Application of Machine Learning to Catalyst Design and Process Design for Methanol Synthesis

Abstract

For methanol/dimethyl ether (DME) synthesis in temperature gradient reactor (TGR) at low pressure, according to step (1) searching for additives to the copper-based catalyst (Cu/Zn/Al), step (2) creating an appropriate temperature gradient in the catalyst layer, and step (3) adjusting the catalyst components at each temperature, high one-pass conversion was achieved. Each step was accelerated by means of machine learning. Principal component analysis and support vector machine (svm) were used in step (1), orthogonal array, svm and genetic algorithm (GA) were used in steps (2) and (3). In addition, in an internal condensation reactor (ICR) where the products methanol and water are condensed in situ to remove them from the catalyst bed and the reaction equilibrium is shifted, GA was applied to determine the reaction network in order to estimate the condensation rate in ICR. Both the experimental results and all codes for machine learning are unvailed.