Rolling element bearing remaining useful life estimation based on a convolutional long-short-term memory network

The rolling element bearing is the leading cause of failures in rotating machinery; on that account, the accurate prediction of its remaining useful life (RUL) using sensor data is an important challenge to improve the reliability and decrease the maintenance costs. Classical data-driven approaches rely on manually extracted features from raw sensor data followed by an estimation of a health indicator, the degradation states and the prediction of RUL using a failure threshold. Based on the recent success of deep neural networks in various artificial intelligence domains, we propose an end-to-end deep framework for RUL estimation based on convolutional and long-short-term memory (LSTM) recurrent units. First the neural network extracts the local features directly from sensor data using the convolutional layer, then an LSTM layer is introduced to capture the degradation process, finally the RUL is estimated using the LSTM outputs and the prediction time value. Experiments are conducted on the ball bearing data provided by FEMTO-ST Institute. The results demonstrate the efficiency of our approach.