Anda di halaman 1dari 7

Proceedings of the ASME 2010

International Mechanical Engineering Congress & Exposition IMECE2010


November 12-18, 2010, Vancouver, British Columbia, Canada

IMECE2010-38089
Tube Failure Analysis and Residual Life Assessment in Water Tube Boiler
Dinkar Nandwana

Srinivasan Periaswamy

Mechanic al Eng. Tech. Group

Mechanic al Eng. Tech. Group

Birla Institute of Technology and Science

Birla Institute of Technology and Science

Pilani, Rajasthan, India

Pilani, Rajasthan, India

ABSTRACT
This paper presents creep analysis to estimate the remaining
life of tubes used in water tube boiler. Three d imensional
fin ite element (FE) models were developed with co mputer
software namely ANSYS Workbench 10.0 to analy ze the
tube temperature and stress distribution. The FE results
showed that the temperature and stress increase as the oxide
scale thickness increases. Formulat ing the creep behavior in
terms of well established creep laws, the remain ing life of
the tube was estimated.
INTRODUCTION
The objective behind the successful operation of boiler is to
take heat energy in its available form (for examp le coal, oil
etc.) and to convert this energy into a more convenient
form. This may be done by heating water in a boiler and
then using the resulting hot water or steam fo r a desired
purpose. There are many types of boiler developed to meet
variety of duties and ever increasing demand. Broadly
boilers can be classified as shell type (fire tube), water tube
and electric boilers.
These boilers involve a high degree of technical risk. Their
deterioration could be induced by the flow of high
temperature fluids. The metallurgical failu re of the tubes in
water tube boiler is due to higher tube temperature than
expected in the original design. The temperature increases
slowly over many years due to internal o xide scale
deposition (long term overheating) or rapidly caused either
by loss of internal steam or water flow (short term
overheating) [1,2,3].But due to high temperature range and
fluctuation, there are no effective measures installed in the
powerplants. Predicting remaining service life can be one
such measure. The remaining life span can be predicted if
average temperature and stress are known. Thus, to prevent

tube failure, detection of tube temperature distribution is


necessary to take proper action. The paper demonstrates
thermal, stress and creep analysis to estimate the residual
life using the creep behavior formu lation in terms of creep
laws.
LITERATURE REVIEW
In water tube boilers, water is converted into steam inside
the tubes, while the hot gases pass around the outside of
tubes. Water tube boilers can operate at higher pressure
than fire tube boilers. The flow of steam and water within a
boiler is termed as circulat ion. This circu lation is crit ical in
preventing tubes fro m overheating. When tubes are
overheated, metal softens, weakens and eventually ruptures.
In simp le water tube circuit, bubbles of steam form in the
heated tubes or risers. The resulting steam and water
mixtu re is lighter than cold water on the unheated side of
the boiler, and rises to a steam dru m at the top of the boiler.
Here the bubbles rise to the surface and steam is released.
The water then flo ws fro m the dru m downwards through
the cooler tubes, or "downcomers", complet ing and
repeating the cycle. The water circu lation in a water tube
boiler is shown in Figure 1.
Ray et al. [4] conducted assessment of service exposed to
boiler tube of the super heater and reheater made of 2.25
Cr-1 Mo steels in a 120 MW boiler o f a thermal power
plant. . The results showed that there was degradation of
ultimate tensile strength (UTS), and stress rupture of the
boiler tubes due to increasing temperature and p rolonged
service exposure. But at the operating condition of
540C/ 40 MPa, all these service exposed tubes had a
remain ing life of more than 100,000 hours, provided that
there were no defects in the materials due to long term
service exposure.

