Approach to Denoising of Local Adaptive Algorithm Based on DT-CWT and Its Application to Auto Transmission Fault Diagnosis

Abstract

To extract weak fault information in auto transmission gear, a denoising approach is proposed using local adaptive algorithm based on complex wavelet. It takes advantage of shift invariance of dual-tree complex wavelet transform, non-Gaussian probability distribution for the wavelet coefficients and the statistical dependencies between a coefficient and its parent. So it can obtain higher SNR (signal-to-noise ratio) than common methods based on discrete wavelet transform. Some actual auto transmission vibration signals are analyzed, and the results show the effectiveness of the proposed method in monitoring and diagnosis of machine conditions, with the capability of early fault detection by extracting periodic impulses from auto transmission vibration signals.