Heat transfer in fixed bed: a model non-linearity approach

Studies have been carried out on the heat transfer in a packed bed of glass beads percolated by air at moderate flow rates. Rigorous statistic analysis of the experimental data was carried out and the traditional two parameter model was used to represent them. The parameters estimated were the effective radial thermal conductivity, k, and the wall coefficient, h, through the least squares method. The results were evaluated as to the boundary bed inlet temperature, T-o, number of terms of the solution series and number of experimental points used in the estimate. Results indicated that a small difference in T-o was sufficient to promote great modifications in the estimated parameters and in the statistical properties of the model. The use of replicas at points of high parametric information of the model improved the results, although analysis of the residuals has resulted in the rejection of this alternative. In order to evaluate cion-linearity of the model, Bates and Watts (1988) curvature measurements and the Box (1971) biases of the coefficients were calculated. The intrinsic curvatures of the model (IN) tend to be concentrated at low bed heights and those due to parameter effects (PE) are spread all over the bed. The Box biases indicated both parameters as responsible for the curvatures PE, h being somewhat more problematic. (C) 2000 Elsevier B.V. Ltd. All rights reserved.