Artificial neural network aided design of catalyst for propane ammoxidation

Abstract

An artificial neural network (BP network) is applied to design a VSbWSn (P, K, Cr, Mo)/SIAL catalyst for acrylonitrile synthesis via propane. The conversion of propane and selectivity of acrylonitrile can be calculated as functions of the catalyst components by the BP network to be trained. After training, the network can simulate the catalytic system very well. If one takes the conversion of propane and selectivity of acrylonitrile as the two sub-objectives, a model for this catalytic system can be given as:Max(y1=X C3), Max(y2=S ACN), y= F(W,X in), 0≤y1≤1.0, 0≤y2≤1.0. A better catalyst could be found through optimization for propane ammoxidation to acrylonitrile. The best yield of acrylonitrile is 55.0%, which is higher than those reported in the literature.