LITH-IKP-EX--05/2302--SE
November 2005
Master Thesis in Solid Mechanics Niklas Karlsson, Per-Henrik Lenander Analysis of Fatigue Life in Two Weld Class Systems LITH-IKP-EX--05/2302--SE Department of Mechanical Engineering Linkping University SE-581 83 Linkping, Sweden Printed in Sweden by UniTryck, Linkping, 2005
ii
iii
iv
Abstract
In current welding standards, there is a lack of connection between acceptance limits and fatigue life. In an ideal standard there should be a clear and consistent connection, assuring that a certain welding class always implies a certain fatigue life of the welded joint. Volvo Construction Equipment is currently involved in reworking the company welding standard STD5605,51, aiming at introducing such a fatigue connection in the standard. The objective of this thesis work is to provide the basic data for reworking the standard, i.e. to calculate fatigue lives for the defect types in the current welding standard. To extend the study, the corresponding ISO standard ISO5817 is studied as well. For the fatigue life calculations, FEM (finite element method) and LEFM (linear elastic fracture mechanics) are used. A few other methods are briefly described and quantitatively compared. The results show a very scattered acceptance limit dependence for the fatigue lives in the different defect cases. This implies that the acceptance limits in most cases need to be revised. Furthermore, some cases should be removed from the standard and some cases from the ISO standard could be included in the Volvo standard.
vi
Preface
This thesis has been performed as the final assignment for the examination as Master of Science at Linkping University. The thesis work was initiated by and carried out at Volvo Articulated Haulers in Bras, Sweden, between June and November of 2005. The intention is to provide a foundation for further work on reviewing the companys welding standard. Finite element modelling and calculation of the fatigue lives of all applicable defect types included in the Volvo standard, as well as in the corresponding ISO standard, has been performed. Applied theories and methods are thoroughly described in theory chapters, but the reader is assumed to have basic knowledge in solid mechanics. We would like to thank the following people at Volvo Articulated Haulers in Bras, who have all contributed to our thesis work: Our supervisor M.Sc. Bertil Jonsson, for invaluable support; weld auditor Stefan Stlberg, for hands-on experience on the shop-floor; Quality engineer Stig Malmqvist, for sharing his expertise on welding standards; everybody at Helfordonsgruppen, for a memorable time. We would also like to thank our examiner Prof. Tore Dahlberg at the Division of Solid Mechanics, Department of Mechanical Engineering, Linkping University, for reading and commenting the entire thesis, and our opponents Mr Mats Andersson and Mr Mattias Danielsson for their valuable opinions. Bras in November 2005
vii
viii
Notations
The following constants, functions and variables are used in this thesis. The constants, functions and variables not listed here are explained in the text. Throughout the thesis variables are written in italic and constants in plain text.
Contents
Abstract ..................................................................................................................................... v Preface ..................................................................................................................................... vii Notations .................................................................................................................................. ix Contents.................................................................................................................................... xi 1 Introduction ...................................................................................................................... 1 1.1 The company .................................................................................................................. 1 1.2 Background .................................................................................................................... 1 1.3 Objectives....................................................................................................................... 1 1.4 Procedure........................................................................................................................ 1 1.4.1 Restrictions............................................................................................................ 2 1.5 Outlines of the report...................................................................................................... 2 2 Fatigue life calculation ..................................................................................................... 5 2.1 Basic theory on LEFM ................................................................................................... 5 2.1.1 Loading of cracks .................................................................................................. 5 2.1.2 Stress intensity factor ............................................................................................ 5 2.1.3 Paris law .............................................................................................................. 6 2.1.4 Requirements......................................................................................................... 7 2.1.5 Plane stress versus plane strain ............................................................................ 7 2.2 Other methods ................................................................................................................ 7 2.2.1 Type of fracture ..................................................................................................... 9 2.2.2 S-N curves ............................................................................................................. 9 2.2.3 The Swedish standard.......................................................................................... 10 2.2.4 The nominal stress method.................................................................................. 11 2.2.5 Example of the nominal stress method................................................................ 12 2.2.6 Comparison with the Swedish standard .............................................................. 13 2.2.7 Problems with interpreting the standard ............................................................ 13 2.2.8 The Hot spot method ........................................................................................... 15 2.2.9 Example of the Hot spot method ......................................................................... 16 2.2.10 The effective notch method.................................................................................. 18 2.2.11 Example of the notch method .............................................................................. 18 2.2.12 Linear Elastic Fracture Mechanics..................................................................... 20 2.3 Comparison of methods ............................................................................................... 20 2.3.1 Advantages with LEFM....................................................................................... 21 2.3.2 Disadvantages with LEFM.................................................................................. 21 3 Procedure and methods ................................................................................................. 23 3.1 Modelling ..................................................................................................................... 23 3.1.1 Basic modelling ................................................................................................... 23 3.1.2 Boundary conditions ........................................................................................... 23 3.1.3 Influence of boundary conditions........................................................................ 25 3.2 Meshing........................................................................................................................ 26 3.2.1 Achieving a meshable model ............................................................................... 26 3.2.2 Elements .............................................................................................................. 27 3.2.3 The crack tip........................................................................................................ 27 3.2.4 The box ................................................................................................................ 29 3.2.5 The centre lines ................................................................................................... 30 3.2.6 Remaining parts of the model.............................................................................. 31 3.2.7 Meshing problems ............................................................................................... 31 xi
3.2.8 Convergence........................................................................................................ 32 3.2.9 Macros................................................................................................................. 33 3.3 Postprocessing.............................................................................................................. 34 3.3.1 How stress intensity factors are calculated in ANSYS ........................................ 34 3.3.2 Numeric integration ............................................................................................ 36 3.4 Verification of integration method ............................................................................... 38 3.4.1 Approximation with upper and lower summations ............................................. 40 3.4.2 Spline values integrated in Excel ........................................................................ 41 3.5 Analytical case for an internal crack ............................................................................ 42 3.5.1 Crack growth rate ............................................................................................... 43 3.5.2 Algorithm............................................................................................................. 44 3.5.3 Discussion about integration limits..................................................................... 44 3.5.4 Comparison with AFGROW................................................................................ 46 Weld auditing.................................................................................................................. 47 4.1 Overall on weld auditing .............................................................................................. 47 4.2 Review of previous weld audits ................................................................................... 47 4.3 Weld audit on a rear frame ........................................................................................... 49 4.3.1 Profile projector measuring of silicone impressions .......................................... 49 4.3.2 Comments ............................................................................................................ 50 Compilation of STD5605,51 .......................................................................................... 51 5.1 About the standard ....................................................................................................... 51 5.2 Modelling ..................................................................................................................... 52 5.3 Results .......................................................................................................................... 52 Compilation of ISO5817 .............................................................................................. 145 6.1 About the standard ..................................................................................................... 145 6.2 Modelling ................................................................................................................... 145 6.3 Results ........................................................................................................................ 145 Study of sheet thickness dependence .......................................................................... 221 7.1 Introduction ................................................................................................................ 221 7.2 Studied geometry - STD5605,51 Case 20.................................................................. 221 7.2.1 Requirements..................................................................................................... 222 7.3 Modelling ................................................................................................................... 223 7.3.1 Dimensions ........................................................................................................ 224 7.4 Influence of sheet length ............................................................................................ 224 7.5 Results ........................................................................................................................ 226 7.6 Comments on the results ............................................................................................ 230 7.6.1 The thickness effect............................................................................................ 231 7.6.2 Effect of absolute acceptance limits .................................................................. 231 7.6.3 Sheet length ....................................................................................................... 231 7.6.4 The bending case............................................................................................... 231 7.7 Conclusions ................................................................................................................ 232 7.8 Theory and calculations on the thickness effect......................................................... 232 Conclusions and discussion ......................................................................................... 237 8.1 Conclusions ................................................................................................................ 237 8.2 General discussion...................................................................................................... 237 8.3 Proposals for revising the standards........................................................................... 238 8.3.1 Primary proposal for guidlines ......................................................................... 239 8.3.2 Secondary proposal for guidelines.................................................................... 239 8.3.3 Other proposals for STD5605,51 ...................................................................... 239 8.3.4 Other proposals for ISO5817............................................................................ 239
xii
8.4 Discussion for STD5605,51 ....................................................................................... 239 8.4.1 Cases which could be added or removed from STD5605,51 ............................ 240 8.4.2 Recommendations for achieving proposals for STD5605,51............................ 241 8.5 Discussion for ISO5817 ............................................................................................. 242 8.5.1 Cases which could be added or removed from ISO5817 .................................. 242 8.5.2 Recommendations for achieving proposals for ISO5817.................................. 243 8.6 Recommendations for further studies ........................................................................ 244 References ............................................................................................................................. 247 Appendix A ........................................................................................................................... 249 Appendix B diagrams for STD5605,51............................................................................ 251 Appendix C diagrams for ISO5817 ................................................................................. 271
xiii
xiv
1 Introduction
1 Introduction
1.1 The company
The thesis work was carried out at Volvo Articulated Haulers in Bras, 30 kilometres northeast of Vxj in southern Sweden. It has recently been merged with Volvo Wheel Loaders into Volvo HLBL (Hauler Loader Business Line), which is a part of Volvo Construction Equipment. In the Bras factory articulated haulers are designed and manufactured. An articulated hauler is most commonly used to transport gravel and rocks in rough terrain during road and building construction, but also other applications are available. Volvo CE markets articulated haulers in a payload range of 25 to 40 tons. In Bras the 650 employees currently produce over 2 000 machines yearly. There is also a factory situated in Pederneiras, Brazil. Volvo CE has approximately 40% of the world market for articulated haulers.
1.2 Background
The background of this thesis is the ongoing efforts to review the Volvo company welding standard, STD5605,511 [1]. An important objective of this is to achieve a clear and consistent connection between the acceptance limits in the standard and the fatigue lives of the welded structures. The welding standard contains descriptions of a number of possible weld defects, with limits for the accepted dimensions of the defects for each welding class. In the Volvo standard, four welding classes, A to D, can be used for assigning suitable requirements when designing a weld joint. They also contain additional designations, for example U for fatigue loaded welds. Reworking the welding standard requires a large amount of quantitative data on current fatigue lives to be calculated. This thesis was initiated to produce all this data as well as additional knowledge on fatigue life issues for weld joints.
1.3 Objectives
The main objective of the thesis work is to provide life calculations for all interesting defect types, in all welding classes, both for the Volvo standard STD5605,51 and for the corresponding ISO standard ISO5817 [2]. The ISO standard is studied in order to take advantage of possibly useful features. The fatigue life data is presented in tables and diagrams, and also thoroughly commented in the text. As an effect of the extensive work on modelling and calculating all cases, a lot of knowledge of various factors affecting the fatigue life has been achieved and will be presented in the report. The influence of sheet thickness and bending loads will be separately investigated. It is the authors objective to provide general recommendations for how the standards could be revised; primarily for the Volvo standard, but since ISO5817 has also been thoroughly reviewed, it will be discussed as well.
1.4 Procedure
A specified set of interesting defect cases in the two standards are being investigated. For each case the geometry is modelled, stress intensity factors are obtained, and finally the fatigue life is calculated.
1 Previously STD5605,51 was called 5.501E. It is now also known as STD181-0001. Throughout this work the Volvo CE weld standard will be referred to as STD5605,51
For a normal type of defect (a few exceptions exist), the weld joint geometry with a defect of the current type and an initial crack is modelled in the FE programme ANSYS. A macro is then used to let a crack propagate in certain steps into the model. For every crack length the FE problem is solved and the stress intensity factors are calculated. The results are used to determine the stress intensity factor as function of crack length for the particular case. Then this function is used when calculating the fatigue life by use of fracture mechanics (integration of Paris law). The life is assumed to be finished when the crack reaches half the sheet thickness. All steps of the procedure are described thoroughly in the following chapters. In a few cases, especially for some internal defects, special methods for estimating the crack propagation have to be applied. This is further discussed in theory chapters and case descriptions.
1.4.1 Restrictions
It should be emphasised that all results in this report are theoretical, based on the exact conditions that are given in the description of each case. In reality, for example, geometries are never perfect, material quality varies and loading conditions can be very complex. All geometries have been modelled in 2-D. Thus, the depth direction is not considered, except in the analytical case used for internal cracks. An important restriction is that an initial crack length of 0.1 mm is assumed. This means that no life during the crack initiation phase is taken into account. All calculations are performed on 10 mm thick sheets. Applying the results to other thicknesses can not be done without consideration. This is investigated and discussed in Chapter 7. Occasionally it is commented in the report that some issue is not fully investigated (normally due to lack of time or due to the problem lying outside the scope of this work). In these cases conclusions are based on reasonable assumptions, given in the text.
1 Introduction Conclusions and discussion are presented in Chapter 8. The authors give their recommendations for how the standards could be reworked to obtain a better connection between welding class and fatigue life. A few other important observations and conclusions are also provided, as well as suggestions for further research in the area. Appendix A contains the MATLAB program for calculating stress intensity factors for the analytical case with an elliptical inner crack. In Appendices B and C, tables and diagrams of results for STD5605,51 and ISO5817 can be found respectively.
Mode I
Mode II
Mode III
Figure 2.1. Three different modes a crack can be loaded in. In Mode I the crack is opened. Mode I is the most dangerous way to load the crack because this loading case generates the greatest stress intensity of all loading cases. In Mode II and III the crack is sheared in two different planes.
where 0 is the nominal stress, a the crack length, and f is a function of geometry and loading. Numerous analytical cases have been derived giving the f-function. Some of them can be found in, for example, Formelsamling i hllfasthetslra, reference [4]. The range of the stress intensity factor can be calculated as
K I = K I max K I min
6 where
K I = K I max
K I = K I max K I min
if K I min 0
if K I min > 0
if K I max 0
K I = 0
The stress intensity range KI is used in Paris law when calculating the number of cycles to failure, see Equation (2.1). Stress intensity factors in the same mode can be added by superposition, i.e., K Itotal = K IA + K IB + K IC + ...
If a crack is loaded in several modes at the same time an effective stress intensity factor, K eff , can be calculated as
2 2 = K I2 + K II + K eff
where
= 3 4
3 1 +
and is Poissons ratio. There are also other suggestions for calculating the effective stress intensity factor.
log K th
log K Ic
log K I
Figure 2.2. The crack propagation rate versus the stress intensity range.
For K I < K th the crack is so short or the load so small that the crack will not propagate. This implies that K th is a threshold value for crack growth. When K I = K Ic the length of the crack or the nominal stress is so large that fracture will occur momentarily at plain strain. Therefore K Ic is the critical value for crack growth. The linear part of the curve in Figure 2.2, between K th and K Ic , is described by Paris law and it is used when calculating the number of cycles to failure for a component. Paris law describes the increment of the crack growth for every cycle, i.e. the crack propagation rate or the increment in length per cycle. The crack propagation rate is a function of the stress intensity range, K I . One has
da n = C (K I ) dN (2.1)
where a is the length of the crack, N is the number of cycles, and C and n are material constants. When calculating the life, one has to integrate Paris law, which is a separable differential equation. Since K I is a function of the above mentioned f-function and the f-function usually is very difficult to integrate analytically, a numerical integration technique is normally preferred.
2.1.4 Requirements
The following requirements must be fulfilled in order to use LEFM at plane strain for the final fracture.
a final K Ic t 2.5 Y W a final
2
If these requirements are not fulfilled, a non-linear model or a linearized non-linear model must be used, for example the Irwin approach or the Dugdale model.
linear elastic fracture mechanics (LEFM). Theory on all methods except LEFM are taken from the reference Svetsutvrdering med FEM [5] All methods in this chapter use data from the International Institute of Welding (IIW) [6] for calculating the life of the weld. Data from IIW are based on real welded structures which have been exposed to cyclic loading. Values from these tests have been plotted in an S-N-diagram and a FAT value (FATigue) of the stress for sustaining 2 million cycles has been calculated. This FAT value gives a failure probability of 2.3% for the weld. Two examples are used when describing the methods. Both are taken from IIWs Fatigue designs for welded joints and components [6]. The first example is a transverse butt weld, Case 213 (see Figure 2.3) and the second example is a cruciform joint, Case 413 (see Figure 2.4). t crack R
t crack t R 0.15t
w Figure 2.4. Case 413, cruciform joint. Both cases are modelled with a transition radius R = 1 mm. This is the smallest radius which can be expected for a normal weld without any subsequent machining [5]. A comparison with the Swedish standard is performed.
log FL
log N
Figure 2.5. S-N curve.
