Gear fault recognition based on recurrence quantification analysis and Gaussian mixture model

In order to overcome the shortcoming of recurrence plot that can only supply the qualitative analysis to signals,the recurrence quantification analysis is used to analysis different fault modes gear's vibration signal.The gear fault pattern recognition method that combined the gaussian mixture model with the feature vector that consists of determinism and laminarity is proposed.Based on the signals that acquired form gear fault experiment table,the proposed method is compared with RBF artificial neural network classification method by Re-substitution test,Jackknife test and Independent data set test respectively.The classification results show that the higher discrimination can be achieved by the proposed method.