An Efficient Method Based on Conditional Generative Adversarial Networks for Imbalanced Fault Diagnosis of Rolling Bearing

Abstract

Fault diagnosis of rolling bearing has always been a vital component in industrial field, and effective fault diagnostic methods can guarantee normal progress of manufacturing production. However, the scarcity of fault samples in practical scenarios is still a vexed question, which will seriously affect the accuracy of data-driven diagnostic methods. For the settlement of above problem, this paper introduces a supervised generation model CGAN (Conditional Generative Adversarial Network) to generate multitudinal fault data, and replaces the real fault data with the generated one to constitute a new dataset to train the classifiers adequately. In order to verify the effectiveness of the proposed method, the experiments are carried out on both artificial dataset and real one. The results show that the generated data of CGAN not only has a high degree of similarity with the real data, but also effectively improves the fault diagnosis accuracy of rolling bearing.