Analytical Void Fraction Profile Near the Walls in Low Reynolds Number Bubbly Flows in Pipes: Experimental Comparison and Estimate of the Dispersion Coefficient
Profil analytique de taux de vide en proche paroi dans les écoulements en conduite à bulles à faible nombre de Reynolds : comparaison expérimentale et estimation du coefficient de dispersion
^{1}
DenService de ThermoHydraulique et de Mécanique des Fluides (STMF), CEA, Université ParisSaclay, 91191
GifsurYvette – France
^{2}
Electricité de France R&D Division, 6 Quai Watier, 78400
Chatou – France
email: olivier.marfaing@cea.fr
^{*} Corresponding author
Received:
27
September
2016
Accepted:
21
December
2016
In a recent paper, we derived an analytical expression for the void fraction profile in low Reynolds number bubbly pipe flows, based on a balance of hydrodynamic forces on bubbles. The objective of the present work is to perform a comparison of this analytical Bubble Force Balance Formula (BFBF) with an experiment from the literature. We begin by simulating this experiment with the NEPTUNE_CFD code. In particular we show that using an R_{ij}ε model to account for the liquid velocity fluctuations yields reasonable results. In order to compare our analytical profile with experimental measurements, we restrict ourselves to the nearwall region. In this region, the void fraction profile results from a balance between dispersion and wall forces, and the dispersion coefficient can be considered as uniform. The analytical BFBF profile is seen to be in good agreement with the measurements. We are also capable to estimate the dispersion coefficient in this nearwall region.
Résumé
Dans une publication récente, nous avons obtenu une expression analytique pour le profil de taux de vide dans un écoulement en conduite à bulles à bas nombre de Reynolds, basée sur un bilan des forces hydrodynamiques sur les bulles. L’objectif du travail est d’effectuer une comparaison de cette Formule de Bilan des Forces sur les Bulles (FBFB) avec une expérience de la littérature. Nous commençons par simuler cette expérience avec le code NEPTUNE_CFD. En particulier, nous montrons qu’en utilisant un modèle R_{ij}ε, nous obtenons des résultats acceptables pour les fluctuations de vitesse liquide. Ensuite, afin de comparer notre profil analytique avec les mesures expérimentales, nous nous restreignons à la zone de proche paroi. Dans cette région, le profil de taux de vide résulte de l’équilibre entre forces de dispersion et de paroi, et le coefficient de dispersion peut être considéré comme uniforme. Le profil analytique FBFB est en bon accord avec les mesures et nous sommes capables d’estimer le coefficient de dispersion dans la zone de proche paroi.
© O. Marfaing et al., published by IFP Energies nouvelles, 2017
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Notations
BFBF: Bubble Force Balance Formula
B : Constant in Equation (16), determined from the knowledge of the average void fraction over a crosssection
C_{W1}, C_{W2}: Wall coefficients
D_{wall}: Nearwall dispersion coefficient (m^{2}/s)
D_{wall}^{*}: Nondimensional nearwall dispersion coefficient
J_{l}: Liquid superficial velocity (m/s)
J_{g}: Gas superficial velocity (m/s)
M_{g}: Interfacial momentum exchange term (Pa/m)
U_{R}: Relative velocity (m/s)
V_{k}: Velocity of phase k (m/s)
y^{*}: Nondimensional distance to the wall
Greek Letters
α_{k}: Volume fraction of phase k
δ_{g}: Gas viscous layer thickness (m)
δ_{g}^{*}: Nondimensional gas viscous layer thickness
Subscripts
Introduction
The capability to understand and simulate twophase flows is of everincreasing importance in applications to nuclear or chemical engineering. At the same time, because of the variety of conditions and flow regimes encountered, an important Validation and Verification (V&V) effort is required to qualify Computational Fluid Dynamics (CFD) codes (Boyd, 2014; Podowski, 2014).
Bubbly flows involve numerous small inclusions of gas in a liquid carrier phase. Because it would be too complex to follow these structures individually, a largerscale description is preferred. In the twofluid formulation (Drew and Passman, 1999; Ishii and Hibiki, 2010; Morel, 2015), information on individual bubbles is eliminated through an averaging procedure. The interaction between the continuous and the dispersed phase is accounted for by closure relations. As an example, the interfacial momentum transfer is decomposed into several components: drag, addedmass, lift, dispersion and wall forces.
Many models exist for each of these forces, with different ranges of validity. These models are often derived in regions far away from the walls. However, nearwall effects can play a crucial role, especially in heat transfer phenomena, like the boiling crisis. Developing closure relations specific to the nearwall region is therefore very important.
