Optimal design of accelerated degradation experiments based on multiple linear process regression

This study develops an optimization design methodology for accelerated degradation testing based on multivariate linear process regression. Initially, multivariate linear process regression equations are formulated, along with confidence upper and lower curves for degradation percentile values. The conventional regression analysis is extended to accommodate heteroskedastic process data by assuming a quadratic variance function within the model. Subsequently, accelerated degradation test condition s are optimized by minimizing the asymptotic variance of life percentiles under standard stress levels. Finally, a case study analysing accelerated degradation test data of an electronic component is presented. This methodology fully exploits longitudinal information across measurement times to statistically infer the complete accelerated degradation process, achieving high informational efficiency and accuracy.