August Whler found that the stress amplitude described the fatigue life better than the maximum stress. The stress amplitude is defined as
max min 2
= max min
For welds it is a more appropriate approach to use the stress range for fatigue analysis. For stresses above the ultimate strength, UTS , the component fractures immediately while for stresses below the fatigue limit, FL , the component has a theoretical infinite life at constant amplitude loading.
10
is used instead of a for welds. The reason for this is that residual stresses close to the
yield limit Y can be expected in both tension and compression when the metal cools in the weld. This means that if, for example, compressive stresses are applied to a weld where the residual stress is near the yield limit in tension, the weld will endure a positive stress range even though weld is loaded in compression. The root for fillet welds often have negative residual stresses while toe cracks have positive residual stresses, but this may vary. A conservative approach when calculating the fatigue life of welds, is thus to use the stress range . When the residual stresses are known, correction factors for compensating the worst case scenario can be found in, for example, [6]. The correction factors are functions of the stress ratio, which is defined as
R= min max
These correction factors are greater than one and are multiplied with the FAT-value in order to receive a longer fatigue life. This may be used for the ground material and to some extent on simple welds. The correction factors are not used for complex structures. To calculate the number of cycles to failure, the straight line with slope m can be used. The equation log N = m log + log C gives the life
N= C C = m N m
The IIW uses a FAT-value of the stress which has been selected so that a certain component should sustain 2 million cycles before failure. This gives the following equation for calculating the number of cycles to failure for a certain range of the stress:
m 6 C = FAT 2 10 FAT m 2 10 6 = m N m C = N
FAT 6 N = 2 10
(2.2)
Equation (2.2) makes it possible to use the tested cases in the IIWs handbook [6] to calculate the life of a component.
C 6 nt = 2 10 f rk
(2.3)
When Equation (2.3) is compared to (2.2) one can see that it is the same equation, only the notations differ. N is the same as nt , f rk equals and C corresponds to the FAT-value.
11
Fatigue lives for a number of cases, as for the IIW standard, have been tested. The major difference is that the Swedish standard also deals with three different welding classes from A to C, where class A gives the longest life. However the descriptions of the classes are vague, without any actual acceptance limits. This makes it difficult to define which class the component belongs to. Therefore the predicted number of cycles may differ a lot, depending on which class is chosen. The BSK also offers the opportunity to predict the fatigue life for varying stress amplitude with a special equation which takes the new slope of the curve into consideration, see Figure 2.5. The BSK can only be used to evaluate the fatigue life for nominal stresses, i.e. it does not support the hot spot and effective notch methods.
nom
Figure 2.6. The nominal stress, nom , extrapolated from the true stress at the surface.
12
For evaluation of fatigue life with nominal stress a large number of cases with associated FAT-values are available from the IIW, [6].
Nominal method for Case 213 220 200 180 1st principal stress [MPa] 160 140 120 100 80 60 40 20 0 0 2 4 6 x [mm] 8 10 12 Start of transistion radius End of transistion radius
nom
nom
200
150
100
50
6 x [mm]
10
12
14
Figure 2.8. Nominal method Case 413 in IIW. Since there is no bending moment here, the nominal stress is equal to the applied stress, i.e. 104 MPa for Case 213 and 82 MPa for Case 413. The FAT-value is 80 MPa for Case 213 respectively 63 MPa for Case 413 but both these values have been multiplied with a factor of
13
1.3 ( Q ) to achieve 50% probability of failure instead of 2.3% which is the case for the FATvalue. The life for the welded structure can be calculated with the modified Equation (2.2) as N nom
Q FAT 6 = 2 10 nom
(2.4)
which gives a life of 2 106 cycles for both Case 213 and 413.
14
nom
nom
Figure 2.9. Free-body diagram of the cruciform joint in Case 413. Note that the possible misalignment has been disregarded. The following stresses can be derived
: = : = = nom t 82 10 = 58 MPa 2 2a 2 2 5
When the stresses, ( rd ) and ( rd ) are known, these can be used to calculate the utilization factor for multiaxial stress state, which can be calculated with the following equation
+ 1.1 z= , allowed , allowed , allowed = 56 MPa ( f rd ) for Case 30 WB [8] and , allowed = 0.6 , allowed ( f rvd ). Both , allowed and , allowed are multiplied with the factor Q = 1.3 for 50% probability of failure, which gives 58 58 z= + 2.18 1.3 56 1.3 0.6 56 The corresponding uniaxial stress can now be calculated as f rk = z 2.18 C = 56 111 MPa 1.1 1.1
2 2
This stress is considerably higher than the nominal stress, nom = 82 MPa. Now the fatigue life can be calculated as C 56 6 6 6 nt = f 2 10 = 111 2 10 0.25 10 cycles rk
3 3
15
which is considerably lower than the 1.40 10 6 cycles achieved by use of nominal stress. This is a very conservative way to calculate the life but enlighten the problems which arises when there are almost no instructions given to the user of the standard. If the probability factor, Q , is inserted when the number of cycles are calculated instead of in the utilization factor, z, as described below 58 58 z = + 2.01 56 0.6 56 f rk =
3 2 2
QC 1.3 56 6 6 6 nt = f 2 10 = 102 2 10 0.72 10 rk the fatigue life becomes almost three times as high. Again this confirms the difficulties with predicting the life.
16
Extrapolation points
hs
0.4t 1.0t Figure 2.10. The Hot spot stress, hs , linear extrapolation. Different equations are used to extrapolate the stress. The following equation describes linear extrapolation. nom = 1.67 0.4t 0.671.0t (2.5)
where 0.4t is the stress at 0.4 times the thickness from the toe, etc. The first extrapolation point, 0.4t , has been chosen since the stress here is not affected by the weld toe geometry.
17
hs
Extrapolation points
Hot spot method for Case 413 300 Start of transistion radius End of transistion radius 250
hs
Extrapolation points
200
150
100
50
10 x [mm]
12
14
16
18
20
Figure 2.12. The hot spot method for Case 413. According to [6] the FAT value for flat butt welds is the same as used for the nominal stress method i.e., 80 MPa, while it should be 100 MPa for fillet welds with crack at toe ground. With these FAT values the equation Q FAT 6 N hs = 2 10 hs gives a life of 1.90 10 6 cycles for Case 213 and 9.16 10 6 for Case 413. The reason why the fatigue life for Case 413 is remarkably long is that the FAT value is relatively large compared to the extrapolated hot spot stress.
3
18
In both examples the 1st principal stress has decreased to a value close to the nominal stress at the distance 0.4t from the toe.
2.2.10
In the effective notch method the notch is replaced by a radius of 1 mm more than the real case. For a welded condition all notches are modelled with a radius of 1 mm (see Figure 2.13) implying that the real radius is zero.
radii
Figure 2.13. All notches modelled by a radius. The FE-model or a handbook with the radii gives the stress concentration at the notch and the stress is then used to calculate the life in the same way for the nominal stress method as for the hot spot method. The effective notch method is advantageous if root cracks are to be evaluated or if different geometries are to be compared. Disadvantages are that the method has not been verified for thicknesses less than 5 mm and the stress must be perpendicular to the weld. Since the stress must be perpendicular to the weld, the 1st principal stress is commonly used in order to simulate the worst case scenario.
2.2.11
The maximum stress at the notch in both Case 213 and 413 is given in Figures 2.14 and 2.15 respectively.
19
notch
Notch method for Case 413 300 Start of transistion radius End of transistion radius
notch
200
150
100
50
3 x [mm]
Figure 2.15. Effective notch stress, notch , for Case 413. Note that only the toe notch is considered in Case 413. For the effective notch method only one FAT-value is available in [3] for modelled radii of 1 mm. This single FAT-value, 225 MPa, does not treat possible misalignment. The same equation as in previous examples Q FAT 6 N notch = 2 10 notch
3
20
gives the life 5.45 10 6 cycles for Case 213 and 1.97 10 6 cycles for Case 413.
2.2.12
Theory about LEFM was given in chapter 2.1 (Basic Theory). How LEFM is applied for Case 213 and 413 is given in chapter 3 (Procedure and methods).
LEFM Notch method Hot spot method Nominal stress method 0 Complexity
Figure 2.16. The diagram describes a qualitative comparison of accuracy and model complexity for the four methods. Figure 2.16 is taken from Modelling and Fatigue Life Assessment of Complex Fabricated Structures by Marquis and Samuelsson [10] and describes the accuracy in calculated life compared to the complexity of the model for the four above described methods. Example of complexity could be whether or not it is a simple 2D model of a fillet weld or an advanced 3D model of the rear frame in an articulated hauler. It can be seen in Figure 2.16 that LEFM is a very accurate method which also requires a lot of work while the other, simpler methods, are ranked depending on accuracy as the notch, hot spot and nominal stress method. The nominal stress method is the least accurate. The notch method gives relatively good result at a lower working effort. The nominal stress method can not always be used at very complex structures since it requires non disturbed nominal stresses [5]. Results for all four methods are shown in Table 2.1 and 2.2 below. Table 2.1. Results for Case 213.
Nominal Life (cycles): % of 2 106 cycles: 2 106 100% Hot spot 1.90 106 95% Effective 5.45 106 273% Fracture 1.48 106 74% Notch Mechanics
21
Notch Mechanics
100%
458%
As can be seen in the tables, the hot spot method have a life close to the expected 2 million cycles for Case 213 while it divert extremely much for Case 413. This result must be questioned. The reason why the life it is extremely long is that the prescribed FAT value for fillet welds is very high compared to the extrapolated hot spot stress. The effective notch method seems to agree for Case 413 while it diverts a lot for Case 213. This is because the stress concentration is quite small compared to the FAT value, which results in a longer life. Also this result must of course be questioned. Fracture mechanics gives a shorter life than all the other methods. When using fracture mechanics and Paris law the life has been integrated from an initial crack length of 0.1 mm. Linear elastic fracture mechanics (LEFM) is not applicable on small cracks, which is a problem since most defects have values below 0.1 mm. To be able to use fracture mechanics one has to assume that there is a known crack which is not too small. If one does not want to assume that there is a crack from the beginning other methods must be used during the initial stage of the formation of the crack, before LEFM can be used. However in reality there are always flaws and pores which serves as initial cracks.
22
results in larger models and longer simulation times. Another problem is the transition region between relatively small element near the crack tip and larger elements far away, which often contains badly shaped elements that are not desirable. Badly shaped elements are rectangular elements which have a large quotient between the long and short side or contain large blunt or small sharp angles. One further disadvantage with LEFM is that a curve of the achieved results from the FEsimulation must be approximated and later integrated numerically. Errors can occur in both these steps. Though there are some disadvantages with LEFM, especially the many extra simulations and the more time-consuming steps, the advantages of more geometry dependent and accurate results outweigh the disadvantages.
23
3.1 Modelling
In this chapter questions about the modelling of the crack and the rest of the geometry will be discussed. All cases in the Volvo CE standard STD5605,51 and the ISO5817 are modelled in the FEprogram (Finite Element) ANSYS [11], [12], except a handful of cases which are solved analytically.
0
y
0
x
Figure 3.1. Natural boundary conditions for a butt weld. Another set of boundary conditions was tested, for the left side and the right side respectively. One end of the model was supported in the x-direction and stress was applied at the other end. To prevent the model from rigid body motion, one node on the side where the model is supported in the x-direction is also locked in the y-direction (see Figure 3.2 and 3.3).
24
0
y 0
x
0
y
0
Figure 3.3. Right side of the weld is supported. Results for all three types of boundary conditions are given in Table 3.1. Table 3.1. Results for different boundary conditions.
K_eff [MPa(mm)] Crack length [mm]: 0,0000 0,0025 0,0050 0,0100 0,0200 0,0400 0,0800 0,1600 0,3200 0,6400 1,2800 2,5600 5,1200 Max difference: Natural: 0,00 19,92 28,13 39,50 54,83 74,59 98,67 125,77 158,98 203,56 279,48 456,16 1182,89 Left side locked: 0,00 19,82 27,87 39,48 54,63 74,55 98,72 126,32 158,77 204,93 280,22 456,51 1183,06 0,9% Right side locked: 0,00 19,92 28,13 39,50 54,83 74,59 98,67 125,77 158,98 203,56 279,48 456,16 1182,89 0,0%
25
As can be seen in Table 3.1, the types of boundary conditions do not influence the results very much. The maximum difference compared to the boundary conditions in Figure 3.1 is around 0.9%. Henceforth boundary conditions given in Figure 3.1 will be used. Reaction forces at the supports have been controlled for all types of boundary conditions. For the boundary conditions selected for further use, the reaction forces were negligible both in xand the y-direction. For the other types of boundary conditions, the total reaction force was the same as the applied force in the x-direction, while it was negligible in the y-direction as expected.
26
As can be seen in Table 3.2 the boundary conditions in the original model have no decisive effect on the result and can be used further on without complications. The maximum difference according to boundary conditions in Figure 3.1 is around 0.9%.
3.2 Meshing
Meshing of the model turned out be one of the largest challenges in this work. How the meshing problems were solved is described in this chapter.
27
crack opening
crack tip
Figure 3.6. Four areas create a box around the crack tip. There is also an area division from the crack tip vertically through the material, see Figure 3.6. This facilitates the meshing.
3.2.2 Elements
The elements used when meshing the models are plane stress eight node serendipity elements and six node triangular elements. Each node has two degrees of freedom; one in the x- and one in the y-direction.
2 1
crack tip
28
To obtain stresses going to infinity at the crack tip, the side nodes on the triangular pie elements are moved to the quarter point of the element towards the crack tip. Figure 3.8 shows an actual ANSYS plot of the crack tip elements.
0.5 10 3 mm
crack tip
Figure 3.8. Screenshot from ANSYS of the crack tip. The first two rows of elements around the crack tip have the length 0.5 10 3 mm.
29
crack tip
Figure 3.9. The mesh in the box around the crack tip. Screenshot from ANSYS. The mesh in the box is of high quality with no badly shaped elements.
30
0.1 mm
Figure 3.10. The centre line below the box. The same situation is applied to the two crack surfaces above the box. One could say that the size and number of elements near the crack tip is overkill but for many cases the crack is set to move from 0.1 mm down to 5 mm. In order to build a general macro which can take care of the entire crack growth, this solution was regarded as a straightforward method.
31
crack tip
32
Figure 3.12. The light grey elements are badly shaped. When using a smaller element side length on the global areas, 0.2 mm instead of 1 mm, the badly shaped elements disappear. The calculation time however is about 10 to 20 times longer and the calculated stress intensity factors improve less than 1.5%. Since more than 100 geometries are to be calculated the coarser mesh is used to save time.
3.2.8 Convergence
In order to guarantee convergence, the two prior test examples - the transverse butt weld and the cruciform joint (Case 213 and 413) - are calculated with the above mentioned mesh and a mesh that has half the element size on all elements except the crack elements. Results in Table 3.3, for Case 213 and Table 3.4, show that the solution has converged. As mentioned before, lack of time is the main factor for choosing the coarser mesh, because over one hundred geometries are to be calculated. The calculation time increase 10 to 20 times with the finer mesh, and that justifies the decision to use the coarser mesh.
3 Procedure and methods Table 3.3. Convergence results for transverse butt weld, Case 213.
K_eff [MPa(mm)] Crack length [mm]: 0,0000 0,0025 0,0050 0,0100 0,0200 0,0400 0,0800 0,1600 0,3200 0,6400 1,2800 2,5600 5,1200 Coarse mesh: 0,00 19,92 28,13 39,50 54,83 74,59 98,67 125,77 158,98 203,56 279,48 456,16 1182,89 Finer mesh: 0,00 19,77 27,86 39,02 54,17 73,99 98,43 126,75 159,85 204,32 279,68 456,26 1183,02 Deviation: 0,00% 0,74% 0,97% 1,22% 1,23% 0,81% 0,25% 0,77% 0,55% 0,37% 0,07% 0,02% 0,01%
33
3.2.9 Macros
After the geometry has been modelled, all of the remaining work - crack growth, area subdivision, meshing, solving and calculation of stress intensity factors - is the same for all models. Therefore generic macros are developed for ANSYS.
34
When the crack has propagated a small distance into the material, a generic macro looks the same for any geometry. The differences between different geometries occur in the beginning of the crack growth, when geometrical differences like undercuts, transition radii, sharp transitions and flat sheets affect the area subdivision. Therefore all the macros with the above mentioned geometries look basically the same and differ only for the initial state of the crack. The advantage of the macros are that they need only three keypoints as input and then automatically move the crack tip into the geometry, divide it into areas and put the correct mesh size on all lines. After solution, the stress intensity factors are calculated and written to a file containing the results for all the steps during the crack growth. The macros also save the meshed models for each crack depth, making it possible to review the mesh afterwards.
