You are on page 1of 27

Available online at www.sciencedirect.

com

Engineering Failure Analysis 16 (2009) 85111


www.elsevier.com/locate/engfailanal

A nite element program for on-line life assessment


of critical plant components
M.K. Samal *, B.K. Dutta, S. Guin, H.S. Kushwaha
Reactor Safety Division, Bhabha Atomic Research Centre, Mumbai 400085, India
Received 21 May 2007; accepted 26 January 2008
Available online 8 February 2008

Abstract
An on-line structural safety evaluation system (BOSSES) has been developed over the years at Bhabha Atomic
Research Centre (BARC), India, which is capable of assessing the damage due to creep and fatigue in critical plant components. The system acquires plant transients in real time and processes them to evaluate the stress, temperature and damage distribution in the components and provides information through a visual module. The aim of the paper is to briey
describe the details of the damage assessment methodology adopted by BOSSES with some case studies of real life plant
components.
2008 Elsevier Ltd. All rights reserved.
Keywords: On-line monitoring; Creep and fatigue; Erosioncorrosion; Critical plant components

1. Introduction
The issue of remaining life prediction has attracted considerable attention in the power generation industry
[111]. Integrity of important components is essential for operational safety, reliability and economic plant
operation. In the design, the life of a plant is estimated based on assumed process transients. However, because
of conservatism in the design the actual lives of these plants are expected to be much more than the estimated
value. The sustained interest in the area of remaining life prediction arises from the need to avoid costly outages, safety considerations and the necessity to extend the component operation life beyond the original design
life [2]. Many of the structural components used in fossil power plant, nuclear industry and chemical process
plants, are subjected to cyclic stresses due to the uctuation of process transients. On the other hand components like steam pipes, superheater headers, turbine rotors, casings, etc., operate at elevated temperatures
[1,2,6,7]. Cyclic stresses with a high mean stress at an elevated temperature lead to a damage mechanism
due to combined creep and fatigue. It is worth to note that among the various aging eects, fatigue, creep,
*

Corresponding author. Tel.: +91 22 25591523; fax: +91 22 25505151.


E-mail addresses: mksamal@barc.gov.in, mksamal@yahoo.com (M.K. Samal).

1350-6307/$ - see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2008.01.007

86

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

creepfatigue interaction and creepfatigue crack growth are commonly responsible for most of the failures in
various industrial components.
Several researchers have stressed the need to develop a life prediction methodology to address the various
aspects of failure mechanisms. Sakurai et al. [6] have presented the life assessment procedure of high temperature components of thermal power plants in Japan. Maekawa et al. [9] and Sakai et al. [12] have discussed the
operating data monitoring and fatigue evaluation system for Japanese Nuclear Power Plants (NPPs). The
development and implementation details of the French transient monitoring Fatiguemeter SYSFAC has been
described by Aufort et al. [5], Morilhat et al. [10] and Balley et al. [13]. FatiguePro, a fatigue monitoring system
developed by EPRI, USA [14], has the additional features of on-line fatigue crack growth monitoring. Bartonicek et al. [15] and Jovanovic and Verelst [16] have also discussed various issues related to fatigue and creep
monitoring of power plant components. An on-line creepfatigue monitoring system (BOSSES) has been
developed over the years in Bhabha Atomic Research Centre, Mumbai, India. The system has many advantages over similar existing systems. The system is being currently used to monitor many critical nozzles and
components of heavy water plants and thermal power plants in India. A brief description of the theoretical
background of damage assessment methodology used in BOSSES is provided in Section 2. Section 3 describes
some typical features of the program. Some case studies of components from heavy water plants and thermal
power plants are provided in Sections 4 and 5, respectively. In addition to the on-line deterministic damage
assessment module, BOSSES also has a probabilistic module for calculating the on-line probability of failure
of components due to various mechanisms such as creep, fatigue and erosioncorrosion, etc., and it is
described in Section 6.
2. Theoretical background
In recent years considerable progress has been made in research on the behaviour of defect-free structures
and structures containing defects at elevated temperature. Developments have been made both in structural
analysis techniques and in methods for collecting relevant material data. For the assessment of defects in plant
that operates in the creep range under cyclic loadings, creepfatigue interaction is the failure mechanism.
Hence a prediction method for the accumulation of damage due to creep and fatigue throughout their lifetime
is important for remaining life assessment of components. The early approaches to high temperature life
assessment describe methodologies that were based on defect-free assessment codes, i.e., ASME code case
N-47 (Now, it is known as Section-III, Subsection-NH) [17] and French RCC-MR [18], which have many similarities and are based on life time assessment of an un-cracked structures. The fracture mechanics based procedures basically followed the low temperature assessment procedures such as R6 [19].
ASME code case N-47 was the rst design code to formally embrace the concept of linear damage summation as a method of predicting material failure at high temperature, consisting of a fatigue (Miners cycle summation) component and a creep (Robinsons creep time summation) component. The French code RCC-MR
incorporates many of the concepts of ASME N-47, but with modied stress analysis procedures. The British
R5 procedure [20] augments and replaces certain sections of RCC-MR and ASME N-47. But, it provides
design and construction rules for mechanical components operating at high temperature. One advantage of
R5 approach is that a notional defect may be postulated and subsequent behaviour predicted for likely service
cycles. Together with the low temperature defect assessment procedure R6, it may be used in any of the three
ways, i.e., safe life (no degradation of property), damage tolerant (life extension) and fail safe (leak before
break) of design. Recently API-579 code [2124] has provided a detailed methodology for predicting damage
due to creepfatigue interaction in components. The dierent techniques for evaluation of creep and fatigue
damage in components with and without defects are described in the following sections.
2.1. Evaluation of damage in defect-free components
According to the conventional method, when any component is subjected to cyclic loading under high temperature and pressure environment, failure is assumed to occur when the summation of damage fractions due
to fatigue and creep becomes unity. In this method, one calculates separately the damage due to the fatigue
and creep and linear damage summation rule is used. Miners cycle summation is used to calculate fatigue

