THE DEVELOPMENT OF Cu-BASED OXIDE CATALYSTS FOR METHANOL SYNTHESIS BY RADIAL BASIS FUNCTION NETWORK AND GENETIC ALGORITHM

摘要

We applied RBFN as a fitness function of GA to develop Gu Zn-Al-Sc catalyst for methanol synthesis. The calculation time off RBFN was very short. RBFN evaluated the activities of Cu-Zn-Al-Sc catalysts for methanol synthesis in the whole search space from the small number of data dispersed randomly. In order to construct reliable RBFN for Cu-Zn-Al-Sc catalyst, at least 69 training data was necessary. The RBFN after training by 92 experimental data obtained by reaction in HTS reactor was successfully used as black-box function to calculate the activity to evaluate the fitness in GA program. Then, evolutions were repeated rapidly and easily. The number of experiments in optimization by GA with RBFN was smaller than that by GA without RBFN. The experimental labor and time was reduced. addition, the optimum catalyst found by GA with RBFN showed higher activity than the optimum catalyst found by GA without RBFN. Thus, GA combined RBFN is robust tool foreffective catalyst development.