In this paper, an analysis and computation technique is


developed for the evaluation of small axisymmet ric creep
deformation of boiler tubes. So me general methods are
discussed, and the finite element method is proposed for
best handling of calculation of creep deformations in a
general comp lex structural co mponents. Using these results
and life damage rule the remaining life o f boiler tube is
estimated.
CREEP CURVES AT CONSTANT STRESS

Figure 1: Water Circulation in water tube boiler

Husain and Habib [5] investigated the steel tubes failure in


a super heater boiler at one of the Kuwait Electrical and
Power p lants which suffered localized overheating. The
tube was made of low alloy steel, SA 213-T12 and it had
been in operation for 109,415 hours before failed. The
investigation indicated that the failure attributed to the
formation of thick scale of magnetite at the inner surface of
the tube wall. This phenomenon prevented the accessibility
of heat to the tube materials and consequently local,
prolong overheating took place, in wh ich the temperature
increased up to 700C in a frequent manner for long period
of time. The properties of the tube materials changed from
its original design values due to the effect of the localized
prolong overheating as a consequence of increasing of the
temperature.
Creep analysis of axisy mmetric bodies using finite elements
for calcu lating the creep strains was developed by
Greenbaum and Rubinstein [6]. The method involved
starting with the elastic solution of the problem and
calculating the creep strains for a s mall time increment.
These creep strains are treated as initial strains to determine
the new stress distribution at the end of the time increment.
Rah man and Suhakar [7] analyzed temperature variation of
tubes in water tube boiler using Finite Element Method.
The results showed that the temperature distribution at the
tube wall decreased due to increasing of mass flow rate of
steam, decreasing thermal conductivity of tube, steam in let
temperature, and increasing scale thickness.
Vishwanathan [8] has done substantial work in the area of
boiler tube life analysis since 1990s. He proposed the time
fractions model to estimate creep damage, for variab le
stress and temperature service conditions, using the results
of constant load, constant temperature. The ru le states that
the creep life has been expended when the sum of the life
fractions, or time fractions, equals unity.

It has been found fro m experiment that if a metal, which


creeps, is subjected to a constant uniaxial stress then the
accumulat ion of creep strains with time has the form as
illustrated Figure 2.

Figure 2: Creep curve using constant load

The variation of strain with time is described by the


following stages.
0A is an instantaneous deformation that occurs immed iately
upon application of the load and may contain both elastic
and plastic deformat ion.
AB is the primary stage in wh ich creep changes at a
decreasing rate as a result of strain hardening. The
deformation is mainly p lastic.
BC is the secondary steady state stage in which the
deformation is plastic. In this stage the creep rate reaches a
minimu m and remains constant as the effect of strain
hardening is counter balanced by an annealing influence.
Here the creep rate is a function of stress level and
temperature.
CD is the tertiary stage in which the creep continues to
increase and is also accompanied by a reduction in cross sectional area and the onset of necking, hence increase in
necking, thereby resulting in fracture.

LIFE FRACTION RULE

Where

Failure due to creep rupture is an important design


consideration under constant stresses and temperature
conditions. Expected service life can be established from
standard creep rupture data. However, since most members
are not subject to either constant stress or constant
temperature, creep-rupture damage criterion have evolved.
This criterion predicts time to rapture in such members
having mu lti-axial states of stress using time to rupture data
obtained from tension tests.

hTP : Heat Transfer coefficient for t wo phase flow.

Consider first the case of uniaxial stressing in which the


magnitude of the stress is varied during the test. If the stress
is changed after a certain amount of creep at a given
constant stress, a question arises as to what the continuing
strain-time path will be. The answer to th is question has not
been completely resolved, but the computation procedure
can be applied to any arbitrary law. Life fraction ru le has
been applied in this particular case.

Fr1 : Frode Nu mber.

The life-fraction ru le is based on the premise that the


expenditure of each indiv idual rupture fraction of the total
life at elevated temperature is independent of all other
fractions of the life to rupture, and that when the fractional
life used up at different stress levels and temperatures are
added up, it will equal unity. It can be simplified as
ti / t Ii =1

h1: Single Phase liquid heat transfer coefficient.


C0: Convection Number.
B0-Boiling Nu mber (q/ GL) where q is heat flu x, G is mass
flow rate, L is latent heat of vaporization.
C1-C6:-Constants.

Ff1 : Flu id dependent parameter.


Using the expressions and operating conditions, the
expression simplifies to

C0=0.522[(1-x)/ x]

(0.8)

Using the exp ressions and assuming the dryness fraction


(X) to vary linearly along the length (0 to 30m) fro m 0 to
0.1, the heat transfer coefficient averages to 55,000
2
W/m C.
B) Material Properties

Where t i is the time at stress or temperature i and t Ii is the


creep rupture time at temperature i.

