Development of hybrid model and optimization of operating conditions for a styrene monomer reactor

Abstract

In the industrial styrene monomer process, maximizing the production rate is needed since the current demand of styrene monomer is more than the production rate. In addition, this plant consumes a large amount of energy. In order to improve this problem, the first principle model has been developed. Although the first principle model of the radial flow reactor gave a reasonable prediction from the industrial point of view, the nature of the process presented some difficulties in the prediction of catalyst activity. The lack of internal or intermediate measurements of the reactor represented a limitation for the detailed model validation. In addition, the lack of experiments of catalyst activity caused the uncertainty of parameters in mathematical equations. Thus the first principle model was not enough to represent changes of catalyst activities. In light of these difficulties for developing the first principle model, an alternative model of the styrene monomer reactor has been developed combining the first principle model with neural networks. Neural networks have given more accurate deactivation factor, and the first principle model has reinforced the insufficient performance owing to lack of data to use neural networks. In this study, a hybrid model, in which the first principle model and neural network model are combined, has been developed for an adiabatic radial-flow styrene monomer reactor. This model has been analyzed via computer simulation to examine important operating variables. The operation performance has been measured in terms of catalyst utilization. Finally, new operating conditions have been proposed to maximize the profit.