Proceedings of the ASME 2011 Pressure Vessels & Piping Division Conference

PVP2011
July 17-21, 2011, Baltimore, Maryland, USA

PVP2011-57531

Extremely Low Probability of Rupture (xLPR) Version 1.0 Code –

Pilot Study Problem Results

*†
D. Rudland C. Harrington
US Nuclear Regulatory Commission Electric Power Research Institute
Office of Nuclear Regulatory Research Dallas, Texas, USA
Mail Stop: CSB-5CA24
Washington, DC 20555-0001

ABSTRACT source code framework to determine the framework and

Nuclear Regulatory Commission (NRC) Standard Review Plan architecture requirements appropriate for building a modular-
(SRP) 3.6.3 describes Leak-Before-Break (LBB) assessment based code with this complexity. The pilot study focused on
procedures that can be used to demonstrate compliance with the PWSCC in pressurizer surge nozzles, and is meant to
10CFR50 Appendix A, GDC-4 requirement that primary demonstrate the feasibility of this code and approach and not to
system pressure piping exhibit an extremely low probability of determine the absolute values of the probability of rupture.
Later development phases will broaden the scope of xLPR to
rupture. SRP 3.6.3 does not allow for assessment of piping
appropriate primary piping systems in pressurized and boiling
systems with active degradation mechanisms, such as Primary water reactors (PWR and BWR), using an incremental
Water Stress Corrosion Cracking (PWSCC) which is currently approach that incorporates the design requirements and lessons
occurring in systems that have been granted LBB approvals. learned from previous iterations.
Along with a series of existing qualitative steps to assure safety This paper specifically examines the xLPR Version 1.0 model,
in LBB-approved lines experiencing PWSCC, NRC staff, the methods and approach used to couple the deterministic
working cooperatively with the nuclear industry through a modules within a probabilistic software framework, and the
memorandum of understanding with the Electric Power results from the pilot study. A comparison of the results
Research Institute, is developing a new, modular based, specific to the surge nozzle sample problem is presented. This
comprehensive piping system assessment methodology to paper concludes with lessons learned from the pilot study.
directly demonstrate compliance with the regulations. This
project, called xLPR (eXtremely Low Probability of Rupture), INTRODUCTION
will model the effects and uncertainties of relevant active 10 CFR Part 50, Appendix A, General Design Criteria (GDC) 4
degradation mechanisms and the associated mitigation states, in part, that the dynamic effects associated with
activities. The resulting analytical tool will be comprehensive, postulated reactor coolant system pipe ruptures may be
vetted with respect to the technical bases of models and inputs, excluded from the design basis when analyses reviewed and
flexible enough to permit analysis of a variety of in-service approved by the NRC demonstrate that the probability of fluid
situations and adaptable such as to accommodate evolving and system piping rupture is extremely low under conditions
improving knowledge. consistent with the design basis. Licensees have typically
demonstrated compliance with this probabilistic criterion
A multi-year project has begun that will first focus on the through deterministic and highly conservative analyses. Given
development of a viable method and approach to address the recent advances in probabilistic methodologies, the NRC staff
effects of PWSCC as well as define the requirements necessary and industry believe that performing a probabilistic analysis of
for a modular-based assessment tool. To meet this goal, the primary system piping that fully addresses and quantifies
first version of this code has been developed as part of a pilot uncertainties and directly demonstrates compliance with GDC 4
study, which leverages existing fracture mechanics based is more appropriate. The NRC and industry expect that a
models and software coupled to both a commercial and an open robust probabilistic software tool, developed cooperatively, will

*
Corresponding author, david.rudland@nrc.gov
†
The views expressed herein are those of the author and do not represent an official position of the USNRC

1 Copyright © 2011 by ASME

This work is in part a work of the U.S. Government. ASME disclaims all interest in the U.S. Government’s contributions.
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 facilitate meeting this goal, and result in improvement in develop the commercial framework version of the xLPR
licensing, regulatory decision-making and design, and will be Model, while the code SIAM-PFM was developed with only
mutually beneficial. Based on the terminology of GDC 4, this open source code to demonstrate this platform. The details of
effort is titled eXtremely Low Probability of Rupture (xLPR). each framework code are given later in this paper.
Development of the xLPR methodology and the corresponding
software tool will involve many challenging technical The propagation of uncertainties is handled by a nested loop
decisions, modeling judgments, and sensitivity analyses. structure where epistemic (lack of knowledge or systematic
uncertainty) and aleatory (irreducible or statistical uncertainty)
The development of a sophisticated probabilistic software tool uncertainties are propagated separately through the model. The
that meets quality assurance (QA) and technical requirements is deterministic kernel for the initial xLPR code is embedded in a
a daunting task. The management structure, the probabilistic time loop structure whose details are discussed in the next
framework, and data handling are just a few of the issues that section. The time loop consists of deterministic models for
need to be addressed early in the software development effort. crack initiation, crack growth, crack stability, leakage, in-
In order to meet this need, a pilot study was conducted to service inspection, and PWSCC mitigation.
demonstrate the feasibility of the proposed NRC-industry
cooperative development process and determine the appropriate The programmatic outcome of the pilot study is intended to be
probabilistic framework for calculating the probability of an optimized development process for the general tool for
rupture for a surge nozzle dissimilar metal weld. The pilot assessing primary piping system safety. In making
study provides relative, order-of-magnitude estimates of piping recommendations for the best computational framework,
rupture probabilities; such analysis will identify areas requiring models, and input distributions for use in the long-term study, a
more focused attention in the long-term study. gap assessment was conducted and lessons learned were
identified, which will be used to identify and prioritize research
Following the pilot study, a more detailed long term study will recommendations. The final outcome of the pilot study will be
be completed to generalize the analysis procedures to all a research plan for moving forward to attain the long term goal
primary system piping. The long-term study will employ the of a fully modularized, probabilistic assessment tool for
same basic organizational, management, and NRC-industry primary piping systems.
cooperative structure as the pilot study. However, technical
and programmatic lessons learned from the pilot study will be
incorporated into the long-term study. Technical issues from xLPR PROCESS
the pilot study left unresolved due to their complexity will be Figure 1 illustrates the time loop for the xLPR Version 1.0
prioritized for resolution in the long-term study. code. The purple boxes in this figure are individual modules
that are summarized later and detailed in Reference [4]. Before
This paper specifically examines the xLPR Version 1.0 model, entering the time loop, the load module is called. This module
the methods and approach used to couple the deterministic calculates all of the appropriate stress values, including the
modules within a probabilistic software framework, and the weld residual stress, for each of the load inputs for the crack
results from the pilot study. A comparison of the results initiation, growth, and stability modules.
specific to the surge nozzle sample problem is presented. This
paper is a continuation of the work presented in [1] and a From main
loop
summary of the work presented in [2].
Load
Module

