Neural Network Technique in the Study of Selected Chemical Engineering Unit Operations Data using MATLAB

The concept of Artificial Neural networks was of McClloch and Pitts in 1943 and since then it has been studied in details by scientists and engineers alike. This is a study of the use of artificial neural network in analysis of selected chemical engineering unit operations. In this paper several networks were developed and trained for three different unit operations. This paper deals with the training of neural networks to perform predictions of several chemical unit operations. The feedforward neural network was trained to model the bubble point temperature and pressure of the water ethanol-water vapor-liquid equilibrium system. It was found that the neural network predicted values with high accuracy. Focused time-delay neural network was used to model and predict the change in concentration of the batch saponification reaction of ethyl acetate. The response of the network in one step in time ahead predictions was quite accurate. The dynamics of a CSTR with a cooling jacket was also modeled with the NARX neural network. The NARX model developed gave multi step on time predictions with enormous aplomb