A Combined UFM and Binary Vapour Mixture Analyser For The Atlas Silicon Tracker PDF

CERN EDMS NO. REV.
VALIDITY
CH1211 Geneva 23 1180058 3.0 Released

Switzerland REFERENCE
2012/010
Date: 2012-04-25
PROJECT REPORT
A COMBINED ULTRASONIC FLOW METER AND

BINARY VAPOUR MIXTURE ANALYSER FOR
THE ATLAS SILICON TRACKER
CFD-2011-04-SONAR
DOCUMENT PREPARED BY: DOCUMENT CHECKED BY: DOCUMENT APPROVED BY:
Gennaro Bozza EN/CV/DC Greg Hallewell /CPPM Greg Hallewell /CPPM

Michele Battistin EN/CV/DC Michele Battistin EN/CV/DC
Enrico Da Riva EN/CV/DC
REFERENCE EDMS NO. REV. VALIDITY
2012/010 1180058 3.0 Released
Page 2 of 111
HISTORY OF CHANGES
REV. NO. DATE PAGES DESCRIPTIONS OF THE CHANGES
1.0 2011-12-06 111

1.1 2012-02-10 112 Table of average velocities for the angled flow meter.
2.0 2012-03-16 111 Spelling and style corrections, some paragraphs
removed, turbulence model equations added, angled
flow meter algorithms revised.
3.0 2012-04-25 111 Typo in the conclusions (p.109) corrected.
2012/010 1180058 3.0 Released
Page 3 of 111
TABLE OF CONTENTS
1. Abstract ......................................................................................... 5
2. Introduction ................................................................................... 6
2.1 The flow meter for lower flow rates applications ..................................................... 8
2.2 The flow meter for higher flow rates applications ................................................... 9
3. Principle of operation of the Acoustic Flow Meter ........................ 11

3.1 Overview ........................................................................................................ 11
3.2 Electronics ...................................................................................................... 11
3.3 The Transducer ................................................................................................ 12
3.4 The algorithms for calculating the sound transit times .......................................... 13
3.4.1 Algorithms for the axial geometry .................................................................... 13
3.4.2 Algorithms for the angled geometry ................................................................. 14
4. Procedure for a CFD analysis ....................................................... 16
4.1 The OpenFOAM CFD software ............................................................................ 16
4.2 Pre-processing ................................................................................................. 16
4.2.1 CAD modelling and export ............................................................................... 16
4.2.2 Mesh generation with BlockMesh ...................................................................... 17
4.2.3 Mesh generation with SnappyHexMesh ............................................................. 21
4.2.4 The y+ parameter ........................................................................................... 23
4.2.5 Remaining case setup ..................................................................................... 24
4.3 Processing ....................................................................................................... 25
4.3.1 The equations ................................................................................................ 25
4.3.2 RANS and turbulence models ........................................................................... 26
4.3.3 Models and solvers used for the flow meter ....................................................... 27
4.4 Post-processing ............................................................................................... 28
4.5 Operating conditions for the simulations ............................................................. 28
5. Simulations for the axial flow meter ............................................ 30

5.1 Problem documentation and simplification of the model ........................................ 30
5.2 The CAD modelling of the axial geometry ............................................................ 34
5.3 The mesh generation ........................................................................................ 35
5.4 Results of the simulations ................................................................................. 38
5.4.1 No transducers, D = 133.7 mm ....................................................................... 38
5.4.2 No transducers, D = 211.6 mm ....................................................................... 44
5.4.3 Transducers at 5 Diameters, D = 133.7 mm ...................................................... 49
5.4.4 Transducers at 10 Diameters, D = 133.7 mm .................................................... 52
5.4.6 Transducers at 5 Diameters, D = 211.6 mm ...................................................... 55
5.4.9 Comparison of the velocity results .................................................................... 59
5.5 Evaluation of the pressure drops ........................................................................ 60
5.6 Average velocities ............................................................................................ 61
6. Simulations for the angled flow meter ......................................... 63

6.1 Problem documentation and simplification of the model ........................................ 63
6.2 The CAD modelling of the angled geometry ......................................................... 66
6.3 The mesh generation ........................................................................................ 67
2012/010 1180058 3.0 Released
Page 4 of 111
6.4 Results of the simulations ................................................................................. 69

6.4.1 Transducers at the maximum distance, α = 15°, D = 133.7 mm .......................... 69
6.4.4 Transducers at the mid distance, α = 15°, D = 133.7 mm ................................... 76
6.4.7 Transducers at the minimum distance, α = 15°, D = 133.7 mm ........................... 81
6.4.19 Comparison of the velocity results .................................................................... 99
6.4.20 Comparison of the pressure results ................................................................ 102
6.4.21 Half flow rate simulations .............................................................................. 105
6.5 Average velocities .......................................................................................... 106
7. Conclusions ................................................................................ 109

