Abstract: Lety=f(\underset\simX, \underset\simZ) （1）where \underset\simX are the Controllable factors of experiment, \underset\simZ are the error factors. y, which has target value m, is the characteristic value of output. Assume that y₁, y₂, …, y_n are the output values of a combination of the controllable factors, and V_e=\sum_i=1^n\left(y_i-m\right)^2,（2） where y₁, y₂, …, y_n are independently and identically distributed with L=E（y_i）, σ=D（y_i)（i=1, 2, …, n）. Obviously, E（V_e）= （m-L）²+σ² （3） From （3）, we can see that the purpose of the design is to find out the best combination of controllable factors so that |m-L| and σ² are minimum. This paper devides controllabe factors into two kinds. One has bigger effect upon σ² than the other. Upon experimental analysis, the first kind which is fixed upon the level of smaller σ² is considered at priority, the second kind which is considered as the faetors of adjust ment minimizes the systematic bias. We give two instances of simulated experiments and get better results.