Fault Diagnosis of Conventional Circuit Breaker Accessories Based on Grayscale Image of Current Signal and Improved ZFNet-DRN

Abstract

Considering that the randomness of the closing phase of the coil circuit for a low-voltage conventional circuit breaker (CCB) will lead to the difference of current signal under the same operation, a fault diagnosis method of switching accessories based on the grayscale image of the current signal and deep residual network (DRN) improved by Zeiler and Fergus Net (ZFNet-DRN) is proposed in this article. First, the signal stacking method is employed to convert the original 1-D current signal into a 2-D grayscale image, so as to retain the information contained in the signal to the greatest extent and avoid the insufficient amount of data input to the subsequent network model due to the limitation of 1-D data length. Second, regarding the diagnosis model, the improved ZFNet network realizes the deep extraction of features, DRN tackles the issue that the recognition rate decreases with the increase in the network depth, and the parametric rectified linear unit (PReLU) activation function and AMSGrad optimization algorithm are introduced to enhance its performance. Finally, the experimental analysis demonstrates that the diagnosis model can overcome the influence of the randomness of the closing phase and effectively complete the fault classification.