All the material properties of 2 Cr-1 Mo steel, t wo phase


flu id, flue gases and Fe 3 O4 o xide scale film are tabulated in
Table 1.

MATHEMATI CAL MODELI NG


Steel

A) Heat Transfer Coefficient


The heat transfer coefficient of flue gas can be obtained
using Ditus Boelter expression for forced convection which
2
comes out to be 57 W/m C [9]. But the heat transfer
coefficient of the water steam two phase flo w (flow boiling)
cannot be directly obtained using the Dieters boiler
expression. Dr. S.G Kandlikar [10] developed a simp le
correlation fo r predict ing saturated flow boiling heat
transfer coefficient inside vertical tubes.
This correlat ion is based on the model utilizing the
contributions due to nucleate boiling and convective
mechanis ms. The correlation proposed is as follo ws
(1)
(2)

Temperature (C)
Convection
Coefficient
2
(W/m C)
Thermal
Conductivity(W/m
K)
Density (Kg/ m3)
Mass Flow rate
(Kg/s)

Flue
Gases

Two
phase
flui d
350
55000

60

427

8700

520
5.164

Oxi de
Fil m

800
57

0.527

Tabl e 1: M at eri al P roperti es

C) Scale Growth Rate


Due to the presence of impurit ies and reaction of water with
iron forming Fe 3 O4 , a film gets deposited in the inner layer
of tubes. Existence of scale reduces heat transfer rate fro m
flue gas to steam in the boiler tube because thermal

conductivity of scale is lo wer than the tube material.


Average rate of the scale formation on the inner surface and
-6
tube thinning was estimated as 4.0372 x 10 mm/hr by
Purbolaksono et al [11].
D) Finite Element Formulation
The method of analysis used here is known as the d irect
stiffness or displacement method. It can best be described
as a variational procedure. In classical elasticity one of the
most widely used variational princip les is the Theorem of
Minimu m Potential Energy [12]. Th is theorem states that of
all d isplacement functions which satisfy the boundary
conditions, the one which satisfies the equilibriu m
equations makes the potential energy an absolute min imu m.
The direct stiffness method is a systematic procedure for
the application of this theorem.
Fro m continuum p roblems, the body is appro ximated by a
set of simple sub-regions, called finite elements, which are
usually taken as geometrically simp le as possible. Within
each element the displacements are assumed to be linear
combinations of functions with undetermined coefficients,
and are chosen so that continuity is preserved along the
edges of adjacent elements. The assumed displacement
functions for any element are related to the displacement at
some particular points of the element. These points are
called the nodal points of the element, and are usually taken
as the corner points. The internal strain energy for each
element of the body is then expressed in terms of the nodal
point displacements. The potential energy for the co mplete
body is determined by summing the strain energy for all the
elements and subtracting the work done by the external
loads.
Finally, minimization of the potential energy with respect to
the nodal displacements yields the desired sys tem of
equations
for determin ing
the unknown nodal
displacements. For programming and computational
reasons, the method of constructing the governing algebraic
equations is to consider each element separately. The
potential energy for each element is minimized. Th is gives
a system of equations in terms of the nodal displacements
for that element and the applied forces acting on the
element. The coefficient matrix for these equations is called
the element stiffness matrix. The governing algebraic
equations for the body are obtained by superposition of the
element stiffness matrices, subject to the condition that the
displacement at any given node must be the same fo r all
elements attached to that node. The resulting equations are
modified to account for displacement boundary conditions,
and are solved to yield the unknown nodal point

