Performance Demonstration Based Probability of Detection (POD) Curves for Fatigue Cracks in Piping

This paper evaluates non-destructive examination (NDE) detection capabilities for fatigue cracks in piping. Industry performance demonstration initiative (PDI) data for fatigue crack detection were used to develop a matrix of statistically based probability of detection (POD) curves that consider various NDE performance factors. Seven primary performance factors were identified — Material, Crack Geometry/Type, NDE Examination Access, NDE Procedure, Examiner Qualification, Pipe Diameter, and Pipe Wall Thickness. A database of 16,181 NDE performance observations, with 18 fields associated with each observation, was created and used to develop statistically based POD curves for 42 stainless steel and 14 carbon steel performance cases. Subsequent comparisons of the POD fits for each of the cases showed that excellent NDE performance for fatigue cracks can be expected for ferritic materials. Very little difference was observed between the POD curves for the 14 carbon steel performance cases considered in this study and NDE performance could therefore be represented by a single POD curve. For stainless steel, very good performance can also be expected for circumferential cracks located on the same side of the weld from which the NDE examination is made. POD depended primarily on component thickness. Three POD curves for stainless steel were prepared. Best estimate and the associated 95% confidence bounds for POD versus through-wall depth logistic regression digital data are provided. Probabilistic fracture mechanics (PFM) calculations were performed to compare best estimate leak probabilities obtained from both the new performance-based POD curves and previous PFM models. This work was performed under joint funding by EPRI and the U.S. Department of Energy (DOE), Office of Nuclear Energy Science and Technology’s Nuclear Energy Plant Optimization (NEPO) program.