Conservative Response Surface Models

The approximate model, obtained from the conventional least square fitting, passes through the observations. Although this can be effective to understand the trend of the observations through the design range, it can cause a serious problem in the convergence of the sequential approximate optimization (SAO) process. In other words, a feasible solution obtained from the SAO is not feasible in the design space in spite of successive updating the approximate models. This can retard the convergence rate of SAO or fail to converge. In order to overcome this difficult, we develop a conservative response surface models.

The formulation of conservative least square models finds  to minimize

while satisfying  or .

References

1.  P.W.M. John, Statistical Design and Analysis of Experiments, 1998, SIAM, Philadelphia

2.  R.H. Myers and D.C. Response Surface Methodology, 1995, Wiley & Sons

3.  M.-S. Kim and S.-J. Heo, “Conservative Quadratic RSM combined with Incomplete Small Composite Design and Conservative Least Squares Fitting”, KSME International Journal, Vol. 17, No.5, pp. 698~707, 2003.