Mechanical fault diagnosis based on redundant second generation wavelet packet transform, neighborhood rough set and support vector machine

This paper investigates the application of the redundant second generation wavelet package transform (RSGWPT), neighborhood rough set (NRS) and support vector machine (SVM) on faulty detection, attribute reduction and pattern classification. On this basis, a novel method for mechanical faulty diagnosis based on RSGWPT, NRS and SVM is presented, which utilizes the RSGWPT to extract faulty feature parameters from the statistical characteristics of wavelet package coefficients to constitute feature vectors, and then makes the attribute reduction by NRS method to obtain the key features, lastly these key features are input into SVM to accomplish faulty pattern classification. The experimental results of the proposed method to fault diagnosis of the gearbox and gasoline engine valve trains show that this method can extract the faulty features, which have better classification ability and at the same time reduce a lot of redundant features in case of assuring the classification accuracy, accordingly improve the classifier efficiency and achieve a better classification performance.

► A new machinery fault diagnosis method based on RSGWPT, NRS and SVM is proposed.
► This approach can select an optimal subset of features from a large database with a lot of features.
► Proposed technique is evaluated by fault diagnosis of gearbox and gasoline engine valve trains.
► This method demonstrates superior diagnosis capability for both rotating and reciprocating machinery.