Chemical Engineering Science 62 (2007) 5870 5879 www.elsevier.

com/locate/ces

Prediction of pressure drop and liquid holdup in trickle bed reactor using relative permeability concept in CFD
Arnab Atta, Shantanu Roy, K.D.P. Nigam
Department of Chemical Engineering, Indian Institute of Technology Delhi, Hauz Khas, New Delhi 110 016, India Received 6 January 2007; received in revised form 2 June 2007; accepted 11 June 2007 Available online 15 June 2007

Abstract A Computational Fluid Dynamics (CFD) model based on porous media concept is presented to model the hydrodynamics of two-phase ow in trickle-bed reactors (TBRs). The aim of this study is to develop a comprehensive CFD based model for predicting hydrodynamic parameters in trickle-bed reactors under cold-ow conditions. The two-phase Eulerian model describing the ow domain as a porous region has been used to simulate the macroscale multiphase ow in trickle beds operating under trickle ow regime using FLUENT 6.2 software. The closure terms for phase interactions have been addressed by adopting the relative permeability concept [Sez, A.E., Carbonell, R.G., 1985. Hydrodynamic parameters for gasliquid cocurrent ow in packed beds. A.I.Ch.E. Journal 31, 5262]. The model has been evaluated by comparing predictions with the data (collected under a varied set of laboratory conditions) available in the open literature. It is shown that while being relatively simple in structure, this CFD model is exible and predictive for a large body of experimental data presented in the open literature. 2007 Elsevier Ltd. All rights reserved.
Keywords: Trickle-bed; Porous media; Relative permeability; CFD

1. Introduction Trickle bed reactors (TBRs) are multiphase reactors in which gas and liquid phases ow co-currently downward over a solid catalyst packing. TBRs nd widespread use in petroleum rening, chemical and process industries, pollution abatement and biochemical industries. Design and scale up of TBRs continues to be a major challenge for chemical engineers. The design and scale-up of trickle bed reactors depend on key hydrodynamic variables such as liquid volume fraction (liquid saturation), particle scale wetting and overall gasliquid distribution. These variables are difcult to determine experimentally and interactions between these are as yet poorly understood. Even though numerous experimental studies have been reported in measurement of these variables, predicting them from rst principle hydrodynamic simulations is difcult as yet and no coherent and conclusive methodology for doing so has yet been espoused.

Corresponding author. Tel./fax: +91 11 26591020.

E-mail address: nigamkdp@gmail.com (K.D.P. Nigam). 0009-2509/$ - see front matter doi:10.1016/j.ces.2007.06.008 2007 Elsevier Ltd. All rights reserved.

Numerous authors have studied and reported experimental data on pressure drop and liquid saturation in trickle-bed reactors (e.g. Specchia and Baldi, 1977; Rao et al., 1983; Szady and Sundaresan, 1991; Al-Dahhan et al., 1997). An exhaustive state of art reviews on hydrodynamic parameters of TBR can be found in Saroha and Nigam (1996). The previous attempts for describing trickle bed hydrodynamics can be categorized mainly into two different classes of work. The classical approach is empirical wherein correlations are developed to t the experimental data (e.g. Larkins et al., 1961; Ellman et al., 1988, 1990; Larachi et al., 1991). Another approach is to describe hydrodynamics in phenomenological manner, i.e., assuming a simple picture of the microscale ow pattern, and then integrate that depiction to address the entire bed (e.g. Holub et al., 1992; Iliuta et al., 2000). Increasing computational power and development of Computational Fluid Dynamics (CFD) has allowed promising applications of numerical simulations to the modeling of multiphase ow in TBR (e.g. Attou and Ferschneider, 1999; Propp et al., 2000; Souadnia and Lati, 2001; Jiang et al., 2002a,b; Gunjal et al., 2003, 2005). Table 1 summarizes these reported attempts and highlights the key features in each of these efforts.

A. Atta et al. / Chemical Engineering Science 62 (2007) 5870 5879

Table 1 Summary of previous CFD models
S. No. Authors 1. Attou and Ferschneider (1999) Key features of model Model developed on the basis of area-averaged mass and momentum balance equations of each uid Capillary pressure gradient was deduced from a momentum balance analysis at the gasliquid interface The liquidsolid and gasliquid interaction forces are formulated on the basis of the KozenyCarman equation 2. Propp et al. (2000) Flow is assumed to be governed by equations of ow in porous media Use of high-resolution nite-difference methods to discretize governing equations. Examined the effects of Ergun equation, capillary pressure and variable porosity 3. Souadnia and Lati (2001) One-dimensional computational model is used with the nite volume technique combined with Godunovs method. Drag forces are accounted for through the equations developed by Sez and Carbonell (1985) Porosity is assumed uniform and constant 4. Jiang et al. (2002a,b) Two uid approach using CFDLIB (Los Alamos National Laboratory) Drag-exchange coefcients are calculated by the model of Attou et al. (1999) Capillary pressure is incorporated via J-function Bed structure implementation is resolved through statistical implementation of sectional porosities 5. Gunjal et al. (2003) Two uid approach using the closure of Attou et al. (1999) have been used in 2D as well as 3D geometry Liquid ow distribution and RTD were studied incorporating the effect of capillary pressure and porosity variation 6. Gunjal et al. (2005) Extension of their own previous model (Gunjal et al., 2003) for simulating the spray ow regime and hysteresis on pressure drop Attempted to simulate the periodic ow The rst effort to simulate the reactor in 3D Remarks