3.3 Postprocessing
This chapter deals with how the ANSYS calculates the stress intensity factors and how the results from the ANSYS are treated in order to calculate the fatigue life.
Figure 3.13. Local coordinate systems at crack front. Picture taken from ANSYS, Inc Theory Reference [14].
35
Further K I , K II and K III are the stress intensity factors for the three different modes, and
3 4 = 3 1 + for plane strain or an axisymmetric case for plane stress
where E is Youngs modulus and is Poissons ratio. (r ) are terms of order r or higher. The stress intensity factors are interesting when r 0 and = 180, therefore the given angle is inserted into equations (3.1:1-3) and the higher order terms are neglected. It gives K II r G u (1 + ) K II = 2 u = 2G 2 1+ r KI r G v (1 + ) K I = 2 v = 2G 2 1+ r w 2 K III r K III = 2G 2 w = r G 2 r (3.2 : 1) (3.2 : 2) (3.2 : 3)
In all equations (3.2:1-3) the following limit values must be evaluated lim
r 0
where is either u , v or w . This is done by using the 5 nodes in the path in ANSYS, mentioned above, see Figure 3.7 and 3.14.
Figure 3.14. (a) displays the crack tip for a half-crack model, while (b) displays the crack tip for a full crack model, which is used in this study. Picture taken from [14]. The following curve is fitted for v .
36
Analysis of Fatigue Life in Two Weld Class Systems v(r ) = r (A + Br ) v(r ) r = A + Br (3.3)
The displacement of the crack tip node is said to be zero, while the known displacement in points K and J are used to determine the constants A and B. The limit value becomes lim
r 0
v(r ) r
= lim(A + Br ) = A
r 0
(3.4)
The same thing is performed for mode II and III, which gives corresponding constants. Combining Equation (3.4) and Equation (3.2) finally gives the stress intensity factors below. GA I K I = 2 1 + GA II K II = 2 1+ K III = 2GA III 2 (3.5 : 1) (3.5 : 1) (3.5 : 1)
i.e., 12 stress intensity factors were calculated for each model. Ten integration points were used below 1 mm and only 3 points above 1 mm. The extra point comes from the stress intensity factor being zero when the crack length is zero. Figure 3.15 displays the stress intensity factor as a function of crack length for Case 213.
37
Keff [MPa(m)]
25 20 15 10 5 0
3 a [mm]
Figure 3.15. The stress intensity factor as a function of crack length for Case 213. The lines in Figure 3.15 at 0.1 mm and 5 mm mark the integration limits. Numerical integration in MATLAB with Simpson quadrature [16] for this case gives a life of 1.48 10 6 cycles. The dots in Figure 3.15 mark the crack lengths where the stress intensity factor has been calculated. The splines are based on these points. To rule out the possibility of errors in life due to too few points for longer cracks, a test with one value for each half mm was carried out. This means that 9 integration points instead of 3 are used (see Figure 3.16). A comparison of the curves and lives shows that differences in life are negligible.
38
Keff (a)
40 35 30
Keff [MPa(m)]
25 20 15 10 5 0
3 a [mm]
Figure 3.16. The difference between many (squares) and fewer (dots) integration points for longer cracks. The dotted curve and the boxes in Figure 3.16 display the previous results, which were presented in Figure 3.15. The solid curve and the squares in Figure 3.16 display what happens if more integration points are used for longer crack lengths. Between 1 and 5 mm in crack lengths, the small difference between the two curves gives a difference in life of 0.2% only.
3 Procedure and methods Table 3.5. Data from Case 213, ANSYS results
a [mm]: 0.0000 0.0025 0.0050 0.0100 0.0200 0.0400 0.0800 0.1600 0.3200 0.6400 1.2800 2.5600 5.1200 Keff [MPa(m)]: 0.00 0.63 0.89 1.25 1.73 2.36 3.12 3.98 5.03 6.44 8.84 14.43 37.41
39
The values in Table 3.5 are plotted together with the approximated splines in Figure 3.17.
Keff (a)
40 35 30 Integration points Curve approximation with splines Curve with straight lines
Keff [MPa(m)]
25 20 15 10 5 0
3 a [mm]
Figure 3.17. Plotted data from Case 213. As can be seen in Figure 3.17, the difference between the spline approximated curve and the straight line curve becomes greater for larger crack depths.
40
No integration point is available for the crack length 0.1 mm, which is used as start defect when integrating with Paris law. However there is a point at a = 0.08 mm which will be used in the verification of the calculated life. MATLAB gives the life 1.60 10 6 cycles when Simpson quadrature [16] is used to integrate the life over the interval 0.08 a 5.12 mm.
dN = C (K )
0
ai
da
is to be integrated numerically in the interval of n C (K I (a )) 0.08 a 5.12 mm. Since there is not the same distance between any of the integration points, the integration is done with both over and under sums. The fatigue life is then approximated with the average of the over and under sums. In Figure 3.18 f (a ) is described.
f (a )
The curve f (a ) =
ai1 ai
ai +1
Figure 3.18. Upper and lower summations. The upper sum becomes Nu =
i =1 n
C (K I )i
(ai +1 ai )
41
Nl =
i =1
C (K I )i +1
n
(ai +1 ai )
This is the same as a Riemann integral. If this is done for integration points 0.08 a 5.12 mm, the fatigue life becomes 1.78 10 6 cycles compared to 1.60 10 6 cycles for the spline approximation. It should be mentioned that this method overestimates the fatigue life since the spline based curve always will have less or equal values than the average line. However it shows that the value calculated with splines in MATLAB is approximately good.
N =
i =1
a ai 1 1 i +1 n 2 C (K I )i
ai-1 ai
ai +1
42
The approximation with Riemann sums gives 1.60 10 6 cycles i.e., the same as with the Simpson quadrature in MATLAB, which means that the integration method is verified.
2W t 2a 2c
t 2c 2a
Figure 3.21. Denotations for the crack. This case, taken from [10], is programmed in Matlab [7]. The Matlab program can be found in the Appendix A. The stress intensity factor is calculated as KI = m F where F can be calculated as
2 4 2a 2a F = M 1 + M 2 + M 4 gf f w t t
a Q
i.e., F is a function of the angle , which means that K I can be calculated at an arbitrary position along the crack tip. In these cases K I will be calculated for the angle 0 ( K Ic ) and 90 ( K Ia ), i.e. where the two extreme values of the stress concentration factor K t can be expected. Further
43
M2 =
M4 =
1. 5
a t g = 1
2.6 2 a 1+ 4 c
a t
cos
The following conditions apply depending on in which direction, a or c, the ellipse is larger. For a c 1 : a Q = 1 + 1.464 c
1.65
a 2 f = cos 2 + sin 2 c
M1 = 1
1/ 4
c 2 f = sin 2 + cos 2 a M1 = c a
1/ 4
(3.6)
44
(2.1)
Paris law (2.1) can be inserted into (3.6:1) and the corresponding for (3.6:2), which gives (3.7)
The next crack lengths which the program uses to calculate the stress intensity factors in the a- and c-directions, depend on the stress intensity factor in the given direction. The constant M determines the resolution i.e., how many integration points that will be used. The greater M is, the fewer integration points. The stress intensity factors and corresponding crack lengths are later integrated with Paris law, in the same way as described in Chapter 3.3.2, in order to achieve the fatigue life.
3.5.2 Algorithm
The Matlab program is based on the following algorithm: 1. Start lengths of the crack in the a- and the c-direction ( ai and ci ) are given from the acceptance limits in the weld classes. 2. ai and ci are used to calculate (K Ia )i and (K Ic )i . 3. The new crack lengths, ai +1 and ci +1 , are calculated based on (K Ia )i and (K Ic )i 4. The interruption criterion is checked. In this case the crack is set to grow up to a given length in the a-direction. If the criterion is fulfilled, all necessary stress intensity factors have been calculated and the fatigue life can be integrated, if not, the program starts over from 2. The interruption criterion is based on the length in the a-direction. The crack grows faster in this direction, and in all other cases in the standard the fatigue life has been integrated from the initial crack length until the crack length is half of the sheet thickness. Therefore the calculations are interrupted when the crack length a is half the sheet thickness.
45
10 8 2acrack [mm] 6 4 2 0 0 2
acrack(N)
4 N [-]
6 x 10
8
5
Figure 3.22. Crack length a(N) for an infinitely deep crack. STD5605 Case 27 D. For an internal elliptical crack the same curve goes more slowly towards infinity at the same crack length, basically because the crack can grow a very long distance in the c-direction (see Figure 3.23).
10 8 2acrack [mm] 6 4 2 0 0 2
acrack(N)
N [-]
10 x 10
5
Figure 3.23. Crack length a(N) for an internal elliptical crack. STD5605 Case 27 C. However, in order to be consistent compared to other cases, the fatigue life is integrated to half of the thickness also for this case.
46
acrack(N)
10 8 2acrack [mm] 6 4 2 0 0 5 10 15 x 10
5
N [-]
Figure 3.24. The points from from AFGROW coincide with the curve from MATLAB. As can be seen in Figure 3.24, the data points from AFGROW coincide with the curve generated with the MATLAB program i.e., the values from the MATLAB program are confirmed.
4 Weld auditing
47
4 Weld auditing
This chapter is an introduction to weld auditing in Volvo CE. Audits are performed to reveal problems in the manufacturing process in order to improve the quality of for example welds. It is also important to discover defects on welds of high criticality in order to prevent disastrous failures.
48
Analysis of Fatigue Life in Two Weld Class Systems Table 4.2. Compilation of occurring defects.
Defect Lack of fusion Non-filled weld Excessive penetration Hole Leg deviation Throat deviation Sharp transition Undercut Weld missing / too short Overlap 1 Category 1 (Security) 2 (Re-work) 3 (No re-work) 1 1 17 3 20 1 18 3 22 1 34 10 45 2 70 16 88 20 8 6 6 1 2 7 4 1 4 22 1 1 1 5 9 10 9 3 5 45 A-stay Body Front frame Rear frame Total 2 2 1 4 2 1 1 6 19 23 19 4 9 88
lle d
pe ne
de
vi ro at Sh a rp
N on -fi
ck
La
iv e
Le
Ex ce
ss
Th
tra
of
4 Weld auditing
49
As seen above, an average object (apart from A-stays) has a number of defects of the order of ten.z The vast majority are of category 2, i.e. they have to be reworked. Defects against safety requirements (category 1), normally referring to joints of higher consequence classes, are luckily not very common. Especially three types of defects are frequent; throat deviation, sharp transition and undercut. The most common defect is the sharp transition. This depends on most welds having the additional designation U, fatigue requirements, where smooth transitions are required.
Figure 4.1. A sharp transition has been marked for making a silicone impression.
50
4.3.2 Comments
The profile projector measuring illuminated a few of the difficulties involved in measuring of weld geometries. The legs and throat lengths are quite easy to measure with sufficient accuracy, but when it comes to angles and radii problems arise. The angle between the sheet and the weld surface is easy to measure with the projector equipment the problem is where it should be measured. If the intention is to measure the angle after the transition (as in this study), then the question is where the transition ends. This is in most cases far from obvious. An even more difficult property to estimate is the nature of the transition. Determining whether the transition is sharp or even is a central issue, as even transition is required in welding classes used for fatigue loaded joints. (Even is the term used in STD5605,51. Smooth would be more correct.) Normally, the even transition is interpreted as a 1 mm radius, but as the curve is always more or less jagged, it is often impossible to decide if there is a radius. Furthermore, features as small as this can vary a lot along the weld. If a nearby cross-section is chosen for examination, the results can be completely different. Thus, measuring of weld dimensions demands more than just measuring one cross-section. To obtain reliable figures, several samples have to be examined and the whole weld has to be inspected to make sure that the picked samples are representative.
5 Compilation of STD5605,51
51
5 Compilation of STD5605,51
This chapter contains overall information on the Volvo CE STD5605,51 weld standard. Also results and some detailed conclusions and recommendations from all calculated cases in the standard are presented in diagrams, tables and on result sheets.
a1.5
5 x 20 (10)
135-D-F [3]
52
to welds, which at breakdown could lead to immediate stand-still of the vehicle. Finally, class [3] apply to welds, which at failure could lead to loss of performance and require repair, but no immediate stand-still. When a defect is found in a weld of consequence class [1] or [2] at the weld audit, there are certain routines within Volvo CE to catch the errors and make sure they are not repeated.
5.2 Modelling
All transverse butt welds in the standard with the requirement not permitted for a certain welding class have been modelled as Case 6A in the STD5605,51 standard, i.e., IIW Case 213 but without the transition radius of 1 mm. The corresponding cases for fillet welds have been modelled as IIW Case 413, but with the difference that a 4 mm transition radius has been used for welding class A, a 1 mm transition radius has been used for classes B, CU and DU, and a sharp transition has been used for class C and D. Other comments on modelling are given in Chapter 3, and in the result sheets for all cases (Chapter 5 and 6).
5.3 Results
Results with pictures and diagrams for each case and welding class can be found on the following pages. All results are also compiled in Figure B.1-B.18 and in Table B.1-B.2, in Appendix B. An overview of the results can be found in Figure B.1. It shows large scatter in the results. However, patterns can be discerned, for example, some cases have nearly no difference in life between the welding classes, while some show large differences. A rough grouping of the cases can be seen in Table 5.1. Table 5.1. Influence of acceptance limits on the fatigue life for a certain welding class.
Large influence 6 7 8 13 14 20 Some influence 3 27 22 23 24 No influence 11 12 25 26 21 Butt welds Fillet welds Butt welds Fillet welds Butt welds Fillet welds
In Figure B.2 and B.3 the transverse butt welds and fillet welds have been separated. For the butt welds in Figure B.2, Case 3, 11 and 12 differ a lot since there are nearly no difference in life between almost all the welding classes. However the lack of differences in Case 3 depends on that no sharp transitions are permitted for the additional designation U. Case 25, 26 and 27 are internal defects and are very hard to compare with the remaining cases for butt welds. Figure B.3, with only fillet welds, show just a little agreement. Case 23 and 24 could be separated from the remaining fillet welds since these cases are guidelines for all fillet welds and not real defects. Case 23 and 24 are modelled with a radius of 1, 2, 3 and 4 mm for DU, CU, B and A respectively. This has been done to study the influence of different radii and is not based on the requirements, except for welding class A. Case 21 show only small variations in fatigue life between the different welding classes, while Case 20 and 22 show influence from the acceptance limits on fatigue life.
5 Compilation of STD5605,51
53
Since welding class A is barely ever used and C and D are meant only for static loading, B, CU and DU could be separated and can be seen in Figure B.4-B.6. These figures reflect what has been said above for the overall results in Figure B.1-B.3 - the classes used for fatigue loaded joints are not more consistent than other classes when it comes to fatigue life. If the ratio of B and CU and of CU and DU is plotted (Figure B.7) one can see that the ratio of B and CU lies between 1 to 2 for almost all the cases, which shows that there is a little bit of consistency in the standard. But the ratio of CU and DU is very scattered, between 1 and 8, with nearly no pattern. Case 25-27 have been left out due to above mentioned differences. Figure B.8 shows the maximum, minimum, average, and median life for each welding class. Preferably these four values should be equal, but due to the present standard they vary a lot. One can see that for many welding classes there is a factor ten or more between the minimum and maximum value. Welding class D and DU appears the most scattered. There is almost a factor of 100 between maximum and minimum in these classes. It is also possible to see in Figure B.8 that the average life for each class is always longer than the median. This shows that there are a few cases with long fatigue lives, which increase the average life. In other words, if these cases with unreasonably long fatigue lives are reworked the average and median lives coincide. Further, all four values; maximum, minimum, average and median, are lower in welding class C compared to CU and in D compared to DU. This is expected since the additional designation U should give the structure a longer fatigue life. It can also be seen that there is a stair with longer life in A than B, B than C and so on, for all welding classes, something which could also be expected. However if also the additional designation U is regarded, one can see that the trend that welding class DU have longer life for the maximum, average and medium value, than class C. This shows that transition radii and smooth notches in general give longer lives even though the defect (acceptance limit) may be larger, which is usually the case between class DU and C. In Figure B.9-B.12 the fatigue life in percent of 2 million cycles has been plotted as a function of the acceptance limit, for different sets of cases (mainly 6, 7, 8, 13, 14, 20) and classes. The reason why these cases are chosen are that the acceptance limit has a major impact on the fatigue life, as described above and presented in Table 5.1. An exponential trendline has been fitted to the points in all four diagrams. Note that there might be coincident (overlapping) points in the diagram. In Figure B.11, also an exponential trendline is fitted to the points. One can see that the trendlines fit the scattered points relatively well, except for small acceptance limits i.e., around 0.5 mm for Figure B.9-B.11. Equations for the trendlines can be found in each diagram. The trendline could serve as a tool when setting new acceptance limits. For example, if the desired life in a certain welding class has been decided, the acceptance limit can be read from the curve. Of course this might not give the exact life for the certain case, but it will at least be possible to get a hint of the size of the desired acceptance limit. It is not possible to create an appropriate trendline for all the calculated data, since it is too scattered and many of the acceptance limits for different defects do not have any influence on the fatigue life. In Figure B.13-B.18, results for each welding class can be found. These results also confirm large scatter in the results.
