Professional Documents
Culture Documents
Seminar Final
Seminar Final
beds”
Seminar Report Submitted in Partial Fulfilment of the Requirement
for the Award of
Degree of
(2020-22)
By
Virendra Kumar
(20CHE232)
(FIRST YEAR)
2.2 Meshing....................................................................................................................................... 14
2.5.3 Result comparison for random and idealized particle arrangements ................................... 25
2.5.4 Wall effects on Sherwood Number and local mass transfer coefficient .............................. 26
3. SUMMARY ...................................................................................................................................... 30
3. REFERENCES ............................................................................................................................. 31
Table of Figures
Table of Tables
1. INTRODUCTION
For optimal design and proper functioning of packed columns for such mass transfer
processes accurate prediction of mass transfer coefficient is essential. Over the years
engineers have been using the correlation published in literature such as Ranz and Marshal,
Gunn, Frossling for getting average mass transfer coefficient in packed beds. But these
correlation which are empirical or semi-empirical are based on experiments which usually
ignored wall effects, end effects, radial variations and which completely ignored the local
effects. Such correlation are only able to estimate average values of mass, momentum and
heat exchange parameters and they are not able to predict spatial distribution of these
parameters and are not valid for systems having poly-disperse and/or non-spherical particles.
For many years modelling of fixed beds have been developed and used but in last two
decades with the advancement of computing technologies CFD has played a vital role in
understanding the processes like heat transfer, hydrodynamics and mass transfer in packed
beds. Conventional approaches for modelling packed beds like pseudo-homogenous or
heterogeneous models does not explain the phenomena at the local or particle scales as they
treat catalyst bed as homogenous continuum. Following are Assumptions and limitations of
conventional models:
5. Flow and mass transfer around individual particle/local effects are neglected
For packed beds with small N (tube to particle diameter ratio) where local and wall effects
play a big role in determining performance of reactor conventional method cannot accurately
determine the performance of the reactor. For small N reactors, CFD is an important tool and
that is why in late 1990s growing number of researchers developed methods for investigating
physical phenomena which takes place in packed beds by applying 3-D CFD. Computational
fluid dynamics (CFD) studies of mass transfer in packed beds is able to address majority of
the limitations of conventional methods. Following are some of the advantages of CFD
studies in packed beds:
Approaches for CFD of packed beds are classified in mainly two parts: Porous Medium
Model (PMM) approach and Particle Resolved Computational Fluid Dynamics (PRCFD)
approach.
PMM treats fixed beds as a continuum and presence of particles is included by adding source
terms accounting for viscous and inertial loss resistance in the incompressible N-S equations.
PMM is able to capture average pressure/temperature change but cannot reproduce secondary
flows and flow anisotropies such as separation, oscillation and vortex formations. It cannot
reproduce radial variations in axial flow which is very important in reactor modelling. Figure
1 clearly express the time and length resolution of PMM model in which catalyst pellet and
fluid flowing through the pores is taken as continuum. PMM model mainly takes into account
the averaging effect throughout the continuum rather than taking into account local effects,
wall effect and end effects. Since PMM models are not much used in recent years because of
advanced methods like PRCFD, further discussion is focused only on PRCFD method.
Department of chemical engineering, ICT, Mumbai Page 6
Computational Fluid Dynamics of Mass Transfer in Packed Beds
data analysis and data visualization. This last step of data extraction , visualization and
analysis is called post-processing.
Figure 3 Overall Workflow and step-wise procedure of PRCFD (Source Jurtz et al.
2018)
2. LITERATURE REVIEW
1. Reconstructive methods
2. Idealized arrangement of particles
3. Random Arrangements of particles
Those methods which reconstruct replicate the arrangements of real beds with the help of
some imaging techniques like MRI or X-Ray tomography are known as reconstructive
methods. Wang et al., (2001),Baker, (2011) and Yang et al., (2013)have used these methods
effectively for generation of packing data. Following are sequential steps used in this method:
1. Real Tubes/cylinders are filled with packing for creation of randomly packed beds.
2. With the help of 3-D imaging/reconstruction techniques actual shape, position and
orientation of each particle are stored in form of voxel data.
3. Extraction of 3-D voxel data to a surface description or directly to volume
representation of numerical domain.