5871

As the model is one-dimensional, it cannot accommodate the variation of radial bed porosity distribution

The code was tested with several test problems, no explicit validation with experimental results for different hydrodynamic parameters were presented

Development in 2D model required for better prediction of parameters

Able to capture some of the key features of the hydrodynamics Bed structure implementation is involved

Though the qualitative prediction of hysteresis was carried out but the development in quantitative comparison is required. Though the simulation of periodic ow can be used to understand some key features, still development is essential in this eld for better prediction

While the CFD models (Table 1) have tried to resolve the existing complexities of the TBR to a great extent, considerable debate still persists on the exact nature of equations to be solved for TBR simulations. Depending on the problem formulation and requirement of specic results, some of the open issues include the number of phases to be solved, the nature of the model (steady state or transient), the formulation of the equations (EulerEuler or EulerLagrangian), the level of details in describing the packed bed (i.e., the particle scale models) and, the appropriate boundary and initial conditions to be used (Ranade, 2002). Most of the literature available (Jiang et al., 2002a,b; Gunjal et al., 2003, 2005) dealing with packed bed ow simulations use a three-phase Eulerian model in which the solid velocity is identically set to zero. Such a calculation is nevertheless computationally demanding and yet experiments

have advanced to a level (except perhaps the MRI imaging studies of Prof. Gladdens group at Cambridge (Sederman and Gladden, 2001, 2005)) wherein the ow features predicted at the particle and liquid-rivulet scale in the TBR can be anticipated with delity. Even in the MRI studies (Sederman and Gladden, 2001, 2005; Lim et al., 2004), which can be indeed intriguingly sophisticated, the studies only aid in high resolution ow visualization and a one-on-one comparison with the CFD computation is not possible. This is because while in the experiments only a single realization of the ow is observed, the averaged equations (which are in principle written for the ensemble moments of the ow variables) represent an average of all possible realizations. Thus in the nal analysis, one is left with comparing the average ow variables (such as overall pressure drop, holdup, etc.) and from a limited number

5872

A. Atta et al. / Chemical Engineering Science 62 (2007) 5870 5879

of experiments, the proles (radial, axial) of these variables. Therefore, whether getting into greater levels of detail with volume and ensemble averaged equations (consequently making the analysis more involved) is a justiable direction of work, is questionable. That being the present scenario, and the fact that even the two-uid CFD models have a lot of questionable assumptions (even though deceptively at a micro-scale), the wisdom of using a highly computationally intensive model for predicting global proles in a TBR may be questioned. In light of that, in this work, a less computationally intensive, yet rst-principle based CFD model has been presented using the porous media concept. The independent experimental data sets, reported by Specchia and Baldi (1977), Szady and Sundaresan (1991), Rao et al. (1983) and the ANN model developed by Iliuta et al. (1999) were selected in the present work to validate the predictions. In addition, we compare our results with the numerical simulation of Gunjal et al. (2005), which is based on a three phase Eulerian concept. The idea was to test our model against the state of the art in TBR modeling and offer it as a viable modeling approach. 2. Modeling The existing hydrodynamic models can be broadly classied into two different categories on the basis of empirical approach and theoretical or semi-empirical approach (presented in Table 2). The model presented here to describe the multiphase ow is based on the Eulerian framework, which consists of continuity and momentum equations of each uid phase with appropriate closures. To relate the drag forces with the ow velocities and volume fractions of each phase, and to the physical properties of the gas, liquid and solid phases, various closure terms have been reported. These have been obtained either from rst principles (e.g. uiduid interfacial force model (Attou et al., 1999)) or by semi-empirical approaches (e.g. relative permeability model (Sez and Carbonell, 1985); slit model (Holub et al., 1992, 1993; Al-Dahhan and Dudukovic, 1994; Al-Dahhan et al., 1998; Iliuta and Larachi, 1999)). Recently, Larachi et al. (2000) have compared these models for liquid holdup and pressure drop to experimental data in the trickling regime. Carbonell (2000) has theoretically reviewed (with the help of mean absolute relative errors calculated by Larachi et al., 2000) all the approaches to obtain drag expressions for twophase ow model and concluded that the relative permeability model (Sez and Carbonell, 1985) and the uiduid interaction model (Attou and Ferschneider, 1999) are based on solid hydrodynamic principles and are able to predict the hydrodynamic parameters with acceptable accuracy. Once again it is noteworthy that the only closure model used frequently in CFD calculations for simulating the ow through packed bed is uiduid interaction model (Attou et al., 1999) in most of the previous simulations (e.g. Jiang et al., 2002a,b; Gunjal et al., 2003, 2005). In line with Carbonell (2000), the approach to be used in this work is based on the assumption that ow domain (xed bed with catalyst particles) can be described as porous media. The two-phase ow phenomena through porous media are very

complex phenomena and it is difcult to develop a purely predictive approach for the ow hydrodynamics. Therefore, the need of some tting parameters is unavoidable. The closures used in this model are the relative permeability model developed by Sez and Carbonell (1985). The uiduid interaction model (developed theoretically, Attou and Ferschneider, 1999) has a notable feature of not having any adjustable parameter in the closure but is found to predict results accurately when incorporated into the CFD framework. This aspect leads to the drawback of complicated incorporation of different particle size and shape effect in that model. Hence, the relative permeability drag force model is rather logical and a promising alternative. This model has a remarkable feature of being exible for different particle sizes which can be incorporated without much complexity. 2.1. Governing equations of the ow The volume-averaged equations for each owing phase can be written as Continuity equation: j( jt s ) + ( u ) = 0, = g, l. (1)