In the literature, many models of dispersion force are dedicated to turbulent bubbly flows. They are derived in regions where velocity gradients are negligible, especially far away from the walls. In some of these models (Laviéville et al., 2015; Lopez de Bertodano, 1998), the dispersion coefficient is expressed as being proportional to the turbulent kinetic energy, and therefore vanishes at the walls.
In a recent paper (Marfaing et al., 2016), we considered the problem of a fully developed “laminar” (low Reynolds number) adiabatic bubbly flow in a vertical pipe. The dispersion force was represented by the following expression, close to Davidson’s model (Davidson, 1990):(1)where D_{eff} is the dispersion coefficient. Assuming D_{eff} was uniform on the entire domain, and performing a balance of hydrodynamic forces on bubbles, we were able to give an analytical expression for the void fraction profile as a function of the liquid velocity and pressure profiles. The obtained analytical void fraction profile Bubble Force Balance Formula (BFBF) vanishes at the wall, as a result of the balance between dispersion and wall forces. It presents a peak near the wall for upward flows, whereas its maximum is reached at the center of the pipe for downward flows. This BFBF was then used as a verification tool for the NEPTUNE_CFD code, a Computational MultiFluid Dynamics (CMFD) code dedicated to the simulation of multiphase flows, primarily targeting nuclear thermalhydraulics applications, such as the Departure from Nucleate Boiling (DNB) or the twophase Pressurized Thermal Shock (PTS). NEPTUNE_CFD has been developed within the joint NEPTUNE R&D project (AREVA, CEA, EDF, IRSN) since 2001.
Naturally, assuming a uniform D_{eff} is a simplifying assumption. However, it is legitimate to consider the dispersion coefficient as uniform to make a comparison with experimental measurements over a localized region of space, in particular near a wall. In this region, the void fraction profile is determined by two dominant effects: dispersion and wall forces.
In this paper, we consider low Reynolds number (“laminar”) bubbly flows in pipes and we investigate the void fraction profile near the walls.
In Section 1, we begin by using the NEPTUNE_CFD code to simulate a series of low Reynolds number (“laminar”) adiabatic bubbly flow experiments performed by Nakoryakov et al. (1996). The calculations are seen to be in correct agreement with the experimental data. In particular, we can see that using the standard R_{ij}ε model of the code yields results for the liquid velocity fluctuations which are of the right order of magnitude.
Then, in Section 2, we focus on the nearwall region. We begin by adapting the argument in (Marfaing et al., 2016) to derive an analytical expression for the void fraction profile near the wall, where the dispersion coefficient is assumed to be uniform and nonzero. Then, the analytical profile is compared to Nakoryakov et al.’s (1996) experimental results. We find a good agreement and we can estimate the dispersion coefficient near the wall D_{wall}.
1 Calculations of “Laminar” AirWater Bubbly Flow Experiments With The NEPTUNE_CFD Code
NEPTUNE_CFD is a Computational MultiFluid Dynamics (CMFD) code dedicated to the simulation of multiphase flows, which has been codeveloped since 2001 in the frame of the NEPTUNE R&D project, supported by the four major stakeholders of the French nuclear industry: AREVA, CEA, EDF and IRSN. The code is able to handle various flow regimes: stratified flows, particle or dropletladen flows, and bubbly flows, in order to address a variety of local phenomena occurring in a nuclear reactor in accident conditions, such as the twophase Pressurized Thermal Shock (PTS), or the Departure from Nucleate Boiling (DNB). In order to demonstrate the applicability of CMFD codes to the prediction of these phenomena, an extensive Validation and Verification (V&V) effort is required. In the case of DNB, this effort is focused on the assessment of the hydrodynamic models used for bubbly flows.
In this section, we use the NEPTUNE_CFD code to simulate a series of low Reynolds number (“laminar”) adiabatic bubbly flow experiments. Although most experimental studies are conducted in the turbulent regime, a few “laminar” experiments are available in the literature.
Nakoryakov et al. (1996) investigated the structure of low Reynolds number bubbly flows in a vertical pipe. Their test section was a vertical cylindrical pipe with internal diameter 14 mm and height 6.5 m, at the bottom of which a mixture of gas and liquid was injected.
