An adaptive prognostics method based on a new health index via data fusion and diffusion process. (April 2022)

• Record Type: Journal Article
• Authors: Li, Peng; Maged, Ahmed; Zhang, Aibo; Xie, Min; Dang, Wei; Lyu, Congmin
• Is Part Of: Measurement. Volume 193(2022)
• Publication Date: 2022-04
• Subjects: Remaining useful life (RUL) -- Genetic algorithm -- Adaptive extended Kalman filter (AEKF) -- Solid-state drives (SSDs); Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.110968
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A fitness metric is proposed for feature selection, and a composite HI is optimally constructed using a genetic algorithm and then transformed to a scaled-corrected HI. A diffusion process-based HI degradation model considering multi-source uncertainties is employed with the fitting-based adaptive EKF algorithm for real-time RUL prediction. The proposed prognostics approach is applied to state-of-the-art 3D triple-level cell solid-state drives in data centers for the first time. Abstract: Remaining useful life prediction (RUL) is critical in predictive maintenance for components or systems prone to deterioration. However, direct RUL prediction methods have difficulties tracking health trends and realizing online prognostics. To address this issue, this paper proposes a novel health index (HI) based adaptive prognostics method by leveraging the advantages of both data fusion to handle multi-dimensional data and the adaptive extended Kalman filter (AEKF) algorithm for parameter identification in the diffusion process. A fitness metric is proposed for feature selection, and then the composite HI sequence is constructed via data fusion using the genetic algorithm. Furthermore, a diffusion process model is built to characterize HI degradation while considering multi-source uncertainties. Model parameters are then updated using the fitting-based AEKF method. Finally, the proposed method is validated on a real-world dataset of solid-state drives in data centers, and prediction results and comparative studies verify its superiority.