Prediction of Bearing Performance Degradation with Bottleneck Feature based on LSTM Network

As an important component of mechanical equipment, the operating status of bearing is directly related to the overall performance of mechanical equipment. Therefore, the prediction of bearing performance degradation is significant for the health monitoring of mechanical equipment. However, the effect of the entire bearing run time and continuous variation are not considered in many traditional prediction methods. To overcome these problems, we propose a novel method which constructs the prediction model based on long short-term memory network, combined with bottleneck feature. Firstly, multiple statistical features are extracted to make up an original feature set. Next, a bottleneck feature obtains by inputting the original feature set into the stacked auto-encoder (SAE) network. Finally, a long short-term memory (LSTM) network is designed for the prediction of bearing performance degradation. An accelerated degradation test of bearings shows performance of the proposed method is better than general methods.