Adaptive TQWT filter based feature extraction method and its application to detection of repetitive transients

作     者：KONG Yun WANG TianYang CHU FuLei 

作者机构：Department of Mechanical Engineering Tsinghua University 

出 版 物：《Science China(Technological Sciences)》 (中国科学（技术科学英文版）)

年 卷 期：2018年第61卷第10期

页      面：1556-1574页

核心收录：

学科分类：0810[工学-信息与通信工程] 08[工学] 0805[工学-材料科学与工程（可授工学、理学学位）] 0802[工学-机械工程] 0702[理学-物理学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 080201[工学-机械制造及其自动化] 

基　　金：supported by the National Natural Science Foundation of China (Grant Nos. 51335006 & 51605244) 

主　　题：tunable Q-factor wavelet transform parameter selection energy-weighted normalized wavelet entropy energy to bandwidth ratio transient detection fault diagnosis 

摘      要：The local defect in rotating machine always gives rise to repetitive transients in the collected vibration signal. However, the transient signature is prone to be contaminated by strong background noises, thus it is a challenging task to detect the weak transients for machine fault diagnosis. In this paper, a novel adaptive tunable Q-factor wavelet transform(TQWT) filter based feature extraction method is proposed to detect repetitive transients. The emerging TQWT possesses distinct advantages over the classical constant-Q wavelet transforms, whose Q-factor can be tuned to match the oscillatory behavior of different signals, but the parameter selection for TQWT heavily relies on prior knowledge. Within our adaptive TQWT filter algorithm, the automatic optimization techniques for three TQWT parameters are implemented to achieve an optimal TQWT basis that matches the transient components. Specifically, the decomposition level is selected according to a center frequency ratio based stopping criterion, and the Q-factor and redundancy are optimized based on the minimum energy-weighted normalized wavelet ***, the adaptive TQWT decomposition can be achieved in a sparse way and result in subband signals at various wavelet ***, the optimum subband signal which carries transient feature information, is identified using a normalized energy to bandwidth ratio index. Finally, the single branch reconstruction signal from the optimum subband is obtained with transient signatures via inverse TQWT, and the frequency of repetitive transients is detected using Hilbert envelope demodulation. It has been verified via numerical simulation that the proposed adaptive TQWT filter based feature extraction method can adaptively select TQWT parameters and the optimum subband for repetitive transient detection without prior knowledge. The proposed method is also applied to faulty bearing vibration signals and its effectiveness is validated.