WR-IMT: A Time Series Predictive Model for Remaining Useful Life Prediction

Haibo Xing,Fei Xiao,Jianxun Li
DOI: https://doi.org/10.1109/ccdc52312.2021.9602062
2021-01-01
Abstract:Deep learning has achieved great success in the remaining useful life (RUL) prediction of machines. However, when it comes to the complex time-dependence data in multi-sensor system, most deep learning approaches such as LSTM (Long Short-Term Memory) and TCN (Temporal Convolutional Network) cannot effectively extract the longterm patterns hidden in time series. In this paper, An end-to-end model named WR-IMT (Weight-restricted Improved MGU-TCN) for long-term RUL prediction is put forward. Firstly, an Improved MGU with bidirectional asymmetric residual structure is designed at first to extract the fine-grained temporal patterns. Then, we proposed the Sequence Segmentation and Occlusion Technology (SSOT) by which the subsequences can be input to our Improved MGU with shared weights. TCN is used for further extraction of coarse-grained patterns. To reinforce the network sensitivity to nonterminal time steps, the Coefficient-restricted Regularization is added to the loss function. Experimental results on the NASA-developed Commercial Modular Aero-Propulsion System Simulation(C-MAPSS) dataset confirm the superiority and effectiveness of our model.
What problem does this paper attempt to address?