Of the 600 starting designs, 580 were used to generate meta-models, while 20 designs were employed as designs of control to verify the affordability of the response surfaces. These 20 were chosen in order to get the right information on the entire range of existence of the output variables.