An automated diagnostics system for eddy current analysis using applied artificial intelligence methods

The purpose of this dissertation research is to develop a diagnostic expert system that integrates database management methods, digital signal processing, artificial neural networks, expert system and fuzzy logic techniques for the automation of steam generator eddy current test (ECT) data analysis. The following key tasks were identified and developed for establishing a robust analysis system: (1) digital eddy current test data calibration, compression, and representation, (2) noise compensation, (3) development of robust neural networks with a low probability of defect misclassification and defect parameter estimation. (4) decision making for flaw detection using fuzzy logic. (5) development of an expert system for database management, compilation of a trained neural network library, and a decision module, and (6) performance evaluation of the integrated approach using eddy current test data. An automated diagnostics system using NDE data is needed because of the necessity to process a large amount of information. and because of the limitations of human processing capability. Different forms of eddy current inspection data were acquired from the EPRI NDE Center and from the Metals and Ceramics Division of ORNL

The integrated approach for automating the analysis of eddy current test data for steam generator tubing diagnosis is an original contribution of this research. The successful implementation of the methodology requires proper data compression, data calibration, data management, fuzzy logic flaw detection and flaw parameters estimation. There are no defined approaches to accomplish this task. The fuzzy flaw detection system, developed in this research, is the first to utilize information from multi-frequency eddy current data for flaw detection. The database management approach developed in this research, is also a unique contribution that would help pave the way for commercial on-line implementation of this nondestructive evaluation technique. A PC WINDOWS- based expert system called EDDYAI was developed using Microsoft Visual C++. This system integrates all the techniques developed in this research project into a user-friendly expert system for automated steam generator multi-frequency eddy current test data analysis.