A method for predicting the remaining useful life of shearer bearings based on improved similarity model

The degradation process of shearer bearings is not a simple linear or exponential relationship. It should be analyzed in different stages. However, the current prediction method for the remaining useful life (RUL) of shearer bearings does not fully consider this factor. In order to solve this problem, a method for predicting the remaining useful life of shearer bearings based on an improved similarity model is proposed. The model uses a universal similarity model to describe the process of equipment degradation. Based on this, through root mean square clustering analysis, the bearing degradation process is divided into the stable operation stage, initial degradation stage, and accelerated degradation stage. With the help of traditional similarity model ideas, the health condition of shearer bearings is calculated by segment. And it is fitted to obtain a degradation curve sample library, Through data preprocessing and similarity analysis on offline sample library data and real-time data of online shearers, the bearing RUL of the shearer is ultimately obtained. The experimental results show that the mean absolute error values of the RUL prediction method for shearer bearings based on improved similarity model are reduced by 30.49%, 7.54%, 16.98%, 24.74%, 17.96% and 9.49% respectively, compared to the convolutional gated recurrent unit (ConvGRU), convolutional long short-term memory neural network (ConvLSTM), convolutional neural networks (CNN), self-organizing map (SOM), recurrent neural networks (RNN), and traditional similarity models. The proposed model can effectively predict bearing RUL. The on-site test results show that after continuous monitoring of the bearing of the shearer for 87 days, the health condition of the bearing is gradually reduced from 0.997 to 0.972. The result is basically consistent with the actual use of the bearing of the shearer on site. It verifies the effectiveness of this method.