xLPR PILOT STUDY DESCRIPTION t=t+1

no

As described above, the xLPR pilot study was conducted to Critical Flaw
Module
demonstrate the feasibility of the proposed developmental yes
t>tf Leak module – COD
Critical?

process and framework for a probabilistic code to address leak rate Module

degradation mechanisms in piping system safety assessments. Preemptive TWC SC yes

Mitigate? Inspection no
The pilot study addresses the specific issue of assessing the Module - POD SC
Critical?
yes TWC
probability of rupture of dissimilar metal (DM), pressurizer Stress
Intensity
Crack
Growth
surge nozzle welds degraded by PWSCC, particularly those Module Module

previously assessed [3] for which a considerable amount of Crack Initiation

publicly available information already exists. The pilot study Module

provides a short term, learning experience that should benefit Crack Coalescence
Module
the longer term program and code development by identifying
areas requiring more focused effort. Figure 1 Time loop flow chart for Version 1.0 xLPR code
In developing the computational implementation for the For each time increment (set by user input), the code checks
Version 1.0 xLPR code, both open-source and commercial whether the analysis is beyond the predefined time period for
software was considered. Working in parallel, two unique the analysis. If it is, the time loop is exited. If not, it continues.
codes were developed to demonstrate the advantages of each
framework. The commercial software GoldSim was used to

2 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 If a pre-emptive mitigation is to be performed at this time (i.e. fatigue cycling is not considered). A set of tensile and
increment, the code will apply mitigation and continue. bending loads are obtained from an average of values for
selected operating plants [5]. The sustained loads from normal
If the crack initiation model dictates that a crack initiates in this thermal, deadweight, and thermal stratification are considered
time step, a single, surface-breaking crack is placed in the and safe shutdown earthquake loads are also defined. The
model with the sampled size and location. The coalescence above loads are considered to be constant for the pilot study.
model will not permit locating a crack at the same location as The as-welded residual stresses are also considered, and are
an existing crack. taken to be axisymmetric. A self-equilibrating through-wall
distribution is defined in terms of a third-order polynomial,
The crack growth module is then used to calculate the crack with random parameters to describe the scatter of residual
growth increment for any existing cracks. As a first step to this stresses.
calculation, the stress intensity factor module is called and
appropriate stress intensity factors are calculated from the Initiation Module: The initiation of PWSCC is a complicated
instantaneous loads, including the weld residual stress. These process where phenomenological models are not matured to the
stress intensity factors are then used with the crack growth point of including them in probabilistic fracture mechanics
model to calculate the crack growth increment. Each of the software. Therefore, for Version 1.0 of the xLPR code, models
existing crack sizes is updated, and if any surface crack has that are empirically driven were considered. Three separate
reached 100% through-wall, it is transitioned to a through-wall models are currently incorporated into the code, see Figure 2.
crack. In addition, a check is made to determine if any surface Of the three models, two are time-based models that are
cracks or through-wall cracks meet the coalescence threshold corrected for temperature and stress. The third is a Weibull
criteria. If they do, they are combined. model that is also corrected for temperature and stress. Any of
these models can be calibrated to either laboratory or service-
Next, the crack stability module is used to determine if any based crack initiation data. For the Version 1.0 code, the
existing cracks have reached a critical size. At any time models were calibrated to the service data found in [3], which
increment, through-wall cracks (TWC), or surface cracks (SC) leads to an arrival rate of approximately 0.01 cracks/year. The
may exist in the analyses. For existing through-wall cracks, if details of the crack initiation models and their calibrations can
the instantaneous crack size is larger than the critical crack size, be found in [4].
a double-ended break (severance of the pipe) is assumed. For
existing surface cracks, if net-section collapse failure is
predicted at the operating loads, the crack transitions to a
through-wall crack. In this case, if the resultant through-wall
crack length is greater than the critical through-wall crack
length, a double ended break is assumed. The size of this
opening at failure is recorded and the time loop is exited.

If a through-wall crack is not critical, the leakage module is

used to determine the level of the leakage. A leakage
calculation is performed for the through-wall crack using the
calculated crack-opening displacement. The calculated leak
rate for this time increment is stored and the time loop
continues.

