Application of Genetic Algorithm to Numerical Experiment in Robust Parameter Design for Signal Response System

Abstract

This paper discusses how to apply the genetic algorithm method with signal response robust parameter design for determination of the best levels of controllable factors effectively. Although there have been many studies on genetic algorithm for robust design, most of them focused on simple response problems, while recently complicated signal response problems have gradually become indispensable. To enhance the effectiveness of genetic algorithm in dynamic robust design, the advantages of traditional orthogonal array experiment concepts i.e. noise factors and the two-step procedure are combined. The average loss function is taken as a suitable performance measure because it has been noticed that maximizing SN ratio does not always guarantee a minimal average loss which is the most desirable result of parameter design. By taking advantages of the unique features of genetic algorithm enhanced by the suitable performance measure, the signal response robust parameter design is achieved within one step. In addition, changing slope target can be accommodated by altering only adjustment factors according to the two-step method. Its application is demonstrated by a "Golf striking design" example to show its effectiveness.