Feature Extraction of Machinery Fault Based on Lifting Wavelet Package Transform and Genetic Algorithm
In order to improve classification precision for different stating of mechanical equipments, a novel method is proposed. Firstly, a biorthogonal wavelet with vibration signals property is constructed via lifting scheme, and the effective features of fault signals are formed by lifting wavelet package transform (LWPT). Secondly, binary system coding scheme is presented. With the distance evaluation criterion, the optimal features are obtained from the statistical characteristics of original signals and envelope spectrum analysis of wavelet package coefficients （WPCs ） of the most sensitive frequency-band by genetic algorithm. The classification precision of different feature combination is implemented by support vector machines (SVMs). Finally, the experiment of feature selection about fault signals has been performed on Bently rotor experimental platform. Testing results show that the proposed method of classification and recognition can effectively find out the best feature subset, and improve the precision and effectiveness of classification on fault signals accurately. Sound foundation of mechanic equipments intelligent fault diagnosis system exploitation has been set up.
Lifting wavelet package transform genetic algorithm(GA) Support vector machines fault diagnose feature selection
JI Xinwei YANG Xiaoqiang HAN Jun ZHOU Chunhua LIU Ying
Engineering Institute of Corps of Engineers, PLA Univ. of Sci. & Tech. , Nanjing, 210007
国际会议
北京
英文
2007-08-05（万方平台首次上网日期，不代表论文的发表时间）