Hydraulics and Mass Transfer
in Randomly Packed Towers Revisited


Simon Piché, Ion Iliuta, Bernard P.A. Grandjean, Faïçal Larachi*

Department of Chemical Engineering and Center for Research on the Properties
of Interfaces and Catalysis (CERPIC)
Laval University, Québec, Canada G1K 7P4

2001 GLS'5 conference in Melbourne, AUSTRALIA
Chem. Eng. Sci., 56, 6003-6013 (2001)


New robust correlations and mechanistic model of macroscopic fluid dynamic and gas-liquid mass transfer characteristics for randomly packed towers were developed based on first principles, neural network computing and dimensional analysis (ANN-DA). These tools concerned the loading and flooding capacities, the total liquid holdup, the irrigated pressure drop, the local volumetric liquid-side, kLa, and gas-side, kGa, mass transfer coefficients, the overall volumetric, KLa and KGa, mass transfer coefficients, and the packing fractional wetted area. Validation of these tools was performed by interrogating a broad experimental database including over 10750 measurements published in the literature over the past seven decades. The fully-predictive mechanistic model proved powerful in forecasting the tower hydraulics below the loading point without requiring any adjustable parameter. On the other hand, the ANN-DA correlations proved highly powerful in correlating the tower fluid dynamics and gas-liquid inter-phase mass transfer regardless of the operating flow regime. These approaches were also benchmarked with respect to the comprehensive Billet and Schultes (1999) phenomenological approach and the classical Onda et al. (1968) mass transfer correlations.

Keywords: Randomly packed towers, hydrodynamics, gas-liquid mass transfer, neural network models

You can get the  packedbedsimulator.zip  file that contains an Excel worksheet simulator to compute pressure drop, liquid holdup, loading/flooding capacities, film and overall volumetric mass transfer  coefficients.

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

The neural correlation was developped with the software NNFit