Remaining Useful Life Prediction Approach Based on Data Model Fusion: an Application in Rolling Bearings

Yonghuai Zhu,Jiangfeng Cheng,Zhifeng Liu,Xiaofu Zou,Zhaozong Wang,Qiang Cheng,Hui Xu,Yong Wang,Fei Tao
DOI: https://doi.org/10.1109/jsen.2024.3477489
IF: 4.3
2024-01-01
IEEE Sensors Journal
Abstract:Data-driven methods based on deep neural networks (DNN) are widely employed for predicting the remaining useful life (RUL) of equipment, yielding remarkable results. However, the performance of DNN relies on the availability and completeness of full lifecycle data. Moreover, problems such as lack of interpretability of prediction results and weak model generalizability still exist. A RUL prediction approach based on data model fusion is proposed in this paper to address these problems. This approach incorporates physics knowledge into the stacked bidirectional long short-term memory network (SBiLSTM) through three ways. Firstly, the full lifecycle data based on the physics degradation model is integrated with sensed data to ensure the integrity of degradation data. Secondly, the degradation trajectory simulated based on the physics degradation model is used as an input feature for the SBiLSTM, enabling the model to better learn the state evolution process of the equipment. Moreover, a multi-objective loss function is constructed by introducing a physics-guided inconsistency loss function alongside the data loss function, to ensure the model predictions consistent with the known physics phenomena and enhance the interpretability of the model. Case studies are conducted for XJTU-SY dataset and PHM2012 dataset to systematically validate the proposed approach. Comparisons with existing data-driven and hybrid methods are made and the results consistently demonstrate the accuracy of the predictions and the robustness of the performance.
What problem does this paper attempt to address?