See discussions, stats, and author profiles for this publication at: [Link]

net/publication/44231711

Experimental investigation and CFD simulation of slug flow in horizontal

channels

Article
Source: OAI

CITATIONS READS

11 902

4 authors, including:

Horst-Michael Prasser Thomas Höhne

ETH Zurich Helmholtz-Zentrum Dresden-Rossendorf
161 PUBLICATIONS 2,432 CITATIONS 142 PUBLICATIONS 1,849 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

TopVox View project

Investigation of coolant mixing at the ROCOM test facility View project

All content following this page was uploaded by Thomas Höhne on 30 May 2014.

The user has requested enhancement of the downloaded file.

Wissenschaftlich-Technische Berichte
FZD-485
September 2007

Christophe Vallée, Thomas Höhne,

Horst-Michael Prasser, Tobias Sühnel

Experimental investigation and CFD simulation of slug flow

in horizontal channels

Technical Report
2
Technischer Fachbericht
Experimentelle Untersuchung und CFD-Simulation von
Schwallströmung in horizontalen Kanälen
Technical Report
Experimental investigation and CFD simulation of slug flow in
horizontal channels
Reaktorsicherheitsforschung-Vorhaben-Nr./ 150 1265
Reactor Safety Research-project No.:

Vorhabenstitel: Aufbau und Durchführung von Experimenten an der

Mehrzweck-Thermohydraulikversuchsanlage
TOPFLOW für generische Untersuchungen von
Zweiphasenströmungen und die Weiterentwicklung
und Validierung von CFD-Codes.

Project Title: Construction and execution of experiments at the

multi-purpose thermal hydraulic test facility
TOPFLOW for generic investigations of two-phase
flows and the development and validation of CFD
codes.

Autoren / Author(s): Christophe Vallée, Thomas Höhne,

Horst-Michael Prasser, Tobias Sühnel

Dienststelle der Autoren / Forschungszentrum Dresden-Rossendorf e.V.

Performing Organisation: Institut für Sicherheitsforschung

Berichtsdatum / September 2007

Publication Date:

Berichts-Nr. / Report-No.: FZD-485

Das diesem Bericht zugrunde liegende Vorhaben wurde mit Mitteln des Bundesministeriums für
Wirtschaft und Technologie unter dem Förderkennzeichen 150 1265 gefördert. Die Verantwortung für
den Inhalt dieser Veröffentlichung liegt bei den Autoren.

3
4
 Berichtsblatt
1. ISBN oder ISSN 2. Berichtsart
Technischer Fachbericht
3a. Titel des Berichts
Experimentelle Untersuchung und CFD-Simulation von Schwallströmung in horizontalen
Kanälen
3b. Titel der Publikation

4a. Autoren des Berichts (Name, Vorname(n)) 5. Abschlussdatum des Vorhabens

Christophe Vallée, Thomas Höhne, Horst-Michael Prasser, 30.09.2006
Tobias Sühnel 6. Veröffentlichungsdatum
4b. Autoren der Publikation (Name, Vorname(n)) September 2007
7. Form der Publikation
Broschüre
8. Durchführende Institution(en) (Name, Adresse) 9. [Link]. Durchführende Institution

*)
Forschungszentrum Dresden-Rossendorf e.V. 10. Förderkennzeichen
Institut für Sicherheitsforschung 150 1265
11a. Seitenzahl Bericht
Postfach 510119
48
01314 Dresden 11b. Seitenzahl Publikation

13. Fördernde Institution (Name, Adresse) 12. Literaturangaben

24
Bundesministeriums für Wirtschaft und Technologie (BMWi) 14. Tabellen
11019 Berlin 1
15. Abbildungen
28
16. Zusätzliche Angaben
17. Vorgelegt bei (Titel, Ort, Datum)

18. Kurzreferat
Für die Untersuchung von geschichteten Zweiphasenströmungen, wurden am Forschungszentrum
Dresden-Rossendorf (FZD) zwei horizontale Kanäle mit rechteckigem Querschnitt gebaut. Diese
Kanäle ermöglichen die Untersuchung Luft/Wasser-Strömungen im Gleich- und Gegenstrom bei
Atmosphärendruck und Raumtemperatur, insbesondere die Entstehung von Schwallströmungen. Die
Teststrecken wurden aus Acrylglas gefertigt, so dass optische Messmethoden angewendet werden
können, wie etwa Hochgeschwindigkeits-Videoaufnahmen bzw. die PIV-Technik (particle image
velocimetry). Die Auswahl des rechteckigen Querschnitts erlaubt einen optimalen Einsatz dieser
Messmethoden. Zudem wurden dynamische Druckmessungen durchgeführt und mit der
Hochgeschwindigkeitskamera synchronisiert. CFD Nachrechnungen geschichteter Strömungen
wurden mit dem kommerziellen Strömungsberechnungscode ANSYS CFX durchgeführt. Das Euler-
Euler Zweifluid-Modell mit der Option für freie Oberflächen wurde auf einem Berechnungsgitter mit
4·105 Kontrollvolumen angewendet. Die Turbulenz lies sich mit einem separaten Ansatz für die
jeweilige Phase des k-ω basierenden SST Modells (shear stress transport) beschreiben. Die
Ergebnisse der Berechnungen stimmen gut im Hinblick auf Schwallentstehung, -geschwindigkeit und
Wellenbrechung überein. Diese qualitativ gute Übereinstimmung zwischen Simulation und
Experiment ist viel versprechend und zeigt, dass CFD als nützliches Tool zur Beschreibung von
horizontalen Zweiphasenströmungen eingesetzt werden kann. Weiterhin wurden Vorrausrechnungen
durchgeführt, um den Einsatz des CFD-Codes für Schwallströmungen in einer realen Geometrie und
bei Parametern zu zeigen, die für die Reaktorsicherheit relevanten sind. Es wurde ein
Zweiphasengemisch von Wasser und Sattdampf bei 50 bar und 263,9°C in einem Modell des
Heißstranges eines deutschen Druckwassereaktors (Typ Konvoi) untersucht. Die Ergebnisse dieser
Simulation zeigen die Entstehung von Wellen im horizontalen Teil des Heißstranges und deren
Anwachsen zu Schwallen im Bogen zum Dampferzeuger.
19. Schlagwörter
Horizontale Zweiphasen-Strömung, Phasengrenze, Schwallströmung
20. Verlag 21. Preis

5
6
 Document Control Sheet
1. ISBN or ISSN 2. Type of Report
Technical Report
3a. Report Title
Experimental investigation and CFD simulation of slug flow in horizontal channels
3b. Title of Publication

4a. Author(s) of the Report (Family Name, First Name(s)) 5. End of Project
Christophe Vallée, Thomas Höhne, Horst-Michael Prasser, 30.09.2006
Tobias Sühnel 6. Publication Date
4b. Author(s) of the Publication (Family Name, First Name(s)) September 2007
7. Form of Publication
Booklet
8. Performing Organisation(s) (Name, Address) 9. Originator's Report No.

*)
Forschungszentrum Dresden-Rossendorf e.V. 10. Reference No.
Institut für Sicherheitsforschung 150 1265
11a. No. of Pages Report
Postfach 510119
48
01314 Dresden 11b. No. of Pages Publication

13. Sponsoring Agency (Name, Address) 12. No. of References

24
Bundesministeriums für Wirtschaft und Technologie (BMWi) 14. No. of Tables
11019 Berlin 1
15. No. of Figures
28
16. Supplementary Notes

17. Presented at (Title, Place, Date)

18. Abstract
For the investigation of stratified two-phase flow, two horizontal channels with rectangular
cross-section were built at Forschungszentrum Dresden-Rossendorf (FZD). The channels
allow the investigation of air/water co-current flows, especially the slug behaviour, at
atmospheric pressure and room temperature. The test-sections are made of acrylic glass,
so that optical techniques, like high-speed video observation or particle image velocimetry
(PIV), can be applied for measurements. The rectangular cross-section was chosen to
provide better observation possibilities. Moreover, dynamic pressure measurements were
performed and synchronised with the high-speed camera system. CFD post-test simulations
of stratified flows were performed using the code ANSYS CFX. The Euler-Euler two fluid
model with the free surface option was applied on grids of minimum 4·105 control volumes.
The turbulence was modelled separately for each phase using the k-ω based shear stress
transport (SST) turbulence model. The results compare well in terms of slug formation,
velocity, and breaking. The qualitative agreement between calculation and experiment is
encouraging and shows that CFD can be a useful tool in studying horizontal two-phase flow.
Furthermore, CFD pre-test calculations were done to show the possibility of slug flow
generation in a real geometry and at relevant parameters for nuclear reactor safety. The
simulation was performed on a flat model representing the hot-leg of the German Konvoi-
reactor, with water and saturated steam at 50 bar and 263.9°C. The results of the CFD-
calculation show wave generation in the horizontal part of the hot-leg which grow to slugs in
the region of the bend.
19. Keywords
horizontal two-phase flow, interfacial area, slug flow
20. Publisher 21. Price

7
8
This report is part of a series, which comprise following reports:

• Construction and execution of experiments at the multi-purpose thermal

hydraulic test facility TOPFLOW for generic investigations of two-phase flows
and the development and validation of CFD codes (Final project report), FZD-
481,
• Experiments on upwards gas-liquid flow in vertical pipes, FZD-482,
• Experiments on two-phase flow in a vertical tube with a moveable obstacle,
FZD-483,
• Experimental investigation of stratified air/water flows in a horizontal channel,
FZD-484,
• Experimental investigation and CFD simulation of slug flow in horizontal
channels, FZD-485,
• CFD models for polydispersed bubbly flows, FZD-486,
• Turbulent Dispersion of Bubbles in Poly-dispersed Gas-Liquid Flows in a
Vertical Pipe, FZD-487,
• Validation of the Multiple Velocity Multiple Size Group (CFX10.0 N x M
MUSIG) Model for Poly-dispersed Multiphase Flows, FZD-487.

