Beruflich Dokumente
Kultur Dokumente
Abstract
Failure analysis of in-service nuclear piping containing defects is an important subject in the nuclear power plants. Considering
the uncertainties in various internal operating loadings and external forces, including earthquake and wind, flaw sizes, material
fracture toughness and flow stress, this paper presents a probabilistic assessment methodology for in-service nuclear piping
containing defects, which is especially designed for programming. A general sampling computation method of the stress intensity
factor (SIF), in the form of the relationship between the SIF and the axial force, bending moment and torsion, is adopted in
the probabilistic assessment methodology. This relationship has been successfully used in developing the software, Safety
Assessment System of In-service Pressure Piping Containing Flaws (SAPP-2003), based on a well-known engineering safety
assessment procedure R6. A numerical example is given to show the application of the SAPP-2003 software. The failure
probabilities of each defect and the whole piping can be obtained by this software.
2004 Elsevier B.V. All rights reserved.
1. Introduction and the core shroud, etc. Thus, the existence of flaws
in the nuclear piping is unavoidable.
With the development of the nuclear power industry, There are benefits of a probabilistic approach as it
more and more pipelines are being used. The mechan- can be used to ensure that the main safety concerns are
ical integrity of the nuclear reactor piping is a matter addressed in an economic manner. The relative val-
of great importance for both economical and safety ues of the failure probabilities of pressure piping may
reasons. Flaws are inherent in many components ow- be used as a guide to the most economic deployment
ing to the processes by which they are manufactured of resources on maintenance, inspection and repair.
or fabricated. For example, in a welded joint defects The assessment results could be used, for example, to
due to the porosity and lack of penetration or fusion concentrate ultrasonic inspection on the locations at
may exist prior to the application of any load. Fur- highest risk, thus gaining maximum benefit from the
thermore, many intergranular stress corrosion crack- inspection. The traditional approach of safety assess-
ing (IGSCC) problems have been observed in these ment and design lies in a deterministic model. For the
plants in areas, such as the primary coolant system deterministic method, when the random distributions
of variables are narrow, the safety margins will be
Corresponding author. Tel./fax: +86-413-6651160. over large. When random distributions of variables are
E-mail address: yjxie@mail.fsptt.ln.cn (Y.J. Xie). wider, the results of the assessment should become un-
0029-5493/$ see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.nucengdes.2003.12.007
238 Y.C. Lin et al. / Nuclear Engineering and Design 229 (2004) 237246
0.6
Plastic collapse (2)
"Safe" region cut-off
0.4
y + u
Lmax
r = (3)
2y
0.2
Lrmax
where y is the yield strength or 0.2% proof stress and
0.0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 u is the materials ultimate tensile stress.
Lr The Option 2 curve is defined by
Lr = yield stress based limit load cracked component. Thus, the FAD in this option is
(1) dependent on the material and the geometry. The es-
3. Random variables and sampling distributions may result in negative values for random
variables. Operating loadings are also fitted by the nor-
3.1. Uncertainties of main assessment parameters mal, log normal and Weibull distributions. The defect
sizes given by non-destructive testing (NDT) can fol-
The operating loadings applied on piping fluctuate low a normal, log normal or exponential distribution.
in most cases, especially for some particular operating
conditions, including internal pressure and tempera- 3.3. Sampling method
ture. These fluctuating loadings may lead to substan-
tial contributions to the axial force, bending moment The R6 procedure provides estimates of failure
and torque moment at the locations of defects. Oc- probability, using detailed integration methods. This
casionally, the external forces, including earthquake complex integration is in terms of the probability
and wind, in many localities also need consideration. density functions (PDFs) of defect size, fracture
So, when the probabilistic analysis of nuclear piping toughness and flow stress. The shortcomings of this
is performed, the operating loadings and the external method are obvious. Firstly, it is necessary to consider
forces should be considered as stochastic variables. more actual uncertainties when calculating the fail-
In addition, defect size, fracture toughness and flow ure probability of assessed pressure piping. Secondly,
stress always have uncertainties. Therefore, the vari- this numerical integration method introduced in R6 is
ability of such parameters as defect size, operating unsuitable for programming.
loads (internal pressure, operating temperature and in- In this paper, the MonteCarlo simulation method
stallation error), the external forces, including earth- is used to sample the stochastic variables. The
quake and wind, fracture toughness and flow stress MonteCarlo simulation method is a simplification.
