Research on Classification Method of Fault Pictures of Pump Equipment Based on FCDNN

Abstract

Accurate classification of equipment faults is one of the effective ways to improve the efficiency of fault analysis and maintenance. How to establish an effective classification model of equipment fault has become a hot topic of current research. The most existing deep neural network (DNN)-based fault classification methods only focus on prediction accuracy without considering the limitation of the size of labeled samples. In this paper, a fault classification deep neural network model (FCDNN) based on the Siamese network architecture is proposed. The model uses the similarity of samples to classify samples. It mainly includes a feature learning module and a metric learning module. By optimizing the feature learning process and combining the metric learning the Siamese network adapts pretrained image models to specific classification tasks. The feature learning module uses a double deep convolutional neural network to extract features, and the metric learning module uses the similarity between the test sample and the sample pair of each labeled sample to complete the classification task. The comparison studies with other models demonstrate the superiority of the proposed model.