Missing Data Bearing Fault Diagnosis Based on SVT-1DCNN

Missing data is a common phenomenon in induction motor fault datasets, which could lead to a loss of statistical ability and/or biased results for fault diagnosis. In this paper, we address the rolling bearing fault diagnosis in the presence of missing data, where a bearing fault diagnosis method based on singular value thresholding (SVT) low-rank matrix filling and one-dimension convolutional neural network (1DCNN) is proposed. Specifically, the SVT low-rank matrix filling is first used to recover the missing fault bearing data, and then 1DCNN is adopted to extract the fault feature and classify each fault automatically. The experimental results show that the classification accuracy of the proposed method can be promoted as high as 96.38%, which is better than that of other methods.