Abstract :
[en] The sustained growth of the industrial sector requires high-efficiency electro-me- chanical energy converters, in particular electrical rotating machines, at the lowest possible cost. The use of modern power electronics converters at all levels of elec- trical power applications, involves, on the other hand, switching components with very low switching times and always increasing current levels. Passive components in these devices (busbars, inductors, transformers) must be designed to be com- pact without compromising their performance (e.g. power losses, electromagnetic interference/compatibility). Automated design optimization methods, in particu- lar shape and topology optimization, used so far mostly in the field of structural engineering, offer a major step evolution in the design of such electro-mechanical and electric energy converters. The objective of this thesis is to provide engineers and practitioners of the field with appropriate methods which allow to carry out such design tasks by numerical optimization in an efficient way, and to extend the design capabilities to electro-mechanical converters.
This thesis exploits a computer aided design (CAD) representation of indus- trial systems and the finite element method (FEM) to solve the partial differential equations (PDEs) that govern their behavior under certain physical conditions. This thesis addresses three main subjects. First, the sensitivity analysis of elec- tromagnetic PDEs solution is revisited in view of being used with gradient-based methods. Classical scalar formulations are extended to a general rigorous frame- work, and expressed analytically prior to discretization, to treat the vector case. Secondly, an iterative solver is designed so as to solve efficiently the large-scale linear systems arising from the design problem. Third, the design improvement capabilities are extended by developing an integrated and unified formalism for simultaneous shape and topology optimization of a system.
