dc.description.abstract
This thesis presents the development of an open-source optimisation framework for the prelim- inary design and comparative analysis of high-performance permanent magnet synchronous motors (PMSMs). The proposed methodology combines multi-objective evolutionary optimi- sation with finite element electromagnetic simulations to systematically explore the trade-offs inherent in electric machine design. The optimisation problem is formulated using the Non-Dominated Sorting Genetic Algorithm II (NSGA-II), enabling the simultaneous optimisation of conflicting objectives such as maximising electromagnetic torque and minimising active mass. Candidate motor geometries are evaluated through automated two-dimensional magnetostatic simulations using FEMM 4.2, with geome- try generation, meshing, solution and post-processing fully scripted via Python. This approach allows thousands of physically consistent motor designs to be analysed under realistic electro- magnetic constraints, including magnetic saturation limits, current density bounds and inverter voltage restrictions. The framework is applied to a representative high-performance traction application based on Formula Student in-wheel motor requirements. Electrical, geometrical and thermal constraints derived from the vehicle powertrain define a realistic and highly coupled design space. The optimisation results produce Pareto fronts that clearly reveal the trade-off between torque ca- pability and machine mass, while highlighting the impact of magnetic loading, slot geometry, magnet thickness and axial length on overall performance. Beyond demonstrating numerical convergence and computational scalability through parallel evaluation, the results confirm that the obtained solutions follow physically meaningful trends consistent with established electric machine design theory. The framework is shown to be suit- able for early-stage motor sizing, topology comparison and design space exploration. By relying exclusively on open-source tools, the proposed methodology offers a transparent, extensible and cost-effective alternative to commercial motor design software. It provides a valuable platform for both academic research and engineering education, while supporting the long-term objective of developing in-house motor design capabilities for high-performance elec- tric propulsion systems.
