Fault detection and classification in cantilever beams through vibration signal analysis and higher-order statistics

Abstract

A method for detecting and classifying faults in an aluminum cantilever beam is proposed in this paper. The method uses features based on second-, third- and fourth-order statistics, which are extracted from the vibration signals generated by the cantilever beam. Fisher’s discriminant ratio (FDR) is used for feature selection, and an artificial neural network is used for fault detection and classification. Three different degrees of faults (low, medium and high) were applied to the cantilever beam, and the proposed pattern recognition system was able to classify the faults, reaching performances ranging from 88 to 100 %. Moreover, the use of higher-order statistics-based features combined with FDR led to a compact feature space and provided satisfactory results.