type of material sampled (i.e. base plate or weld metal).
5.4.2.4 CrMo lowalloy steels
Chromiummolybdenum lowalloy steels are widely used in the petroleum industry and in
highpowergenerating equipment because of their good corrosion and oxidation resistance
(a)
Peak temperature, 0C
Thin plate welding (1 D heat flow)
(b)
Peak temperature, 0C
Fig. 5.24. Grain growth diagrams for reheated CMn steel weld metal; (a) Thick plate welding (2D heat
flow), (b) Thin plate welding (1D heat flow). No preheating (T0 = 200C).
HAZ
Base plate
Dissolution of NbC
Peak temperature, 0C
Fig. 5.26. Predicted variation in austenite grain size across the fully transformed HAZ at different locations along the periphery of the weld (Example 5.6).
and their high creep strength. They are readily weldable, although reheat cracking and cold
cracking in the HAZ and fusion zone may be a problem.30 In practice, these difficulties can be
overcome by the choice of an appropriate welding procedure (e.g. preheating in combination
with a low heat input), which reduces austenite grain growth and maximises the proportion of
the HAZ refined by subsequent weld passes.3031
The grain growth data of Miranda and Fortes31 provide a basis of calibrating the kinetic
constants in equations (522) and (545), as shown in Fig. 5.27. Information about steel chemical
composition and parameters used to construct the maps are contained in Table 5.7 and 5.8,
respectively.
It is seen from the nomograms in Fig. 5.28(a) and (b) that considerable austenite grain
growth occurs in CrMo lowalloy steel weldments due to dissolution of molybdenum carbide
(Mo2C) at elevated temperatures. However, the maximum HAZ austenite grain size is signifiTable 5.7 Chemical composition of CrMo steel used by Miranda and Fortes31 (in wt%).
C
Si
Mn
Cr
Mo
Ni
Cu
Al
0.10
0.24
0.46
0.01
0.01
2.19
0.94
0.23
0.1
0.01
0.01
Calibration points
Symbol * w s >
n
0.32
T ~ F I MO* I QT~
(kJmol1)
(kJmol1)
(kJmor"1)
(^m)
(fim)
(IJLm1711S1)
(IXm2S1)
180
71
300
20
1.3 X 109
7.8 X 1017
cantly smaller than that observed during welding of Nbmicroalloyed steels, in spite of the fact
that Mo2C is less stable than NbC. This situation can probably be attributed to drag from
alloying and impurity elements in solid solution, which retard grain growth through elastic
attraction of the atoms towards the open structure of the grain boundary (indicated by a time
exponent n = 0.32 in Table 5.8).
5.4.2.5 Type 316 austenitic stainless steels
The 316 series of stainless steels have good general corrosion resistance, but are sensitive to
stress corrosion cracking in the presence of chlorides. They are used extensively in the chemical industry and in power plants, particularly in advanced nuclear technology. During welding
dissolution of the base metal chromium carbides (e.g. C^C 6 ) will occur in parts of the HAZ
where the peak temperature of the thermal cycle has been above about 8500C. This gives rise
to considerable austenite grain growth.24
HAZ grain growth diagrams for this type of steel (with chemical composition as in Table
5.9) can be constructed on the basis of the experimental data of Ashby and Easterling,24 using
the parameters listed in Table 5.10. Figure 5.29 shows a correlation between predicted and
observed grain sizes after calibration of the model against two experimental data points.
(a)
Peak temperature, 0C
Thin plate welding (1D heat flow)
Prior austenite grain size (urn)
Relative volume fraction of pinning precipitates (f/y
(b)
Peak temperature, 0C
Fig. 5.28. HAZ grain growth diagrams for CrMo lowalloy steel; (a) Thick plate welding (2D heat
flow), (b) Thin plate welding (1D heat flow). No preheating (T0 = 200C).
Table 5.9 Chemical composition of type 316 austenitic stainless steel used by Ashby and Easterling24
(in wt%).
C
Mn
Si
Cr
Ni
Mo
0.05
2.0
1.0
19
11
2.3
0.045
0.030
Table 5.10 Data used to construct welding maps for type 316 austenitic stainless steel (compiled from
miscellaneous sources).
I Qapp.
I Atf*+ I Q/ I Dlm.
1
I K
o*
(kJmol )
(kJmol )
(kJ mol" )
(jim)
(fxm)
(1JLm2S1)
224
60
240
18
18
3.4 X IQ9
0.5
I ^*
(1Xm2S1)
2.1 X IQ14
!Estimated from data quoted by Kou32 for Cr23C6 in 304 stainless steel.
^Activation energy for diffusion of Cr in austenite.
Calibration point
Fig. 5.29. Comparison between measured and predicted HAZ austenite grain sizes after calibration of
model to data reported by Ashby and Easterling24 for type 316 stainless steel (simulated thick plate
welds).
It follows from the nomograms in Fig. 5.30(a) and (b) that the presence of Cr23C6 has little
influence on the HAZ grain coarsening behaviour because it dissolves at a fairly low temperature. This results in a rather coarse austenite grain structure adjacent to the fusion boundary,
which in certain cases may exceed 100 jim.
Peak temperature, 0C
Peak temperature, 0C
Fig. 5.30. HAZ grain growth diagrams for type 316 austenitic stainless steel; (a) Thick plate welding
(2D heat flow), (b) Thin plate welding (1D heat flow). No preheating (T0 = 200C).
Example (5.7)
Consider GTA butt welding of a 2mm thin sheet of type 316 austenitic stainless steel under the
following conditions:
Provided that the conditions for one dimensional (1D) heat flow are met, estimate on the
basis of the nomograms in Fig. 5.30(b) the variation in the austenite grain size across the HAZ
after welding. Estimate also the total width of the HAZ (referred to the fusion boundary)
under the prevailing circumstances.
Solution
First we calculate the net heat input per mm2 of the weld:
Readings from Fig. 5.30(b) give the HAZ austenite grain size profile shown in Fig. 5.31.
The total width of the grain growth zone can be estimated from equation (549) by using
data from Table 1.1 (Chapter 1). Taking the peak temperature for incipient dissolution of
Cr23C6 equal to 9000C, we obtain:
Peak temperature, 0C
Fig. 5.31. Predicted variation in austenite grain size across the HAZ of a 316 austenitic stainless steel
weld (Example 5.7).
D x 10 (arbitrary units)
(a)
D x 10 (arbitrary units)
(b)
D x 10 (arbitrary units)
(C)
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
(a)
(b)
(C)
(d)
Fig. 5.33. The evolution of 2D grain structures during normal grain growth; (a) Initial Voronoi structure,
485 grains, (b) 456 grains, (c) 349 grains, (d) 251 grains. Data from Saetre and Ryum.36
23.
24.
25.
26.
27.
28.
29.
30. RJ. Alberry and W.K.C. Jones: Metals Technology, 1977, 4, 4551; ibid., 360364; ibid., 557566.
31. R.M. Miranda and M.A. Fortes: Mater. ScL Eng., 1989, A108, 18.
32. S. Kou: Welding Metallurgy, 1987, Toronto, John Wiley & Sons.
33. V.E. Fradkov and D.G. Udler: in Simulation and Theory of Evolving Microstructures (Eds M.R
Anderson and A.D. Rollett), 1990, TMS, Warrendale, Pa, pp. 1529.
34. HJ. Frost, CV. Thompson and D.T. Walton: ibid., pp.3139.
35. K. Kawasaki, T.Nagai and S. Ohta: ibid., pp.6577.
36. T.O. Saetre and N. Ryum: in Modelling of Coarsening and Grain Growth (Eds C S . Pande and
S.R Marsh), 1993, The Minerals, Metals & Materials Society, pp.281294.
37. M.A. Fortes and A. Soares: ibid., pp. 257270.
38. A.D. Rollett, DJ. Srolovitz and M.R Anderson: Acta Metall, 1989, 37, 12271240.
39. G.S. Grest, M.R Anderson, DJ. Srolovitz and A.D. Rollett: Scripta Metall Mater., 1990, 24,
661665.
40. DJ. Srolovitz, G.S. Grest and M.R Anderson: Acta Metall, 1986, 34, 18331845; ibid, 1988,
36,21152128.
41. T.O. Sartre, O. Hunderi and E. Nes: Acta Metall, 1986, 34, 981 987.
42. K. Marthinsen, O. Lohne and E. Nes: Acta Metall, 1989, 37, 135145.
43. A.D. Rollett, MJ. Lutony and DJ. Srolovitz: Acta Metall. Mater., 1992, 40, 4355.
44. A. Kern, W. Reif and U. Schriever: Proc. Int. Conf. on Grain Coarsening, 1821 June, 1991,
Rome, Mater. ScL Forum, 1992, vols 9496, 709714, Trans. Tech Publications.
45. B. Radhakrishnan and T. Zacharia: Proc. Int. Conf. on Modeling and Control of Joining Processes, 810 Dec, 1993, Orlando, FL, American Welding Society.
Appendix 5.1
Nomenclature
thermal diffusivity (mm2 s"1)
amperage (A)
Zener coefficient
time (s or min)
temperature (0C or K)
another complex function of the impurity diffusivity and the interaction energy between the grain boundary and
the impurity atoms
arc efficiency
6
Solid State Transformations in Welds
6.1 Introduction
The majority of phase transformations occurring in the solid state take place by thermally
activated atomic movements. In welding we are particularly interested in transformations that
are induced by a change in temperature of an alloy with a fixed bulk composition. Such
transformations include precipitation reactions, eutectoid transformations, and massive transformations both in the weld metal and in the heat affected zone.
Since welding metallurgy is concerned with a number of different alloy systems (including
low and high alloy steels, aluminium alloys, titanium alloys etc.), it is not possible to cover all
aspects of transformation behaviour. Consequently, the aim of the present chapter is to provide the background material necessary for a verified quantitative understanding of phase
transformations in weldments in terms of models based on thermodynamics, kinetics, and
simple diffusion theory. These models will then be applied to specific alloy systems to illuminate the basic physical principles that underline the experimental observations and to predict
behaviour under conditions which have not yet been studied.
Stable (B)
Stable (a)
Temperature
Fig. 6.1. Schematic representation of the molar free energies of two solid a and P phases as a function of
temperature (allotrophic transformation no compositional change).
Temperature
reaction can proceed when the driving force AG becomes greater than the righthand side of
equation (61).
For an allotropic transformation, in which there is no compositional change, AG will be a
simple function of temperature, as illustrated in Fig.6.1. For alloys the situation is slightly
more complex, since there is an additional variable, i.e. the composition. In such cases the
temperature at which the aphase becomes thermodynamically unstable (Teq) corresponds to a
fixed point on the a(3 solvus boundary in the equilibrium phase diagram, as shown
schematically in Fig. 6.2. Since phase diagrams are available for many of the important industrial alloy systems, it means that the driving force for a transformation reaction can readily be
obtained from such diagrams in the form of a characteristic undercooling (AT).
%B
Fig. 6.2. Schematic representation of the a(3 solvus boundary in a simple binary phase diagram.
(62)
where v is a vibration frequency factor, Nv is the total number of heterogeneous nucleation
sites per unit volume, AG^, is the energy barrier against nucleation, and Qd is the activation
energy for atomic migration across the nucleus/matrix interface.
Grain boundary
Vacancy clusters
Dislocations/stacking faults
Inclusions
Grain
boundaries
Inclusion
Grain
corners
Free
surfaces
AG* /AG*
het.
horn.
Free
surface
Nucleation site
Fig. 6.3 Schematic diagram showing the most potent sites for heterogeneous nucleation in metals and
alloys.
It follows from the graphical representation of equation (62) in Fig. 6.4 that the nucleation
rate Nhet is highest at an intermediate temperature due to the competitive influence of
undercooling (driving force) and diffusivity on the reaction kinetics. This change in Nhet with
temperature gives rise to corresponding fluctuations in the transformation rate, as shown
schematically in Fig. 6.5. Note that the peak in transformation rate is due to two functions,
growth and nucleation (which peak at different T) whereas peak in Nhet is due to nucleation
only.
6.2.2.2 Determination
of AGhet and Qd
During the early stages of a precipitation reaction, the reaction rate may be controlled by the
nucleation rate Nhet. Under such conditions, the time taken to precipitate a certain fraction of
the new phase t* is inversely proportional to Nhet\
(63)
where C1 and C2 are kinetic constants.
By taking the natural logarithm on both sides of equation (63), we obtain:
(64)
If the complete Ccurve is known for a specific transformation reaction, it is possible to
evaluate AG*het and Qd from equation (64) according to the procedure described by Ryum.3
In general, a plot of In t* vs HT will yield a distorted Ccurve with welldefined asymptotes, as
shown in Fig. 6.6. At high undercoolings, when &G*het is negligible, the slope of the curve
becomes constant and equal to QdIR. The mathematical expression for this asymptote is:
T
Low undercooling
High diffusivity
High undercooling
Low diffusivity
%B
Nhet.
Fig. 6.4. Schematic diagram showing the competitive influence of undercooling (driving force) and
diffusivity on the heterogeneous nucleation rate.
Fraction transformed
logt
logt
Fig. 6.5. Fraction transformed as a function of time referred to the Ccurve (schematic).
(65)
At the chosen reference temperature Tr the time difference between the real Ccurve and
the extension of the lower asymptote amounts to (see Fig. 6.6):
(66)
from which
(67)
It follows that equations (65) and (67) provide a systematic basis for obtaining quantitative information about Qd and AGhet from experimental microstructure data through a simple
graphical analysis of the shape and position of the Ccurve in temperaturetime space.
1/r
lnt
Fig. 6.6. Determination of AG*het and Qd from the Ccurve (schematic).
where TV4 is the Avogadro constant, ^ n is the interfacial energy per unit area between the nucleus and the matrix, AGV is the driving force for the precipitation reaction (i.e. the volume free
energy change associated with the transformation), and 5(0) is the socalled shape factor which
takes into account the wetting conditions at the nucleus/substrate interface.
For a particular alloy, AGV is for small Ar proportional to the degree of undercooling:l
(69)
where C3 is a kinetic constant. This equation follows from the definition of AGv in diluted alloy
systems and the mathematical expression for the solvus boundary in the binary phase diagram.
By substituting equation (69) into equation (68), we get:
(610)
It follows that A0 is a characteristic material constant which is related to the potency of the
heterogeneous nucleation sites in the material. The value of A0 is, in turn, given by equations
(67) and (610):
(611)
(612)
Equation (612) can further be modified to allow for compositional and structural variations in the parent material by using the calibration procedure outlined in Fig. 6.7. Let tr
denote the time taken to precipitate a certain fraction of (3 at a chosen reference temperature
T= Tr in an alloy containing Nv nucleation sites per unit volume. If we take the corresponding
solvus temperature of the (3phase equal to T*q , the expression for t* becomes:
(613)
(614)
Equation (614) provides a basis for predicting the displacement of the Ccurve in temperaturetime space due to compositional or structural variations in the parent material. In genT
Ccun/e(Nv=N^)
logt
Fig. 6.7. Method for eliminating unknown kinetic constant in expression for t*.
eral, an increase in Nv will shift the nose of the Ccurve to the left in the diagram (i.e. towards
shorter times), as shown schematically in Fig. 6.8, because of the resulting increase in the
nucleation rate. Moreover, in solutedepleted alloys the critical undercooling for nucleation
will be reached at lower absolute temperatures where the diffusion is slower. This results in a
marked drop in Nhet, which displaces the Ccurve towards lower temperatures and longer times
in the ITdiagram, as indicated in Fig. 6.9.