(obtained from yield stress and ultimate tensile stress), The only requirement of the method is that the phys-
are taken into account in the SAPP-2003 application ical (or mathematical) system should be described by
software. PDFs, i.e. by giving the PDFs of defect size, fracture
In the above-described random parameters, fracture toughness, flow stress and operating loadings, respec-
toughness, for example, is one of the most important tively. Then, the MonteCarlo simulation can proceed
and sensitive parameters. However, accurate fracture by sampling from these PDFs, which necessitates a
toughness data cannot be obtained from the current fast and effective way to generate random numbers
materials of piping, and there are many uncertainties uniformly distributed within the interval [0, 1]. In
during experiments. Therefore, it is necessary to sam- this program, the mean value, the distribution type
ple these stochastic variables according to the appro- and standard deviation of every random variable are
priate distribution. needed when performing the failure analysis of the
studied piping.
3.2. Distribution of main random parameters
The normal, log normal, exponential and Weibull 4. Calculating stress intensity factor
distributions are four different types of distribution
that have been used to describe the variations of the In many safety assessment codes, such as BS 7910
above-mentioned random variables in this software. (British Standards Institution, 1999), R6 procedure,
The parameters defining these distributions should be the fracture assessment procedures are mainly defined
specified for normal operating and fault conditions for Mode I loading. For mixed Mode II and Mode III
and for the initial loading and proof test as neces- loadings, calculations of effective SIF are introduced
sary. The R6 procedure proposed that the log normal in the appendix of BS 7910 and the R6 procedure.
and Weibull distributions are suitable for the fracture For the ductile tearing, it is recommended that the ap-
toughness, yield stress and ultimate tensile stress. This proach is restricted to initiation toughness (e.g. J0.2 ),
is because the use of a normal distribution for fracture unless it is known that the tearing follows a clearly
toughness and flow stress has the disadvantage that the defined path. Among the planar flaws of the pressure
mathematical transformations carried out with these piping, the main defects are circumferential and axial
Y.C. Lin et al. / Nuclear Engineering and Design 229 (2004) 237246 241
M M 2q
Tn N N Tn
R t
cracks. For circumferential cracks, the axial forces, crack, respectively. Due to the uncertainties of oper-
bending moments and torsions imposed on the pipe ating pressure, operating temperature and installation
element where the defects exist, contribute to the SIF. error, the axial force and bending moment in Eqs. (7)
For axial cracks, the contribution to the SIF mainly and (8) are also uncertain. They can be calculated by
comes from the internal pressure. The magnitude
of the axial force, bending moment and torsion are N = P NP + T NT + U NU + NG + QNQ + W NW
associated with the operating pressure, operating tem- (9)
perature and installation error of the pressure piping.
When a probabilistic failure analysis of piping con- M = P MP + T MT + U MU + MG
taining defects is carried out, the uncertainties of these + QMQ + W MW (10)
random variables should be treated by one appropriate
where NP , NT , NU , NQ and NW are the axial forces
sampling method. Furthermore, it is critical to com-
at the location of the crack for normalized loads (in-
bine the calculating requirements of the MonteCarlo
cluding operating pressure, operating temperature,
method and the computing characteristics of the SIF,
installation error, seismic and wind force). NG is the
and then give the proper sampling expression for the
contribution from gravity of the piping and its internal
SIF.
