Handling Multi-Source Uncertainty in Accelerated Degradation Through a Wiener-Based Robust Modeling Scheme.

Area of Science:

• Reliability Engineering
• Materials Science
• Statistical Modeling

Background:

• Accelerated lifetime modeling faces challenges from complex degradation, limited data, and external factors.
• Existing methods struggle with unit-to-unit variability and external perturbations, impacting prediction accuracy.

Purpose of the Study:

• To develop a robust framework for accurate accelerated lifetime modeling and prediction.
• To address challenges of heterogeneous degradation, limited samples, and external disturbances.

Main Methods:

• A Wiener process-based framework incorporating random-effect structures for variability.
• Interval-based inference to handle sampling limitations.
• A hybrid estimator (Huber loss + Metropolis-Hastings) to mitigate external influences.
• Quantitative stress-parameter linkage for transferring accelerated test data to normal conditions.

Main Results:

• The framework provides more stable parameter inference compared to traditional methods.
• Improved reliability in lifetime predictions was demonstrated.
• Validation on connector stress relaxation data confirmed the methodology's effectiveness.

Conclusions:

• The proposed Wiener process framework offers a robust solution for complex accelerated lifetime modeling.
• This approach enhances the accuracy and reliability of predictions in engineering applications.
• The methodology effectively handles unit variability and external disturbances for better lifetime forecasting.