Neda Kalantari

The process of converting methanol to propylene is influenced by many parameters. Using smart techniques can be an effective way to investigate variable parameters and find optimal conditions. In this work, optimal design of ZSM-5 catalysts with different combinations of templates and operating conditions in the methanol-to-propylene process was performed using response surface methodology and hybrid artificial neural network-genetic algorithm methods. Objective functions for optimization were methanol conversion and propylene selectivity. Effects of different variables in the dual-responses system, including molar ratios of tetra propyl ammonium bromide (TPABr), cetyltrimethylammonium bromide (CTAB), and Pluronic F127, as well as weight hourly space velocity of feed and process temperature on the performance of catalysts, werestudied both experimentally and theoretically. Modeling results showed that the designed neural network structure for theprocess had superior accuracy compared to the response surface method (RSM) with correlation coefficients of 0.9976, 0.9950, and 0.9946 for training, validation, and testing, respectively. By combining optimal templates, an optimum operating temperature of 420 °C and WHSV of 1h−1 were obtained based on the genetic algorithm applied on a trained artificial neural network to achieve maximum selectivity of propylene and the highest possible conversion of methanol. The optimal catalyst had stable performance under the optimal conditions.