Hybrid analytical-surrogate modelling & multi-objective optimization of PMSMs

Thumbnail Image

Files

master_Kamran_Muhammad_2026.pdf (1.48 MB)

URL

Journal Title

Journal ISSN

Volume Title

School of Electrical Engineering | Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2025-12-10

Department

Major/Subject

Energy Systems and Markets

Mcode

Degree programme

Master's Programme in Advanced Energy Solutions
Kehittyneiden energiaratkaisujen maisteriohjelma
Magisterprogrammet i avancerade energilösningar

Language

en

Pages

57

Series

Abstract

This thesis contributes to the electrical machine design by creating a state-of-the-art, surrogate accelerated optimization algorithm to the surface-mounted per-anent magnet synchronous motors (PMSMs). Among the foundations of the field of analytical rigor and machine learning, the work attempts to tackle one of the most critical weaknesses of motor research to date: the trade-off between computational efficiency and model fidelity in the iterative design process. The study develops a validated, analytical model of PMSM and uses it to produce a rich, physically based data set based on which training high-performance neural network ensemble surrogates is done. These surrogate models, which are carefully checked against analytical calculations, exhibit almost perfect predictive accuracy and consistency in quantifying uncertainty, enabling the risk-conscious and fast design exploration. Through the combination of the surrogate and the multi-objective genetic algorithms, the study provides the best solutions globally without interrupting the tradeoffs between the torque and the efficiency and the loss with an unprecedented level of computational speed. The results map the Pareto-optimal design frontier appro-riately, and shed light on the natural tradeoffs of high-performance PMSM design engineering, and give a single framework to design optimization times in weeks down to minutes. This will not only democratize the use of complex design solutions - the reliance on high cost simulation tools can be eliminated - but also create a transparent, reproducible workflow of the next-generation motor research and industrial implementation.

Description

Supervisor

Belahcen, Anouar

Keywords

PMSMs, pareto front, analytical, surrogate modelling, neural networks, multi-objective optimization

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202601201646

Collections

[dipl] Sähkötekniikan korkeakoulu / ELEC