If the surface and/or through-wall cracks are found to be stable,

the inspection module is used to calculate a probability of
detection for each crack in the analysis. The time loop then Figure 2 Flow Diagram of Version 1.0 xLPR Crack
continues. Initiation Module.
The crack initiation models treat size effects by breaking the
weldment up into segments of length ℓ. Only one crack can
MODULE DESCRIPTIONS FOR xLPR VERSION 1.0 initiate in each segment. Each of the segments can have a
The xLPR Version 1.0 software was constructed as a collection different stress and temperature. The uncertainty in the
of modules. Descriptions of the modules for Version 1.0 are “constants” in the relation defining the initiation time has
summarized below and are discussed in detail in [4]. This within-heat and heat-to-heat contributions (or within-weld and
version concentrates on initiation and growth of primary water weld-to-weld). The weld-to-weld uncertainty defines the
stress corrosion cracks (PWSCC) in a dissimilar metal scatter in the time until the first crack initiates in a particular
pressurizer surge nozzle weld. weld. The within-weld uncertainty parameter defines the
subsequent scatter in the time to cracking for segments in that
Loads: The loads on the pipe weldment are defined before particular weld. It would be expected that the time to cracking
beginning the analysis. Loads contributing to stresses normal is more tightly correlated for the wetted material of a particular
to circumferential cracks are considered. Since the Version 1.0
code concentrates on PWSCC, only sustained loads are needed

3 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 weld than for wetted material for multiple welds given that instability that employs an elastic-plastic formulation for
some key welding parameters do not vary for a particular weld. evaluation of the applied J-integral that is based on a reduced
thickness analogy to estimate the compliance of cracked
The relative variances of heat-to-heat versus within-heat may elastic-plastic tubes subject to tension and bending [12].
control the propensity to form long cracks as compared to
cracks that are isolated around the circumference. The choice Transition from part-through to through-wall cracks is handled
of ℓ could be based on specimen size (such as one inch or two by determining the through-wall crack length where the
inches) with modifications based on calibrations to field data. cracked area is equal to the part-through wall crack area at
For the pilot study, 19 segments (each about 2-inches in length) through-wall penetration. Once a through-wall crack becomes
were chosen. Sensitivity studies conducted demonstrate that unstable, a double-ended break is considered to occur.
the leak probabilities are not highly influenced by small
changes in the number of segments. An initiated crack is Crack Opening Displacement: Calculation of crack opening
randomly placed within its associated segment. displacement (COD) is essential in the prediction of leakage
through cracks. There is a range of COD prediction
Stress Intensity Factors: Stress intensity factor solutions for performance as documented in [13], where the GE/EPRI
part-through circumferential semi-elliptical surface cracks (SC) method is suggested to be the “best” method for calculating
[6] and straight-fronted through-wall circumferential cracks COD. The basis for the recommendation in [13] was presented
(TWC) [7] are included. The part-through solutions consider in [14], where the details of a comparison of a variety of COD
through-wall stress distributions described by a fourth order analysis methods with finite element calculations are presented.
polynomial. These stresses vary only through the thickness. The conclusion was that the GE/EPRI method is the best choice
Local stress intensity factors for the deepest point and surface for COD calculation. Based on these results and
point are provided. For through-wall cracks, tension, through- recommendation, the decision was made to use the GE/EPRI
thickness, and global bending stresses are considered. COD analysis in xLPR. While implementing this process,
errors were found in the combined tension and bending
Crack Growth: For the pilot study, PWSCC of DM butt solutions described in [14]. The work conducted within the
welds is the only subcritical cracking mechanism that is xLPR program [4] and which is currently coded in Version 1.0,
considered. For each time increment, PWSCC growth is corrected the errors in the proceeding publication.
calculated using the loads and stress intensity solutions
presented in the previous section. The PWSCC crack growth Leak Rates: Leak rates for straight-fronted through-wall
model incorporated into the xLPR Version 1.0 code was cracks (complex cracks will be addressed in later versions of
developed from Alloy 82/182 DM weld laboratory the code) are evaluated based on the SQUIRT software [15],
experimental data used to measure the growth rate of PWSCC. which, in turn, is based on the Henry-Fauske two-phase flow
The collective experimental data has been gathered and model. Pressure drops due to entrance effects, friction, phase
analyzed by EPRI in MRP-115 [8]. The PWSCC growth change (liquid to gas), and bends and protrusions are
model developed in that effort included a temperature corrected considered. The effects of leak detection are determined during
crack growth rate with uncertainty in the rate both from a post-processing. Leak rates are calculated for each active TWC
within weld and weld-to-weld perspective. The effects of in the analysis.
dissolved hydrogen on the MRP-115 crack growth rate were
detailed in MRP-263 [9] and implemented into the xLPR
Version 1.0 code. A modification to the MRP-115 equation FRAMEWORK DESCRIPTIONS FOR xLPR VERSION
was developed in the MRP-263 study adding the dependence 1.0
on dissolved hydrogen concentration. One of the objectives of the xLPR pilot study was to investigate
the computational frameworks that are capable of meeting the
Inspection: The influence of in-service inspections (ISI) is objectives of this investigation. The choice of the appropriate
treated through the probability of detection (POD) [10]. The computational framework is essential to assure that the code
POD is a function of crack size that is estimated from test data; can be developed in a modular fashion and still be able to
therefore, there is uncertainty in the POD. At each time step, provide the structure for uncertainty handling. To meet this
the POD for each of the cracks is recorded, and the influence of objective, both commercial and open source framework
inspection on leak probabilities (leaks of various sizes) is software were considered. In this section, a brief description of
evaluated during post-processing. For cracks that grow to leak each of the two frameworks selected for use in developing
during the lifetime, the crack contributes (1-POD) leaks, rather xLPR Version 1.0 is presented. Details for each framework can
than one leak. If more than one inspection takes place then the be found in References [16,17].
influence is the product of the (1-POD)s (independent
inspections) or the (1-POD) of the last inspection (dependent GoldSim
inspections). The xLPR Version 1.0 model and pilot study case was
constructed using the models described above coupled to a
Crack Stability: The stability of part-through cracks is based commercial software framework. After a comprehensive
on net-section collapse for tension and bending loading [11]. review of the commercial software available for this effort, the
The stability of through-wall cracks is based on tearing Version 1.0 xLPR model was constructed using the commercial

