The role of coarse models in space-mapping: A study on an IPM motor optimization

Space-Mapping (SM) is a recognized and successful engineering optimization paradigm. It replaces the optimization of a high-fidelity model by an iterative process of optimization and update of a low fidelity, but cheap to evaluate, model. In spite of its success, its performance greatly depends on how fast and on how accurate the cheap model is. In this work, utilizing the optimization of an IPM motor as a test bed, we make a thorough experimental study of a number of coarse models of varying accuracy and execution times. The main goal is to show how the characteristics of the coarse models affect the SM performance.