8. Bibliography ............................................................................... 110
9. Index ......................................................................................... 111
2012/010 1180058 3.0 Released
Page 5 of 111
1. Abstract
The fluid dynamic design of a new acoustic flow meter is shown in this report. The
acoustic flow meter is a new addition to the ATLAS silicon tracker cooling system which
is being upgraded soon.
In this instrument, which can also operate with gas blends, ultrasonic sound waves are
produced with special transducers, and sound transit times are continuously measured
in opposite directions in flowing gas for continuous real-time measurement of both the
gas blend mixture ratio and the flow rate.
Two possible implemented geometries of the flow meter were studied through CFD
(Computational Fluid Dynamics) techniques: the axial geometry and the angled
geometry. 18 simulations for the angled geometry and 8 simulations for the axial
geometry were performed, by varying a number of geometrical parameters such as the
tube diameter, the distance of the transducers from the tube elbow (for the axial
geometry), the length and the intersecting angle of the secondary pipe (for the angled
geometry). The optimal configurations, in terms of both measurement reliability and
pressure drop in the device were found, and the results will be shown in the following
chapters.
Figure 1-1: Velocity contours and streamlines in an angled flow meter as a result of a CFD
simulation.
2012/010 1180058 3.0 Released
Page 6 of 111
2. Introduction
The acoustic flow meter is a new addition to the ATLAS silicon tracker (Figure 2-1)
cooling control system which is being upgraded soon. With the upgrade, the present
underground compressor-driven C3F8 circulation plant will be replaced by a
thermosiphon [1] (Figure 2-2) equipped with a surface-mounted condenser, allowing
the approximately 90m depth of the ATLAS experimental cavern to be exploited to
hydrostatically generate the required fluorocarbon liquid delivery pressure.
The upgrade may require a change from C3F8 (octafluoro-propane) to a blend

containing 10-30% of C2F6 (hexafluoro-ethane) to reduce the evaporation temperature
and better protect the silicon from cumulative radiation damage with increasing LHC
luminosity.
Figure 2-1: View of the ATLAS inner tracker. Three different layers of the inner tracker can be
noticed in this figure: the Pixel detectors, the Semi-Conductor Tracker SCT and the Transition
Radiation Tracker TRT.
Indeed, the temperature of the silicon substrates of the ATLAS Semi-Conductor Tracker
(SCT) and pixel detectors must be maintained at -7°C or lower. This requires, at full
power dissipation, a coolant evaporation temperature of approximately -25°C. With
pure octafluoro-butane C3F8, due to excessive pressure drops in inaccessible regions of
the exhaust vapour return system, the minimum achievable evaporation temperature is
approximately -15°C. The addition of the more volatile hexafluoro-ethane C2F6 to make
a binary blend, would allow a lower evaporation temperature, (Table 2-1) the required
-25°C, at the same saturation pressure. The cooling system has to remove
approximately 60 kW of heat, and the mass flow rate is approximately 1.2 kg/s.
2012/010 1180058 3.0 Released
Page 7 of 111
Table 2-1 : Evaporation pressure at -25°C evaporation temperature in C3F8 and several C3F8/C2F6
blends.
Fluorocarbon Average (Minimum) Sound velocity (ms-1)
Coolant Evaporation Pressure in superheated vapour
for evaporation at -25°C (20°C, 1 barabs)
C3F8 1.7 (1.7) barabs 114,95
90%C3F8/10%C2F6 2.3 (1.8) barabs 116,78
80%C3F8/20%C2F6 2.7 (2.1) barabs 118,68
70%C3F8/30%C2F6 3.2 (2.3) barabs 120,65
Central to this upgrade is a new ultrasonic instrument in which sound transit times are
continuously measured in opposite directions in flowing gas for continuous real-time
measurement of the C3F8/C2F6 mixture ratio and the flow rate.
The instrument, illustrated in different possible geometries in the following chapters, is

composed of a pair of ultrasonic transducers mounted within a gas enclosure intended
for operation at elevated atmospheric or sub-atmospheric pressures.
The flow meter will be installed in the vapour return tubing (red part of Figure 2-2), far
from the on-detector evaporative cooling channels, where the vapour mixture is in the
superheated state, allowing sound velocity to be predicted according to single-phase
equations of state.
Figure 2-2: Basic scheme of the Two-Phase Full Scale Thermosiphon with the Chiller and Brine
circuits.
2012/010 1180058 3.0 Released
Page 8 of 111
2.1 The flow meter for lower flow rates applications

For lower flow rate (0.016 kg/s) applications (which means applications different from
the ATLAS inner tracker cooling), acoustic flow meters have already been constructed:
the geometry used was the “pinched” axial flow meter, Figure 2-3, and no CFD studies
were done for it. In the pinched axial mechanical envelope the transducers are centred
and mounted flush with the inboard ends of the wide bore sections of the tube,
between which are welded a pair of diameter reducing cones and a ‘pinched’ region (of
inner diameter 44.3 mm, compared with the transducer foil diameter of 44 mm),
through which all the vapour is channelled. For this configuration it is important that
the pinched tube stretch have approximately the same diameter of the transducers, so
that the sound path covers the whole tube diameter. In this way, the top measurement
reliability can be achieved. Unfortunately, a pinched axial flow meter is not realizable
for the ATLAS silicon tracker cooling system since the flow rate is higher (1.2 kg/s), the
only available transducer diameter is 44 mm, and the diameter of 44.3 mm of the
pinched tube stretch would therefore be too small, causing an excessive velocity of the
flow.
Figure 2-3: Views of the ‘pinched axial’ mechanical envelope, showing an ultrasonic transducer,
its mounting and axial flow-deflecting cone, together with tubes for pressure sensing and the
evacuation and the injection of calibration gas.
2012/010 1180058 3.0 Released
Page 9 of 111
2.2 The flow meter for higher flow rates applications

