An Application of Nelder-Mead Algorithm in Response Surface Methodology: CCD

Abstract

A novel method using the Nelder-Mead algorithm is proposed to be used instead of the first order model in moving the experiment in the response surface methodology toward the neighbor of the optimum. A second order model similar to the second order model in the CCD is constructed to estimate the optimum design point and the optimum response. From the simulation using five published test functions and five different normal generators, it can be concluded that the proposed method outperforms the traditional CCD in terms of the number of experiments, the MAPEs of the estimated optimum design point and estimated optimum response and the coverage probabilities.