The gas was air. The liquid used was an aqueous solution of sodium hydroxide, potassium ferri and ferrocyanide, to which glycerine was added to increase the viscosity. In the reported experiments, the Reynolds number was comprised between 875 and 1860. For such low Reynolds numbers, the corresponding singlephase flow would be “laminar”. The observed liquid velocity fluctuations (“pseudoturbulence”) are due to the presence of bubbles.
The flow was observed to be statistically steady and axisymmetric – statistically meaning that the average flow, not its fluctuations, is steady and axisymmetric.
A measuring station, located at the top of the pipe, recorded the radial profiles of void fraction, liquid velocity and velocity fluctuations. Measurements were made with an electrodiffusional method, based on the rate of mass transfer from a microelectrode to the liquid. The probe was able to get to a distance y from the wall of one tenth of a bubble diameter.
Nakoryakov et al. (1996) conducted nine experimental runs called Ub101Ub109. His experimental data are available online https://scholar.lib.vt.edu/ejournals/JFE/data/JFE/DB96377/. Experimental conditions are summarized in Table 1. The average bubble diameter is known at the recording station, but there is little information about the bubble size distribution. In a previous paper by the same team (Kashinsky et al., 1993), a similar test section is used and the authors report that for most of the flow regimes, the bubbles are rather uniform in size.
Experimental runs conducted by Nakoryakov et al. (1996).
This experiment is simulated with the NEPTUNE_CFD code.
At present, the NEPTUNE_CFD code is devoted to the simulation of turbulent bubbly flows. No specific model for liquid velocity fluctuations in low Reynolds number flows has been implemented yet. For this reason, we make use of the standard R_{ij}ε model of the code (Mimouni et al., 2009) to account for the liquid velocity fluctuations.
In Figures 13, the experimental and computed profiles are compared for run Ub102.
Figure 1
Computed (green) and experimental (black) void fraction profiles run Ub102. 
Figure 2
Computed (green) and experimental (black) liquid velocity profiles run Ub102. 
Figure 3
Computed (green) and experimental (black) axial liquid velocity fluctuations run Ub102. 
The void fraction and liquid velocity profiles are well predicted by our simulations.
In Figure 3, the axial liquid velocity fluctuations are defined as being the axial component of the liquid fluctuating velocity. We can see that using the standard R_{ij}ε model of the code yields results which are of the right order of magnitude.
For a higher gas superficial velocity (run Ub103), the same comparison is carried out in Figures 46. The location of the void fraction peak is well captured but the calculations underestimate its magnitude: the predicted profile is too spread. The computed liquid velocity profile is in good agreement with the experiment. Again, the liquid velocity fluctuations are of the right order of magnitude, which confirms the R_{ij}ε model yields reasonable results in these low Reynolds number conditions.
Figure 4
Computed (blue) and experimental (black) void fraction profiles run Ub103. 
Figure 5
Computed (blue) and experimental (black) liquid velocity profiles run Ub103. 
Figure 6
Computed (blue) and experimental (black) axial liquid velocity fluctuations run Ub103. 
Even for low Reynolds number flows, taking into account liquid velocity fluctuations is essential, because they drive the dispersion force. This force counterbalances the effects of lift and wall forces, and contributes to smoothing the void fraction field. For comparison, we carry out computations of run Ub102 without using a model for liquid velocity fluctuations. The obtained void fraction profile displays strong oscillations (Fig. 7).
Figure 7
Computed (blue) and experimental (black) void fraction profiles run Ub102. The computations are performed without taking into account liquid velocity fluctuations. 
2 Analytical and Experimental NearWall Void Fraction Profiles – Comparison – Estimate of the Wall Dispersion Coefficient
In this section, we focus on the nearwall void fraction profiles.
In a recent paper (Marfaing et al., 2016), we studied analytically, with the twofluid formalism, the problem of a fully developed “laminar” adiabatic bubbly flow in a vertical pipe. The exchange of momentum between the phases was described as the sum of drag, lift, wall and dispersion contributions, with uniform coefficients over the entire crosssection. Under this assumption, we were able to express analytically the void fraction distribution over the entire domain by a BFBF. This BFBF can be used as a test case for the verification of multiphase CFD codes (see BFBF case in the V&V report of NEPTUNE_CFD).
The dispersion force was represented by Equation (1).
Here, in order to make experimental comparisons, we no longer assume these coefficients are uniform over the entire domain. However, it is legitimate to consider them uniform over a localized region of space, in particular near a wall.