Two possible geometries of the acoustic flow meter for the ATLAS silicon tracker
cooling system, where the flow rates are higher, are the (open) axial geometry and the
angled geometry. In the case of the Axial Flow Meter, Figure 2-4, the transducers are
placed as twin cylinders between two consecutive pipe elbows, facing each other.
Figure 2-4: Layout of an in-line flow meter.
In the case of the Angled Flow Meter, Figure 2-5, the set-up consists of a transversal
secondary pipe, of internal diameter d, intersecting at a certain angle α with the main
tube of internal diameter Dmain. The secondary pipe is closed at both ends by a pair of
transducers, wafer-like devices responsible for creating the sound wave. In order for
this sound pressure wave to propagate optimally, static vapour is needed in the
transversal secondary pipe, or, at least, the influence of axial direction velocity
components should be minimized.
Figure 2-5: Configuration with transducer inter-distance = 3*Dmain/sinα.

2012/010 1180058 3.0 Released
Page 10 of 111
The aim of this work is finding the best geometry in terms of both measurement
reliability and pressure drop in the device. By using CFD techniques, simulations for
different transducers distances, different angles of the secondary pipe, different main
pipe diameters and different distances of the transducers from the elbow (for the axial
geometry) were done.
In chapter 3 the principle of operation of the Flow Meter will be explained. An overview
of the major components, such as the transducers and the electronics will be shown.
Chapter 4 is dedicated to Computational Fluid Dynamics. The models used, the

equations solved, the meshing tools, the settings and the software employed will be
covered with.
In chapters 0 and 6 the results for all the configurations will be explained in detail
through velocity and pressure contours and a range of plots and tables.
2012/010 1180058 3.0 Released
Page 11 of 111
3. Principle of operation of the Acoustic Flow Meter

3.1 Overview
The sound velocity in a binary gas mixture at known temperature and pressure is a
unique function of the molar concentration of the two components of differing
molecular weight. Ultrasonic gas analysis was first used in particle physics for the
analysis of the N2/C5F12 (dodecafluoro-pentane: mw = 288) Cherenkov gas radiator of
the SLD Cherenkov Ring Imaging Detector 1 , as a simpler alternative to on-line
refractive index monitoring. Since then it has been adopted in all the major ring-
imaging Cherenkov detectors, including DELPHI 2 , COMPASS 3 and LHCb. The molar
concentration of the two component vapours is determined from comparison of sound
velocity measurements with velocity-composition look-up table data gathered from
prior measurements in calibration mixtures or from theoretical data derived according
to an appropriate equation of state. Due to on-going improvements over a number of
years, the NIST-RefProp refrigerant-oriented extended Benedict–Webb–Rubin (BWR)
equation of state is presently the optimum for calculating sound velocity in
fluorocarbon mixtures, including those containing C2F6 and C3F8. For mixtures of
fluorocarbons with other gases, (N2/C3F8 in our current interest) optimum sound
velocity predictions are made using PC-SAFT (Perturbed Chain – Statistical Associating
Fluid Theory) equation of state. Predictions made with these equations of state are
compared with measurements in calibration mixtures, for the setup of the instrument.
3.2 Electronics
In custom microcontroller-based electronics, the transmitting transducer is excited with
a short burst of high voltage 50 kHz square wave pulses generated in a driver circuit
from a TTL4 precursor pulse train, itself generated in the microcontroller. The receiving
transducer is connected to a DC biasing circuit followed by an amplifier and
comparator. A fast (40 MHz) transit time clock, generated in the same microcontroller,
is started in synchronism with the rising edge of the first transmitted 50 kHz sound
pulse. The first received sound pulse crossing the comparator threshold level stops this
clock (Figure 3-1). The time between the transmitted and first received sound pulses is
1
The SLC Large Detector (SLD) experiment at the SLAC Linear Collider (SLC) was equipped with Cherenkov
Ring Imaging Detectors (CRIDs) in both barrel and end cap regions to provide particle identification over a
broad momentum range. The SLD no longer exists.
2
DELPHI was a Particle Physics experiment at the CERN. It studied the products of electron-positron
collisions at the LEP circular accelerator. The main parts are still kept in the original state in the cavern
where it was operating and can be visited.
3
COMPASS is a high-energy physics experiment at the Super Proton Synchrotron (SPS) at CERN in Geneva,
Switzerland. The purpose of this experiment is the study of hadron structure and hadron spectroscopy with
high intensity muon and hadron beams.
4
Transistor–transistor logic (TTL) is a class of digital circuits built from bipolar junction transistors (BJT) and
resistors. It is called transistor–transistor logic because both the logic gating function (e.g., AND) and the
amplifying function are performed by transistors (contrast with RTL and DTL). TTL is notable for being a
widespread integrated circuit (IC) family used in many applications such as computers, industrial controls,
test equipment and instrumentation, consumer electronics, synthesizers, etc.
2012/010 1180058 3.0 Released
Page 12 of 111
measured by the microcontroller, which also handles communication via a USB/RS232

interface.
Figure 3-1: Principle of measurement of transit time between the first transmitted sound pulse
and the first over-threshold detected pulse
3.3 The Transducer

