Global optimization of cis-cyclooctene with very few experiments

Abstract

Response surface methodology (RSM) is dependent on the design of experiments (DoE) and empirical modelling techniques to find the optimum conditions for a process under the circumstances where by knowledge of the underlying process are unknown in majority. An iterative RSM framework was proposed to model and optimize the catalytic epoxidation of cis-cyclooctene with the use of Cobalt (II)-exchanged zeolite X catalyst. The Gaussian process (GP) regression model, which is a flexible, non-parametric methodology, was selected to the empirical model for RSM to approximate the relationship between the process factors and response as they are capable of providing a high accuracy of approximation and thus exhibiting a higher potential in finding the optimum process conditions. The HSS algorithm was applied to obtain the design points for the evaluation of the design points. Finally, optimization was conducted to obtain the best response value based on the empirical method obtained. The effects of the process factors o the response variables are also illustrated by the response surface plots. Concluding, the developed GP model was applied to the optimization of the cis-cyclooctene epoxidation process successfully.