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

87

damage and Robinson time summation is used for creep [25]. The method is outlined in Fig. 1. Let
r1, r2, . . . , rn, are the maximum applied stress intensities for n1, n2, . . . , nn number of cycles. S1, S2, . . . , Sn are
the corresponding stress amplitudes. Corresponding to these stress amplitudes, N1, N2, . . . , Nn are the allowable cycles calculated from the SN diagram (Fig. 2). The fractional damage due to fatigue is calculated as
n
X
ni
:
1
Df
Ni
i1
Similarly, rm1, rm2, . . . , rmn are average values of equivalent stresses applied for hold times t1, t2, . . . , tn, respectively. Corresponding to these stresses, T1, T2, . . . , Tn are the creep rupture lives calculated from the creep rupture time curve of the material for the given operating temperature (Fig. 3). The fractional damage due to
creep is calculated as
n
X
ti
:
2
Dc
T
i
i1
The creep rupture life of the material is expressed by the equation: T = Arv, where A and v are material
constants.
2.1.1. R5 method
Here, linear damage summation rule is used. The component will fail if
n
X
Df Dc  6 Dallowable :
Dtotal
k1

Assessment of damage due to creep and fatigue interaction


(for defect free components)

Assessment of damage due to fatigue

Assessment of damage due to creep

For defect free structures, evaluate fatigue


damage (S-N curve)
(ASME III-NH/ RCC-MR/ R5/ API-579)

For defect free structures, evaluate creep


damage (creep-rupture curve)
(ASME III-NH/ RCC-MR/ R5/ API-579)

Combine damage due to creep and fatigue


(ASME III-NH/ RCC-MR/ R5/ API-579)

Fig. 1. Flowchart for assessment of damage due to creepfatigue interaction in defect-free components.

(Si, Ni)

N
Fig. 2. Schematic SN curve of a material.

88

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

temperature = constant

( mi ,Ti )
m

Fig. 3. Schematic stress rupture curve (due to creep) of a material.

2.1.2. API method


Here, the following rule is used:
n h
i
X
f
f
Df Dc 6 Dallowable ;
Dtotal

k1

where f = 0.3 for carbon and low alloy steels and f = 0.5 for austenitic and high alloy steels.
2.2. Evaluation of damage in components with defect (fracture mechanics based methods)
Creepfatigue interaction tends to be a low cycle phenomenon. Once a signicant dwell time is included, it
takes a long tome to accomplish the 105 plus cycles needed to get into the high cycle regime. Consequently, low
cycle fatigue is commonly dominated by crack propagation, with crack initiation taking place early in the life.
The owchart for the assessment of damage by fracture mechanics based method for components with defects
is shown in Fig. 4. From the assumption that crack growth governs creep/fatigue damage accumulation, several research workers have developed theories based on tracing the process of a crack, growing by the twin
mechanics of creep during dwell periods and by a Paris type law based on fracture mechanics during rapid
cycling. This approach assumes that damage is represented by the crack depth a. The cyclic crack growth
is then added to the creep crack growth during the dwell period (thold) to calculate the total growth per transient as [26,27]:



da 
da 
da 

 thold :
5
dN 
dN 
dN 
cycle

fatigue

creep

2.2.1. Crack growth rate due to fatigue


(a) R5 method: If conditions during the transient are in the linear static range, cyclic crack growth would be
calculated using Pariss law:

da 
CDK eff f ;
6
dN fatigue
where DKe is an eective stress intensity range adjusted for crack closure
DK eff qDK:
R5 gives the following empirical expressions for q:
8
for R < 2:5;
>
< 0:5
q 1 R=5 2:5 6 R < 0;
>
:
1
R P 0;

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

89

Assessment of damage due to creep and fatigue interaction


(Fracture mechanics based method for components with defect)

Assessment of damage due to fatigue


crack growth

Assessment of damage due to creep crack


growth

Evaluate stress intensity factor range


(RCC-MR A16/ API-579)

Evaluate reference stress and stress


intensity factor
(RCC-MR A16/ API-579)

Use modified Paris law to evaluate crack


growth per cycle
(RCC-MR A16/ R5/ API-579)

Evaluate C* or C(t) integral


(RCC-MR A16/ R5/ API-579)

Use proper creep crack growth law to


evaluate crack growth for given holding
periods in a cycle
(RCC-MR A16/ R5/ API-579)

Combine crack growth due to creep and


fatigue in each cycle
(RCC-MR A16/ R5/ API-579)
Fig. 4. Flowchart for assessment of damage due to creepfatigue interaction in components with defect.

where R is the ratio of minimum and maximum values of stress intensity factors, i.e., Kmin/Kmax. If the strain
ranges falls in elasto-plastic range, DKe then is replaced by DJe (J being Rices path independent integral
used as a loading parameter in elasticplastic fracture mechanics).
(b) API-579 method:


n
da 
Eamb
C DK
;
ET
dN fatigue

where Eamb and ET are Youngs moduli at ambient temperature and working temperatures, respectively.
2.2.2. Crack growth rate due to creep
(a) Use of C* integral for steady state condition: It is based on the assumption that a cracked body is subjected to a steady loading at elevated temperature and the load has been applied for suciently long time for
steady state creep to engulf the remaining ligament. The stress and strain rates of the component are related by
creep constitutive equation, which is analogous to the relationship between plastic stain and stress in the subcreep temperature regime. In the analogy, strain is replaced by strain rate and the stress coecient is replaced
by the creep stress coecient and the plastic strain hardening constant is replaced by creep strain rate hardening constant. The integral C* is dened analogous to Rices J-integral as
 
