Optimization of Cu-Zn-Al Oxide Catalyst for Methanol Synthesis Using Genetic Algorithm and Neural Network as Its Evaluation Function

Abstract

Genetic algorithm was applied to optimize the Cu/Zn/Al ratio of mixed oxide catalyst for methanol synthesis from syngas. A layered neural network was used instead of experiments to evaluate the "fitness" of the catalyst code. This procedure eliminated laborious steps, such as catalyst preparation and activity testing, from the optimization loop. The calculated activity (STY) was almost identical to the original one and could be used as an evaluation function in the genetic algorithm program. The combination of catalyst design by genetic algorithm and activity evaluation by a layered neural network is a promising method for highly efficient catalyst screening.