Integrated Statistical Processing of NDE Signals for Class Discrimination

Abstract

An integrated methodology for the analysis of experimental NDE signals for the classification of test-objects according to their mechanical characteristics is presented. The methodology consists of several techniques, adopted from various multi-disciplinarial fields, and incorporates concepts from the theory of information, pattern recognition techniques, cluster analysis, and data compression. Statistical tests are performed to verify the significance of the differences between the signals obtained from different groups and to estimate the confidence level associated with the classification procedure. The techniques are outlined and discussed, and experimental examples are presented to demonstrate their successful application.