4. Mesh is created from the 3-D voxel data.
Drop and roll technique and sequential arrangements/deposition of particles based on pre-
defined seed-pellets/clusters are two main techniques under SDA. Collective rearrangements
methods are fundamentally statistical methods like Monte-Carlo based methods where
specified number of particles are randomly placed and then moved under influence of gravity
to minimize the configurational potential until a ground state called mechanical equilibrium is
attained. Atmakidis and Kenig, (2012) and Deen and Kuipers, (2014) have used Monte-Carlo
based method called ballistic deposition method for packing generation. Drawback of Monte-
Carlo methods which don’t take into account particle collision is that these algorithms lead to
non-physical particle arrangements for non-spherical particles.
𝑑𝑉𝑝
𝑚𝑝 = 𝐹𝑠 + 𝐹𝑔 + ∑ 𝐹𝑐𝑜𝑛𝑡𝑎𝑐𝑡 + ∑ 𝐹𝑐𝑜𝑛𝑡𝑎𝑐𝑡
𝑑𝑥
𝑛𝑒𝑖𝑔ℎ𝑏𝑜𝑟 𝑝𝑎𝑟𝑡𝑖𝑐𝑙𝑒 𝑛𝑒𝑖𝑔ℎ𝑏𝑜𝑟 𝑤𝑎𝑙𝑙
𝑉𝑝 is velocity of particle
𝐹𝑐𝑜𝑛𝑡𝑎𝑐𝑡 is contact force between particles and between particle and wall
According to Jurtz et al., (2019) DEM can be classified into two different approaches:
In hard sphere framework particles are considered elastic and collision are instantaneous.
This approach is applicable mainly when system is not dominated by multi-particle contacts
and recently have been used by Boccardo et al., (2015) to generate beds with spherical,
cylindrical and trilobe particles.
More general approach is soft sphere approach where contact forces are proportional to
overlap of particles, particle material and geometrical properties and contact forces can be
modelled by approaches like linear spring model, non-linear spring-dashpot models etc.
DEM Figure 5 explains the detailed procedure to carry out DEM with the help of a flow
chart. Recently Bale et al., (2017), Bale et al., (2018), Bale et al., (2019) have successfully
used DEM method for achieving good results for void fraction and creating realistic particle
packed beds.
Figure 5 Flow Diagram for DEM method (Source: Ma, Wang, Ren and Shi (2018))
For voxel-based methods particles are characterised by number of voxels and during
collision of particles a restitution force proportional to overlap volume is calculated based
on linear spring-dashpot model. Caulkin et al., (2015) used this method and achieved good
results for void fraction and particle orientation compared to NMR/XMT measurements.
Niegodajew and Marek, (2016), Boccardo et al., (2015), Niegodajew and Marek, (2016)
used surface mesh-based particle collision models in which particles are represented by
vertices, edges and faces and restitution forces are calculated based on intersection of these
representations.
2.2 Meshing
Above discussed packing generation methods gives information either on particle position
and orientation or voxel representation of the bed geometry. After generation of CAD
model of fixed bed, depending upon the type of numerical solver geometrical
representation is discretized into different mesh types:
1. Structured Mesh
2. Unstructured Mesh
For simple geometries curvilinear structured mesh as shown in figure 6(A) are used but for
fixed bed geometries unstructured mesh as shown in figure 6(B-F) are used.
For Lattice-Boltzmann Method cartesian cut-cell type meshes (Figure 6D) are used and for
FEM commonly tetrahedral (6B) or hexahedral (6C) type mesh are used. For FVM based
solvers all mesh type except Cartesian cut-cell approach are used.
1. Global Gaps Method in which all particles are shrunk by a certain amount.
2. Global Overlaps Method in which all particles are inflated by certain amount.
Recently Bale et al., (2017), Bale et al., (2018), Bale et al., (2019) have successfully used
global modification approach to generate realistic packed beds with prefixed void fraction.
Major disadvantage of global modification approach is that deviation in overall radius
result in large deviation in pressure drop/porosity which can be further dealt by local
modification methods.
