Simulation And Optimization Of Reactive Distillation

Reactive distillation is a process with chemical reaction and separation. It has attracted many researchers and has been used widely for the virtues of high selectivity, low capital investment and little energy consumption. Because of the violently interactions between reaction and separation, many models have shortcomings when simulating reactive distillation and there are few methods for optimizing the process. In this thesis, artificial neural network (ANN) models are presented to simulate reactive distillation with no consideration of the specific process and mechanism in the columns. The results show that good effects can be achieved using ANN models with high simplicity. More data are produced by software of ASPEN to make up the shortage of experimental data when setting up the ANN models. By comparing the different predictive results of different ANNs to the training set and testing set, the transfer function and nodes of hidden layer are ascertained, thus the optimal structure of ANN is established. Genetic algorithm is used broadly in engineering optimization with the excellent adaptive character and the distinguished ability in search of global solution. In this thesis, a multi-objective genetic algorithm based on different weights is presented. In company with the ANN models, the multi-objective genetic algorithm is used to optimize the process of reactive distillation. Results optimized show that the proposed method is feasible and has high precisions. Processes of catalytic distillation for hydrolyzing ethyl acetate and reactive distillation for synthesizing ethyl acetate are taken as examples to demonstrate the proposed simulation and optimization methods.