Introduction

Biohydrogen (BioH₂)
is a 2^nd and a 3^rd generation biofuel that presents the
advantage to be renewable, sustainable and environmentally friendly because it generates
no toxic by-products and greenhouse gases when burnt. Second generation BioH₂
can be obtained by anaerobic digestion, also denoted by dark fermentation, of agricultural and food
waste, provided methanogenesis is prevented [1]. However,
the production of BioH₂ is necessarily coupled with that of volatile
fatty acids (VFA), while both soluble hydrogen and VFA are able to inhibit BioH₂
production; this means that these must be removed as far as they are produced. Mixing
plays, therefore, a key role, as a three-phase system is involved: namely, an
aqueous solution enriched in soluble sugars and fatty acids, a suspended lignocellulosic waste and a gas phase that is composed of
CO₂ and H₂. Agitation must enable the suspension of
organic waste, homogenize local concentrations, and enhance hydrogen desorption
at the same time [2, 3], but this is limited because high shear may disturb the
biological process [4].

The aim of this project is to
set up an anaerobic submerged membrane reactor (AnMBR)
devoted to biohydrogen (BiOH₂) production
using dark fermentation process applied to lignocellulosic
waste. The AnMBR consists of an unbaffled
stirred tank reactor equipped with a two-stage radial impeller, coupled with an
external hollow fibre membrane module placed in a forced circulation loop. The
objective of the membrane is to remove soluble VFA when fresh substrate is
added. The membrane is ?submerged?, which means that the permeate
is pumped back through the membrane, so that the stirred tank can be operated
at atmospheric pressure to avoid inhibition by soluble H₂. As a result, the
position of the inlet and the outlet of the forced circulation loop in the tank
must be defined carefully, for example to avoid the constriction of the inlet
by lignocellulosic materials, or without inducing
concentration gradients that could impair reaction rates.

Materials and methods

In this work, mixing has been
studied both experimentally and numerically. In the experiments, a laboratory-scale
anaerobic
submerged membrane reactor (AnMBR with a 5-L
stirred tank) has been used. As the viscosity of the aqueous
solution is close to that of water, vortex formation was studied first by flow
visualization using two cameras, as an unbaffled tank
was preferred to avoid the presence of stationary solids close to the baffles;
particle image velocimetry (PIV) was used to analyse
the single-phase flow for different axial positions of two impellers as a
function of rotation speed. Macro-mixing in the stirred tank and the AnMBR was analysed using both a conductivity tracer
technique and dye injection coupled to flow visualization, so that mixing and
circulation times could be defined in the tank and in the loop. Solid
suspension was also investigated as a function of rotation speed using flow
visualisation as function of the amount of solid (from 5 to 10 g/L) and
impeller position.

In parallel, Computational
Fluid Dynamics (CFD) has been applied to establish a model able to capture the details
of the flow field, including velocity and turbulent power dissipation, but also
the spatial distribution of species concentrations and of the phases in
multiphase flows. The 3D numerical simulations were carried out using the commercial
software package, PHOENICS^® (CHAM,
UK). Parallel computing was
used to minimize computation time for predicting the flow structure in the stirred tank. This
solves the Reynolds Average Navier-Stokes equations,
using a modified k-ε model for closure. Standard single-phase simulations were
driven to simulate the dispersion of a passive tracer, a VOF (volume of fluid)
approach was applied to predict vortex formation and a two-fluid Euler-Euler approach
was retained to simulate solid suspension as a function of reactor geometry and
operating conditions. The objective was to validate the physical models using
experimental data and to establish a numerical methodology that could be used in
anaerobic fermentation during which all the measuring techniques described
above cannot be used.

Results

The experimental results show
that a laminar flow pattern prevailed in the membrane unit and remained
unaffected by the permeate flow. Conversely, turbulent flow prevailed in the
stirred tank reactor and mixing time decreased roughly as the inverse of the
impeller rotation speed. The value of the dimensionless mixing time was close
to 30. As a rough approximation, the bioreactor was shown to correspond to a
perfectly mixed tank with space time t, connected to a
laminar loop of space time t/10. Figure 1 shows that the vortex
formation in the stirred tank was observed only when rotation speed was between
150 and 200 rpm, while the suspension of straw grinded at a
characteristic size of 2 mm started about 150 rpm and was complete
above 250 rpm. This was sensitive to the position of the bottom impeller, but slightly
influenced by the amount of straw between 5 and 10 g/L.

Fig.1. Pictures of vortex
formation at rotation speed the impeller equal to 200 rpm (A) and of the evolution of solid suspension
with 100 rpm (B) and 150 rpm (C)

CFD simulations gave access to the critical features of the
local flow and were shown to be able to
predict correctly the formation of the vortex with the VOF approach and the
suspension of the solid phase using the algebraic slip model, as illustrated
in Figures 2(A) and 2(B). In single-phase flow, the details of the flow
field deduced from the computations were compared to the local velocity field
obtained by PIV and a fair agreement was reported. This provided a better
understanding of the influence of the position of the two impellers. For mixing
analysis, the simulations using transient flow and the k-e turbulence model
showed a fairly good agreement with the experimental data in the tank, using
the transient evolution of species concentrations. Finally, the axial positions
of the two impellers could be optimized, accounting simultaneously for
distributive and dispersive mixing and the opposite objectives to homogenize
concentrations, prevent vortex and enhance solid suspension; the same stood for
the positions of the inlet and of the outlet of the circulation loop as a
function of rotation speed.

Fig.2. MOFOR (Moving Frame of
Reference) simulations coupled to VOF (Volume of Fluid) approach
to investigate vortex formation at 200 rpm (A) and to solid suspension using the Algebraic Slip model after 1 s at 200 rpm (B)

Conclusion

As a conclusion, the results confirm the potential
interest to couple experiments with CFD simulations for the optimization of the
operating conditions and of the design anaerobic submerged membrane reactor in
terms of mixing, mass transfer and permeate flow. In the future, this approach seems
also able to provide useful guidelines for the scale-up industrial bioreactors
and to predict any modification that could be difficult to validate
experimentally. The perspectives of this work are,
now, to implement a simplified biokinetic model of
BioH₂ production by dark fermentation in the CFD code using the
transport equation of additional scalar variables, first with a simple
substrate such as glucose for witch metabolic pathways have been already
described in the literature.

References

[1] A. Pandey,
C. Larroche, S. C. Ricke,
C.-G. Dussap, E. Gnansounou,
Biofuels: alternative
feedstocks and conversion processes, Oxford UK: Academic
Press, 2013.

[2] J. Lindmark, E. Thorin, R. Bel Fdhila, E. Dahlquist, Effects of mixing on the result of anaerobic
digestion: Review. Renewable and Sustainable Energy Reviews, vol 40,  pp
1030-1047, 2014.

[3]
M. Eng, A. Rasmuson, Large eddy simulation of the influence of solids on macro instability frequency in a stirred tank. Chemical Engineering Journal, In Press, Accepted Manuscript, Available online 26 August 2014.

[4] S.I. Padmasiri, J. Zhang, M. Fitch, B. Norddahl, E.Morgenroth, Lutgarde Raskin, Methanogenic population dynamics and performance of an anaerobic membrane bioreactor (AnMBR) treating swine manure under
high shear conditions. Water Research, vol 41, pp 134-144, 2007.