Figure 8 Global and Local Modification Approaches (Source Jurtz et al., (2019))
Surface Re-meshing algorithm automatically detects proximate faces/edges and create a small
gap which can be filled with volume cells of good quality. Both local bridges and local caps
method shows good agreement with experimental/correlational values. Eppinger et al.,
(2014), Wehinger et al., (2015) and Wehinger et al., (2016) have used local caps method for
shapes that are not cylinder-like as trilobes and the results generated agree very well with
experimental data. Local bridges method also gives good result for pressure drop and heat
transfer predictions but choice of thermal conductivity of bridges is matter of further
research. Following table shows a list of research paper for different approaches, mesh types,
contact modification and different solver:
1. Reε < 1: Viscous flow regime. Pressure drop is a linear function of the interstitial
velocity.
2. 10 ≤ Reε ≤ 150: Steady laminar inertial regime. Pressure drop is a non-linear function
of the interstitial velocity, and boundary layers are forming.
3. 150 ≤ Reε ≤ 300: Unsteady laminar inertial regime. The flow shows oscillating
behaviour in the wake within the voids.
4. At Reε = 250, laminar vortices start to form.
5. Reε > 300: Turbulent flow. Characterized by an unsteady and chaotic flow
1. DNS
2. LES
3. RANS
Figure 9 Different Turbulence Models A) DNS B) LES C) RANS (Source Jurtz et al.,
(2019))
There are mainly three types of methods to solve these set of governing equations:
1. FEM
2. FVM
3. LBM
FEM is very rarely used for CFD of mass transfer in packed beds and in previous decade only
Motlagh and Hashemabadi, (2008) used FEM for simulation of fluid dynamics and heat
transfer for fixed beds with N=2 and used only 10 cylindrical particles.
Figure 10 Volume Cells for finite volume method. (Source Jurtz et al., (2019))
As we can see in our tables of the various publications majority of the work done is by using
FVM method. FVM is an ideal method which is most popularly used in works of fixed bed
because of its beneficial characteristics like non-invasive boundary conditions and FVM
conserves mass, energy and momentum by its definition. FVM is more popular than other
methods because of its applicability on unstructured mesh and the development in the domain
of automated meshing algorithms in the last decade.
1. D/d ratio
2. Schmidt number
3. Reynolds number
4. Wall effects
After the discretization of NS equations, set of algebraic equations are solved by various
methods of linear algebra and concentration and velocity profiles are generated. Taking the
reference of Bale et al., (2018) following solution strategy is adopted in majority of the recent
publications which use DNS as mode of simulation. Bale et al., (2018) separately solves
continuity, momentum equations and species transport equations. In the first step only
continuity and momentum equation is solved and in the second step species transport
equation is solved by assigning saturation values once the convergence of continuity and
momentum is achieved. Following are typical assumptions and boundary conditions used:
Boundary conditions(hydrodynamics)
Once the concentration profiles generated both local and Overall Sherwood number can be
calculated as follows.
Out of all these quantities only 𝐶𝐴,𝑜𝑢𝑡 is calculated by simulation rest all parameters are
known values. After getting 𝐶𝐴,𝑜𝑢𝑡 , overall mass transfer is found out from the above given
formula and then overall Sherwood number can be found out.
𝜕𝐶𝐴,𝑓
𝑘𝑙 𝑑 𝜔 − ∬𝑆 (𝐷𝐴,𝑓 )𝑑𝑆
𝜕𝑟
𝑆ℎ𝑜 = where 𝑘𝑙 = where 𝜔 = 2 and 𝜑 = (𝐶𝐴,𝑠 − 𝐶𝐴,𝑓 )
𝐷𝐴,𝑓 𝜑 4𝜋𝑅𝑝
Calculation of 𝐶𝐴,𝑓 is the biggest concern which is addressed by Deen and Kuipers, (2014a) by
defining locally defined bulk concentration given by following formula
𝑟
Where 𝑔(𝑟) = exp [−(𝑅 )] , Where 𝑙 = 2 ∗ (1 + 𝛽)𝑅𝑝 , where it is observed that volume
𝑝
concentration steadily decreases from the surface of any particle. 𝑙 is the dimension of cube
around any spherical particle and Value at which the volume averaged concentration
converges with respect to β is considered as the bulk concentration around that particle.
