Publication Type

Journal Article

Version

publishedVersion

Publication Date

10-2025

Abstract

Multivariate degradation data are increasingly common due to advances in sensor technology. They provide rich information for reliability analysis and remaining useful life (RUL) prediction. The multivariate measurement typically contains actual degradation of multiple performance characteristics (PCs) and measurement errors. The multiple PCs often have different physical meanings with different orders of magnitude, making the associated measurement errors vary in the orders of magnitude as well. It is highly desirable to have a parsimonious model accounting for the possible correlation in degradation measurements, so that more flexibility can be reserved for modeling the actual multivariate degradation. This study proposes a statistical framework that uses multiplicative bias factor for multivariate degradation modeling. The framework is applicable to many existing univariate and multivariate degradation models, including general path models and various stochastic degradation processes. Statistical inference for the resulting models can be readily done using EM-type algorithms. When the framework is applied to a univariate monotone degradation process, statistical inference for the resulting models is easier than the existing additive measurement error models for these processes. To mitigate model misspecification, a Bayesian model averaging scheme is developed to dynamically predict the system RUL when new degradation data are available. The advantages of our model are demonstrated through extensive simulation studies and three real-life datasets.

Keywords

Bayesian model averaging, MCEM algorithm, multivariate degradation, remaining useful life

Discipline

Management Sciences and Quantitative Methods | Operations Research, Systems Engineering and Industrial Engineering

Research Areas

Integrative Research Areas

Publication

Journal of Quality Technology

Volume

57

Issue

4

First Page

331

Last Page

349

ISSN

0022-4065

Identifier

10.1080/00224065.2025.2522412

Publisher

Asqc American Society for Quality Control

Copyright Owner and License

Authors

Additional URL

https://doi.org/10.1080/00224065.2025.2522412

Download

Share

COinS