All these reports are published as reports of the Forschungszentrum Dresden-

Rossendorf.

Dieser Bericht ist Teil einer Serie, die folgende Einzelberichte umfasst:

• Aufbau und Durchführung von Experimenten an der Mehrzweck-

Thermohydraulikversuchsanlage TOPFLOW für generische Untersuchungen
von Zweiphasenströmungen und die Weiterentwicklung und Validierung von
CFD-Codes (Abschlussbericht), FZD-480,
• Experimente zu aufwärtsgerichteten Gas-Flüssig Strömungen in vertikalen
Rohren, FZD-482,
• Experimente zur Zweiphasenströmung in einem vertikalen Rohr mit
verschiebbarem Hindernis, FZD-483,
• Experimentelle Untersuchung von geschichteten Luft/Wasser Strömungen in
einem horizontalen Kanal, FZD-484,
• Experimentelle Untersuchung und CFD-Simulation von Schwallströmung in
horizontalen Kanälen, FZD-485,
• CFD Modelle für polydisperse Blasenströmungen, FZD-486,
• Turbulente Blasendispersion in einer polydispersen Rohrströmung, FZD-487,
• Validierung des N x M MUSIG Modells für polydisperse Mehrphasen-
strömungen, FZD-487.

Alle Berichte sind als Berichte des Forschungszentrums Dresden-Rossendorf

veröffentlicht.

9
10
Contents

1. Introduction ................................................................................................ 13

2. Previous modelling of slug flow ............................................................... 14

3. CFD model with time-dependent inlet boundary conditions for

slug generation in a short channel........................................................... 16
3.1 The horizontal channel with a separator at the inlet..................................... 16
3.2 Interface capture .......................................................................................... 17
3.2.1 Interface capture method ............................................................................. 17
3.2.2 Water level history during slug flow.............................................................. 18
3.3 CFD model of the duct ................................................................................. 19
3.3.1 Flow solver................................................................................................... 19
3.3.2 Multiphase flow model ................................................................................. 19
3.3.3 Turbulence modelling................................................................................... 20
3.4 Geometry, grid generation and boundary conditions ................................... 21
3.4.1 Geometry and grid generation ..................................................................... 21
3.4.2 Fluid properties, boundary and initial conditions .......................................... 21
3.4.3 Discretization schemes, time step and convergence ................................... 22
3.5 Results......................................................................................................... 22
3.5.1 Qualitative results ........................................................................................ 22
3.5.2 Slug propagation across the duct................................................................. 24
3.5.3 Dynamic pressure ........................................................................................ 25
3.5.4 Particle Image Velocimetry (PIV) of a slug................................................... 27
3.6 Conclusions ................................................................................................. 28

4. CFD model for a self-generation of slug flow.......................................... 29

4.1 The Horizontal Air/Water Channel (HAWAC)............................................... 29
4.2 Influence of the inlet blade position on the two-phase flow generation ........ 30
4.3 Flow pattern map of the HAWAC ................................................................. 31
4.4 CFD model of the duct, turbulence modelling in the interfacial area
and boundary conditions.............................................................................. 32
4.5 Results......................................................................................................... 33

5. Hot-leg model pre-test calculations ......................................................... 36

5.1 Background: the reflux condenser mode and the counter-current flow
limitation....................................................................................................... 36
5.2 The hot-leg model in the pressure chamber of the TOPFLOW-facility......... 37
5.3 CFD model of the hot-leg............................................................................. 38
5.3.1 Modelled scenario........................................................................................ 38
5.3.2 Geometry and grid generation ..................................................................... 39
5.3.3 Boundary conditions .................................................................................... 39
5.3.4 Fluid properties ............................................................................................ 40
5.3.5 Discretization schemes, time step and convergence ................................... 40
5.4 Results......................................................................................................... 40

11
6. Summary and conclusions ....................................................................... 43

7. References ................................................................................................. 44

8. Indexes ....................................................................................................... 46
8.1 Nomenclature and abbreviations ................................................................. 46
8.2 Figures......................................................................................................... 47
8.3 Tables .......................................................................................................... 48

12
1. Introduction
Slug flow as a multiphase flow regime can occur in the main cooling lines of
pressurized water reactors, for instance after a small break Loss of Coolant Accident
(LOCA), and also in other fields, like chemical plants or oil pipelines. The slug flow
regime is usually characterized by an acceleration of the gaseous phase and by the
transition of fast liquid slugs, which carry a significant amount of liquid with high
kinetic energy. It is potentially hazardous to the structure of the system due to the
strong oscillating pressure levels formed behind the liquid slugs as well as the
mechanical momentum of the slugs.

For the experimental investigation of air/water flows, two horizontal channels with
rectangular cross-section were built at Forschungszentrum Dresden-Rossendorf
(FZD). The first channel, allowing co- and counter-current flows, was used for
preliminary studies. As some uncertainties were noticed at the inlet of the test-
section, the HAWAC (Horizontal Air/Water Channel) dedicated to co-current flows
was built. Its improved inlet device provides defined inlet boundary conditions. Both
channels allow in particular the study of air/water slug flow under atmospheric
pressure.

Parallel to the experiments, CFD calculations were carried out. The aim of the
numerical simulations is the validation of prediction of the slug flow with the existing
multiphase flow models built in the commercial code ANSYS CFX. Further it is of
interest to prove the understanding of the general fluid dynamic mechanism leading
to slug flow and to identify the critical parameters affecting the main slug flow
parameters (like e.g. slug length, frequency and propagation velocity; pressure drop).

13
2. Previous modelling of slug flow
According to Manfield (2000), only few attempts using CFD to study slug flow have
been made. He simulated a single, isolated liquid slug using a three-dimensional
CFD Volume-of-Fluid (VOF) method. Three models of varying complexity were
developed: a simulation of the tail of an “infinitely long” slug, a model of short slug
whose front is represented by a solid barrier and a simulation of a slug front as a gas-
liquid interface.

Other groups have also reported studies of slug flow using CFD. Mao and Dukler
(1991) wrote a dedicated CFD programme to simulate the liquid flow around a Taylor
bubble in vertical flow. Pan (1996) reported that this was limited in ability and
flexibility. Issa and Tang (1990) used a similar approach, also for vertical flow. Moe
(1992) simulated the propagation of elongated bubbles in channels initially filled with
liquid, closed at one end and open to the atmosphere at the other. Moe also
simulated the motion of a single bubble in a pipe containing stagnant liquid, but did
not extend this to systems with a net liquid flow. Ubbink (1997) included a case study
of the shape and rise velocity of Taylor bubbles in vertical flow as part of his general
investigation into the modelling of gas-liquid interfaces. He used a moving-wall
boundary condition and simulated a “wedge” of axisymmetric vertical flow. The initial
position of the interface was taken from experimental observations and Ubbink
simulated the development of the interface shape with time.

Recently, Anglart and Podowski (2002) investigated the fluid mechanics of Taylor
bubbles and slug flows in vertical, circular channels using detailed, three-dimensional
CFD simulations. The Volume-of-Fluid model with the interface-sharpening algorithm,
implemented in the commercial CFX-4 code, was used to predict shape and velocity
of Taylor bubbles moving along a vertical channel. Several cases were investigated,
including both a single Taylor bubble and a train of bubbles rising in water. The
computer simulations that were performed revealed the importance of properly
modelling the three-dimensional nature of phenomena governing the motion of Taylor
bubbles.

Lun et al. (1996) used the commercial CFD programme FIDAP to simulate the
growth of a wave on the interface of two immiscible fluids with different properties in
horizontal stratified gas-liquid flow. The method used to simulate the liquid surface
was not described. A coarse two-dimensional grid was developed to represent the
flow domain. Significant numerical instabilities during the calculation led to
convergence difficulties. These were ascribed to the coarseness of the meshes.
Results show the growth of a wave until it completely broaches the pipe and thus
initiates slug flow. However, it seems unlikely that the simulated wave growth was
completely unbiased by the numerical instabilities.

A three-dimensional slug flow study has been performed by Ujang (2003) for the
movement of large Taylor bubbles in a downward inclined pipe. The free surface
simulation was conducted using a moving wall boundary. By adjusting the wall speed
in accordance with the Taylor bubble rise velocity, the bubble front could be artificially
frozen in the computational domain and detailed studies on the formation of the
bubble front could be carried out. Close agreement was obtained with air-water

14
experimental data in a 1-inch diameter pipe for a downward angle of -15° and for
different fluid velocities.