Momentum balance equation: ju + u u jt = (p + ( g) (2)

+R )+F ,

where F is the total drag force per unit of bed volume exerted by the phase ; and R are, respectively, the volume averaged viscous stress tensor and the turbulence stress tensor of phase . Note that in general, p is phase-specic and the same pressure is not shared by both phases. In order to solve these equations the assumptions taken in this model are 1. There is no inter-phase mass transfer. 2. Both the owing uids are incompressible. 3. The porous medium is taken to be isotropic, i.e., permeabilities are independent of direction. 4. The porosity is uniform and constant. 5. Trickle ow regime is the operating ow regime, i.e., gasliquid interaction is low so capillary pressure force can be neglected. This means that we assume same pressure for both phases at any point in time and space. 6. The contribution of the turbulent stress terms to overall momentum balance equation (2) is not signicant. This assumption has also been used by other authors (e.g. Jiang et al., 2002a). Inter-phase coupling terms accounted by Eq. (2) are based on relative permeability concept developed by Sez and Carbonell (1985), as discussed below.

A. Atta et al. / Chemical Engineering Science 62 (2007) 5870 5879

Table 2 Various approach to model hydrodynamics of TBR
Approach Empirical References Larkins et al. (1961), Sai and Varma (1987), Ellman et al. (1988, 1990), Larachi et al. (1991), Xiao et al. (2000, 2001) and Pina et al. (2001) Relative permeability model Sez and Carbonell (1985) Remarks

5873

Based on dimensional analysis to produce explicit correlations for pressure drop and liquid holdup using ow variables and packing characteristics or using LockhartMartinelli parameter which was proposed for open horizontal tubes Ergun equation has been modied to account for the existence of a second owing phase by incorporation of relative permeability which has been correlated as a function of liquid saturation of each phase depending on the experimental results It is also a modied form of the Ergun equation. The ow through complex geometry of the actual void space in the catalyst bed at the pore level has been modeled by the much simpler ow inside a rectangular slit assuming the width of the slit is a function of bed porosity the angle of inclination of the slit to the vertical axis is related to a tortuosity factor for the packed bed. The surface area per unit volume of solid in this rectangular slit was made equal to the surface area per unit volume of solid in the reactor The liquidgas interfacial drag has been taken into account by rst developing an expression for the drag on a single bubble/slug and then multiplying it by the number of bubbles/slugs per unit volume of the porous layer To avoid the empiricism of the relative permeability and slit models, this model was developed in which the drag force on each phase has contributions from the particle-uid interactions as well as from uiduid interactions. The model was derived from a momentum balance on the uid ow around fully wetted particles

Theoretical or semiempirical

Slit model

Holub et al. (1992, 1993), Al-Dahhan and Dudukovic (1994), Al-Dahhan et al. (1998), Iliuta and Larachi (1999)

Model based on fundamental force balance

Tung and Dhir (1988), Narasimhan et al. (2002)

Fluiduid interfacial force model

Attou et al. (1999)

2.2. Relative permeability model The concept of relative permeability is very commonly used to the various problems of multiphase ow through porous media. Essentially, it is a concept that stems from the classical Darcys Law, a macroscopic equation based on average quantities for modeling pressure drop through a porous medium at xed supercial velocity for one phase ow. If a uid of viscosity is passing through a porous medium of absolute permeability k in a homogeneous gravitation eld with the ow rate U, then the pressure gradient p across the medium is given by Darcys equation U = (p g), k (3)

cause when two uids are simultaneously present in a porous medium, one uids ability to ow will be guided by the microscopic conguration of the other uid. In general, the expression for drag force for single-phase ow is modied using certain parameter (named as relative permeability, k , of that phase) to accommodate the presence of second phase (Sez and Carbonell, 1985). Sez and Carbonell (1985) have modied the Ergun equation for the single-phase pressure drop to calculate the two-phase ow pressure drop which can be represented in dimensionless form with the help of Reynolds and Galileo numbers: F = 1 k A Re Re2 +B Ga Ga g. (4)

where g denotes the acceleration due to the gravitational forces and is the density of that single phase uid. For uid ow in horizontal direction g can be neglected. While describing two-phase ow in porous media, it becomes very essential to modify the above stated equation be-

This model accounts for the reported non-linearity in the pressure gradient as a function of velocity (MacDonald et al., 1979), using the concept of relative permeability (Sez and Carbonell, 1985). The constants A and B in Eq. (4) are the Ergun equation coefcients for single-phase ow in the packed bed (Ergun,