We begin by presenting, in Section 2.1, the main ideas that lead to the derivation of an analytical expression. Some of the assumptions used in (Marfaing et al., 2016) are removed in this work. In particular, dispersion and drag coefficients are only assumed to be uniform near the wall. As a consequence, the analytical expression is restricted to the nearwall region. We show that the void fraction profile results from a balance between dispersion and wall forces.
In Section 2.2, we compare the experimental void fraction profiles obtained by Nakoryakov et al. (1996) with the analytical form derived in Section 2.1. And we use this data to estimate the dispersion coefficient at the wall.
2.1 Analytical NearWall Void Fraction Profile
Consider a low Reynolds number (“laminar”) steady, axisymmetric, fully developed bubbly flow in a vertical pipe with internal radius R (Fig. 8). In particular, no phase change occurs, and the variations in gas and liquid densities are neglected.
Figure 8
Bubbly flow in a vertical pipe. 
The flow being steady and fully developed, it follows from the mass balance that liquid and gas velocities for k = g or l are purely axial and only depend on the radial coordinate r:(2)
Because the Reynolds number is low, we neglect Reynolds stresses in the gas phase, a common assumption in theoretical studies of “laminar” flows (Antal et al., 1991; Azpitarte and Buscaglia, 2003). It is also common in these studies to neglect viscous stresses in the gas phase. Strictly speaking, this may be possible only some distance away from the wall. Let δ_{g} denote the thickness of the boundary layer where gas viscosity is not negligible. Denoting by d_{b} the bubble diameter and y = R − r the distance to the wall, we assume:(3)and we place ourselves in the region(4)
In this region, viscous and Reynolds stresses are neglected, and the momentum balance for the gas phase can be written:(5)where stands for the interfacial momentum exchange term. Because of Equation (2), the lefthand side of the momentum balance is zero.
The interfacial momentum transfer term can be written as the sum of four contributions (Drew and Passman, 1999; Ishii and Hibiki, 2010; Morel, 2015): drag, lift, wall, and dispersion force. Because of Equation (2), virtual mass contribution is zero. As we shall see, wall and dispersion forces are dominant.
Drag and lift forces are respectively expressed as(6)where C_{D} is the drag coefficient, d_{b} the bubble diameter, the relative velocity between the phases, and C_{L} the lift coefficient. Unlike in (Marfaing et al., 2016), coefficients C_{D} and C_{L} are no longer supposed to be uniform over the entire domain. We assume the drag coefficient is uniform in the region δ_{g} < y < d_{b}. Consequently, because gas viscosity is negligible in this region, the relative velocity is also uniform, and nonzero.
The dispersion force, proportional to the void fraction gradient, results in the migration of bubbles from high to low void fraction regions. Unlike in (Marfaing et al., 2016), the dispersion coefficient is considered as uniform only in the region δ_{g} < y < d_{b}, with value D_{wall}. In this region, the contribution of the dispersion force therefore writes:(7)
The wall force pushes the bubbles away from the wall, hence ensuring the zero void fraction condition observed experimentally. Let denote the unit vector normal to the wall pointing towards the fluid. Antal et al. (1991) express the wall force contribution as(8)
Coefficients C_{W1} and C_{W2} are determined by Antal et al. (1991) as follows: C_{W1} = −0.104 − 0.06U_{R} (with the relative velocity in m/s) and C_{W2} = 0.147.
The projection of Equation (5) onto the radial direction writes:(9)
Rearranging terms in Equation (9), and simplifying by the void fraction, it follows:(10)
We recall that this expression has been derived in the region δ_{g} < y < d_{b}, where the relative velocity is uniform and nonzero. For small distances to the wall, the dominant terms are the wall force, which goes to infinity, and the logarithmic derivative of α_{g}, because α_{g} tends to zero. Before simplifying Equation (10), we begin by rewriting it in nondimensional form. Let(11)be the nondimensional distance to the wall,(12)be the nondimensional liquid velocity and the nondimensional pressure, and(13)be the nondimensional bubble dispersion coefficient – we recall that, in the region δ_{g} < y < d_{b}, the drag and dispersion coefficients C_{D} and D_{wall} are constant, and the relative velocity U_{R} is constant and nonzero. Equation (10) can then be rewritten:(14)
From Equation (14), the void fraction profile α_{g} (y^{*}) depends on nondimensional numbers which fall into three categories:

nondimensional space coordinate: δ_{g}^{*} defined by . From Equation (3), δ_{g}^{*} ≪ 1. We can neglect gas viscosity for δ_{g}^{*} < y^{*},

nondimensional dependent variables V_{l}^{*} and P^{*} defined by Equation (12),

nondimensional hydrodynamic coefficients: D_{wall}^{*}, C_{L}, C_{W1} and C_{W2}.
For small distances to the wall, the dominant terms in Equation (14) are the wall force, which goes to infinity, and the logarithmic derivative of α_{g}, because α_{g} tends to zero. So Equation (14) can be approximated by:(15)
This simplification reduces the number of nondimensional parameters involved in the void fraction profile α_{g} (y^{*}):

the nondimensional space coordinate δ_{g}^{*},

the ratio of hydrodynamic coefficients C_{W2}/D_{wall}^{*}.
Integration of (15) yields the nearwall void fraction profile:(16)where B is a nondimensional integration constant, depending on the gas flow rate. This expression is of course consistent with the BFBF derived in (Marfaing et al., 2016). However, some of the assumptions used in (Marfaing et al., 2016) are removed in the derivation of Equation (16).
In Section 2.2, we compare Equation (16) with Nakoryakov et al.’s (1996) measurements. We will also be able to estimate the dispersion coefficient.
2.2 Experimental Comparison – Estimate of the Wall Dispersion Coefficient
For each run carried out by Nakoryakov et al. (1996), we examine the void fraction measurements in the region y < d_{b}, y standing for the distance to the wall, and d_{b} for the bubble diameter. In order to test Equation (16), we plot the graph of the logarithm of the void fraction ln α_{g} as a function of ln y^{*}. We calculate the leastsquares line and the correlation coefficient R^{2}. The example of run Ub103 is displayed in Figure 9. It must be noted that for this run, the measured void fraction vanishes nowhere.
Figure 9
Graph of ln α_{g} as a function of ln y ^{*} for run Ub103. The squares are the experimental points, and the solid line is the calculated leastsquares line. 
Figure 9 reveals that the nine experimental points of run Ub103 are almost aligned. The correlation coefficient R^{2} is 97% and the slope of the leastsquares line is 2.1. So the void fraction profile in the nearwall region is satisfactorily represented by Equation (16), with . Using the value C_{W2} = 0.15 recommended by Antal et al. (1991), we can estimate the nondimensional dispersion coefficient as D_{wall}^{*} = 0.070.
The same procedure is applied for the other runs.
No zero values of the void fraction were reported by the experimenters for runs Ub102, Ub103, Ub106, so it is possible to calculate the logarithm of the void fraction ln α_{g}.
Experimental profiles Ub105, Ub108 and Ub109 present isolated points (2 out of 9) where the reported void fraction is zero. This is illustrated in Table 2, which shows the experimental values of α_{g} as a function of y^{*} in the case of run Ub105.
Nearwall void fraction measurements for run Ub105.
In view of the recorded values, we can assume that for y^{*} = 0.2 or 0.3, the void fraction in the experiment was not strictly zero but too low to be detected by the probe.
In runs Ub101, Ub104 and Ub107, the gas flow rate at the bottom of the pipe is very low (see Tab. 1). For this reason, the recorded values of void fraction are one order of magnitude lower than in the other cases, and half the points were 0. This is illustrated in Table 3, which shows the experimental values of α_{g} as a function of y^{*} in the case of run Ub107.
Nearwall void fraction measurements for run Ub107.
Again, we can reasonably assume that void fractions below 0.001 were not detected by the probe.
Consequently, our analysis for runs Ub101, Ub104, Ub105, Ub107, Ub108 and Ub109 is conducted without taking into account these zero values.
The example of run Ub105 and Ub107 is displayed in Figures 10 and 11.
Figure 10
Graph of ln α_{g} as a function of ln y^{*} for run Ub105. The squares are the experimental points, and the solid line is the calculated leastsquares line. 
Figure 11
Graph of ln α_{g} as a function of ln y^{*} for run Ub107. The squares are the experimental points, and the solid line is the calculated leastsquares line. 
