Efficient Simulative Pass/Fail Characterization Applied to Automotive Power Steering

Abstract

Any component should optimally serve its application by providing exactly the right quantity of features and performances. This is called application fitness of a component. Application fitness can be assessed by simulating component models in an application model. Varying the component's performances may end up in a pass-or fail-behavior with regard to the application requirements. Characterizing the border between this pass and fails states is extremely helpful in the definition of the component's properties. With a number of component properties, this characterization problem gets complex. In this paper, we propose an approach for the planning of simulative experiments, to efficiently characterize this pass/fail border in n dimensions. Especially, smart sampling helps a lot to keep the simulation effort at bay, even if the pass or fail domain falls into a number of unconnected regions. The proposed approach is evaluated taking into account semiconductor components in an automotive electric power steering application. The smart sampling as proposed shows substantial improvements in the number of simulation runs while maintaining a comparable resolution at the border. 1