Influence of the selection method on the solution of an inverse Stefan problem using genetic algorithms

This paper presents the influence of a tournament selection, proportional selection and proportional selection supplemented with linear scaling or sigma truncation scaling on the accuracy of a solution of a two-phase design inverse Stefan problem using genetic algorithms. In the problem to be solved, the coefficient of convective heat transfer on one boundary had to be so selected that the moving interface of the phase change (freezing front) would take the given position. The solution consists in minimization of the functional whose value is the norm of the difference between the given interface position and the position reconstructed for the selected convective heat-transfer coeffi-cient. For the minimization of the functional genetic algorithms were used, whereas the Stefan problem was solved by an alternating phase truncation method.