System-level remaining useful life prediction methodology based on the dynamic health index of multi-indicator fusion: Two cases of subsea equipment

Abstract

System-level remaining useful life (RUL) prediction is challenging because multisensor data fusion requires an integrated representation of multicomponent degradation, and the uncertainty of model parameters is often overlooked during prediction. To address these issues, this study proposes a system-level RUL prediction methodology based on multisensor and multi-indicator information fusion. The main contribution is the development of a dynamic and reliability-driven health index (HI) fusion framework that enables accurate and stable assessment of overall system degradation. In the HI construction process, a system reliability model is established using prior information, and sensitivity analysis is performed to derive the allocation matrix. A dynamic HI prediction model is then formulated by integrating the allocation matrix, indicator-level health states, and a historical-data-based HI degradation model. Furthermore, a dynamic Bayesian network is employed to update degradation parameters and propagate uncertainty, thereby improving prediction robustness under varying operating conditions. The effectiveness and accuracy of the proposed methodology are demonstrated through a field case of a subsea Christmas tree and an experimental case of a subsea blowout preventer system. The predicted system-level RUL and its uncertainty range provide quantitative guidance for optimizing maintenance scheduling, determining inspection intervals, and planning life-extension strategies.