54
With additional designation U: 2.48 106 cycles 124% of 2 106 cycles 2.48 106 cycles 124% of 2 106 cycles
CU
DU
C D
1
5 Compilation of STD5605,51
55
This case does not supply so much additional information to the standard. Since no sharp transitions are allowed in CU and DU and incomplete root penetration is not permitted for A and B, the acceptance limits cannot be reworked for A and B, because then CU and DU will have a longer life than A and B and this is not wanted. C and D are meant only for static loading. Geometrically Case 3 is the same as Case 8 in the STD5605,51 standard but the two cases does not have the same acceptance limits for welding class B. Since the cases are similar and it could be hard for others than experts to see the differences between incomplete root penetration (Case 3) and root concavity (Case 8), it is recommended that these two cases are united into one case. It is not consistent to allow a different acceptance limit, which is the case for welding class B, when the cases are so similar. Further, it should be mentioned that the additional designation U implies a full penetration.
56
l
Modelling and boundary conditions:
Modelling:
A = initial crack
40 30 20 10 0 0
6 5 acrack [mm] 4 3 2 1
3 4 acrack [mm]
0 0
0.5
N [-]
1.5
2.5 x 10
6
5 Compilation of STD5605,51
57
STD5605,51 Case 3 C
Description: Incomplete root penetration Requirements and results: A 0.1 t, but max. 1 mm, l t but max. 25 mm. A = 1 mm, l = 10 mm initial crack = 1.0 mm nom = 104 MPa
l
Modelling and boundary conditions:
Modelling:
A = initial crack
40 30 20 10 0 0
6 5
acrack [mm]
1 2 3 4 acrack [mm] 5 6
4 3 2 1 0 0 1 2 N [-] 3 x 10 4
5
58
STD5605,51 Case 3 D
Description: Incomplete root penetration Requirements and results: A 0.2 t, but max. 2 mm. L 100 mm A = 2 mm initial crack = 2.0 mm nom = 104 MPa
l
Modelling and boundary conditions:
Modelling:
A = initial crack
40 30 20 10 0 0
6 5
acrack [mm]
1 2 3 4 acrack [mm] 5 6
4 3 2 1 0 0 2 4 6 N [-] 8 10 12 x 10
4
5 Compilation of STD5605,51
59
A Not permitted
B C CU D
1
DU
60
5 Compilation of STD5605,51
61
STD5605,51 Case 6 A
Description: Undercut Requirements and results: Not permitted initial crack = 0.1 mm nom = 104 MPa
6 N = 1.58 10 cycles 6 N f = 79% of 2 10 cycles
Modelling:
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
1 2 3 4 acrack [mm] 5 6
62
STD5605,51 Case 6 B
Description: Undercut Requirements and results: Permitted locally if A 0.05 t, but max. 0.5 mm. l 25 mm A = 0.5 mm initial crack = 0.5 mm nom = 104 MPa
6 N = 0.73 10 cycles 6 N f = 37% of 2 10 cycles
Modelling:
A = initial crack
40 30 20 10 0 0
6 5
Keff [MPa(m)]
acrack [mm]
1 2 3 4 acrack [mm] 5 6
4 3 2 1 0 0 2 4 N [-] 6 x 10 8
5
5 Compilation of STD5605,51
63
STD5605,51 Case 6 C
Description: Undercut Requirements and results: Permitted locally if A 0.1 t, but max. 1 mm. A = 1 mm initial crack = 1.0 mm nom = 104 MPa
6 N = 0.37 10 cycles 6 N f = 18% of 2 10 cycles
Modelling:
A = initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
2 N [-]
3 x 10
4
5
64
STD5605,51 Case 6 CU
Description: Undercut Requirements and results: Permitted locally if A 0.1 t, but max. 1 mm. A = notch depth + initial crack = 0.9 + 0.1 = 1 mm nom = 104 MPa
6 N = 0.42 10 cycles 6 N f = 21% of 2 10 cycles
Modelling: R1
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
4 x 10
5
5
5 Compilation of STD5605,51
65
STD5605,51 Case 6 D
Description: Undercut Requirements and results: Permitted locally if A 0.2 t, but max. 2 mm. A = 2 mm initial crack = 2.0 mm nom = 104 MPa
6 N = 0.11 10 cycles 6 N f = 5% of 2 10 cycles
Modelling:
A = initial crack
40 30 20 10
6 5
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
6 N [-]
10
12 x 10
4
66
STD5605,51 Case 6 DU
Description: Undercut Requirements and results: Permitted locally if A 0.2 t, but max. 2 mm. A = notch depth + initial crack = 1.9 + 0.1 = 2 mm nom = 104 MPa
6 N = 0.13 10 cycles 6 N f = 6% of 2 10 cycles
Modelling: R1
initial crack
100 80
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
60 40 20 0 0 2 4 6
acrack [mm]
0 0
N [-]
10
15 x 10
4
5 Compilation of STD5605,51
67
7. 8.
A 0.05t, but max. 0.5 mm. 0.92 106 cycles 46% of 2 106 cycles
A Not permitted
B C
1
CU D DU
68
Modelling:
A = initial crack
40 30 20 10 0 0
6 5
Keff [MPa(m)]
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
2 N [-]
3 x 10
4
6
5 Compilation of STD5605,51
69
Modelling:
A = initial crack
40 30 20 10 0 0
6 5
acrack [mm]
1 2 3 4 acrack [mm] 5 6
4 3 2 1 0 0 2 4 6 8 10 x 10
5
N [-]
70
Modelling:
A = initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
N [-]
4 x 10
5
5
5 Compilation of STD5605,51
71
Modelling: R1
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
4 x 10
5
5
72
Modelling:
A = initial crack
40 30 20 10 0 0
acrack [mm]
1 2 3 4 acrack [mm] 5 6
4 3 2 1 0 0 2 4 6 N [-] 8 10 12 x 10
4
5 Compilation of STD5605,51
73
Modelling: R1
initial crack
100 80
Keff [MPa(m)]
6 5 acrack [mm] 4 3 2 1
60 40 20 0 0 2 4 6
acrack [mm]
0 0
N [-]
10
15 x 10
4
74
A The weld reinforcement shall be removed and the surface machined to the level of the parent metal
CU
DU
5 Compilation of STD5605,51
75
It is not recommended, however, to entirely delete the case from the standard since some kind of limitation of the weld reinforcement could be important for reasons of appearance or for technical reasons (too large weld reinforcements could interfere with other parts etc.). However, it should be clear that the weld reinforcements have no influence on the fatigue properties. Further, the requirement for welding class D Overlap is permitted to a small extent should be replaced by something more concrete with a length in actual figures. Otherwise the requirement could be widely interpreted.
76
STD5605,51 Case 11 A
Description: Weld reinforcement Requirements and results: The weld reinforcement shall be removed and the surface machined to the level of the parent metal initial crack = 0.1 mm nom = 104 MPa
6 N = 3.25 10 cycles 6 N f = 163% of 2 10 cycles
Modelling:
A = initial crack
40 30 20 10 0 0
6 5
Keff [MPa(m)]
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
2 N [-]
3 x 10
4
6
5 Compilation of STD5605,51
77
STD5605,51 Case 11 B
Description: Weld reinforcement Requirements and results: A 1.5 + 0.05 b. Overlap is not permitted. A = 2 mm, b = 10 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 1.73 10 cycles 6 N f = 86% of 2 10 cycles
Modelling:
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
78
STD5605,51 Case 11 C
Description: Weld reinforcement Requirements and results: A 1.5 + 0.15 b. Overlap is permitted to a small extent. A = 2.5, b = 10 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 1.63 10 cycles 6 N f = 81% of 2 10 cycles
Modelling:
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
5 Compilation of STD5605,51
79
STD5605,51 Case 11 CU
Description: Weld reinforcement Requirements and results: A 1.5 + 0.1 b. Overlap is not permitted. A = 2.5 mm, b = 10 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 1.76 10 cycles 6 N f = 88% of 2 10 cycles
Modelling: R1
initial crack
acrack [mm]
1 2 3 4 acrack [mm] 5 6
80
STD5605,51 Case 11 D
Description: Weld reinforcement Requirements and results: A 1.5 + 0.15 b. Overlap is permitted to a small extent. A = 3 mm, b = 10 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 1.56 10 cycles 6 N f = 78% of 2 10 cycles
Modelling:
initial crack
40 30
6 5
acrack [mm]
4 3 2 1
20 10 0 0
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
5 Compilation of STD5605,51
81
STD5605,51 Case 11 DU
Description: Weld reinforcement Requirements and results: A 1.5 + 0.15 b. Overlap is permitted to a small extent. A = 3 mm, b = 10 mm nom = 104 MPa initial crack = 0.1 mm 6 N = 1.70 10 cycles 6 N f = 85% of 2 10 cycles Modelling and boundary conditions: Modelling: R1
A = initial crack
40 30 20 10
6 5
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
82
A The penetration bead shall be removed and the surface machined to the level of the parent metal
CU
DU
5 Compilation of STD5605,51
83
It is not recommended to delete the case from the standard entirely since some kind of limitation of the penetration bead is important for reasons of appearance as well as for technical reasons (too large penetration beads could interfere with other parts etc.). However, it should be clear that the penetration bead has no influence on the fatigue properties.
84
STD5605,51 Case 12 A
Description: Penetration bead Requirements and results: The penetration bead shall be removed and the surface machined to the level of the parent metal initial crack = 0.1 mm nom = 104 MPa
c
Modelling and boundary conditions:
Modelling:
A = initial crack
40 30 20 10 0 0
6 5 acrack [mm] 4 3 2 1
3 4 acrack [mm]
0 0
0.5
N [-]
1.5
2.5 x 10
6
5 Compilation of STD5605,51
85
STD5605,51 Case 12 B
Description: Penetration bead Requirements and results: A 1.5 + 0.1 c A = 1.8 mm, C = 3 mm nom = 104 MPa initial crack = 0.1 mm 6 N = 1.69 10 cycles 6 N f = 85% of 2 10 cycles Modelling:
c
Modelling and boundary conditions:
initial crack
40 30
6 5
acrack [mm]
4 3 2 1
20 10 0 0
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
86
STD5605,51 Case 12 C
Description: Penetration bead Requirements and results: A 1.5 + 0.2 c A = 2.1 mm, C = 3 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 1.69 10 cycles 6 N f = 84% of 2 10 cycles
c
Modelling and boundary conditions:
Modelling:
initial crack
40 30
6 5
acrack [mm]
4 3 2 1
20 10 0 0
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
5 Compilation of STD5605,51
87
STD5605,51 Case 12 CU
Description: Penetration bead Requirements and results: A 1.5 + 0.2 c A = 2.1 mm, C = 3 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 1.77 10 cycles 6 N f = 89% of 2 10 cycles
c
Modelling and boundary conditions:
Modelling:
initial crack
40 30 20 10
6 5
Keff [MPa(m)]
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
88
STD5605,51 Case 12 D
Description: Penetration bead Requirements and results: A 1.5 + 0.3 c A = 2.4 mm, C = 3 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 1.69 10 cycles 6 N f = 84% of 2 10 cycles
c
Modelling and boundary conditions:
Modelling:
initial crack
40 30
6 5
acrack [mm]
4 3 2 1
20 10 0 0
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
5 Compilation of STD5605,51
89
STD5605,51 Case 12 DU
Description: Penetration bead Requirements and results: A 1.5 + 0.3 c A = 2.4 mm, C = 3 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 1.76 10 cycles 6 Nf = 88% of 2 10 cycles
c
Modelling and boundary conditions:
Modelling:
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
1 2 3 4 acrack [mm] 5 6
90
CU
DU
x r
crack
Figure 5.2. The radius is derived from the known distances A and x. The radius becomes
5 Compilation of STD5605,51 r= x 2 + A2 2A
91
where x is constant and x = 10 mm, i.e. the gap between the sheets before welding. The lower side of the weld has been modelled to resemble the picture in the standard (see the top right side of this result sheet. A line fillet with a radius of 2 mm has been used. A, B, C and D have been modelled with a sharp transition, while CU and DU have been modelled with a transition radius of 1 mm. Class A and B have the same acceptance limits and therefore the same results. As can be seen in the results, CU and DU have only insignificantly longer life than C and D.
Boundary conditions
Two types of boundary conditions are available for this case, see Figure 5.3 and 5.4.
crack
Figure 5.3. Boundary conditions for generating the results. The arrow indicates the direction of rotation.
crack
Figure 5.4. Boundary conditions which gives longer fatigue life. The arrow indicates the direction of rotation. Boundary conditions according to Figure 5.3 causes the sheets to perform a counter-clockwise rotation, which opens the crack. There are both a bending moment and an uniaxial tension which contributes to the crack growth. This results in a considerably lower fatigue life compared to the boundary conditions in Figure 5.4. In Figure 5.4 the tension causes a clockwise rotation on the sheets which results in a bending moment closing the crack. The crack would not grow if the uniaxial tension was not high enough to overcome the bending moment and open the crack.