𝑙 = 2(1 + 𝛽)𝑅𝑝
𝑘𝑜 𝜗
Where 𝐽′ = (𝐷 𝑓 )0.58 is the 𝐽′ factor dimension number which can be compared with
𝐿 𝐴,𝑓
experimental values. Regular arrangements have higher J factor because void fraction of
regular arrangements FCC (32%) and BCC (26%) is lower than Wilson 1963 (43.1%),
Wilson 1966 a (43.6%) and Wilson 1966 b (40.1%) which results in more intensified and
uniform mixing in FCC and BCC arrangements resulting into higher mass transport. Close
packing in regular arrangements with respect to experimental setup results into higher
interstitial velocity and high turbulence in comparison to the experiment. High velocity and
turbulence are more favourable to higher mass transfer coefficient which further results into
higher J factor in comparison to the experimental setup. On the other hand for random
arrangements prefixed void fraction equal to experiments are set which give almost identical
results for both experiments and simulations. Except Klöker et al., (2005) recently almost all
the publications in this domain have used mainly random particle arrangements.
2.5.4 Wall effects on Sherwood Number and local mass transfer coefficient
Figure 14 Effect of wall and Schmidt number on Sherwood number (Bale and Joshi
(2019))
Figure 15 Effect of wall on mass transfer coefficient (Deen and Kuipers, (2014))
Figure 14 overall Sherwood number increases with Reynolds number mass flux increases
which decreases the average bulk concentration inside the packed bed. Hence the difference
between surface concentration and average bulk concentration increases with increase in
Reynolds number which increases overall Sherwood number. Second observation is that
slope of the graph between Sherwood number and Reynolds number is low for low Schmidt
number and high for higher Schmidt number. Reason for this is that at Sc and Re <=100
diffusion is dominant and increasing mass flux does not produce much difference between
surface concentration and average bulk concentration resulting into lower slope and converse
is also true. Figure 15 clearly shows that near the walls local mass transfer coefficient are
higher because of a preferred pathway for the fluid which increases local Reynolds number
around particles near wall resulting into higher local mass transfer phenomena.
From the above graph we see that till D/d=10.8 slope of Sherwood number
versus Reynolds number increases but after that there is no change in slope which shows
that effect of confining wall on mass transfer process exists till D/d=10.8 beyond which
wall effect ceases to exist. From the above analysis it can be inferred that D/d has a
significant effort on the Sherwood number and hence on the performance of the bed.
Figure 16 shows that graph of void fraction with the axial distance of a packed bed
can be divided into three parts. Left most part shows the effect of flat bottom on the packing
fraction, right most part shows the effect of top layer packing and middle part represents the
steady void fraction in the middle portion of a packed bed. Due to flat bottom large interstitial
portion is vacant and hence fluctuation in void fraction results into fluctuated mass transfer
through those voids. Similarly, at the top part absence of a layer of spherical particles results
into loose packing at the top which results into sudden increase of void fraction. Since mass
transfer is a function of void fraction and because of these end effects overall mass transfer
and hence the Sherwood number get affected. Another observation is that for D/d<=10.8
fluctuations of void fraction even in middle portion of bed are large which results into
fluctuations in Sherwood number as well.
D/d=4, Re=100
D/d=11, Re=100
Increase of spread
Figure 18 Distribution of Sherwood number with Reynold number (Bale et al., (2019))
Above figure clearly shows that as Reynold number increases spread of the Sherwood
number increases because the variance of distribution of Reynold number at different points
with average Reynold number increases resulting into same trend by Sherwood number.
Increase of spread shows that difference between lows and highs of Sherwood number
increases with increasing Reynold number.