Z
Z e_ ij
ou_ i



C
ds where W
W dy  T i
rij deij ;
10
ox
C
0
where C is the line contour, W* is the strain energy density associated with the point stress and the strain rate
i
is the gradient of velocity components in x-direction, ds is the incremental
e_ ij . Ti is the traction in i-direction, ou
ox
*
length of contour. The C integral can be also explained in terms of energy release rate (per unit time).
(b) Use of C(t) integral for transient condition: The crack tip stress elds under small scale creep can be
characterized by a time dependant C(t) integral, whose value is determined along a contour taken very close
to the crack tip, i.e.,

90

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

Ct

W dy  T i

C!0


ou_ i
ds:
ox

11

It may be noted that C(t) is the same as C*, but its value is determined close to crack tip within a region where
the creep strain dominates over the elastic strains. C* is only valid as a measure of creep crack load once a
stationary state has been reached. In the transient phase of stress redistribution leading to the steady state,
C(t) is more appropriate criterion. Since C(t) is generally larger than C*, ignoring the transient phase can
underestimate the severity of creep induced crack growth. Transient conditions are usually important under
repeated loading conditions, where new high stresses are reestablished at the beginning of each cycle. A criterion to know whether transient conditions need to be considered is to compare the cycle period, tcyc, with
the Reidel transition time [28], ttr. If tcyc > ttr, steady state dominates and C* is acceptable and if tcyc < ttr,
the transient state dominates and C(t) is the preferred measure.
(c) R5 method: In this method, the reference stress is calculated as, rref = ryP/PL where is the yield stress of
the material, P is the applied load and PL is the limit load at which stress in the net section at the cracked
location reaches yield stress. Under dierent combination of loads and moments, the reference stress can
be calculated by nite element modeling of the component or using Handbook solutions [29]. The reference
strain rate can be evaluated using the creep constitutive equation:
e_ c C 1 rn1 ;
where C1 and n1 are the creep strain rate coecient and hardening exponents, respectively. The C* integral can
be represented as
C  rref e_ ref R0 ;
12

2
eff
and e_ ref is the reference strain rate evaluated using. The crack growth rate due to creep can
where R0 DK
rref
be expressed in terms of C* integral as
da
D  C / ;
dt

13

where D and / are material constants. To know that steady sate creep condition has been achieved, the stress
K2
redistribution time tred is evaluated using the equation, tred EC
 , where K is the stress intensity factor and E is
Youngs modulus. In case of transient creep condition, the C(t) integral is evaluated using Reidels approximation, i.e., and C(t) integral is used in the creep crack growth equation instead of C*.
(d) API-579 method: This method is also based on calculation of reference stress and use of C* and C(t)
integral. The only dierence is that these are to be evaluated using the specied expressions given in the document and the stress intensity factor and reference stresses for dierent geometries have to be calculated using
the appendices of API-579 (e.g. Appendix C for stress intensity factor K, Appendix D for reference stresses
and Appendix F for creep and fatigue material properties). The time tred is calculated as
tred ai ; ci

0:91Kai ; ci 2
;
nBN 1  EY  C  ai ; ci

14

where ai and ci correspond to crack depth and crack width for a semi-elliptical surface crack, EY is Youngs
modulus at mean temperature of the cycle, K is the stress intensity factor and nBN is the BaileyNorton coefcient evaluated at the reference stress in the current load increment. C* and C(t) are evaluated as


3
2



 nnBN 3
2
1
BN
_
Kai ; ci 
eref
7
6 tred ai ; ci
C  ai ; ci
and Ctai ; ci C  ai ; ci 4
15; 15
rref
ti
1  Dbc  Diac
where ti is the cycle time, Dbc is the local creep damage before initiation of crack and this is computed using the
net section stress considering the pre-crack loading history. Dac is the local creep damage after initiation of
crack and this is computed using the reference stress (Appendix D of API-579) considering the post-crack
D  Ctl where D and l
loading history. The creep crack growth rate is computed using the equation da
dt
are creep crack growth rate constants for the material.

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

91

3. BARC on-line structural safety evaluation system (BOSSES)


The various modules of the BARC on-line structural safety evaluation system (BOSSES) are explained
through a owchart in Fig. 5. The transient acquisition module collects the plant transients, e.g. pressure, temperature and ow rate, at certain time intervals through a data acquisition system interfaced to a PC. In online fatiguecreep monitoring, the conversion of plant transients to the temperature/stress responses in the
structure is one of the most important tasks. The technique generally used is Greens function technique
(GFT). This is a method, which transforms the plant transients to temperature/stress responses using a predetermined transfer function. The primary advantage of this method is the less computation time. However,
the GFT has a number of limitations. Strictly the GFT cannot be applied to the nonlinear analysis because of
the inherent assumptions of linear superposition. The GFT provides information on some predetermined
points of the structure. If the number of points is increased, the computation time also increases. Various
investigators, e.g. Chen and Kuo [30], Mukhopadhyay et al. [8,11,31] and Maekawa et al. [9], have discussed
these aspects in detail. Sakai et al. [12] have outlined an approximate strategy to account for the eect of
change of convective heat transfer coecient using GFT.
To overcome the limitations of GFT and also to utilize the computing power of modern computers, the
present system BOSSES employs on-line nite element technique to compute temperature transients and the
thermal stresses in the structure due to uctuation of the process transients. The system can take care of
the variations of heat transfer coecient and can provide whole-eld information. The damage evaluation
module nds out the fatigue usage factor, creep damage and fatiguecreep interaction eect. Various algorithms are available for cycle counting in an irregular stress history. The most widely used is the rain-ow cycle
counting technique. The stress time history is converted to stress frequency spectrum using the rain-ow cycle
algorithm as introduced by Socie [32]. Later various investigators, e.g. Downing and Socie [33], Rychlik [34]
and Glinka and Kam [35], have presented various algorithms of rain-ow cycle counting technique. In this

Fig. 5. Flowchart of various modules of BOSSES.