A systematic study of numerical simulation of slug flow in horizontal pipes using CFX-
5 was carried out by Frank (2003). It was shown that the formation of the slug flow
regime strongly depends on the wall friction of the liquid phase. In simulations using
inlet/outlet boundary conditions it was found, that the formation of slug flow regimes
strongly depends on the agitation or perturbation of the inlet boundary conditions.
Furthermore Frank showed that the length of the computational domain plays an
important role in slug formation.

Experimental data was delivered by Woods and Hanratty (1999) for a horizontal
pipeline with a diameter of 76.3 mm and a length of 20 m. The Perspex pipe
segments make it possible to film the flow. Further, liquid holdup and dynamic
pressure were measured. Similar experiments were performed by Soleimani and
Hanratty (2003) for a 18.3 m long pipe of 25.4 mm diameter.

Nevertheless, the direct comparison between CFD calculations and measurements

of the slug generation mechanisms and its propagation in horizontal pipes was not
analysed. The recent improvements of the multiphase flow modelling in the ANSYS
CFX code make it now possible to simulate in detail these mechanisms. In order to
validate the implemented models, high-resolution measurement data is needed in
time and also in space. Thanks to the optical access of the test channels built at
Forschungszentrum Rossendorf, it is possible to study detailed local stratified
air/water flow phenomena. These experimental results give an important input for
two-phase flow CFD model validation (i.e. interfacial momentum transfer, turbulent
profiles of each phase).

15
3. CFD model with time-dependent inlet boundary conditions
for slug generation in a short channel
3.1 The horizontal channel with a separator at the inlet

Experiments were carried out at a horizontal duct mounted between two separators
(Fig. 3.1). This configuration allows both co-current and counter-current flow
experiments. The two-meter long acrylic glass test section has a rectangular cross-
section (height x width: 250 x 50 mm²).

air inlet air outlet

V& G 50
P

250
2350

zf

pump

V& L jet pump

nP

Fig. 3.1: Schematic view of the horizontal channel with inlet separator and
modelled part (red)

The flow rates for both air and water are controlled separately by:
• an air flow controller in a range of up to 1650 l/min (i.e. 2.2 m/s superficial
velocity in the test section);
• a frequency transformer for the pump motor to control the water flow rate.
A jet pump is driving the water flow, while the air is being injected at the top of one of
the separators depending on the type of experiment (co-current or counter-current
flow). The directly pumped water flow rate is measured by an ultrasonic flow meter,
which is based on the sound wave propagation time measurement. A jet pump is
used to amplify the water flow rate in order to enhance the operation range of the
facility. The increase of the water flow rate in the jet pump is calculated from the
dynamic pressure difference measured at two positions within the convergent part of
the jet pump. The best-achieved amplification factor of the jet pump (i.e. the total
water flow divided by the directly pumped water flow) was about 1.8. The maximum
total water flow rate reaches about 10 l/s (i.e. 0.8 m/s superficial velocity), whereas
the measurement accuracy is about ±10%.

16
First tests have shown that in co-current mode the flow rates are sufficient to
establish the following flow regimes: stratified flow (smooth and wavy flows), slug
flow and elongated bubble flow.

Due to the rectangular cross-section, the flow can be observed very well from the
side of the duct. So, optical techniques, like high-speed video observation and
particle image velocimetry (PIV) can be applied to record the flow pattern and to
extract quantitative information. For dynamic pressure measurements, piezoelectric
transducers were mounted on top of the duct. Their data acquisition system was
synchronised with the high-speed video camera.

3.2 Interface capture

3.2.1 Interface capture method

To make quantitative observations, the flow was filmed with a high-speed video
camera from the side of the duct (see Fig. 3.2, corresponding to the modelled region
of Fig. 3.1). Then the pictures have been analysed to capture the gas-liquid interface.

Fig. 3.2: Example of an image taken from the high-speed camera during
counter-current stratified wavy flow (superficial velocities: JL = 0.14 m/s;
JG = -2.2 m/s)

The capture method consists in the following steps:

1. a background subtraction with an image of the completely filled duct;
2. a pixel detection in each vertical line:
• of the darkest pixel;
• of the minimum of a grey-level time variation;
3. picking out of the pixels detected in step 2, the one which fits best into a
continuous interface line (see Fig. 3).

Fig. 3.3: Detected interface line for the image given in Fig. 3.2 with subtracted
background

Finally, the interface is represented by a water level as function of the duct length z
and the time t: h(z, t). The spatial resolution of the pictures is about 1.1 mm/pixel.

17
The accuracy of the interface detection algorithm depends on the thickness of the
interface in the images, which is less than 6 pixels for most of the observed stratified
flows. Therefore, the accuracy of the water level measurement is about ±3 pixels (i.e.
±3.4 mm).

3.2.2 Water level history during slug flow

The superficial velocities of the modelled slug flow (see Fig. 3.4) were as follows:
0.69 m/s for the water flow (i.e. 8.68 l/s) and 2.2 m/s for the air flow (i.e. 1650 l/min).

Fig. 3.4: Slug flow - studied cross-sections (blue and red lines) and characteristic
interface step (yellow box)

The water level in a cross-section as a function of time can be extracted from the
function h(z, t) by keeping z constant. The accuracy of the water level detection is in
the range mentioned above except in the region of the slug front, where the definition
of a water level is difficult due to the generated two-phase mixture. Fig. 3.5 shows the
resulting water level history for the two cross-sections indicated in Fig. 3.4, whose
axial positions are as follows:
• at zf = 0.75 m, after the first flange (left line in Fig. 3.4);
• just before the duct outlet, at zf = 2.17 m (right line in Fig. 3.4).
The water level was measured with a frequency of 100 Hz, which corresponds to the
frame rate of the high-speed camera. From Fig. 3.5 can be seen that the average
water level decreases in flow direction because of the pressure drop of approximately
500 Pa. Each slug passing the cross-sections is recognised by a significant peak
marked by the boxes. After the passage of the slug, the water level decreases by
about 50 mm compared to the level in front of the slug. Then it is increasing slowly
until it reaches the critical level and the next slug appears. This phenomenon also
shows that a slug expels a significant amount of water from the duct to the second
separator.

18
 250

200
Water level [mm]

150
cross-section:
zf = 0.7 m
outlet
100

slug peaks
50

0
0 5 10 15 20
Time [s]

Fig. 3.5: Time-dependent water level during slug flow (measuring frequency:
100 Hz)

3.3 CFD model of the duct

3.3.1 Flow solver

The CFD calculation was performed using the software package CFX-5.7 (ANSYS
CFX, 2004). The main characteristics of the solver are:
• Finite element based finite volume method for hybrid unstructured
meshes;
• Coupled solver for momentum and mass;
• Implicit pressure based formulation;
• Co-located variables;
• Algebraic multi-grid method;
• High parallelisation;
• A selection of different time and space discretization schemes;
• Availability of different turbulence models (k-ε model, k-ω model, SST
model, Reynolds stress models, LES with Smagorinsky subgrid model);
• Multi-phase formulations: Lagrangian particle tracking, homogenous multi-
phase model, Euler-Euler multi-phase formulation (particle model, mixture
model).
3.3.2 Multiphase flow model

For slug flow simulations in horizontal pipes and channels, an appropriate multiphase
flow model has to be selected in combination with the ANSYS CFX free surface
model. Due to the geometrical scale of the horizontal channel it is not possible to
resolve the spatial structure of the free surface into the micro-scale with the CFD

19
model. A slug flow simulation can therefore not cover onset and evolution of Kelvin-
Helmholtz instabilities at the free surface. The interface drag law applied to the
location of the free surface must consider the influence of these free surface
instabilities on the macro-scale flow properties. Additionally, the movement of the
free surface at the inlet boundary condition introduces instabilities.

For free surface simulations, the inhomogeneous multiphase model recommended

by Frank (2003) was used, where the gaseous and liquid phases can be partially
mixed in certain areas of the flow domain. In this case, the local phase demixing after
gas entrainment is controlled by buoyancy and interphase drag and is not hindered
by the phase interface separating the two fluids. A surface capturing method using a
compressive advection scheme is used in ANSYS CFX. Explicit interface tracking
methods (e.g. VOF or Level Sets methods) were not applied. A further decision has
to be made regarding the applied fluid morphology and interphase drag law for the
multiphase flow. The total drag force Fd is most conveniently expressed in terms of
the dimensionless drag coefficient CD,

1
⋅ ρ LG ⋅ (UL − U G ) ⋅ A ⋅ C D
2
Fd = (3.1)
2

where ρ is the fluid density, (UL - UG) is the relative speed and A is the projected area
of the interphase in flow direction. In (1), L describes the liquid phase and G the
gaseous phase.

The interfacial length scale was set to dLG = 0.001 m, according to a similar
assumption made by Frank (2003). The drag law for slug flow was taken from Ishii
(1990, p. 37) and the model constant was fixed to CD = 0.44. The surface tension
coefficient was set to 0.072 N/m.

