Network Model for Estimation of Hydrodynamics of Packed Column

Different types of packing materials are used to increase gas-liquid contact area in packed columns. The hydrodynamic study of packed column includes the variation of pressure drop and liquid hold up as a function of liquid and gas flow rates. Artificial neural network is an upcoming modeling tool & present work is aimed at optimizing the topology of artificial neural network model for estimation of pressure drop, minimum liquid wetting flow rate and flooding velocity as a function of type and size of packing, liquid flow rate and gas flow rate. The linguistic variable for five types and ten sizes of packing is also incorporated as input parameters with appropriate codes assigned. ANN models with fifteen different topologies have been developed and three models S-50, M-100 and C-10 are shortlisted based on better RMSE values. Further, model M-100 has been observed to be highly acceptable based on comparison of relative error all the data points