92

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

method the eect of small stress excursions are separated and the relevant dominant cycles capable of producing damage are computed. The cycle counting algorithm of the present system is based on this method. The
fatigue usage factor is evaluated from the computed cycles using material fatigue data. The creep damage
index is evaluated from the computed temperature and stress histories and the material creep curve, using
the procedure shown in API reports [2124]. To account for the combined damage mechanism, the damage
accumulation approach is adapted [26,27]. Availability of an users friendly menu driven graphical display
module for plant transients, computed temperature/stress responses, damage related information and remaining life assessment are another features of the present system. There is a provision for ecient data storage,
data backup and data restoration. The system has been integrated to a data acquisition system and hence the
fatiguecreep degradation of several components of a plant can be monitored using a single PC.
4. Monitoring of critical nozzles of heavy water plants
There are dierent types of heavy water plants in India depending upon the type of process used. The heavy
water plant Tuticorin is based on a mono-thermal process whereas the heavy water plant Kota is based on a
bi-thermal process of heavy water production. In the mono-thermal process, the feed gas is taken usually from
a fertilizer plant in the form of synthesis gas mixture (73% H2 and 24.4% N2, rest other gases). The gas is
passed through a drier and purication column to remove impurities. The cooled gas is then introduced into
the exchange column. Subsequently further processing produces heavy water. In this type of plant, the drier
unit is subjected to regular thermal cycling.
There are usually two drier vessels, which operate in tandem. The ow diagram of gases from and to the
two drier vessels along with the sensor signals for BOSSES is shown in Fig. 6a. The drier vessel performs four
dierent functions, i.e., service, regeneration, cooling and standby. The operation stage is typically 8 h of duration. The total time for regeneration, cooling and standby is also 8 h. During operation, the temperature of the
process gas entering into the vessel is around 0 C. In the regeneration, it is heated to around 185 C and it is
cooled down again to around 2 C during cooling operation. The drier is exposed to fatigue damage due to
this thermal cycling. As the maximum temperature (i.e., 185 C) is low, creep eect is not signicant. The

Fig. 6a. Line diagram of ow of process gas and associated instrumentation for the drier unit of a heavy water plant.

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

93

objective of the on-line monitoring system is to calculate the stress and thermal transients seen by the drier
nozzle and then to calculate the fatigue damage over the period of monitoring. This value of fatigue damage
can be extrapolated to know the life to which the component can be operated safely. The nite element (FE)
models of the top and bottom end of the drier vessel (with nozzles) used for computation of material temperature; stress and damage in BOSSES are shown in Figs. 6b and 6c. The typical plant transients of various
parameters as acquired by the data acquisition system are plotted in Fig. 7.
BOSSES computes the material temperature, stress and damage due to fatigue at all the Gauss points of the
FE models and stores the information for further use by the plant operator. A typical computed information
history for a shell-nozzle junction of the bottom end of drier vessel is shown in Fig. 8. Similarly, FE model of
some of the components of a bi-thermal process heavy water plant (i.e., heavy water plant Kota) are shown in
Figs. 9a and 9b. The computed stress (von Mises equivalent stress in MPa) and damage (i.e., usage factor due
to fatigue) contour at any instant of time (e.g., after 16 h of online data processing) is shown in Figs. 10 and
11, respectively. From the damage information, one can easily estimate the remaining life of a component
assuming that the design life of the component is over when the total damage fraction becomes unity. In
the following section, various information regarding stress and damage state of some components of a thermal
power plant as computed by BOSSES are provided.
5. Monitoring of critical components of thermal power plants
In a thermal power plant, high energy piping, boiler headers, turbine rotors, casings, steam chests, valves, etc.,
are the most critical components subjected to creep and fatigue damage under service conditions. To prevent any
unforeseen outages of the plant due to steam leakage preceded by crack formation leading to plant unavailability
and down-time cost, it is necessary to generate some information on localized degradation due to various damage
mechanisms in real-time basis. For this purpose, BOSSES has been implemented in two units of a coal-red
power plant (each unit generates 210 MW). The units are in service for approximately 10 years. The main steam
pressure and temperatures are approximately 140 kg/cm2 and 535 C, respectively. The corresponding pressure
and temperature in the hot reheat lines are 34 kg/cm2 and 535 C, respectively. Some of the components under
monitoring are superheater outlet header, reheater inlet and outlet header and the hot reheat pipe bends, etc. The
material of these headers and pipe bends is ASME SA 335 P22 (a low alloy ferritic steel).
5.1. Monitoring of creep and fatigue damage
The nite element models of some of the components monitored by BOSSES are shown in Figs. 12a12c.
The actual plant transients used by BOSSES are main steam pressure, temperature and ow and similar
parameter in other circuits such as hot and cold reheat lines. Some of the plant transients of the thermal power
plant (e.g. main steam pressure, temperature and ow) are shown in Figs. 1315, respectively. It may be noted
that the piping load is an important loading on a component apart from internal pressure and thermal stresses.
The piping loads on a component are evaluated by carrying out stress analysis of the whole piping loop
(with associated supports and restraints) using various types of straight and bend pipe nite elements due

Fig. 6b. FE model of shell-nozzle junction of the top end of a drier column.

94

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

Fig. 6c. FE model of shell-nozzle junction of the bottom end of a drier column.

Fig. 7. Actual service data recorded for the drier unit of a heavy water plant.

to various load such as dead weight, pressure and thermal loads. The layout of such a piping loop (where the
hot reheat piping bend is a part) is shown in Fig. 16a and the associated piping loads computed on both ends
of the pipe bend are shown in Fig. 16b.
With all the loads known, BOSSES evaluates the material temperature, stress and damage (due to creep and
fatigue) in all the components. The computed values of material temperature, maximum stress intensity and
damage history at a shell-nozzle junction of a superheater outlet header of the thermal power plant are shown
in Fig. 17. Similarly, one can obtain the whole-eld information on stress and damage in a component at any
instant of time and as an example, the stress and damage contours of the hot reheat pipe bend after 24,000 h of
on-line computation are shown in Figs. 18a and 18b, respectively. Calculation procedure of crack growth due

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

95

Fig. 8. Computed material temperature, stress intensity and accumulated damage for the shell-nozzle junction.

