ARTIFICIAL NEURAL NETWORKS FOR ESTIMATION OF HEAT TRANSFER PARAMETERS IN A TRICKLE BED RERACTOR USING RADIAL BASIS FUNCTIONAL NETWORKS

The effective heat transfer parameters in a trickle bed reactor (gas-liquid co-current downflow through packed bed reactors), ker, the effective radial thermal conductivity of the bed, and hw, the effective wall-to-bed heat transfer coefficient are estimated for air-water system over a wide range of flow rates of air (0.01-0.898 kg/ms) and water (3.16-71.05 kg/ms) covering trickle, pulse, and dispersed bubble flow regimes in a 50 mm I.D. column employing ceramic spheres (2 mm), glass spheres (4.05 and 6.75 mm) and ceramic raschig rings (4 and 6.75 mm) as the packing materials. Radial temperature profile method is used for the estimation of ker and hw using the experimentally measured radial temperature profile in conjunction with the solution of the two-dimensional pseudo-homogeneous two-parameter model. Radial Basis Functional Networks (RBFN) of Artificial Neural Networks (ANN) are also applied for modeling the effective heat transfer parameters in the trickle bed reactor. The RBFN designed to suit the present system has 7 inputs (four temperatures at four radial positions in the bed, liquid and gas rates, and the ratio of column to particle diameter) and 3 outputs (ker, hw, and the flow regime). The network predictions are found to be in a very good agreement with the experimentally observed values of ker, hw, and the flow regime