Example (6.1)
Isothermal transformation (IT) or continuous cooling transformation (CCT) diagrams are available for many of the important alloy systems.4 In the case of aluminium, socalled temperatureproperty diagrams exist for different types of wrought alloys.45 Suppose that the Ccurve
in Fig. 6.10 conforms to incipient precipitation of [3'(Mg2Si) particles at manganesecontaining dispersoids in 6351 extrusions. Use this information to estimate the values of A0 and Qd in
equation (63) when the solvus temperature of (3'(Mg2Si) is 5200C.
Solution
The parameters A0 and Qd can be evaluated from the Ccurve according to the procedure
shown in Fig. 6.6. Referring to Fig. 6.11, the value of AG^ at the chosen reference temperature Tr = 35OC (623K) is equal to:
When AGhet is known, the parameter A0 can be obtained from equation (611):
T
iogt
Fig. 6.8. Effect of Nv on the shape and position of Ccurve in temperaturetime space (schematic).
logt
\e,
Fig. 6.9. Effect of solute content on the shape and position of Ccurve in temperaturetime space (schematic).
%B
Similarly, Qd can be read from Fig. 6.11 by considering the slope of the lower asymptote:
This value is in good agreement with the reported activation energy for diffusion of magnesium in aluminium.6
Temperature, 0C
AA 6351  T6
Time, s
Fig. 6.10. Ccurve for 99.5% maximum yield strength of an AA6351T6 extrusion. After Staley.5
103/T, K"1
lnt
Fig. 6.11. Determination of kG*heU and Qd from the Ccurve in Fig. 6.10 (Example 6.1).
Incoherent Interface
Coherent interface
(a)
(b)
a/p
Fig. 6J2. Schematic illustration of atom transfer across different kinds of interfaces; (a) Incoherent
interface, (b) Coherent interface.
(616)
Temperature
Liquid
Concentration
%B
Diffusion of solute
Distance
Fig. 6.14. Schematic representation of concentration profile ahead of advancing interface during precipitation of (B from a supersaturated aphase.
Temperature
Concentration
%B
Diffusion of solute
Distance
Fig. 6.15. Schematic representation of concentration profile ahead of advancing interface during growth
of solutedepleted P into a metastable aphase.
(618)
where Dm is the diffusivity of the solute in the matrix, and erfc(u) is the complementary error
function (defined previously in Appendix 1.3, Chapter 1).
Similarly, for growth of spherical precipitates, the variation in the radius r with time can be
written as:8
(619)
where e 2 *s t n e corresponding parabolic thickening constant for a spherical geometry, defined
as:
(620)
The parabolic relations in equations (617) and (619) imply that the growth rate slows
down as the (3phase grows. This is due to the fact that the total amount of solute partitioned
during growth decreases with time when the diffusion distance increases. Moreover, the form
of equations (618) and (620) suggests that the maximum in the growth rate is achieved at an
intermediate temperature because of the competitive influence of undercooling (driving force)
and diffusivity on the reaction kinetics. Consequently, a plot of E1 or 2 vs temperature will
reveal a pattern similar to that shown in Fig. 6.4 for the nucleation rate, although the thickening constants generally are less temperaturesensitive.
In addition to the models presented above for plates and spheres, approximate solutions
also exist in the literature for thickening of needleshaped precipitates, based on the Trivedi
theory for diffusioncontrolled growth of parabolic cylinders.9 However, because of space limitations, these solutions will not be considered here.
6.2.4 Overall transformation kinetics
The progress of an isothermal phase transformation may be conveniently represented by an ITdiagram of the type shown in Fig. 6.5. Among the factors that determine the shape and position of the Ccurve are the nucleation rate, the growth rate, the density and the distribution of
the nucleation sites as well as the physical impingement of adjacent transformed volumes.
Due to the lack of adequate kinetic models for diffusioncontrolled precipitation, we shall
assume that the overall microstructural evolution with time can be described by an Avramitype of equation:10
(621)
where X is the fraction transformed, n is a time exponent, and k is a kinetic constant which
depends on the nucleation and growth rates.
The exponential growth law summarised in the Avrami equation is valid for linear growth
under most circumstances, and approximately valid for the early stages of diffusioncontrolled
growth.10 Table 6.1 gives information about the value of the time exponent for different experimental conditions.
In general, the value of n will not be constant, but change due to transient effects until the
steadystate nucleation rate is reached and n attains its maximum value. Subsequently, the
nucleation rate starts to decrease as the sites become filled with nuclei and eventually approach zero when complete saturation occurs. This is because the heterogeneous nucleation
sites are not randomly distributed in the volume, but are concentrated near other nucleation
sites leading to an overall reduction in n. From then on, the transformation rate is solely
controlled by the growth rate.
6.2.4.1 Constant nucleation and growth rates
For a specific transformation reaction, the value of k in equation (621) can be estimated from
Table 6.1 Values of the time exponent n in the Avrami equation. After Christian.10
Polymorphic changes, discontinuous precipitation, eutectoid reactions, interface
controlled growth, etc.
Increasing nucleation rate
Constant nucleation rate
Decreasing nucleation rate
Zero nucleation rate (saturation of point sites)
Grain edge nucleation after saturation
Grain boundary nucleation after saturation
Diffusion controlled growth
All shapes growing from small dimensions, increasing nucleation rate
All shapes growing from small dimensions, constant nucleation rate
All shapes growing from small dimensions, decreasing nucleation rate
All shapes growing from small dimensions, zero nucleation rate
Growth of particles of appreciable initial volume
Needles and plates of finite long dimensions, small in comparison with their separation
Thickening of long cylinders (needles) (e.g. after complete end impingement)
Thickening of very large plates (e.g. after complete edge impingement)
Precipitation on dislocations (very early stages)
kinetic theory, using the classic models of nucleation and growth described in the previous
sections. In practice, however, this is a rather cumbersome method, particularly if the base
metal is of a heterogeneous chemical nature. Alternatively, we can calibrate the Avrami equation against experimental microstructure data, e.g. obtained from generic ITdiagrams. A convenient basis for such a calibration is to write equation (621) in a more general form:
(622)
where k* is a new kinetic constant (equal to kr1/n). In the latter equation the parameter k* can be
regarded as a time constant, which is characteristic of the system under consideration. Note
that this form of the Avrami equation is mathematically more appropriate, as the dimensions of
the k* constant are not influenced by the value of the time exponent n.
During the early stages of a transformation reaction, the reaction rate is controlled by the
nucleation rate. Let f denote the time taken to precipitate a certain fraction of P (X = Xc) at
an arbitrary temperature T (previously defined in equation (614)). It follows from equation
(622) that the value of k* in this case is given as:
(623)
A combination of equations (622) and (623) then gives:
(624)
from which
(625)
Equation (625) represents an alternative mathematical description of the Avrami equation,
and is valid as long as the nucleation and growth rates do not change during the transformation. It has therefore the following limiting values and characteristics: X=O when t = 0, X =
Xc when t = t*, and X>1 when r> <*>.
6.2.4.2 Site saturation
If the nucleation rate is considered to be zero by assuming early site saturation, the subsequent
phase transformation simply involves the reconstructive thickening of the player. In the onedimensional case, the process can be modelled in terms of the normal migration of a planar
a/p interface, as shown schematically in Fig. 6.16. Let Aa/^ denote the interfacial area between a and (3 per unit volume and Ua/^ the growth rate of the incoherent a/pinterface. From
Fig. 6.16 we see that the volume fraction of the transformed (3phase is given as:
(626)
By using the standard JohnsonMehl correction for physical impingement of adjacent transformation volumes, we may write in the general case:
(627)
which after integration yields:
(628)
This specific form of the Avrami equation is valid under conditions of early site saturation
where the a/pinterface is completely covered by P nuclei at the onset of the transformation.
6.2.5 Nonisothermal transformations
So far, we have assumed that the phase transformations occur isothermally. This is, of course,
a rather unrealistic assumption in the case of welding where the temperature varies continuously with time. From the large volume of literature dealing with solid state transformations in
Fig. 6.16, Schematic illustration of the planar geometry assumed in the site saturation model.
metals and alloys, it appears that the bulk of the research has been concentrated on modelling
of microstructural changes under predominantly isothermal conditions.1"411 In contrast, only
a limited number of investigations has been directed towards nonisothermal transformations. 51012 " 18 However, these studies have clearly demonstrated the advantage of using analytical modelling techniques to describe the microstructural evolution during continuous cooling, instead of relying solely on empirical CCTdiagrams.
6.2.5.1 The principles of additivity
From the literature reviewed it appears that there is considerable confusion regarding the application of isothermal transformation theory for prediction of nonisothermal transformation
behaviour. These difficulties are mainly due to the independent variations of the nucleation
and growth rate with temperature. In fact, it can be shown on theoretical grounds that the
problem is only tractable when the instantaneous transformation rate is a unique function of
the fraction transformed and the temperature.10 This leads to the additivity criterion described
below.
The principles of additivity are based on the theory advanced by Scheil.12 He proposed that
the start of a transformation under nonisothermal conditions could be predicted by calculating
the consumption of fractional incubation time at each isothermal temperature, with the transformation starting when the sum is equal to unity. The Scheil theory has later been extended to
phase transformations to predict continuous cooling transformation kinetics from isothermal
microstructure data.10'17'18
Let t* again denote the time taken to precipitate a certain fraction of P (X  Xc) at an arbitrary temperature T. If the reaction is additive, the total time to reach Xc under continuous
cooling conditions is obtained by adding the fractions of time to reach this stage isothermally
until the sum is equal to unity. Noting that t* varies with temperature, we may write in the
general case:
(629)
Cooling curve
logt
Fig. 6.17. Schematic illustration of the Scheil theory.
2 1
Nucleation rate (N*het), cm s
Temperature, 0C
Fig. 6.18. Predicted variation in N*het and E1 with temperature during the austenite to ferrite transformation in a CMn steel (0.15 wt% C, 0.40 wt% Mn). Data from Umemoto et al.19
(633)
where dt is the time increment at T, and f is the corresponding hold time required to reach Xc
at the same temperature (given by equation (614)). The derivation of equation (632) is shown
in Appendix 6.2
The principles of additivity are also applicable under conditions of early site saturation. If
only U^p varies with temperature, it is possible to rewrite equation (628) in an integral form:
(634)
This equation can readily be integrated by numerical methods when the temperaturetime
programme is known.
6.2.5.4 Nonadditive reactions
If the additivity condition is not satisfied, it means that the fraction transformed is dependent
on the thermal path, and the differential equation has no general solution. This, in turn, implies
that the Ccurve concept breaks down and cannot be applied to nonisothermal transforma
Next Page
tions. Solution of the differential equation then requires stepwise integration in temperaturetime space, using an appropriate numerical integration procedure. As already pointed
out, this will generally be the case for diffusioncontrolled precipitation reactions, since the
evolution parameter X is a true function of temperature. Under such conditions, experimentally based continuous cooling transformation (CCT) diagrams must be employed.
The microstructural constituents listed above are indicated in Fig. 6.19, which shows
photomicrographs of typical regions within lowalloy steel weldments.
6.3.2 Currently used nomenclature
Quantification of microstructures in steel welds is most commonly done by means of optical
microscopy. Several systems have been introduced throughout the years for the classification
of the various constituents, with each system reflecting different investigator's views and
Previous Page
tions. Solution of the differential equation then requires stepwise integration in temperaturetime space, using an appropriate numerical integration procedure. As already pointed
out, this will generally be the case for diffusioncontrolled precipitation reactions, since the
evolution parameter X is a true function of temperature. Under such conditions, experimentally based continuous cooling transformation (CCT) diagrams must be employed.
The microstructural constituents listed above are indicated in Fig. 6.19, which shows
photomicrographs of typical regions within lowalloy steel weldments.
6.3.2 Currently used nomenclature
Quantification of microstructures in steel welds is most commonly done by means of optical
microscopy. Several systems have been introduced throughout the years for the classification
of the various constituents, with each system reflecting different investigator's views and
(a)
(b)
(C)
(d)
Fig. 6.19. Optical micrographs showing various microstructural constituents commonly found in lowalloy steel weldments; (a) Coarse grained HAZ (low heat input welding), (b) Coarse grained HAZ (high
heat input welding), (c) Asdeposited weld metal (low heat input welding), (d) Reheated weld metal (low
heat input welding). Letters in micrographs are defined in the text.
discretions. This controversy in terminology has been a source of confusion, and the work by
SubCommission IXJ of the International Institute of Welding (HW)23 for developing guidelines for quantification of microstructures is, therefore, an important step towards a standardised system of nomenclature.
The IIW recommendations are based on the scheme originally proposed by Abson and
Dolby.23 The IIW system involves a simplified classification procedure compared with the
outline used in Fig. 6.19, since the distinction between acicular ferrite and the various sideplate
structures is based on features such as aspect ratio, relative lath size, and number of parallel
laths. This has led to the introduction of the FSconstituent (ferrite with aligned second phase),
which, in principle, comprises both Widmanstatten ferrite and upper bainite.
In contrast to the HW approach to classifying microstructural elements based on their appearance in the optical microscope, other investigators rank the various constituents solely in
terms of their transformation behaviour, according to the scheme originally proposed by Dube
et ai26 From a scientific point of view, this classification system is more correct, since it does
not violate common terminology based on thermodynamics and kinetics of transformation
reactions. However, with the omission of the FSconstituent grouping utilised by the HW, the
Dube system is more inconvenient to use in practice because the different transformation products often cannot readily be identified on the basis of their transformation characteristics. Consequently, both classification systems appear to have their weaknesses, which, in turn, limit
their applicability.
6.3.3 Grain boundary ferrite
Grain boundary (or allotriomorphic) ferrite is the first phase to form on cooling below the Ae3temperature. It nucleates preferentially at austenite grain corners and boundaries, since these
sites generally provide the lowest energy barrier against nucleation (see Fig. 6.3). The fundamental aspects of grain boundary ferrite have been reviewed in detail by Bhadeshia,26 where
many of the original references can also be found.
