Real-time monitoring of bearings is of great significance to the safety and reliability of rotating machinery operation.Existing research focuses on the extraction of bearing fault feature frequency,which is limited by the demodulation spectrum resolution and sampling time,and cannot determine the type of bearing faults in a more real-time manner.In order to realize the effective recognition of bearing fault state,the expansion strategy of enhancing dictionary completeness and sparse coefficient sparsity is proposed,and the adaptive feature vector extraction is established based on dictionary learning,which can recognize four kinds of bearing faults under different rotational speeds and mixed loads,and the results show that only a small number of samples are needed(500 samples,250 samples),and the classification accuracy can be realized with a higher accuracy of 90％or more.
维普
万方数据