92
5 Compilation of STD5605,51
93
Modelling:
initial crack
60 50
Keff [MPa(m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
4 N [-]
6 x 10
8
5
94
STD5605,51 Case 13 C
Description: Requirements and results: Edge displacement, one-sided welding A 0.15 t, but max. 3 mm. A = 1.5 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 0.32 10 cycles 6 N f = 16% of 2 10 cycles
Modelling:
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
2 N [-]
3 x 10
4
5
5 Compilation of STD5605,51
95
STD5605,51 Case 13 CU
Description: Requirements and results: Edge displacement, one-sided welding A 0.15 t, but max. 3 mm. A = 1.5 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 0.34 10 cycles 6 N f = 17% of 2 10 cycles
Modelling: R1
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
2 N [-]
3 x 10
4
5
96
STD5605,51 Case 13 D
Description: Requirements and results: Edge displacement, one-sided welding A 1.5 + 0.25 t, but max. t or 4 mm. A = 4 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 0.04 10 cycles 6 N f = 1.9% of 2 10 cycles
Modelling:
initial crack
100 80
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
60 40 20 0 0 2 4 6
acrack [mm]
0 0
2 N [-]
3 x 10
4
4
5 Compilation of STD5605,51
97
STD5605,51 Case 13 DU
Description: Requirements and results: Edge displacement, one-sided welding A 1.5 + 0.25 t, but max. t or 4 mm. A = 4 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 0.043 10 cycles 6 N f = 2.1% of 2 10 cycles
Modelling: R1
initial crack
100 80
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
60 40 20 0 0 2 4 6
acrack [mm]
0 0
N [-]
4 x 10
5
4
98
B C CU D DU
5 Compilation of STD5605,51
99
STD5605,51 Case 14 A
Description: Edge displacement, double-sided welding Requirements and results: A 0.1 t, but max. 2 mm A = 1 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 0.60 10 cycles 6 N f = 30% of 2 10 cycles
Modelling:
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
3 N [-]
5 x 10
6
5
100
STD5605,51 Case 14 B
Description: Edge displacement, double-sided welding Requirements and results: A 0.15 t, but max. 3 mm A = 1.5 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 0.32 10 cycles 6 N f = 16% of 2 10 cycles
Modelling:
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
2 N [-]
3 x 10
4
5
5 Compilation of STD5605,51
101
STD5605,51 Case 14 C
Description: Edge displacement, double-sided welding Requirements and results: A 0.2 t, but max. 5 mm A = 2 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 0.18 10 cycles 6 Nf = 9% of 2 10 cycles
Modelling:
initial crack
80 60 40 20
6 5
4 3 2 1
0 0
0 0
0.5
1 N [-]
1.5 x 10
2
5
102
STD5605,51 Case 14 CU
Description: Edge displacement, double-sided welding Requirements and results: A 0.2 t, but max. 5 mm A = 2 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 0.19 10 cycles 6 N f = 10% of 2 10 cycles
Modelling: R1
initial crack
80 60
6 5
acrack [mm]
4 3 2 1
40 20 0 0
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
5
5 Compilation of STD5605,51
103
STD5605,51 Case 14 D
Description: Edge displacement, double-sided welding Requirements and results: A 1.5 + 0.25 t, but max. t or 5 mm A = 4 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 0.04 10 cycles 6 N f = 1.9% of 2 10 cycles
Modelling:
initial crack
100 80
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
60 40 20 0 0 2 4 6
acrack [mm]
0 0
2 N [-]
3 x 10
4
4
104
STD5605,51 Case 14 DU
Description: Edge displacement, double-sided welding Requirements and results: A 1.5 + 0.25 t, but max. t or 5 mm A = 4 mm initial crack = 0.1 mm nom = 104 MPa
6 N = 0.044 10 cycles 6 N f = 2.3% of 2 10 cycles
Modelling: R1
initial crack
6 5 acrack [mm] 4 3 2 1
60 40 20 0 0 2 4 6
acrack [mm]
0 0
N [-]
4 x 10
5
4
5 Compilation of STD5605,51
105
Locally A 0.05t, but max. 0.5 mm. l 25 mm 0.91 106 cycles 45% of 2 106 cycles
A Not permitted
B CU C D
1
DU
106
STD5605,51 Case 20 B
Description: Undercut Requirements and results: Locally A 0.05 t, but max. 0.5 mm. l 25 mm A = notch depth + initial crack = 0.4 + 0.1 = 0.5 mm nom = 82 MPa
6 N = 0.91 10 cycles 6 N f = 45% of 2 10 cycles
Modelling:
R1
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
1 2 3 4 acrack [mm] 5 6
4 3 2 1 0 0 2 4 6 8 10 x 10
5
N [-]
5 Compilation of STD5605,51
107
STD5605,51 Case 20 C
Description: Undercut Requirements and results: Locally A 0.1 t, but max. 1 mm. A = 1 mm initial crack = 1.0 mm nom = 82 MPa
6 N = 0.34 10 cycles 6 N f = 17% of 2 10 cycles
Modelling:
A = initial crack
z Stress intensity factor as a function of crack length: Crack length as a function of number of cycles:
Keff(acrack)
40 30 20 10 0 0
6 5
acrack(N)
acrack [mm]
1 2 3 4 acrack [mm] 5 6
4 3 2 1 0 0 1 2 N [-] 3 x 10 4
5
108
STD5605,51 Case 20 CU
Description: Undercut Requirements and results: Locally A 0.1 t, but max. 1 mm. A = notch depth + initial crack = 0.9 + 0.1 = 1 mm nom = 82 MPa
6 N = 0.54 10 cycles 6 N f = 27% of 2 10 cycles
Modelling: R1
initial crack
Stress intensity factor as a function of crack length: Crack length as a function of number of cycles:
Keff(acrack) acrack(N)
50 40
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
3 N [-]
5 x 10
6
5
5 Compilation of STD5605,51
109
STD5605,51 Case 20 D
Description: Undercut Requirements and results: Locally A 0.2 t, but max. 2 mm. A = 2 mm initial crack = 2.0 mm nom = 82 MPa
6 N = 0.13 10 cycles 6 N f = 7% of 2 10 cycles
Modelling:
A = initial crack
40 30 20 10 0 0
6 5 acrack [mm] 4 3 2 1
3 4 acrack [mm]
0 0
N [-]
10
15 x 10
4
110
STD5605,51 Case 20 DU
Description: Undercut Requirements and results: Locally A 0.2 t, but max. 2 mm. A = notch depth + initial crack = 1.9 + 0.1 = 2 mm nom = 82 MPa
6 N = 0.20 10 cycles 6 N f = 10% of 2 10 cycles
Modelling:
R1
initial crack
80 60 40 20 0 0
6 5
N [-]
5 Compilation of STD5605,51
111
B C
CU D
DU
112
leg does not affect the life of the joint substantially and therefore the leg could be very long and still have sufficient life. Thus the acceptance limits should not be relaxed too much (if at all). The current acceptance limits for C and D might intentionally have been set identically just to avoid too long legs. The present acceptance limits have no influence on the life and therefore, from a fatigue point of view, this case could be removed from the standard. It is not recommended to delete the case from the standard entirely since some kind of limitations for the leg deviation is important for reasons of appearance as well as for technical reasons. However, it should be clear that the leg deviation has no influence of the fatigue properties.
5 Compilation of STD5605,51
113
STD5605,51 Case 21 A
Description: Leg deviation Requirements and results: A 0.5 + 0,15 a A = 1.25 mm, a = 5 mm initial crack = 0.1 mm nom = 82 MPa
6 N = 1.62 10 cycles 6 N f = 81% of 2 10 cycles
Modelling:
R4
initial crack
30 25 [MPa(m)]
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
eff
0 0
0.5
1 N [-]
1.5 x 10
2
6
114
STD5605,51 Case 21 B
Description: Leg deviation Requirements and results: A 1.5 + 0.15 a A = 2.25 mm, a = 5 mm initial crack = 0.1 mm nom = 82 MPa
6 N = 0.90 10 cycles 6 N f = 45% of 2 10 cycles
Modelling:
R1
initial crack
30 25
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
10 x 10
5
5 Compilation of STD5605,51
115
Modelling:
initial crack
40 30 20 10
6 5
Keff [MPa(m)]
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
4 N [-]
6 x 10
8
5
116
Modelling:
R1
initial crack
30 25
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
10 x 10
5
5 Compilation of STD5605,51
117
B C CU D DU
118
throat becomes too small [5]. This has to be considered when reviewing the acceptance limits for throat deviation. Furthermore, a too small throat should always be handled with precaution. The model used has its crack in the sheet, and the only effect of the deviated throat is a somewhat higher stress intensity around the crack tip. On the other hand, a crack directly in the weld could have serious consequences if the weld is substantially weakened by throat deviation.
5 Compilation of STD5605,51
119
STD5605,51 Case 22 C
Description: Throat deviation Requirements and results: Locally -0.1 a a = 5-0.5 mm = 4.5 mm initial crack = 0.1 mm nom = 82 MPa
6 N= 0.59 10 cycles 6 N f = 30% of 2 10 cycles
Modelling:
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
3 N [-]
5 x 10
6
5
120
STD5605,51 Case 22 CU
Description: Throat deviation Requirements and results: Locally -0.1 a a = 5-0.5 mm = 4.5 mm initial crack = 0.1 mm nom = 82 MPa
6 N = 0.69 10 cycles 6 N f = 34% of 2 10 cycles
Modelling:
R1
initial crack
30 25
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
4 N [-]
6 x 10
8
5
5 Compilation of STD5605,51
121
STD5605,51 Case 22 D
Description: Throat deviation Requirements and results: Locally -0.2 a a = 5-1 mm = 4 mm initial crack = 0.1 mm nom = 82 MPa
6 N= 0.45 10 cycles 6 N f = 23% of 2 10 cycles
Modelling:
initial crack
40 30 20 10
6 5
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
N [-]
4 x 10
5
5
122
STD5605,51 Case 22 DU
Description: Throat deviation Requirements and results: Locally -0.2 a a = 5-1 mm = 4 mm initial crack = 0.1 mm nom = 82 MPa
6 N = 0.53 10 cycles 6 N f = 26% of 2 10 cycles
Modelling:
R1
initial crack
30 25
Keff [MPa (m)]
Keff (acrack)
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
3 N [-]
5 x 10
6
5
5 Compilation of STD5605,51
123
CU
DU
124
effect on the results. In the normal case with no penetration, the force has to flow around the slit, and then concentrating towards the surfaces (where the initial crack is). Here, the force is more uniformly distributed over the cross section, and that gives a lower stress intensity factor for small crack lengths. Approaching half the thickness, the stress intensity exhibits more normal values, as compared to other cases. The low stress intensity for small cracks results in long life. However, considering the significant geometrical differences, the fatigue lives are difficult to compare with other cases.
5 Compilation of STD5605,51
125
STD5605,51 Case 23 A
Description: Connecting radius. Fully penetrated Tweld joint Requirements and results: r 4 min initial crack = 0.1 mm nom = 82 MPa
6 N = 3.11 10 cycles 6 N f = 155% of 2 10 cycles
Modelling:
R4
initial crack
30 25
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
2 N [-]
3 x 10
4
6
126
STD5605,51 Case 23 B
Description: Connecting radius. Fully penetrated Tweld joint Requirements and results: Even transition r = 3 mm is used initial crack = 0.1 mm nom = 82 MPa
6 N = 2.92 10 cycles 6 N f = 146% of 2 10 cycles
Modelling:
R3
initial crack
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
0.5
1.5 N [-]
2.5 x 10
3
6
5 Compilation of STD5605,51
127
STD5605,51 Case 23 CU
Description: Connecting radius. Fully penetrated Tweld joint Requirements and results: No requirements in welding class r = 2 mm is used initial crack = 0.1 mm nom = 82 MPa
6 N = 2.74 10 cycles 6 N f = 137% of 2 10 cycles
Modelling: R2
initial crack
30 25
Keff [MPa(m)]
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
0.5
1.5 N [-]
2.5 x 10
3
6
128
STD5605,51 Case 23 DU
Description: Connecting radius. Fully penetrated T-weld joint Requirements and results: No requirements in welding class r = 1 mm is used initial crack = 0.1 mm
6 N = 3.14 10 cycles 6 N f = 157% of 2 10 cycles
Modelling:
R1
A = initial crack
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
0.5
1.5 N [-]
2.5 x 10
3
6
5 Compilation of STD5605,51
129
A B CU DU
130
In most cases this resulted, as expected, in shorter life. In all this cases the initial crack was moved to a higher x-value. However, in Case 24 for welding classes A, B and CU the highest stress concentration, and thus the place for the initial crack, was situated at lower x-coordinate values than in the original models. This resulted, somewhat surprisingly, in longer life. After all, if the crack grows from the point with the highest stress it could be expected to give approximately the shortest life, as in the other cases. But apparently there are some geometrical characteristics that make the stress concentration at the crack tip increase considerably slower for a crack at a lower x - value. Because of this, one can doubt if the x-coordinate for the maximum stress which the FEprogram provides is correct. There is, however, no indication of any error in the FE-model. It can be discussed whether it is correct to use the results from the models with the initial crack at the current locations. They obviously imply longer life than in the worst case. On the other hand, the initial crack is situated at the point where it is most likely to start growing. Despite the results in this case, putting the initial crack at the point with the highest stress concentration still seems to be the best idea for a general procedure. Neither in the cases with better (more probable) results is it known for sure how close they are to the actual minimum life.
5 Compilation of STD5605,51
131
STD5605,51 Case 24 A
Description: Connecting radius, fillet weld Requirements and results: r = 4 mm initial crack = 0.1 mm nom = 82 MPa
6 N = 1.47 10 cycles 6 N f = 74% of 2 10 cycles
Modelling:
R4
initial crack
30 25
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
10
15 x 10
5
132
STD5605,51 Case 24 B
Description: Connecting radius, fillet weld Requirements and results: r = 3 mm initial crack = 0.1 mm nom = 82 MPa
6 N = 1.27 10 cycles 6 N f = 64% of 2 10 cycles
Modelling: R3
initial crack
30 25
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
10
15 x 10
5
5 Compilation of STD5605,51
133
STD5605,51 Case 24 CU
Description: Connecting radius, fillet weld Requirements and results: r = 2 mm nom = 82 MPa
Modelling:
R2
initial crack
30 25
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
20 15 10 5 0 0 1 2 3 4 acrack [mm] 5 6
0 0
6 N [-]
10
12 x 10
5
134
STD5605,51 Case 24 DU
Description: Connecting radius, fillet weld Requirements and results: r = 1 mm nom = 82 MPa
Modelling: R1
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
N [-]
10 x 10
5
5 Compilation of STD5605,51
135
Inner crack with A 0,2t but max. 4 mm, and l t is permitted. Crack which reaches the surface or crack in the heat-affected zone is not permitted 0.92 106 cycles 46% of 2 106 cycles
C Not permitted
B Not permitted
A Not permitted
136
STD5605,51 Case 25 D
Description: Crack Requirements and results: Inner crack with A 0.2 t but max. 4 mm, and l t is permitted. Crack which reaches the surface or crack in the heat-affected zone is not permitted. A = 2 mm, l = 2c = 10 mm nom = 104 MPa
6 N= 0.92 10 cycles 6 N f = 46% of 2 10 cycles
l
Modelling and boundary conditions:
Modelling:
2a 2c
12 10
Keff [MPa (m)]
10 8 2acrack [mm] 6 4 2
8 6 4 2 0 2 3 4 5 6 2acrack [mm] 7 8 9
0 0
N [-]
10
15 x 10
5
5 Compilation of STD5605,51
137
A 0,2t, but max. 4 mm, and l t is permitted. Lack of fusion must not reach the surface 0.92 106 cycles 46% of 2 106 cycles
C Not permitted
B Not permitted
A Not permitted
138
STD5605,51 Case 26 D
Description: Lack of fusion Requirements and results: A 0,2 t, but max. 4 mm, and l t is permitted. Lack of fusion must not reach the surface. A = 2 mm, l = 2c = 10 mm nom = 104 MPa
6 N = 0.92 10 cycles 6 N f = 46% of 2 10 cycles
l
Modelling and boundary conditions:
Modelling:
2a 2c
10 8 2acrack [mm] 6 4 2
10 8 6 4 2 0 2 4 6 2acrack [mm] 8 10
0 0
N [-]
10
15 x 10
5
5 Compilation of STD5605,51
139
A 0,2t, but max. 2 mm. l 2 t, but max. 50 mm. Incomplete root penetration must not occur closer than 100 mm to the end of the weld or the crossing point respectively 0.73 106 cycles 6 36 % of 2 10 cycles
A 0,1 t, but max. 2 mm. l 0,4 t, but max. 20 mm. Incomplete root penetration must not occur closer than 100 mm to the end of the weld or the crossing point respectively 2.95 106 cycles 6 147% of 2 10 cycles
A Not permitted
140
Internal cracks are not favourable since they are hard to detect before they have reached the surface. When the crack has reached the surface it might in some cases be too late do anything about it.
5 Compilation of STD5605,51
141
STD5605,51 Case 27 B
Description: Incomplete root penetration Requirements and results: A 0,1 t, but max. 2 mm. l 0,4 t, but max. 20 mm. Incomplete root penetration must not occur closer than 100 mm to the end of the weld or the crossing point respectively. A = 1 mm, l = 4 mm nom = 104 MPa
6 N = 2.95 10 cycles 6 N f = 147% of 2 10 cycles
l
Modelling and boundary conditions:
Modelling:
2a 2c
10 8
Keff [MPa (m)]
10 8 2acrack [mm] 6 4 2
6 4 2 0 2 4 6 2acrack [mm] 8 10
0 0
N [-]
4 x 10
5
6
142
STD5605,51 Case 27 C
Description: Incomplete root penetration Requirements and results: A 0,2 t, but max. 2 mm. l 2 t, but max. 50 mm. Incomplete root penetration must not occur closer than 100 mm to the end of the weld or the crossing point respectively. A = 2 mm, l = 20 mm nom = 104 MPa
6 N = 0.73 10 cycles 6 N f = 36% of 2 10 cycles
l
Modelling and boundary conditions:
Modelling:
2a 2c
10 8 2acrack [mm] 6 4 2
10
0 2
6 2acrack [mm]
10
0 0
N [-]
10 x 10
5
5 Compilation of STD5605,51
143
STD5605,51 Case 27 D
Description: Incomplete root penetration Requirements and results: A 0,2 t, but max. 4 mm. Incomplete root penetration must not occur closer than 100 mm to the end of the weld or the crossing points respectively. A = 2 mm, l = infinite nom = 104 MPa
6 N = 0.61 10 cycles 6 N f = 31% of 2 10 cycles
l
Modelling and boundary conditions:
Modelling:
A = initial crack
10 8 2acrack [mm] 6 4 2
25 20 15 10 5 0 2 3 4 5 6 2acrack [mm] 7 8 9
0 0
4 N [-]
6 x 10
8
5
144
6 Compilation of ISO5817
145
6 Compilation of ISO5817
This chapter contains overall information on the CEN ISO5817 weld standard. Also results and some detailed conclusions and recommendations from all calculated cases in the standard are presented in diagrams, tables and on result sheets.