6.3.3.1 Crystallography of grain boundary ferrite
The grain boundary ferrite allotriomorphs nucleate having a KurdjumowSachs type orientation relationship with one of the austenite grains:20
This orientation relationship, which lies within the socalled Bain orientation region,27 is
adopted in order to minimise the increase in the strain energy resulting from lattice distortion
AGEby formation of a lowenergy interface between the ferrite nucleus and the parent austenite
phase.1 Subsequent growth of the ferrite may then occur into the adjacent austenite grain with
which the ferrite has a random orientation relationship,28 since a disordered (incoherent) interface generally has a higher mobility than an ordered (coherent/semicoherent) interface at low
undercoolings.
6.3.3.2 Nucleation of grain boundary ferrite
As shown in Fig. 6.20, allotriomorphic ferrite can nucleate both at grain corners, grain edges
and grain faces, the former ones being the most potent sites for ferrite nucleation.1 If we
assume that ferrite nucleation occurs preferentially at austenite grain faces, the A^_/ Nv term
in equation (614) may be replaced by the inverse grain size ratio Dy /Dy, where Dy and Dy
refer to the austenite grain size in the actual and the reference material, respectively. This
leads to the following expression for t*:
(635)
Edge
Corner
Fig. 6,20. Sketch of an austenite grain showing different sites for ferrite nucleation.
Based on equation (635) it is possible to predict the displacement of the Ccurve in temperaturetime space due to structural or compositional variations in the parent material. As an
illustration, we shall assume that the parameters listed in Table 6.2 are representative of nucleation of grain boundary ferrite in a lowalloy steel with an initial austenite grain size of 10
jim. In addition, we need information about the Ae3temperature in the equilibrium phase
diagram. This temperature can readily be obtained from thermodynamic calculations, even for
multicomponent systems.29 Alternatively, we can use the empirical relationship quoted by
Leslie:30
(636)
Parameter I T
Value
(K)
(s)
T*eq I
(K)
T0
(Jmol"1)
823
0.6
1108
700
Q]
I Dy
(kJmol1)
(jim)
135
10
Example (6.2)
Consider SA welding on a thick plate of a Nbmicroalloyed steel under the following conditions:
Evaluate on the basis of the grain growth diagram in Fig. 5.21 (a) (Chapter 5) and equations
(632), (633), and (635) the conditions for ferrite formation at two different positions within
the HAZ corresponding to a peak temperature Tpof 13500C and 10000C, respectively. Assume in these calculations that the ferrite may form within the temperature range from 800 to
6000C, and that the equilibrium volume fraction of ferrite (f^qFe) in the fully transformed
steel is 0.9.
Solution
First we calculate the net heat input per unit length of the weld:
Readings from Fig. 5.21 (a) give the following HAZ grain sizes:
As expected, the theoretical Ccurves in Fig. 6.21 reveal a strong effect of the austenite
grain size on the HAZ transformation kinetics. By considering the superimposed weld cooling curve, it is possible to estimate the volume fraction of ferrite/""7^ which forms in each case
from equations (632) and (633). Taking the time exponent n in the Avrami equation equal
to 5/2 for nucleation of ferrite at austenite grain boundaries10 and 1 Xc = 0.98, we obtain:
Grain refined region (Dy = 10 jlm):
Temperature, 0C
Time, s
Fig. 6.21. Effect of austenite grain size on the HAZ transformation kinetics (Example 6.2). The superimposed cooling curve corresponds to a cooling time, A%5, of 21s.
and
and
From this we see that polygonal ferrite dominates the microstructure within the grain refined region, whereas ferrite hardly forms within the grain coarsened HAZ under the prevailing circumstances. Although experimental data are not available for a direct comparison, the
predicted effect of the austenite grain size on the HAZ transformation kinetics is reasonable
and consistent with general experience (e.g. see experimental CCTdiagrams in Fig. 6.22).
Effect of austenite and ferrite stabilising elements
In practice, the transformation behaviour of steel weldments is affected both by the prior
austenite grain size and by alloying additions. In welding metallurgy, the combined effects are
often discussed by considering their influence on hardenability. Broadly speaking, additions of
hardenability elements may serve two purposes:
(i)
(ii)
To control the microstructure through modification of the nucleation and growth rates of
proeutectoid ferrite.
Temperature, 0C
Time, s
Fig. 6.22. CCTdiagrams for a lowcarbon CuNi containing steel. Superimposed on the CCTdiagrams
are two cooling curves corresponding to Af875 of 10 and 100 s, respectively. Austenitising conditions;
Heavy solid lines: 9000C for 5 min, Heavy broken lines: 13000C for 5 s. Data from Cross et al?x
In the latter case the effect is related to a shift in the Ae3temperature of the steel, which
alters the undercooling and hence, the driving force for the austenite to ferrite transformation.
This point is illustrated by the following example.
Example (6.3)
By substituting data from Table 6.2 into equation (635), we arrive at the following expression
for r *50:
It is evident from the graphical representation of the above equation in Fig. 6.23 that an increase in the v4e3temperature (e.g. from 1108 to 1136K) displaces the Ccurve towards higher
temperatures and shorter times in the ITdiagram. This, in turn, gives rise to improved conditions for ferrite nucleation. Taking the time exponent n in the Avrami equation equal to 5/2
and (1 Xc) = 0.98 as in the previous example, we obtain after integration:
Temperature, 0C
Time, s
Fig. 6.23.Effect of steel chemical composition (Ae3temperature) on the HAZ transformation kinetics
(Example 6.3). The superimposed cooling curve corresponds to a cooling time, Ar875, of 21s.
and
Although the calculated value offa'Fe is rather uncertain, the trends predicted in the present
example are reasonable and consistent with general experience.
During welding, quantitative information about the extent of boron segregation which occurs to the austenite grain boundaries under various thermal programmes can be obtained on
the basis of a well established theoretical model for quenchinduced segregation of boron in
steel.3839 At peak temperatures above 1000 to 11000C, borocarbides and nitrides present in
the base plate will rapidly dissolve in the matrix,4041 leading to a significant increase in the
amount of free diffusible boron. Generally, solute atoms in a crystal lattice will have an associated strain energy,38 which implies that it is energetically feasible to pair the solute boron
atom with a vacancy. Since the formation of vacancies is a thermally activated process, it follows that the fraction of boron occupying such sites, [B]v, increases exponentially with temperature:39
(637)
Here m contains various geometric and entropy terms, Zy is the vacancy formation energy,
Eb is the vacancyboron binding energy, and [B] is the bulk concentration of free boron. By
substituting reasonable average values for ra, Ep and Eb into equation (637), we arrive at the
following expression for [B]V:39A2
(638)
when Tp Ae3.
In addition to the diffusible boron content, the grain size is also an important variable in
steel hardenability. A quantitative estimate of the combined effect of boron segregations and
austenite grain size on the HAZ transformation kinetics can be obtained by assuming that the
ferrite nucleates primarily on grain faces. In such cases the total number of heterogeneous
nucleation sites per unit volume Nvis given as:
(640)
where na is the number of nucleation sites per unit grain boundary area, and Sv is the grain
boundary surface area per unit volume (equal to 2/Dy).
If we, as a second approximation, assume that na is inversely proportional to the amount of
boron which diffuses to the grain boundaries on cooling, the N* / Nv ratio in equation (614)
can be written as:
(641)
Equation (641) predicts that the position of the ferrite Ccurve in temperaturetime space
depends on the l[B]gb Z)7) I l[B]*gb Z>y J ratio, as shown schematically in Fig. 6.24. Consequently,
this ratio can be regarded as a measure of the HAZ hardenability during welding of boroncontaining steels.
In Fig. 6.25 the microstructure data of Akselsen et al.42 have been replotted vs the
hardenability parameter [B]gbDy, taking the product [B]*gbDy in the reference steel equal to
unity for a direct comparison between theory and experiments. It is evident from the graph
that the HAZ martensite content of the two boroncontaining steels can be represented by one
single curve under the prevailing circumstances. This result is to be expected if the displacement of the ferrite Ccurve in temperaturetime space is determined by a relationship of the
type shown in equation (641).
Temperature
Cooling curve
log time
Fig. 6.24. Effect of boron alloying on the shape and position of ferrite Ccurve in temperaturetime space
according to the site blocking mechanism (schematic).
Martensite, vol%
Consider GMA welding on a thick plate of a boroncontaining steel under the following conditions:
/=300A, U = 30V, v = 4mm s1, TI = 0.8, T0 = 200C
Suppose that the free (diffusible) boron content of the steel at elevated temperatures is 40
ppm. Estimate on the basis of the theory outlined in the previous sections the conditions for
ferrite/martensite formation in the grain refined HAZ (Tp = 11000C) when the austenite grain
size is 15Lim. In these calculations we shall assume that the [B]*gb Dy product in the reference
steel (with thermodynamic properties as in Table 6.2) is close to unity.
Solution
First we estimate the amount of boron which segregates to the austenite grain boundaries
during cooling from equation (639). When Tp = 11000C (1373K), we get:
This gives the following value of the Af* / Nv ratio (equation 641):
The resulting displacement of the ferrite Ccurve can now be calculated from equation (614), using input data from Table 6.2:
As expected, the theoretical Ccurves in Fig. 6.26 reveal a strong effect of boron alloying
on the HAZ transformation kinetics. By considering the superimposed weld cooling curve, it
is possible to estimate the volume fraction of ferrite f^~Fe which forms from equations (632)
and (633). Taking the time exponent n in the Avrami equation equal to 5/2 for nucleation of
ferrite at austenite grain boundaries and (1  Xc) = 0.98 as in the previous examples, we obtain after integration within the temperature range from 800 to 600 C:
and
If the same calculations are performed for the reference steel in Table 6.2 (characterised by
and
From this we see that boron, even in small quantities, can have a dramatic effect on the
HAZ transformation kinetics by promoting the formation of bainite and martensite at the expense of grain boundary ferrite. This is in good agreement with general experience (see experimental CCTdiagrams in Fig. 6.27).
Effect of solidificationinduced segregation
As shown in Chapter 3 of this textbook, the characteristic solidification pattern of mild and
lowalloy steel weld metals leads to extensive segregation of alloying and impurity elements
to grain boundaries and interdendritic spaces. Of particular interest in this respect is phosphorus segregations at columnar austenite grain boundaries. Referring to Figs. 3.38 and 3.39
(Chapter 3), the phosphorusrich zone is seen to extend about 7jam on either side of the grain
boundaries. Since phosphorus is among the strongest ferrite stabilising elements in steel (see
equation (636)), the existence of such solidificationinduced segregations would be expected
Temperature, 0C
Time, s
Fig. 6.26. Predicted displacement of ferrite Ccurve in temperaturetime space due to segregation of
boron to prior austenite grain boundaries (Example 6.4). The superimposed cooling curve corresponds
to a cooling time, Ats/5, of 9.2s.
to enhance the nucleation rate of grain boundary ferrite at these sites due to the associated
increase in the Ae3temperature. The above phenomenon should not be confused with equilibrium segregation of phosphorus to austenite grain boundaries during heat treatment of steel,
which stems form attraction of the atoms towards the open structure of the boundary. In the
latter case phosphorus may act as a hardenability element by occupying favourable sites for
ferrite nucleation at the austenite grain boundaries analogous to that documented for boron in
steel.43'44
Example (6.5)
Suppose that the local phosphorus content adjacent to the columnar austenite grain boundaries
in a lowalloy steel weld metal is 500 ppm, whereas the bulk concentration of phosphorus is
100 ppm. Estimate on the basis of the theory developed in the previous sections the resulting
displacement of the ferrite Ccurve in temperaturetime space when the columnar austenite
grain size is 80jun. In these calculations we shall assume that the transformation characteristic
of the bulk metal is similar to that of the reference steel in Table 6.2.
Solution
First we estimate the actual Ae3temperature within the phosphorusrich region adjacent to the
columnar austenite grain boundaries from equation (636). Taking the A<?3temperature of the
bulk phase equal to 835C (1108K), the local phase boundary temperature becomes:
By substituting data from Table 6.2 into equation (635), we arrive at the following expression for ^ 0 :
(a)
Temperature, 0C
Temperature, 0C
It is evident from the graphical representation of the above equation in Fig. 6.28 that the
observed increase in the yl^temperature from 1108 to 1136K displaces the ferrite Ccurve
towards higher temperatures and shorter times in the ITdiagram. The resulting effect on the
weld metal transformation kinetics is obvious, since an increase in the nucleation rate of ferrite will favour early site saturation at the austenite grain boundaries. On this basis it is not
surprising to find that allotriomorphic ferrite in lowalloy steel weld metals tends to form continuous veins of blocky ferrite along the columnar austenite grain boundaries, as shown in Fig.
6.19(c). Outside the fusion zone the conditions for early site saturation are less favourable,
since modern steelmaking practice implies that solidificationinduced segregations are removed by prolonged hightemperature annealing prior to the welding operation. Hence, ferrite veining of the type shown in Fig. 6.19(c) is not commonly observed within the reheated
regions of the base plate, unless the heat input is extremely large (see Fig. 6.19(a) and (b)).
Temperature, 0C
Time, s
Fig. 6.28. Effect of solidificationinduced phosphorus segregations on the austenite to ferrite transformation in lowalloy steel weld metals (Example 6.5).
Equation (643)
Consider multipass welding with covered electrodes (SMAW) on a thick plate of lowalloy
steel under the following conditions:
Previous experience has shown that polygonal ferrite forms within the lowtemperature
reheated regions of the weld metal, typically 1.0 to 1.7 mm beneath the surface (fusion boundary) of subsequent weld passes. Estimate on the basis of the grain growth diagram in Fig.
5.24(a) (Chapter 5) and equation (643) the maximum variation in the ferrite grain size across
the weld HAZ under the prevailing circumstances. Thermal data for lowalloy steels are given
in Table 1.1 (Chapter 1).
Solution
First we need to convert the depths at which polygonal ferrite appears beneath the fusion
boundary to an equivalent (characteristic) peak temperature range. If we neglect the contribution from heat flow in the welding direction, this conversion can be done on the basis of
equation (547) in Chapter 5:
and
The prior austenite grain size Dy at these two locations can now be read from Fig. 5.24(a)
and Table 5.6, respectively:
Since the cooling rate (CR.) at a given temperature is essentially the same across the weld
HAZ, the maximum variation in the ferrite grain size can be evaluated directly from equation
(643) without further background information:
From this we see that the variation in the ferrite grain size is significantly smaller than the
corresponding change in the prior austenite grain size. This result is in good agreement with
general experience.
6.3.3.3 Growth of grain boundary ferrite
If the austenite grain boundaries become rapidly decorated with a continuous layer of ferrite
(so that the subsequent transformation involves the reconstructive thickening of these layers),
the evolution of allotriomorphic ferrite is determined solely by its growth kinetics.37 As shown
in Example 6.5, this is a realistic assumption in the case of asdeposited steel weld metals,
where the presence of solidificationinduced phosphorus segregations at the columnar austenite
grain boundaries favours an early site saturation.