Fig. 9a. FE model of shell-nozzle junction of a waster stripper column of a heavy water plant.

to creep and fatigue are described in Section 2.2. In the following section, some details of creep and fatigue
crack growth for a hot reheat pipe bend is described.
5.2. On-line computation of crack growth for assessing tness for purpose of service of components
The geometrical details of the pipe bend are given as: internal radius = rj = 224 mm; external radius
re = 254 mm; average radius = rm = 239 mm; bend radius = Rc = 1500 mm; thickness = h = 30 mm; initial
crack size a0 = 0.5, 0.1 and 1.5 mm; c0/a0 = 3
X

Rm
;
h

hRc
;
r2m

z pr2m h:

Nomenclature for loading: pressure = P; bending moment = M2 = M (in-plane). The nomenclature for various
geometrical parameters of the pipe bend is provided in Fig. 19. The dierent postulated crack congurations

96

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

Fig. 9b. FE model of shell-nozzle junction of a hot tower column of a heavy water plant.

Fig. 10. Computed von Mises equivalent stress (in MPa) contour of a cylindrical shell-nozzle junction due to dierent loading conditions.

Fig. 11. Computed damage contour of a cylindrical shell-nozzle junction due to dierent loading conditions.

used in the analysis are shown in Fig. 20. The details of stress intensity factor and limit load calculation used by
BOSSES for the pipe bend are described in the following section. The variation of stress intensity factors (at
bottom and side end of the semi-elliptical crack) and limit load ratios (i.e., ratio of reference stress to the material
yield stress) for four cases (Cases 14 as shown in Figs. 20a20d) with initial crack size to thickness ration are
plotted in Figs. 2123, respectively. The stress intensity factors for dierent crack congurations were calculated

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

97

Fig. 12a. FE model of a superheater outlet header.

Fig. 12b. FE model of a reheater outlet header.

Fig. 12c. FE model of a hot reheat pipe bend.

using RCC-MR A16 procedure. The nominal stresses were evaluated using on-line FE analysis for the combined
internal pressure and in-plane bending moment acting on the elbow. The stress intensity factor is expressed as

 a
a2
a3
a4 p
pa;
16
r3 i3
r4 i4
K r0 i0 r1 i1
r2 i2
h
h
h
h
where a is crack depth and r0 and r1 are the derived stresses and calculated in the following way for dierent
crack congurations with nominal longitudinal membrane stress (r1m), nominal circumferential stress (r2m),
nominal longitudinal bending stress (r1b), nominal circumferential bending stress (r2b) and internal pressure
P. The stresses for various crack congurations of the pipe bend are given as:
For
For
For
For

circumferential crack (internal): r0 = r1m  r1b + P and r1 = 2r1b.


circumferential crack (external): r0 = r1m + r1b and r1 = 2r1b.
longitudinal crack (internal), r0 = r2m  r2b + P and r1 = 2r2b.
longitudinal crack (external), r0 = r2m + r2b and r1 = 2r1b.

The coecients i0 and i1 in Eq. (16) are obtained from the tables provided for straight pipes with corresponding crack congurations in A16. r2, r3 and r4 are the stresses corresponding to second- to fourth-order

98

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

Fig. 13. Recorded history of main steam pressure of a thermal power plant.

Fig. 14. Recorded history of main steam temperature of a thermal power plant.

nonlinear stress distributions across the thickness and are neglected in this analysis. The reference stress can be
evaluated using modied limit load option or elasto-plastic stress option of A16. Both the options are used
here and the maximum among the two limit loads is selected for further calculation. The crack growth computed by BOSSES for various crack congurations with initial crack size of 1.5 mm are shown in Fig. 24 where
on-line computation has been carried out for a period of 24,000 h of plant operation. The maximum crack
growth is for the longitudinal semi-elliptical crack at external surface of the intrados of the pipe bend. Once,

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

99

Fig. 15. Recorded history of main steam ow of a thermal power plant.

Fig. 16a. Piping layout of the reheat line.

the damage fraction and crack growth are known, BOSSES can estimate the life consumed for the component
for the entire service life using suitable extrapolation techniques. As an example, the life for the hot reheat pipe
bend under consideration has been extrapolated for a service life of around 12 years and is plotted in the form

100

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

Fig. 16b. Computed piping loads on a reheat pipe bend.

Fig. 17. Computed material stress intensity, temperature and damage history at a shell-nozzle junction of the superheater outlet header.

of a bar chart in Fig. 25. Life consumed by the pipe bend for both conventional and fracture mechanics based
methods are shown in the above gure. This information is available for all the critical components of a thermal power plant and serves as an important information base for scheduling and prioritizing maintenance of
various components.

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

101

Fig. 18a. Snapshot of contour of maximum stress intensity (MPa) in a hot reheat pipe bend at any time instant (e.g. after 24,000 h of
processing of online data).

Fig. 18b. Current damage contour in a hot reheat pipe bend at any time instant (e.g., after 24,000 h of processing of online data).

Fig. 19. Geometric and loading details of the a typical pipe bend.

6. On-line evaluation of failure probabilities due to creep, fatigue and erosioncorrosion


The crack growth due to creep and fatigue depends upon the stress intensity factor, C* integral, etc., of the
assumed cracked congurations and the methods to evaluate them are already described in Section 2.2. The
method to incorporate them in the calculation of various state probabilities of pipe bend will be described here.
On the other hand, the erosioncorrosion rate depends upon the process parameters such as uid velocity, pH,
temperature and oxygen content [36,37]. When water reacts with steel, an oxide lm is formed on the metal surface. When water is deoxygenated, this layer consists of magnetite (Fe3O4). According to experiments, four different ferrous ion complexes are formed upon dissolution of this magnetite, as written below [38]:

102

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

Fig. 20a. Pipe bend with circumferential part-through external crack at the intrados.

