Adversarial Adaptation Based on Bidirectional Temporal Convolutional Network for RUL Prediction

Remaining useful life (RUL) prediction is one of the key techniques in prognostics and health management (PHM). Deep learning-based prognostics methods, which can automatically mine useful degradation information from monitoring data and infer causal relationships, have received a lot of attention in RUL prediction of machinery. However, in some industrial application scenarios, the operating conditions of the actual data often differ significantly from those of the training data, which greatly limits the predictive performance of the prediction methods. To overcome the above limitations, an anti-adaptive residual life prediction framework is proposed for RUL prediction under different operating conditions. First, a new network, named bidirectional temporal convolutional network (BDTCN), is proposed to capture the interdependence of the input data on the time scale through forward and reverse convolution operations. Then, an anti-adaptive training strategy is developed to help the BDTCN further extract the operating condition invariant degradation features so that it can perform RUL prediction across operating conditions. The proposed framework is evaluated through ablation experiments and comparison with existing methods. The experimental results demonstrate the effectiveness and superiority of the framework in RUL prediction.