抄録
Design optimization in mechanical engineering has been very popular nowadays and number of algorithms and softwares are proposed. However, it is not so widely recognized that before finding improved optimum solution, finding and establishing feasible region is necessary. And,in some cases, it would be hard work rather than improving objective function. Most of non-linear programing algorithms need initial design point being set within constraints or just allow a little violation of constrains. Popular meta-heuristic algorithms, such like Genetic Algorithm(GA) and Simulated Annealing(SA) , are in same situation regarding demand of constraints satisfaction. So, to success optimization in such infeasible problem, it is necessary to establish feasible region before starting. We proposes non linear discriminant analysis as such method. Discriminant analysis is one of common methods in multi-valiate analysis in statistical analysis.
