A multi-objective evolutionary algorithm with integrated response surface functionaltities for configuration optimization with discrete variables

In this paper a multi-objective evolutionary algorithm is introduced. This algorithm is especially adapted to handle discrete and continuous design variables simultaneously making it suitable for configuration optimization problems, which are often characterized by a mix of discrete and continuous variables. Furthermore response surface approximation functionalities have been integrated to increase efficiency and performance. The response surface approximations are built over continuous subspaces using the current population design points. A clustering technique is applied for dividing the population in groups in order to built more accurate response surfaces. Subsequently an optimization is performed on these response surface models and the resulting optima are fed back in the population. The capabilities of the algorithm have been demonstrated on a number of real world problems two of which are presented in this paper. First the buckling load of a stringer stiffened plate is optimized and second the configuration optimization of the satellite encompassing the general structural layout, arrangement of equipment and mechanical dimensioning ^§}.