Research On Blind Source Separation Algorithm For Fault Diagnosis Of Rotating Machinery Under Underdetermined Measurement

As an effective method of signal separation and extraction,blind source separation is widely used in the fault diagnosis of traditional machinery.In the general blind separation model,the number of signal receivers is usually not less than the number of sources.But in fact,due to the problems of cost and monitoring environment,it may not be able to install multiple sensors at the same time,and sometimes even in the process of mechanical equipment operation,only a single channel can be used to monitor the equipment,that is,the so-called underdetermined measurement.At the same time,the characteristics of weak fault signal are often difficult to extract.All of these restrict the application of blind source separation in rotating machinery fault diagnosis.In this paper,the blind source separation theory is applied to the research of underdetermined measurement(i.e.underdetermined blind source separation)and source number estimation,and the fault feature extraction under low SNR and underdetermined conditions is realized,and the algorithm is successfully applied to the fault diagnosis of rotating machinery.The details are as follows:(1)This paper summarizes the research status of blind source separation and rotating machinery fault diagnosis,analyzes the research results and existing problems of other scholars,and studies the core theoretical framework of blind source separation.(2)The estimation of the source number involved in blind source separation is studied.For the low signal-to-noise ratio(SNR)conditions,the traditional eigenvalue decomposition method is optimized by using the Toeplitz matrix.The simulation results show that the new algorithm can improve the estimation accuracy of the source number of noncoherent and coherent sources at low SNR.(3)The underdetermined blind source separation method is studied.For one dimensional observation signal,the underdetermined blind source separation method is optimized by using EEMD,and the correlation coefficient of the source signal and the source signal is used as the evaluation standard.Compared with the traditional underdetermined blind source separation method,the optimization algorithm improves the restoration accuracy of the source signal.(4)For underdetermined rotating machinery fault diagnosis,the underdetermined blind source separation algorithm is optimized by using Hankel matrix,and the multi-dimensional signal is constructed by Hankel matrix to realize blind source separation.The simulation results show that the new algorithm is more effective for extracting weak fault features.(5)This paper studies the fault mechanism of the rotor and typical signal characteristics,and build the rotor experimental platform.The single channel fault signal is analyzed and diagnosed by using the fault diagnosis algorithm proposed in this paper,the experiment result shows that the new algorithm can be more effective to diagnose the fault.In this paper,the underdetermined blind source separation method is applied to fault diagnosis through the combination of theory and experiment,and the method of source number estimation and underdetermined blind source separation are completed.The research provides a reference for the diagnosis and analysis of rotating machinery fault signals.