6.2 Modelling
All transverse butt welds in the standard with the requirement not permitted have been modelled with Case 6A in the STD5605,51 standard, i.e., Case 213 without the transition radius of 1 mm. Case 213 with the transition radius of 1 mm has been used when smooth transition is required. Corresponding cases for fillet welds with the requirement smooth transition have been modelled with Case 413, while ISO5817 Case 1.20B has been used if smooth transition is not prescribed. Further comments on modelling are given in Chapter 3 and in the result sheets for each case and class.
6.3 Results
Results with pictures and diagrams for each case and welding class can be found on the following pages. All results are also compiled in Figure C.1-C.11 and in Table C.1-C.2, in Appendix C. An overview of the results can be found in Figure C.1. It shows large scatter in the results. However, patterns can be discerned. For example, some cases have nearly no difference in life between the welding classes, while some show large differences. Heavy grouping of the cases can be seen in Table 6.1.
146
Analysis of Fatigue Life in Two Weld Class Systems Table 6.1. Influence of acceptance limits on the fatigue life for a certain welding class.
Large influence Some influence No influence
Butt welds Fillet welds Butt welds Fillet welds Butt welds Fillet welds
1.7
As can be seen in Table 6.1 especially the acceptance limits for the fillet welds have very little influence on the fatigue life. It can also be seen in Table 6.1 and Figure C.1 that there are only five cases out of 19 calculated where the acceptance limits have a major impact on the fatigue life. This is quite remarkable. In other words, fatigue properties have not been thoroughly evaluated when the standard was written. The same conclusion can easily be drawn when looking at Figure C.2 and C.3 for transverse butt welds and fillet welds respectively. The poor fatigue properties are most clearly seen in Figure C.3 for fillet welds, where no class show the desired pattern of a gradually decreasing life with a gradually decreasing welding class. In a few cases the undesirable pattern can be explained by the fact that the defect for classes B and C are not permitted. They have therefore been modelled with Case 213 or 413, which give the same lives for both B and C. The ratios between B and C and C and D have been plotted in Figure C.5. Many cases have the ratio one between the classes while a few cases show a different pattern. A ratio of one between the welding classes simply means no difference in fatigue life between the welding classes. The deviant value for Case 2.12, lack of fusion, depends on an internal defect where the initial crack is 4 mm. Thus, the crack can only be integrated 1 mm before half the thickness has been reached. This results in the short fatigue life and large ratio between welding class C and D. Case 1.14, 1.17 (fillet weld), 3.1 (one-sided and double sided welding) and 3.2 can be lifted out from the standard along with Case 1.7 (fillet and butt weld). The results for these cases are plotted in Figure C.4. Case 2.12 and 2.13 are left out because they are internal defects. The cases presented in Figure C.4 have a connection between the acceptance limits and the fatigue life, something which separates them from the rest. If the ratios between the welding classes are plotted for these cases, see Figure C.6, one can see that many of the cases have a ratio of 1.5-2.5 between B and C. The ratio between C and D varies more. Half of the cases have a ratio of about 1 to 2 while the other half have a ratio of about 2.5 to 3.5. It should be said that the same pattern as described above was found in STD5605,51. Many cases had a ratio of approximately 2 between B and CU (compared to B and C in ISO5817) The ratio between CU and DU varied a lot (compare with C and D in ISO5817). In Figure C.7 the maximum, minimum, average and median value for each welding class has been plotted. The results are more less scattered than for the STD5605,51 standard i.e., the maximum and the minimum value in each welding class do not differ more than a factor of ten. This is true if the maximum value in welding class D is disregarded. The maximum value
6 Compilation of ISO5817
147
in welding class D comes from Case 1.13, overlap (or coldlap), where the defect is a horizontal crack which makes a 90 turn before growing vertically into the material. In welding class B and C, this type of defect is not permitted and Case 213 without the transition radius gives much shorter life than the cold lap. This makes it hard to compare this case with the others and the deviating large maximum value can be disregarded when studying Figure C.7. Further, if the maximum value in welding class D is disregarded, the maximum, minimum, average and median value decreases when moving from welding class B to D, something which is expected, since a higher class ought to give a longer life. Cases 1.7, 1.14, 1.17, 3.1 and 3.2 all show influence from the acceptance limits on the fatigue life. The lives of these cases have been plotted versus the acceptance limit in Figure C.8. An exponential trendline has been fitted to the points and the equation for the line is fitted in Figure C.8. The trendline could serve as a tool when setting new acceptance limits. For example, the acceptance limit can be read on the curve if the desired life in a certain welding class has been decided. Of course this might not give the exact life for the certain case, but it will at least be possible to get a notion about in what hundred the desired acceptance limit might be. Similar more extensive data can be found in Figure B.9-B.12 for the STD5605,51. These data can be a great help if new acceptance limits are to be set. It is not possible to create an appropriate trendline for all the calculated data, since they are too scattered and many of the acceptance limits for different defects do not have any influence on the fatigue life. In Figure C.9-C.11, results for each welding class can be found. These results also confirm large scatter in the results.
148
6 Compilation of ISO5817
149
R1
initial crack
40 30
6 5
acrack [mm]
4 3 2 1
20 10 0 0
3 4 acrack [mm]
0 0
N [-]
10 x 10
5
150
Modelling:
R1
initial crack
50 40
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
3 N [-]
5 x 10
6
5
6 Compilation of ISO5817
151
152
Modelling: R1
initial crack
50 40
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1 0 0 2
30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
4 N [-]
6 x 10
8
5
6 Compilation of ISO5817
153
Modelling: R1
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
4 x 10
5
5
154
h = 2.5 mm, b = 10 mm
1.76 106 cycles 88% of 2 106 cycles
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
6 Compilation of ISO5817
155
Modelling: R1
initial crack
40 30 20 10 0 0
6 5
156
R1
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
6 Compilation of ISO5817
157
Modelling: R1
initial crack
40 30 20 10
6 5
acrack [mm]
1 2 3 4 acrack [mm] 5 6
4 3 2 1
0 0
0 0
0.5
1 N [-]
1.5 x 10
2
6
158
6 Compilation of ISO5817
159
Modelling:
initial crack
40 30 20 10
6 5
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
4 N [-]
6 x 10
8
5
160
Modelling:
initial crack
40 30
6 5
acrack [mm]
4 3 2 1
20 10 0 0
3 4 acrack [mm]
0 0
4 N [-]
6 x 10
8
5
6 Compilation of ISO5817
161
Modelling:
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
4 N [-]
6 x 10
8
5
162
6 Compilation of ISO5817
163
Modelling:
initial crack
40 30 20 10
6 5
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
164
Modelling:
initial crack
40 30 20 10 0 0
6 5
Keff [MPa(m)]
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
6 Compilation of ISO5817
165
Modelling:
initial crack
40 30 20 10
6 5
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
166
0.1 mm
initial crack
Figure 6.1. Modelling of the transition at the weld toe. This case could provide interesting information about the influence of the transition angle on the fatigue life, especially since there is no corresponding case in STD5605,51. However, the acceptance limits for this case give no significant differences at all, even though the angle
6 Compilation of ISO5817
167
difference of 20 degrees between the classes is a seemingly distinct difference that could have been expected to have a noticeable effect on the results.
168
Modelling:
initial crack
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
4 N [-]
6 x 10
8
5
6 Compilation of ISO5817
169
Modelling:
initial crack
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
4 N [-]
6 x 10
8
5
170
0.1 mm
initial crack
Figure 6.2. Modelling of the transition at the toe. The transition between the weld reinforcement and the sheet metal is a singular point. I.e., the finer the mesh the larger stress is obtained. The stress goes towards infinity at a sharp corner. Therefore there is no need to evaluate any stress concentration at the transition. In real life there will never be a sharp transition but there is always a transition radius. Since the crack initiation phase is not studied, no investigation will be performed on how much longer life there is in this phase. This case could provide interesting information about the influence of the transition angle on the fatigue life, especially since there is no corresponding case in STD5605,51. However, the
6 Compilation of ISO5817
171
acceptance limits for this case give no significant differences at all, even though the difference in angle between 90 and 150 degrees is seemingly distinct and could have been expected to have noticeable effects on the results.
172
Modelling:
initial crack
initial crack
40 30
6 5
acrack [mm]
4 3 2 1
20 10 0 0
3 4 acrack [mm]
0 0
N [-]
10
15 x 10
5
6 Compilation of ISO5817
173
Modelling:
initial crack
initial crack
40 30
6 5
acrack [mm]
4 3 2 1
20 10 0 0
3 4 acrack [mm]
0 0
N [-]
10
15 x 10
5
174
Modelling:
initial crack
initial crack
40 30
6 5
acrack [mm]
4 3 2 1
20 10 0 0
3 4 acrack [mm]
0 0
N [-]
10
15 x 10
5
6 Compilation of ISO5817
175
Not permitted
1.58 106 cycles 79% of 2 106 cycles
B Not permitted
[21] is used. The equation describes the kink angle the crack will grow with. Equation (6.1) combines the stress intensity factors in mode I and II ( K I and K II ) to calculate the angle.
Step 1
Starting condition is that the crack, or the overlap, is horizontal (see Figure 6.3).
176
initial crack
Figure 6.3. The horizontal initial crack. First the stress intensity factors are calculated for the initial crack (they become almost equal in mode I and II). The stress intensity factors are then used in Equation (6.1) to calculate the angle, at which the crack will continue to grow. In this case the angle becomes 1 = 54 .
Step 2
The FE-model is rebuilt and the crack is chosen to become 0.1 mm longer in the 1 = 54 . The increment of the crack does not follow any given equation but is chosen freely. The model now looks like in Figure 6.4. overlap
1 = 54
crack tip
Figure 6.4. The crack has turned down into the sheet. The stress intensity factors are once again calculated. Now K I becomes approximately twice the size of K II , because the crack is loaded more in mode I now. If these values are inserted into Equation (6.1) the angle becomes 2 = 36 . This means that the crack has now
6 Compilation of ISO5817
177
turned tot = 1 + 2 = 54 36 = 90 , i.e. it is now moving vertically down. The crack is once again set to grow with an increment of 0.1 mm (this time vertically, see Figure 6.5). Then a general macro can take care of the remaining calculations of the crack growth. The stress intensity factors for mode I and II are calculated and combined using Equation (6.1). The life is then integrated, but this time the overlap is taken into account when calculation the life.
overlap
1 = 54 2 = 36
crack tip
Figure 6.5. The crack has turned 90 and is now propagating vertically.
178
Modelling:
h = initial crack
40 30
8 6 acrack [mm] 4 2 0 0
20 10 0 2
5 6 acrack [mm]
0.5
1.5 N [-]
2.5 x 10
3
6
6 Compilation of ISO5817
179
C D
180
Modelling: R1
initial crack
50 40
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
10 x 10
5
6 Compilation of ISO5817
181
Modelling: R1
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
4 x 10
5
5
182
Modelling: R1
initial crack
100 80
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
60 40 20 0 0 2 4 6
acrack [mm]
0 0
N [-]
10
15 x 10
4
6 Compilation of ISO5817
183
ISO5817 Case 1.16 Excessive asymmetry of fillet weld (excessive unequal leg length)
Requirements and results
D h 2 mm + 0.2a C h 2 mm + 0.15a B h 1.5 mm + 0.15a
184
Modelling:
initial crack
40 30 20 10 0 0
6 5
acrack(N)
acrack [mm]
1 2 3 4 acrack [mm] 5 6
4 3 2 1 0 0 2 4 N [-] 6 x 10 8
5
6 Compilation of ISO5817
185
Modelling:
initial crack
40 30 20 10 0 0
6 5 acrack [mm] 4 3 2 1
acrack(N)
3 4 acrack [mm]
0 0
4 N [-]
6 x 10
8
5
186
Modelling:
initial crack
40 30
6 5
acrack [mm]
1 2 3 4 acrack [mm] 5 6
4 3 2 1
20 10 0 0
0 0
4 N [-]
6 x 10
8
5
6 Compilation of ISO5817
187
188
Modelling: R1
initial crack
Stress intensity factor as a function of crack length: Crack length as a function of number of cycles:
Keff(acrack)
50 40
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
acrack(N)
30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
10 x 10
5
6 Compilation of ISO5817
189
Modelling: R1
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
4 x 10
5
5
190
Modelling: R1
initial crack
100 80
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
60 40 20 0 0 2 4 6
acrack [mm]
0 0
N [-]
10
15 x 10
4
6 Compilation of ISO5817
191
192
Modelling:
initial crack
40 30 20 10
6 5
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
4 N [-]
6 x 10
8
5
6 Compilation of ISO5817
193
Modelling:
initial crack
40 30 20 10
6 5
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
3 N [-]
5 x 10
6
5
194
Unlimited.
6 Compilation of ISO5817
195
Modelling:
initial crack
40 30 20 10 0 0
6 5
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
N [-]
10
15 x 10
5
196
Modelling:
initial crack
40 30
6 5
acrack [mm]
4 3 2 1
20 10 0 0
3 4 acrack [mm]
0 0
0.5
1 N [-]
1.5 x 10
2
6
6 Compilation of ISO5817
197
ISO5817 Case 2.12 Lack of fusion (incomplete fusion) - Lack of side wall fusion
Requirements and results
D Short imperfections permitted, but not breaking of the surface h 0.4s, but max. 4 mm h = 4 mm, s = 10 mm C B Not permitted
Not permitted
D
1
198
remarkable that the acceptance limit is 4 mm for Case 2.12, while it is only 2 mm for Case 2.13, when the defects in Case 2.12 and 2.13 (butt weld) are very similar.
6 Compilation of ISO5817
199
initial crack
10 8 2acrack [mm] 6 4 2
25 20 15 10 5 0 2 3 4 5 6 2acrack [mm] 7 8 9
0 0
0.5
N [-]
1.5
2.5 x 10
5
200
Not permitted
B Not permitted
6 Compilation of ISO5817
201
initial crack
slit
Figure 6.6. Placement of the initial crack in welding class D The crack growth direction is calculated with the equation
3K 2 + K K 2 + 8 K 2 I I II = arccos II 2 K I2 + 9 K II
(6.1)
[21] which has previously been used in Case 1.13. However, one problem arises; the Equation (6.1) can only turn in one direction because K I is always greater than zero in this case. The x reason why K I > 0 is that the crack is never closed and further, 0 arccos 180 . The r crack strives to be loaded solely in mode I and the problem is that Equation (6.1) is unstable if the crack turns too much. For example, if the value = 5.6 is received and is rounded down to = 6 , then K II might grow more than K I , resulting in a greater change of angle for the next step. In the end this leads to an unrealistic result. This proves to be the case when Equation (6.1) is used for several steps in the crack growth process. After a few steps the crack has turned almost parallel to the horizontal tension. This is unrealistic and therefore the process is stopped. In order to overcome this problem, Equation (6.1) is used as long as the angle is decreasing with each step. When a calculation results in greater change in angle than the previously calculated angle, the process is stopped, and the crack is assumed to continue growing in the direction of previously calculated angle until half of the throat has been reached. The crack is set to grow with an increment of 0.1 mm for the first steps. In the first step the crack makes a turn of about 20. In the second step the turning angle has decreased to about 05, and in the following step the angle starts to grow again. The process is stopped at the second step, see Figure 6.7.
202
Figure 6.7. Step 1 and 2 which lead to the final crack growth direction. As can be seen in Figure 6.7 the desired crack growth direction from Equation 6.1 is unrealistic.
6 Compilation of ISO5817
203
Modelling:
initial crack
12 10
Keff [MPa (m)]
0 0
4 N [-]
6 x 10
8
5
204
initial crack
15
3 2.5
acrack [mm]
10
2 1.5 1 0.5
0 0
0.5
2.5
0 0
2 N [-]
3 x 10
4
5
6 Compilation of ISO5817
205
Not permitted
1.58 106 cycles 6 79% of 2 10 cycles
B Not permitted
206
Modelling:
h = initial crack
10 8 2acrack [mm] 6 4 2
25 20 15 10 5 0 2 3 4 5 6 2acrack [mm] 7 8 9
0 0
4 N [-]
6 x 10
8
5
6 Compilation of ISO5817
207
208
Modelling:
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
3 N [-]
5 x 10
6
5
6 Compilation of ISO5817
209
Modelling:
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
2 N [-]
3 x 10
4
5
210
Modelling:
initial crack
80 60 40 20
6 5
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
6 N [-]
10
12 x 10
4
6 Compilation of ISO5817
211
212
Modelling:
initial crack
60 50
Keff [MPa(m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
4 N [-]
6 x 10
8
5
6 Compilation of ISO5817
213
Modelling:
initial crack
60 50
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
40 30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
2 N [-]
3 x 10
4
5
214
Modelling:
initial crack
80 60 40 20
6 5
acrack [mm]
4 3 2 1
0 0
3 4 acrack [mm]
0 0
6 N [-]
10
12 x 10
4
6 Compilation of ISO5817
215
216
Analysis of Fatigue Life in Two Weld Class Systems high bending stress initial crack
Figure 6.9. The initial crack on the left side. The angular misalignment results in bending stresses at the fixed right side of the structure. These bending stresses decline towards the left end of the structure in Figure 6.8 and 6.9. This results in higher stresses if the crack is placed on the right side of the weld reinforcement, which further leads to higher stress intensity factors and a shorter life. Neither of the ways of modelling this case is wrong - it depends on the situation in real life. Here, the worst case i.e., the initial crack on the right side is chosen when modelling the rest of the classes.
