Research and Application of Regularized Sparse Filtering Model for Intelligent Fault Diagnosis Under Large Speed Fluctuation

Abstract

The speed of mechanical rotating parts often fluctuates during the working process. Vibration signals collected under constant speed have a strong correlation with the corresponding fault types. However, the mapping relationship becomes complex under large speed fluctuation, which is an urgent research subject in intelligent fault diagnosis. As an effective unsupervised learning method, sparse filtering (SF) has been successfully used in intelligent fault diagnosis. However, the generalization capability of this method to deal with large speed fluctuation remains poor. To overcome this deficiency, this study adds regularization to the loss function of SF to obtain regularized SF methods. The frequency domain signals under large speed fluctuation are directly input to regularized SF for feature extraction, and softmax regression is used as a classifier for fault type identification. Experimental results of gearbox and bearing datasets show that L1/2 regularized sparse filtering (L1/2-SF) model can solve the problem of large speed fluctuation more effectively than other regularized SF models can.

Keywords

• Rotating parts
• fault diagnosis
• large speed fluctuation
• sparse filtering
• regularization