displacements. The element strains and stresses are then


calculated fro m the known displacements.
The inclusion of creep behavior in a fin ite element
approach is best handled by using an incremental approach
and treating the creep strains as initial strains. The solution
of the problem begins by obtaining the elastic solution.
Based on these stresses, the creep strains for a small time
interval are co mputed. The results obtained are taken as
initial strains for the next time interval and are included as
fictit ious creep forces at the nodal points in the evaluation
of the nodal displacements and element stresses and strains
for the next time step. The process is repeated for each
increment of t ime. The basic assumption used in this
approach is that the change in stress during any time
increment is small co mpared to the stress at the beginning
of that increment. The erro r introduced by this type of
procedure can be made as small as one wishes by reducing
the time increment.
RESULTS AND DISCUSSIONS
A) Temperature Distribution at Different Scale
Thickne ss
Existence of scale reduces heat transfer fro m flue gas to
steam in the boiler tube because thermal conductivity of
scale is lower than the tube material. As the scale thickness
increases, the temperature of steam decreases. As a
consequence, energy transferred fro m steam decreases. In
order to make equilibriu m condition, temperature of flue
gas has to be increased. However, increasing the
temperature of flue gas needs time and combustion of
additional fuel. The disadvantage of increasing the
temperature o f the flue gas is decreasing rupture strength of
material. Therefore, existence of scale must be eliminated.
Figure 3 shows the temperature distribution for a tube with
no scale. The temperature varies fro m 350C inside to
366C outside. Figure 4 shows the temperature distribution
of the boiler tube for 2 mm of scale thickness, in which the
minimu m temperature is found to be 452C at the inner
tube (scale region), and maximu m is 502C at the outer
surface of the tube.
The scale thickness was varied fro m 0 mm to 7 mm and the
average temperature of the tube was plotted. Figure 5
shows the temperature plot of the boiler tube for different
scale thickness. This plot clearly implies that the existence
of scale is significant to the temperature d istribution in the
boiler tube. For examp le, temperature of the tube for scale
thickness of 1.5 mm is found to be 435C while for tube

without scale is 350C, hence there is 85C increment due


to an increase in scale thickness by 1.5 mm.

B) Creep stre ss and Creep strain


The stresses and creep strains for each of the discrete
element of the tube are displayed in the output. Figure 6
shows the stress distributions for no scale and Figure 7
shows the same for 0.5mm th ick scale. Average stress and
strain in the tube are plotted as the scale thickness is varied
fro m 0 mm to 7 mm. Figure 8 shows the plot of stress
against the scale thickness . As the scale thickness increases
with t ime, the stress increases from 121.6 MPa to 825 MPa.
Figure 9 shows the strain variation with time as the scale
grows. It can be seen that the strain gradually increases with
time.

Figure 3: Temperature variation for no scale

Figure 6: Stress plot for no scale (all values in 1e8 Pa)


.

Figure 4: Temperature Variation for 2mm scale

Figure 7: Stress plot for 0.5 mm scale (all values in 1e8 Pa)
Figure 5: Plot of temperature with oxide thickness

Figure 8: Stress variation with scale thickness

Figure 10: Life Damage Rule

2) Life Damage Rule


Using the life damage rule as described earlier, the
remain ing life can be predicted. The rupture time fo r the
tube for different tube temperatures can be obtained using
Larson Miller parameter [8]. It can be obtained using the
relation
LMP= T (20 + log t)
Figure 9: Strain variation with time (in yrs)

where LMP is Larson miller parameter, T is temperature of


tube in Rankine and t is the rupture time in hours.

Two approaches have been discussed here to obtain the


residual life: one using the creep curve and the other using
the life damage rule.

For this case the LM P is 3700.Thus the rupture time can be


obtained. Table 2 shows the rupture time corresponding to
different scale thickness, the ratio of service life to rupture
time and the cumulat ive sum. Since the cu mulative sum
reaches 1 after 26 years, the life co mes out to be 26 years
using this approach.

1) Creep Curve

CONCLUSIONS AND RECOMMENDATIONS

Based on the stresses, the creep strains for a small time


interval are co mputed. The results obtained are taken as
initial strains for the next time interval and are included as
fictit ious creep forces at the nodal points in the evaluation
of the nodal displacements and element stresses and strains
for the next time step. The process is repeated for each
increment of time and the creep curve is plotted (Figure
10). It fairly matches with exact curve as shown in Figure 2.
Tertiary stage is marked as the onset of boundary voids and
micro cracks, thus the failure of tube. On comparing the
two figures it can be concluded that the tertiary stage for
this particular tube reaches after 26 years. Thus the life of
the tube is approximately 26 years.

