Statistical multivariate degradation modeling– A systematic review

By analyzing the historical health status data and identifying the degradation patterns and trends, degradation modeling helps to predict the system reliability, formulate reliable maintenance plans, and prolong the service life of systems. With the increasing structural complexity and functional diversity of modern engineering systems, it is not enough to determine the overall health status of a system merely relying on a single component or performance characteristic (PC). Meanwhile, considering the coupling of components and the impact of the common operating environment, dependence is present between the degradation processes of different components or PCs in the product. Therefore, the multivariate degradation modeling of dependent PCs has attracted widespread attention in recent years. Focusing on the data-driven approaches and especially statistical models, this paper systematically reviews the existing multivariate degradation modeling methods, including general path models, stochastic process models, copula-based methods and other approaches. This review also discusses factors that may affect the modeling of multivariate degradation processes, such as measurement errors and covariate effects. This paper helps to understand the existing studies in a comprehensive and systematic way, which can also inspire further development and applications in multivariate degradation modeling.