Modeling and Simulation of an Industrial Ethylene Oxide (EO) Reactor Using Artificial Neural Networks (ANN)

Abstract

In the present work, a one-dimensional heterogeneous model was used for dynamic simulation of an industrial fixed-bed catalytic ethylene oxide (EO) reactor in the presence of long-term catalyst deactivation. In order to determine the level of optimum ethylene dichloride (EDC), a multilayer perceptron (MLP) neural network was used. In addition, the effect of inlet gas velocity on EO mole fraction of gas and solid phases was investigated. The model validation was carried out by comparison of model results with corresponding industrial conditions and over a period of three operating years. A good agreement was found between the simulation results of the dynamic model and historical process data. The error of simulation was found to be less than 5%. The results of the artificial neural network (ANN) modeling showed that the maximum selectivity occurs in the range of 0.37–0.42 ppm of EDC. Also, it was observed that with decrease of gas velocity the difference between the EO mole fraction of gas phase and solid phase increases. This behavior was attributed to the distinct resistances of kinetically controlled and the mass transfer resistance of gas film around the catalyst pellets.

Distillation of Bulk Chemicals

“…To moderate the reaction, ethylene dichloride can be used [73,74]. The use of air instead of oxygen requires a much larger gas purge and about 1.5 times more catalyst.…”

“…Selectivity has increased from w65% in the early 1950s to current 85e90% values. Catalyst lifespan is 2e3 years, during which the ethylene conversion decreases by approximately 3e4% per year due to deactivation [73]. EO processes are licensed by the various companies that also sell catalyst.…”

Heterogeneous Reactor Modeling of an Industrial Multitubular Packed‐Bed Ethylene Oxide Reactor

“…The complete oxidation of ethylene is inhibited by EDC, resulting in partial oxidation; therefore, the amount of EDC injection significantly affects the reactor performance. EDC injection sensitivity was studied by Rahimpour et al 13 using artificial neural networks (ANNs) in developing a heterogeneous one‐dimensional transient model. These modeling studies of EO reactors have led to a significant improvement in the understanding of the reaction dynamics and identified numerous performance‐affecting variables.…”

“…As ethylene oxidation is highly exothermic, a fluidized bed should be an appropriate choice; however, problems of abrasion, sintering, and much lower selectivity discourage its commercialization 20, 21. Therefore, all commercial EO processes currently employ fixed‐bed multitubular reactors 13, 22. The design features of fixed‐bed multitubular reactors for EO like tube diameter, number of tubes, tube length, and/or catalyst geometry (Tabs.…”

“…The reactor system and the process flow diagram of a conventional EO production process are shown in Fig. 1 13, 22. In a multitubular packed‐bed reactor, the tubes are packed with Ag catalyst pellets and boiling water circulates in the shell side to remove the exothermic reaction heat.…”