Optimization of the reaction conditions of ethanol dehydration to ethylene based on rbf neural network simulation

Experiments of ethanol dehydration to ethylene over La modified HZSM-5 zeolite catalyst were carried out according to orthogonal design. The effects of ethanol mass fraction (in feed), reaction temperature, space velocity and catalyst particle size on the yield of ethylene were investigated. Based on the RBF neural network theory, a model simulated the process of ethanol dehydration to ethylene was established by using the above test data as training samples. Then the network system trained was used to predict the effects of various factors and their interactions on ethylene yield. Three-dimensional graphs produced by the network could effectively express the relationships between reaction conditions and catalytic activity. Under the optimal reaction conditions of a reaction temperature of 250 °C, a space velocity of 0.5 h^-1, an ethanol mass fraction (in feed) of 74% and a catalyst particle size of 70 mesh, the simulated ethylene yield was 98.87%, which was very close to the test result of 97.12% with a relative error of -1.77%.