Session: MF-04-01 Fitness-For-Service and Failure Assessment

Paper Number: 62046

Start Time: Tuesday, July 13, 2021, 09:00 AM

62046 - Comparative Assessments of Different Leakage-Rate Models for ATLAS+ 

In support of the Euratom research and training program Advanced Structural Integrity Assessment
Tools for Safe Long-Term Operation (ATLAS+), this paper presents a comparative assessment of both
conservative and best-estimate leakage-rate models carried out using the recently-developed Leak
Analysis of Piping – OCI (LOCI™) code, an application in the Structural Integrity Assessments
Modular – OCI (SIAM-OCI™) framework. Oakridge Consulting International (OCI) performed the
comparative assessment via its role as a supporting international member of the ATLAS+ consortium.
Benchmarking studies were also carried out comparing the LOCI™ results with the XPIPE™ code
developed by Materialprüfungsanstalt Universität Stuttgart (MPA) and a new leakage-rate code
developed by Jacobs Engineering, Cheshire, UK.

In the context of an LBB assessment where emphasis is placed on the detection of small leakage rates, the
distinction between conservative and best-estimate models could be very important. This classification
can provide guidance on where these models can be appropriately applied in terms of understanding
margins and uncertainties. Country-specific regulatory requirements for an LBB assessment can also play
an important role in model selection. A conservative leakage-rate model is required to consistently
underestimate the actual leakage rate and is typically applied in deterministic procedures, e.g., the
deterministic approach in the German standard KTA 3206. A best-estimate leakage-rate model places
greater emphasis on predictive accuracy. In a probabilistic framework, the model’s estimated solutions
can serve as a measure of central tendency (mean, median, or mode) for a prescribed statistical
distribution characterizing the uncertainties in the leakage-rate predictions.

In a more general probabilistic analysis, model inputs and their associated statistical distributions are
sometimes proposed which have conservative values for parameters to ensure that some extreme cases are
not under-represented. However, conservatism can lead to over-estimates of the probability of leakage or
rupture and, thereby, invalidate the model. Moreover, conservatism at a subsystem (i.e., model) level can
potentially become non-conservative when implemented at the system or framework level.

Presenting Author: Paul Williams Oakridge Consulting International, Inc.

Authors:

Paul Williams Oakridge Consulting International, Inc.
Richard Bass Oakridge Consulting International, Inc.
Peter Dillström Kiwa Inspecta Technology AB
Peter Gill Jacobs Engineering
Brian Daniels Jacobs Engineering
Robert Lammert Materials Testing Institute (MPA), University of Stuttgart
Georg Wackenhut Material Testing Institute, University of Stuttgart (MPa)