Acoustic Emission and Feature Selection based on Sensitivity Analysis

Abstract

An alternative methodology based on the sensitivity analysis of pre-trained BP-networks is analysed. The power of this new approach to feature selection is demonstrated on the classification of simulated AE-data.