Several investigators have modelled the evolution of allotriomorphic ferrite in lowalloy
steel weld metals along the lines indicated above.46"48 The most thorough analysis is probably
that of Bhadeshia et al.46 who were able to account for the combined effect of temperature and
steel chemical composition on the growth kinetics. However, in order to illustrate the competition between the different variables that contribute to the formation of grain boundary ferrite
in asdeposited steel weld metals, the simplified treatment of Liu and Olson47 and Fleck et al.4S
has been adopted here.
As a starting point the Avrami equation17 is considered for the limiting case of zero nucleation rate by assuming early site saturation. If subsequent growth of the grain boundary ferrite
allotriomorphs occurs in both directions perpendicular to the austenite grain boundaries at a
rate which is controlled by diffusion of carbon in austenite, the time dependence of the ferrite
thickness, AZa, is given by equation (617):
(644)
where dy and I1 are the diameter and length of the inscribed cylinder, respectively.
By substituting the above expressions for Ua and Sv into equation (634), we obtain:
(647)
Fig. 6.30. Hexagonal prism model for the columnar austenite grain morphology in lowalloy steel weld
metals. The inscribed cylinder in the figure has approximately the same surface to volume ratio as the
hexagonal prisms.
Temperature, 0C
Fig. 6.31. Effect of temperature and steel chemical composition on the onedimensional parabolic thickening constant for allotriomorphic ferrite. (1): 0.03 wt% C, (2): 0.06 wt% C, (3): 0.08 wt% C and (4):
0.10 wt% C. Data from Bhadeshia et al.46
(648)
from which
(649)
Equation (649) predicts that the volume fraction of grain boundary ferrite in the asdeposited weld metal depends on the combined action of the following three main variables:
(i) The onedimensional parabolic thickening constant ei which is determined by the weld
metal chemical composition (i.e. the content of austenite and ferrite stabilising elements).
(ii) The columnar austenite grain size dy which is controlled by the weld metal solidification
microstructure (i.e. the weld metal chemistry, the weld pool geometry, and the thermal
conditions under which solidification occurs).
(iii) The retention time within the critical transformation temperature range for allotriomorphic
ferrite, as determined by the applied heat input and the mode of heat flow (i.e. thick plate,
medium thick plate, or thin plate welding, respectively).
It can be seen from the microstructure data of Grong et al.50 reproduced in Fig. 6.32 that the
influence of these variables are adequately accounted for in the present model. However, the
calculated volume fractions of grain boundary ferrite are consistently lower than the measured
ones. This discrepancy can probably be attributed to the use of constant values for the onedimensional parabolic thickening constant. Consequently, if proper corrections are made for
the inherent variation in E1 with temperature and steel chemical composition, the agreement
between theory and experiments is significantly improved, as shown by the data of Bhadeshia
et al.46 reproduced in Fig. 6.33.
Example (6.7)
Consider GMA welding on a thick plate of a lowalloy steel under the following conditions:
Suppose that the grain boundary allotriomorphs form within the temperature range from
750 to 6000C at a constant rate ei of 3 jam s~1/2. Estimate on the basis of the theory developed
in the previous section the volume fraction of allotriomorphic ferrite in the weld deposit when
the columnar austenite grain size is 80Lim. In these calculations we shall assume that the
equilibrium volume fraction offerritefq~Fe in the fully transformed steel is 0.9. Thermal data
for lowalloy steels are given in Table 1.1 (Chapter 1).
Solution
The situation is described in Fig. 6.34. In this case the problem is to estimate the retention time
/Str within the critical temperature range for ferrite formation from the Rosenthal thick plate
solution (Chapter 1). From equation (166), we have:
Volume fraction of GF
Regression line
Fig. 6.32. Experimental verification of equation (649). Data from Grong et al.5Q
Calculated volume
fraction of GF
Temperature
Fig. 6.33. Comparison between measured and predicted volume fractions of grain boundary ferrite in
CMn steel weld metals. Data from Bhadeshia et al.46
log time
Fig. 6.34. Conditions for allotriomorphic ferrite formation in lowalloy steel weld metals (Example 6.6).
Although experimental data are not available for a direct comparison, the calculated value
of fa'Fe is reasonable and of the expected order of magnitude.
6.3.4 Widmanstdtten ferrite
With increasing degree of undercooling the redistribution of carbon becomes insufficient to
maintain a planar growth mode, and hence, further growth of the ferrite can only take place by
lateral movement of ledges along a lowenergy interface*.120 This, in turn, implies a
KurdjumowSachstype orientation relationship between the austenite and the ferrite, i.e.
{Ill }1_Fe parallel with {110}a_F6, and <110> 7 .^ parallel with <11 l>a_Fe, which is a characteristic feature of the Widmanstatten ferrite structure. The ferrite sideplates, once nucleated,
grow very rapidly under the prevailing conditions because of an efficient redistribution of
carbon to the sides of the advancing interface.1 As a result, parallel arrays of ferrite laths of
high aspect ratios (typically 10:1 to 20:1) are often found in the areas adjacent to the austenite
grain boundaries, as shown by the optical micrographs in Fig. 6.19(b) and (c).
Following the treatment of Bhadeshia et al.,46 the lengthening rate of Widmanstatten ferrite
can be estimated using the Trivedi theory9 for diffusioncontrolled growth of parabolic cylinders with correction for the assumed displacive character of the Widmanstatten ferrite transformation. As shown in Fig. 6.35, the pertinent growth rates are normally so large that the
formation of Widmanstatten ferrite is essentially complete within a fraction of a second. This
implies that the transformation, for all practical purposes, can be treated as being isothermal.37
When the growth rate is known, the volume fraction of Widmanstatten ferrite in the asdeposited weld metal may be estimated by assuming that the ferrite sideplates nucleate at a
constant rate at the yFelaFe boundaries and subsequently grow into the interior of the columnar austenite grains until they physically impinge with intragranularly nucleated acicular ferrite or allotriomorphic ferrite.46 It is seen from the microstructure data of Bhadeshia et al.46
reproduced in Fig. 6.36 that the calculated volume fractions of Widmanstatten ferrite are in
reasonable agreement with experiments, although the discrepancy in certain cases is admittedly large.
*A different view is suggested by Bhadeshia et al.31'46'51'52 who claim that growth of Widmanstatten ferrite occurs in
a displacive manner analogous to that documented for martensite in steel, with the exception that carbon must diffuse
during growth.
Next Page
Temperature, 0C
Fig. 6.35. Predicted growth rates of Widmanstatten ferrite in CMn steel weld metals; (a) Growth rate
calculations for weld compositions listed in Fig. 6.31, (b) Growth rate calculations after modifying compositions to allow for carbon enrichment due to grain boundary ferrite formation. Data from Bhadeshia
etai46
Previous Page
Temperature, 0C
Fig. 6.35. Predicted growth rates of Widmanstatten ferrite in CMn steel weld metals; (a) Growth rate
calculations for weld compositions listed in Fig. 6.31, (b) Growth rate calculations after modifying compositions to allow for carbon enrichment due to grain boundary ferrite formation. Data from Bhadeshia
etai46
Calculated volume
fraction of WF
Fig. 6.36. Comparison between measured and predicted volume fractions of Widmanstatten ferrite in CMn steel weld metals. Data from Bhadeshia et al.46
(a)
(b)
Fig. 6.37. The development of transformation textures in asdeposited steel weld metals containing acicular
ferrite; (a) (200) pole figure showing the crystallographic orientations of acicular ferrite referred to the
original cell/dendrite growth direction, (b) Backscattered electron channeling contrast image of delta
ferrite/austenite columnar grain (the metallographic section is normal to the cell/dendrite growth direction). After Kluken et al.63
parallel with the cell/dendrite growth direction (Fig. 6.38(a)). The other plates have a <111>
direction aligned in the same crystal growth direction (Fig. 6.38(b)). These data can be represented by two subcomponents which are displaced with respect to each other by a 60 rotation
about a common <111> axis.
It is evident from the measurements of Kluken et al.63 that the acicular ferrite plates in asdeposited steel weld metals exhibit an orientation relationship with both the austenite and the
prior delta ferrite columnar grains in which they grow. This 'memory' effect arises from the
characteristic solidification pattern and transformation behaviour of lowalloy steel welds (e.g.
see discussion in Section 3.8.2, Chapter 3). As shown schematically in Fig. 6.39, the columnar
grain region will exhibit a sharp <100> solidification texture which has its origin in the phenomenon of preferred crystal growth. At the onset of the peritectic reaction, the austenite
adopts a KS type of orientation relationship with the delta ferrite in order to minimise the
energy barrier against nucleation.64 The austenite subsequently grows around the periphery of
the primary phase until impingement occurs on neighbouring columnar grain boundaries. During
the yFe to aFe transformation, this memory effect gives rise to the formation of acicular ferrite
plates which have a <100> direction approximately parallel with the original cell/dendrite
growth direction. The presence of the two other texture components within the weld metal is
thus a result of complementary crystal rotations taking place within the same orientation region.
The proposed sequence of reactions is in excellent agreement with the texture analysis of
Hu65 who made theoretical calculations of the resulting orientations of iron after Fe to yFe and
yFe to aFe transformations in succession according to the scheme outlined in Fig. 6.39.
6.3.5.3 Nature of acicular ferrite
Currently, the mechanisms of acicular ferrite formation in lowalloy steel weld metals are not
(a)
(b)
Fig. 6.38. Schematic diagrams showing the three main texture components in acicular ferrite according
to the KurdjumowSachs orientation relationship; (a) The <100> texture component, (b) The two complementary <111> texture components. After Kluken et a/.63
fully understood. However, detailed TEM studies performed by Bhadeshia et aL37>59>62>66 have
clearly demonstrated that acicular ferrite is a form of intragranularly nucleated bainite. In
practice, this means that the microconstituent may be present either as 'upper' or 'lower'
acicular ferrite in the weld deposit (depending on the carbon concentration), as shown
schematically in Fig. 6.40.
In general, the ferrite component of upper bainite is composed of groups of thin parallel
laths (subplates) with a welldefined crystallographic habit.2061 Although the growth mechanism of upper bainite is still a subject of considerable controversy, it has been postulated that
the subplates advance into the austenite with their own tip configurations. One model is shown
in Fig. 6.41, where each subplate forms as a ledge upon the adjacent subplates through a
nucleation and growth process. These ferrite laths possess the same variant of the KS orientation relationship, which means that they are separated by lowangle grain boundaries. A
typical austenite grain will contain numerous sheaves of bainitic ferrite exhibiting different
variants of the KS orientation relationship. This implies that the boundary between adjacent
plates of acicular ferrite should alternately be of the lowangle and highangle type, a feature
which also has been observed experimentally (see data in Fig. 6.42). Hence, both the morphology and the crystallography of acicular ferrite bear a close resemblance to upper bainite.
Columnar grain
Fig. 6.39. Schematic diagram showing the sequence of reactions occurring during cooling of a lowalloy
steel weld through the critical transformation temperature ranges. After Kluken et al.63
Carbon diffusion
into austenite
Carbide precipitation
from austenite
'Upper' acicular
ferrite
'Lower' acicular
ferrite
Fig. 6.40. Schematic illustration of the transition from 'upper' to 'lower' acicular ferrite in lowalloy
steel weld metals. The diagram is based on the ideas of Bhadeshia and Svensson.37
Successive
nucleation and
growth of
parallel plates
Misorientation, degrees
Fig. 6.41. Proposed model for nucleation and growth of upper bainite in steel (schematic). After
Verhoeven.61
Plate number
Fig. 6.42. Measured spatial misorientation between adjacent plates of acicular ferrite in a lowalloy steel
weld. Data from Kluken et a/.63
A more realistic approach would be to estimate the volume fraction of acicular ferrite
via the equation:37
(650)
where f%FFe and f%FFe are the corresponding volume fractions of grain boundary ferrite and
Widmanstatten ferrite, respectively (note that in equation (650) the formation of microphases
has been disregarded).
The method outlined above has shown to work well for numerous welds (e.g. see Fig. 6.43),
but fails when the primary microstructure consists of a mixture of acicular ferrite and martensite,
as is the case in high strength steel weld deposits6768 In spite of this shortcoming, equation (650) expresses in an explicit manner the real essence of the problem, namely that the evolution
of the acicular ferrite microstructure depends on the interplay between several competing nucleation and growth processes which occur consecutively during cooling from the Ae3temperature. This important point is often overlooked when discussing the conditions for acicular
ferrite formation in lowalloy steel weld metals.
Size effects in heterogeneous nucleation
The important influence of secondphase particles on the austenite to ferrite transformation
has been examined theoretically by Ricks et al69 using classical nucleation theory. By assuming inert, incoherent, and nondeformable inclusions and constant values for the volume free
energy change and the surface free energy of both the yFe /yFe and yFe /aFe boundaries, the
normalised energy barrier to nucleation has been calculated and plotted against the particle
radius as shown in Fig. 6.44. It can be seen from the figure that the nucleation of ferrite at
inclusions is always energetically more favourable than homogeneous nucleation, but less
favourable than nucleation at austenite grain boundaries, irrespectively of the inclusion size.
The most potent nucleation sites are particles of a radius greater than about 0.20.5 jum, which
Volume fraction
Allotriomorphic
ferrite
Widmanstatten
ferrite
Acicular
ferrite
AG* /AG*
het.
horn.
Inclusion
, Austenite
grain boundary
is within the typical size range of most weld metal inclusions (see Figs. 2.572.61 in Chapter
2). This finding is in excellent agreement with the results of Barbaro et al.70 reproduced in Fig.
6.45, showing that a certain minimum inclusion size (say 0.20.3 jam) is required for acicular
ferrite nucleation in steel weld deposits.
It should be noted, however, that Ricks et al69 omitted a consideration of the effects of
plastic strain produced as a result of differences in thermal contraction between the austenite
and the particles as well as the possibility for the ferrite to adopt reasonable orientation relationships with both the austenite and the catalyst particles. Based on nucleation theory it can
be argued that these factors will influence the transformation process.1 This, in turn, may
explain why certain types of inclusions appear to be more favourable nucleation sites for acicular
ferrite than other (to be discussed below).
Catalyst effects in heterogeneous nucleation
As mentioned above, there is considerable circumstantial evidence available in the scientific
literature that intragranular nucleation of acicular ferrite is preferentially associated with specific types of nonmetallic inclusions (i.e. 7Al2O3, MnOAl2O3, TiN). 365571 " 78 Different mechanisms have been proposed over the years to explain these phenomena, including:3655'73'76'77
(i)
Nucleation resulting from a small lattice disregistry between the inclusions and the
ferrite.
(ii)
(iii) Nucleation in the vicinity of inclusions resulting from favourable strain or dislocation
arrays due to differences in the thermal contraction between the particles and the
matrix.
Probability of nucleation
Because of the complexity of the weld metal inclusions, and the experimental difficulties
involved in performing controlled in situ measurements, it cannot be stated with certainty
which of these three mechanisms that are operative during the acicular ferrite transformation.
