Artificial Neural Network (ANN) Modeling of Reactive Distillation

As a result of good modeling capabilities, neural networks have been used extensively for a number of chemical engineering applications such as sensor data analysis, fault detection and nonlinear process identification. However, only in recent years, with the upsurge in the research on nonlinear control, its use in process control has been widespread. But there is almost no work which has tested the capacity of Artificial Neural Network (ANN) to model Reactive Distillation (RD). The superiority of ANN was tested for the synthesis of 2-methylpropylacetate from 2-methylpropanol and acetic acid by RD. The effects of various parameters like feed, temperature, reflux ratio, and boil-up ratio, each varied at three levels, were studied. At each combination, the composition of distillate and bottom was predicted by Aspen simulation. An appropriate ANN model was fitted using generated data. A three-layered feed-forward back-propagation model, with 3, 5 and 4 neurons in the layers, was fitted. Log-sig function was used in each layer. The model was found to have high R2 (= 0.993) value and good predicting capability