A knowledge-based automatic nondestructive test for CIMS using the eddy-current method

This dissertation presents a heuristic pattern recognition signal processing method to automate defect detection and classification processes. Testing is performed by the eddy current nondestructive testing technique. Eddy current testing consists of measuring impedance changes in a test coil caused by eddy currents induced in an electrically conductive material.;The methodology presented in this dissertation utilizes an expert's knowledge in interpreting the impedance plane plots. Since this approach utilizes an expert's cognitive processes in impedance plane interpretation, its problem domain centers on representing the expert's knowledge. This is achieved through the knowledge engineering process.;The knowledge engineering process involves two steps. One is detecting defects; the other is classifying defect types. The pseudo run-length test is performed to check the signal data randomness to detect defects. Then the classification process begins. Each defect type's signal features are extracted based on experiments to classify the defect types. The rules for both the defect detection and classification process are developed based on the feature extraction results.;This approach is less difficult and expensive in developing a pattern recognition signal processing method for eddy current analysis in comparison to existing signal processing methods.