Enhancing performance of modified central composite designs in robust parameter designs

This study proposes an enhanced methodology for modified central composite designs (MCCDs) in robust parameter designs (RPDs) by integrating the principles of minimum aberration designs. The approach ensures high resolution levels for control factors and improved resolution levels for noise factors, while deriving optimal axial distances for fractional and mixed-factor designs. Compared to standard MCCDs, the proposed designs achieve lower or equal values of scaled prediction error variance under specified resolution constraints, verified through theoretical justifications and simulation studies. We evaluated the proposed designs against several benchmarks, such as composite mixed-resolution designs, small composite designs, and classical central composite designs. Using metrics including scaled prediction error variance, slope variance, run size, and prediction coverage probability, we show that the proposed designs maintain desirable geometric and statistical properties while reducing the number of experimental runs. This work provides a cost-efficient and theoretically sound alternative for RPDs applications involving noise factors.