5874

A. Atta et al. / Chemical Engineering Science 62 (2007) 5870 5879

1952). The Reynolds and Galileo numbers are dened as Re = Ga = u de , (1 ) 2 gd 3 3 e

2 (1

)3

, (5)

where de =

6Vp . Ap

While deriving these equations, Sez and Carbonell (1985) assumed that ow is one-dimensional and the liquid holdup does not change along the bed length, which implies that pressure gradients in the liquid and gas phases are equal. Therefore, the capillary pressure terms are neglected. Again by subtracting the momentum equation of gas and liquid phases from each other, and assuming that the liquid density is much greater than gas density, the equation for calculating liquid holdup can be found as Re2 Rel 1 A +B l kl Ga l Ga l Re2 Reg 1 g A +B kg Ga g Ga g
g l

In order to consider the microscopic/local conguration of the second uid and to dene the ability to ow of one uid in presence of other uid, the term relative permeability (k ) was introduced. From the previous discussions as well as from Eq. (4), it can be observed that the relative permeability corrects the drag force expression for single-phase ow conditions to account for the ow of two phases. With this perception it can be stated that it is a different approach to account uiduid interfacial drag forces in order to achieve the same goal i.e., pressure drop and holdup. Again since the relative permeability parameter has been incorporated to accommodate the presence of a second phase, essentially it will be a function of phase saturation or holdup of that corresponding phase. To determine the dependence of the relative permeability on the saturation for each phase Sez and Carbonell (1985) analyzed different data sets for liquid holdup and pressure drop over a wide range of Reynolds and Galileo numbers in packed beds available in the literature till that time. They made the hypothesis that liquid relative permeabilities are only a function of reduced saturation ( l ), which is represented by the ratio of effective volume of ow of the liquid phase to the available volume of ow considering that the static liquid holdup ( 0 ) represents a portion of the void fraction occupied l by stagnant liquid. Thus
l

= 1. (10)

More detailed derivation of these equations can be found elsewhere (Sez and Carbonell, 1985). Regarding the dependency as well as sensitivity of relative permeabilities on different possible parameters, very recently, Nemec and Levec (2005) have shown through their extensive experimentation and analysis for a wide rage of operating conditions and with typical shapes and sizes of particles encountered in commercial trickle-bed reactors, that relative permeabilities are solely functions of the corresponding phase saturation. Before making this conclusion these authors have carefully explored the effects of uncertainties associated with the phase relative permeabilities and also have carried out the detailed study on the phenomenological insights of the suitable correlations, e.g. the effect of particle shape and size, effect of ow rate and reactor pressure. Interestingly, they have opposed the observation by Lakota et al. (2002) on the particle shape dependency of the gas phase relative permeability. They have argued that the effect of shape factor is accounted by Ergun constants however the relative permeability being the ratio between single and two-phase pressure drop, this shape effect has been already taken care in that respect. 3. Boundary conditions and numerical solution Considering a two-dimensional axisymmetric domain (Fig. 1), the above set of model equations were solved using commercial software FLUENT 6.2 (of ANSYS. Inc., USA) dening the solution domain as porous. The gas phase was treated as primary phase and liquid phase was considered as secondary phase. At the inlet, at velocity prole for gas and liquid phases was assumed and implemented. No slip boundary condition was set for all the impermeable reactor walls. At the bottom of the column, an outlet boundary condition was specied. The reference pressure equal to atmospheric pressure was xed at the outlet. As a patch for the initial conditions, the overall volume fraction of the liquid phase was estimated using the correlation given by Eq. (10). Unsteady state simulations were carried out with the time step of 0.005 s. Some preliminary numerical experiments were carried out to identify the required number of computational cells to obtain grid independent results. It was also ensured with the preliminary numerical experiments to have discretization scheme independent results. These simulations conrmed that the grid size taken was satisfactory, as further

0 l . 0 l

(6)

The gas phase relative permeability was correlated as a function of the gas phase saturation. The empirical correlations were reported to be (Sez and Carbonell, 1985): kl =
2.43 , l 4.80 kg = sg ,
0

(7)

where sg = 1 l . The static liquid holdup ( 0 ) can be calculated by the foll lowing correlation given by Sez and Carbonell (1985)
0 l

1 (20 + 0.9Eo )

where Eo =

2 2 l gd p l (1

)2

(8)

After simplifying these expressions for a given particle diameter and the ow rates of gas and liquid ows, the following equation of motion can be used to compute the pressure drop: p l = Fg
g

Re2 Reg 1 g A +B kg Ga g Ga g

g g.

(9)

A. Atta et al. / Chemical Engineering Science 62 (2007) 5870 5879

5875

Air-water inlet

1.6 m

500 cells for 1m Axis B.C.

1.2 m

b
A

1.8 1.6 1.4 B

25 cells radially

height of the bed (m)

1.2
Inlet

1 0.8 0.6 0.4

This work_outlet Anderson and Sapre (1991)_outlet

Outlet
Fig. 1. Computational domain.

