Design and Development of Cu-Zn Oxide Catalyst for Direct Dimethyl Ether Synthesis Using an Artificial Neural Network and Physicochemical Properties of Elements

A hybrid catalyst consisting of Cu-Zn-X oxide and (gamma)-alumina was prepared for single-stage dimethyl ether synthesis. An artificial neural network (ANN) was applied to find the effective additives for the hybrid catalyst. For the training of ANN, elements (X)-B, K, Nb, Re, Cd, Ce, Sm, and Tl-were selected. The activity change with time on stream of the hybrid catalyst was fitted to a generalized power law equation (GPLE). The resultant GPLE parameters and the physicochemical properties of the eight elements were used as training data for ANN. After the training, the trained ANN was used to predict the activity of the hybrid catalyst containing various X elements as Cu-Zn-X. Elements Al, Ti, V, and Nb were predicted as promising, and the composition of hepternary oxide catalyst was optimized by the combination of design of experiment, ANN, and grid search. The catalyst with the optimized composition showed stable and high activity.