3.3.3 Turbulence modelling

Generally, it is possible to apply a homogeneous turbulence model to both phases. In

this study, the fluid-dependent shear stress transport (SST) turbulence models
(Menter, 1993) were selected for each phase. Damping of turbulent diffusion at the
interface has not been considered. The k -ω based SST model accounts for the
transport of the turbulent shear stress and gives highly accurate predictions of the
onset and the amount of flow separation under adverse pressure gradients. The SST
model combines the advantages of the Wilcox (1994) and of the k-ε model.

To take buoyancy effects into account, a buoyancy production term was used in the
turbulence model and was included in the k- and ω-equations for additional
turbulence production and dissipation. If the full buoyancy model is being used, the
buoyancy production term Pkb is modelled according to ANSYS CFX (2004) as:

μt r
Pkb = − ⋅ g • ∇ρ (3.2)
ρ ⋅ Prt

where ρ is ther density of the fluid, Prt the turbulent Prandtl number, μt the turbulent
viscosity and g the gravity vector.

20
3.4 Geometry, grid generation and boundary conditions

3.4.1 Geometry and grid generation

Due to very high numerical efforts of transient slug flow simulations, the modelling of
the complete test facility is not feasible. In order to keep computational time within
acceptable limits, only the horizontal test section with its rectangular cross-section
was modelled. The model dimensions are: length x height x width = 2000 x 250 x
50 mm³. The grid was created with CFX-MESHBUILD and transferred to CFX-5 PRE
(Fig. 3.6). It consists of 4x105 hexahedral elements (number of nodes: length x
height x width = 400 x 100 x 10), which reduces the computational time and improves
the quality of the mesh compared to tetrahedrons. Two different grid densities were
studied.

a) Mesh density b) Fluid domain (channel)

Fig. 3.6: CFD model of the duct

3.4.2 Fluid properties, boundary and initial conditions

The gaseous phase is air and the liquid phase is water, both at 25°C and at a
reference pressure of 1 bar. The phases have been treated as isothermal and
incompressible. Buoyancy effects between the two phases are taken into account by
the directed gravity term.
In the experimental setup the inlet boundary conditions into the horizontal channel
are not constant. Especially the liquid and gas velocities depend on the conditions
before the entrance into the channel. Therefore, the inlet boundary conditions of the
CFD calculation (water level, water mass flow rate and air mass flow rate) should be
time-dependent. For the calculation, air and water velocities were set to constant
values over the whole inlet cross-section (at zm = 0.0 m). The test conditions of the
recalculated slug flow experiment were as follows: water flow of 8.68 l/s and air flow
of 1650 l/min. In accordance with these conditions and the initial water level of
194 mm, the inlet velocities were set to 8.93 m/s for the air phase and to 0.895 m/s
for the water phase. These velocities were kept constant during the simulation. The
turbulent kinetic energy and dissipation rate at the inlet were set to the standard
values of ANSYS CFX.
Thanks to the interface capture, the water level history of the recalculated experiment
is known at zf = 0.75 m (see paragraph 3.2.2 and Fig. 3.5). At this position, the

21
influence of the duct inlet geometry on the flow is not so prominent anymore. For this
reason, the experimental water level history measured at zf = 0.75 m (see the blue
curve in Fig. 3.5) was used as time-dependent boundary condition for the
calculations at the inlet cross-section (at zm = 0.0 m) of the modelling domain.

A pressure controlled outlet boundary condition has been applied to the cross section
at the other end of the horizontal channel (at zm = 2.0 m). The inner surface of the
channel walls has been defined as hydraulically smooth with a non-slip boundary
condition applied to both gaseous and liquid phases.

In accordance with the experiment, the initial water level and therefore the entire free
surface was set to y0 = 0.194 m, so the liquid phase covers 78% of the channel (see
Fig. 3.7-a). A hydrostatic pressure was assumed for the liquid phase (see Fig. 3.7-b).

a) Air volume fraction, initial state [-] b) Hydrostatic pressure, initial state [Pa]
Fig. 3.7: Initial conditions of the mass fractions and the hydrostatic pressure in
the channel

3.4.3 Discretization schemes, time step and convergence

The parallel transient calculations on 4 processors lasted 3 weeks on a 76 processor

MSC-LINUX cluster (dual CPU compute nodes at 3.2 GHz, each containing 2 GB
RAM).

Due to the high numerical efforts for the transient slug flow simulation, it was not
possible to carry out a full numerical study including the evaluation of the
dependence of numerical solutions on time integration method, time steps and
convergence criteria according to the Best Practice Guidelines (Menter, 2002).

A high resolution discretization scheme was used. For time integration, the fully
implicit second order backward Euler method was applied with a constant time step
of dt = 0.001 s and a maximum of 15 coefficient loops. A convergence in terms of the
RMS values of the residuals to be less than 10-4 could be assured most of the time.

3.5 Results

3.5.1 Qualitative results

The picture sequence (see Fig. 3.8) shows the flow observed during the experiment
in comparison to the corresponding CFD calculation. Since the filmed test section

22
length and the modelled part of the channel are different, the last part of the model
can not be compared with experimental data. This explains the length difference
between the camera frames and the visualisations of calculation results in Fig. 3.8.

ANSYS CFX calculation Experiment

pressure sensors: left right

t = 0.75 s:

t = 0.80 s:

t = 0.85 s:

t = 0.90 s:

t = 0.95 s:

t = 1.00 s:

t = 1.05 s:

t = 1.10 s:

t = 1.15 s:

Fig. 3.8: Measured and calculated image sequences at slug flow conditions

In both cases, a slug is developing. In the CFD calculation, the slug develops
approximately at t = 0.90 s due to a high peak of the experimental water level at the
inlet cross-section of the model. The tail of the calculated slug and the flow behind it
is in good agreement with the experiment. The entrainment of droplets in front of the
slug could not be observed in the calculation. However, the rolling over and breaking
of the wave, characteristic of a slug, are clearly to be seen in Fig. 3.8. These are
created by the shear generated by the high air velocity.

23
In contrast to the measurement, the slug length is increasing in time in the
calculation. This could be a result of the different amount of water in the channel at
the beginning of the experiment and the simulation. While in the simulation, the liquid
phase covered 78% of the channel, it represented about 70% at the beginning of the
experiment. Furthermore, in the experiment, this value was also reduced by a first
slug, which carried a significant amount of water out of the channel. This first slug
could not be simulated. This is probably an effect of the simplified initial conditions
assumed in the calculation. Because of the constant initial water level, it took quite a
long simulation time to establish a wavy flow along the channel, which is necessary
for slug formation. Whereas at the beginning of the measurements, the test channel
was already in a fully established intermittent slugging regime.

3.5.2 Slug propagation across the duct

The slug position was extracted from the experimental pictures after an interface
capture. Since the interface capture algorithm detects only one pixel in each vertical
line, the interface function always presents a discontinuity where the slug is rolling
over (c.f. the yellow box in Fig. 3.4). This point was considered as the slug axial
position.

As remarked in the previous paragraph, the visualised calculation results show that
the slug gains length continuously. For this reason, the slug front propagates
significantly faster than its tail, which is in a qualitative good agreement with the
experiment. Therefore, the slug position was taken as the first axial position (from the
left) where the recalculated slug reaches the top of the duct.

The slug axial position obtained in this way was plotted in function of time in Fig. 3.9
for both experiment and simulation. This was performed with an accuracy of about
±30 mm. As shown in Fig. 3.9, in the experiment the slug moves along the duct with
a constant velocity, in average at 3.7 m/s.

The slug propagation of CFD calculation is divided into two characteristic parts:
• between 0.75 and 0.90 s, as the slug is still in preparation, the wave which
is generating the slug is accelerating (doted part of the red line in Fig. 3.9).
• from t = 0.90 s, when the wave blocks the whole cross-section of the
channel, the slug propagates with a nearly constant velocity in the
calculation, too. The average velocity of the calculated slug is 4.4 m/s,
which is 18% higher than in the experiment.

24
 1600

1400

1200
Slug abscissa z [mm]

Model equation (measurement): Slug abscissa:

1000
measured
y = 3743,5x - 2800,7
2 calculated
800 R = 0,9987
Model equation (calculation): Linear model:
600 y = 4400x - 3482,5 measured
2
R = 0,9986 calculated
400

200

0
0,75 0,80 0,85 0,90 0,95 1,00 1,05 1,10 1,15
Time [s]

Fig. 3.9: Propagation of a slug - comparison between measurement and

calculation

3.5.3 Dynamic pressure

a. Pressure measurements

Since slug flow is a transient phenomenon, suitable pressure sensors have to be

fast. That is why existing piezoelectric pressure transducers with a rise time of 2 µs
were chosen. They were mounted on top of the last two sections of the duct (i.e. at
zf = 1.45 m and 1.95 m, c.f. Fig. 3.1). Moreover, the pressure measurements were
synchronised with the high-speed video camera system. A disadvantage of the
piezoelectric transducers is the drift due to the discharge time of the quartz crystals.
This causes a static error of about 20%. Unfortunately, alternatives were not
available for the present studies.

