Switched Reluctance Motors (SRMs) are gaining recognition due to their robust design, cost-effectiveness, fault tolerance, and reliable high-speed performance, positioning them as promising alternatives to traditional electric motors. However, SRMs face high torque ripples, vibration, acoustic noise, and nonlinear modeling complexities. Through careful geometry design optimization, these drawbacks can be mitigated. Design optimization for SRMs is a multi-objective and nonlinear problem that requires an accurate finite element analysis (FEA) model to relate designable parameters to output objectives. The geometric design process follows a multi-stage and iterative approach, leading to prohibitive computational time until the optimal design is reached.
Machine learning algorithms (MLAs) have recently acquired attention in electric machine design. This study introduces an extensive analysis of various MLAs applied to SRM modeling and design. Additionally, it presents a robust framework for a comprehensive evaluation of these MLAs, facilitating the selection of the optimal machine learning topology for SRM design. Existing research on the geometry optimization of SRMs using MLAs has focused only on the machine’s static characteristics.
This thesis introduces an advanced optimization method utilizing an MLA to act as a surrogate model for both static and dynamic characteristics of the SRM. The dynamic model incorporates conduction angles optimization to enhance the torque profile. The proposed MLA is applied to map out the SRM geometrical parameters, stator and rotor pole arc angles and their dynamic performance metrics, such as average torque and torque ripples. The optimal design improves the average torque and significantly reduces the torque ripples.
Radial forces constitute a critical objective that should be considered alongside average torque, efficiency, and torque ripple in the design optimization of SRMs. Accurate modeling of radial forces is a prerequisite for optimizing motor geometry to mitigate their adverse effects on vibrations and acoustic noise. This work presents an MLA-based surrogate model for the most influential radial force harmonic components, facilitating the integration of radial force reduction into a multi-objective optimization framework.
The proposed optimization framework employs two MLA-based surrogate models: the first correlates SRM pole arc angles with average torque and torque ripples, while the second models the most significant radial force harmonics. A genetic algorithm leverages these surrogate models to predict new geometrical parameters that enhance the SRM's torque profile and reduce radial forces. The optimization framework significantly reduced torque ripples and radial forces while slightly increasing average torque. The optimal design candidates were verified using FEA and MATLAB simulations, confirming the effectiveness of the proposed method, which offers significant computational time savings compared to traditional FEA techniques. / Thesis / Doctor of Philosophy (PhD)
| Identifer | oai:union.ndltd.org:mcmaster.ca/oai:macsphere.mcmaster.ca:11375/30074 |
| Date | January 2024 |
| Creators | Omar, Mohamed |
| Contributors | Bakr, Mohamed, Emadi, Ali, Electrical and Computer Engineering |
| Source Sets | McMaster University |
| Language | English |
| Detected Language | English |
| Type | Thesis |
Page generated in 0.0015 seconds