Optimal measurement points evaluation for friction plate via a comprehensive analysis of correlation and clustering

As an important component of the transmission chain, it is of great significance to monitor the running status of the friction plate in real time. Aiming at the problems of high cost of friction plate fault diagnosis, difficult fault signal acquisition, mutual coupling of vibration signals and the ambiguous selection and evaluation rules of the measuring points, an optimal measurement points evaluation for friction plate based on a comprehensive analysis of correlation and clustering is proposed. Firstly, wavelet packet decomposition (WPT) is processed on the vibration signals to construct a WPE flow image feature for an energy flow distribution, where the unusual characteristics would be described in a clear way. Secondly, considering the ability of correlation and clustering, a comprehensive index, named as the redundant and clustering index (RC-index), is calculated for each measuring point, then the measuring points are sorted according to RC-index. The final optimal measurement points sets are evaluated and built via the identification accuracy based on k-nearest neighbor (KNN) classifier. Experiments on the actual friction plate test bench are analyzed to indicate that the optimal measurement points set can maintain relatively high fault type identification accuracy with fewer measuring points.