You are on page 1of 3

APPLYING NDE RELIABILITY

Greg Selby, EPRI

ABSTRACT
Often we perform NDE in order to comply with regulatory or other requirements, without taking credit for
the reduction in flaw population that it provides. This reduction in flaw population can form part of the
technical basis for the integrity of a system and potentially for a reduced inspection schedule in the future.
One way to perform this analysis is by using probabilistic fracture mechanics. A current example in the
US is the Extremely Low Probability of Rupture (xLPR) project, whose purpose is to verify the integrity
of PWR piping systems that include nickel-alloy welds. The xLPR analysis will calculate the rupture
probability by Monte Carlo simulations using probability distributions for flaw introduction, flaw growth,
flaw coalescence, inspection reliability, leak detection reliability, and application of mitigations. The
inspection reliability inputs include both probability of detection (POD) and probability of correct sizing.
The POD has been calculated using a logistic statistical analysis of thousands of data points acquired
during NDE qualifications. Questions remain as to whether, and how, to adjust this POD curve to account
for the differences between the qualification environment and the environment in the power plant. These
questions have greater urgency after the Spring 2012 failure of a qualified, manual ultrasonic examination
of a dissimilar metal weld in a US PWR failed to detect five axial cracks, which were discovered a few
days later when two of the leaked after machining of the outside surface.

INTRODUCTION
Primary piping systems in US pressurized water reactor (PWR) nuclear power plants (NPP) have been
evaluated since the 1990s using an approach called Leak Before Break (LBB). This approach allows a
significant reduction in the number of pipe supports and hangers. One of the fundamental assumptions of
LBB is that the piping system is not subject to any known active degradation system. This assumption has
been obviated by the emergence in the 2000s of primary water stress corrosion cracking (PWSCC) in
dissimilar metal welds in PWR primary piping. Industry need a new analysis to supplement or replace
LBB.
A large, joint project was initiated by industry and the US Nuclear Regulatory Commission
(USNRC) with the objective of developing a new safety analysis approach for PWR primary piping, one
that would allow for the possibility of PWSCC. This project is called Extremely Low Probability of
Rupture (xLPR). The technical approach is based on probabilistic fracture mechanics.
Probabilistic fracture mechanics calculations are used for problems that may be too complex for
deterministic solutions, for example when there are several key input variables that all may vary
substantially independently of one another. In the case of xLPR, the input variables include the material’s
PWSCC susceptibility; the probability of crack initiation, as a function of time at operating conditions;
crack growth rate; residual and operating stresses; the frequency of volumetric examination, and the
probability that an existing crack would be detected; NDE sizing accuracy; the frequency and
effectiveness of leak detection; leak rates expected through cracks; and so on. A probability distribution is
developed for each input variable. Then a series of thousands of simulations begin for each weld. For
each simulation, a single value of each input variable is selected at random, weighted according to its
probability distribution (highly improbable values will be selected occasionally, but not often). An
iterative calculation is begun, simulating the performance of the weld from the beginning of plant
operation through 60 years of life. At each time increment, several evaluations occur: a new crack may
initiate; existing cracks may grow in length and depth; multiple existing cracks may coalesce; volumetric
inspections, performed at specific times during life, may detect the cracks, leading to repair or
replacement; cracks may be detected by leakage; mitigations may be applied; and if the cracking is severe,
the applied loads may rupture it. The simulation ends when the weld ruptures or reaches a life of 60 years.
After thousands of simulations the rupture probability is developed as a function of operating time.

24
The xLPR calculation is complex and the input variable distributions are difficult to achieve. The
many organizations collaborating in xLPR have formed several working groups to develop these input
variable distributions, and other working groups to develop the simulation and evaluation approaches.