fluids. The following six parameters, MP , MT , MU ,
In this paper, a sampling expression of the SIF for
MQ , MG and MW , have similar meanings. For the
circumferential internal surface cracked pipe is intro-
mixed Mode II and Mode III loadings, the torques in
duced (see Fig. 2). The following discussion is based
the vicinity of defects, TP , TT , TU , TQ , TG , and TW ,
on Mode I loadings. Firstly, assume that the axial force
should also be considered and the meanings of these
is N and bending moment is M. The Mode I linear
six parameters are similar to those of NP , NT , NU ,
elastic SIF can be evaluated from
NQ , NG and NW . However, due to the installation
KI = KIN + KIM (6) complexity of the industry piping and the uncertain-
ties of the operating loadings and material properties,
where KIN and KIM represent the Mode I linear elastic it is often a complicated system or condition. In ad-
SIF components from the axial force N and bending dition, the pressure piping is often subjected to the
moment M, respectively, which can be determined as combination of axial force, bending moment, torsion
and the difference in temperature of different parts of
KIN = N KIN (a, b) (7)
the pipelines. So, performing the structural analysis
KIM = M KIM (a, b) (8) of actual piping is not a simple and easy task. In
order to make the SAPP-2003 application software
where KIN and KIM are the non-dimensional SIF com- easy to operate and obtain more accurate risk analysis
ponents from the normalized axial force and the nor- results, a structural analysis module for this software
malized bending moment, respectively. They can be has been developed, which is a secondary developed
calculated according to the methods introduced in the product of ANSYS engineering analysis software. In
Ductile Fracture Handbook (Zahoor, 1989) and the this module, the nuclear piping can be easily mod-
proposals made by Xie (1998). Parameters, a and b, eled and be exactly analyzed, simulating the actual
represent the crack length and height of the studied operating condition. Therefore, the above-mentioned
242 Y.C. Lin et al. / Nuclear Engineering and Design 229 (2004) 237246
normalized parameters, NP , NT , NU , NG , NW , NQ , mated as the quotient of the failure counts to the num-
MP , MT , MU , MG , MQ and MW , can be easily ber of simulations, i.e.
calculated using this structural analysis module. n
From Eqs. (6)(10), the following expression can Pfi = (13)
m
be derived
When the MonteCarlo algorithm is used to eval-
KI = [(P NP + T NT + U NU + NG uate the structural reliability of nuclear components
+ QNQ + W NW )]KIN (a, b) with very low failure probabilities, a very large number
of simulations have to be taken in order to achieve high
+ [(P MP + T MT + U MU + MG computational accuracy. This can be accomplished
+ QMQ + W MW )]KIM (a, b) (11) with the developing computer technology.
For the Mode I loading, the flow chart to compute
As shown in Eq. (11), a, b, P, T, U, Q and W are ran-
the failure probability of a separate defect is shown
dom variables. If these stochastic variables are sam-
in Fig. 3. The processes to perform the probabilistic
pled according to the given standard deviations and
distributions, Eq. (11) will be the desired sampling
expression for the Mode I linear elastic SIF. If the re- Start
p
where KI (a, b) represents the non-dimensional SIF Draw the sample of random
variables by the Monte Carlo
resulting from the operating pressure. method
It is noted that if an embedded crack is considered, Repeat m times
Characterize the defect (the
the depth of the crack, p, should be also included in compound cracks and interaction
Eqs. (7), (8), (11) and (12). defects need recharacterisation)
Under many engineering conditions, the failure Plot assessment point on the
probability and a more accurate description of in- failure assessment diagram(FAD)
In a MonteCarlo simulation, based on a selected Fig. 3. The flow chart to compute failure probability of separate
failure criterion, the failure probability can be esti- defect.
Y.C. Lin et al. / Nuclear Engineering and Design 229 (2004) 237246 243
fracture failure analysis of nuclear piping using SAPP- In actual nuclear piping, the failure events of
2003 are briefly described as follows. Firstly, the welded joints containing defects are often correlated
parameters, including the detail geometry and the ma- to each other because the random variables of one
terial property parameters of the studied pipe, should certain welded joint containing defects may be cor-
be input in the interface of the secondary developed related with those of another welded joint containing
ANSYS module so as to model the actual piping and defects. These random variables are the operating
perform the structural analysis. Secondly, output the pressure, operating temperature, installation error,
axial force, bending moment and torque induced by and earthquake and wind (Zhou et al., 1998b). So,
normalized loads at the location of the assessed defects the axial force, bending moment and torque of one
from structural analysis module. Thirdly, back to the weld joint are correlated with those of another weld
main interface of the SAPP-2003 software and input joint. Because each welded joint containing defects
the pipes material property parameters, the defects is series connected in the logical chart of pressure
and welding lines characteristic parameters, which piping containing defects, the lower bound (Pf1 ) and
include the category, sizes and locations of defects, the upper bound (Pf2 ) of failure probability of the as-
material types and directions of welding lines. The sessed piping can be obtained by Eqs. (14) and (15),
software will automatically save all data to the system respectively.
