An AB-CNN intelligent fault location recognition model for induction motor

Abstract

Aiming at the problems of complex structure, fault prone and low accuracy of fault identification of asynchronous motor, this paper proposed a fault location recognition model of asynchronous motor based on AB-CNN (AdaBelief-CNN) intelligent algorithm. Taking asynchronous motor as the application background, an experimental platform is built to collect different types of fault data of asynchronous motor. Firstly, the dimension of the data is reduced by principal component analysis, the convolutional neural network (CNN) is used to carry out forward propagation fault recognition training for asynchronous motor fault data, and output the trained data prediction value. Secondly, through the loss function and AdaBelief optimization algorithm, the asynchronous motor fault location recognition model is optimized. Finally, the accuracy of the fault location identification model is tested with the test data. Comparing with the original CNN network identification model and Adam optimized CNN network (AD-CNN), the proposed solution outperforms other schemes by about 10%.