Topology optimization methods for the design of electromagnetic actuators

(eng) Topology optimization methods aim to design devices in an automated way. The designer defines a design space which is divided in cells, a library of materials and an objective function. The method goal is then to maximize or minimize this objective function by distributing optimally the materials in the cells. This thesis aims to develop topology optimization methods based on deterministic algorithms for the design of electromagnetic devices. Deterministic algorithms are indeed characterized by a fast convergence compared to heuristic algorithms, but are usually faced with two main challenges: the presence of local minimizers and the handling of intermediate materials, i.e. mixes of different materials in the same cell, at the end of the optimization. Through the consideration of study cases, this thesis introduces two groups of methods. The isotropic methods are based on materials characterized by isotropic properties. Each material is then described by a unique design variable per cell giving its proportion in the cell. These methods are usually simple to apply, but the range of problems on which they are effective is limited. The second group is called the anisotropic methods, in which an anisotropic permeability is considered for the iron material. These methods can be applied to a wider range of problems, but involve a larger number of design variables in order to represent the iron magnetic anisotropy, which increases the computation time. The thesis besides studies a mesh refinement algorithm aiming to increase the resolution of the produced topology while limiting the computation time. Eventually, the proposed methods are applied to the design of a switched reluctance motor, a permanent magnet synchronous motor and magnetic thrusts, in order to highlight their performances.