A robust intelligent fault diagnosis method for rolling element bearings based on deep distance metric learning

Abstract Intelligent data-driven fault diagnosis methods for rolling element bearings have been widely developed in the recent years. In real industries, the collected machinery signals are usually exposed to environmental noises, and the bearing operating condition changes in different working scenarios. That leads to distribution discrepancy between the labeled training data and the unlabeled testing data, and consequently the diagnosis performance deteriorates. This paper proposes a novel deep distance metric learning method for rolling bearing fault diagnosis based on deep learning. A deep convolutional neural network is used as the main architecture. Based on the learned representations through multiple hidden layers, a representation clustering algorithm is proposed to minimize the distance of intra-class variations and maximize the distance of inter-class variations simultaneously. A domain adaptation method is adopted to minimize the maximum mean discrepancy between training and testing data. In this way, the robustness of the fault diagnosis method can be significantly improved against noise and variation of working condition. Extensive experiments on a popular rolling bearing dataset are carried out to validate the effectiveness of the proposed method, and the diagnosis performance is widely evaluated in different scenarios. Comparisons with other approaches and the related works on the same dataset demonstrate the superiority of the proposed method. The experimental results of this study suggest the proposed deep distance metric learning method offers a new and promising tool for intelligent fault diagnosis of rolling bearings.