Multilevel Variability and Epistemic Uncertainty Analysis of Reliability Physics Models Through Hierarchical Probabilistic Modelling

Variability is a critical issue in modern reliability prediction, with small discrepancies between devices and lots significantly impacting failure rates, leading to uncertainty and risk that current methods struggle to quantify. We develop a probabilistic modelling framework for Bayesian analysis of semi-conductor reliability capable of capturing simultaneous device, chip, and lot-level variability effects alongside epistemic uncertainty analysis. The novel framework and technical infrastructure required for effective use is explained, with provided case studies demonstrating the methodological benefits of the framework over existing statistical approaches to semiconductor reliability.