Fault diagnosis of rotating machinery based on auto-associative neural networks and wavelet transforms

This paper presents a new technique for monitoring the condition of rotating machinery from vibration analyses. The proposed method combines the capability of wavelet transform (WT) to treat transient signals with the ability of auto-associative neural networks to extract features of data sets in an unsupervised mode. Trained and configured networks with WT coefficients of nonfaulty signals are used as a method to detect the novelties or anomalies of faulty signals. The effectiveness of the proposed technique is evaluated using the numerical data and experimental vibration data of a gearbox. Despite the fact that noise is present in both cases, results demonstrated that the proposed method is a good candidate to be used as an online diagnosis tool for rotating machinery.