In Table 4, we sum up the results of this procedure for all runs.
Evaluation of the nondimensional dispersion coefficient D_{eff}^{*}. Constant C_{W2} is set to 0.15.
High correlation coefficients R^{2} (above 87%) are found for all runs. So the void fraction profile in the nearwall region is satisfactorily represented by our analytical model (Eq. 16). The nondimensional dispersion coefficient D_{wall}^{*} keeps the same order of magnitude over all runs, and lies in the range 0.0530.082.
2.3 Discussion
As explained in Introduction, most existing models of dispersion force are dedicated to turbulent bubbly flows. They are derived in regions where velocity gradients are negligible, especially far away from the walls. In some of these models, the dispersion coefficient is expressed as being proportional to the turbulent kinetic energy of the liquid phase, and vanishes at the walls.
In our work, the focus is on low Reynolds number flows, and we show that assuming a nonzero dispersion coefficient near the wall leads to good predictions when compared to the experiments.
The present experimental comparison is based on a series of nine experimental runs in “laminar” conditions. Additional measurements over a wider range of conditions would be required to examine the dependence of D_{wall}^{*} on the flow parameters. Direct Numerical Simulations (DNS) may also help, taking advantage of the low Reynolds numbers involved.
We can conjecture an important parameter the dispersion coefficient depends on is the wall shear – or to be precise the nondimensional shear . Understanding this relationship is probably easier in low Reynolds number flows. The knowledge gained could shed some light on more general situations, including the high Reynolds number case – existing high Reynolds number dispersion models being developed for the limit case of no shear.
“Laminar” bubbly flows have been given less attention than turbulent flows in the literature, and for the above reasons, it is important to conduct more experiments in this regime.
Having a wider experimental database would also help assess the possible effect of the bubble size distribution. As mentioned in Section 1, Nakoryakov et al. (1996) measure the average bubble diameter at the recording station. There is little information about the distribution of bubble diameters, but in a previous paper by the same team (Kashinsky et al., 1993), a similar test section is used and the authors report that for most of the flow regimes, the bubbles are rather uniform in size.
Conclusion
In this paper, we consider low Reynolds number (“laminar”) bubbly flows in pipes and we investigate the void fraction profile near the walls.
We begin by using the NEPTUNE_CFD code to simulate a series of low Reynolds number (“laminar”) adiabatic bubbly flow experiments performed by Nakoryakov et al. (1996). The calculations are seen to be in correct agreement with the experimental data. In particular, we can see that using the standard R_{ij}ε model of the code yields results for the liquid velocity fluctuations that are of the right order of magnitude.
In the second part, we focus on the nearwall region. We begin by adapting the argument in Marfaing et al. (2016) to derive an analytical expression for the void fraction profile in the local region δ_{g} < y < d_{b}, where the dispersion coefficient is assumed to be uniform and nonzero, and where the gas viscosity and Reynolds stresses are neglected. In this region, the void fraction profile is found to result from a balance between wall and dispersion forces.
Then, the analytical profile is compared to Nakoryakov et al.’s (1996) experimental results. High correlation coefficients R^{2} (above 87%) are found for all runs, thus validating the analytical derivation. At the same time, the calculated leastsquares lines enable us to evaluate the nondimensional coefficient D_{wall}^{*} to be used in our dispersion force model. D_{wall}^{*} is consistently found to be about 0.07 for the considered runs.