4 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 software, GoldSim Pro, which is compatible with the free history of all of the changes that have occurred over the life of
downloadable version of the GoldSim Player ‡ software. the model. Providing this configuration management capability
GoldSim Pro includes both a model developer’s version and the is particularly useful for coordinating model changes when
simulation software. multiple people have the ability to access and modify the same
model file, and as a Quality Assurance/Quality Control feature
The GoldSim software is dynamic, probabilistic simulation allowing for verification and documentation of where and when
software developed by GoldSim Technology Group, LLC. This changes have been made to a model file. Details of the
general-purpose simulator is a hybrid of several simulation GoldSim framework can be found in Reference [16].
approaches, combining an extension of system dynamics with
some aspects of discrete event simulation, and embedding the SIAM-PFM
dynamic simulation engine within a Monte Carlo simulation The SIAM-PFM Framework is a problem-solving environment.
framework [18]. The acronym SIAM-PFM (SIAM for short) stands for
Structural Integrity Assessment Modular – Probabilistic
The modular-based GoldSim framework model for the Version Fracture Mechanics. SIAM-PFM is an Object Oriented Open
1.0 xLPR model manages input variables (e.g., material Source (OOOS) framework, within which a wide range of
properties) and model output (e.g., results), as well as the flow nuclear power plant safety issues can be addressed in a
of information that includes the system level model logic. The systematic and consistent way by using modern principles of
GoldSim framework for xLPR was constructed with an option probabilistic risk assessment. Probability techniques are
to use standard Microsoft Excel spreadsheets to define the applied to problems in fracture mechanics in order to predict
inputs as well as dynamically pass simulation results to fracture behavior, and thus to assess the structural integrity of a
Microsoft Excel for advanced post-processing. The commercial variety of nuclear power plant components that passively bear
framework simulation software serves as the integrating shell large loads over long periods of time. This platform is readily
that links various modules used in the xLPR Model. extensible to different problem classes. A common feature of
the different applications is that they are all the subjects of
The GoldSim xLPR model framework controls the order in probabilistic risk assessment and, therefore, represent “risk-
which the modules are called and the passing of data into and informed” analyses.
out of modules. The xLPR approach is to define all of the
modules independently, so that each one can be created by Every SIAM-PFM component is written using the Python
collaborators in any common programming language. Both programming language and Python frameworks. They are
simple and complex calculations are coded as modules and are easily installed and un-installed on Windows operating
directly coupled to the xLPR GoldSim framework using systems, and all components are potentially portable to other
dynamic link libraries (DLLs) by wrapping the original module operating systems such as UNIX and OS X.
source code in a simple standard DLL shell [18].
In the SIAM-PFM problem solving environment, all
The framework utilizes the GoldSim software libraries of components use the same working principle: workspaces that
probability distribution functions and the capability to correlate contain projects (SIAM projects), or directories in which all
variables and perform multiple-realization stochastic analyses inputs and outputs of a given test case, are saved in both binary
in a Monte Carlo approach. The framework benefits from the and in text files. Users can navigate through the different
GoldSim software's ability to store simulation data from large projects in the project-explorer panel as on any IDE (Integrated
numbers of realizations and generate statistics on global Development Environment) or windows explorer and create
probability distributions. GoldSim permits each run to be saved projects that represent test cases. Convenient plots of the
in a single action, including all input data and results from outputs are provided to visualize data, and users can also
Monte Carlo analysis. Finally, the GoldSim framework has extract the raw data in text files to create custom plots.
built in graphical user interface (GUI) functions that allow the
developer to quickly assemble specific model runs and to create The implementation of xLPR within SIAM (SIAM-xLPR)
interactive player files for end-users, which allow for viewing, presents a series of tabs to define the case conditions. In these
navigating, and even modification of input values and model tabs, pre-defined input values have been set by default to the
options to run the xLPR model using the GoldSim Player probabilistic base case. The SIAM-xLPR main GUI framework
software application [18]. has seven tabs. Input data can be entered on the first six tabs,
in any order. Default input values are provided, and can be
The GoldSim framework for xLPR includes the software’s modified as needed. The seventh tab displays the SIAM-xLPR
ability to track changes that have been made to a model file. “Execute Utility” window that presents a command line view of
This feature (referred to as versioning) allows the differences program executions, where realizations are created and
between the current version and a previous version of a model executed according to the program flow.
file to be quickly determined [18]. The version history is an
integral part of the model file, providing an easy-to-access The SIAM-xLPR framework has been developed primarily
using the Python v.2.6 programming language and Python
‡
GoldSim Player allows the user to view and navigate through the model frameworks scipy, numpy and PyQt. Details of the SIAM
logic, run an existing GoldSim model, and display the results, without having to framework can be found in Reference [17].
purchase GoldSim Pro.

5 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 • Revised mean weld residual stress inputs as shown in
PILOT STUDY PROBLEM STATEMENT Figure 3.
To demonstrate the feasibility of conducting analyses to assess o ID stress: mean = -345 MPa, standard
the probability of rupture in pressurizer surge nozzles resulting deviation = 34 MPa.
from PWSCC, and to compare the results from the two o Xc: mean = 0.38, standard deviation = 0.038)
frameworks developed in this effort, a pilot study problem Chemical Mitigation
statement was developed. This problem statement consisted of • Two cases, where the base case hydrogen value of 25
a probabilistic base case, and a series of sensitivity analyses cc/kg-STP was increased to 50 and 80 cc/kg-STP,
that are intended to demonstrate the features of the Version 1.0 respectively.
code. Crack Initiation
• Direct Model I was used in place of Direct Model II
The base case and five sensitivity cases are summarized in Safe End Evaluation
Table 1. The analyses were conducted using the controlled • Revised mean weld residual stress inputs as shown in
versions of both the GoldSim and SIAM framework models Figure 3.
developed for the xLPR pilot study. o ID stress: mean = -16.2 MPa, standard
Table 1 xLPR Version 1.0 Analyses. deviation = 117 MPa.
o Xc : mean = 0.18, standard deviation = 0.036
Analysis Description
Probabilistic Base Probabilistic base case analysis using In addition to the sensitivity analyses, the base case and the safe
Case Monte Carlo sampling. end evaluation case were both post-processed to take credit for
Sensitivity Study leak detection and inspection. For each case, leak rate
detection limits of 0.1, 1, 10 and 50 gpm were considered. The
Analyses evaluate different mitigation inspection intervals assumed were 2, 5, 10, 20 and 30 years.
Stress Mitigation times, for the same stress-based
mitigation.
Xc
Chemical effects of increasing the
Chemical hydrogen concentration in the water on
Mitigation the crack growth module. Three hydrogen
concentrations were evaluated.
Considers the crack initiation model
Crack Initiation
uncertainty.
Considers stainless steel safe end weld,
Safe End which causes a through-thickness bending
Evaluation stress that can reduce the tensile inner-
diameter stress.

