A Robust Deep Learning-Based Fault Diagnosis Method for Rotating Machinery

In the recent years, intelligent data-driven fault diagnosis methods on gearboxes have been successfully developed and popularly applied in the industries. Currently, most of the machine learning techniques require that the training and testing data are from the same distribution. However, this assumption is difficult to be met in the real industries, since the gearbox operating conditions usually change in practice, which results in significant data distribution gap and diagnostic performance deteriorations in applying the learned knowledge on the new conditions. This paper proposes a deep learning-based domain adaptation method to address this issue. The raw current signals are directly used as the model inputs for diagnostics, which are easy to collect in the real industries and facilitate practical applications. The maximum mean discrepancy metric is introduced to the deep neural network, the optimization of which guarantees the extraction of generalized machinery health condition features across different operating conditions. The experiments on a real-world gearbox condition monitoring dataset validate the effectiveness of the proposed method, which offers a promising tool for cross-domain diagnosis in the real industries.