Fig. 20b. Pipe bend with circumferential part-through internal crack at the intrados.

Fig. 20c. Pipe bend with longitudinal part-through external crack at the intrados.
2b

Fe3 O4 32  bH 3FeOHb

4  3bH2 O; where b 0; 1; 2; 3:

The equilibrium constants Kb are calculated as K b FeOHb2b =H P H2 1=3 , where FeOHb2b ] is the
concentration of bth ferrous ion, [H+] is that of hydrogen ion and P H2 is the partial pressure of molecular
hydrogen gas. The corrosion rate depends upon two factors, i.e.,
(a) Oxide dissolution: This rate is governed by the Arrhenius relationship: Rk = R0 exp(Ek/RT), where
Ek = activation energy = 31,580 cal/mol, R0 = 9.55  1032 atoms/cm2 S, T = uid temp in K, R = universal
gas constant = 2 cal/mol/K.
(b) Mass transfer based on oxide dissolution: This transfer rate is given by: RMT = K(CS  CB), where
K
mass transfer coefficient 0:0791DO2 =dUd=nx n=DO2 0:335 , (d = inner diameter of pipe, U = ow
velocity, n = kinematic viscosity, x = 0.54 for fully turbulent ow)

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

Fig. 20d. Pipe bend with longitudinal part-through internal crack at the crown.

Fig. 21. Variation of stress intensity factor at bottom of the semi-elliptical crack with various a/h ratio.

Fig. 22. Variation of stress intensity factor at side tip of the semi-elliptical crack with a/h ratio.

D O2
CS
CB

0.5
0.6
oxygen diusivity = 7.4  108T (2.6  18)
P /(290) 2b
P
surface concentration of ferrous ions FeOHb
 K b H P H2 1=3
a given bulk concentration

103

104

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

Fig. 23. Variation of limit load ratio of the pipe bend (for dierent crack congurations) with a/h ratio.

Fig. 24. Computed crack growth (in mm) due to creepfatigue due to various crack congurations for the hot reheat pipe bend.

Equivalent erosioncorrosion rate can be dened as: Rate = [(Rk)1 + (RMT)1]1. It may be noted that Kb
and P H2 are functions of temperature. To calculate the occurrence rates of aw, leak and rupture in a component, Markov model [39] is used (Fig. 26). To nd out the erosioncorrosion rate, various process parameters (such as ow velocity, pH, temperature, etc.) are used. Figs. 27a27d show the variation of corrosion rate
with ow velocity; pipe inner diameter; uid pH and temperature, etc., for a typical feed water pipe. The four
states of the Markov model for stable (aw and leak) and unstable (rupture) crack propagation mechanisms as
shown in Fig. 26 are:

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

105

Fig. 25. Predicted life consumed (in %) by a typical hot reheat pipe bend using various methods.

R
Fig. 26. Flowchart of the Markov model.

(a) S = success (depth of corrosion less than 0.125t),


(b) F = aw (depth of corrosion 0.1250.45t),
(c) L = leak (depth of corrosion 0.450.8t),
(d) R = rupture (depth of corrosion beyond 0.8t),
where t is thickness of the component (e.g. a pipe). The various state transition rates used in the Markov
model (Fig. 26) are:
(i)
(ii)
(iii)
(iv)
(v)
(vi)

/ = aw occurrence rate;
k = leak failure rate;
qF = rupture failure rate given aw;
qL = rupture failure rate given leak;
x = aw repair rate;
l = leak repair rate.

106

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111


0.020

Corrosion Rate (mm/year)

0.018
0.016
0.014
0.012
0.010

OTHER PARAMETERS:
PIPE INNER DIAMETER = 7 cm
TEMPERATURE = 553 K
FLUID pH = 10.2

0.008
0.006
0.004
0.002
0.000
500

750

1000 1250 1500 1750 2000 2250 2500 2750 3000

Flow Velocity (cm/s)


Fig. 27a. Variation of corrosion rate with velocity of uid ow.
0.020

Corrosion Rate (mm/year)

0.018
0.016

OTHER PARAMETERS:
FLOW VELOCITY = 1500 cm/s
TEMPERATURE = 553 K
FLUID pH = 10.2

0.014
0.012
0.010
0.008
0.006
0.004
0.002
0.000
5

10

15

20

25

30

35

40

45

50

Diameter (cm)
Fig. 27b. Variation of corrosion rate with diameter of pipe.

Dierent limit state functions to calculate the state transition rates for a pipe are shown in Table 1.
Notations: rate = erosioncorrosion rate in mm/year; T = time of inspection (usually 10 years);
Pop = operating pressure; Pf = failure pressure (dened as in Shell-92 model [38]) which is given as
!
dT
1:8UTSt 1  t
Pf
;
17

D
1  dT
Mt


q
dT 0:45t rate  tLSF 3
L2
where M 1 0:805 Dt and
, D = pipe outer diameter; L = axial length
dT 0:8t rate  tLSF 4
of corrosion defect; t = pipe thickness; UTS = ultimate tensile strength of the piping material; YS = yield

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

107

Corrosion Rate (mm/year)

10

0.1

0.01

OTHER PARAMETERS:
FLOW VELOCITY = 1500 cm/s
PIPE INNER DIAMETER = 7 cm
FLUID pH = 10.2

1E-3

1E-4
340 360 380 400 420 440 460 480 500 520 540 560 580 600

Temperature (K)
Fig. 27c. Variation of corrosion rate with uid temperature.

0.10

Corrosion Rate (mm/year)

0.09
0.08
0.07

OTHER PARAMETERS:
FLOW VELOCITY = 1500 cm/s
TEMPERATURE = 553 K
PIPE INNER DIAMETER = 7 cm

0.06
0.05
0.04
0.03
0.02
0.01
0.00
9.5

9.6

9.7

9.8

9.9

10.0

10.1

10.2

10.3

10.4

10.5

pH
Fig. 27d. Variation of corrosion rate with pH of uid.

