Modelling of packed column using artificial neural networks

Abstract

Packed column is an established gas liquid contacting device with many industrial applications. Estimation of liquid hold up and pressure drop as a function of gas and liquid flow rates for single as well as two phase flow are of significance from operational as well as design point of view. Various empirical models have been suggested for the same. Artificial Neural Networks (ANN) are gaining importance in modelling of multivariable, non-linear relationship with high accuracy and even in presence of inadequate data. In present work multi layer perceptron (MLP) ANN with GDR based learning have been developed for estimation of liquid hold up and pressure drop for packed column. The ANN models thus developed are observed to be of good accuracy level, both for training and test data set.