Modeling of multi-variable correlation degradation process based on copula function and improved particle swarm optimization algorithm

With the increasing complexity of modern equipment systems, ignoring the correlation between multi-component and multi-performance degradation processes can lead to significant deviations in reliability assessment. Aiming at the limitation of current condition-based maintenance research, which predominantly focuses on single-component or single-performance degradation process systems, this paper proposes a multivariate degradation modeling method based on copula functions. This study first analyzes the correlations between multivariate degradation processes using the copula function. Subsequently, it compares the characteristics and applicability of univariate degradation models based on the gamma process and Wiener process. Building on this foundation, copula functions are utilized to establish multivariate correlated degradation models and system reliability models, along with parameter estimation methods and copula function selection strategies. The particle swarm optimization algorithm is improved through asymmetric learning factors and dynamic adjustment of inertia weights, enhancing its computational performance. Case analysis demonstrates that compared to gamma and Wiener degradation models, the copula-based multivariate correlated degradation model achieves 3.53% and 11.39% better fits for real-world transmission degradation data, respectively, validating the superiority of the proposed method and model.