3. SUMMARY
CFD of mass transfer in packed beds is an important alternative for costly experiments and
very important to find out both local and overall Sherwood numbers, fundamentally
understand the mass transfer phenomena locally in the packed bed and by using this
knowledge to accurately design packed beds. Over the years packed beds have been used in
many applications like solid catalysed heterogeneous reactions, gas absorption, stripping,
distillation, chromatographic separation and ion exchange processes. Design of packed beds
before CFD was done by using various correlation and conventional approaches which
ignored wall effects, end effects, local effects, radial and axial variations but with the use of
CFD more accurate predictions of Sherwood number is possible because CFD has the ability
to carry out virtual experiments close to real physical experimentations. Finite volume
method is the most prominent for simulation of packed beds and in majority of the literature
Direct numerical simulation is used to avoid any sub-grid modelling. But DNS is
computationally very expensive but with the advent of high computational facilities it has
become to simulate a section of packed beds by activating few particles. Random packing
generation has been automated recently and can be done by DEM-CFD through use of
commercial software like ANSYS-FLUENT/CFX etc. For low D/d ratio conventional
approaches fail to describe the process and very less number of experiments done in this
regime. CFD can be effectively used to create Sherwood numbers where there are
experimental gaps and recently it is being increasingly used to design packed beds with
higher accuracy. Not only mass transfer but coupled heat and mass transfer can also be found
using CFD for reactions which are exothermic in nature. CFD can also be used to design new
catalyst and test their mass transfer and heat transfer characteristics to compare different
types of catalysts. CFD methods discussed here can be extended further for incorporation of
intraparticle transport of species and heat transfer and modifications for liquid-solid systems.
There are many challenges to be addressed using CFD for mass transfer in packed beds.
3. REFERENCES
1. Atmakidis, T., Kenig, E.Y., 2012. Numerical analysis of mass transfer in packed-bed
reactors with irregular particle arrangements. Chem. Eng. Sci. 81, 77–83.
https://doi.org/10.1016/j.ces.2012.06.048
2. Atmakidis, T., Kenig, E.Y., 2009. CFD-based analysis of the wall effect on the
pressure drop in packed beds with moderate tube/particle diameter ratios in the
laminar flow regime. Chem. Eng. J. 155, 404–410.
https://doi.org/10.1016/j.cej.2009.07.057
3. Baker, M.J., 2011. CFD simulation of flow through packed beds using the finite
volume technique. Univ. Exet. 208.
4. Bale, S., Sathe, M., Ayeni, O., Berrouk, A.S., Joshi, J., Nandakumar, K., 2017.
Spatially resolved mass transfer coefficient for moderate Reynolds number flows in
packed beds: Wall effects. Int. J. Heat Mass Transf. 110, 406–415.
https://doi.org/10.1016/j.ijheatmasstransfer.2017.03.052
5. Bale, S., Tiwari, S., Sathe, M., Berrouk, A.S., Nandakumar, K., Joshi, J., 2018. Direct
numerical simulation study of end effects and D/d ratio on mass transfer in packed
beds. Int. J. Heat Mass Transf. 127, 234–244.
https://doi.org/10.1016/j.ijheatmasstransfer.2018.07.100
6. Bale, S., Tiwari, S.S., Nandakumar, K., Joshi, J.B., 2019. Effect of Schmidt number
and D/d ratio on mass transfer through gas-solid and liquid-solid packed beds: Direct
numerical simulations. Powder Technol. 354, 529–539.
https://doi.org/10.1016/j.powtec.2019.05.067
7. Boccardo, G., Augier, F., Haroun, Y., Ferré, D., Marchisio, D.L., 2015. Validation of
a novel open-source work-flow for the simulation of packed-bed reactors. Chem. Eng.
J. 279, 809-820–820. https://doi.org/10.1016/j.cej.2015.05.032
8. Bu, S.S., Yang, J., Zhou, M., Li, S.Y., Wang, Q.W., Guo, Z.X., 2014. On contact
point modifications for forced convective heat transfer analysis in a structured packed
bed of spheres. Nucl. Eng. Des. 270, 21–33.
https://doi.org/10.1016/j.nucengdes.2014.01.001
9. Caulkin, R., Ahmad, A., Fairweather, M., Jia, X., Williams, R.A., 2009. Digital
predictions of complex cylinder packed columns. Comput. Chem. Eng. 33, 10–21.
https://doi.org/10.1016/j.compchemeng.2008.06.001
10. Caulkin, R., Jia, X., Fairweather, M., Williams, R.A., 2008. Lattice approaches to
flow and heat transfer characteristics in pebbles through CFD methodology. Nucl.
Eng. Des. 258, 66–75. https://doi.org/10.1016/j.nucengdes.2013.02.009
21. Jurtz, N., Kraume, M., Wehinger, G.D., 2019. Advances in fixed-bed reactor
modeling using particle-resolved computational fluid dynamics (CFD). Rev. Chem.
