Shape optimization of magnetic devices using genetic algorithms with dynamically adjustable parameters

Abstract

An improved method for the inverse shape optimization of magnetic devices using genetic algorithms (GAs) with dynamically adjustable parameters is presented. The proposed method starts from an initial population using large number of bits per chromosome enabling searching for the optimal solution in a wider region without aggravating the computational speed. Later, as the optimization process evolves, the search space is gradually decreased by restriction of the number of bits and by translation and reduction of the searching space according to the values of the objective function; therefore, dynamically adjusting to the best fit solution decreases the computation resources to a minimum. The obtained results exhibit an acceleration of the optimization process and an increase of the solution accuracy.