Gas residence time testing and model fitting : a study of gas-solids contacting in fluidised beds.

Abstract

This work is concerned with the effect of vessel geometry on the hydrodynamics of fluidisation of a bed of milled iron oxide. The effect of going from a cold model representative of a typical pilot plant reactor to one simulating a semi-commercial unit is quantified, and various reactor internal configurations on the latter are evaluated. The experimental approach is one based on residence time testing and model fitting with parameter optimisation. A model screening aimed at identifying the most reasonable modelling approach is included, and altogether seven models in two categories are formulated and solved in the dynamic mode. Three of these models are considered novel at present, along with the dynamic solutions to two of the others. The residence time technique involves methane as an inert tracer in air, and continuous analysis of gas withdrawn from the bed via sample probes by a pair of flame ionisation detectors. The process stimulus is governed by a pseudo-random binary sequence, and correlation analysis is employed for noise reduction. A Fourier transform routine, developed from first principles, converts a pair of correlation functions to a process frequency response, and model predictions are compared with the experimental data in this form. Two parameters per model are fitted, and the residual error at the optimum parameter combination provides a means of identifying the best-fitting model. The optimised parameters of this model are regarded as estimates of those of the actual process. Five models compete in the first screening category. Four of these have appeared in the literature in one form or another, and the fifth is novel in that it accounts for axial mixing in the bubble phase by employing multiple plug flow units. This model, referred to as the multiple bubble-track or MBT model, is shown to fit the experimental data better than any of the other models in both bubbling and slugging systems. This suggests that employing multiple plug flow units in parallel for the bubble phase is mechanistically more correct than employing a single plug flow unit. The second screening category is related to the situation in which gas is sparged into an already fluidised bed at some height above the main distributor. The two models in this category are both considered novel, and describe opposite extremes of possible behaviour in one particular sense: one assumes rapid coalescence between grid and sparger bubbles, and the other none at all. The laterally segregated bubble phase or LSBP model emerges as the better process description.The formulation of this model suggests that physically, bubbles from the sparger tend to retain their identity as they pass through the bed. Crossflow ratios estimated on the basis of the best-fitting model in each category point to the existence of a very strong scale-up effect. From the shape of the crossflow profiles it appears that most of the interphase mass transfer occurs in the bottom meter or so of the bed, and it is suggested that grid design is the most significant controlling factor. The presence or otherwise of vertical coils in the bed is shown to have no significant effect on crossflow, and mass transfer between sparger bubbles and the dense phase is shown to be similar to that between grid bubbles and the dense phase. Finally, it is demonstrated that the axial crossflow profile in the bubbling bed is consistent with the concept of an axially invariant mass transfer coefficient based on bubble to dense phase interfacial area.