It would be interesting in future work to study the dependence of D_{wall}^{*} on the flow conditions. “Laminar” bubbly flows have been given less attention than turbulent flows in the literature. So it would be important to carry out more experiments.
It is also important to mention that, in the literature, most models of dispersion force are dedicated to turbulent bubbly flows. They are derived in regions where velocity gradients are negligible, especially far away from the walls. In some of these models, the dispersion coefficient is expressed as being proportional to the turbulent kinetic energy of the liquid phase, and vanishes at the walls. Here, on the contrary, the use of a nonzero dispersion coefficient is shown to yield good predictions. Although our comparisons have been carried out for “laminar” flows, we believe that using a dispersion coefficient which does not vanish at the wall is also relevant in turbulent flows.
Acknowledgments
This work was performed in the frame of the NEPTUNE Project, financially supported by CEA (Commissariat à l’Énergie Atomique), EDF (Électricité de France), IRSN (Institut de Radioprotection et de Sûreté Nucléaire) and AREVANP.
References
 Antal S.P., Lahey R.T., Flaherty J.E. (1991) Analysis of phase distribution in fully developed laminar bubbly twophase flow, Int. J. Multiphase Flow 17, 5, 635–652. [CrossRef] [Google Scholar]
 Azpitarte O.E., Buscaglia G.C. (2003) Analytical and numerical evaluation of twofluid model solutions for laminar fully developed bubbly twophase flows, Chem. Eng. Sci. 58, 3765–3776. [Google Scholar]
 Boyd C. (2014) Perspectives on CFD analysis in nuclear reactor regulation, Keynote Lecture, CFD4NRS5 Conference, 911 September, Zurich, Switzerland. [Google Scholar]
 Davidson M.R. (1990) Numerical calculations of twophase flow in a liquid bath with bottom gas injection: the central plume, Appl. Math. Modelling 14, 67–76. [CrossRef] [Google Scholar]
 Drew D.A., Passman S.L. (1999) Theory of multicomponent fluids, Springer, New York, ISBN 0387983805. [Google Scholar]
 Ishii M., Hibiki T. (2010) Thermofluid dynamics of twophase flows, Springer, New York. [Google Scholar]
 Kashinsky O.N., Timkin L.S., Cartellier A. (1993) Experimental study of “laminar” bubbly flows in a vertical pipe, Exp. Fluids 14, 308–314. [Google Scholar]
 Laviéville J., Mérigoux N., Guingo M., Baudry C., Mimouni S. (2015) A generalized turbulent dispersion model for bubbly flow numerical simulation in NEPTUNE_CFD, Proc. NURETH2015 Conference, August 30September 4, 2015, Chicago. [Google Scholar]
 Lopez de Bertodano M. (1998) Twofluid model for twophase turbulent jet, Nucl. Eng. Des. 179, 11, 65–74. [CrossRef] [Google Scholar]
 Marfaing O., Guingo M., Laviéville J., Bois G., Méchitoua N., Mérigoux N., Mimouni S. (2016) An analytical relation for the void fraction distribution in a fully developed bubbly flow in a vertical pipe, Chem. Eng. Sci. 152, 579–585. [Google Scholar]
 Mimouni S., Archambeau F., Boucker M., Laviéville J., Morel C. (2009) A secondorder turbulence model based on a Reynolds stress approach for twophase flow – part I: adiabatic cases, Sci. Technol. Nucl. Install. 2009, 792395. [CrossRef] [Google Scholar]
 Morel C. (2015) Mathematical modeling of disperse twophase flows, Springer, New York. [CrossRef] [Google Scholar]
 Nakoryakov V.E., Kashinsky O.N., Randin V.V., Timkin L.S. (1996) Gasliquid bubbly flows in vertical pipes, J. Fluids Eng. 118, 377–382, Experimental data are available online https://scholar.lib.vt.edu/ejournals/JFE/data/JFE/DB96377/. [Google Scholar]
 Podowski M.Z. (2014) Model verification and validation issues for multiphase flow and heat transfer simulation in reactor systems, Keynote Lecture, CFD4NRS5 Conference, 911 September, Zurich, Switzerland. [Google Scholar]
Cite this article as: O. Marfaing, M. Guingo, J.M. Laviéville and S. Mimouni (2017). Analytical Void Fraction Profile Near the Walls in Low Reynolds Number Bubbly Flows in Pipes: Experimental Comparison and Estimate of the Dispersion Coefficient, Oil Gas Sci. Technol 72, 4.
All Tables
Evaluation of the nondimensional dispersion coefficient D_{eff}^{*}. Constant C_{W2} is set to 0.15.
All Figures
Figure 1
Computed (green) and experimental (black) void fraction profiles run Ub102. 