The details of the pilot study problem statement can be found in

[2], with some of the pertinent inputs given below:
Figure 3 Comparison of base case and mitigated mean weld
Base case residual stress
• Normal operating loads and geometry from MRP-216 [5].
• Material property distributions were derived from Uncertainty in parameters is handled using probability
distributions, defined and categorized by the Models and Inputs
database of nuclear pipe material property experiments
[4]. Group. For the pilot study, expert judgment was used to
categorize the uncertainties in the inputs as either epistemic or
• Mean weld residual stress shown in Figure 3
aleatory. The general uncertainty characterization for the
o ID stress: mean = 300 MPa, standard inputs is given in Table 2. Detailed input information is given
deviation = 110MPa.
in the pilot study problem statement [2].
o Xc §: mean = 0.25, standard deviation = 0.05
• Direct Method II crack initiation model was used [4].
• No inspections or leak detection.

Stress mitigation case

• Three cases were run with mitigation at time 10, 20, and
40 years.

§
Xc, as defined in Figure 3, is the distance through wall where the weld
residual stress changes sign.

6 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 2 General variable uncertainty characterization for
pilot study **
Aleatory Epistemic
Some crack initiation
Crack size at initiation
parameters
Material strength properties Weld residual stress
Some crack growth
Material toughness properties
parameters
Some crack growth
Probability of detection
parameters
Pressure
Temperature

PILOT STUDY RESULTS

Each of the problems in the pilot study was run with both the Figure 4 Probability of rupture for the base case
GoldSim and SIAM framework. Throughout this section,
comparisons between the results from the two codes are Switching the crack initiation parameters from epistemic to
presented. In certain cases, results from only one code are aleatory greatly influences the behavior of the epistemic
shown since the trends between the codes were similar. When realizations, as shown in Figure 5. With reclassification of the
this occurs, the results from the code that is used in any crack initiation parameter to aleatory uncertainty, the
particular figure are identified by a model result designation, probability of rupture varies with time, i.e., the grey lines are no
i.e., GSxLPRv1.0 (GoldSim) or SIAM_v1.0 (SIAM). In longer vertical. This change leads to a smoother probability
addition, for the base case analyses, the total number of estimate over the 60-year timeframe. An interesting
realizations was chosen to be 50,000 (1,000 epistemic and 50 consequence from these assessments is that the quantile curves
aleatory). are now completely different. Because the time of crack
occurrence was not fixed for each epistemic set, it is more
The probability of rupture results for the base case are shown in likely to have at least one crack for each epistemic realization
Figure 6. In this figure, the light grey vertical lines represent (but a smaller chance that ALL realizations within an epistemic
the probability of rupture for each epistemic realization, i.e., set leads to rupture). While the estimate of the mean probability
representing aleatory uncertainty. In this case, for each of rupture gives similar results, it is not the case for the quantile
epistemic realization, there were 50 aleatory realizations. The values. Their interpretation changes considerably from one
fact that each grey line is nearly vertical indicates that for each assumption to the other. In Figure 5, for instance, a median of
epistemic realization, all 50 aleatory realizations either ruptured 0.32 at 60 years means that half of the epistemic realizations
(value = 1), or did not rupture (value = 0), i.e., there was no have a 32% chance of a rupture in the future, while in the base
effect of the aleatory uncertainty. The vertical lines also occur case, there was absolutely no chance of rupture for half of the
well distributed across the 60-year time period which indicates epistemic realizations (median in Figure 4 was zero).
that the epistemic uncertainty is controlling the behavior. In
addition to the light grey lines, the mean and standard
quantiles †† are also shown in Figure 4. For this case, the mean
value suggests that there is a 41% chance of pipe rupture in 60
years (720 months).

Figure 5 Probability of rupture for base case with crack

initiation uncertainty changed from epistemic to
** aleatory
Variables/inputs not shown are considered constant.
††
Points taken at regular intervals from the cumulative distribution
function of a random variable

7 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 A comparison of the mean probability of rupture using the This modification of the stress profile affects both the crack
GoldSim and SIAM xLPR framework codes is shown in Figure initiation and growth models. For the example shown in
6. The results show a good comparison between the codes, Figure 8, the mitigation was applied at 10, 20 and 40 years.
with the main difference resulting from slight differences ‡‡ in The results from these analyses demonstrate that for the inputs
the crack initiation probabilities. considered, the application of the stress-based mitigation causes
the mean probability of rupture to no longer increase with time.
Since the data shown in Figure 8 represents the cumulative
probability of rupture, a horizontal line represents no additional
ruptures during that time period. However, upon a close
investigation of the results from Figure 8, it is clear that before
the probabilities cease to increase, they rise slightly above the
non-mitigated rupture probabilities. This is due to the tensile
zone in the mitigated weld residual stress distribution. For a/t
values between 0.5 and 0.9, the weld residual stress becomes
tensile. In those realizations where a crack is present that is at
least 50% deep at the time the mitigation occurs, the crack
growth rate is increased. This effect causes the slight increase
in the rupture probabilities before the mitigation effects become
apparent.

