Fault diagnosis of rod pumping system based on deep conditional domain adaption network

Abstract

The generalization ability of traditional fault diagnosis methods is insufficient. This paper presents a fault diagnosis method of sucker rod pumping system based on deep condition domain adaption network (DCDAN). Firstly, the convolution neural network is used for feature extraction. Then, the weighted maximum mean discrepancy (WMMD) is used to adjust the characteristic distribution of relevant subclasses in different domains to realize the fine-grained domain adaptation of subclasses. At the same time, the fault classification ability of the model is guaranteed by optimizing the classification loss. The results show that this method can improve the generalization performance of the model in the target domain.