SIMPLIFIED MODELS FOR THE RESIDENCE TIME OF FLUID-SOLID REACTIONS WITH NON-UNIFORM PARTICLE SIZE

Abstract

The estimation of the residence time required to achieve a desired conversion in a fluid-solid reaction demands a great computational effort. This effort is even greater when the solid particles are of different sizes. For this reason, it is of interest to develop simplified models that allow such estimation with less effort. In this work, simplified models are proposed for the case of heterogeneous reactions that can be represented with the shrinking-core model, with controlling resistance in the mass transfer through the film, in the chemical reaction or in the diffusion through the ash. The proposed models were developed for plug flow reactors and for continuous perfectly mixed reactors. These models are of two types: correlations and artificial neural networks. To develop the correlations, functions that meet certain mathematical conditions were selected. These functions were then fitted by regression. In addition, different structures of artificial neural networks were tested until the one that minimizes the test error was found. Both to adjust the correlations and to train the neural networks, data obtained from solving the shrinking-core model considering different controlling stages were used.