April 23, 2024
16:15
Louvain-la-Neuve
Place Sainte Barbe, auditorium BARB 92
![](http://cdn.uclouvain.be/styles/one_third_width/groups/cms-editors-immc/events/DeGreef_picture.png?itok=fsW_pXdm)
This work addresses the challenge of designing a synchronous reluctance machine with a focus on on optimizing mean torque, torque ripple and structural integrity. Simultaneous optimization of all parameters results in computing times of several months, a consequence of the large number of parameters and the long evaluation times of finite element models required to capture all the phenomena involved in the operation of these machines. To overcome the curse of dimensionality, it is proposed to decompose the complex optimization problem into smaller, more manageable subproblems. The decomposition methodology is based on a comprehensive variance-based sensitivity analysis to account for the effects of complex parameter interactions. By interpreting Sobol indices, the developed time-efficient design approaches can effectively reduce the total optimization time by eightfold while maintaining the quality of the solution. In contrast, the literature-inspired approaches benchmarked in this thesis failed to consistently limit torque ripple while ensuring structural integrity. In addition, this work demonstrates that the database collected for sensitivity analysis can also be used to compare different topologies for a wide range of objectives without requiring resource-intensive optimizations. The proposed methodology involves utilizing probability density functions to identify the topology most likely to outperform others. This is illustrated by comparing topologies featuring one or four internal flux barriers, with or without a cut-off barrier, in terms of iron losses, mean torque, torque ripple, power factor, structural integrity, and suitability for sensorless control.
Jury members :
- Prof. Bruno Dehez (UCLouvain, Belgium), supervisor
- Prof. Hervé Jeanmart (UCLouvain, Belgium), chairperson
- Dr. François Henrotte (ULiège, Belgium)
- Prof. Carole Henaux (Université de Montpellier, France)
- Prof. Hamid Ben Ahmed (Ecole Normale Supérieure de Rennes, France)
- Dr. Virginie Kluyskens (UCLouvain, Belgium)
- Dr. Christophe Versèle (Alstom Charleroi, Belgium)