Méthode de décomposition de domaine mixte pour la simulation magnétostatique de machines électriques

The transport industry has been moving towards a strong electrification of propulsion chains for several years. This trend imposes the search for innovative electrical machine structures and their optimization. Thus, this virtual testing approach remains very demanding because the complex operation of these machines can only be modeled from expensive 3D finite element simulations, including the consideration of the rotor part movements. This computing time constraint is incompatible with the implementation of an optimization approach for these new machine topologies since the optimization process can lead to numerous evaluations of the numerical model. This is why such optimization methods are currently only implemented for relatively simple models in the industrial environment. The objective of the thesis is then to propose an innovative computational strategy that allows to strongly reduce the cost of these numerical simulations. For this purpose, a domain decomposition technique based on a mixed writing of equations at interfaces and adapted to the geometrical particularities of rotating machines will be developed. This strategy is based on a discretization using Whitney elements to solve 3D magnetostatic problems, a multi-scale approach taking advantage of parallel computers, and an adapted management of the rotor motion.