Remaining Useful Life Prediction Method of Offshore Equipment Bearings Based on Convolutional Neural Network-Bidirectional Gated Recurrent Unit-Squeeze and Excitation

Abstract

The remaining useful life forecast (RUL) of rolling bearings, a crucial part of offshore equipment, is one of the most troublesome equipment because it may avoid equipment failure and lessen equipment failure loss. This paper proposes a method to build CNN-BIGRU bearing health indicators based on the SE attention mechanism, and combines primary linear regression to predict the RUL of bearings in order to address the issues of low accuracy and poor generalization performance in the current bearing RUL prediction. The proposed method combines the spatial feature extraction capability of convolutional neural networks with the temporal feature extraction capability of bidirectional gated recurrent units, allowing it to effectively use feature information from the spatial and temporal dimensions of vibration signals to improve prediction accuracy and stability. The suggested technique is validated in this research using experimental data from the 2012 IEEE PHM Challenge for the whole life cycle bearing. The experimental findings reveal that the approach can more accurately estimate the RUL of the bearing than the standard model, proving the usefulness and viability of the suggested method.