database. Then, define the simulation number and the Pf1 = max(Pfi ) for i = 1, 2, . . . , n (14)
R6 options used in the analysis. Finally, perform the
n
safety assessment and review the analysis results.
Pf2 = 1 (1 P fi ) for i = 1, 2, . . . , n (15)
i=1
5.2. Computing the failure probability of a whole
piping by the MonteCarlo method where Pfi is the failure probability of the ith welded
joint containing defects and n is the number of all
The absolute values of failure probability can be weld joints in the assessed piping. The upper bound,
used as a part of an overall piping safety analysis, to Pf2 , is calculated under the condition that the failure
make a case for a new piping, or continued operation event of every welded joint containing defects is in-
of the old piping. Similarly, the results can be used in dependent of each other, based on the weakest link
the piping design or modification studies to investigate model (Dai et al., 1987). So, if the Pf2 of the pressure
the effect of changing the various design parameters piping is smaller than its acceptable failure probabil-
and thereby optimize the design. ity Pa , the assessed piping is reliable and can be used
However, the structure of piping is very complex continuously; while if Pf1 of the pressure piping is
in most cases. Huang (1990) and Zhou et al. (1998a) greater than the acceptable failure probability Pa , the
proposed that the physical chart of the assessed sys- assessed piping is unreliable and should be discarded
tem should be changed into a logical chart before the or repaired; if
reliability analysis of the system is performed. Here,
Pf1 < Pa < Pf2 (16)
a physical chart means the practical structure of the
assessed system, and a logical chart implies the rela- and Pf1 and Pf2 are of the same order of magnitude,
tionship between the failure event of the system and it is suggested that the assessed pressure piping is
the failure events of every component. So, to carry out unreliable and should be discarded or repaired (this is
the fracture risk analysis of pressure piping, the first safe and convenient); while if Eq. (16) is satisfied and
step is to change the complex structure chart into a Pf1 and Pf2 are not of the same order of magnitude,
logical chart for the fracture risk analysis, and then the the failure probability of the assessed piping must
correlation among failure events of every welded joint be known in order to make a reliability assessment.
should be considered. Through NDT, most defects of So, theoretical analysis of the correlation among fail-
assessed piping can be found in some welded joints. ure events of the welded joints containing defects
Therefore, the failure of the whole nuclear pressure must be carried out, and a corresponding method of
piping will occur only if any one of the welded joints computing the failure probability of a piping contain-
containing defects fails. ing defects, Pf , should be set up. Because the other
244 Y.C. Lin et al. / Nuclear Engineering and Design 229 (2004) 237246
stochastic variables (material properties, defect sizes (5) Repeat step 4, n times to assess all other weld
etc) of defected welded joints are independent of each joints. If the assessment result is unsafe, increment
other in the assessed nuclear piping, it is critical to the failure counter nn by 1.
consider the contribution from the operating pressure, (6) Repeat steps 35, mm times. Calculate the failure
operating temperature, installation error, earthquake probability of the assessed piping, Pf = nn/mm
and wind. For Mode I loading, for example, the linear and save it to the system database.
elastic SIF can be obtained by Eq. (17)
Table 2
Main parameters needed in the analysis
Parameters Mean Distribution Standard deviation
Table 3
Analysis results of three cracks and the studied piping
Fracture toughness Failure probability The piping
deviation (kJ m2 )
The first crack The second crack The third crack Lower bound Upper bound Failure probability
the values of the random variables standard devia- given in the present paper to show SAPP-2003s ap-
tions and the partial safety factors need to be given plication in risk analysis of in-service nuclear pressure
when performing the deterministic assessment using piping containing defects.
this program. For the above example, the determin-
istic assessment results are the failure probabilities
of three cracks, 0, and the failure probability of the Acknowledgements
whole piping, 0.
