Optimal design of electromechanical actuators: a new method based on global optimization

The aim of this paper is to show the advantage of a deterministic global-optimization method in the optimal design of electromechanical actuators. The numerical methods classically used are founded either on nonlinear programming techniques (i.e., augmented Lagrangian, sequential quadratic programming) or on stochastic approaches which are more satisfactorily adapted to global optimum research (i.e., genetic algorithm, simulated annealing). However, the latter methods only guarantee reaching this global optimum with some probability. The present paper proposes a deterministic branch and bound algorithm associated with interval arithmetic which is then applied to the dimensioning of a slotless permanent magnet machine. The problem is formulated as a multiobjective-optimization problem with mixed variables. Original and unexpected results in this optimal design are obtained, and comparisons with previous works are presented.