However, based on simple theoretical calculations it can be argued that the contribution from
the elastic strain fields around the particles due to differential contraction effects probably is
too insignificant to influence the free energy of transformation and that the resulting punchingstress at the particle/matrix interface is well below the critical value required to generate new
dislocations in the austenite.47'81'82 Moreover, detailed STEM/EDS microanalyses have failed
to reveal detectable variations in the matrix composition in the vicinity of the inclusions.83
Hence, nucleation resulting from a small lattice disregistry between the inclusions and the
ferrite appears to be the most likely explanation to the observed effects of deoxidation practice
on the weld metal transformation behaviour.
From a theoretical standpoint, the development of a faceted ferrite nucleus which exhibits a
rational orientation relationship with both the austenite and the inclusions would require that
the substrate and the austenite have similar crystal structures and identical lattice orientations.
The catalyst particles must therefore be cubic and bear an orientation relationship with the
austenite which lies within the Bain region.27 However, considering the fact that the weld
metal inclusions form in the liquid state prior to the solidification process, the latter requirement cannot generally be met.37 Nevertheless, even if the orientation of the inclusions were
perfectly random, it is apparent that orientation relations within the Bain region would be
observed purely by chance. In view of the high symmetry of the cubic system, the probability
of this happening must be calculated. An approximative estimate is given below for single
phase cubic inclusions, based on the method described by Ryder et al?4
Figure 6.46 contains a standard stereographic projection of the austenite crystal, showing
the <100> rFe poles (squares) and the <110>rFepoles (dots). The Bain orientation region is
represented by small circles of radius 11 centered on the <100> r / v  and the <110> rFe poles.84
Fig. 6.46. Stereographic projection of Bain regions represented by 11 circles, round <100>7_Fepoles
(squares) and <110>7_Fepoles (dots) of the austenite lattice. After Ryder et aiS4
The austenite/inclusion(i) orientation relationship is within the orientation region derived
from the Bain correspondence if one <100>rpole lies within a <100>7_Feregion and the other
two <100>rpoles lie within <110>rF<?regions. Assuming a random inclusion orientation, the
probability P 1 that a given <100>rpole lies within a given <100> r ^region is (o/47T, where o>
is the solid angle enclosed by one <100> r ^region (equal to 27t(lcos 11) = 0.115). Since
there are three <100>/ directions and six <100>7_Feregions, the probability that at least one
<100>rpole lies within a <100>7_F,region is equal to:
(651)
Imagine now that the inclusion lattice is rotated through 360 about the <100>raxis which
lies within a <100> y . Fe region. The other two <100> r poles will then lie within
<110>7.Feregions for at the most 4 X 22 = 88 of this rotation, since the diameter of the
<100>7_Feregions is 22. Hence, the probability P2 that, if one <100> r pole lies within a
<100>7.Feregion, the other two will lie within <110>7.Feregions is given in the upper limit
by:
(652)
The total probability that a given orientation relationship lies within the Bain region purely
by chance is thus:
(653)
Therefore, assuming random orientation, about 4% of the weld metal inclusions would lie
within the Bain region purely by chance if they were single phase cubic crystals. In practice,
however, inclusions commonly found in lowalloy steel weld metals are of a very heterogeneous chemical and crystalline nature. As shown in Fig. 2.72 in Chapter 2, a typical inclusion
may contain up to six different constituent phases, including the three cubic phases 7Al2O3,
MnOAl2O3, and TiN. This implies that at least 12% of the inclusions may contain a cubic
phase which lies within the Bain orientation region.
Orientation Relationship
(s)
No.
Plane Combinations
Ratio^
(100) s ~//(011) a _ Fe
[011]s ~ // [533] a . Fe
{011}a_Feand{400}s
1.02
MnOAl2O3
(200) s ~//(110) a F e
{110}a_Feand {400}s
0.99
(spinel)
(011)s // (010)a Fe
[Oil],//[QOlLp 6
0.99
(TlO) 8 // (100)a Fe
(112)s//(011)a_Fe
[lll] s //[011] a . F e
0.97
1.17
(101) s //(103) a _ Fe
(320)s//(112)a_Fe
[232] s //[351] a _ Fe
{310} a . Fe and{330} s
{112}a.Fe and {320}s
0.91
0.99
1.10
1.01
7Al2O3
(distorted spinel)
TiN
(NaCl)
Variant
Interplanar Spacing
(101) s //(133) a _ Fc
[221]s//[200]a_Fe
Defined as d(hkl)a.Fe /d(hkl)s
Fig. 6.47. Standard (100) stereographic projections of the orientation relationships listed in Table 6.3; (a)
Variant (1) and (3), (b) Variant (4), (5) and (6). The Bain orientation region is indicated by the 11 circles
in the graphs (see Fig. 6.46 for details).
the advantage of partial lattice coherence in a second (independent) direction, which further
contributes to a reduction of AG^e/ through a minimisation of the interfacial energy between
the two phases. This makes TiN an extremely efficient nucleant for acicular ferrite.
Microstructure data available for submerged arc (SA) steel weld deposits clearly support
the above findings that nucleation of acicular ferrite occurs preferentially at inclusions which
contain aluminium or titanium. As shown in Fig. 6.50, a high volume fraction of acicular
ferrite is always achieved when sufficient amounts of titanium are added either through the
filler wire or the flux, irrespectively of the aluminium and oxygen concentrations. This is in
sharp contrast to welds produced with welding consumables containing low levels of titanium,
where the acicular ferrite content drops rapidly with decreasing [A%Al]weld/[%O]anal ratios
Fig. 6.48, Example of a faceted crystal delimited by {100} and {110} planes (schematic).
TiN
MnOAI2O3
TAl2O3
AG
;et.
Nucleation site
Fig. 6.49. Qualitative ranking of different inclusion constituent phases with respect to nucleation potency of acicular ferrite.
due to the presence of lower fractions of 7Al2O3 and MnOAl2O3 in the inclusions (see Fig.
2.72 in Chapter 2). Similar observations have also been made by other investigators.3672"
76,78,79
It should be noted that the weld metal transformation behaviour in practice depends on
complex interactions between a number of important variables, including alloying and
deoxidation practice, the solidification microstructure, the prior austenite grain size, and the
weld thermal cycle.36'5358 This means that the presence of 7Al2O3, MnOAl2O3 or TiN at the
surface of the inclusions is perhaps a necessary but not a sufficient criterion for formation of
acicular ferrite in steel weld metals.
Example (6.8)
Consider a partly TiAl deoxidised steel weld metal which contains a total number of 4 X 107
inclusions per mm 3 . Based on Fig. 2.72 in Chapter 2 and the theory developed above, estimate
an upper limit for the volume of a typical plate of acicular ferrite when the weld metal [A%Al]weld/
\yO\anaL r a t i o i s  80 
Vol% AF
[A%Alweld/[%o]anal
Fig. 6.50. Effect of deoxidation practice (inclusion chemistry) on the acicular ferrite transformation in
lowalloy steel weld metals. Data compiled from miscellaneous sources.
Solution
From Fig. 2.72 it is seen that the total number of constituent phases in the inclusions is six,
including the three cubic phases 7Al2O3, MnOAl2O3, and TiN. If we assume a random orientation and only one nucleation event per inclusion, the following upper limit for the acicular
ferrite plate volume is obtained:
The above volume corresponds to an acicular ferrite plate which has the shape of a square
lath of side lOjim and thickness 2im. Although this estimate is in reasonable agreement with
experimental observations,37 the prediction is conservative in the sense that it assumes only
one nucleation event per inclusion. In practice, an oxide inclusion which is orientated within
the Bain region has the capability of nucleating several acicular ferrite plates, as shown by the
SEM micrograph in Fig. 6.51. In addition, the acicular ferrite plates may nucleate
autocatalytically at aFe /yFe boundaries, a process which also is referred to as sympathetic
nucleation in the literature.36'37'7083 At present, it is not clear to what extent autocatalytic
nucleation plays a role in the development of the acicular ferrite microstructure.
Hardenability effects
Since acicular ferrite is one of the last phases to form after the growth of allotriomorphic and
Widmanstatten ferrite, it is bound to be influenced by the prior transformation products, as
indicated by equation (650). The strong dependence of the acicular ferrite content on the
austenite grain size must therefore be understood on this basis. 3747 ' 48 ' 7088 " 90 It is evident from
the data of Barbaro et al.70 reproduced in Fig. 6.52 that a coarse austenite grain size favours
intragranular nucleation of acicular ferrite at the expense of formation of allotriomorphic and
Fig. 6.51. SEM micrograph of a carbon extraction replica showing evidence of multiple nucleation of
acicular ferrite at a weld metal inclusion.
Widmanstatten ferrite. This effect is most pronounced during slow cooling, since the combination of a small austenite grain size and a slow cooling rate implies that much of the yFephase already has transformed to allotriomorphic ferrite before the temperature for intragranular
nucleation of acicular ferrite is reached.
Example (6.9)
Consider a lowalloy steel weld metal which contains a total number of 5 X 107 inclusions per
mm3 with an average radius of 0.25 Jim. Use this information to evaluate the conditions for
acicular ferrite formation within the asdeposited weld metal and the lowtemperature reheated
region of the weld when the austenite grain size is 100 and 10 iim, respectively. In these
calculations we shall assume that the equilibrium volume fraction of ferrite ff~Fe in the fully
transformed steel is 0.9.
Solution
First we need to estimate the total surface area per unit volume available for ferrite nucleation at austenite grain boundaries, Sv (GB), and nonmetallic inclusions, Sv (/), respectively.
Asdeposited weld metal
(as before)
Volume fraction of AF
On this basis it is not surprising to find that the microstructure within the grain refined
region of lowalloy steel welds is usually polygonal ferrite, while the asdeposited weld metal
also contains high proportions of acicular ferrite (see Fig. 6.19(c) and (d), respectively).
It is important to realise, however, that the presence of allotriomorphic ferrite at the austenite
grain boundaries has the benefical effect of suppressing the formation of bainitic sheaves at the
austenite grain boundaries, which, in turn, allows the acicular ferrite to develop on intragranular
nucleation sites.9091 Consequently, due to the number of competing nucleation and growth
processes involved, the volume fraction of acicular ferrite is often seen to pass through a maximum when the weld metal hardenability is successively increased by additions of alloying
elements, as shown in Fig. 6.53.
Microstructural component, %
GF and PF
WF
63.7 Bainite
Bainitic microstructures (besides acicular ferrite) are frequently observed during welding, particularly in the HAZ of lowcarbon microalloyed steels,22'93"95 but also within the fusion region of the weld if the nucleation conditions are favourable.37'55'5991 Two main forms can be
identified, i.e. upper and lower bainite, as indicated in Fig. 6.19(a).
6.3.7.1 Upper bainite
In general, the morphology of upper bainite bears a close resemblance to Widmanstatten ferrite (both are typical sideplate structures), while its crystallography is more like that of lowcarbon martensite (the KS orientation relationship is usually less precise than for WF).20 In
spite of this similarity, the formation of upper bainite cannot be fully understood in terms of
HAZ
WF/UB
HAZ
Fig. 6.54. Schematic illustration of the HAZ transformation behaviour during high heat input welding;
(a) Tioxide containing steel, (b) Conventional AlTi microalloyed steel. The diagrams are based on the
ideas of Homma et al.92
the classic theory of martensite nucleation assuming a pure invariant plane strain deformation.1'20'61
HAZ transformation behaviour
As shown in Fig. 6.19(a), upper bainite is frequently observed within the grain coarsened HAZ
of lowcarbon microalloyed steels. The bainite sheaves nucleate preferentially at austenite
grain boundaries and subsequently grow into the interior of the grains until they physically
impinge with other transformation products such as martensite or Widmanstatten ferrite. Because of these competitive nucleation and growth processes, the highest volume fractions of
upper bainite are normally attained at an intermediate cooling rate, as indicated by the microstructure data in Fig. 6.55.
Weld metal transformation behaviour
Upper bainite may also form within the columnar grain region of high heat input SA steel
welds, as shown by the optical micrograph in Fig. 6.56. The indications are that the observed
Microstructural component, %
iUBj
Ferrite with
aligned second dhase (FS)
WF
Mill
Cooling time, At 8/5 , s
Fig. 6.55. Effect of cooling time, Af8/5, on the grain coarsened HAZ transformation behaviour (simulated
thick plate heat cycles, Tp 13500C). Data from Akselsen et al96
Fig. 6.56. Optical micrograph showing formation of upper bainite within the columnar grain region of a
SA steel weld.
shift in the weld metal transformation behaviour is related to a change in the deoxidation
practice which alters the kinetics of the subsequent solid state transformation reactions through
a modification of the solidification microstructure.64 Solidification induced phosphorus
segregations are of particular interest in this respect, since previous examinations have shown
that phosphorus can strongly enhance the formation of grain boundary ferrite by raising the
local Ae3 temperature of the steel97 (see also Example 6.5).
Following the discussion in Section 6.3.5.4, the evolution of allotriomorphic ferrite at the
austenite grain boundaries has the beneficial effect of suppressing the formation of bainitic
sheaves at these sites, which, in turn, allows the acicular ferrite to develop on intragranular
Microstructural component, %
Carbides
Lower bainite
Martensite
Fig. 6.58. TEM micrograph showing the formation of lower bainite within the HAZ of a lowcarbon
microalloyed steel.
6.3.8 Martensite
At very high undercoolings, the austenite decomposes to martensite by means of an invariant
plane strain deformation, which implies that there is no change in the steel chemical composition. The reaction product will either be lath or plate (twinned) martensite, depending on the
alloy level. Lath martensite is commonly found in plain carbon and lowalloy steels up to
about 0.5wt% C, and is formed by a slip mechanism, as shown schematically in Fig. 6.59(a).
When the carbon content exceeds this threshold, the martensite transformation occurs rather
by formation of deformation twins (Fig. 6.59(b)). The crystal structure of plate martensite is
bet (bodycentred tetragonal), while lath martensite reveals a bcc (bodycentred cubic) structure,61 which becomes increasingly distorted with increasing steel carbon contents.20 Both
transformation products exhibit the characteristic KurdjumowSachs orientation relationship
with the austenite, but this relationship tends to be less precise at high carbon levels.1'20
6.3.8.1 Lath martensite
Lath martensite is frequently observed within the grain coarsened HAZ during low heat input
welding of microalloyed steels. As shown in Fig. 6.60, the microstructure will be fully
martensitic for values of Ar875 up to about 510s, depending on the steel hardenability. The
hardness of the martensite formed is usually below 400 VPN, which is significantly lower than
the corresponding peak value achieved after waterquenching.98 This means that considerable
autotempering occurs during cooling from the M^temperature, which in the case of lowcarbon microalloyed steels can be as high as 5000C.
