Optimization of methanol production via CO2 hydrogenation: comparison of sampling techniques for process modeling

Abstract

Abstract This paper compares performances of three different sampling techniques in representing nonlinear response surface for optimization problem, with five decision variables. Three studied sampling techniques are Central Composite Design (CCD), Box – Behnken Design (BBD) and Latin Hypercube Sampling ( LHS). Simulation – based optimization of methanol production via CO2 hydrogenation process was used as a case study. The objective of this optimization problem is to minimize the methanol production cost. The nonlinear model ( the objective function) represents the relationship between operating conditions and methanol production cost. The results show that the response surface of this case study is trend to be the third order model ( with R – squared value greater than 0.97) for all sampling techniques. BBD is the most suitable sampling technique in this study, resulting in the percent error less than 1.2. The LHS shows comparable performance with BBD if it has suitable number of sample points.