6 Compilation of ISO5817
217
Modelling:
initial crack
40 30 20 10 0 0
6 5
Keff [MPa(m)]
acrack [mm]
4 3 2 1
3 4 acrack [mm]
0 0
6 N [-]
10
12 x 10
5
218
Modelling:
initial crack
50 40
Keff [MPa(m)]
6 5 acrack [mm] 4 3 2 1
30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
N [-]
10 x 10
5
6 Compilation of ISO5817
219
Modelling:
initial crack
50 40
Keff [MPa (m)]
6 5 acrack [mm] 4 3 2 1
30 20 10 0 0 1 2 3 4 acrack [mm] 5 6
0 0
3 N [-]
5 x 10
6
5
220
221
7.1 Introduction
The aim of this study is to investigate the effect of the sheet thickness on the fatigue life. The study is performed on a range of thicknesses from 5 to 40 mm for one particular case in STD5605,51; Case 20, undercut (fillet weld). An important objective is to review the applicability of the 10 mm results on other sheet thicknesses. As the thickness 10 mm is used throughout the whole work, it is very important to know to what extent these results are valid for other thicknesses, which factors that affect this, and how they can be considered. Along with variation in thickness, also the throat thickness was varied. This was done so that the ratio between the throat thickness and the sheet thickness was 0.5, i.e., a t = 0.5 . The same ratio has previously been used for all cases in both standards investigated. However, this leads to an unrealistic throat thickness for the thicker sheets. For example, the throat is 20 mm for a 40 mm thick sheet. In order to examine how the fatigue life is affected by other loads than tension (worst case), also a bending stress is investigated.
t t
notch
222
Two types of load conditions are tested; one with uniaxial tension (see Figure 7.1) and one with a bending moment applied on both ends (see Figure 7.2). In both cases the same boundary conditions are applied, see Figure 7.1 and 7.2. t t
notch
h Figure 7.2. Cruciform joint subjected to bending. The maximum stress at bending, as well as the stress at pure tension, is the same as the FATvalue for the corresponding Case 413 in IIW [6], i.e., = 82 MPa.
7.2.1 Requirements
According to the STD5605,51 standard, the requirements according to Table 7.1 apply to undercuts for cruciform joints,. Table 7.1. Requirements for Case 20 in STD5605,51.
D C B A
Not permitted
Note that the maximum value in classes C, D and B are reached already for a thickness of 10 mm. This means that for t > 10 mm, the relative measure of the acceptance limit decreases with increasing thickness.
223
7.3 Modelling
The undercut is modelled differently depending on the welding class. In Figure 7.3 the notch area in Figure 7.1 and 7.2 is described for each welding class. Since undercut is not permitted for welding class A, this class has been modelled with an initial crack of 0.1 mm at the transition radius. According to the standard, welding class A should have a transition radius of 4 mm. The initial crack has been inserted at the place where the stress concentration has its maximum. This means that the initial crack has been inserted at different places depending on if the structure is subjected to bending or tension.
R4
R1
A
R1
initial crack CU
initial crack D
A = initial crack
DU
R1
A
R1
R1
A
initial crack
A = initial crack
initial crack
Figure 7.3. Modelling of the undercut for each welding class. Classes B, CU, and DU have been modelled as a notch with the transition radius 1 mm. The depth of the notch is the acceptance limit A minus the initial crack length of 0.1 mm. Classes C and D have been modelled the same way, i.e. with an initial crack starting after the end of the transition radius. The accepted measure A has served as the length of the initial crack. The reaction forces at the supports are checked for each model and type of boundary conditions. Results show that the reaction forces are very small, below 1 10 5 N, and can therefore be neglected.
224
7.3.1 Dimensions
Six different thicknesses are tested: 5, 10, 15, 20, 30 and 40 mm. Each thickness and each welding class has been subjected to both tension and bending, resulting in 72 different simulations. The throat thickness is always half of the sheet thickness. (This is in fact not fully appropriate for thicker plates, where weld preparation is often used resulting in other throat thicknesses.)
3000
2500
Keff [MPa(mm)]
2000
1500
1000
500
10
15
Figure 7.4. Stress intensity factor as a function of crack length for different sheet thicknesses (bending).
225
The results show that the length of the horizontal sheet has a negligible influence on the stress intensity factors when the distance is 25 mm or more from the weld toe. However, it can be seen that the length of the horizontal sheet has an influence on the stress intensity factors for distances up to around 20 to 25 mm at bending.
Keff (a) for different distances from edge to weld, for tension
6000 5 mm 10 mm 15 mm 125 mm
5000
4000
3000
2000
1000
10
15
Figure 7.5. Stress intensity factor as a function of crack length for different sheet lengths (tension). The results at tension are similar to the results at bending. The sheet length affects the stress concentration at the crack tip when the distance from weld to end is less than approximately 25 mm. However the differences for small cracks are negligible. Only at a crack depth of about 0.5 mm significant differences appear. Because of the small differences in stress intensity factor for small cracks, the differences in life are quite modest compared to the bending case. Thus it, would not have any large influence if the sheets were modelled a bit too short. The conclusion of the sheet length study is that the modelled sheet lengths, giving a distance of 25 mm or more between weld toe and sheet end, are sufficient for both bending and tension. (As mentioned above, models with thicknesses 30 and 40 mm have been prolonged to obtain a distance of 25 mm from weld to sheet end. The 20 mm thickness model has a distance of approximately 26 mm from weld to end, and thinner sheets will have even longer distances.) It should be noticed that the case reviewed is welding class DU, which has the largest defect. Thus it is most probably the case that demands the longest sheets to give correct results. Smaller defects may not demand these sheet lengths. Another conclusion is that the modelled sheet length used in the entire report - 100 mm from end to end, giving a distance of 38 mm from weld to end - is by far on the safe side.
226
7.5 Results
The results of the parameter study are presented in Table 7.2-3 and Figure 7.6-11 below. Table 7.2. Fatigue life [106 cycles]; bending.
Welding class A B C CU D DU 5 10 Sheet thickness [mm] 15 20 30 40
227
10
35
40
45
Fatigue life as a function of thickness - welding class B 4.5 4 3.5 Number of cycles x 106 [-] 3 2.5 2 1.5 1 0.5 0 Bending Tension
10
35
40
45
228
1.5
0.5
10
35
40
45
10
35
40
45
229
1.5
0.5
10
35
40
45
Fatigue life as a function of thickness - welding class DU 1.6 1.4 Number of cycles x 106 [-] 1.2 1 0.8 0.6 0.4 0.2 0 Bending Tension
10
35
40
45
230
Life as function of sheet thickness (tension) 3,0 2,5 Life [1e6 cycles] 2,0 1,5 1,0 0,5 0,0 0 10 20 30 40 50 Thickness [mm]
A B C CU D DU
Life as function of sheet thickness (bending) 7,0 6,0 Life [1e6 cycles] 5,0 4,0 3,0 2,0 1,0 0,0 0 10 20 30 40 50 Thickness [mm]
A B C CU D DU
231
232
Another remarkable result is the small difference in life between 5 and 10 mm for class A. Since theses cases could be expected to give results in accordance with the thickness effect, the 5 mm sheet should have considerably longer life. There is no obvious explanation for this. The magnitude of the thickness dependence at bending should be further investigated, but this is not done in this report. However, the thickness effect, judging from the 5 and 10 mm results, seems to be somewhat smaller for bending than for tension.
7.7 Conclusions
An important objective of the thickness study was to investigate the applicability of the results from 10 mm sheets on other thicknesses. Since the thickness 10 mm is used throughout the in this work, it is very important to know to what extent these results are valid for other thicknesses. The main conclusion is that it is necessary to be cautious when applying a result to another thickness. The two counteracting effects described above makes the thickness dependence look different from case to case. Under certain circumstances it is possible to quantify the thickness effect. This, however, requires that thicknesses have proportional geometrical properties. For example, if two models with different thicknesses have a defect of the same depth, this will distort the result of the thickness effect. Handbook rules seem to be useful for calculating the thickness effect. Though, before using such, the choice of method and parameter settings should be further investigated. (In this case, tref = 15 mm and n = 0.25 in Equation 9.1 appear to give reliable results for class A.) Due to the absolute acceptance limits for many defect types, it is often difficult to generalize the results to larger thicknesses. As the results in this parameter study show, the life for thicknesses 15 and 20 mm is often longer than that for 10 mm. However, this totally depends on the combination of defect type and acceptance limits, and varies from case to case.
(7.1)
In Equation 7.1 t is the thickness, while the parameters tref (reference thickness) and n are obtained from some suitable handbook. Different standards use the thickness factor
233
differently. For example, in the IIW recommendations it is used only to reduce the allowed stress range for sheet thicknesses larger than tref, while in BSK 99 [8] it is used in the opposite way, i.e., to increase allowed stress for thinner sheets (tref = 25 mm in both standards). A thickness factor can also be used for both thinner and thicker plates (see below). It should be noticed that this applies to toe cracks, when the crack propagates through the sheet. For cracks through the weld (root cracks), the thickness effect is less investigated, but it is probably smaller (partly because there is no technological effect) [5]. In [5] is given one example of a method to calculate the thickness factor. It can be used for sheets both thinner and thicker than the reference thickness t ref , which for this method is 15 mm. For fillet welds the exponent n is recommended to be 0.15 for 4 < t 15 mm and 0.25 for t > 15 mm. The thickness factor, intended for increasing the allowed stress range, can also be used to calculate a longer life for the original stress range. If the thickness factor t is multiplied with the allowed stress range, the same life as for the reference thickness is obtained. But the life is inversely proportional to the third power of the stress, and thus (for identical stress) the factor in life is 3 t . The thickness factors and the life factors (compared to t ref = 15 mm), for the thicknesses used in this parameter study are listed in Table 7.4. Table 7.4. Thickness factor t (see Equation 7.1) and life factor as the function 3 t .
Life factor t n t (= t 3)
5 mm 10 mm 15 mm 20 mm 30 mm 40 mm
The life factor for each thickness compared to tref = 15 mm, as obtained from the results in this parameter study, are shown in Table 7.5 for all welding classes. These factors are also plotted in Figure 7.14.
234
Table 7.5. Normalized life factors computed from the results in the parameter study.
Life factor for welding class Thickness A B C CU D DU Theoretical value from Table 7.4
5 mm 10 mm 15 mm 20 mm 30 mm 40 mm
Figure 7.14. Plot of results from table 7.5. As can be seen in Figure 7.14 and Table 7.5, the life factors in a few cases correlate very well to the values obtained by computing backwards from the design recommendations. This is the case especially for thick sheets in welding class A. This is expected, since A is the ideal case without defect and therefore has the expected stress gradient causing the thickness effect. In other cases the defects may modify the stress distribution, possibly affecting the magnitude of the thickness effect. Furthermore, as discussed above, the effect of the absolute acceptance limits for t 10 mm has a very large impact on the lives in some cases. The deviation for thinner sheets in class A suggests that the value of n might be slightly incorrect (or maybe just conservative, in order to be on the safe side). If n = 0.25 is used also for 4 < t 15 mm, the thickness factors will give life factors that correlate well to the results obtained from the parameter study.
235
An interesting observation is that the ratio in life between thicknesses 5 and 10 mm is approximately the same for all welding classes, i.e. for the whole range of defect sizes (0-0.2 t). Since the defect is proportional to the thickness for thicknesses 5 and 10 mm, this implies that for the thickness effect calculation to be valid, the whole geometry should be scaled by the same factor as the thickness. It can also be noted that the results correlate well with the handbook rule, despite the fact that the statistical and technological effects, dealing with material properties, are not considered. This may be interpreted as these two factors having little influence, and thus the geometrical factor playing the major role in the thickness effect. However, this limited study does not support any extensive conclusions on this.
236
237
8.1 Conclusions
The main conclusions on the studied welding standards are the following: Today, both STD5605,51 and ISO5817 show large scatter in the results and there is no general relationship between the acceptance limits and the fatigue life. It would be possible, however not easy, to create a general standard. In addition, a few general conclusions can be drawn: The angle between weld and sheet does not have any significant influence on the fatigue life. Internal defects ought to be excluded from a general standard. Arbitrary denotations, for example locally permitted, smooth transition etc. should be avoided and replaced by numbers. These conclusions are thoroughly discussed in the following section. After this, a few proposals for revising the standards are presented, followed by discussions and recommendations for each standard based on the proposals.
initial crack
Figure 8.1. The transition angle between weld and sheet. In Case 1.12, incorrect weld toe, in the ISO5817 standard the acceptance limit is defined as an angle between the sheet and the weld. For transverse butt welds there is almost no difference in fatigue life for angles (see Figure 8.1) between 30 and 90. For fillet welds the angle can vary between 70 and 90, and it gives only one percentage point in difference in fatigue life.
238
This variation in angle is not so large for the fillet weld. In Case 1.20 B the transition angle is 45. If the result for Case 1.20 B is comperad to the results in Case 1.12, one can see that the transition angle is of no interest of the fatigue life for sharp transitions when the 0.1 mm initial crack is present. This is also reflected in many other cases, for example weld reinforcement, penetration bead and leg deviation. This conclusion is confirmed by for example Martinsson in Fatigue Strength of welded cruciform joint with cold laps [23]. In the article, the fatigue life is calculated on different geometries for different initial cracks.. The geometries have different stress concentration factor K t , for no initial crack available. For a 0.1 mm initial crack there are nearly no differences in fatigue life. One can see that the longer initial crack is, the less is the impact on the fatigue life. This can be explained by the fact that once the crack is there, the crack propagation is governed by the stress intensity factor at the crack, and the stress concentration due to the transition angle plays a minor role. It is possible that the transition angle has an effect during the crack initiation phase, and that the life is shorter for larger angles during this stage. This should be the case because the stress concentration factor increases with an increasing transition angle, under the condition that there is a transition radius present between weld and sheet, which there always is in real life. If, on the contrary, there is a sharp transition the stresses goes towards infinity for a singular point in the FE-model. The crack initiation phase, which could give a larger difference in life for different angles, is not included in this study. All internal defects are too difficult to deal with. Internal defects should be avoided since they cannot be detected with the human eye, but require some kind of other method for detection, for example ultrasound or x-ray. The internal defects may result in cracks which reach the surface and causes failure before they have been detected and taken care of. Therefore it is the standpoint of this work that the internal defects are excluded from the standard. Another aspect why internal defects are hard to include in a general standard is that it is difficult to predict the life of internal cracks with a limited extension in the length direction of the weld. Once they have reached the surface, they can still grow in the length direction, thus making it hard to find reasonable integration limits. If these defects are included, theory and reality might divert too much from each other and the standard will be useless from that point of view. Possibly, it could be better to have internal defects in a special section in the standard, and then omit fatigue properties. All vague denotations, for example locally permitted, permitted to a small extent, short imperfection, smooth transition etc., must be changed to give real information in the standard. If the standard is to be consistent, important factors influencing for the fatigue life can not be replaced by notations which can be interpreted arbitrarily.