Table 3 Simulation parameters

Discretization scheme Pressurevelocity coupling Relaxation parameters Pressure Momentum Volume fraction Body forces Convergence criterion First order UPWIND SIMPLE algorithm 0.3 0.7 0.2 1.0 105
-0.9

0.2 G' -0.6 C -0.3 E 0 F 0 D 0.3 G 0.6 0.9

Distance from the centre of the bed (m)

Fig. 2. (a) Computational domain used by Anderson and Sapre (1991). (b) Quantication of liquid spreading for constricted inlet at the top.

increase in number of grids did not appreciably affect the predicted results. The simulation parameters are summarized in Table 3. 4. Results and discussion Initially some numerical experiments were carried out to endorse the fact that in the real ow, the ow characteristic is 2D or 3D even if the initial liquid inlet is uniform (1D). This was to establish the versatility of our model and to make sure that we can indeed use the model in 2D and 3D scenarios without compromising its robustness, even though most of the experimental validation we present later in this work relate to 1D effects. One of the case studies involves liquid ow introduced over only a small part of the top of the bed which is identical to

the geometry considered by Anderson and Sapre (1991) (where the ratio between bed diameter to the particle diameter is 400) (Fig. 2a). Following the correlations proposed by Cohen and Metzner (1981) for incorporation of porosity distribution, we have observed that the porosity variation/uctuation ceases to exist after almost eight particle diameters from the bed wall. Therefore, it can be concluded that incorporation of porosity variation in this particular case will be ineffective as there is a minimum possibility of liquid spreading near the wall for constricted inlet at the top. Taking the similar ow condition of Anderson and Sapre (1991), we have completed our case study and it (Fig. 2b) shows that for constricted inlet at the top (line AB) there is some degree of spreading of liquid and the width has been quantied by this present CFD model (line CD). EF represents the result of Anderson and Sapre (1991) through which 80% of the ow is taking place where GG depicts the width of the bed. This prediction of liquid spreading would not have been possible without taking 2D model equations. However, it may be noted that these simulations were carried

5876

A. Atta et al. / Chemical Engineering Science 62 (2007) 5870 5879

Table 4 Details of operation conditions used for simulations

S. No. Source Bed diameter, D (m) Bed length, l (m) 1.49 1.05 1.835 1.49 1 l/D ratio Particle diameter, dp (m) 0.003 0.0027 0.00627 0.003 0.006 D/dp ratio Bed porosity Gas velocity (m s1 ) Liquid velocity (m s1 ) 0.0020.008 0.0028 0.004128 0.00040.006 0.00170.0092

1. 2. 3. 4. 5.

Szady and Sundaresan (1991) Specchia and Baldi (1977) Rao et al. (1983) ANN Model Gunjal et al. (2005)

0.165 0.08 0.0924 0.165 0.114

9.03 13.13 19.86 9.03 8.77

55 29.63 14.77 55 19

0.37 0.38 0.373 0.37 0.37

0.22 0.20.8 0.130.95 0.22 0.22

18 This work Szady and Sundaresan (1991) ANN Gunjal et al. (2005) simulation Liquid saturation (-)

0.6

This work Szady and Sundaresan (1991) ANN Gunjal et al. (2005) simulation

15 Pressure drop (kPa.m-1)

0.5

12

0.4

0.3

0.2

0.1
0 0 0.002 0.004 0.006 (m.s-1) 0.008 0.01

Liquid Velocity

0 0 0.002 0.004 0.006 Liquid Velocity (m.s-1) 0.008 0.01

Fig. 3. Comparative study of pressure drop with literature data (Vg = 0.22 m s1 ).

Fig. 4. Comparative study of liquid saturation with literature data (Vg = 0.22 m s1 ).

out without taking porosity distribution in the bed. With this understanding and perception, further numerical simulations were carried out for different geometries and ow conditions on Sun Fire V880 server accounting uniform and constant porosity of the bed for simplicity of the computation/model. The simulated results were rst validated against various published experimental data of pressure drop and liquid holdup. The details of the experimental data set are given in Table 4. Two-phase pressure drop per unit length and dynamic liquid holdup at different liquid and gas ow rates were obtained by a series of numerical simulations to compare with the above mentioned data sets. The values of Erguns constants (A and B) used in the closure model are 180 and 1.8 for all the numerical simulations. Ad hoc tting of these parameters was not required. Figs. 3 and 4 show the comparison of simulated results with the experimental data of Szady and Sundaresan (1991) and numerical prediction of Gunjal et al. (2005) for observed pressure drop per unit length and total liquid saturation. While adopting the experimental data set from Szady and Sundaresan (1991), only the upper branch of pressure drop curve was taken, as that was the case for prewetted bed where capillary pressure can be neglected. Both the model results show reasonable agreement

of variation of the pressure drop and liquid saturation with different liquid supercial velocity at a constant gas supercial velocity of 0.22 m s1 . Figs. 5 and 6 show the model prediction for pressure drop and liquid saturation for experimental data of Specchia and Baldi (1977) for different gas supercial velocities at a constant liquid velocity of 0.0028 m s1 . In all cases the results were also veried for accuracy with the numerical prediction based on the three-phase Eulerian simulation carried out by Gunjal et al. (2005). It is noteworthy that while carrying out the simulations Gunjal et al. (2005) varied their choice of Erguns constants for different data sets (e.g., A = 180 and B = 1.8 for the data set of Szady and Sundaresan (1991); A = 500 and B = 3 for the data set of Specchia and Baldi (1977)). Again, they seem to have used the denition of Etvos no. (Eo ) while calculating static liquid holdup as Eo =
l 2 2 l gd p (1 2 )2

