Prediction Method for Remaining Useful Life of Rolling Bearings Based on a Multi-layer Long Short-Term Memory Model

The traditional prediction method of remaining useful life (RUL) of rolling bearings is tedious in characteristic calculation and weak in model fitting ability. This paper proposes a prediction method for the remaining useful life of rolling bearings based on a multi-layer long short-term memory (LSTM) model in multi-working conditions to address this issue. To establish this prediction model, the input parameters for the model consist of the horizontal vibration acceleration and the operating time of the bearings at the degradation stage, while the output parameter represents the remaining useful life. The XJTU-SY dataset covering multi-working conditions is employed as an example to compare with a recurrent neural network model. The results indicate that the prediction accuracy of the method in this paper surpasses that of the recurrent neural network model.