A typical pressure measurement of a slug flow is shown in Fig. 3.10. The maximum
pressure level behind a slug varies from slug to slug, but it is typically in the order of
kilopascals.

The zoom in on the first slug passage shows that the pressure increases rapidly:
typical rise times are 2 ms (left sensor) and 10 ms (right sensor). An analysis of the
camera pictures taken during the pressure measurements could explain the large
spectrum of rise times. They are in fact an indicator of the slug length: the longer the
slug, the longer the rise time. This means that in the presented measurement (Fig.
3.10), the slug has grown between the two sensors. The pressure decreases
simultaneously on both measuring positions. The time of about 20 ms needed to
reach atmospheric pressure again, is longer than the rise time. The pictures taken by
the camera show that the pressure decreases when the slug leaves the duct towards
the outlet separator.

25
 5000
5000

Pressure [Pa]
4000
4000 3000
2000
1000
3000 0
Pressure [Pa]

-1000
0,1 0,2 0,3 0,4 0,5 0,6
Sensor:
2000 Time [s] left
right
1000

-1000
0 1 2 3 4 5 6 7 8 9 10
Time [s]

Fig. 3.10: Typical pressure measurement by slug flow and zoom in on the first
pressure increase

b. Comparison with the calculation

Because the slug also closes the channel cross-section in the calculation, an area
with high pressure is created after the slug (see Fig. 3.11).

Pressure [kPa]

Fig. 3.11: Calculated pressure field and interface at t = 0.95 s

Fig. 3.12 shows the comparison of the time-dependant pressure at both sensor
positions for the experiment and the CFD calculation. Since it was not possible to
simulate the first slug, there is no pressure peak in the calculation that corresponds
to the first experimentally observed peak. During the travelling of the second slug, the
pressure at the first sensor position increases later in the calculation than in the
experiment, which is due to the lower velocities in the acceleration phase. At the
second sensor position, the pressure increase is simultaneous. Then, the pressure
decreases earlier in the calculation than in the experiment, i.e. the calculated slug

26
becomes faster than in reality. This is in agreement with the remarks made on the
slug position diagram (see paragraph 3.5.2 and Fig. 3.9).

5000

4000

Experiment:
3000
Pressure [Pa]

left sensor
right sensor
2000 Calculation:
left position
right pos.
1000

-1000
0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4
Time [s]

Fig. 3.12: Transient pressure at the left and right sensor positions for experiment
and ANSYS CFX calculation

Furthermore, for the recalculated slug, the pressure peak value is higher in the
calculation (4.4 kPa) than in the measurement (2.3 kPa), whereas the peak value of
the first slug observed in the experiment reaches between 3.6 and 5.0 kPa
(respectively at the first and second sensor positions), which is in much better
agreement with the calculated value.

This difference of the experimental pressure peak value is due to the fact, that the
boundary conditions for the second slug are already influenced by the first slug,
which has cleared a certain amount of water from the channel before. The lower
pressure peak of the second slug can be explained by the smaller amount of water to
be driven by the air pressure. Furthermore, the effective free area on top of the slug
where air flows is bigger. This effect is missing in the calculation, since the first slug
was not reproduced.

3.5.4 Particle Image Velocimetry (PIV) of a slug

To show the inner structure of the velocity field of a slug, a PIV-system was used.
The laser-light sheet was focused to the vertical mid-plane of the test section. The
PIV-camera was directed from the side to the upper part of the duct, in the region of
the first pressure sensor (zf = 1.45 m). The limited size of the observation area
(140 x 105 mm²) did not allow to cover the entire height of the channel of 250 mm.

Fig. 3.13 shows a PIV-picture of a slug with the calculated velocity field. The slug
propagation velocity was 3.8 m/s (in the same region as chosen in paragraph 3.5.2),

27
which is similar to the slug velocity of the recalculated slug. The vector colour shows
that the entrained air flow on top of the slug accelerates the water to absolute local
velocities of about 6 m/s.

Fig. 3.13: Velocity field of the secondary flow inside of a slug (the vector colour
shows the absolute velocity; the vector length and direction show the
relative velocity after subtracting the propagation velocity of the slug)

The vector direction and length show the relative velocities after subtraction of the
propagation velocity of the slug itself. It reveals the secondary flow pattern inside of
the slug. At the slug front, the water steams upwards. On top, it is accelerated in
direction of the slug movement and becomes faster than the slug itself. Due to
gravity, the portions that roll over fall down again. In this way, the roll-over vortex
structure is formed. On the left side of the image, water seems to flow against the
main flow direction. This is true in the coordinate system of the slug, since this water
flow is slower than the slug. In reality it flows in the right direction, of course.

3.6 Conclusions

This preliminary study shows that the slug flow regime can be qualitatively
reproduced with CFD. In order to perform reasonable quantitative comparisons, the
experimental boundary conditions at the inlet have to be constant and well defined.
Because of the separator at the test-section inlet, important inlet parameters such as
water level and velocity profiles cannot be controlled properly and are not stationary
in the present facility. Therefore, detailed investigations require improved inlet
geometry. This was realised in a new test facility, which is described in section 4.

28
4. CFD model for a self-generation of slug flow
4.1 The Horizontal Air/Water Channel (HAWAC)

The Horizontal Air/Water Channel (HAWAC) (Fig. 4.1) is devoted to co-current flow
experiments. A special inlet device provides defined inlet boundary conditions by a
separate injection of water and air into the test-section. The cross-section of this
channel is smaller than the previous one (described in section 3): its dimensions are
100 x 30 mm² (height x width). Thus it fits the field of view of the PIV-system and
makes it possible to cover a wider range of superficial velocities. The test-section is
about 8 m long, and therefore the length-to-height ratio L/H is 80. Alternatively,
related to the hydraulic diameter, the dimensionless length of the channel is L/Dh =
173.
air inlet air outlet

V& G

8m

inlet device
V& L

pump

Fig. 4.1: Schematic view of the Horizontal Air/Water Channel (HAWAC)

The inlet device (Fig. 4.2) is designed for a separate injection of water and air into the
channel. The air flows through the upper part and the water through the lower part of
this device. Because the inlet geometry produces a lot of perturbations in the flow
(bends, transition from pipes to rectangular cross-section), 4 wire mesh filters are
mounted in each part of the inlet device. The filters are made of stainless steel wires
with a diameter of 0.63 mm and have a mesh size of 1.06 mm. They aim at providing
homogenous velocity profiles at the test-section inlet. Moreover, the filters produce a
pressure drop that attenuate the effect of the pressure surge created by slug flow on
the fluid supply systems.

wire mesh filters control rod

air inlet
inclinable blade
water inlet

Fig. 4.2: The HAWAC inlet device

29
Air and water come in contact at the final edge of a 500 mm long blade that divides
both phases downstream of the filter segment. The free inlet cross-section for each
phase can be controlled by inclining this blade up and down. Both, filters and
inclinable blade, provide well-defined inlet boundary conditions for the CFD model
and therefore offer very good validation possibilities.

The water flow rate is measured with a paddle-wheel flow transmitter and adjusted
via the frequency inverter of the pump motor. The air flow rate is measured and
controlled with the thermal mass flow meters of the TOPFLOW-facility. These are
mounted in parallel in order to ensure a high precision over a large measuring range.
The flow rates are measured with an accuracy of ±0,2 l/s for the water and ±1,5% for
the air. The maximum superficial velocities achieved in the test-section are 2 m/s for
the water and 8 m/s for the air. Optical measurements were performed with a high-
speed video camera.

4.2 Influence of the inlet blade position on the two-phase flow generation

The adjustable inclination of the inlet blade separating the phases allows to influence
the generation of the two-phase flow regime. The first contact between gas and liquid
can be modified with the blade position: if the velocities at the end of the blade are
similar, a smooth come together will be managed, else a perturbation can be
introduced in the channel. For example, observations have shown that the inlet
length needed for slug generation can vary from 0.5 to 3 m just by changing the
blade inclination from an opening ration of 20/80 (air/water) to 80/20. Nevertheless,
and despite a short channel length, the flow regime established at the end of the
channel is not significantly affected by the plate inclination.

Furthermore, at high water flow rates, especially when the inlet blade is inclined
down, a hydraulic jump can be realised in the test-section. As shown in Fig. 4.3, the
hydraulic jump is the turbulent transition zone between super- and subcritical flow. In
the supercritical region, the flow is always stratified, whereas after the hydraulic jump
(i.e in the subcritical region) the typical two-phase flow regimes were observed (e.g.
elongated bubble flow, slug flow). The position of the hydraulic jump in the channel
depends on both flow rates and on the inlet blade inclination. When a hydraulic jump
is formed, its position strongly influences the inlet length needed for the generation of
slug flow.

supercritical flow hydraulic jump subcritical flow

inlet blade

Fig. 4.3: Hydraulic jump in the test-section (jL = 0.37 m/s; jG = 0.10 m/s, inlet
blade = 26 mm)

It was observed that the position of the inlet blade not always defines the water level
at the channel inlet. For example when the inlet blade is inclined upwards and the
water flow rate is very low, the water flow separates from the lower surface of the
blade. In this case, the initial water level is lower than the height of the outlet edge of

30
the blade. To create proper inlet boundary conditions at these flow rates, the blade
has to be inclined down.

