Surrogate Based Adaptation of Existing PMSM Designs to Changing Requirements Considering Material Cost and Influence on Production

This work presents an efficient method for mapping the electromagnetic characteristics of an electric machine, with low errors, using multiple neural networks and physical dependencies. To estimate the impact of design changes on production, an analysis of interdependencies between product parameters and production processes was conducted. The surrogate models allow a fast computation of the adapted machine speed torque characteristics and derived targets like corner point or drive cycle efficiency. In combination with the developed material cost function and the production influence function a paretooptimization is carried out. This work demonstrates how changes in target values affect material costs and the necessary production adjustments, using the position of the corner point and different efficiency levels as an example. The presented approach offers the opportunity to adapt existing machine designs towards a wide range of changed requirements, which lowers development costs and increases the utilization of existing production facilities for small quantities.