239
240
questions will arise whether or not some of the cases should be deleted from the standard or completely revised. Some of the scattered data depend on only limitations when modelling several cases. This is, for example, because of the additional designation U requiring smooth transitions, that in some cases require completely different modelling of the different classes. In, for example Case 3, incomplete root penetration, welding classes C and D have been modelled with a crack according to the acceptance limits, i.e. 1 mm for C and 2 mm for D. Classes CU and DU however was modelled only with the initial crack length 0.1 mm, since it was considered impossible to have an smooth transition in a defect of this type. This resulted in very long lives for classes CU and DU, while C and D got significantly shorter lives. The standard must be revised, or the additional designation must be dropped, for cases like these if conformity is to be reached. The standard is inconsistent in the sense that Case 24 state that cruciform joints should have transition radii for welding class A (r = 4 mm) and B (r = 1 mm), while there is no corresponding requirement for transverse butt welds. This means that butt welds can never be assigned fatigue requirements in a higher welding class than CU! In order to emphasize the transition radius requirements for welding classes A and B it is recommended to move Case 24 from its current position as the last case for cruciform joints and place it either in the beginning of the standard or first among the cruciform joints. Similarly, a new case or some other type of definition ought to be introduced for transverse butt welds.
Linear misalignments for transverse butt welds are included in the standard, but there are no corresponding cases for fillet welds. As can be seen for Case 13 and 14, misalignment has a major effect on the fatigue life for transverse butt welds. A butt weld without any defect and transition radius, has a fatigue life of 1.58 million cycles, which corresponds to 79% of 2 million cycles (Case 6A in STD5605,51). For Case 13 and 14 the corresponding fatigue life of welding class A is only 30%. From the fatigue point of view the effect of misalignment would probably be a large threat for fillet welds as well. Another case which does not exist in the standard is angular misalignment obtained when the sheets connected by the weld do not lie in the same plane, see Case 3.2 in the ISO5817 standard, Chapter 6. Results from Case 3.2 show that this defect affects the life, since the introduction of bending aggravates the situation for the crack. Therefore it is suggested that angular misalignments for both transverse butt welds and fillet welds are added to the standard. One case that could be considered added to the standard is cold lap or overlap, see Case 1.13 in the ISO5817 standard, Chapter 6. Even if this case is not fully investigated here. Just one simulation of a horizontal crack turning vertical. Further studies with variations of the turning angle and crack growth length must be conducted before it can be decided whether or not this case should be included.
241
It could be difficult to fit cold laps into a general standard, because of the long lives. If included it could be placed among the cases which do not affect the fatigue life. Further, it is worth raising the question whether or not the size of the cold lap matters? Maybe it does not. A cold lap of 1 or 2 mm seems to give the same fatigue life. However further studies have to answer this question.
Cases which could be removed
Case 11, weld reinforcement, Case 12, penetration bead and Case 21, leg deviation, all deal with an angle between weld and sheet. As has been said above, the angle between the two surfaces does not affect the fatigue life noticeably when there is an initial crack present. These cases could be deleted from the standard if fracture mechanics with a known initial crack of 0.1 mm is regarded. However, before deleting these cases, the fatigue life in the crack initiation stage should be further investigated. Note that these three cases have an impact on the fatigue life but it is not the size of the defect but rather the initial crack that reduces the life. Case 3, incomplete root penetration and Case 8, root concavity are very similar. Therefore it is recommended to join these cases into one case i.e., remove either Case 3 or Case 8. Maybe the name should be changed to something that describes the new case better.
The undercut is the only defect which exists both for butt and fillet welds, and this case has the same acceptance limits i.e., no changes have to be done in order to achieve the same
242
acceptance limits for butt welds and fillet welds. Of course some acceptance limits might have to be changed if they dont affect the fatigue life as wanted. Achieved results from the FE simulations could easily be scaled in order to achieve the correct stress level for a given fatigue life and acceptance limit.
Secondary proposal for guidelines
Case 3, incomplete root penetration, needs major reworking for welding classes B, CU and DU, but it has some values which fit with the remaining cases. Further Case 22, throat deviation, needs extensive rework since the fatigue life is far too long and this case diverts too much from the other cases. Finally, Case 23 and 24, which both deal with the transition radius, are not ordinary cases but rather guidelines to the other cases since they have no real defects. Therefore they should be excluded when working for achieving the primary guideline. All other cases need minor revision.
Other proposals for guidelines
Since welding classes B, CU, and DU are all meant for fatigue loading, these classes are interesting to study more deeply. When plotting the ratio of fatigue life between B and CU, and between CU and DU (Figure B.7), one can see that the ratio between B and CU lie between 1 and 2 for all cases. It could be relatively easy to modify the acceptance limits to achieve a common ratio for these two classes. The ratio should of course should be greater than 1, preferably around 1.5-2. It is more difficult to find consistency when studying the ratio between CU and DU. The values are very scattered. If possible, the ratio should be the same or almost the same as the ratio between B and CU.
Angular and linear misalignments for transverse butt welds are available in the standard but there are no corresponding cases for fillet welds. As can be seen for Case 3.1 and 3.2 both angular and linear misalignment have a major effect on the fatigue life for transverse butt welds. A butt weld without any defect and transition radius have a fatigue life of 1.58 million
243
cycles, which is 79% of 2 million cycles (Case 6A in STD5605,51). For Case 3.1, the life of welding class B is 30% while it is 52% for the same class in Case 3.2. The effect of misalignment would probably have a large impact on the fatigue life for fillet welds as well.
Cases which could be removed
Case 1.9, excess weld metal (or weld reinforcement) and Case 1.11, excess penetration (or penetration bead) have proven - both in the ISO5817 and the STD5605,51 standard - to be useless from a fatigue point of view. The average life is very long and it does not differ much between the different classes. If a general life for each welding class should be set, the defect would be so large that it would probably not be tolerated from a production point of view. Thus, from a fatigue point of view the defect could be excluded from the standard. Case 1.10, excessive convexity, Case 1.12, incorrect weld toe and Case 1.16, excessive asymmetry of fillet weld (excessive unequal leg length or leg deviation) all deal with an angle between weld and sheet. As stated above (Chapter 8.1), the angle does not affect the fatigue life noticeably when there is an initial crack present. This means that these cases could be deleted from the standard if an initial crack of 0.1 mm is presumed. However, before deleting these cases the fatigue life in the crack initiation stage should be investigated. Note that these three cases have an impact on the fatigue life but it is not the size of the defect but the initial crack which reduces the life. Case 1.21, excessive throat thickness, should not be adopted in the standard since the life becomes longer with a worse welding class. This is impossible to fit with any of the guidelines for a new standard. Case 2.13, lack of penetration for a transverse butt weld, should be removed from the standard since internal defects are not possible to handle. Internal defects cannot be discovered by the human eye before breaking of the surface. This might lead to unexpected failures. Another undesirable factor is that it is hard to determine the integration limits for an internal crack which do not have an infinite length in the depth directions. The crack could continue growing after breaking the surface. This means that many cycles might remain and this fact makes it hard to determine the fatigue life for a structure with this defect.
Uncertain cases
Calculations on Case 1.13, overlap (or coldlap), show that this type of defect gives a longer life than might have been expected. However, this case is not fully investigated after just one simulation of a horizontal crack turning vertical. Further studies with variations of the turning of the crack must be conducted before it can be decided whether or not this case belongs in the standard. Case 2.13, lack of penetration for a fillet weld, could with changed acceptance limits easily be included in a revised standard. However with internal defects the crack might grow from the root. This is dangerous because such cracks are hard to discover before failure. Therefore, this case should perhaps be brought out into a special section of the standard, and the defect should not be permitted for any welding class.
244
Primary proposal for guidelines
A part of the primary prosal is already implemented in the ISO5817 i.e., the same acceptance limit apply for both butt and fillet welds. Only Case 1.12, incorrect weld toe and Case 2.13, lack of penetration, show some differences but these cases have already been decided to be dropped from the fatigue part of the standard. The acceptance limits in Case 1.12 does not have any impact on the fatigue life and Case 2.13 is an internal defect. Apart from these two cases, new acceptance limits have to be decided. If old acceptance limits are used, it will be easy to scale the results from the FE simulation conducted in this study. This will give the correct stress level for a given fatigue life.
Secondary proposal for guideline
If the secondary proposal of a clear-cut standard are to be achieved, all the remaining cases, Case 1.7, 1.14, 1.17, 1.20, 3.1, and 3.2, need some minor revision according to below. Case 1.7, undercut, in order to create a wider range in fatigue life needs to relaxe the maximum requirement in welding class B and C. Also the acceptance limits need rework. Case 1.14, sagging or incompletely filled groove and Case 1.17, root concavity, need minor revision regarding the acceptance limits. Case 1.20, insufficient throat thickness, can not have the same acceptance limit in welding class C and D. The only difference is on the maximum deviation. Except from this, only minor revision needs to be done. Case 3.1, linear misalignment, needs only minor revision regarding the acceptance limits. Case 3.2, angular misalignment, the acceptance limits can be increased i.e., greater angles can be tolerated in order to reduce the fatigue lives. However, this might be incompatible with production aspects.
Other proposals for guidelines
The cases mentioned above already have a good ratio between welding class B and C and between C and D. This as can be seen in Figure C.5 and C.6. A ratio of 1 depends on badly set acceptance limits. This could easily be adjusted in order to get better ratios. A ratio of 1.5 to 2.5 between B and C and between C and D is possible to achieve, with only some minor changes of the acceptance limits. If all cases were included it would be impossible to set acceptance limits which would give the same ratio between all the welding classes In some cases the angle between weld and sheet does not matter for the life.
245
Investigation of internal cracks to determine if it is possible to change the acceptance limits in order to fit them into the standards. This requires that, for example, integration limits in Paris law are investigated. Proposals for a new standard require many simulations before new acceptance limits can be determined. Different sheet thicknesses should be investigated. This is an extensive task that also need input from production engineers.
246
References
247
References
[1] Volvo Group Standard (1989). 5.501E Welding Manual Design and Analysis (contains the STD5605,51 standard also referred to as STD181-0001). Gothenburg, Sweden: Volvo Comit Europen de Normalisation (2003). ISO5817:2003, Welding Fusion-welded joints in steel, nickel, titanium and their alloys (beam welding excluded) Quality levels for imperfections. Brussels, Belgium: CEN Dahlberg T and Ekberg A (2002). Failure, Fracture, Fatigue an Introduction. Lund, Sweden: Studentlitteratur. ISBN 91-44-02096-1 Sundstrm B et al. (1999). Handbok och formelsamling i hllfasthetslra (in Swedish). Stockholm, Sweden: Institutionen fr hllfasthetslra KTH. Eriksson and Lignell A-M, Olsson C, Spennare H: Svetsutvrdering med FEM (in Swedish). Stockholm, Sweden: Industrilitteratur. ISBN 91-7548-636-9 Hobbacher A et al. (1996). Fatigue design and welded joints components. Cambridge: The International Institute of Welding. ISBN 1-8557-3315-3 Holm L, Mrtensson A, et al. (1987). Bestmmelser fr stlkonstruktioner, BSK (in Swedish). Stockholm, Sweden: Statens planverk och AB Svensk Byggtjnst. ISBN 91-7332-328-4 Gransson L, kerlund S et al. (2001). Boverkets handbok om Stlkonstruktioner, BSK 99 (in Swedish). Karlskrona, Sweden: Boverket. ISBN 91-7147-527-3 Per-Olof Danielsson, Lic. Eng. Volvo CE, previously Bombardier Transportation in Kalmar, Sweden Marquis G and Samuelsson J (2005). Modelling and Fatigue Life Assessment of Complex Fabricated Structures, Symposium on Structural Durability, in Darmstadt, 910 June 2005. Darmstadt, Germany
ANSYS Version 8.1. SAS IP, Inc ANSYS Version 9.0. SAS IP, Inc
[2]
Paris P.C. and Sih G.C. (1965). Stress Analysis of Cracks. Fracture Toughnesss and Testing and its Applications STP 381: 30-85 ANSYS (1995). ANSYS Theory Reference, seventh edition. USA: SAS IP, Inc
MATLAB Version 7.0.1.24704 (R14) Service Pack 1. The Mathworks, Inc
Eldn L and Wittmeyer-Koch L (2001). Numeriska berkningar analys och illustrationer med MATLAB. Lund, Sweden: Studentlitteratur. ISBN 91-4402-0074
Newman J.C. and Raju I.S. (1986). Stress Intensity Factor Equations for Cracks in Three-Dimensional Bodies Subjected to Tension and Bending Loads. Computational Methods in the Mechanics of Fracture Chapter 9. Hampton, USA: Elsevier Science Publishers B.V.
AFGROW Version 4.0009e.12. http://afgrow.wpafb.af.mil/index.php access date 2005-11-03
von Essen W et al. (1981). Byggsvetsnorm utgva 2, StBK-N2 (in Swedish). Stockholm, Sweden: Statens stlbyggnadskommitt. ISBN 91-7332-126-5 Volvo Corporate Standards (2004). STD105-0001 Critical Characteristics for Design Products. Gothenburg, Sweden: Volvo Richard H.A. (2001). In: CD-ROM Proceedings of ICF10. Honolulu, USA. Martinsson J. (2002). Fatigue Strength of welded cruciform joint with cold laps. Proc. Design and Analysis of Welded High Strength Steel Structures, pp. 163-185, in Stockholm June 2002. Stockholm, Sweden: EMAS
Appendix A
249
Appendix A
%Function which calculates stress intensity factors for analytical case function [a dK_eff c dK_I_c] = invandig_spricka(a_start,a_end,c_start) %=========== Constants =========== W = 1e99; %Sheet depth t = 10; %Sheet thickness t = t/2; C = 5e-12; n = 3; sigma = 104; a = a_start; c = c_start; delta_M = 10; fi_0 = 0; fi_90 = 90; fi_0 = (fi_0*pi)/180; fi_90 = (fi_90*pi)/180; i = 1; while a(i) <= a_end f_w(1,i) = sqrt(sec( ((pi*c(i))/(2*W)) * sqrt(a(i)/t) )); if a(i)/c(i) <= 1 Q(i,1) = 1 + 1.464*(a(i)/c(i))^1.65; M_1(i,1) = 1; f_fi_0(i,1) = ( (a(i)/c(i))^2*cos(fi_0)^2 + sin(fi_0)^2 )^(1/4); f_fi_90(i,1) = ( (a(i)/c(i))^2*cos(fi_90)^2 + sin(fi_90)^2 )^(1/4); else Q(i,1) = 1 + 1.464*(c(i)/a(i))^1.65; M_1(i,1) = sqrt(c(i)/a(i)); f_fi_0 = ( (c(i)/a(i))^2*sin(fi_0)^2 + cos(fi_0)^2 )^(1/4); f_fi_90 = ( (c(i)/a(i))^2*sin(fi_90)^2 + cos(fi_90)^2 )^(1/4); end M_2(i,1) = 0.05/( 0.11 + (a(i)/c(i))^(3/2) ); M_3(i,1) = 0.29/( 0.23 + (a(i)/c(i))^(3/2) ); g_0(i,1) = 1 - ( ( (a(i)/t)^4 * sqrt(2.6 - 2*(a(i)/t)) )*abs(cos(fi_0)) ) / (1 + 4*(a(i)/c(i)) ); %Angle in c-direction %Angle in a-direction %Transformation into radians %Transformation into radians %Constant in Paris' law %Constant in Paris' law %Tension [MPa]
250
g_90(i,1) = 1 - ( ( (a(i)/t)^4 * sqrt(2.6 - 2*(a(i)/t)) )*abs(cos(fi_90)) ) / (1 + 4*(a(i)/c(i)) ); F_a(i,1) = ( M_1(i) + M_2(i)*(a(i)/t)^2 + M_3(i)*(a(i)/t)^4 ) * g_90(i) * * f_fi_90(i) * f_w(i); F_c(i,1) = ( M_1(i) + M_2(i)*(a(i)/t)^2 + M_3(i)*(a(i)/t)^4 ) * g_0(i) f_fi_0(i) * f_w(i); dK_I_a(i,1) = sigma * sqrt((pi*a(i))/Q(i)) * F_a(i); dK_I_c(i,1) = sigma * sqrt((pi*a(i))/Q(i)) * F_c(i); a(i+1) = a(i) + C*dK_I_a(i)^n*delta_M; c(i+1) = c(i) + C*dK_I_c(i)^n*delta_M; i = i + 1; end dK_I_a = dK_I_a/sqrt(1000); dK_I_c = dK_I_c/sqrt(1000); a = a(1:end-1)'/1000; c = c(1:end-1)'/1000; dK_eff = dK_I_a;
Appendix B
251
252
Appendix B
253
254
Appendix B
255
256
Appendix B
257
258
Appendix B
259
260
Appendix B
261
262
Appendix B
263
264
Appendix B
265
266
Appendix B
267
268
Appendix B
269
270
Appendix C
271
272
Appendix C
273
274
Appendix C
275
276
Appendix C
277
278
Appendix C
279
280
Appendix C
281
282