Research On Bearing Compound Fault Diagnosis Based On Blind Source Separation

In the 1990s,a technology called blind source separation began to emerge,and it began to develop rapidly in the field of signal analysis and processing.Because of the excellent adaptability of blind source separation in vibration signal analysis,this article focuses on the composite faults of rolling bearings in mechanical equipment.Under underdetermined conditions,the method of composite fault blind source separation based on the combination of EEMD preprocessing and CICA technology is studied.The feasibility is verified in the simulation experiment.Finally,it was successfully applied to the separation of the mixed fault signal of the single channel of the rolling bearing of the experimental platform.This article mainly contains the following items:First,the paper analyzes and studies the vibration mechanism of rolling bearings in mechanical equipment and the characteristic forms of rolling bearing failures.The ICA algorithm and Fast ICA algorithm are analyzed.Later,in order to verify that the Fast ICA algorithm can play a blind source separation effect on the signal,a simulation experiment was designed.Secondly,the preprocessing of the original signal under the under-determined condition is essentially for the purpose of extending the channel and reducing signal noise.The results of the simulation comparison test show that compared with the EMD algorithm,the IMF component generated by the EEMD algorithm decomposition signal has a better anti-aliasing effect.The single-channel signal is preprocessed by the EEMD algorithm,and the selection criteria of the IMF components obtained by the EEMD decomposition are studied emphatically.The screening criteria based on the combination of kurtosis value and cross-correlation coefficient can improve the accuracy of screening IMF components;then,the number of original signal sources is estimated based on the singular value decomposition algorithm to determine whether the number of IMFs screened is appropriate.The IMF components containing fault information are screened,and the components containing false information are eliminated to ensure the smooth progress of the subsequent blind source separation algorithm.Simulation experiments are used to verify the feasibility of blind source separation under a single channel.The test results show that EEMD combined with Fast ICA algorithm can effectively realize the blind source separation of rolling bearing composite faults.Finally,this paper draws on the advantages of EEMD and CICA algorithms in the field of single-channel hybrid fault diagnosis under underdetermined conditions,and proposes an algorithm that combines EEMD and CICA,and then verifies it through simulation experiments.The experimental results show that the method is feasible in the separation and extraction of single-channel composite fault signals.Later,through the experimental analysis of the measured data of the rolling bearing composite fault of the power transmission fault comprehensive test bench,it is concluded that the algorithm based on the combination of EEMD and CICA is feasible and effective in the blind source separation of single-channel composite faults under underdetermined conditions.Analyzing the experimental results,through the comparison of Fast ICA and CICA two algorithms in the experiment,both of these two algorithms can meet the requirements of blind source separation of rolling bearing composite faults.As an improved algorithm of Fast ICA,CICA can extract the target fault signal of the bearing by using the known information of the rolling bearing.It is verified through experiments that the combination of EEMD and CICA can effectively realize single-channel signal compound fault diagnosis,which has certain practical value.