MDSM
← All presentations
MDSM 2027 presentation

AI-Assisted Multi-Objective Optimization of Axial-Flux Motors: Methodology, Constraints, and Performance Trade-Offs

Type: Session

Schedule to be announced.

About this presentation

Axial-flux permanent-magnet motors offer significant potential for compact, high-torque-density electric drives. However, their design involves complex trade-offs among motor mass, magnet usage, electromagnetic losses, and electrical loading. Exploring these trade-offs requires extensive numerical simulations and carefully formulated optimization strategies.

This presentation introduces an AI-assisted multi-objective optimization methodology combining quasi-3D finite-element analysis (FEA), MATLAB automation, surrogate-based optimization, and AI coding agents. A practical rotor–stator–rotor axial-flux motor with a yokeless stator is used to demonstrate the complete workflow, from defining normalized design variables and optimization objectives to evaluating Pareto-optimal solutions.

Particular attention is given to the role of AI coding agents in developing and refining MATLAB optimization scenarios. Rather than replacing conventional numerical simulation, AI assists engineers in creating evaluation functions, implementing constraints, automating simulations, and processing optimization results while maintaining engineering oversight. A central part of the study compares two approaches to handling current-density constraints: a penalty-based formulation and a fixed-current-density formulation with dependent stator-height adjustment. The comparison demonstrates how different constraint implementations influence the resulting Pareto fronts and the achievable trade-offs between motor mass and electromagnetic losses.

A subsequent focused optimization study combines promising designs from the initial studies and refines the search within a reduced design space. The selected optimized design achieves approximately 44% lower magnet mass and 15% lower electromagnetic losses, with an efficiency improvement of 3.7 percentage points at the specified operating point, while maintaining nearly the same total active-material mass as the baseline design.

The presentation will discuss practical implementation challenges, simulation verification, and the limitations of AI-assisted engineering workflows. Attendees will gain insights into structuring automated optimization studies, selecting appropriate constraint formulations, interpreting multi-objective optimization results, and integrating AI coding agents into engineering workflows without compromising simulation accuracy and reliability.

Speaker

Vladimir Kuptsov — Speaker.

View the program ↗