For the engineering application, the probabilistic as- This research project was funded by CNPC. The
sessment method can generally provide a more accu- authors would like to express appreciation for their
rate description of in-service pressure piping integrity supporting the researchs (Project No.99081414).
status, which leads to full utilization of the existing
structures especially those containing defects. Uncer-
References
tainties in defining the distributions and deviations of
material properties, operating loadings and particu-
Ainsworth, R.A., et al., 2001. Assessment of the Integrity of
larly defects sizes mean that the calculations may be Structures Containing Defects. British Energy Generation Ltd,
more beneficial in producing relative failure probabili- UK, R6-Revision 4.
ties for ranking assessment locations and for the exam- British Energy Generation Ltd, 1999. Assessment of the integrity
ination of operating conditions. Relative results may of structures containing defects. R6-Revision 3, Amendment
be used as a guide to the most economic deployment 10.
British Standards Institution, 1999. Guide on the Method for
of resources on analysis, maintenance, inspection and Assessing the Acceptability of Flaws in Fusion Welded Struc-
repair. But, most traditional approaches to safety as- tures. London, BS 7910.
sessment lie in a deterministic model that invariably Budden, P.J., Shaples, J.K., Dowling, A.R., 2000. The R6
involves a large safety factor usually assigned from procedure: recent developments and comparison with alternative
heuristic and somewhat arbitrary decisions. approaches. Int. J. Pres. Ves. Pip. 77, 895903.
Dai, S.H., et al., 1987. Reliability Engineering and its Applications
in Chemical Equipment (Chinese). Beijing.
Dowling, A.R., Townley, C.H.A., 1975. The effects of defects on
7. Conclusions structural failure: a two criteria approach. Int. J. Pres. Ves. Pip.
3, 77137.
As special structural components, piping is subject Harrison, R.P., Loosemore, K., Milne, I., 1976. Assessment of
the Integrity of Structures Containing Defects. CEGB Report,
to many kinds of load, such as the internal pressure, Central Electricity Generating Board, UK, R/4/R6.
operating temperature and gravity force, which give Huang, X.R., 1990. Reliability Engineering (Chinese). Beijing.
substantial contributions to the SIF in piping. In many Milne, I., Ainsworth, R.A., Dowling, A.R., Stewart, A.J., 1986.
places, the external forces, including earthquake and Assessment of the Integrity of Structures Containing Defects.
wind, are also very important. In the present paper, a CEGB Report, Central Electricity Generating Board, UK,
R/H/R6-Rev3.
general sampling expression of the SIF, in the form Rastogi, R., Bhasin, V., et al., 2002. Assessment of integrity of
of the relationship between the SIF and the axial components in piping of 500 MWe PHWR: using R-6 method.
force, bending moment and torsion, is proposed. This Nucl. Eng. Des. 212, 109114.
expression provides a critical basis for the developed Xie, Y.J., 1998. A theory on cracked pipe. Int. J. Pres. Ves. Pip.
software, SAPP-2003. The normalized axial force, 75, 865869.
Zahoor, A., 1989. Ductile Fracture Handbook, vols.13. Electric
bending moment, torsion in the location of flaws, Power Research Institute, Palo Alto, EPRI NP6301-D,
which are needed in the general sampling expression London.
of the SIF, can be easily obtained from the struc- Zhou, Z.G., et al., 1996. The Application of Probability Fracture
tural analysis module of SAPP-2003 software. The Mechanics in the Pressure Vessels (Chinese). Beijing.
failure probabilities of piping are determined by the Zhou, J., et al., 1998a. The Reliability Assessment of System in
Engineering-principle and Method (Chinese). Chang Qin.
MonteCarlo method, considering the uncertainties Zhou, J.Q., et al., 1998b. A study on reliability assessment metho-
of operating loading, defect sizes, material fracture dology for pressure piping containing circumferential defects I
toughness and flow stress. A numerical example is & II. Int. J. Pres. Ves. Pip. 75, 679691.