Noise emission analysis a way for early detection and classification faults in rotating machines

Nowadays, rotating machines (RM) plays an important role in industry. Therefore detection a precise faults in main part of RM will avoid non programmed stops of production by real time management of machine status. Different detection methods using vibration signature analysis, noise signature analysis and lubricant signature analysis were presented in literature reviews however there is no much techniques using bio-inspired features. In this work acoustic signal analysis and processing based on speech recognition techniques were used to detect early faults in REB namely: faults in rolling ball, in inner race, outer race and protecting cage. Commonly used Speech recognition features were selected, also two classifiers, used in ASR, were tested Euclidian distance and K-NN method, the overage results obtained using combination of features and ED is 92% while the results are improved using the K-NN methods to the average of 94%.