Three Approaches to Quantification of NDE Uncertainty and a Detailed Exposition of the Expert Panel Approach Using the Sheffield Elicitation Framework

The ASME Boiler & Pressure Vessel Code Section XI Committee is currently developing a new Division 2 nuclear code entitled the “Reliability and Integrity Management (RIM) program,” with which one is able to arrive at a risk-informed, NDE-based engineering maintenance decision by estimating and managing all uncertainties for the entire life cycle including design, material selection, degradation processes, operation and non-destructive examination (NDE). This paper focuses on the uncertainty of the NDE methods employed for preservice and inservice inspections due to a large number of factors such as the NDE equipment type and age, the operator’s level and years of experience, the angle of probe, the flaw type, etc. In this paper, we describe three approaches with which uncertainty in NDE-risk-informed decision making can be quantified: (1) A regression model approach in analyzing round-robin experimental data such as the 1981–82 Piping Inspection Round Robin (PIRR), the 1986 Mini-Round Robin (MRR) on intergranular stress corrosion cracking (IGSCC) detection and sizing, and the 1989–90 international Programme for the Inspection of Steel Components III-Austenitic Steel Testing (PISC-AST). (2) A statistical design of experiments approach. (3) An expert knowledge elicitation approach. Based on a 2003 Pacific Northwest National Laboratory (PNNL) report by Heasler and Doctor (NUREG/CR-6795), we observe that the first approach utilized round robin studies that gave NDE uncertainty information on the state of the art of the NDE technology employed from the early 1980s to the early 1990s. This approach is very time-consuming and expensive to implement. The second approach is based on a design-of-experiments (DEX) of eight field inspection exercises for finding the length of a subsurface crack in a pressure vessel head using ultrasonic testing (UT), where five factors (operator’s service experience, UT machine age, cable length, probe angle, and plastic shim thickness), were chosen to quantify the sizing uncertainty of the UT method. The DEX approach is also time-consuming and costly, but has the advantage that it can be tailored to a specific defect-detection and defect-sizing problem. The third approach using an expert panel is the most efficient and least costly approach. Using the crack length results of the second approach, we introduce in this paper how the expert panel approach can be implemented with the application of a software package named the Sheffield Elicitation Framework (SHELF). The crack length estimation with uncertainty results of the three approaches are compared and discussed. Significance and limitations of the three uncertainty quantification approaches to risk assessment of NDE-based engineering decisions are presented and discussed.