This paper discusses round-robin exercises using the NRC steam generator (SG) mock-up at Argonne National Laboratory to assess inspection reliability. The purpose of the round robins was to assess the current reliability of SG tubing inservice inspection, determine the probability of detection (POD) as function of flaw size or severity, and assess the capability for sizing of flaws. For the round robin and subsequent evaluation completed in 2001, eleven teams participated. Bobbin and rotating coil mock-up data collected by qualified industry personnel were evaluated. The mock-up contains hundreds of cracks and simulations of artifacts such as corrosion deposits and tube support plates that make detection and characterization of cracks more difficult in operating steam generators than in most laboratory situations. An expert Task Group from industry, Argonne National Laboratory, and the NRC have reviewed the signals from the laboratory-grown cracks used in the mock-up to ensure that they provide reasonable simulations of those obtained in the field. The mock-up contains 400 tube openings. Each tube contains nine 22.2-mm (7/8-in.) diameter, 30.5-cm (1-ft) long, Alloy 600 test sections. The flaws are located in the tube sheet near the roll transition zone (RTZ), in the tube support plate (TSP), and in the freespan. The flaws are primarily intergranular stress corrosion cracks (axial and circumferential, ID and OD) though intergranular attack (IGA) wear and fatigue cracks are also present, as well as cracks in dents. In addition to the simulated tube sheet and TSP the mock-up has simulated sludge and magnetite deposits. A multiparameter eddy current algorithm, validated for mock-up flaws, provided a detailed isometric plot for every flaw and was used to establish the reference state of defects in the mock-up. The detection results for the 11 teams were used to develop POD curves as a function of maximum depth, voltage and the parameter mp , for the various types of flaws. The POD curves were represented as linear logistic curves, and the curve parameters were determined by the method of Maximum Likelihood. The effect of both statistical uncertainties inherent in sampling from distributions and the uncertainties due to errors in the estimates of maximum depth and mp was investigated. The 95% one-sided confidence limits (OSL), which include errors in maximum depth estimates, are presented along with the POD curves. For the second round robin a reconfigured mock-up is being used to evaluate the effectiveness of eddy current array probes. The primary emphasis is on the X-Probe. Progress with the X-Probe round robin is discussed in this paper.
Probability of detection models for eddy current NDE methods
