Fault Feature Extraction Of Flexible Thin-walled Bearings Based On CEEMD And LRSD-TNNSR

Flexible thin-wall bearing is the most important component of harmonic reducer in robot.the life of flexible thin-wall bearing determines the life of harmonic reducer,so its major or minor faults will have a fatal impact on the precise transmission of the reducer.In order to promote the development of industrial robot or its core part harmonic reducer,it is very important to carry out the research of vibration signal fault feature extraction of flexible thinwall bearing.In this paper,takes the flexible thin-wall bearing as the research object.According to the characteristics of the vibration signal of the flexible thin-walled bearing with low rank,large amount of noise and the periodic impact component with high amplitude produced by the alternation of long axis and short axis,a fault feature extraction method based on CEEMD and LRSD-TNNSR is proposed.Firstly,the principle of CEEMD algorithm is introduced,and the simulation results show that CEEMD algorithm has the advantage of denoising compared with EMD and EEMD algorithm.Then it introduces the principle and advantages of the innovative algorithm LRSDTNNSR.This algorithm is an innovation and improvement of the original LRSD algorithm.It uses truncated kernel norm and improves the sparsity of low-rank components of real data to optimize the LRSD objective function.At the same time,a two-stage iterative optimization method is proposed to solve the optimization problem.Compared with the original LRSD algorithm,it can recover the low-rank components more accurately and reduce the difficulty of application.Finally,the bearing experiment is carried out,and the CEEMD algorithm is used for preprocessing.The redundant and irrelevant parts of the original signal which are not related to the fault impact are filtered out,and the amplitude of fault frequency signal is amplified at the same time.And then the LRSD-TNNSR algorithm is used for further processing to accurately extract the low-rank outer ring fault frequency of 128.95 Hz and inner ring fault frequency of 190.2Hz of flexible thin-wall bearing at the speed of 750 rpm and 900 rpm.Combined with the advantages of the two algorithms,a method based on CEEMD and LRSDTNNSR is proposed.Through the comparative tests of four groups in bearing vibration signal extraction,it is proved that the method based on CEEMD and LRSD-TNNSR is more effective and the fault feature signals are more prominent than the LRSD-TNNSR algorithm.