A Structural Causal Model for Electronic Device Reliability: From Effects to Counterfactuals

Abstract

Electronic devices exhibit changes in electrical resistance over time at varying rates, depending on the configuration of certain components. Since measuring overall electrical resistance requires partial disassembly, only a limited number of measurements are performed over thousands of operating hours. This leads to censored failure times, whether under natural stress or under accelerated stress conditions. To address these challenges, including device-specific failure thresholds, a parametric structural causal model is developed to extract information from both observational and experimental data, with the aim of estimating causal effects and counterfactuals, regardless of the applied stress regime. Synthetic data are used to illustrate the main findings.