Table 1
Limit state function used by BOSSES for various date transition rates of a pipe
State transition rate

Limit state function

Flaw occurrence rate (/)


Leak failure rate (k)
Rupture failure rate given aw (qF)
Rupture failure rate given leak (qL)

LSF1
LSF2
LSF3
LSF4

(d, T) = 0.45t  (0.125t + rate t)


(d, T) = 0.8t  (0.45t + rate  t)
(Pf) = Pf  Pop
(Pf) = Pf  Pop (with a dierent Pf)

108

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

strength of the piping material; d(T) = depth of corrosion, x = PPFD/(TFI + TR); P = probability that the
piping element with aw will be inspected per inspection interval, PFD = probability that a aw will be detected, TFI = mean time between inspection of aw, TR = mean time to repair a aw or a leak,
l = PlkPLD/(TLI + TR); Plk = probability that the piping element with leak will be inspected per inspection
interval, PLD = probability that a leak will be detected, TLI = mean time between inspection of leak.
The state transition probabilities are calculated using rst-order reliability methods (FORM). The calculation procedure followed in Markov model is briey described here. Once the state transition rates are known,
one can set four dierential equations, one for each state, as follows:
dS
/S xF lL;
dt
qF F qL L:

dF
/S  kF qF xF ;
dt

dL
kF F  qL lL;
dt

dR
dt
18

In vector form these equations can be rewritten as


dX
AX;
dt
2

19
3

St
/
x
l
0
6 / kF qF x
7
6 F t 7
0
0
6
7. The solution of this linear matrix dier7
where Xt 6
4 Lt 5; A 4 0
qL l 0 5
kF
qL
0
0
qF
Rt
ential equation is given by: Xftg C 0 E0 ez0 t C 1 E1 ez1 t C 2 E2 ez2 t C 3 E3 ez3 t where the eigen-values are given
by: z = {z0, z1, z2, z3}, and the eigenvectors are given by: {E0, E1, E2, E3}. The constants Cj, j = 0, 1, 2, 3 are constants of integration and can be determined by the following boundary conditions (assuming that at t = 0 the
pipe segment is free of detectable aws):
2 3
0
607
6 7
Xt ! 1 6 7
20
405
1
and
2 3
1
607
6 7
Xt 0 6 7:
405

21

0
The eigen-values are determined by solving the fourth-order polynomial equation given by
detA  zI 0;

22

where I is the identity matrix. All the expressions of the corresponding eigen-values, eigenvectors and the integration constants are calculated. Fig. 28a shows the variation of dierent state probabilities due to erosioncorrosion damage in the feed water pipe with increasing plant operational time (using an oine calculation), using
the parameter values given in Tables 2 and 3, respectively. Fig. 28b shows the on-line calculation of various
state probabilities over time of 2 years due to creep and fatigue crack growths in the reheat pipe bend (Fig. 12c).
7. Discussion
The prevailing approach of the plant operator to estimate the need for inspection (i.e., inspection schedule)
is often on the basis of oine inspection and past operation and maintenance (O&M) experience. Such an
approach results in frequent inspection of problem-free equipments and many times critical shell-nozzle junction are overlooked due to lack of detailed information. The application of the present system helps in making
realistic decisions to schedule maintenance intervals on certain selected components and hence it can be used

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

109

State probabilities

0.1

SUCCESS
FLAW
LEAK
RUPTURE

0.01

1E-3

10

20

30

40

50

60

70

80

90

100

Plant operational time (years)


Fig. 28a. Oine calculation of various state probabilities for the erosioncorrosion model.

Table 2
Process parameters of the ow through a feed water pipe
Process parameter

Mean value

Standard deviation

Fluid ow velocity (cm/s)


Pipe inner diameter (cm)
Pipe thickness (mm)
Fluid temperature (K)
Fluid pH
Pipe outer diameter (cm)
Material ultimate strength (MPa)
Operating uid pressure (MPa)

1500
7
7
553
10.2
8.4
455
8.7

50
0.14
0.14
25
0.51
0.168
32
0.87

Table 3
Other plant operational parameter used in the Markov model
Other plant operational parameters

Value

Mean time between inspection of aws (year)


Mean time between inspection of leaks (year)
Mean time to repair aw or leak (h)
Pipe element aw inspection probability per inspection
Flaw detection probability
Leak detection probability
Pipe element leak inspection probability per inspection

10
1
200
0.25
0.9
0.9
0.9

to save lot of resources. The system can be merged with existing O&M planning and scheduling activities for
ecient plant management and thus provide a cost eective solution. The system can also be used to rank
various components of the plant based on the Risk associated with their failure (Risk is dened as the product of probability of failure and consequences of that failure mechanism). Such a risk-based monitoring
should be a part of the overall concept of risk-based life management of dierent plant components.

110

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111


1
success
flaw
leak
rupture

State probabilities

1E-4

1E-8

1E-12

1E-16

1E-20
0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

Time (year)
Fig. 28b. On-line calculation of various state probabilities for the creepfatigue crack growth model.

