INVERSE DESIGN AND MULTI-PURPOSE OPTIMIZATION METHOD BASED ON A STACKED SUBORDINATE MODEL FOR ELECTRIC MOTORS

Abstract

The invention is a computer-implemented method for determining the design parameters of an electric motor based on target performance constraints. The method samples design parameter sets from the design space of the electric motor, feeds each parameter set into a numerical simulation to obtain performance outputs including efficiency, torque, and power, and creates a training dataset. Using the training dataset, at least two machine learning models that learn the relationship between design parameters and performance outputs are trained and validated separately; physics-based constraints related to the electric motor are incorporated into the loss functions of the models. The trained models are combined in a batch structure to create a surrogate model.Target performance constraints are given to the multi-objective optimization algorithm, the performance outputs of the candidate parameter sets in the design space are calculated through the surrogate model, and the optimum design parameters are determined from among the candidates that satisfy the target performance constraints. Optionally, parameter importance and interaction analysis is performed.