A predictive maintenance strategy for multi-component systems based on uncertain process and CEEMDAN-LS: A case study on lithium-Ion Batteries

Predictive maintenance (PdM) is an effective approach to enhancing system availability and reducing maintenance costs. However, existing PdM strategies often focus on maintenance decisions for individual components, with limited optimization of system-level maintenance strategies. To address this gap, this paper proposes a multi-component system PdM strategy based on uncertain process and complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN)-least square (LS). First, considering the impact of epistemic uncertainty, a Remaining Useful Life (RUL) prediction method based on an uncertain process is proposed. CEEMDAN is employed to denoise the available measurements, followed by the estimation of the unknown parameters of the uncertain degradation model using the least squares method. Second, uncertain simulation is combined with kernel density estimation (KDE) to obtain the probabilistic RUL of the equipment. Subsequently, a preventive replacement strategy is designed by integrating the RUL PDF and the configuration of multi-component systems. Finally, the proposed approach is validated using the NASA lithium-ion battery degradation dataset. Experimental results demonstrate that the proposed RUL prediction method achieves higher prediction accuracy compared to traditional uncertain process, and nonlinear Wiener methods. Moreover, the proposed multi-component system maintenance strategy effectively prevents system failures.