Figure 6 Comparison of GoldSim and SIAM Framework

for mean probability of rupture for the base case
The effect of the combination of leakage detection and
inspection on the base case mean probability of rupture is
shown in Figure 7. In this case, a 2-yr inspection interval and
a leak detection limit of 1gpm is shown. These results illustrate
an almost four-order of magnitude decrease in the probability
of rupture when credit is taken for in-service inspection and
leak detection.

Figure 8 Effect of pre-emptive mitigation on the mean

rupture probability
As described earlier, the effects of hydrogen on the PWSCC
growth rate were implemented in the xLPR Version 1.0 code.
The effects of hydrogen and zinc on the initiation of PWSCC
was considered, but not implemented. The effects of increasing
the hydrogen concentration on the mean rupture probabilities
are shown in Figure 9. The increase in hydrogen caused a
decrease in the mean probability of rupture. This change is
attributed solely to the change in the crack growth rate due to
the increased hydrogen concentration. A large change in
rupture probability occurred when the hydrogen content was
increased from 25 cc/kg to 50 cc/kg. However, only a marginal
Figure 7 Effect of inspection and leak detection on base
increase in rupture probability occurred when the hydrogen
case mean probability of rupture
concentration was increased from 50 cc/kg to 80 cc/kg.
The effect of pre-emptive mitigation using a stress-based Overall, the decrease in rupture probability is only about a
technique on the mean rupture probabilities is shown in Figure factor of two when the hydrogen concentration increases from
8. The only mechanical mitigation method that is incorporated 25 cc/kg to 80 cc/kg at 60 years.
into xLPR Version 1.0 is a pre-emptive stress-based mitigation.
For this option, the user inputs a mitigated weld residual stress
distribution and a time at which that mitigation is to occur.

‡‡
See [2] for more details

8 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Figure 9 Effect of hydrogen on the mean probability of
rupture
Figure 11 Mean probability of rupture for safe end
The mean probability of rupture for the safe end sensitivity case sensitivity case with mitigation, leak detection
is compared to the base case in Figure 10. The only difference and inspection
in the inputs for this case is the change in the weld residual
stress. Inspections or leak detection are not credited in either
case. The change in weld residual stress caused a two order of The confidence for the mean value of the probability of rupture
magnitude decrease in the mean probability of rupture. is calculated using the bootstrap method [2], which consists of
sampling with replacement over the response generated by the
original analyses. The central limit theorem states that when
the mean and variance of the initial distribution are finite, the
mean distribution should be asymptotically normal. Therefore,
if the distribution of the mean values is normal, the sample size
should be large enough to represent a stable solution. The
confidence in the mean probability of rupture for the safe end
case with a 10 year ISI and a 1 gpm leak detection limit is
shown in Figure 12. The distribution in Figure 12 is nearly
normal suggesting a stable solution.
80
Probability Distribution Function

70
60
50
40

Figure 10 Mean probability of rupture for base and safe 30

end case 20
Taking mitigation, a 1 gpm leak detection limit and a 10 year 10
inspection interval into account reduces the mean rupture 0
probability by six orders of magnitude at 60 years, see Figure
4.6E-08 6.7E-07 1.3E-06 1.9E-06 2.6E-06 3.2E-06
11. For this example, a mitigation time of 20 years was chosen.
10,000 realizations were used with importance sampling on Mean Probability of Rupture
both the weld residual stress and the crack initiation Figure 12 Confidence in the mean probability of rupture
parameters, which were found to control the uncertainty in this for safe end case with ISI and leak detection
problem [2]. The results of the analyses are shown in Figure
11. As shown in this figure, the mitigation at 20 years reduced The confidence for the safe end case with a 20-yr mitigation, a
the rupture probability at 60 years by two orders of magnitude, 10 year ISI and a 1gpm leak detection limit is shown in Figure
while the leak detection and ISI reduced the rupture 13. Since only a few of the 10,000 realizations produced
probabilities by about three orders of magnitude. The rupture with mitigation, inspection, and leak detection, the
combined effect caused a reduction of six orders of magnitude confidence in the mean value is low. As with the previous
on the rupture probability at 60 years. example, the confidence is estimated using the bootstrap
method, which would produce a normal distribution of the
mean if the analysis was stable. However, the results in Figure
13 suggest an exponential distribution and indicate a lack of

9 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 stability in the analyses. Additional realizations would be all aspects of degradation of LBB-approved lines are
needed to obtain a better estimate of the mean value for the required before this code can be used to assess full
probability of rupture. However, the distribution reported in compliance with regulations.
Figure 13 spans approximately one order of magnitude, which • Focus should be placed on data storage, data handling, and
is considered good based on the sample size (105) compared to post-processing to ease the difficulty of running this code
the calculated probability of rupture ([10-9-10-8]). Based on the for the end user.
previous results, increasing the number of realizations leading • Importance sampling is necessary for the calculation of the
to rupture by a factor of 2 or 3 may be enough to significantly probability of rupture in piping systems. Emphasis needs to
increase the accuracy. Therefore, it may not be necessary to be placed on processes and procedures for identifying the
increase the sample size by an order of magnitude. variables that need to be importance sampled. Adaptive
sampling, or other reliability methods, needs to be
considered.
250
• The classification of uncertainty is very important to
GSxLPRv1.02_M02_SE_003.gsm;
GSxLPRv1.02_M02_CFO_IN10L_exp.txt; understanding the uncertainty in the probability of rupture.
200 bootstrap_LHS_Importance_Safe_End_Mitig.xlsx Uncertainty is not trivial and needs involvement at all levels
GSxLPRv1.01_M02.gsm; of development of a complex system. Knowing which
GSxLPR_v1.01_M02_CFO_001_MAX_exp.txt variables control the rupture and what part of the uncertainty
150
PDF

