Feature-based imaging system: The Peach Bottom field trials

Feature-based systems that combine imaging and signal analysis capabilities may be useful for nondestructive evaluation (NDE) of plant components. This report describes the field evaluation of an integrated system at a plant site during a scheduled outage for pipe weld examination to discriminate intergranular stress corrosion cracking (IGCSS) from benign, geometrical reflectors. The integrated system consisted of a personal computer (PC)-based system capable of detailed analysis of ultrasonic signal data and an in-service inspection (ISI) imaging system used in many commercial pipe examinations for IGSCC. The integrated system was trained to discriminate automatically IGSCC from other reflectors using EPRI NDE Center's inventory of field-removed samples. The methods and results of this training are described. Several classifiers were synthesized using mathematical features derived from signals that were collected to simulate field conditions. Data collection procedures were developed that required minimal operator training. Field data were collected and analyzed before and after pipe decontamination prior to pipe replacement. Automatic decision maps were generated for easier interpretation and comparison. The field trial was conducted in October 1987--January 1988. Select pipe specimens were decontaminated and shipped to a contractor's facility for detailed metallurgical and dye penetrant analysis. A comparison between reflector calls from the ultrasonic data and penetrant test results are provided in the final section. As a basis for comparison, the performance of the automated system was compared with manual calls. The automated technique results were better than manual; although both were well below acceptable standards as defined for IGSCC qualification. 2 refs., 174 figs., 10 tabs