6.3.8.2 Plate (twinned) martensite
In the intercritical HAZ (i.e. partly transformed region), carbonrich austenite, formed by decomposition of pearlite or isolated carbides, may transform to high carbon (twinned) martensite
on cooling, as shown schematically in Fig. 6.61. However, since the Mf temperature in this
case will be far below room temperature, significant amounts of retained austenite may be
present in the areas adjacent to the martensite islands. This transformation product is therefore
Slip
Twin
Fig. 6.59. Mechanisms of martensite formation in steel (schematic); (a) Slip along parallel planes, (b)
Generation of deformation twins. The diagram is based on the ideas of Verhoeven.61
Steel A
Steel B
Steel C
Rapid growth of austenite into pearlite until the dissolution process is completed.
(ii)
Slower growth of austenite into ferrite at a rate which is either controlled by carbon
diffusion in the austenite or by diffusion of substitutional elements such as manganese in the ferrite, depending on the applied annealing temperature.
wt% C
Fig. 6.61. Schematic illustration of the formation of plate (twinned) martensite within the intercritical
HAZ of lowcarbon microalloyed steels; Heating leg of thermal cycle: Pearlite > yFe (>0.5 wt% C),
Cooling leg of thermal cycle: yFe(> 0.5 wt% C) > a'Fe (twinned martensite).
(iii) Very slow final equilibration of solute concentration gradients within the austenite or
the ferrite through diffusion.
At temperatures above say 7700C, the rate of pearlite decomposition is sufficiently high for
complete dissolution, even during lowheat input welding. Hence, in hot rolled and normalised steels all pearlite will transform to austenite when the peak temperature of the thermal
cycle exceeds this threshold. However, if the starting microstructure is a mixture of martensite
and upper bainite (as often will be the case in multipass welding), the austenite will nucleate
both at prior austenite grain boundaries and along the interfaces between laths of bainite or
martensite, as shown by the optical micrographs in Fig. 6.62. A similar pattern has also been
observed during intercritical annealing of dualphase steels.104'105 Considering the kinetics,
carbides precipitated within autotempered lowcarbon martensite or between the laths of bainite
would be expected to dissolve at a rate comparable with that of pearlite.105 Hence, it is reasonable to assume that no retained carbides will be present intragranularly after reheating to say
Tp 7700C when the starting microstructure is a mixture of martensite and bainite.
Further growth of the austenite into the ferrite requires, however, that the peak temperature
of the thermal cycle is raised significantly above 7700C in order to reach the kinetic
(paraequilibrium) stage where the reaction is controlled by diffusion of carbon in austenite."
Consequently, the possibilities of obtaining growth of the austenite colonies within the low
peak temperature regions of the intercritical HAZ are strongly limited, which implies that the
carbon content of the austenite islands in these regions should be close to the saturation level
of carbon in austenite at the temperature of dissolution.
Conditions for MA formation
On subsequent cooling, the carbonrich austenite can decompose to a variety of microstructures (ranging from twinned martensite to pearlite) or remain untransformed, depending on the
base plate hardenability, the weld cooling programme, and the initial size of the austenite
islands.102103'106 When the starting microstructure is a mixture of ferrite and pearlite (as frequently observed in normalised steels), twinned martensite forms readily along the prior base
(a)
(b)
Fig. 6.62. Optical micrographs showing favourable sites for austenite formation during twopass weld
thermal simulation. (First cycle: Tp 13500C, Atm 12 s, Second cycle: Tp 775C, Atm 12 s);
(a) Intergranular, (b) Intragranular. After Akselsen et aim
metal pearlite bands or within isolated pearlite colonies at high cooling rates, as shown by the
TEM micrographs in Fig. 6.63(a) and (b). With decreasing cooling rates, however, the transformation product shifts from twinned martensite to predominatly pearlite (see Fig. 6.63(c)
and (d)) in the absence of strong hardenability elements such as molybdenium and boron which
can stabilise the MA constituent.102'107"109
In controlled rolled and accelerated cooled steels, where the carbides are mainly present in
the form of submicroscopic colonies or isolated particles located at ferrite/ferrite grain boundaries
(see Fig. 6.64(a)), the situation becomes slightly more complex. Under such conditions, the
majority of the austenite islands formed within the low peak temperature region of the
intercritical HAZ will be of a size below 1 jam. It has been verified experimentally that such
(a)
(C)
(b)
(d)
Fig. 6.63. Examples of transformation products formed after intercritical thermal cycling of a ferritic/
pearlitic starting microstructure to a peak temperature of 775C; (a) MA islands (black areas) surrounded by ferrite (At6/4 ~ 5 s), (b) Closeup of twinned martensite within a MA island (At6/4 ~ 5 s),
(c) Isolated pearlite colony formed at an intermediate cooling rate (Ar674 ~ 12 s), (d) Pearlite colonies
formed during slow cooling (At614 ~ 35 s). After Akselsen et al.103
small particles do not readily transform to martensite, but will largely remain in the steel in the
form of retained austenite.100'103'110'111 This can be explained by the lack of nucleation opportunities for martensite or by the volume restraint provided by the surrounding ferrite matrix.1 x x
Consequently, twinned martensite in controlled rolled and accelerated cooled steel is seen to
form within a few, relatively large austenite islands which stem from decomposition of single
carbide colonies, as illustrated by the TEM micrograph in Fig. 6.64(b).
During multipass welding, the initial carbide distribution in the base plate will be of less
importance because of the transformations imposed by the heat of previous weld passes. Hence,
after full reaustenitising of the steel the MA constituent may form both intergranularly and
intragranularly, depending on the starting microstructure, as shown previously in Fig. 6.62(a)
and (b). An example of intergranularly nucleated twinned martensite is contained in Fig. 6.65.
Volume fraction of MA constituent
At present, an adequate kinetic model for partial reaustenitising of steel during continuous
heating and cooling is lacking. However, in view of the previous discussion it is obvious that
the volume fraction of the MA constituent within the intercritical HAZ of single pass steel
weldments must be closely related to the initial base plate pearlite content. This is also in
agreement with experimental observations (see Fig. 6.66).
Next Page
(a)
(b)
Fig. 6.64. Conditions for martensite formation within the intercritical HAZ of a controlled rolled and
accelerated cooled steel containing copper and nickel; (a) TEM micrograph of the initial base plate carbide distribution, (b) TEM micrograph showing evidence of twinned martensite within a l(im large
austenite colony formed during intercritical thermal cycling to T 775C (Ar674 35 s). After Akselsen
etal^
Previous Page
(a)
(b)
Fig. 6.64. Conditions for martensite formation within the intercritical HAZ of a controlled rolled and
accelerated cooled steel containing copper and nickel; (a) TEM micrograph of the initial base plate carbide distribution, (b) TEM micrograph showing evidence of twinned martensite within a l(im large
austenite colony formed during intercritical thermal cycling to T 775C (Ar674 35 s). After Akselsen
etal^
Fig. 6.65. TEM micrograph showing evidence of intergranularly nucleated twinned martensite in doublecycled specimens (First cycle: Tp ~ 13500C, A;6/4 12s, Second cycle: Tp  775C, A;6/4  12s) After
Akselsen et al.103
Fig. 6.66. Effect of base plate pearlite content on the volume fraction of MA constituent in thermally
cycled specimens (simulation conditions: Tp 775C, Atm 5s). Data from Akselsen et al.m
Temperature, 0C
Grain boundary
Cr carbide precipitate
Fig. 6.68. Schematic illustration OfCr23C6 precipitation at grain boundaries in austenitic stainless steels.
The diagram is based on the ideas of Kou.112
(654)
A graphical representation of equation (654) in Fig. 6.69 shows that the nose of the
Ccurve for the reference steel (characterised by Teq = T*eq and D1 = D 7 ) is strongly shifted to
the left in the ITdiagram, thereby providing favourable conditions for Cr23C6 precipitation in
the heat affected zone during welding. The problem becomes less imminent if the base plate
carbon content is reduced from say 0.07 to 0.04 wt%. In that case the associated reduction in
the solvus temperature Teq from 920 to 8000C will displace the Ccurve towards longer times
in the diagram which by far exceed the duration of a typical weld thermal cycle, as shown in
Fig. 6.69.
6.4.2 Area of weld decay
Experience shows that corrosion attact within the HAZ of single pass austenitic stainless steel
welds usually starts in regions where the peak temperature of the thermal cycle has been close
to the nose of the Ccurve in Fig. 6.69. This observation cannot be explained by just considering the HAZ temperaturetime pattern, since the retention time within the critical temperature
range for chromium carbide precipitation is virtually the same for both high and low peak
temperature thermal cycles, as indicated in Fig. 6.70. The explanation is, of course, that the
Ccurve becomes increasingly shifted to the right in the ITdiagram as the peak temperature
increases due to austenite grain growth occurring during welding. This point is more clearly
illustrated below.
Example (6.10)
Consider single pass butt welding of 2mm sheets of 304 austenitic stainless steels with covered electrodes under the following conditions:
Evaluate on the basis of equations (632), (633), and (654) in combination with the grain
growth diagram in Fig. 5.30(b) (Chapter 5) the conditions for chromium carbide formation
within the HAZ during welding when the base plate carbon content is 0.07 and 0.04 wt% C,
respectively. Thermal data for austenitic stainless steels are given in Table 1.1 (Chapter 1).
Temperature, 0C
Time, s
Fig. 6.69. Effect of carbon content on the isothermal precipitation of C ^ C 6 in 304 austenitic stainless
steels. Broken curve: 0.07 wt% C, Solid curve 0.04 wt% C. The diagrams are constructed on the basis
of equation (654).
T
Value
tr
T*q
A0
Q/
>*
(K)
(s)
(K)
(Jmol 1 )
(kJmol1)
(jxm)
923
60
1193
150
200
18
Solution
If we, as a first approximation, neglect the contribution from heat flow in the welding direction, the temperaturetime pattern is given by equation (1100) in Chapter 1. Taking n = 4 and
(1 Xc) = 0.98 (assuming interfacecontrolled growth), the extent of chromium carbide precipitation occurring within the HAZ of the weld can be calculated from equations (632) and
(633) by numerical integration over the actual thermal cycles. The results of such computations are presented graphically in Fig. 6.71.
As expected, precipitation of chromium carbides occurs readily within the low peak temperature region of the weld when the carbon content is 0.07 wt%. For points located within
the grain growth zone, the resulting displacement of the Ccurve towards longer times in the
CCTdiagram will gradually retard the reaction, which implies that the fraction transformed
starts to decline when the peak temperature of the thermal cycle exceeds about 11500C. In
contrast, the lowcarbon steel reveals no sign of chromium carbide precipitation within the
HAZ. This result is in good agreement with general experience, and explains why the carbon
content of modern austenite stainless steels has been gradually lowered to values below 0.03
wt % in step with the progress in the steel manufacturing technology.20112
Fraction transformed
Agehardenable AlMgSi alloys are widely used as structural components in welded assemblies. They offer tensile strength values higher than 350 MPa in the artificially aged T6 condition owing to the presence of very fine, needle shaped (3"(Mg2Si) precipitates along <100>
directions in the aluminium matrix.113 Although AlMgSi alloys are readily weldable, they
suffer from severe softening in the heat affected zone (HAZ) because of reversion (dissolution) of the 3"(Mg2Si) precipitates during the weld thermal cycle. 6112 This type of mechanical
impairment represents a major problem in engineering design.114
Peak temperature, 0C
Fig. 6.71. Conditions for Cr23C6 precipitation within the HAZ of a single pass austenitic stainless steel
butt weld (Example 6.10).
Fig. 6.72. TEM micrograph showing nucleation of nonhardening (3'(Mg2Si) precipitates at Mnbearing
dispersoids in an AA 6082 alloy.
Table 6.5 Input data used to construct Ccurve for p' (Mg2Si) precipitates in reference aluminium alloy.
Parameter
Value
Tr
t\
T*q.
A0
(K)
(s)
(K)
(Jmol"1)
573
20
766
350
Qj
S^
100
(655)
A graphical representation of equation (655) in Fig. 6.73 shows that the nose of the Ccurve for the reference alloy (characterised by Teq = r and Sv = S^ is strongly shifted to
the left in the ITdiagram, thereby providing favourable conditions for (3'(Mg2Si) formation
dunng welding. In general, an increase in T or Sv will enhance the quenchsensitivity of the
material because of the resulting increase in the nucleation rate. This will be the case if the
alloy contams large amounts of excess silicon in solid solution or is homogenised at a temperature lower than 5800C.118119
Example (6.11)
Consider plasmaMIG butt welding of a 10 mm thick AlMgSi plate under the following
conditions:
qo = 10 kW, v = 10 mm s~\ T0 = 200C
Evaluate on the basis of equations (632), (633), and (655) the conditions for P'(Mg,Si)
precipitation within the high peak temperature region of the HAZ during welding (T7 > T )
In these calculations we shall assume that the transformation behaviour of the base'metafis
similar to that of the reference alloy in Table 6.5. Relevant thermal data for AlMgSi alloys
are given in Table 1.1 (Chapter 1).
Solution
Temperature,C
Referring to Fig. 1.43 in Chapter 1, the mode of heat flow becomes essentially onedimensional if the net arc power is kept sufficiently high compared with the plate thickness (i.e.
Time, s
Fig. 6.73. Ccurve for precipitation of 3'(Mg2Si) in the reference AA 6082 alloy. The diagram is constructed on the basis of equation (655).
qold>0.05 kW mm"1). Since this requirement is met in the present case, the HAZ temperaturetime pattern is given by equation (1100).Taking n = 0.75 and (1  Xc) = 0.84 for precipitation
of p'(Mg2Si) in AA6082 aluminium alloys,118 we obtain after integration of equations (632)
and (633) over the weld cooling cycle:
and
The above calculations show that precipitation of (3'(Mg2Si) particles at dispersoids is, indeed, a significant process under the prevailing circumstances. Since the mode of heat flow
during single pass butt welding of aluminium plates is essentially onedimensional, it is possible to construct general transformation diagrams which give the fraction transformed as a
function of the applied heat input. An example of such a diagram is contained in Fig. 6.74.
Fraction transformed
logt
Net heat input ( q /vd), kJ/mm2
Fig. 6.74. Conditions for 0'(Mg2Si) precipitation within the HAZ of single pass AA 6082 butt welds
(Tp>Teq).
Hardness (VPN)
Waterquenched
specimens
57 days
Log time
Fig. 6.75. Typical ageing curve for an AA 6082 aluminium alloy at room temperature (schematic).
2D kinetic (cell) model, assuming that the reaction is interfacecontrolled. Let r denote the
radius of the growing precipitates (defined in Fig. 6.76(a)). Since we are only interested in the
terminal value of r at a fixed temperature, the time t in the expression for r can be regarded as
constant. Hence, we may write (when C0 Ca and Cp C0):
(656)
where C4 is a kinetic constant.