PROBABILITY OF DETECTION
The calculation takes account of the reliability of the periodic volumetric examinations that are performed
in compliance with ASME Code and USNRC regulations. It does so by applying crack detection
probability distributions. The xLPR team decided to use probability of detection (POD) statistics
developed from the extensive database accumulated by EPRI during its administration of the qualification
program established by the Performance Demonstration Initiative (PDI) for implementation of the
requirements of ASME Code, Section XI, Appendix VIII. This database was selected for several reasons:
the population of observations is high; the ultrasonic responses of the qualification cracks have been
benchmarked against the responses of PWSCC in the field; the inspection procedures, personnel, and
equipment used in qualification are the same as those used in the field; and the qualifications were
performed and the database was accumulated under EPRI’s quality assurance (QA) program. The team
developed POD curves – POD as a function of crack through-wall depth – for use in developing crack
detection probability inputs for the xLPR calculation [1].
A pilot calculation showed that the inspections, in combination with leak detection, provided a
significant reduction in the probability of pipe rupture. Figure 1 shows the rupture probability over the 60-
year lifetime of a pressurizer surge line weld. The solid curve, ending with a rupture probability of about
half a percent through 60 years of life, takes no credit for inspection or leak detection. The dashed curve
takes credit for leak detection and for inspection, using the EPRI POD statistics; it shows a reduction of
three orders of magnitude in rupture probability.
A question remained: what is the relation between the POD documented during qualification, and
the POD that is actually achieved in the field inspection environment?

Figure 1. Pilot calculation: rupture probability in a pressurizer surge line.

25
A FAILURE TO DETECT CRACKING IN THE FIELD
During the Spring of 2012 a US PWR chose to install a weld overlay for mitigation of a dissimilar metal
weld to the steam generator. The weld was about 130mm thick, with the outside surface tapered at an
angle of 11°. The purpose of the overlay was only mitigation; no cracking was known to exist. In
preparation for the overlay the weld was examined using a qualified ultrasonic procedure and examiner.
The inspection that was directed in the axial direction, to detect circumferential cracks, was performed
using a manually-driven, encoded technique, which produces the same results as an automated
examination. The inspection that was directed in the circumferential direction, to detect axial cracks, was
performed using a purely manual technique. No cracks were reported by either examination.
A few days later, the outside surface of the weld was machined to remove the 11° taper. Visual
inspection of the machined surface revealed two leaking, axially-oriented cracks in the weld. The
circumferential examination was repeated using the same probes, and also other available probes that were
considered to be reasonably applicable but were not qualified for this weld. The two leaking cracks, and
three other axial cracks that were deep but not leaking, were all detected using the original qualified
probes and also some of the unqualified probes.1
The utility conducted a root cause evaluation. The evaluation found that the utility, the plant site,
and the examiner all contributed to the error. The utility had not fully installed into its procedures the
published industry guidance for dissimilar metal weld examination [2]. The plant site had not adequately
prepared the examiner for the examination, and had not provided adequate oversight; and the examiner did
not take advantage of all his opportunities to prepare and practice for the examination.
Executive and technical industry groups were formed to assimilate the lessons from this failure and
to develop industry actions and guidance intended to prevent another occurrence. This guidance is
expected to be issued before the end of 2012.

THE IMPACT ON XLPR


Clearly, at least for manual examination the probability of detection achieved in the field may vary from
that demonstrated during qualification. The xLPR team will need to determine an adjustment to the
qualification-based crack detection probability distributions, recalculate the rupture probability, and
evaluate the significance of the effect. For fully encoded examinations –both robotic and manually-driven
– perhaps only minimal adjustments, or no adjustments, need be made to the qualification-based detection
probabilities. Coverage and quality of coupling can be assessed by review of the recorded data;
interpretations of the data can be performed by many analysts if needed.

CONCLUSION
This is a work in progress. Clearly we can conclude that non-encoded, manual examination may benefit
from additional oversight before and during the examination. Other improvements to industry practices
will be implemented through existing industry initiatives. Those are mostly human issues; the difficult
issue that is both technical and human –how to quantify crack detection reliability for manual examination
in the field –is unsolved.

REFERENCES
1. Materials Reliability Program: Development of Probability of Detection Curves for Ultrasonic
Examination of Dissimilar Metal Welds (MRP-262, Revision 1) – Typical PWR Leak-Before-Break
Line Locations. EPRI, Palo Alto, CA: 2009. 1020451.
2. Nondestructive Evaluation: Guideline for Conducting Ultrasonic Examinations of Dissimilar Metal
Welds. EPRI, Palo Alto, CA: 2009. 1018181.
1
The utility redeveloped the analysis for the overlay, taking into consideration the presence of the cracks. Because the cracks were
all oriented axially the overlay was still feasible. The utility excavated the tops of the leaking cracks, welded up the excavations
manually, and proceeded to apply the weld overlay successfully.

26

You might also like