(11)

in which 2 has been used instead of in the denominator. The simulations in this work are then validated with the experimental results of Rao et al. (1983) for pressure drop with varying gas ow rates (Fig. 7). As is clear, the same

A. Atta et al. / Chemical Engineering Science 62 (2007) 5870 5879

5877

45 40 Pressure drop (kPa.m-1) 35 30 25 20 15 10 5 0 0 0.2 0.4 0.6 0.8 1 Gas velocity (m.s-1)
Fig. 5. Comparative study of pressure drop with literature data (Vl = 2.8 kg m1 s1 ).
Specchia and Baldi (1977) This work Gunjal et al. (2005) simulation

5 4.5 Pressure drop (kPa.m-1) 4 3.5 3 2.5 2 1.5 1 0.5 0 0 0.002 0.004 0.006 0.008 (m.s-1) 0.01 0.012 Liquid velocity
Gunjal et al. (2005) experiment Gunjal et al. (2005) simulation This work

Fig. 8. Comparative study of pressure drop with 3-phase simulation (Vg = 0.22 m s1 ).

0.45 0.4 Liquid saturation (-) 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0 0.2 0.4 0.6 0.8 1 Gas velocity (m.s-1)
Fig. 6. Comparative study of liquid saturation with literature data (Vl = 2.8 kg m1 s1 ).

Specchia and Baldi (1977) This work Gunjal et al. (2005) simulation

18 Pressure drop (kPa.m-1) 16 14 12 10 8 6 4 2 0 0 0.2 0.4 0.6 0.8 1 1.2 Gas velocity (m.s-1)
Fig. 7. Comparative study of pressure drop with literature data (Vl = 4.128 kg m1 s1 ).

This work Rao et al. (1983)

model seems to predict experimental observations from various sources collected over a wide variety of conditions. Finally, we compare our model prediction against the ANN models of Iliuta et al. (1999). Iliuta et al. (1999) developed state-of-art explicit correlations for hydrodynamic parameters of TBR based on a large database of experimental observations upon the combination of dimensionless analysis and articial neural networks. This large data base was then used to train a hybrid articial neural network (ANN) model so that the best predictions of pressure drop, holdup and wetting efciency to a given TBR could be predicted. This ANN model is also able to predict the trickle-pulse transition boundary. The comparison of our simulated results with this model (Figs. 3 and 4) shows that the predictions are satisfactory. In the present study, the assumption of neglecting the capillary pressure (i.e., the liquid holdup is practically independent of the axial position in the bed) leads to the fact that this model is strictly valid for low interaction regime i.e., trickling ow regime. This is evident from the validation of predicted results with the results of ANN model (Figs. 3 and 4). It shows that at low liquid velocity the prediction is more accurate and as the liquid velocity increases the regime moves towards the transition region where the model equations cease to hold. It is noteworthy that through this comparison, we are effectively comparing our model against the extensive experimental database of Iliuta et al. (1999). Thus, even without changing the Erguns constants, the present model can be used to predict pressure drop and liquid saturation in low interaction regime with enough condence. Figs. 8 and 9 support this statement, which show the comparison of predicted pressure drop and liquid holdup for different liquid velocity with the CFD results of Gunjal et al. (2005) and their in-house experimental data for spherical particle of 6 mm diameter with A=500 and B =2.4. In these cases (Figs. 8 and 9) the gas supercial velocity was kept constant at 0.22 m s1 and diameter of the bed was 0.114 m. Our model functions as arguably well as their model in predicting their experimental data, in fact, the pressure drop comparisons are better.

5878

A. Atta et al. / Chemical Engineering Science 62 (2007) 5870 5879

0.29

Notation
Gunjal et al. (2005) experiment Gunjal et al. (2005) simulation

0.24 Liquid holdup

A Ap B de dp Eo F

This work

0.19

0.14

0.09

0.04 0 0.002 0.008 0.004 0.006 Liquid velocity (m.s-1) 0.01 0.012

Fig. 9. Comparative study of liquid holdup with 3-phase simulation (Vg = 0.22 m s1 ).

