Reliability inference for field conditions from accelerated degradation testing

TLDR

Accelerated degradation testing typically uses deterministic stress levels, yet the standard reliability prediction approach ignores stress variability, potentially causing significant errors. This study extends a constant‑stress ADT model to predict field reliability by incorporating stress variations. The authors develop a statistical inference procedure based on ADT data and demonstrate it with a case study. Simulation results confirm the procedure’s accuracy across various field‑stress distributions. © 2006 Wiley Periodicals, Inc.

Abstract

Abstract Accelerated degradation testing (ADT) is usually conducted under deterministic stresses such as constant‐stress, step‐stress, and cyclic‐stress. Based on ADT data, an ADT model is developed to predict reliability under normal (field) operating conditions. In engineering applications, the “standard” approach for reliability prediction assumes that the normal operating conditions are deterministic or simply uses the mean values of the stresses while ignoring their variability. Such an approach may lead to significant prediction errors. In this paper, we extend an ADT model obtained from constant‐stress ADT experiments to predict field reliability by considering the stress variations. A case study is provided to demonstrate the proposed statistical inference procedure. The accuracy of the procedure is verified by simulation using various distributions of field stresses. © 2006 Wiley Periodicals, Inc. Naval Research Logistics, 2006.