A) Conclusions

C) RESI DUAL LIFE ASS ESSMENT

In this paper, Finite Element solution have been presented


for obtaining the creep behavior of axisymmetric boiler
tubes which are subjected to axisymmet ric pressure and
temperature loading. The temperature variation was plotted
with respect to different scale thickness and the results
show that as the scale growth progresses with time, the
temperature keeps on increasing. The same is the case with
creep stress and strain.
Finally the residual life was estimated using two
approaches- using creep curve and life damage rule. Both
the approaches gave the same result. The life of the tube
came out to be 26 years. Th is result correlates with the

actual powerplant where service life varies fro m 25-35


years. Thus, the results are fairly accurate.

4.

During the course of this investigation, two major areas


were encountered for which further work is needed. The
first area relates to the creep behavior and life analysis of
cracked boiler tubes. In our analysis we considered very
simp lified assumptions that the surface of the tube was
clean and no crack in itiated or pitting formed on the
surface. If we consider actual boiler tubes, soot deposit,
scale formation and cracks are co mmon ly encountered on
the surface of tubes. Transversal and longitudinal through
crack or cracks to some depth are practically possible.
Thus, the analysis of cracked tubes should be the next area
of investigation.
The second major area of further study may be the cases of
surface pitted and corroded boiler tubes. While operation,
boiler tubes are exposed to abrasion and corrosion by the
particles in the flue gas and steam and/or water
respectively. Therefore, somehow similar to the crack
problems, the already developed finite element method may
be further developed to study the creep behavior of such
eroded or corroded tubes.
REFERENCES

2.

Port R. D. and Herro H. M., The NALC O Guide to


Boiler Failure Analysis, Nalco Chemical
Co mpany, McGraw-Hill, Inc, 1991.
Bailey R. W., Design Aspect of Creep,
Transaction of the ASME Journal of Applied
Mechanics, Vol. 1.

3.

Farr J. R. and Jawad M. H., Guidebook for the


Design of ASME Section VIII Pressure Vessels,
ASME p ress, 1998.

Scale T hickn ess


( in mm)
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5

Service Ti me
( in years)
0
2.60
5.21
7.81
10. 42
13. 03
15. 63
18. 24
20. 84
23. 45
26. 06

B.

Engineering Failure Analysis, 10: 645-654, 2003.

B) Recommendations for future research

1.

Ray, A.K., Sahay, S.K. & Goswami,


Assessment of service exposed boiler tubes,

5.

Husain, A. & Hab ib, K. Investigation of tubing


failure of super-heater boiler fro m Kuwait
Desalination Electrical Power Plant, Desalination,
183: 203-208, 2005.

6.

Greenbaum G. A. and Rubinstein M. F., Creep


Analysis of Axisymmetric Bodies Using Finite
Elements,
Nuclear
Engineering
and
Design, 1968.

7.

Rah man, Sukahar. Tube Temperature Distribution


in water tube boiler- a parametric study by finite
element method. ICCBT 2008 - 170 F - (14)
pp157-170.

8.

Vishwanathan R. Damage Mechanisms and Li fe


Assessment o f High temperature components.ASM
International (1989).

9.

Incropera, F.P., DeWitt, D.P.: Introduction to Heat


Transfer, 3rd edition. W iley, New York (1996).

10. S.G. Kandlikar. A General Correlation for


saturated two-phase flow boiling heat transfer
inside horizontal and vertical tubes.1983.
11. Purbolaksono et al. Failure Investigation on
Reheater Tube Due to Deposit and Wall Thinning.
Journal of Failure Analysis and Prevention (2009)
9:365 369.
12. Zienkiewicz O. C. Finite Element Method, 3ed.
McGraw-HILL Book Co mpany Ltd., 1977.

Rupture Ti me
(in years)
4.57E+12
4.31E+10
1.89E+09
1.66E+08
2.41E+07
4.98E+06
1.36E+06
4.54E+05
1.77E+05
8.46E+04
3.85E+04

Service
Tim e/Rup tu re Tim e
0
5.22E-07
1.18E-05
1.35E-04
9.31E-04
4.51E-03
1.64E-02
4.95E-02
1.26E-01
2.66E-01
5.83E-01

Cu mu la t iv e s u m of
the rat io
0
5.22E-07
1.23E-05
1.48E-04
1.07E-03
5.59E-03
2.20E-02
7.16E-02
1.98E-01
4.64E-01
1.048

Table 2: Life Damage Rule

Anda mungkin juga menyukai