An Investigation of Artificial Intelligence Methodologies in the Prediction of the Dirty Amine Flow Rate of a Gas Sweetening Absorption Column

Adaptive neuro-fuzzy and artificial neural networks (ANN) were used for the prediction of dirty amine flow rate of a refinery adsorption column. Gas flow rate and gas pressure were the experimental inputs and dirty amine flow rate was selected as output. Recursive least square and error back propagation algorithm have been applied for training adaptive neuro-fuzzy system and multi layer perceptron neural network. Comparison of prediction errors showed that both models predict dirty amine flow rate with high accuracy and results are in good agreement with the experimental data; nonetheless the neuro-fuzzy model predicted this system better than ANN.