4.3 Flow pattern map of the HAWAC

A flow pattern map (Fig. 4.4) was arranged on the basis of visual observations of the
flow structure at different combinations of the gas and liquid superficial velocities.
The observed flow patterns are:
• 1 stratified flow
• 2 wavy flow
• 3 elongated bubble flow
• 4 slug flow
Further, sub-categories were defined to consider the slug generation frequency and
the appearance of elongated bubbles:
• a sporadic (transition regime)
• b periodic, but only one type of structure (either slug or elongated bubble)
• c periodic and several types of structure simultaneously

10

elongated bubble flow slug flow

1
1 2
3-a
JL [m/s]

3-b
3-c
4-a
0,1 4-b
4-c
stratified flow wavy flow

0,01
0,01 0,1 1 10
JG [m/s]

Fig. 4.4: Flow pattern map for the rectangular channel (inlet blade in horizontal
position)

In the flow map, some transitions are different from those found by Mandhane (1974)
and Baker (1954). Especially the transitions from stratified to elongated bubble flow
and from wavy to slug flow were observed at higher liquid superficial velocities than
in the literature. This can be attributed to the shorter channel length and would
indicate that the flow regime is not fully established at the end of the test-section.
Another source of discrepancy could be the cross-section geometry: the results from
Mandhane are based on pipe measurements, whereas ours were performed in a
channel with rectangular cross-section.

31
4.4 CFD model of the duct, turbulence modelling in the interfacial area and
boundary conditions

The second channel with rectangular cross-section was modelled using ANSYS
CFX. The model dimensions are 4000 x 100 x 30 mm³ (length x height x width),
which corresponds only to the first half of the test-section (Fig. 4.5-a). The grid
consists of 6x105 hexahedral elements.

a) Fluid domain (channel) b) Air volume fraction, initial state (Zoom)

Fig. 4.5: Model and initial conditions of the mass fractions

The slug flow experiment at a superficial water velocity of 1.0 m/s and a superficial
air velocity of 5.0 m/s was chosen for the CFD calculations. In the experiment, the
inlet blade was in horizontal position. Accordingly, the model inlet was divided into
two parts: in the lower 50% of the inlet cross-section, water was injected and in the
upper 50% air. An initial water level of y0 = 50 mm was assumed for the entire model
length (Fig. 4.5-b).

In the simulation, both phases have been treated as isothermal and incompressible,
at 25°C and at a reference pressure of 1 bar. A hydrostatic pressure was assumed
for the liquid phase. Buoyancy effects between the two phases are taken into
account by the directed gravity term. At the inlet, the turbulence properties were set
using the “Medium intensity and Eddy viscosity ratio” option of the flow solver. This is
equivalent to a turbulence intensity of 5% in both phases. The inner surface of the
channel walls has been defined as hydraulically smooth with a non-slip boundary
condition applied to both gaseous and liquid phases. The channel outlet was
modelled with a pressure controlled outlet boundary condition.

As the goal of the CFD calculation was to induce surface instabilities, which are later
generating waves and slugs, the interfacial momentum exchange and also the
turbulence parameters had to be modelled correctly. Without any special treatment of
the free surface, the high velocity gradients at the free surface, especially in the
gaseous phase, generate too high turbulence throughout the two-phase flow when
using the differential eddy viscosity models like the k-ε or the k-ω model (ANSYS
CFX, 2006). Therefore, certain damping of turbulence is necessary in the interfacial
area because the mesh is too coarse to resolve the velocity gradient in the gas at the
interface. On the gas side of the smooth free surface, this damping should be similar
to that used near a solid wall. Moreover, on the liquid side the advanced model
should take the anisotropy between the normal and the tangential Reynolds stresses
into account. A simple symmetric damping procedure provides for the solid wall-like
damping of turbulence in both gas and liquid phases. Therefore, a damping of

32
turbulent diffusion at the interface was applied in this calculation by an appropriate
source term in the ω-equation:
2
⎛ 6 ⋅ μi ⎞
A ⋅ Δy ⋅ β ⋅ ρ i ⋅ ⎜⎜ B ⋅ ⎟
⎟
(4.1)
⎝ β ⋅ ρ i ⋅ Δn
2
⎠

where A the interface area density, ∆n is the typical grid cell size across the interface,
ρi and µi are the density and viscosity of the phase i. Further, the model parameter B
should have a value of 100.

The parallel transient calculation of 5.0 s of simulation time on 4 processors last 4

days. A high-resolution discretization scheme was used. For time integration, the fully
implicit second order backward Euler method was applied with a constant time step
of dt = 0.001 s and a maximum of 15 coefficient loops. A convergence in terms of the
RMS values of the residuals to be less then 10-4 could be assured most of the time.

4.5 Results

In the following picture sequences (Fig. 4.6 and Fig. 4.7) a comparison is presented
between CFD calculation and experiment: the calculated phase distribution is
visualized and comparable camera frames are shown. In both cases, a slug is
generated. The sequences show that the qualitative behaviour of the creation and
propagation of the slug is similar in the experiment and in the calculation. In the CFD
calculation, the slug develops at approximately t = 1.30 s, induced by instabilities.

t = 1,10 s
t = 1,30 s
t = 1,35 s
t = 1,40 s
t = 1,45 s
t = 1,50 s
t = 1,55 s
t = 1,60 s
t = 1,65 s
t = 1,70 s
t = 1,75 s
t = 1,80 s

Fig. 4.6: Calculated picture sequence at JL = 1.0 m/s and JG = 5.0 m/s (depicted
part of the channel: 1.4 to 4 m after the inlet)

33
Fig. 4.7: Measured picture sequence at JL = 1.0 m/s and JG = 5.0 m/s with ∆t =
50 ms (depicted part of the channel: 0 to 3.2 m after the inlet)

The single effects leading to slug flow that can be simulated are shown in details in
Fig. 4.8. These phenomena are:
• Instabilities and small waves are generated by the interfacial momentum
transfer randomly (Fig. 4.8-a). As a result bigger waves are generated.
• The waves can have different velocities and can merge together (Fig. 4.8-
b and c).
• Bigger waves roll over (Fig. 4.8-c) and can close the channel cross-section
(Fig. 4.8-d).
The needed entrance length for slug generation was defined as the length between
the inlet and there a wave closes nearly the entire cross-section for the first time.
This was observed at about 1.5 m in the experiment and 2.5 m in the calculation.

In contrast to the measurement, the stratified flow after the slug calculated with
ANSYS CFX is too smooth which defers the generation of the next slug. Since the
slug cleared an important amount of water from the channel, the next slug appears
after the channel is slowly filled up again by the transport of liquid from the inlet. This
process takes approximately 1.5 s. In the experiment, small waves are generated
immediately after the slug and create the next one within 0.3 to maximum 0.7 s.
Sources of instabilities are here the high air velocity but also the pressure surge
created by the slugs, particularly when they leave the channel. This effect has not
been proper simulated since just the half of the channel was modelled. Therefore, the
slug frequency cannot be compared at this stage of the simulation.

34
 a) t = 1.27 s b) t = 1.37 s

c) t = 1.47 s d) t = 1.51 s
Fig. 4.8. Details of the slug generation calculated with ANSYS CFX at different
simulation time

35
5. Hot-leg model pre-test calculations
5.1 Background: the reflux condenser mode and the counter-current flow
limitation

In the event of a loss-of-coolant-accident (LOCA) in a pressurised water reactor

(PWR), emergency strategies have to be mapped out, in order to guarantee a safe
removal of heat from the reactor core, also in case of component breakdown. During
a small break LOCA with failure of the high pressure emergency core cooling system
(ECC) and of the main feed pumps, a natural circulation starts in the primary circuit.
This allows the heat removal, also if steam is generated in the reactor core due to the
depressurisation of the primary circuit. But if further the water level in the reactor
pressure vessel (RPV) falls below the hot-leg inlet, only steam will flow to the steam
generator. Therefore the natural circulation breaks down and switches to the reflux
condenser mode.

In the reflux condenser mode, the steam coming from the RPV condensates in the
vertical U-tubes of the steam generator. In each half of the steam generator, the
condensate flows down the tube in which it has been formed. Therefore, about one
half of the condensate flows as usual over the pump to the downcomer, whereas the
other half flows over the hot-leg back to the upper plenum. In the hot-leg, the
condensate have to flow in counter-current to the steam (Fig. 5.1).

Fig. 5.1: Schematic view of the flow in the hot-leg and steam generator during
reflux condenser mode (GRS, 1993)

The horizontal stratified counter-current flow of condensate and steam is only stable
for a certain range of flow rates. If the steam flow increases too much, the
condensate is clogged in the hot-leg. This is the begin of the counter-current flow
limitation (CCFL): the liquid is carried over by the steam and partially entrained in
opposite direction to the steam generator (Fig. 5.2). As a consequence, the hot-leg
and steam generator are flood, which decreases the water level in the RPV and
reduces the core cooling. In case of an additional increase of the steam flow, the

36
condensate is completely blocked and the cooling of the reactor core from the hot-leg
is impossible.

