A deep adversarial learning prognostics model for remaining useful life prediction of rolling bearing

Remaining useful life (RUL) prediction for condition-based maintenance decision making plays a key role in prognostics and health management (PHM). Accurately predicting RUL of rotating components of complex machines becomes a challenging task for PHM. It is known that for many existing methods the current prediction error of RUL prediction may be accumulated into the future predictions, and this can lead to a prediction error superposition problem. In this paper, the formation mechanism of prediction error superposition is analyzed, and for the first time a deep adversarial LSTM prognostic framework is proposed to overcome the major issue related to prediction error superposition. In the proposed framework, a generative adversarial network (GAN) architecture combining long short-term memory (LSTM) network and auto-encoder (AE) is investigated for bearing RUL monitoring. In the proposed deep adversarial learning prediction framework, due to the potential involvement of long-term and complex tasks, the LSTM (generator) is used to predict the degradation process of rolling bearings based on available historical data, and a simple but useful AE (discriminator) is used to determine and refine the accuracy of the prediction. Therefore, the discriminator plays the adversarial role of the generator (LSTM), and in this way, the prediction accuracy of the LSTM network can be significantly improved. For illustration purpose, two practical case studies are presented to show the prediction performance of the proposed method. Experimental results show the proposed method works well for vibration monitoring and performs better in comparison with the reference machine learning and deep learning approaches.