A hybrid algorithm for predicting the remaining service life of hybrid bearings based on bidirectional feature extraction

The reliability of bearings is crucial for the safe operation of industrial machinery. However, many current algorithms mainly focus on the forward features and ignore the information contained in the reverse features, resulting in an unstable prediction process. Addressing this problem, this paper proposed a hybrid forecasting model integrating two-way feature merging, basing a bidirectional time convolutional network (BiTCN). Firstly, to enhance the extraction of degeneration information, the initial data is processed using variational mode decomposition (VMD) optimized by the black-winged kite algorithm (BKA). BiTCN subsequently isolates the bidirectional features and inputs them into bidirectional gated recurrent unit (BiGRU). Finally, employing an attention mechanism to enhance focus on crucial information. Experiments show that our method outperforms existing state-of-the-art methods, with an average reduction of 0.006 and 0.001 in Mean Absolute Error and Root Mean Square Error on the three bearing datasets and a more stable performance throughout the prediction process.