Bayesian Averaging Evaluation Method of Accelerated Degradation Testing Considering Model Uncertainty Based on...

Area of Science:

• Reliability Engineering
• Statistical Modeling

Background:

• Accelerated degradation testing (ADT) is crucial for evaluating long-life product reliability under resource constraints.
• Bayesian methods enhance ADT by incorporating historical data and mitigating small sample size limitations.
• Traditional ADT Bayesian evaluation suffers from model uncertainty, potentially leading to inaccurate reliability predictions.

Purpose of the Study:

• To address the challenge of model uncertainty in Accelerated Degradation Testing (ADT) Bayesian evaluation.
• To propose a novel method for quantifying and mitigating the impact of model uncertainty on reliability assessments.
• To enhance the accuracy and robustness of Bayesian evaluations in ADT.

Main Methods:

• Analysis of the ADT Bayesian modeling process.
• Development of a new model averaging evaluation method for ADT Bayesian analysis.
• Application of relative entropy to weigh different models in the evaluation process.

Main Results:

• The proposed model averaging method effectively reduces inaccuracies stemming from model selection uncertainty.
• Demonstrated ability to provide more reliable lifetime and reliability estimations for high-reliability products.
• Quantified the impact of model uncertainty on ADT Bayesian evaluation outcomes.

Conclusions:

• The novel relative entropy-based model averaging method offers a significant improvement over traditional ADT Bayesian approaches.
• This method provides a valuable tool for theoretical research and practical engineering applications in ADT.
• Enhances the reliability of lifetime predictions for critical components and systems.