The 50 kHz capacitive ultrasonic transducer used in the Flow Meter, Figure 3-2, was
originally developed during the 1980s for the Polaroid autofocus instant cameras. Now,
it is marketed by SensComp as ultrasonic transducer, model 600 series, principally for
use in robotic applications. The transducer is 44 mm in diameter and comprises a gold
coated Mylar foil stretched over a spiral backing plate with a machined spiral groove.
The backing plate is biased and excited at a high voltage (80-360V) with the case and
foil grounded. Although the transducer was originally developed for operation in air at
ambient temperature and atmospheric pressure, the spiral groove allows gas to access
both sides of the diaphragm, allowing operation over a wide pressure range.
Figure 3-2: The SensComp 50 kHz capacitive ultrasonic transducer originally developed during
the 1980s for the Polaroid autofocus instant camera.
2012/010 1180058 3.0 Released
Page 13 of 111
3.4 The algorithms for calculating the sound transit times

3.4.1 Algorithms for the axial geometry
Rolling average transit times in both directions are used to compute the gas flow rate
while the averages of the sound velocities in both directions are used together with
temperature and pressure, to compute the binary gas composition. Transit times,
5
temperature and pressure data continually stream from a FIFO memory to a
supervisory computer.
Gas mixture is continuously analysed using SCADA 6 software, by comparing the

average sound velocity in both directions with stored velocity vs. concentration look-up
tables. These tables may be created from prior measurements in calibration mixtures
or from theoretical thermodynamic calculations. Flow rates are calculated from the
difference in transit time in the two directions. In future versions these calculations
may be made in an on-board microcontroller.
In the axial geometry the vapour flow rate is calculated from the sound transit times
measured parallel, tdown, and anti-parallel, tup, to the flow direction, according to the
following algorithm:
(3-1)
where v is the linear flow velocity (m/s), c the speed of sound in the gas and L the
distance between transducers.
The gas volume flow V (m3/s) can therefore be inferred from the two transit times by:
(3-2)
where A is the internal cross sectional area of the axial flow tube between the two
ultrasonic transducers (m2).
The sound velocity c can also be inferred from the two transit times via:
(3-3)
It can be seen from equations above that the knowledge of the temperature of the gas
is not necessary for flowmetry.
5
FIFO is an acronym for First In, First Out, an abstraction related to ways of organization and manipulation
of data relative to time and prioritization. This expression describes the principle of a queue processing
technique or servicing conflicting demands by ordering process by first-come, first-served (FCFS)
behaviour: what comes in first is handled first, what comes in next waits until the first is finished,
analogous to the behaviour of persons standing in line, where the persons leave the queue in the order
they arrive, or waiting one's turn at a traffic control signal.
6
SCADA (supervisory control and data acquisition) generally refers to industrial control systems (ICS):
computer systems that monitor and control industrial, infrastructure, or facility-based processes.
2012/010 1180058 3.0 Released
Page 14 of 111
3.4.2 Algorithms for the angled geometry
The “angled crossing” geometry is the second possible configuration modelled for
measurements at high fluorocarbon flow rate. In its simplest implementation, an
angled ultrasonic flow meter consists of a pair of transducers of diameter d separated
by a distance L, and aligned on a sound path intersecting the main tube, of internal
diameter DMain, at an angle α. The vapour flow rate is calculated from the opposed
sound transit times, tdown, and tup, according to the following algorithm:
, 3-4
where v is the linear flow velocity (m/s), c the speed of sound in the gas and the
distance L between transducers is entirely contained within the main flow tube.
The gas volume flow V (m3/s) is inferred from the two transit times by:
3-5
while the sound velocity c can be inferred from the two transit times via:
3-6
If the transducers are attached to the inner wall of the main tube they will impinge on
the flow, creating eddies that will affect the measured gas flow velocity. A preferable
geometry has the transducers either minimally backed-off a distance L’ (counting both
sides) to position their inner edges flush with the internal surface of the main tube.
This minimal non-impinging transducer spacing can be defined as:
, where
3-7
Alternatively the transducers may be withdrawn a longer distance L’ with respect to the
internal surface of the main tube. The latter configuration allows the changing of a
transducer without interruption of the main gas flow: the transducers can be backed off
outboard of quarter-turn ball valves, which when open allow the sound path to traverse
the main tube.
In configurations where the transducer spacing includes an element L’ containing

quasi-static vapour in communication with the flow in the main tube the formalism for
calculating the flow velocity is modified, since only DMain/sinis used in the calculation
of gas flow velocity via equations above.
; 3-8
From either equations above the velocity v can be expressed as

2012/010 1180058 3.0 Released
Page 15 of 111
( ( )
) 3-9
or as
( ( )
) 3-10
The sound velocity, c, is the physical root derived from the equations above in terms of
measurable L, DMain, α, tup and tdown:
( )( ) ( )( )
√( ) ( ) 3-11
2012/010 1180058 3.0 Released
Page 16 of 111
4. Procedure for a CFD analysis

