A Novel Mechanism-Equivalence-Based Tweedie Exponential Dispersion Process for Adaptive Degradation Modeling and Life Prediction

• Published in: Sensors (Basel, Switzerland) Vol. 25; no. 2; p. 347
• Main Authors: Wu, Jiayue, Liu, Yujie, Wang, Han, Ma, Xiaobing, Zhao, Yu
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 01.01.2025; MDPI
• Subjects: Accuracy; Analysis; degradation analysis; Machine learning; mechanism equivalence; Methods; Parameter estimation; RUL prediction; Stochastic models; Stochastic processes; Stress; Trends; Tweedie exponential dispersion process; China; RUL prediction; Tweedie exponential dispersion process; degradation analysis; mechanism equivalence
• ISSN: 1424-8220; 1424-8220
• DOI: 10.3390/s25020347

Accurately predicting the remaining useful life (RUL) of critical mechanical components is a central challenge in reliability engineering. Stochastic processes, which are capable of modeling uncertainties, are widely used in RUL prediction. However, conventional stochastic process models face two major limitations: (1) the reliance on strict assumptions during model formulation, restricting their applicability to a narrow range of degradation processes, and (2) the inability to account for potential variations in the degradation mechanism during modeling and prediction. To address these issues, we propose a novel mechanism-equivalence-based Tweedie exponential dispersion process (ME-based TEDP) for adaptive degradation modeling and RUL prediction of mechanical components. The proposed model enhances the original Tweedie exponential dispersion process (TEDP) by incorporating degradation mechanism equivalence, effectively capturing the correlation between model parameters. Furthermore, it improves prediction accuracy and interpretability by employing a dynamic testing–modeling–predicting strategy. Application of the ME-based TEDP model to high-speed rail bogie systems demonstrates its effectiveness and superiority over existing approaches. This study advances the theory of degradation modeling and significantly improves the precision of RUL predictions.