If the distribution of the precipitates is approximated by that of a 2D facecentered cubic
space lattice (see Fig. 6.76(b)), the parameter, AXp, is simply given as:
(657)
where rm denotes the maximum particle radius which forms within the system if all alloying
elements are present in solid solution at the onset of the ageing reaction (C0 = C*).
Because of the stoichiometry of the precipitation reaction, C0 and C* in the expression for
AXp may be taken proportional to the magnesium concentration in solid solution. Hence, we
may write:
(658)
is the corresponding resistivity in the fully annealed condition (i.e. when all Mg and Si are tied
up in precipitates).
A comparison between equation (658) and the electrical resistivity data in Fig. 6.77 confirms the relevance of this powerlawrelationship, although the deviation in certain cases in
admittedly large.
Example (6.12)
Consider plasmaMIG butt welding of a 10 mm thick AlMgSi plate under conditions similar
to those employed in Example 6.11. Estimate on the basis of equation (658) the relative
(b)
Concentration
(a)
Distance
AXp
Fig. 6.76. Simplified 2D kinetic (cell) model for precipitation of hardening particles in AlMgSi alloys
during natural ageing; (a) Particle/matrix concentration profile, (b) Cell geometry.
Mg /C*Mg
fraction of hardening precipitates which forms within the fully reverted region of the HAZ
after prolonged room temperature ageing.
Solution
Referring to Example (6.11), the relative fraction of (3'(Mg2Si) precipitates which forms at
dispersoids during the weld cooling cycle amounts to:
from which
This gives:
Since the resulting precipitation strength increment is directly proportional to AXp,ul loss
of solute in the form of (3'(Mg2Si) particles during the weld cooling cycle will inevitably lead
to a reduced HAZ strength in the naturally aged (T4) condition.6 We shall return to this question in Chapter 7 (Section 7.4.3).
6.5.2 Sub grain evolution during continuous drive friction welding
Continuous drive friction welding is a solidstate joining process that produces coalescence by
the heat developed between two surfaces by mechanically induced rubbing motion. When the
appropriate rotation velocity is reached, the two workpieces are brought together and an axial
force is applied. The two surfaces are held under pressure, and due to the heat developed, a
plasticised layer forms at the interface. After a predetermined time the rotation stops and the
pressure is increased to facilitate forging or local upsetting of the heated metal. Filler metal,
flux, or shielding gas is not required with this process.
The structural changes within the fully plasticised region of friction welded AlMgSi
alloys arise from the combined action of work hardening and softening due to dynamic recovery.
When the steady state conditions for deformation in friction welding are reached, the rate of
dislocation generation in the plasticised material balances the dislocation annihilation rate,
which means that large strains can be imposed without any changes in theflowstress or subgrain
size. Referring to Fig. 6.78, the subgrain structure within the fully plasticised region consists
of sheaves of virtually equiaxed grains (23 jim in size) that are separated from each other by
lowangle grain boundaries. This type of microstructure is characteristic of hot worked aluminium alloys.120
In general, the ZenerHollomon (Zh) parameter provides a basis for evaluating the evolution of the subgrain structure during hot working. As shown by McQueen and Jonas,121 the
following relationship exists between the subgrain diameter ds and the Z^parameter in
aluminium alloys:
(660)
Partly deformed
region
Fully plasticized
region
Partly deformed
region
(a)
(b)
Fig. 6.78. Micrographs showing the subgrain structure within the fully plasticised and partly deformed
region of a friction welded AlMgSi alloy; (a) Overview, (b) Closeup of the subgrain structure (EBSP
image).
where
(661)
The peak temperature T and the strain rate i distributions within the fully plasticised region of a friction weld can be computed on the basis of the generic models developed by
Midling and Grong.122 Examples of such calculations are shown in Fig. 6.79(a). Plots of the resulting ZenerHollomon parameter and subgrain diameter at different locations within the
HAZ are contained in Fig. 6.79(b).
Radial position:
Peak temperature, 0C
Radial position:
ZenerHollomon parameter, s
Axial distance, mm
Axial distance, mm
Fig. 6.79. Modelling of the subgrain evolution in the fully plasticised region of a friction welded AA
6082 aluminium alloy; (a) Predicted peak temperature Tp and strain rate (e) distributions, (b) Plots of
the ZenerHollomon Zh parameter and resulting subgrain diameter ds at different locations within
the HAZ. Data from Midling and Grong.122
It is evident from the graphs that the value of the ZenerHollomon parameter is of the order
of 10141012 s"1 within the fully plasticised region of the HAZ during continuous drive friction
welding of AlMgSi alloys. The predicted range in Zh corresponds to a subgrain size of 23
jam, in agreement with experimental observations (see EBSP image in Fig. 6.78(b)). Outside
the fully plasticised region the ZenerHollomon parameter drops from about 1012 to 1010 s"1
due to a sudden change the strain rate from 102 to 10 s"1 as the contribution from the material
flow field in the rotational direction ceases. This value is outside the validity range of equation
(660), and leads to the formation of coarse subgrains at the boundary between the fully plasticised and the partly deformed region, as shown by the TEM micrographs in Fig. 6.80.
Fig. 6.80. TEM micrographs showing evidence of coarse subgrains at the boundary between the fully
plasticised and partly deformed region of a friction welded AA 6082 aluminium alloy.
References
1.
2.
3.
4.
D. A. Porter and K.E. Easterling: Phase Transformations in Metals and Alloys, 1981, Wokingham
(England), Van Nostrand Reinhold.
M.E. Fine: Phase Transformations in Condensed Systems, 1964, New York, The Macmillian Company.
N. Ryum: Compendium in Physical Metallurgy, 1973, Trondheim (Norway), The University of
Trondheim, The Norwegian Institute of Technology.
G.F. Vander Voort (ed.): Atlas of TimeTemperature Diagrams for Nonferrous Alloys, 1991, Materials Park (Ohio), ASM International.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
97.
98.
99.
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
116.
117.
118.
119.
120.
121.
122.
RE. Marth, H.I. Aaronson, G.W. Lorimer, T.L. Bartel and K.C. Russell: Metall. Trans., 1976, 7A,
15191528.
CM. Dallam and D.L. Olson: Weld. J., 1989, 68, 198s2O5s.
G. Thewlis: HW Doc. IXJ16590 (1990).
R.A. Farrar, Z. Zhang, S.R. Bannister and G.S. Barrite: J. Mater. ScL, 1993, 28, 13851390.
S.S. Babu and H.K.D.H. Bhadeshia: Mater. ScL Technol, 1990, 6, 10051020.
H. Homma, S. Ohkita, S. Matsuda and K. Yamamoto: Weld. J., 1987, 66, 301s309s.
R.E. Dolby: Weld. J., 1979, 58, 225s238s.
B.C. Kim, S. Lee, NJ. Kim and D.Y. Lee: Metall Trans., 1991, 22A, 139149.
R.H. Phillip: Weld. J., 1983, 62, 12s18s.
O.M. Akselsen, J.K. Solberg, G. R0rvik and AJ. Paauw: Technical Report STF34 F87013, 1987,
Trondheim (Norway), SintefDivision of Metallurgy.
A.O. Kluken and 0. Grong: Proc. Int. Conf. on Recent Trends in Welding Science and Technology,
Gatlinburg, TN, May, 1989, pp. 781786. Publ. ASM International, Materials Park, OH, 1990.
O.M. Akselsen, G. R0rvik, M.I. Ons0ien and 0. Grong: Weld. J., 1989, 68, 356s362s.
G.R. Speich, V.A. Demarest and R.L. Miller: Metall. Trans., 1981,12A, 14191428.
C.A.N. Lanzillotto and FB. Pickering: Metal ScL, 1982, 16, 371382.
J.H. Chen, Y. Kikuta, T. Araki, M. Yoneda and Y. Matsuda: Acta Metall, 1984, 32, 17791788.
M. Ramberg, O.M. Akselsen and 0 . Grong: Proc. Int. Conf. on Advances in Welding Science and
Technology, Gatlinburg, TN, May, 1986, pp. 679684. Publ. ASM International, Materials Park,
OH, 1986.
O.M. Akselsen, 0. Grong and J.K. Solberg: Mater. ScL Technol, 1987, 3, 649655.
X, L. Cai, AJ. GarrattReed and W.S. Owen: Metall. Trans., 1985, 16A, 543557.
JJ. YI, LS. Kim and H.S. Choi: Metall. Trans., 1985, 16A, 12371245.
O.M. Akselsen, J.K. Solberg and 0 . Grong: Stand. J. Metall, 1988,17, 194200.
A.R. Marder: Metall. Trans., 1981, 12A, 15691579.
X.P Shen and R. Priestner: Metall Trans., 1990, 21 A, 25472553.
NJ. Kim and YG. Kim: Mater. ScL Eng., 1990, A129, 3544.
K.R. Kinsman, G. Das and R.F Hehemann: Acta Metall, 1977, 25, 359365.
N.K. Balliger andT. Gladman: Met. ScL, 1981, 15, 95108.
S. Kou: Welding Metallurgy, 1987, Toronto (Canada), John Wiley & Sons.
J.E. Hatch (ed.): Aluminium Properties and Physical Metallurgy, 1984, Metals Park (Ohio),
American Society for Metals.
FM. Mazzolani: Aluminium Alloy Structures, 1985, Boston (USA), Pitman Publish. Inc.
O. Lohne and A.L. Dons: Scand. J. Metall, 1983, 12, 3436.
R.P Martukanitz: Proc. Int. Conf. on Advances in Welding Science and Technol, Gatlinburg
(Tenn.), May, 1986, pp. 193201. Publ. ASM International, Materials Park (OH).
H.R. Shercliff, 0. Grong, O.R. Myhr and M.F Ashby: Proc. 3rd. Int. Conf. on Aluminium Alloys
Their Physical and Mechanical Properties, Trondheim (Norway), June, 1992, Vol. Ill, pp. 357369. Publ. The Norwegian Institute of Technology, Trondheim, Norway.
D.H. Bratland, 0. Grong, O.R. Myhr and H.R. Shercliff: Proc. Int. Conf on ComputerAssisted
Materials Design and Process Simulation, Tokyo (Japan), September, 1993, pp. 135141. Publ.
The Iron and Steel Institute of Japan.
D.H. Bratland, 0 . Grong, H.R. Shercliff, O.R. Myhr and S.Tjotta:,4cta Metall Mater., Overview
No. 124,1997,45,122.
CM. Sellars: Proc. 3rd Int. Conf. on Aluminium AlloysTheir Physical and Mechanical
Properties, Trondheim (Norway), June, Vol. Ill, pp. 89106. Publ. The Norwegian Institute of
Technology, Trondheim, Norway.
HJ. McQueen and JJ. Jonas: Treatise on Mater. ScL Technol, 1975,6, 393493.
O.T. Midling and 0 . Grong: Acta Metall Mater., 1994,42,15951609: ibid., 16111622.
Appendix 6.1
Nomenclature
start temperature of ferrite to
austenite transformation (0C or
K)
acicular ferrite
integration parameter
amperage (A)
modified Z1 integral
arbitrary function of X
KurdjumowSachs orientation
relationship
isothermal transformation
free energy donated to the system when nucleation occurs heterogeneously (J mol"1 or J nr 3 )
increase in strain energy
resulting from lattice distortion
(J mol 1 or J m 3 )
increase in surface energy due to
nucleus formation (J mol"1 or J
ITT3)
martensiteaustenite constituent
metal inert gas
polygonal ferrite
integration parameter
half thickness of plate (jam, m)
peak temperature ( C or K)
frequency factor (s"1)
chosen reference temperature
(0C or K)
voltage (V)
migration rate of incoherent
interface (im s"1)
upper bainite
1
Widmanstatten ferrite
arc efficiency
fraction transformed (corrected
for physical impingement of adjacent volumes)
fraction transformed (not corrected for physical impingement
of adjacent volumes)
fraction transformed defined by
the Ccurve
maximum fraction of hardening
precipitates which form during
natural ageing
Appendix 6.2
Additivity in relation to the Avrami Equation
In order to prove that equation (632) is the isokinetic version of the Avrami equation, we first
need to rewrite it in a differential form. From equation (625), we have:
(A61)
Substituting
(A62)
(A63)
(A64)
(A65)
Provided that equation (A65) contains separable variables of X and t* (T), it can be integrated as follows:
(A66)
7
Properties of Weldments
7.1 Introduction
Weldments are prime examples of components where the properties obtained depend upon the
characteristics of the microstmcture. Since failure of welds often can have dramatic consequences, a wealth of information is available in the literature on structureproperty relationships. However, in order to fit some of the apparently conflicting results into a more consistent
picture, a theoretical approach is adopted here rather than a review of the literature. This
procedure also involves the use of phenomenological models for the quantitative description
of structureproperty relationships in cases where a full physical treatment is not possible.
dislocation density and small lath size. The formation of large proportions of upper bainite,
Widmanstatten ferrite, or grain boundary ferrite, on the other hand, are considered detrimental
to toughness, since these structures provide preferential crack propagation paths, especially
when continuous films of carbides are present between the ferrite laths or plates. Attempts to
control the weld metal acicular ferrite content have led to the introduction of welding
consumables containing complex deoxidisers (Si, Mn, Al, Ti) and balanced additions of various alloying elements (Nb, V, Cu, Ni, Cr, Mo, B).
The final weld metal microstructure will depend on complex interactions between several
important variables such as:3"5
(i)
(ii)
The relative contribution from each is determined by the steel chemical composition and
the weld thermal history. Because of the number of variables involved, a full physical treatment of the problem is not possible. Consequently, the simplified treatment of Gladman and
Pickering8 has been adopted here.
Figure 7.1 shows the individual strength contributions in lowcarbon bainite, which is the
dominating microconstituent in asdeposited steel weld metals (includes both upper and lower
bainite as well as acicular ferrite). Firstly, there are the solid solution strengthening increments
from alloying and impurity elements such as manganese, silicon and uncombined nitrogen,
which in the present example correspond to a matrix strength of about 165 MPa. Secondly, the
grain size contribution to the yield stress is shown as a very substantial component, the magnitude of which is determined by the bainite lath size. Finally, a typical increment for dispersion
strengthening is indicated. This contribution is negligible at large lath sizes typical of upper
bainite, but becomes significant at small grain sizes because of a finer intralath carbide dispersion.8 Hence, in steel weld deposits containing high proportions of acicular ferrite or lower
bainite carbides will make a direct contribution to strength, even at relatively low carbon levels.