In the text 
Figure 2
Computed (green) and experimental (black) liquid velocity profiles run Ub102. 

In the text 
Figure 3
Computed (green) and experimental (black) axial liquid velocity fluctuations run Ub102. 

In the text 
Figure 4
Computed (blue) and experimental (black) void fraction profiles run Ub103. 

In the text 
Figure 5
Computed (blue) and experimental (black) liquid velocity profiles run Ub103. 

In the text 
Figure 6
Computed (blue) and experimental (black) axial liquid velocity fluctuations run Ub103. 

In the text 
Figure 7
Computed (blue) and experimental (black) void fraction profiles run Ub102. The computations are performed without taking into account liquid velocity fluctuations. 

In the text 
Figure 8
Bubbly flow in a vertical pipe. 

In the text 
Figure 9
Graph of ln α_{g} as a function of ln y ^{*} for run Ub103. The squares are the experimental points, and the solid line is the calculated leastsquares line. 

In the text 
Figure 10
Graph of ln α_{g} as a function of ln y^{*} for run Ub105. The squares are the experimental points, and the solid line is the calculated leastsquares line. 

In the text 
Figure 11
Graph of ln α_{g} as a function of ln y^{*} for run Ub107. The squares are the experimental points, and the solid line is the calculated leastsquares line. 

In the text 