Stress-test Based Transition Model for Lifetime Drift Estimation and RUL Prediction of Discrete Parameters in Semiconductor Devices

In recent years, self-driving technologies in cars have become more and more mature. This affects the whole automotive industry. Autonomous cars are expected to have more up-time and more total usage time compared to the current generation of non-autonomous vehicles.

In semiconductor industry for automotive applications, functionality over lifetime is a quality target. With the increasing usage time in self-driving cars, new challenges arise in the prediction of remaining useful life (RUL) in the context of prognostics and health management (PHM). Predictions of remaining useful life are both important for on-line monitoring and product testing before shipping. For this, statistical models for lifetime based on accelerated stress tests are needed.

We propose a semi-parametric transition model for the calculation of the lifetime drift of discrete electrical parameters based on accelerated stress tests. We further discuss methods for extrapolation of projected drift to calculate interval estimators for the remaining useful life.

Keywords: Lifetime drift model, Quantile regression methods, Remaining useful life prediction, Semiconductor industry, Transition model.

Download PDF