Advances in automatic detection of failures in electric machines using audio signals

Abstract

in this paper nonlinear chaotic features have been obtained from audio signals of different kinds of electric machines as a first step in order to develop a personal computer (PC) based artificial intelligence system for the fault identification and diagnosis of electric machines. These techniques can be applied in fault identification and diagnosis in industrial scenarios by mean of expert systems. Different nonlinear features (based on chaos theory) to detect changes of the audio signal were studied: maximal Lyapunov exponent, correlation dimension and correlation entropy. We also studied related measurement such as the time delay and the value of the first minimum of the mutual information function, the first zero of the autocorrelation function and Shannon entropy. We used different recordings from different scenarios (PC fans, an iron cutter and an electric drill).