5. Conclusions In this study, we have developed a two-phase Eulerian CFD model based on porous media concept to simulate gasliquid ow through packed beds. The porous media model is advantageous to handle gasliquid interaction terms due to its ability to lump the adjustable parameters as compared to the conventional k-uid CFD treatment of the problem. The closures used in this model are the relative permeability model developed by Sez and Carbonell (1985). The predicted results are veried for different sets of independent experimental data (Szady and Sundaresan, 1991; Specchia and Baldi, 1977; Rao et al., 1983) and results obtained by ANN model (Iliuta et al., 1999) which incorporates a myriad variety of experimental information. Predicted values showed good agreement with the experimental data. The predicted results are also compared with the numerical results of Gunjal et al. (2005), which are based on the three-phase Eulerian simulation. As discussed earlier, we can propose this model for future studies on prediction of hydrodynamic parameters under highpressure operation provided the suitable correlations are available (e.g. namely the new correlations developed for relative permeabilities by Nemec and Levec, 2005), which can be incorporated in this present CFD model as a modication of the closure. One must however approach the problem of ow modeling in TBRs with caution. Many decades have been spent in development of TBR technology, prior to the development of CFD, and the practice in the industry for designing TBRs is well established (even if based on heuristics in many cases and not totally scientically based). Any further developments that we may want to propose with CFD must be consistent with that prior knowledge. As such, the present contribution is the rst of a series of papers wherein we demonstrate the use of porous media approach as a viable CFD method that is both consistent with prior know-how as well as reveals new information from a design and scale-up perspective.

g Ga k l p Re s u U Vp

constant in the viscous term of the Ergun type equation particle surface area, m2 constant in the inertial term of the Ergun type equation equivalent particle diameter, 6Vp /Ap particle diameter, m modied Etvos number, l gd 2 2 / l (1 )2 p drag force on the phase per unit volume, kg m2 s2 gravitational acceleration, m s2 Galileo number of the phase, 2 gd 3 3 / 2 (1 )3 e relative permeability of phase length of the reactor pressure, Pa Reynolds number of the phase, u de / (1 ) saturation of the phase supercial velocity of the phase, m s1 ow velocity, m s1 particle volume, m3

Greek letters
l 0 l

reduced saturation of liquid phase, static liquid holdup bed voidage holdup of phase viscosity, Pa s density of the phase, kg m3 surface tension, N m1

0/ l

0 l

Subscripts g l gas/liquid phase gas phase liquid phase

Acknowledgment A. Atta is indebted to the All India Council for Technical Education (AICTE), India, for providing National Doctoral Fellowship. References
Al-Dahhan, M.H., Dudukovic, M.P., 1994. Pressure drop and liquid holdup in high-pressure trickle bed reactors. Chemical Engineering Science 49, 56815698. Al-Dahhan, M.H., Larachi, F., Dudukovic, M.P., Laurent, A., 1997. Highpressure trickle-bed reactors: a review. Industrial & Engineering Chemistry Research 36, 32923314. Al-Dahhan, M.H., Khadilkar, M.R., Wu, Y., Dudukovic, M.P., 1998. Prediction of pressure drop and liquid holdup in high-pressure trickle-bed reactors. Industrial & Engineering Chemistry Research 37, 793798. Anderson, D.H., Sapre, A.V., 1991. Trickle-bed reactor ow simulation. A.I.Ch.E. Journal 37, 377382. Attou, A., Ferschneider, G., 1999. A two-uid model for ow regime transition in gasliquid trickle-bed reactors. Chemical Engineering Science 54, 50315037.

A. Atta et al. / Chemical Engineering Science 62 (2007) 5870 5879 Attou, A., Boyer, C., Ferschneider, G., 1999. Modeling of the hydrodynamics of the cocurrent gasliquid trickle ow through a trickle-bed reactor. Chemical Engineering Science 54, 785802. Carbonell, R.G., 2000. Multiphase ow models in packed beds. Oil & Gas Science and Technology 55, 417425. Cohen, Y., Metzner, A.B., 1981. Wall effects in laminar ow of uids through packed beds. A.I.Ch.E. Journal 27, 705715. Ellman, M.J., Midoux, N., Laurent, A., Charpentier, J.C., 1988. A new improved pressure drop correlation for trickle-bed reactors. Chemical Engineering Science 43, 22012206. Ellman, M.J., Midoux, N., Wild, G., Laurent, A., Charpentier, J.C., 1990. A new improved liquid holdup correlation for trickle bed reactors. Chemical Engineering Science 45, 16771684. Ergun, S., 1952. Fluid ow through packed columns. Chemical Engineering Progress 48, 8994. Gunjal, P.R., Ranade, V.V., Chaudhari, R.V., 2003. Liquid distribution and RTD in trickle bed reactors: experiments and CFD simulations. Canadian Journal of Chemical Engineering 81, 821830. Gunjal, P.R., Kashid, M.N., Ranade, V.V., Chaudhari, R.V., 2005. Hydrodynamics of trickle-bed reactors: experiments and CFD modeling. Industrial & Engineering Chemistry Research 44, 62786294. Holub, R.A., Dudukovic, M.P., Ramachandran, P.A., 1992. A phenomenological model for pressure drop, liquid holdup, and ow regime transition in gasliquid trickle ow. Chemical Engineering Science 47, 23432348. Holub, R.A., Dudukovic, M.P., Ramachandran, P.A., 1993. Pressure drop, liquid holdup, and ow regime transition in trickle ow. A.I.Ch.E. Journal 39, 302321. Iliuta, I., Larachi, F., 1999. The generalized slit model: pressure gradient, liquid holdup and wetting efciency in gasliquid trickle ow. Chemical Engineering Science 54, 50395045. Iliuta, I., Larachi, F., Grandjean, B.P.A., Wild, G., 1999. Gasliquid interfacial mass transfer in trickle-bed reactors: state of-art correlations. Chemical Engineering Science 54, 56335645. Iliuta, I., Larachi, F., Al-Dahhan, M.H., 2000. Double-slit model for partially wetted trickle ow hydrodynamics. A.I.Ch.E. Journal 46, 597609. Jiang, Y., Khadilkar, M.R., Al-Dahhan, M.H., Dudukovic, M.P., 2002a. CFD of multiphase ow in packed-bed reactors: I. k-uid modeling issues. A.I.Ch.E. Journal 48, 701715. Jiang, Y., Khadilkar, M.R., Al-Dahhan, M.H., Dudukovic, M.P., 2002b. CFD of multiphase ow in packed-bed reactors: II. Results and applications. A.I.Ch.E. Journal 48, 716730. Lakota, A., Levec, J., Carbonell, R.G., 2002. Hydrodynamics of trickling ow in packed beds: relative permeability concept. A.I.Ch.E. Journal 48, 731738. Larachi, F., Laurent, A., Midoux, N., Wild, G., 1991. Experimental study of a trickle bed reactor operating at high pressure: two-phase pressure drop and liquid saturation. Chemical Engineering Science 46, 12331246. Larachi, F., Iliuta, I., Al-Dahhan, M.A., Dudukovic, M.P., 2000. Discriminating trickle-ow hydrodynamic models: some recommendations. Industrial & Engineering Chemistry Research 39, 554556.