4.1 The OpenFOAM CFD software
The fluid dynamical design of the flow meter was done with the open source CFD
software OpenFOAM. OpenFOAM is written in C++ and uses an object oriented
approach which makes it easy to extend. The package includes modules for a wide
range of applications.
OpenFOAM uses the Finite Volume discretization method on unstructured meshes. It

provides many capabilities, including free-surface and multi-phase flow modelling,
Lagrangian spray model and automatic mesh motion.
4.2 Pre-processing
4.2.1 CAD modelling and export
The CAD modelling, which is the first step of a CFD simulation, was done for the flow
meter with the 3D CAD Software CATIA. After completing it, the CAD geometry is
exported into a format readable by OpenFOAM, the STL (Stereolithography) format.
The next step is applying a mesh on the STL geometry. For the flow meter the mesh
was generated by using two OpenFOAM meshing tools: BlockMesh and
SnappyHexMesh.
BlockMesh is a simple meshing tool which allows to generate simple meshed

geometries starting from vertices and blocks. The mesh is hexahedral.
SnappyHexMesh doesn’t generate geometries, but allows to generate a mesh on

geometries imported from CAD software. The total number of cells varies from 1.5
million to 10 million, depending on the simulation.
2012/010 1180058 3.0 Released
Page 17 of 111
4.2.2 Mesh generation with BlockMesh

The BlockMesh tool consists in typing a code, by using OpenFOAM syntax, in order to
generate geometry and mesh. For a more detailed description, refer to [2].
Figure 4-1: BlockMesh dictionary: in this zoom, the syntax for defining the vertices is shown.
In Figure 4-1 a zoom of the first part of a BlockMesh dictionary is shown. The syntax is
very simple, and for defining a vertex is sufficient to type the three coordinates x, y
and z. After defining the vertices it is necessary to define the blocks and the mesh
wanted (Figure 4-2). Every block consists of 8 vertices, which are the first 8 numbers
of a row (each row corresponds to a block, Figure 4-2). The next 3 numbers of a block
correspond to the number of cells along the three directions. The last three numbers,
finally, refer to the grading along the three directions. The grading is defined as the
ratio between the longest and the shortest cell of a block along one direction:
4-1
2012/010 1180058 3.0 Released
Page 18 of 111
Figure 4-2: BlockMesh dictionary: in this zoom, the syntax for defining blocks and arcs is shown.
In order to generate a tube, 5 blocks are sufficient: one central and 4 lateral. The 4
lateral blocks present one curved face, each one to form a quarter of tube. The curved
face is obtained through the edges entry in the BlockMesh dictionary. So, the first row
in Figure 4-2 can be read in the following way: an arc from vertex 1 to vertex 2 passes
through a third point of coordinates (25.58, -61.76 0). All the blocks, correctly defined,
form the whole geometry. Finally, after block and mesh generation, the last step is
defining the boundaries. Figure 4-3 shows the entries in the dictionary file. The surface
of the block Mesh geometry is made of faces of blocks, and every face must be defined
as wall, inlet, outlet, symmetry plane and so on.
2012/010 1180058 3.0 Released
Page 19 of 111
Figure 4-3: BlockMesh dictionary: in this zoom, the syntax for defining the boundaries is shown.
4.2.2.1 BlockMeshDict with Excel

In order to correctly calculate the entries for the BlockMeshDict file, an Excel spread
sheet was used. Figure 4-4 shows the table for the vertices and arcs entries. By simply
typing the values in the orange squares, the whole pipe is calculated. The formulas are
slightly different for axial and angled geometry.
After the geometry generation, also the blocks entry is implemented in Excel. For the
blocks entry, it was necessary to find an algorithm to calculate the number of cells in
every block as a function of the direction, the grading (as defined in paragraph 4.2.2)
and the adjacent blocks, so that the overall mesh was uniform. In other words, the
meshes of every block had to fit together like in a jigsaw. This was made possible by
the formula below:
4-2
( )
2012/010 1180058 3.0 Released
Page 20 of 111
Where
● n is the number of cells along a direction (value to type in the BlockMeshDict)
● g is the grading wanted
● L is the length of the whole block along the direction
● lb is the length of the longest cell along the direction
Figure 4-4: Vertices entry and arcs entry on an Excel spread sheet for BlockMesh. In yellow
squares the values calculated, in orange squares the values inserted.
Figure 4-5: Calculation of the total number of cells, calculation of the number of cells along a
direction for a block and for a certain grading, and finally calculation of the blocks entry. The
formula was used for this calculation.
2012/010 1180058 3.0 Released
Page 21 of 111
4.2.3 Mesh generation with SnappyHexMesh

