Artificial Neural Network (ANN)-Aided Optimization of ZSM-5 Catalyst for the Dimethyl Ether to Olefin (DTO) Reaction from Neat Dimethyl Ether (DME)

Abstract

A ZSM-5 catalyst for light olefin synthesis from dimethyl ether (the reaction of dimethyl ether (DME) to olefin, abbreviated as DTO) was developed. In comparison with the reaction of DTO from diluted dimethyl ether (abbreviated as Diluted-DTO), lower light olefin selectivity and shorter catalyst life are the drawbacks of the reaction of DTO from neat (90 vol %) DME (abbreviated as Neat-DTO). After the effects of Si/Al ratio of zeolite, DME concentration, gas hourly space velocity (GHSV) of gas feed, and Si/Al ratio on the activity of calcium-incorporated ZSM-5 were examined, additives were screened using the physicochemical properties of the additive elements and an artificial neural network. The addition of boron or phosphorus to ZSM-5 improved the catalyst life. The catalyst composition such as Si/Al, Si/Ca, Si/P, and Si/B was then optimized for longer catalyst life, using an L9 orthogonal array and an artificial neural network (ANN). Grid search was applied to find the maximum catalyst life. The catalyst life of H−Ca-ZSM-5 (Si/Al = 250, Si/Ca = 20, Si/P = 400, Si/B = 200) was 146 h when Neat-DTO was performed at 803 K and GHSV = 1000 h−1. The life is comparable to that of the catalyst supplied to Diluted-DTO.