A Comparison of the Performances between a Genetic Algorithm and the Taguchi Method over Artificial Problems

In the manufacturing industry, to produce the best quality product, it is important to define several levels of inputs. The Taguchi method is proposed for the solution of this problem and is widely used. In this study, a steady-state genetic algorithm (GA), called Genmak, is developed for the solution of the experimental design problems. In order to compare the performance of the suggested algorithm with that of the Taguchi method, 3 sets with different characteristics are carefully designed. Each set has 1000 test problems. Each of these test problems is an experimental design problem having 4 factors with 3 levels. Significant effects and optimum solution are determinated by statistical methods for every problem generated. Two methods are applied to the problems and the number of problems in which the optimum solution is reached is recorded. Then, the methods are compared with respect to these records. The results show that the performance of the GA is as high as that of the Taguchi method. Another important result is that the performance of the methods decreases as the amount of interaction in a problem increases. Overall, it is concluded that GAs are suitable for finding the optimum solution to this kind of problem and can be used as an alternative to the Taguchi method.