If the geometry is very simple, for example a simple tube, BlockMesh is sufficient. If
the geometry is more complicated, like for the flow meter different configurations, a
further step is necessary in the mesh generation, by using SnappyHexMesh. The
SnappyHexMesh utility generates 3-dimensional meshes containing hexahedra (hex)
and split-hexahedra (split-hex) automatically from triangulated surface geometries in
Stereolithography (STL) format. The mesh approximately conforms to the surface by
iteratively refining a starting mesh and morphing the resulting split-hex mesh to the
surface. An optional phase will shrink back the resulting mesh and insert cell layers.
The specification of mesh refinement level is very flexible and the surface handling is
robust with a pre-specified final mesh quality.
In order to run SnappyHexMesh, the user requires the following:
● Surface data files in STL format, either binary or ASCII, located in a

constant/triSurface sub-directory of the case directory.
● A background hex mesh typically generated using BlockMesh, which defines the
extent of the computational domain and a base level mesh density.
● A SnappyHexMeshDict dictionary, with appropriate entries, located in the system
sub-directory of the case.
The SnappyHexMeshDict dictionary includes switches at the top level that control the
various stages of the meshing process and individual sub-directories for each process.
The three stages of the meshing process are:
1. Mesh castellation
2. Mesh snap
3. Mesh layer addition (optional)
After running SnappyHexMesh, the user checks the mesh both visually and through the
CheckMesh tool. CheckMesh tool verifies that parameters like maximum skewness and
maximum non-orthogonality can fit a smooth and fast iteration process for the
numerical integration. On the other hand, when one runs a simulation that didn’t pass
the CheckMesh test, the convergence is much more difficult to achieve and the risk of
simulation abortion due to numerical instability is much higher. Unfortunately there is
no formula for evaluating the optimal SnappyHexMeshDict setup for a given geometry.
Therefore, the only one way to achieve it is starting from the default values and
changing them according to the user’s experience and sensitivity. Achieving the best
setup may require a relatively long time, and this is certainly the step that takes longer
in a whole CFD simulation with OpenFOAM. This is one of the major weaknesses of
SnappyHexMesh. In order to help the author to achieve the best setup, the Excel
spread sheet in Figure 4-6 was created. This spread sheet helps from one side to
visualize in a single sheet all the changes done and from the other side to apply, if
necessary, DOE 7 techniques for searching the best solution. For further information
about SnappyHexMesh refer to the user guide [2], available on the OpenFOAM website.
7
DOE means Design Of Experiments. In the design of experiments, the experimenter is often interested in
the effect of some process or intervention (the "treatment") on some objects (the "experimental units").
2012/010 1180058 3.0 Released
Page 22 of 111
Figure 4-6: Excel spread sheet to help achieving the best setup for SnappyHexMesh.
2012/010 1180058 3.0 Released
Page 23 of 111
4.2.4 The y+ parameter

A good mesh is so that:
● It passes the CheckMesh test.
● In case the wall function approach is adopted for turbulence modelling, the y+ on
the walls is included within the range [50,150].
If no wall functions are used, the y+ is generally < 1.
y+ parameter is a non-dimensional wall distance for a wall-bounded flow, and it can be

defined in the following way:
4-3
Where is the friction velocity8 which is generally defined as:
4-4
y is the distance of the cell centroid from the nearest wall and is the local kinematic
+
viscosity of the fluid. y is commonly used in boundary layer theory and in defining the
law of the wall.
y+ was calculated through an Excel spread sheet, as a function of the thickness of the
cell near the wall, as shown in Figure 4-7. is the wall shear stress, which allows the
computation of the friction velocity.
Figure 4-7: calculation of the y+ on an Excel spread sheet. The value on the left is for the
diameter of 133.7 mm, the value on the right is for the diameter of 211.6 mm.
8
Friction velocity, also called shear velocity, is a form by which a shear stress may be re-written in units of
velocity. Shear velocity is used to describe shear-related motion in moving fluids.
2012/010 1180058 3.0 Released
Page 24 of 111
4.2.5 Remaining case setup

Apart from mesh generation and boundaries definition, which have already been
introduced in the previous paragraphs, the dictionary files in Table 4-1 need to be
filled, in the simpleFoam case, in order to complete the setup and add boundary
conditions and fluid properties.
Table 4-1: Dictionary files for a case setup with the simpleFoam solver
file directory role
k 0 k value for the turbulence model
omega 0 ω value for the SST k-ω turbulence model
epsilon 0 ϵ value for the k- ϵ turbulence model
u 0 Velocity value at the boundaries
p 0 Pressure value at the boundaries
nut 0 Turbulent viscosity at the boundaries
RASproperties constant Choice of the turbulence model (paragraph 4.3.2)
transportProperties constant Choice of the fluid kinematic viscosity
controlDict system Choice of the application (paragraph 4.3.3),
number of iterations, time step, sampling probes
decomposeParDict system Choice of the number of processors for solver
and/or SnappyHexMesh
fvSchemes system Choice of numerical integration schemes for each
term of the equations solved
fvSolution system Choice of solvers sub dictionary, non-orthogonal
correctors, under-relaxation factors
sampleDict system Choice of the sampling lines for plotting graphs
In order to quickly evaluate boundary conditions like velocity, nut, omega, epsilon and
k, the Excel spread sheet in Figure 4-8 was built.
2012/010 1180058 3.0 Released
Page 25 of 111
Figure 4-8: Fluid dynamical parameters for the flow meter in an Excel Spread sheet
4.3 Processing
The processing step consist in integrating numerically the equations of mass,
momentum and energy
4.3.1 The equations