Fig. 5.2: Possible flow phenomena in the hot-leg bend during CCFL (GRS, 1993)

5.2 The hot-leg model in the pressure chamber of the TOPFLOW-facility

For the experimental investigation of stratified two-phase flow in a complex geometry,

the “hot-leg model” was constructed at Forschungszentrum Dresden-Rossendorf.
The hot-leg model is built in the pressure chamber of the TOPFLOW facility, which is
used to perform experiments with steam/water at pressures up to 50 bars and
temperatures up to 264°C, but under pressure equilibrium with the chamber inside
(Fig. 5.3). A special heat exchanger condenses the exhaust steam from the test
section in the chamber. Its cold end is permanently connected to the inner
atmosphere of the chamber in order to guarantee full pressure equilibrium at all times
(Fig. 5.4). Therefore, the test section does not have to support overpressures and
can be designed with thin materials.

Fig. 5.3: The pressure chamber and the test-section components

37
In the hot-leg experiment, the test-section is a steal frame with glass side walls for
high-speed video observations. Its outlines represent a 1:3 hot-leg model of the
German Konvoi-reactor and corresponds to a channel height of 250 mm (straight part
of the hot-leg). The experiments focus on the flow regimes observed in the region of
the elbow and of the steam generator inlet chamber during counter-current flow
limitation.

Fig. 5.4: Scheme of the pressure chamber setup

Because of technical problems with the pressure chamber of the TOPFLOW facility,
no experiments could be performed in the frame of the current project. However CFD
pre-test calculations were done to show the possibility of slug flow simulation in the
geometry of the hot-leg model in a steam/water flow.

5.3 CFD model of the hot-leg

5.3.1 Modelled scenario

The aim of the numerical simulation is the investigation of the feasibility of numerical
prediction of stratified two-phase flow with ANSYS CFX in a real geometry and at
relevant parameters for nuclear reactor safety.

For the present CFD calculation, a hypothetical scenario was selected with a typical
stratified flow in the hot-leg. At the beginning of the simulation, the primary circuit is
filled with a constant coolant level at approximately 3/4 of the main lines diameter.
Nevertheless, steam is produced in the RPV and is streaming through the remaining
cross-section towards the hot-leg into the steam generator. The assumed pressure is
50 bar and the steam is at saturation conditions. The test should show the resulting
surface behaviour of the interphase, for example the formation of waves and/or slugs
in the hot-leg.

38
5.3.2 Geometry and grid generation

Due to very high numerical efforts of transient slug flow simulations, the modelling of
the complete test-section including the separators is not feasible. In order to keep
computational time within acceptable limits, in a first approach only the horizontal
test-section was modelled (Fig. 5.5). The grid was created with CFX-MESHBUILD and
transferred to CFX-PRE. It consists of 1.4 x 105 hexahedral elements, which reduces
the computational time and improves the quality of the mesh compared to
tetrahedrons.

a) simplified model: hot-leg and steam b) grid: zoom on the bend and steam
generator inlet chamber generator inlet chamber
Fig. 5.5: CFD model of the hot-leg test-section

5.3.3 Boundary conditions

Corresponding to the described scenario, the initial water level in the horizontal part
of the hot-leg was set to y0 = 0.194 m over the entire free surface. A hydrostatic
pressure was assumed for the liquid phase.

The inlet of the CFD model is the vertical cross-section of the horizontal part of the
hot-leg, which is opened to the RPV in the original primary circuit. Following inlet
boundary conditions were chosen for the calculation and were kept constant during
the simulation time:
• height of the steam inlet: 0.194 ≤ y ≤ 0.25 m
• steam superficial velocity: jG = 3.0 m/s (i.e. mass flow rate: 0.213 kg/s)
• water superficial velocity: jL = 0.0 m/s
The inlet turbulence properties have been set using the “Medium intensity and Eddy
viscosity ratio” option of the flow solver, which is equivalent to a turbulence intensity
of 5% in both phases.

A pressure controlled outlet boundary condition was applied to the vertical cross-
section of the steam generator inlet chamber, which corresponds to the connection to
the separator in the experiment. The inner surface of the channel walls was defined
as hydraulically smooth with a non-slip boundary condition applied to both gaseous
and liquid phases.

39
5.3.4 Fluid properties

The gaseous phase is steam and the liquid phase is water, both saturated at 263.9°C
and at a reference pressure of 50 bar. The fluid densities and viscosities are taken
from Table 1 (values are highlighted for 50 bar). The phases have been treated as
isothermal and incompressible. Buoyancy effects between the two phases are taken
into account by the directed gravity term.

Table 1: Water and steam properties

pressure saturation density kinematic viscosity
temperature [kg/m³] [m²/s]
[bar] [°C] water steam water steam
-7
7.5 167.76 899.7 3.91 1.80·10 3.73·10-6
10.0 179.89 887.1 5.15 1.69·10-7 2.92·10-6
30.0 233.86 821.9 15.00 1.39·10-7 1.13·10-6
50.0 263.94 777.4 25.35 1.29·10-7 7.11·10-7

5.3.5 Discretization schemes, time step and convergence

A high resolution discretization scheme was used. For time integration, the fully
implicit second order backward Euler method was applied with a constant time step
of Δt = 0.001 s and a maximum of 15 coefficient loops. A convergence in terms of the
RMS values of the residuals to be less then 10-4 could be assured most of the time.

5.4 Results

The results of the transient CFD-calculation are shown in image sequences of the
phase distribution (Fig. 5.6) and of the steam velocity field (Fig. 5.7). After about 0.2 s
of simulation time, the water level begins to incline upwards in direction of the steam
generator. This is the result of the interfacial momentum exchange between the
steam and the stagnant water. At the water surface, small waves are generated
randomly in the horizontal part of the hot-leg and propagate in direction of the bend.

The velocity fields show that the waves crest constrict the steam cross-section and
causes an acceleration of the steam velocity, especially in the bended region of the
hot-leg model (Fig. 5.7). Moreover, swirls are formed downstream of the biggest
waves (e.g. at 0.27 s).

From 1,52 s, slugs are generated and transport liquid over the steam generator
bottom to the outlet. The slugs cause a water loss in the horizontal part of the
channel and the water level is decreasing. As a consequence, the cross-section for
the steam is widening and the slugging finishes. From about t = 4,5 s, the flow regime
in the hot-leg is again wavy flow.

40
 t = 0.27 s t = 0.52 s

t = 0.77 s t = 1.02 s

t = 1.27 s t = 1.52 s
Fig. 5.6: Steam volume fraction in the hot-leg model over the time, calculated
with ANSYS CFX

41
 t = 0.27 s t = 0.52 s

t = 0.77 s t = 1.02 s

t = 1.27 s t = 1.52 s
Fig. 5.7: Steam velocity distribution in the bended region of the hot-leg model
over the time, calculated with ANSYS CFX (water: black labelled)

42
6. Summary and conclusions
For the investigation of an air/water slug flow, two horizontal channels with
rectangular cross-section were built at Forschungszentrum Dresden-Rossendorf
(FZD). In the first channel with a separator at the inlet, optical measurements were
performed with a high-speed video camera and were complemented by simultaneous
dynamic pressure measurements. By an interface capture method, the water level
history can be extracted from the image sequences. Furthermore, the velocity field of
a slug was measured using Particle Image Velocimetry (PIV). It reveals the inner flow
circulation of the slug, around the point there the slug is rolling over. A CFD
simulation of stratified co-current flow was performed using the code ANSYS CFX
(version 5.7) applying the two-fluid model with the free surface option. No explicit
interface tracking methods (e.g. VOF or Level Sets methods) were used. Slug flow
was successfully simulated with a transient simulation, using a time-dependent inlet
water level from the experimental data.

In the HAWAC test facility, a special inlet device provides well defined as well as
variable boundary conditions, which allow very good CFD-code validation
possibilities. A flow pattern map was arranged and shows the potential of the facility.
A picture sequence recorded during slug flow was compared with the equivalent CFD
simulation made with the code ANSYS CFX. The two-fluid model was applied with a
special free surface treatment. Due to an interfacial momentum transfer, it was
possible to generate slugs based on instabilities. The behaviour of slug generation
and propagation at the experimental setup was qualitatively reproduced, while
deviations require a continuation of the work. The creation of small instabilities due to
pressure surge or an increase of interfacial momentum should be analysed in the
future. Furthermore, pressure and velocity measurements should be performed in the
HAWAC facility to allow quantitative comparisons, also at other superficial velocities.

CFD pre-test calculations were done to show the possibility of slug flow generation in
a real geometry and at relevant parameters for nuclear reactor safety. The simulation
was performed on a flat model representing the hot-leg of the German Konvoi-reactor
at the scale of 1:3. The fluids were water and steam at a pressure of 50 bar and the
corresponding saturation temperature of 263.9°C. The results of the CFD-calculation
show first an upwards inclination of the stagnant water in direction of the steam
generator, driven by the momentum exchange from the injected steam. Further,
waves are generated randomly in the horizontal part of the hot-leg and grow to slugs
in the region of the bend.

43
7. References
ANSYS CFX (2006). User Manual. ANSYS Inc.

