Affiliations: [a] Laboratoire des Plasmas et de la Conversion de l’Energie (LAPLACE), ENSEEIHT-INPT, Toulouse, France | [b] Mitsubishi Electric R&D Centre Europe (MERCE), Rennes, France
Correspondence:
[*]
Corresponding author: F. Messine, Laboratoire des Plasmas et de la Conversion de l’Energie (LAPLACE), ENSEEIHT-INPT, Toulouse, France. E-mail: messine@laplace.univ-tlse.fr
Abstract: Numerical optimization algorithms based on finite element analysis are widely used to design electrical machines, especially with complex geometries. The advantage of these approaches is that they have high accuracy by taking into account nonlinear properties of the materials. However, the computational time is important. In this paper, quasi-Newton based algorithm using fmincon Matlab solver and mesh adaptive direct search based algorithm using NOMAD solver are compared by testing several relaxed design problems in order to find an appropriate algorithm for our optimization problem. Finally, we achieve the optimal solution and the shape of the optimal machine is then validated.
