Faïçal Larachi*, Lamia Belfares, Ion Iliuta,  Bernard P.A. Grandjean
Department of Chemical Engineering & CERPIC,
Laval University, Québec, Canada G1K 7P4
*Corresp. author. Phone: 1-418-656-3566;
Fax: 1-418-656-5993. email

Industrial & Engineering Chemistry Research; 2001; 40(3); 993-1008.


- In Table 1 and Glossary, coalescence index CI=1 for coalescing liquid and CI=2 for noncoalescing liquid
- In Table 3, J=14
- In Table 4: Input, Output and Fitting Parameters of the ANN-DA Correlation for  EL

Abstract:   The state-of-the-art tools for the evaluation of the macroscopic hydrodynamics of co-current upflow three-phase fluidization are critically evaluated by thoroughly interrogating the broadest fluidization database ever built.  The database is compiled through worldwide conjoint initiatives as a result of a decade of compilation efforts by the groups of professors L. S. Fan (Ohio, USA), S. D. Kim (Seoul, South Korea), G. Wild (Nancy, France) and our ULaval group.  The database represents almost the whole heritage of the non-proprietary data released in the open literature in the field of gas-liquid-solid fluidization (23,000 experiments on bed porosity, liquid, gas and solids hold-ups).  It is dedicated to embrace wide-ranging fluids’ properties, particle and vessel sizes, and operating conditions.  The database contains 55 Newtonian (20,500 data), 19 non-Newtonian liquids (2,500 data), 110 various particles, 17 different column diameters, and includes wall effect ratios Dc/dp and grain sizes ranging, respectively, from 8 to 800 and 0.25 to 15 mm.  Two novel approaches in the field of three-phase fluidization modeling are proposed to reconcile the formidable diversity of patterns and the wide variability of hydrodynamic parameters encountered in this advanced database.  Both of them exhibit a substantial gain in their forecasting ability with respect to the currently known prediction methods.  The first approach relies on the combination of multi-layer perceptron artificial neural networks and dimensional analysis (ANN-DA approach) to derive three highly accurate correlations for bed porosity, and liquid and gas hold-ups.  The second is based on a phenomenological hybrid k-x generalized bubble wake model (k-x GBWM) in which the wake parameters k and x are beforehand extracted by solving an inverse k-x GBW model.  The ANN-DA approach is then applied to correlate k and x in terms of the accessible fluidization input characteristics, and fed into the k-x GBWM to forecast the phase holdups.  The robustness of the proposed ANN-DA correlations and k-x GBWM is assessed and the limitations of the correlations with regard to their generalization capabilities are discussed.
Keywords : Three-phase fluidization, generalized bubble wake model, hybrid modeling, neural network, database, hydrodynamics

You can get the  file (version 28/06/2001) that contains an Excel worksheet simulator to compute phase hold-ups and bubble wake model parameters.

You may also download our 
Excel worksheets simulators for  Trickle-bed or Flooded Bed reactors.

The neural correlation was developped with the software NNFit