Research On Blind Separation And Intelligent Diagnosis Of Rolling Bearing Fault

With the progress of science and the development of industrial modernization,more and more large-scale mechanical equipment in production,automation,more and more sophisticated equipment parts,the stable operation of mechanical equipment and fault feedback play a more important role.Rolling bearings are widely used in rotating machinery and are typical and representative.Their operating conditions directly affect the performance of the whole system.In-depth study on condition monitoring and fault diagnosis of rolling bearings(CMFD)can effectively improve the production efficiency of machinery and reduce accidents The occurrence rate is of great academic significance and engineering application value.In this paper,the vibration signal of rolling bearing as the research object,the work done as follows:(1)For the bearing running multiple failures occur frequently in many failing to separate the more complex scenario,the less the case bearing composite failure research,an approach based on frequency domain bearing signal due to the blind source separation research methods.The use of the bearing signal frequency domain has a sparse,using frequency domain sparse component analysis of bearing composite failure to disengage.The bearing simulation experiments,the signal for compound the problem of separation,separation from the source signal is similar to a high degree,proof that the program can be more precise and bearing composite failure to disengage.(2)Bearing signal in the frequency domain that has a weak and sparse,resulting in the mixing matrix estimate accuracy is poor,in response to this problem,based on the frequency domain complex point-source-point detection method and cell density detection algorithm,through a combination of improved mixing matrix estimate accuracy.The actual bearing signal simulation experiments,the normalized mean square error smaller mixing matrix estimate that the program can accurately improve the mixing matrix estimate accuracy,and be able to accurately to composite failure to disengage.(3)Commonly used for this stage of the smart diagnostic methods will still need to rely on traditional signal processing technology to extract a feature limitations,this chapter introduces the depth study,based on the depth of the trusted network of CMFD program by Case Western Reserve University bearing data set for experiments to be better recognition results,and also studied the different sample length and different iterations of the deep conviction of network diagnostic programs,the program greater awareness,and finally,in conjunction with other diagnostic methods are compared,the Program's recognition accuracy has improved to verify the depth of belief network troubleshooting scenarios.