On Simulation Knowledge Acquisition Using Gaussian Processes For The Design Of Electric Motors

In the automotive industry, high efforts are made to design high-efficiency, low-cost, requirement-optimized electric drives. Traditionally, time-consuming and expensive 3D finite element analyses are used to assess the effects of different design parameters from various engineering disciplines such as the electromagnetics, heat transfer, and mechanics. Often, to decrease the calculation efforts, scalable, physics-based reduced order models are used. For setting up and interpreting such reduced order models, simulation experts employ implicit knowledge by applying heuristic correction factors. This paper proposes a novel method to formalizing such implicit simulation knowledge using Gaussian Processes. Furthermore, the successful application of the method to a case study of electric motors is reported by validating the aforementioned correction factors for different motor designs. (C) 2019 The Authors. Published by Elsevier B.V.