Variable predictive model based class discriminate application in gearing fault diagnosis

Variable predictive model based class discriminate (VPMCD) is present recently as a new approach of pattern recognition. VPMCD hypothesizes that the values of feature vectors have linear of non-linear inter-relations. Based on this inter-relations, mathematics model can be constructed and the feature vectors can be classified through training and testing the models founded. In the paper, a new gear fault diagnosis approach based on VPMCD, EMD and singular value decomposition (SVD) is proposed. Namely, firstly, by using EMD the gear vibration signal can be decomposed into several IMFs (intrinsic mode function, IMF). Then the singular values of the feature vector matrix consisting of the first several IMFs, which contain main fault information are calculated. After that the singular values are taken as feature vector, based on which, an VPMCD-classifier is created to distinguish the gear's work status and fault categories. Finally the proposed approach is also applied to experimental data, and the analysis results show that the proposed approach is feasible and can fulfill the classification of gear fault categories effectively.