Comparison of Sensitivity Analysis Methods in High- Dimensional Verification Spaces

The modern electronic systems grow in complexity as the number of potential factors that affect the systems behavior is increasing. The analysis of such systems can be simplified if the most important factors are identified upfront. Sensitivity analysis addresses this problem. The paper describes and compares six sensitivity analysis methods based on: variance decomposition, OneFactor-at–a-Time screening and regression. The methods are applied on a set of custom functions and a real system from the field of automotive electrical systems: a battery control model. The best performances were obtained by the following methods: the variance-based methods (Jansen and EFAST) and the regression based method.