Acknowledgement
The authors acknowledge the cooperation extended by operation and maintenance personnel of various
heavy water and thermal power plants for successful implementation and operation of the above on-line structural safety evaluation system.
References
[1] Saxena A, Sherlock TP, Viswanathan R. Evaluation of remaining life of high temperature headers: a case history. In: Proceedings of
EPRI workshop on life extension and assessment of fossil plant; 1988.
[2] Liaw PK, Saxena A, Schaeer J. Estimating remaining life of elevated-temperature steam pipes Part II. Fracture mechanics
analyses. Eng Fract Mech 1989;32:70922.
[3] Thoraval G. Creep of high temperature steam piping: EDF experience with fossil red power plants from 1955 to 1987. Nucl Eng Des
1989;116:38998.
[4] Miksch M, Schucktanz G. Evaluation of fatigue of reactor components by on-line monitoring of transients. Nucl Eng Des
1990;119:23947.
[5] Aufort P, Bimont G, Chau TH, Fournier I, Morilhat P, Souchois T, et al. On line fatiguemeter: a large experiment in French nuclear
plants. Nucl Eng Des 1991;129:17784.
[6] Sakurai T, Sugai S, Aoki M. Life assessment of high-temperature components of thermal power plants in Japan. Nucl Eng Des
1993;139:3117.
[7] Liaw PK, Saxena A, Perrin J. Life extension technology for steam pipe systems Part I. Development of material properties: Part II.
Development of life prediction methodology. Eng Fract Mech 1993;45:75998.
[8] Mukhopadhyay NK, Dutta BK, Kushwaha HS, Mahajan SC, Kakodkar A. On line fatigue life monitoring methodology for power
plant components. Int J Pressure Vessel Piping 1994;60:297306.
[9] Maekawa O, Kanazawa Y, Takahashi Y, Tani M. Operating data monitoring and fatigue evaluation systems and nding for boiling
water reactors in Japan. Nucl Eng Des 1995;153:13543.
[10] Morilhat P, Chau TH, Savoldelli D, Mallez C. Intelligent system for transient data collection and fatigue monitoring of pressurised
water reactors nuclear steam supply system. Nucl Eng Des 1999;192:1039.
[11] Mukhopadhyay NK, Dutta BK, Kushwaha HS, Mahajan SC, Kakodkar A. Implementation of nite element based fatigue
monitoring system at Heavy Water Plant Kota. Nucl Eng Des 1999;187:15363.
[12] Sakai K, Hojo K, Kato A, Umehara R. On-line fatigue-monitoring system for nuclear power plant. Nucl Eng Des 1994;153:1925.
[13] Balley J, Bertagnolio D, Faidy C, Kappler K, Kergoat M, LHuby Y, et al. New developments in French transient monitoring system:
SYSFAC from the experiments to the industrial process. Nucl Eng Des 1995;153:14551.

M.K. Samal et al. / Engineering Failure Analysis 16 (2009) 85111

111

[14] Gerber D. EPRIs fatigue monitoring system, EPRI fatigue seminar operating power plant fatigue damage assessment, Burlington
VT, USA; August 1995.
[15] Bartonicek J, Zaiss W, Hienstorfer W, Kocklemann H, Schockle F. Monitoring system and determination of actual fatigue usage.
Nucl Eng Des 1995;153:12733.
[16] Implementation of power plant component life assessment technology using a knowledge-based system. In: Jovanovic A, Verelst L,
editors. Proceedings of the SP 249 Project Final workshop; February 1997.
[17] ASME boiler and pressure code. Case N-47 (29) Class 1 components in elevated temperature service, Section 3, Division 1. New York:
ASME; 1991.
[18] RCC-MR Appendix A16, AFCEN, Paris; 2002.
[19] R6, Assessment of the integrity of structures containing defects. British energy, Report R6-Rev.4; 2001.
[20] Goodall IW, editor. Assessment procedure for the high temperature response of structures, R5. Nuclear electric procedure, Issue 2;
1998.
[21] Assessment of components operating in the creep regime, Section-10 of API-579 document; June 2001.
[22] Stress intensity factor solutions for FFS assessment of a component, Appendix-C of API-579 document; June 2001.
[23] Reference stress solutions for FFS assessment of a component, Appendix-D of API-579 document; June 2001.
[24] Material properties for FFS assessment of a component, Appendix-F of API-579 document; June 2001.
[25] ASME boiler and pressure vessel code, New York; 1995.
[26] Nuclear Electric R5: Assessment procedure for high temperature response of structures, Nuclear Electric plc Report R5 Issue 1: vol. 4,
Assessment procedure for defects under steady loading; vol. 5, Creepfatigue crack growth. UK: Berkeley Nuclear Laboratories;
1990.
[27] A16: Guide for defect assessment and leak before break analysis, Commissariat A Lenergie Atomique, Direction des Reacteurs
Nucleaires, Departement de Mecanique et de Technologie, 3rd Draft, Rapport DMT 96.096, EPAC 5450, che cooperative 4557,
France; 1995.
[28] Reidel R, Rice JR. Tensile cracks in creeping in solids. In: Proceedings of the 12th ASTM fracture mechanics conference. ASTM STP
700. Philadelphia: ASTM; 1980.
[29] Kumar V, German D, Shih CF. An engineering approach for elasticplastic fracture analysis, EPRI nal report NP 1931; 1981.
[30] Chen KL, Kuo AY. Greens function technique for structures subjected to multiple site thermal loading. Struct Mech Reactor
Technol 1989:3538.
[31] Mukhopadhyay NK, Dutta BK, Kushwaha HS. On-line creepfatigue monitoring system for components at elevated temperature.
In: Third workshop on creep, fatigue and creepfatigue interaction. Kalpakkam: IGCAR; 1999. p. C34963.
[32] Socie DF. Fatigue life prediction using local stress strain concepts. Exp Mech 1977;17:506.
[33] Downing SD, Socie DF. Simple rainow counting algorithm. Int J Fatigue 1982;4:3140.
[34] Rychlik I. A new denition of the rainow cycle counting method. Int J Fatigue 1987;9:11921.
[35] Glinka G, Kam JC. Rainow counting algorithm for very long stress histories. Int J Fatigue 1987;9:2238.
[36] Abdulsalam M, Stanley JT. Steady-state model for erosioncorrosion of feed water piping. Corrosion 1992;48:58793.
[37] Caleyo F. A study on reliability assessment methodology for pipelines with active corrosion defects. Int J Pressure Vessels Piping
2002:779.
[38] Sweeton FH, Baes CF. J Chem Thermodyn 1970;2:479.
[39] Fleming KN. Markov models for evaluating risk-informed in-service inspection strategies for nuclear power plant piping systems.
Reliab Eng Syst Safety 2004;83:2745.

You might also like