A novel machinery RUL prediction method based on exponential model and cross-domain health indicator considering first-to-end prediction time

Remaining useful life (RUL) prediction is of great significance in the mechanical transmission system as it provides a basis for the predictive maintenance of bearings and gears. Recently, intelligent data-driven methods of tag-based end-to-end and based-health indicator multi-step iterative for RUL prediction have been developed. However, they still mostly suffer from some limitations: 1) The reasonable failure threshold of mechanical system is rarely considered. 2) First-to-end prediction time (FEPT) is ignored during the construction of degradation trend. To address the above drawbacks, a novel machinery RUL prediction method based on Exponential model and cross-domain health indicator (CDHI) considering FEPT is proposed. The grey target decision (GTD) is first applied to the performance evaluation of mechanical system and a weighted GTD is proposed to construct sensitive health indicators and sensitive failure indicators for determining FPT and EPT respectively. Then, to generate the CDHI considering FEPT under different working conditions, a transfer quadratic function-based ConvLSTM framework is proposed based on the degradation trend of quadratic function and maximum mean discrepancy. Finally, the Exponential model is used to perform RUL regression prediction based on the generated CDHIs. The proposed RUL prediction method has been demonstrated by five datasets and compared with several state-of-the-art approaches. Extensive experiments show that the proposed method outperforms other approaches in terms of superiority, robustness and generalization.