Bearing Remaining Life Prediction Based on Temporal Convolutional Networks with Hybrid Dilated Convolutions

Rolling bearings play a crucial role in mechanical equipment, impacting both running time and costs significantly. Accurate prediction of remaining life allows timely maintenance actions to be taken, thereby reducing downtime and maintenance costs, and is essential for effective equipment maintenance and operation. In this paper, we propose a new bearing life prediction model based on hybrid dilated time convolutional network structure. We address the issue of convolutional kernel discontinuity caused by empty convolutions in the Temporal Convolutional Network by introducing the Hybrid Dilated Convolution technique. This alleviates the problem of local feature loss to some extent and enables more global features to be captured. Finally, the effectiveness of the model was validated on the publicly available FEMTO-ST dataset.