Fault Recognition of Induction Motor Based on Convolutional Neural Network Using Stator Current Signal

An induction motor is an integrated electrical device that includes plenty of systems and consequently exhibits a diversity of faults. This chapter proposes a fault identification method based on convolutional neural network (CNN) to overcome the complexity of feature extraction and feature selection, which require much experience and professional knowledge. The stator current signal is obtained and converted into time domain image for CNN inputs. The CNN model is established with different size of convolution kernel, different number of convolutional layers, different value of learning rate, and different batch size to train the network. The influence of different parameters on the performance of the network model is systematically analyzed and an optimal model is obtained to accurately identify induction motor fault. Compared with support vector machine and artificial neuron network, the results illustrate that the method is more efficient and the recognition accuracy is higher.