基于多尺度图域特征的轴承故障诊断方法OA

Bearing possesses important functions of load transfer,support and positioning,and is the key components of common mechanical equipment.Its health directly affects the reliability and performance of the equipment,so its monitor-ing and diagnosis is of great significance.Due to the complex operating conditions and strong background noise of bearings,the accuracy of conventional fault diagnosis methods is low,and misdiagnosis is easy to occur.In this paper,a bearing fault diagnosis method based on multi-scale graph feature was proposed.Firstly,the transmission relationship of bearing vibration signal was analyzed,the transmission relationship was quantified as visual edge,and the visual edge was optimized by using the filtering concept to construct the graph signal.Then,the multi-scale spectral graph wavelet transform was used to decom-pose the graph signal into several layers,and the dynamic entropy and spectral amplitude entropy of different layers were ex-tracted respectively.Combined with covariance,the features of different layers were screened,and then the feature space was constructed.Finally,the fault recognition of bearing was realized based on the Mahalanobis Distance similarity of multi-scale graph features.The bearing fault dataset was employed to verify this method.The results show that the proposed meth-od can effectively identify different bearing faults,and the recognition accuracy is much better than that of the traditional time-domain and frequency-domain features method,and has better robustness.