in those variables is epistemic and can be reduced will not

100 only inform the regulators, but will also help direct future
research in this area.
50
SUMMARY
This paper summarizes the technical basis and results from the
0 pilot study for Version 1.0 of the xLPR code. The xLPR code
5.0E-10 2.0E-09 4.0E-09 6.0E-09 8.0E-09 9.9E-09 is being written cooperatively through an addendum to the US
Mean probability of rupture NRC and EPRI Memorandum of Understanding, and is
intended to aid in assessing whether LBB-approved systems
Figure 13 Confidence in the mean probability of rupture experiencing PWSCC quantitatively comply with regulations.
for safe end case with mitigation, ISI and leak In addition, the code is intended to be used to calculate the
detection probability of various failure modes for RCS components. This
first version of the code is meant to demonstrate the feasibility
of both the management and coding structure needed for the
LESSONS LEARNED
development of a more robust and comprehensive code but not
The xLPR task groups worked very closely over an
to determine absolute values for the probability of rupture. To
approximately 1.5-year period to develop two separate
meet this limited goal, the pilot study focused on a single
framework codes. This effort included not only developing the
location with a single degradation mechanism, i.e., pressurizer
framework structure and flow, but also development of the
surge nozzle dissimilar metal weld with PWSCC.
configuration management program, documentation and
verification of the framework, and the modules development, as
For the pilot study, experts were gathered to determine the
well as understanding and implementing the physics-based
appropriate models and inputs needed to determine the
modules necessary for the calculation of probability of rupture.
probability of rupture for the pressurizer surge nozzle DM
Over that time period, the team developed an appreciation for
weld. These experts compiled, coded, and verified, using a pre-
the complexity of this problem, and the project structure needed
defined configuration management process, the modules
for successful completion of the pilot study. Through the
needed for the stated purpose. These modules included loads
process, many important lessons were learned:
(with weld residual stress), crack initiation, crack growth, crack
coalescence, crack stability, crack opening displacement,
• Configuration management is only a small portion of quality leakage, inspection, and mitigation. In addition, these self-
assurance. Establishing a program QA plan and controls for contained modules were incorporated into two distinct
xLPR is the essential first step in the continuing probabilistic frameworks developed using open source and
development process. Each organization involved in the commercial software respectively. The frameworks were
development of this code must then comply with the developed to control the time flow of the analyses, while
program QA requirements. The xLPR program needs to linking the modules and properly accounting for and
have a transparent and traceable CM system that will cover propagating the problem uncertainty.
the xLPR code lifecycle.
• A well-written unambiguous software requirements To demonstrate the capabilities of the codes developed, a
document needs to be developed and followed for future problem statement was developed that consisted of a base case
xLPR versions. and sensitivity studies. These runs focused on demonstrating
• From the models standpoint, certain assumptions were made the calculation of the probability of rupture with and without
due to the limited scope of the pilot study. Best-estimate mitigation, with and without inspection, and with and without
models with properly characterized uncertainties that cover

10 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 leakage detection, while also accounting for irreducible and Gabriel Ilevbare – EPRI
reducible uncertainties. To propagate these uncertainties, the Frank Ammirato – EPRI
code used nested loops to separate the uncertainty types. Patrick Heasler – Pacific Northwest National Laboratory
Importance sampling was used for cases of extremely low Bruce Bishop - Westinghouse
probability of rupture.
Inputs Group
The complete xLPR pilot study effort, which includes not only Eric Focht – U.S. NRC
the code development efforts, but the management structure, Mark Kirk – U.S. NRC
the pilot statement problem, and the detailed analysis of the Guy DeBoo – Exelon
results clearly demonstrate that it is feasible to develop a Paul Scott – Battelle
modular-based computer code for the determination of Ashok Nana – AREVA NP Inc.
probability of rupture for LBB approved piping systems. John Broussard – Dominion Engineering
Nathan Palm – Westinghouse
ACKNOWLEDGEMENTS Pat Heasler – Pacific Northwest National Laboratory
The development of the Version 1.0 xLPR code was a group Gery Wilkowski – Emc2
effort that spanned a variety of experts across many fields of
expertise from the U.S. NRC, EPRI, and their contractors. The Acceptance Group
success of the program emphasizes the dedication of the xLPR Mark Kirk – U.S. NRC
team, the strength of its leadership, and the generous support Glenn White – Dominion Engineering Inc.
from both the NRC and EPRI. There are many people to Aladar Csontos – U.S. NRC
thank, including members from the computational, models, Robert Hardies – U.S. NRC
inputs and acceptance groups, as well as the Program David Rudland – U.S. NRC
Integration Board (PIB). Every person on this team provided Bruce Bishop – Westinghouse
valuable contributions, and their efforts were sincerely Robert Tregoning – U.S. NRC
appreciated.
Program Integration Board
Computational Group Craig Harrington – EPRI
David Rudland – U.S. NRC Aladar Csontos – U.S. NRC
Bruce Bishop – Westinghouse Robert Hardies – U.S. NRC
Nathan Palm – Westinghouse Denny Weakland - Ironwood Consulting
Patrick Mattie – Sandia National Laboratories David Rudland – U.S. NRC
Cedric Sallaberry – Sandia National Laboratories Bruce Bishop – Westinghouse
Don Kalinich – Sandia National Laboratories Eric Focht – U.S. NRC
Jon Helton – Sandia National Laboratories Guy DeBoo – Exelon
Hilda Klasky – Oak Ridge National Laboratory Marjorie Erickson – PEAI
Paul Williams – Oak Ridge National Laboratory Gary Stevens – U.S. NRC
Robert Kurth – Emc2 Howard Rathbun – U.S. NRC
Scott Sanborn – Pacific Northwest National Laboratory Mark Kirk – U.S. NRC
David Harris – Structural Integrity Associates Glenn White – Dominion Engineering Inc.
Dilip Dedhia – Structural Integrity Associates
Anitha Gubbi – Structural Integrity Associates
REFERENCES
Models Group
Marjorie Erickson – PEAI 1 Rudland, D., Mattie, P., Klasky, H., Harris, D., Bishop, B.,
Gary Stevens – U.S. NRC and Kurth, R., “Development Of Computational
Howard Rathbun – U.S. NRC Framework And Architecture For Extremely Low
David Rudland – U.S. NRC Probability Of Rupture (xLPR) Code,” PVP2010-25963,
John Broussard – Dominion Engineering Proceedings of the ASME 2010 Pressure Vessels & Piping
Glenn White – Dominion Engineering Division / K-PVP Conference, PVP2010, July 18-22, 2010,
Do-Jun Shim – Emc2 Bellevue, Washington, USA.
Gery Wilkowski – Emc2 2 Rudland, D., “xLPR Version 1.0 report, Technical Basis
Bud Brust – Emc2 and Pilot Study Problem Results.” NRC Letter Report,
Cliff Lange – Structural Integrity Associates xLPR Computational Group, ML110660292, 2011.
Dave Harris – Structural Integrity Associates 3 Materials Reliability Program: Advanced FEA Evaluation
Steve Fyfitch – AREVA NP Inc. of Growth of Postulated Circumferential PWSCC Flaws in
Ashok Nana – AREVA NP Inc. Pressurizer Nozzle Dissimilar Metal Welds (MRP-216,
Rick Olson – Battelle Rev. 1) EPRI, Palo Alto, CA: 2007. 1015383. MRP-216,
Darrell Paul – Battelle Rev. 1.
Lee Fredette – Battelle
Craig Harrington – EPRI

