Process Modelling of Gas–Liquid Stirred Tank with Neural Networks

Abstract

Stirred tank reactors are widely used as the major processing unit in environmental and waste management engineering. It also finds its applicability in many chemical, pharmaceutical and petroleum industries. Due to the dynamic nature of the chemical reactions involved and the non-linear functional relationship between the input and output variables, it is difficult to correctly predict a universal empirical correlation for the process variables, i.e. mass transfer coefficient and gas hold-up rate. As such, intelligent modelling using neural networks was adopted in the present experimental work. Experiments were conducted with two types of impeller such as Rushton and curved blade to observe the mass transfer rate and gas hold-up characteristics. The Levenberg-Marquardt optimization algorithm was used to train the neural network so that the error between the desired output and actual output is reduced. The predictive capability of the model has been found satisfactorily and also independent of the impeller type.