Semiconductor Component Fault Assessment and Probability Impact Estimation on Application Level

Failures at component level can affect the application behavior in many different ways. During product development it is crucial to know the severity and the probability of such influence. The current methodologies for dealing with such problems are based on engineering judgment, but these are limited by the complexity of the applications and its components. In this paper, we present an automated approach in which we model the failures at component level, propagate them in application through simulation, cluster the failures and estimate the overall probability that different application failure modes have. This approach is applied on an automotive Electric Power Steering application while the components of interest are an analog-to-digital converter and a current sensor. The results show that the large number of failure modes on component level boils down to a very low number of application failure modes. For each of these failure modes the probability of occurrence is computed starting from the related root causes on component level1.