Eng. 35, 139–190. https://doi.org/10.1515/revce-2017-0059
22. Klöker, M., Kenig, E.Y., Piechota, R., Burghoff, S., Egorov, Y., 2005. CFD-based
study on hydrodynamics and mass transfer in fixed catalyst beds. Chem. Eng.
Technol. 28, 31–36. https://doi.org/10.1002/ceat.200407048
23. Lau, Y.M., Deen, N.G., Kuipers, J.A.M., 2013. Development of an image
measurement technique for size distribution in dense bubbly flows. Chem. Eng. Sci.
94, 20–29. https://doi.org/10.1016/j.ces.2013.02.043
24. Lee, J.J., Yoon, S.J., Park, G.C., Jae, W.L., 2007. Turbulence-induced heat transfer in
PBMR core using LES and RANS. J. Nucl. Sci. Technol. 44, 985–996.
https://doi.org/10.1080/18811248.2007.9711338
25. Motlagh, A.H.A., Hashemabadi, S.H., 2008. 3D CFD simulation and experimental
validation of particle-to-fluid heat transfer in a randomly packed bed of cylindrical
particles. Int. Commun. Heat Mass Transf. 35, 1183–1189.
https://doi.org/10.1016/j.icheatmasstransfer.2008.07.014
26. Niegodajew, P., Marek, M., 2016. Analysis of orientation distribution in numerically
generated random packings of Raschig rings in a cylindrical container. Powder
Technol. 297, 193–201. https://doi.org/10.1016/j.powtec.2016.04.024
27. Ookawara, S., Kuroki, M., Street, D., Ogawa, K., 2007. High-fidelity DEM-CFD
modeling of packed bed reactors for process intensification. Chem. Eng. 16–20.
28. Shams, A., Roelofs, F., Komen, E.M.J., Baglietto, E., 2014. Large eddy simulation of
a randomly stacked nuclear pebble bed. Comput. Fluids 96, 302–321.
https://doi.org/10.1016/j.compfluid.2014.03.025
29. Shams, A., Roelofs, F., Komen, E.M.J., Baglietto, E., 2013a. Large eddy simulation
of a nuclear pebble bed configuration. Nucl. Eng. Des. 261, 10–19.
https://doi.org/10.1016/j.nucengdes.2013.03.040
30. Shams, A., Roelofs, F., Komen, E.M.J., Baglietto, E., 2013b. Quasi-direct numerical
simulation of a pebble bed configuration, Part-II: Temperature field analysis. Nucl.
Eng. Des. 263, 490–499. https://doi.org/10.1016/j.nucengdes.2013.02.015
31. Wang, Z., Afacan, A., Nandakumar, K., Chuang, K.T., 2001. Porosity distribution in
random packed columns by gamma ray tomography. Chem. Eng. Process. 40, 209–
219. https://doi.org/10.1016/S0255-2701(00)00108-2
32. Wehinger, G.D., Eppinger, T., Kraume, M., 2015. Evaluating Catalytic Fixed-Bed
Reactors for Dry Reforming of Methane with Detailed CFD. Chemie Ing. Tech. 87,
734–745. https://doi.org/10.1002/cite.201400153
33. Wehinger, G.D., Heitmann, H., Kraume, M., 2016. An artificial structure modeler for
3D CFD simulations of catalytic foams. Chem. Eng. J. 284, 543–556.
https://doi.org/10.1016/j.cej.2015.09.014
34. Yang, J., Wang, Q., Zeng, M., Nakayama, A., 2010. Computational study of forced
convective heat transfer in structured packed beds with spherical or ellipsoidal
particles. Chem. Eng. Sci. 65, 726–738. https://doi.org/10.1016/j.ces.2009.09.026
35. Yang, X., Scheibe, T.D., Richmond, M.C., Perkins, W.A., Vogt, S.J., Codd, S.L.,
Seymour, J.D., McKinley, M.I., 2013. Direct numerical simulation of pore-scale flow
in a bead pack: Comparison with magnetic resonance imaging observations. Adv.
Water Resour. 54, 228–241. https://doi.org/10.1016/j.advwatres.2013.01.009