11 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 4 Materials Reliability Program: Models and Inputs 17 H. B. Klasky, P. T. Williams, S, Yin, and B. R. Bass,
Developed for Use in the xLPR Pilot Study (MRP-302). “SIAM-xLPR Version 1.0 Framework Report” ,
EPRI, Palo Alto, CA 2011. 1022528 ORNL/NRC/LTR-248, Letter report to the Office of
5 Materials Reliability Program: Advanced FEA Evaluation Nuclear Regulatory Research, US NRC, ML110700026,
of Growth of Postulated Circumferential PWSCC Flaws in September 2010
Pressurizer Nozzle Dissimilar Metal Welds (MRP-216, 18 Goldsim User’s Guide, Version 10.0; Goldsim Technology
Rev. 1) EPRI, Palo Alto, CA: 2007. 1015383. MRP-216, Group: Issaquah, WA, February 2009.
Rev. 1.
6 Anderson, T.L., Thorwald, G., Revelle, D.A., and Lanaud,
C., “Stress Intensity Solutions for Surface Cracks and
Buried Cracks in Cylinders, Spheres, and Flat Plates,”
Structural Reliability Technology final report to The
Materials Property Council, Inc., March 14, 2000.
7 Anderson, T.L., “Stress Intensity and Crack Opening Area
Solutions for Through-wall Cracks in Cylinders, and
Spheres,” Structural Reliability Technology final report to
The Materials Property Council, Inc., January 29, 2003.
8 G.A. White, N.S., Nordmann, J. Hinkling, and C.D.
Harrington, “Development of Crack Growth Rate
Disposition Curves for Primary Water Stress Corrosion
Cracking (PWSCC) of Alloy 82, 182 and 132 weldments,”
TMS (The Minerals, Metals and Materials Society),
Proceedings of the 12th International Conference on
Environmental Degradation of Nuclear Power Systems –
Water Reactors, (2005)
9 Materials Reliability Program: Technical Bases for the
Chemical Mitigation of Primary Water Stress Corrosion
Cracking in Pressurized Water Reactors (MRP-263), EPRI,
Palo Alto, CA: 2009. 1019082. [Proprietary]
10 Materials Reliability Program: Materials Reliability
Program: Development of Probability of Detection Curves
for Ultrasonic Examination of Dissimilar Metal Welds
(MRP-262, Rev. 1) EPRI, Palo Alto, CA: 2009. 1020451.
MRP-262, Rev. 1.
11 Rahman, S.; "Net-Section-Collapse Analysis of
Circumferentially Cracked Cylinders - Part II: Idealized
Cracks and Closed-Form Equations"; Engineering Fracture
Mechanics; Vol. 61; 1998; pp. 213-230.
12 Gilles, P., and Brust, F. W., “Approximate Methods for
Fracture Analysis of Tubular Members Subjected to
Combined Tension and Bending Loads,” Proceedings of
the 8th OMAE Conference, Hague, The Netherlands, 1989.
13 Scott, P., Olson, R. and Wilkowski, G.; “Development of
Technical basis for Leak-Before-Break Evaluation
Procedures”; NUREG/CR-6765, May 2002.
14 Rudland, D., Wang, Y., and Wilkowski, G.; “Comparison
of Crack-Opening Displacement Predictions for LBB
Applications”; International Journal of Pressure Vessels
and Piping; Vol 79; 2002: pp. 209-217.
15 “User’s Guide for SQUIRT (Windows Version 1.1),”
Battelle Memorial Institute, March 2003.
16 P. D. Mattie, C. J. Sallaberry, J. C. Helton and D. A.
Kalinich, “Development, Analysis, and Evaluation of a
Commercial Software Framework for the Study of
Extremely Low Probability of Rupture (xLPR) Events at
Nuclear Power Plants,” SAND2010-8480, Letter report to
the Office of Nuclear Regulatory Research,
ML110700019, US NRC, December 2010

12 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use