The results in Fig. 7.1 are of significant practical importance, since they show the inherent
limitations of the system with regard to the maximum strength that can be achieved through
control of the microstructure. As shown in Section 6.3.5.4 (Chapter 6), the typical lath size
(width) of acicular ferrite in lowalloy steel weld metals is about 2 jam. According to Fig. 7.1,
this corresponds to a maximum yield strength of approximately 650 MPa, which is in good
agreement with the observed threshold strength of acicular ferrite containing steel weld deposits.9 If higher strength levels are required, it is necessary to decrease the grain (lath) size
through a refinement of the microstructure, i.e. by replacing acicular ferrite with either lower
p 1/4
f/o
Number of carbides per mm Nv (mm" )
Dispersion
Grain size
Matrix strength
1/2
Bainitic ferrite grain size, mm
Fig. 7.1. Contributions to strength in lowcarbon bainite. Data from Gladman and Pickering.8
bainite or martensite. Development along these lines has led to the introduction of a new
generation of high strength steel weld metals with a yield strength in the range from 650 to 900
MPa.10'11
7.2.1.2 Weld metal resistance to ductile fracture
It is well established that the weld metal resistance to ductile fracture is strongly influenced by
the volume fraction, shape, and size distribution of nonmetallic inclusions.12"15 Although a
verified quantitative understanding of the fracture process is still lacking, there seems to be
general agreement that it involves the following three basic steps:16
(i)
(72)
where 7^ is the surface energy of the particle, Ep is the Young's modulus of the particle, A is the
stress concentration factor at the particle, and dv is the particle diameter.
GMAW
(E=1.6kJ/mm)
Fig. 7.2. Variation of true fracture strain /with inclusion volume fraction Vv. Data from Widgery.12
SAW
inclusions appear to be preferential nucleation sites for microvoids in lowalloy steel weld
metals (see Fig. 7.5(b)). Although the combined effect of particle size and local stress concentration on the ductile fracture behaviour cannot readily be accounted for in a mathematical
simulation of the process, the CVN data in Fig. 7.6 suggest that the content of large inclusions
(e.g. of a diameter greater than about 1.5 Jim) should be minimised in order to maintain a high
resistance against dimpled rupture. In practice, this requires careful control of the weld metal
aluminiumoxygen balance and the heat input applied during welding (see Section 2.12 in
Chapter 2).
Effect of strength level
The toughness of a material reflects its ability to absorb energy in the plastic range. One way of
looking at toughness is to assume that it scales with the total area Uj under the stressstrain
curve. Several mathematical expressions for this area have been suggested. For ductile materials we may write:19
(73)
where Rm is the ultimate tensile strength (UTS).
If Uj is regarded as a material constant, one would expect that Rm and y are reciprocal
Total inclusion
population
SAW
Frequency, %
Frequency, %
Inclusions associated
with dimples
Frequency, %
Total inclusion
population
(b)
SAW
Inclusions associated
with dimples
Frequency, %
(a)
Fig. 7.5. Histograms showing the size distribution of nonmetallic inclusions in the weld metal and in the
centre of microvoids at the fracture surface, respectively; (a) Low aluminium level (Alcontaining manganese silicate inclusions), (b) High aluminium level (AI2O3 inclusions). Data from Andersen.18
High Ti levels
CVN upper shelf energy, J
Medium Ti levels
Low Ti levels
SAW
Nv (d v >1.5 um)105
Fig. 7.6. Correlation of CVN upper shelf energy with number of particles per mm3 greater than 1.5 urn,
Nv(dv > 1.5 um). Data from Grong and Kluken.15
It is evident from the tensile test data in Fig. 7.7 that the fracture strain is a true function of
Rm, although the relationship appears to be linear rather than nonlinear, as predicted by equation (74). These results are of considerable practical importance, since they imply that the
upper shelf energy absorption, and hence, the shape of the CVN transition curve is strongly
affected by the weld metal strength level. Accordingly, control of the weld metal microstructure becomes particularly urgent at high strength levels to avoid problems with the cleavage
fracture resistance (to be discussed below).
7.2.1.3 Weld metal resistance to cleavage fracture
Cleavage fracture is characterised by very little plastic deformation prior to the crack propagation, and occurs in a crystallographic fashion along planes of low indicies, i.e. of high atomic
density.1 Bodycentred cubic (bcc) iron cleaves typically along {100} planes, which implies
that the cracks must be deflected at high angle grain (or packet) boundaries, as shown
schematically in Fig. 7.8. Consequently, in steel weld metals the ferrite grain size and the
bainite packet width are the main microstructural features controlling the resistance to cleavage crack propagation.
Since the microstructure which forms within each single austenite grain will not be uniform
but a complex mixture of two or more constituent phases, it is difficult, in practice, to define
a meaningful grain size or packet width. For this reason, most investigators have attempted to
correlate toughness with the presence of specific microconstituents in the weld metal.3"5 For
example, an increase in the volume fraction of acicular ferrite will result in a corresponding
increase in toughness (i.e. decrease in the CVN transition temperature), as shown in Fig. 7.9.
(a)
(b)
Fig. 7.8. Schematic diagrams showing cleavage crack deflection at interfaces; (a) High angle ferriteferrite grain boundaries, (b) High angle packet boundaries (bainitic microstructures).
Transition temperature, 0C
(75)
where En is the Young's modulus of the matrix, ye^ is the effective surface energy (equal to
the sum of the ideal surface energy and the plastic work), and c is the half crack length.
Since c is proportional to the particle diameter dv, equation (75) predicts that welds containing large inclusions should be more prone to cleavage cracking than others. This result is
also in agreement with general observations. For example, in selfshielded FCA steel weld
metals it has been demonstrated that cleavage crack initiation is usually associated with large
aluminiumcontaining inclusions which form in the molten pool before solidification (see Fig.
7.10). Consequently, control of the inclusion size distribution is essential in order to ensure an
adequate lowtemperature toughness.
7.2.1.4 The weld metal ductile to brittle transition
In addition to the parameters mentioned above, there are several other factors, some interrelated, which play an important part in the initiation of cleavage fracture. These are:1
(i)
(a)
(b)
(C)
Fig. 7.10. Initiation of cleavage fracture in a selfshielded FCA steel weld from an aluminiumcontaining
inclusion; (a) Initiation site short distance ahead of the notch, (b) Detail of initiation site showing cracked
inclusion, (c) Detail of cracked inclusion (remnants of particle are left in the hole).
(ii)
(iii)
(iv)
In general, this picture is too complicated to establish a physical framework within which
the various theoretical models for the ductile to brittle transition in steel can be embedded. We
are therefore forced to base our judgement and understanding of how key parameters affect the
position and shape of the CVN transition curve solely on scattered phenomenological observations and empirical models (e.g. see the reviews of Grong and Matlock3 or Abson and Pargeter4).
An example of how far the latter approach has been developed is given below.
Akselsen and Grong20 have established a series of empirical equations which relates toughness to the weld metal acicular ferrite content and the ultimate tensile strength (UTS). Figures
7.11 and 7.12 show how each of these parameters influences the CVN transition curve. It is
evident from the diagrams that control of the weld metal acicular ferrite content becomes
particularly important at high strength levels to avoid problems with the fracture toughness. In
cases where undermatch is aimed at (i.e. a weld metal to base plate strength ratio less than
unity), the weld metal tensile strength is typically of the order of 450 to 550 MPa. Within this
range a volume fraction of acicular ferrite beyond 25 vol% will generally be sufficient to meet
current toughness requirements (35 J at 40 0 C). If overmatch is desired, the volume fraction
of acicular ferrite becomes more critical, partly because of a higher weld metal strength level
and partly because of more stringent toughness requirements (e.g. 45 J rather than 35 J at
40 0 C). Process diagrams of the type shown in Figs. 7.11 and 7.12 can therefore serve as a
basis for proper selection of consumables for welded steel structures.
It should be noted that Akselsen and Grong20 in their analysis omitted a consideration of the
important influence of free (uncombined) nitrogen and nonmetallic inclusions on the CVN
transition curve. Based on the experimental data in Fig. 7.13 it can be argued that such
compositional variations can be equally detrimental to toughness as a decrease in the acicular
ferrite content. Consequently, further refinements of the models are required if a verified
quantitative understanding of the ductile to brittle transition in lowalloy steel weld metals is
to be obtained.
Example (7.1)
Consider multipass FCA steel welding with two different electrode wires, one with titanium
additions and one without. Table 7.1 contains a summary of weld metal chemical compositions. Provided that the microstructure and the inclusion size distribution are similar in both
cases, use this information to evaluate the lowtemperature toughness of the welds, as revealed
by CVN testing.
Solution
Since the nitrogen content of both welds is quite high (0.011 wt%), the risk of a toughness
deterioration due to strain ageing is imminent, particularly at low Ti levels. Taking the atomic
weight of titanium and nitrogen equal to 47.9 and 14.0 g mol"1, respectively, the stoichiometric
amount of titanium that is necessary to tieup all nitrogen as TiN can be calculated as follows:
WeIdA
In weld A most of the nitrogen is free (uncombined) due to an unbalance in the titanium content. This means that the risk of a toughness deterioration due to strain ageing is high.
(a)
Absorbed energy, J
35 Joules
Test temperature, 0C
(b)
Absorbed energy, J
35Joules.
Test temperature, 0C
Fig. 7.11. Predicted effect of weld metal acicular ferrite content on the CVN transition curve at two
different tensile strength levels; (a) Rm = 600 MPa, (b) Rm = 800 MPa. Data from Akselsen and Grong.20
WeIdB
Weld B contains 0.030 wt% Ti, which is not far from the stoichiometric amount of titanium
necessary to tieup all nitrogen. Although some titanium also is bound as Ti2O3, it is reasonable to assume that the free nitrogen content in this case is sufficiently low to eliminate problems with strain ageing. Consequently, weld B would be expected to exhibit the highest toughness (i.e. the lowest CVN transition temperature) of the two, as indicated in Fig. 7.14.
(a)
25 vol% acicular ferrite
Absorbed energy, J
UTS
35 Joules
Test temperature, 0C
(b)
Absorbed energy, J
75 v o l % acicular ferrite
UTS
..35J.Q.ute$
Test temperature, 0 C
Fig. 7.12. Predicted effect of weld metal ultimate tensile strength (UTS) on the CVN transition curve at
two different volume fractions of acicular ferrite; (a) 25 vol% AF, (b) 75 vol% AF. Data from Akselsen
and Grong.20
Table 7.1 Chemical composition of FCA steel weld metals considered in Example (7.1).
Element
Weld
wt% C
wt% Si
wt% Mn
wt% Al
wt% Ti
wt% S
wt% N
wt% O
0.10
0.40
1.50
0.005
0.006
0.008
0.011
0.031
0.10
0.40
1.50
0.005
0.030
0.008
0.011
0.031
(a)
SMAW (basic electrodes)
Absorbed energy, J
Absorbed energy, J
Absorbed energy
WeIdB
WeIdA
35 J
Test temperature, 0C
Fig. 7.14. Schematic drawings of the CVN transition curves for welds A and B (Example (7.1)).
(a)
(b)
Fig. 7.15. Typical lowtemperature fracture modes of TiB containing steel weld metals; (a) Quasicleavage (aswelded condition), (b) Intergranular fracture (after PWHT).
(ii)
For these reasons local PWHTs are commonly required for all structural parts above a
specified plate thickness (e.g. 50 mm according to current North Sea offshore specifications).
Postweld heat treatment is usually carried out in the temperature range from 550 to 6500C.
In practice, however, the toughness achieved will depend on the weld metal chemical composition, and in some cases deterioration rather than improvement of the impact properties is
observed after PWHT. In such cases the reduction in toughness can be ascribed to:3'4
(i)
(ii)
Segregation of impurity elements (e.g. P, Sn, Sb and As) to prior austenite grain
boundaries, which, in turn, can give rise to intergranular fracture. The indications
are that this type of embrittlement is strongly enhanced by the presence of second
phase particles at the grain boundaries.
Experience shows that TiB containing steel weld metals often fail by intergranular fracture in the columnar grain region after PWHT,23 as evidenced by the SEM fractographs in Fig.
7.15. The observed shift in the fracture mode is associated with a significant drop in toughness
(Fig. 7.16) and arises from the combined action of solidificationinduced phosphorus
segregations and borocarbide precipitation along the prior columnar austenite grain bounda
ACVN, J
SAW
Base line
Fig. 7.17. TEM micrograph showing precipitation of borocarbides, Fe23(B,C)6, along prior austenite
grain boundaries in a TiB containing steel weld metal after PWHT (600 0 CIh).
ries (Fig. 7.17). Since borocarbides are brittle and partly incoherent with the matrix, they can
be regarded as microcracks (of length dp) ready to propagate. In such cases there is virtually
no plastic deformation occurring before crack propagation, which implies that the intergranular
fracture stress is given by the Griffith's equation:24
(76)
where 7 ^ is again the effective surface energy (equal to the sum of the ideal surface energy
and the plastic work), and dp is the particle diameter.
Although the value of yeg_ would be expected to be low in the presence of solidificationinduced phosporus segregations,24 this alone is not sufficient to initiate intergranular fracture
in the weld metal. However, during PWHT the borocarbides will start to grow from an initially small value to a maximum size of about 0.1 to 0.2jim (Fig. 7.17), following the classic
growth law for grain boundary precipitates dpatl/4?5 This implies that the intergranular fracture stress, Oj(I), will gradually decrease with increasing annealing times, as indicated in Fig.
7.18. When the matrix fracture strength, Cj(M), is reached, the fracture mode shifts from
predominantly quasicleavage in the aswelded condition (Fig. 7.15(a)) to intergranular rupture after PWHT (Fig. 7.15(b)). This is observed as a marked reduction in the CVN toughness,
as shown previously in Fig. 7.16.
7.2.2 HAZ mechanical properties
The last twenty years have seen a revolution in the metallurgical design of steel. Whereas old
steels relied on the use of carbon for strength, the trend today is to rely more on grain refinement in combination with microalloy precipitation to meet the current demand for an improved weldability. This includes both the sensitivity to weld cracking and the HAZ mechanical properties required by service conditions and test temperatures. The latter aspect is of
particular interest in the present context and will be discussed later.
Stress
Intergranular
fracture mode
Quasicleavage
fracture mode
Particle diameter
[Annealing time]174
Fig. 7.18. Schematic illustration of the mechanisms of temper embrittlement in TiB containing steel
weld metals (Gf(M): matrix fracture strength, (*/(/): intergranular fracture strength).
p0.2> R nv M P a
R
HV 5 ,kp/mm 2
Cooling time, At 8 / 5 , s
Fig. 7.19. Structureproperty relationships in the grain coarsened HAZ of lowcarbon microalloyed steels
(vol% M: martensite content, Rp : 0.2% proof stress, Rm: ultimate tensile strength, HV5: Vickers hardness, A%5.* cooling time from 800 to 5000C). Data from Akselsen et al.26
A number of different empirical models exist in the literature for prediction of HAZ peak
hardness and strength.26"31 However, the aptness of some of these models is surprisingly
good, which justifies construction of isohardness and isostrength diagrams for specific grades
of steels.32 Examples of such diagrams are given in Fig. 7.20. It is
Lebih dari sekadar dokumen.
Temukan segala yang ditawarkan Scribd, termasuk buku dan buku audio dari penerbitpenerbit terkemuka.
Batalkan kapan saja.