Anglart, H., and Podowski, M. Z. (2002). Fluid mechanics of Taylor bubbles and slug
flows in vertical channels. Nuclear science and engineering, 140, 165-171.

Baker, O. (1954). Simultaneous flow of oil and gas. Oil & Gas Journal, 53, 184-195.

Cook, M., and Behnia, M. (2001). Bubble motion during inclined intermittent flow. Int.
J. of Heat and Fluid Flow, 22, 543-551.

Frank, T. (2003). Numerical simulations of multiphase flows using CFX-5. CFX Users
conference, Garmisch-Partenkirchen, Germany.

Hope, C. B. (1990). The development of a water soluble photocromic dye tracing

technique and its application to two-phase flow. PhD thesis, Imperial College of
Science, Technology and Medicine, London.

Ishii, M. (1990). Two-fluid model for two-phase flow. Multiphase Science and
Technology, 5, Edited by G. F. Hewitt, J. M. Delhaye, N. Zuber, Hemisphere
Publishing Corporation, 1-64.

Issa, R. I., and Tang, I. F. (1990). Modelling of vertical gas-liquid slug flow in pipes.
Proc. Symp. On Advances in Gas-Liquid flows, Dallas.

Lun, I., Calay, R. K., and Holdo, A. E. (1996). Modelling two-phase flows using CFD.
Applied Energy, 53, 299-314.

Mandhane, J. M., Gregory, G. A., and Aziz, K. (1974). A flow pattern map for gas-
liquid flow in horizontal pipes: predictive models. Int. J. Multiphase Flow, 1, 537-
553.

Manfield, P. D. (2000). Experimental, computational and analytical studies of slug

flow. PhD thesis, Imperial College of Science, Technology and Medicine,
London.

Mao, Z., and Dukler, A. E. (1991). The motion of Taylor bubbles in vertical tubes. II.
Experimental data and simulations for laminar and turbulent flow. Chem. Eng.
Sci., 46, 2055-2075.

Menter, F. R. (2002). CFD best practice guidelines for CFD code validation for
reactor safety applications. EC project ECORA, report EVOL-ECORA-D01.

Menter, F. R. (1993). Zonal two-equation k-ω turbulence models for aerodynamic

flows. AIAA, 93-2906.

Moe, R. (1992). Transient simulation of 2-3D stratified and intermittent two-phase

flows. PhD thesis, University of Oslo.

Odozi, U. A. (2000). Three-phase gas/liquid/liquid/ slug flow. PhD thesis, Imperial

College of Science, Technology and Medicine, London.

44
Pan, L. (1996). High pressure three phase (gas/liquid/liquid) flow. PhD thesis,
Imperial College of Science, Technology and Medicine, London.

Soleimani, A., and Hanratty, T. J. (2003). Critical liquid flows for the transition from
the pseudo-slug and stratified patterns to slug flow. Int. J. of Multiphase Flow,
29, 51-67.

Sonnenburg, H. G., Palazov, V., and Dräger, P. (1993)., UPTF-TRAM

Versuchsgruppe A, Analyse ausgewählter Versuche. UPTF Fachtagung IV,
25/03/1993, GRS, Mannheim.

Sühnel, T. (2003). Aufbau und Inbetriebnahme eines horizontalen Luft-Wasser-

Strömungskanals. Diplomarbeit, Institut für Sicherheitsforschung, FZ
Rossendorf, Dresden.

Ubbink, O. (1997). Numerical prediction of two fluid systems with sharp interfaces.
PhD thesis, Imperial College of Science, Technology and Medicine, London.

Ujang, P. M. (2003). A three-dimensional study of Taylor bubble turning in two-phase

down flow. Second MIT conference on computational fluid and solids
mechanics, Elsevier Science Ltd., Ed.: K.J. Bathe, 1176-1180.

Wilcox, D. C. (1994). Turbulence modelling for CFD. La Cañada, California: DCW

Industries Inc.

Woods, B. D., and Hanratty, T. J. (1999). Influence of Froude number on physical

processes determining frequency of slugging in horizontal gas-liquid flows. Int.
J. of Multiphase Flow, 25, 1195-1223.

45
8. Indexes
8.1 Nomenclature and abbreviations

Notations

Notation Unit Denomination

A m2 projected area of the body in the direction of flow
A m-1 interfacial area density
B - model parameter (damping of turbulent diffusion)
CD - dimensionless drag coefficient
d m interfacial length scale
Dh m hydraulic diameter
Fd N drag force
g m/s² gravity vector
h m water level
H m channel height
J m/s superficial velocity
L m channel length
Pkb kg/s² buoyancy production term (full buoyancy model)
Pr - Prandtl-number
t s time
U m/s velocity
x m x coordinate (channel width)
y m y coordinate (channel height)
z m z coordinate (channel length)

Greek notations

Notation Unit Denomination

Δn - typical grid cell size across the interface
ρ kg/m³ fluid density
μ kg.m-1.s-1 viscosity

Indexes

Notation Denomination
f index for test facility
G index for gaseous phase
L index for liquid phase
m index for CFD-model
t index for turbulent

46
Abbreviations

Notation Denomination
CCFL Counter-Current Flow Limitation
CFD Computational Fluid Dynamics
ECC Emergency Core Cooling
FZD Forschungszentrum Dresden-Rossendorf
HAWAC Horizontal Air/Water Channel
LES Large Eddy Simulation
LOCA Loss Of Coolant Accident
PIV Particle Image Velocimetry
PWR Pressurised Water Reactor
RPV Reactor Pressure Vessel
SST Shear Stress Transport
TOPFLOW Transient Two Phase Flow Test Facility
VOF Volume Of Fluid

8.2 Figures

Fig. 3.1: Schematic view of the horizontal channel with inlet separator and
modelled part (red) .............................................................................. 16
Fig. 3.2: Example of an image taken from the high-speed camera during
counter-current stratified wavy flow (superficial velocities: JL =
0.14 m/s; JG = -2.2 m/s)....................................................................... 17
Fig. 3.3: Detected interface line for the image given in Fig. 3.2 with
subtracted background ........................................................................ 17
Fig. 3.4: Slug flow - studied cross-sections (blue and red lines) and
characteristic interface step (yellow box)............................................. 18
Fig. 3.5: Time-dependent water level during slug flow (measuring
frequency: 100 Hz) .............................................................................. 19
Fig. 3.6: CFD model of the duct......................................................................... 21
Fig. 3.7: Initial conditions of the mass fractions and the hydrostatic pressure
in the channel ...................................................................................... 22
Fig. 3.8: Measured and calculated image sequences at slug flow conditions ... 23
Fig. 3.9: Propagation of a slug - comparison between measurement and
calculation ........................................................................................... 25
Fig. 3.10: Typical pressure measurement by slug flow and zoom in on the
first pressure increase ......................................................................... 26
Fig. 3.11: Calculated pressure field and interface at t = 0.95 s............................ 26

47
 Fig. 3.12: Transient pressure at the left and right sensor positions for
experiment and ANSYS CFX calculation............................................. 27
Fig. 3.13: Velocity field of the secondary flow inside of a slug (the vector
colour shows the absolute velocity; the vector length and direction
show the relative velocity after subtracting the propagation velocity
of the slug)........................................................................................... 28
Fig. 4.1: Schematic view of the Horizontal Air/Water Channel (HAWAC).......... 29
Fig. 4.2: The HAWAC inlet device ..................................................................... 29
Fig. 4.3: Hydraulic jump in the test-section (jL = 0.37 m/s; jG = 0.10 m/s,
inlet blade = 26 mm)............................................................................ 30
Fig. 4.4: Flow pattern map for the rectangular channel (inlet blade in
horizontal position) .............................................................................. 31
Fig. 4.5: Model and initial conditions of the mass fractions ............................... 32
Fig. 4.6: Calculated picture sequence at JL = 1.0 m/s and JG = 5.0 m/s
(depicted part of the channel: 1.4 to 4 m after the inlet) ...................... 33
Fig. 4.7: Measured picture sequence at JL = 1.0 m/s and JG = 5.0 m/s with
∆t = 50 ms (depicted part of the channel: 0 to 3.2 m after the inlet) .... 34
Fig. 4.8. Details of the slug generation calculated with ANSYS CFX at
different simulation time....................................................................... 35
Fig. 5.1: Schematic view of the flow in the hot-leg and steam generator
during reflux condenser mode (GRS, 1993) ........................................ 36
Fig. 5.2: Possible flow phenomena in the hot-leg bend during CCFL (GRS,
1993) ................................................................................................... 37
Fig. 5.3: The pressure chamber and the test-section components.................... 37
Fig. 5.4: Scheme of the pressure chamber setup.............................................. 38
Fig. 5.5: CFD model of the hot-leg test-section ................................................. 39
Fig. 5.6: Steam volume fraction in the hot-leg model over the time,
calculated with ANSYS CFX ................................................................ 41
Fig. 5.7: Steam velocity distribution in the bended region of the hot-leg
model over the time, calculated with ANSYS CFX (water: black
labelled)............................................................................................... 42

8.3 Tables

Table 1: Water and steam properties ................................................................ 40

48

View publication stats