Integration of NDI Reliability into FAD-Based Probabilistic Analysis of Pressure Vessel Integrity

The paper presents a methodology that aims towards the integration of nondestructive inspection (NDI) quality as quantified by the probability of detection (POD), into a probabilistic failure assessment diagram (FAD) for analysis of pressure vessel integrity. The first part of the paper introduces the common models for POD estimation (e.g., Marshall, log-odds) along with a new POD model developed by the authors. The FAD procedure coupled with POD models is developed into a computer code called FADSIM. The statistic of input includes flaw size and material characteristics such as fracture toughness, UTS, and yield strength. The developed rationale enables the estimation of the gain in reliability against fracture failure when applying a specific NDI procedure characterized by ist POD - flaw size dependence in the context of deterministic or statistically variable FAD input parameters. Extended probabilistic sensitivity analyses are thus enabled. Along this line, the accuracy in the underlying statistical sampling simulation methodology was checked against a compact analytic probabilistic model. Acceptable accuracy was demonstrated for a number of simulations (10 (exponent5) and greater) that can be performed in a reasonable time with current computing hardware (PC). For the purpose of illustration, representative parametric results have been produced for the case of a cylindrical pressure vessel model having anaxial or transverse semi-elliptic crack at the inner surface. The beneficial role of NDI has been quantified in terms of the decrease of failure probability as function of applied stress or a FAD defined safety index.