This paper investigates the capability of a Quasi-Newton optimization algorithm, the L-BFGS-B one, to reduce the drawbacks of the direct coupling with FEM models. After a short description of the L-BFGS-B algorithm, the authors have tested it to the TEAM Workshop Problem 25. L-BFGS-B algorithm issues linked to FEM estimation of the objective function have been analyzed. Finally the authors present the results of an electromagnetic device optimal design, using a large number of parameters (47).

As the refrigeration technology consumes more than 15% of worldwide consumption of electricity, the development of technologies with improved efficiency remains an important challenge. In this context, magnetic refrigeration is an interesting solution and this paper proposes an original magnetic refrigerator. The aim -and also the originality -of this new device is to integrate the magnetocaloric regenerator in the air gap of a Permanent Magnet Synchronous Machine (PMSM). The conventional design of PMSM stator and rotor is presented and the results lead to the construction of a prototype that enables the validation of the proposed concept. Finally, thanks to the coupling of an optimization algorithm and a 2D FEM simulation, the rotor is optimized and the airgap flux density is greatly improved (from 0.9 T to 1.35 T).

No abstract