| In practice, design variables are always disturbed or change because of different problems after optimal operating conditions confirmed. A global optimal solution which is sensitive to some variable perturbation in its vicinity may be somewhat different from the theoretical global optimal solution and make manufacture device not work satisfyingly. Thus, as the final target searching robustness optimal solutions which are immune to the adjacent disturbance is more significance than global optimal solutions in practical application. Based on Pareto optimal solutions of multi-objective optimization problems, this paper presents a method that can search robustness optimal solutions more effectively by combining the concept of degree of robustness and NSGA-II.Firstly, to get better optimal solutions under disturbed condition, the degree of robustness is defined more accurately. For a solution, its neighbor samples not only have to meet the external constraint condition in the objective function space, but also have to satisfy the new internal constraint condition that can improve the quality of robustness optimal solutions by more strict robustness requirement. Then, as the fitness function, the value of mean effective objective function of neighbor samples is computed according to the degrees of robustness of solutions. In the process of evolution, degree of robustness enters in consideration when selecting solutions for the next generation. From a macro point of view, it promotes the request of optimal solutions'robustness. Finally, by doing a lot of simulation experiments on two different test functions, the parameters'impacts on searching results are analyzed and the comparisons with the methods of searching robustness optimal solutions proposed by Deb et al. are made. Simulation results show that the measure in this paper can find the optimal solutions with better robustness and provide the distribution of solutions with different degrees of robustness. |
PDF Full Text Request