5879

Larkins, R.P., White, R.R., Jeffery, D.W., 1961. Two-phase concurrent ow in packed beds. A.I.Ch.E. Journal 7, 231239. Lim, M.H.M., Sederman, A.J., Gladden, L.F., Stitt, E.H., 2004. New insights to trickle and pulse ow hydrodynamics in trickle-bed reactors using MRI. Chemical Engineering Science 59, 54035410. MacDonald, I.F., El-Sayed, M.S., Mow, K., Dullien, F.A.L., 1979. Flow through porous mediaThe Ergun equation revisited. Industrial & Engineering Chemistry Fundamentals 18, 199208. Narasimhan, C.S.L., Verma, R.P., Kundu, A., Nigam, K.D.P., 2002. Modeling hydrodynamics of trickle-bed reactors at high pressure. A.I.Ch.E. Journal 48, 24592474. Nemec, D., Levec, J., 2005. Flow through packed bed reactors: 2. Two-phase concurrent downow. Chemical Engineering Science 60, 69586970. Pina, D., Tronconi, E., Tagliabue, L., 2001. High interaction regime LockhartMartinelli model for pressure drop in trickle-bed reactors. A.I.Ch.E. Journal 47, 1930. Propp, R.M., Colella, P., Crutcheld, W.Y., Day, M.S., 2000. A numerical model for trickle bed reactors. Journal of Computational Physics 165, 311333. Ranade, V.V., 2002. Computational Flow Modeling for Chemical Reactor Engineering. Academic Press, London. Rao, V.G., Ananth, M.S., Varma, Y.B.G., 1983. Hydrodynamics of two-phase cocurrent downow in packed beds. A.I.Ch.E. Journal 29, 467483. Sez, A.E., Carbonell, R.G., 1985. Hydrodynamic parameters for gasliquid cocurrent ow in packed beds. A.I.Ch.E. Journal 31, 5262. Sai, P.S.T., Varma, Y.B.G., 1987. Pressure drop in gasliquid downward ow through packed beds. A.I.Ch.E. Journal 33, 20272036. Saroha, A.K., Nigam, K.D.P., 1996. Trickle bed reactors. Reviews in Chemical Engineering 12, 207347. Sederman, A.J., Gladden, L.F., 2001. Magnetic resonance imaging as a quantitative probe of gasliquid distribution and wetting efciency in trickle-bed reactors. Chemical Engineering Science 56, 26152628. Sederman, A.J., Gladden, L.F., 2005. Transition to pulsing ow in trickle-bed reactors studied using MRI. A.I.Ch.E. Journal 51, 615621. Souadnia, A., Lati, M.A., 2001. Analysis of two-phase ow distribution in trickle-bed reactors. Chemical Engineering Science 56, 59775985. Specchia, V., Baldi, G., 1977. Pressure drop and liquid holdup for two phase concurrent ow in packed beds. Chemical Engineering Science 32, 515523. Szady, M.J., Sundaresan, S., 1991. Effect of boundaries on trickle-bed hydrodynamics. A.I.Ch.E. Journal 37, 12371241. Tung, V.X., Dhir, V.K., 1988. A hydrodynamic model for two phase ow through porous media. International Journal of Multiphase Flow 14, 4765. Xiao, Q., Anter, A.M., Cheng, Z.M., Yuan, W.K., 2000. Correlations for dynamic liquid holdup under pulsing ow in a trickle-bed reactor. Chemical Engineering Journal 78, 125129. Xiao, Q., Cheng, Z.M., Jiang, X., Anter, A.M., Yuan, W.K., 2001. Hydrodynamics behavior of a trickle bed reactor under forced pulsing ow. Chemical Engineering Science 56, 11891195.