Modeling and Optimization Strategy for Heterogeneous Catalysis Based on Support Vector Regression and Genetic Algorithm

Abstract

This paper presents a soft computing based heterogeneous catalysis modeling and optimization strategy, namely SVR-GA, for the discovery and optimization of dimethyl ether synthesis on new catalytic materials. In the SVR-GA approach, a support vector regression model is constructed for correlating process data comprising values of input variables of catalyst compositional, operating conditions and output variables of performance of catalyst. Next, model inputs variables are optimized using genetic algorithms (GAs) with a view to maximize the performance of catalyst. Moreover, the SVR model is employed as an approximate model for fitness function in SVR-GA architecture. The SVR-GA is a novel strategy for heterogeneous catalysis modeling and optimization. The major advantage of the hybrid strategy is that modeling and optimization can be conducted exclusively from the historic small sample space data wherein the detailed knowledge of process phenomenology (reaction mechanism, rate constants, etc.) is not required and difficult to get, and simultaneously constructed for the Cu-Zn-Al-Zr slurry catalysts compositional model and kinetic model in the synthesis of DME. Finally, new catalysts, the optimum compositions and optimum preparation conditions leading to maximized CO conversion and DME selectivity were obtained. The optimized solution was verified experimentally to be feasible.