The equations solved by the software can be expressed in the following vectorial
compact form, in the hypothesis of steady state and incompressible flow:
Conservation of mass
( ⃗) 4-5
Conservation of momentum (Navier Stokes)
(⃗ ⃗) ⃗ ⃗ ⃗ 4-6
Where ⃗ represents the shear stress term for incompressible flows, and ⃗
represents the body forces acting on the fluid such as the centrifugal force. Gravity is
neglected for this study.
2012/010 1180058 3.0 Released
Page 26 of 111
4.3.2 RANS and turbulence models

If complicated phenomena like turbulence are intended to be studied, the equations
above become very difficult to integrate. Indeed, the turbulent fields are characterized
by fluctuations of velocity. Such fluctuations mix mass, energy, momentum within the
domain, and since they can be very small and their frequency very high, they would
determine an excessively high computational cost for conventional engineering
problems. Therefore, a simplification is needed. A widespread technique consists in
time-averaging the momentum equation (Reynolds Averaged Navier Stokes, RANS), in
order to remove the smallest turbulence scales. Now, in order to compute the Reynolds
Stresses in the RANS, an appropriate turbulence model is needed. The two turbulence
models used in the set of simulations of this project are
 The RNG k-ϵ model
 The SST k-ω model
For further information about these models, refer to [3], [4] and [5].
4.3.2.1 The RNG k-ϵ model

The RNG k-ϵ model is one of the most common turbulence models, although it just
doesn't perform well in cases of large adverse pressure gradients. It is a two equation
model that means it includes two extra transport equations to represent the turbulent
properties of the flow. This allows a two equation model to account for history effects
like convection and diffusion of turbulent energy. The first transported variable is the
turbulent kinetic energy k. The second transported variable in this case is the turbulent
dissipation rate ϵ. The first variable, k, determines the energy in the turbulence,
whereas the second variable, ϵ, determines the scale of the turbulence. The RNG model
was developed to account for the effects of smaller scales of motion. In the standard k-
ϵ model the eddy viscosity is determined from a single turbulence length scale, so the
calculated turbulent diffusion is that which occurs only at the specified scale, whereas
in reality all scales of motion will contribute to the turbulent diffusion. The RNG
approach is a mathematical technique that can be used to derive a turbulence model
similar to the k-epsilon. It results in a modified form of the epsilon equation which
attempts to account for the different scales of motion through changes to the
production term. The turbulent kinetic energy k and its rate of dissipation ϵ are
obtained from the following set of two equations:
( ) ( ) ( )
{ 4-7
( ) ( ) ( ) ( )
Where represents the generation of turbulent kinetic energy due to the mean
velocity gradients, is the generation of turbulent kinetic energy due to buoyancy,
represents the contribution of the fluctuating dilatation in compressible turbulence to
2012/010 1180058 3.0 Released
Page 27 of 111
the overall dissipation rate, and are the inverse effective Prandtl numbers for k
and ϵ, respectively.
4.3.2.2 The SST k-ω model

The SST k-ω model is one of the most commonly used turbulence models. As well as
the RNG k-ϵ model, the SST k-ω model is a two equation model, which means it
includes two extra transport equations to represent the turbulent properties of the flow.
This allows a two equation model to account for history effects like convection and
diffusion of turbulent energy. The first transported variable is turbulent kinetic energy
k. The second transported variable in this case is the specific turbulent dissipation rate
ω. The first variable k, determines the energy in the turbulence, whereas the second
variable ω determines the scale of the turbulence. The SST k-ω turbulence model [4] is
a two-equation eddy-viscosity model which has become very popular. The shear stress
transport (SST) formulation combines the best of k-ϵ and k-ω. The use of a k-ω
formulation in the inner parts of the boundary layer makes the model directly usable all
the way down to the wall through the viscous sub-layer. The SST formulation also
switches to k-ε behaviour in the free-stream and thereby avoids the common k-ω
problem that the model is too sensitive to the inlet free-stream turbulence properties.
The SST k-ω model has the following form:
( ) ( ) ( )
{ 4-8
( ) ( ) ( )
Where represents the generation of turbulent kinetic energy due to the mean
velocity gradients, represents the generation of ω, and represent the effective
diffusivity of k and ω, respectively. and represent the dissipation of k and ω due
to turbulence.
4.3.3 Models and solvers used for the flow meter

Although the fluid in the flow meter is a gas, since the maximum speed of the flow is
approximately u = 22 m/s (see Table 5-2), which corresponds to Mach 0.19 < Mach
0.3, an incompressible fluid approach can be adopted.
Also, the system is treated as steady. Indeed, during the normal operation, the effects
of a transient over the measurements are negligible.
A suitable solver for incompressible steady-state flows is the SIMPLE (Semi-Implicit

Method for Pressure-Linked Equations) solver algorithm. SIMPLE algorithm, which in
OpenFOAM is called SimpleFoam, was used for the whole set of simulations.
SIMPLE algorithm exploits the fact that if a steady-state problem is being solved
iteratively, it is not necessary to fully resolve the linear pressure-velocity coupling, as
the changes between consecutive solutions are no longer small.
Characteristics of the SIMPLE algorithm are:

2012/010 1180058 3.0 Released
Page 28 of 111
● An approximation of the velocity field is obtained by solving the momentum

equation. The pressure gradient term is calculated using the pressure distribution
from the previous iteration or an initial guess.
● The pressure equation is formulated and solved in order to obtain the new pressure
distribution.
● Velocities are corrected and a new set of conservative fluxes is calculated.
4.4 Post-processing
Post-processing is the last step of a CFD simulation and consists in analysing the
results of a simulation. In fact, the raw result consists of the fields of velocity and
pressure in the domain, which means a value of velocity and pressure for each cell
constituting the domain. Such data can be processed to generate velocity and pressure
contours and plots.
The software used in this set of simulations for the post-processing is Paraview for the
contours and for calculating the average velocities and Gnuplot for the plots.
4.5 Operating conditions for the simulations

Operating conditions refer to the point of the full scale thermosiphon cooling system for
ATLAS where the flow meter will be installed. In particular, they correspond to Table
4-2 and point A of Table 4-3 below. These tables are contained, for a more detailed
description, in [1].
Figure 4-9: p-h Diagram of the full scale thermosiphon under operation.
2012/010 1180058 3.0 Released
Page 29 of 111
Table 4-2: Main design parameters.

Parameter Design specification
Flow of C3F8 at full power 1.2 kg/s
Nominal pressure of the C3F8 at the return rack (plant side) 0.5 bar (abs)
Temperature at the return rack 20°C
Liquid pressure at the supply distribution lines (plant side) 15 bar (abs)
Temperature at the supply rack 20°C
Maximum operating pressure PN40
Table 4-3: Nominal operating points.

Operating Pressure Temperature Density Enthalpy Physical
point (bar) (°C) (kg/m3) (kJ/kg) State
A 0.5 20 3.90 310.8 Superheated vapour
B 0.51 -20 4.65 280.6 Superheated vapour
C 0.309 -25 2.85 277.4 Superheated vapour
D 0.309 -60 1699 140.3 Saturated liquid
E 0.4 -65 1717 135.7 Sub-cooled liquid
F 16.1 -62 1712 139.0 Sub-cooled liquid
G 16.1 -51 1672 149.2 Sub-cooled liquid
H 16 -20 1552 179.4 Sub-cooled liquid
I 16 20 1365 222.0 Sub-cooled liquid
I’ 0.5 -51 7.89 222.0 Two–phase 66% vapour
2012/010 1180058 3.0 Released
Page 30 of 111
5. Simulations for the axial flow meter

5.1 Problem documentation and simplification of the model
The axial flow meter is the first of the two geometries studied. In this chapter all the
results about the axial geometry will be presented. The conclusion of this study is that
this configuration would not be capable to provide a reliable measurement of the
velocity, due to the not negligible influence of the transducers over the flow, positioned
at the centre of the tube.
In the axial flow meter the transducers are placed as twin cylinders between two
consecutive pipe elbows, facing each other. Figure 5-1 illustrates the geometry of an
in-line sonar flow meter. The transducers will be modelled as cylinders with deflection
cones (Figure 5-3). The deflection cones must be positioned as illustrated in Figure 5-1,
at a distance of 660 mm from each other, and a distance L from the elbows. Since the
elbows modify the velocity field until a certain development length, it is intended to
check whether the L distance can affect the measurement.
Figure 5-1: Layout of an in-line flow meter.
The optimal L distance is intended to be studied in this set of simulations. Therefore,

three L distances will be modelled:
 L = 5D
 L = 10D
 L = 20D
The influence of the supports was considered for the first simulations made. Then,
since they don’t significantly affect the flow, the supports were removed. The supports
are fin shaped, as illustrated in Figure 5-2.
2012/010 1180058 3.0 Released
Page 31 of 111
Figure 5-2: Transducer with a fin-shaped support
In order to achieve symmetry in the domain and, as a result, speed up the simulations
by only simulating half domain, only the position of the supports with a fin perfectly
vertically positioned will be considered. Also, given that the transducer’s wiring will not
pass inside the fins, but will rather pass close to them, the influence of the wiring will
be neglected. Figure 5-3 shows in detail the dimensions of the transducer (in yellow)
and the deflection cone (in green).
Figure 5-3: Geometry and dimensions of a transducer with a deflection cone applied
Simulations will be performed for two different inner diameters of the main pipe:
 D = 133.7 mm
 D = 211.6 mm
2012/010 1180058 3.0 Released
Page 32 of 111
A smaller diameter would reduce the cost of the pipe and would improve the
measurement precision, since the velocity component parallel to the sound wave would
be higher, and the velocity itself would be more sensitive to flow rate variation. On the
other hand, the resulting higher pressure drop might not meet the design requirements
of the whole system. A greater diameter would therefore allow a considerably lower
pressure drop.
Finally, two simulations without transducers will be run for the two diameters in order
to compute the wake downstream of the elbow.
The total number of simulations is 2 + 2 x 3 = 8.

Table 5-1: List of the eight simulations for the axial flow meter
Simulation No Transducers Main Tube Diameter Transducers distance
[ mm]
1 No 133.7
2 No 211.6
3 Yes 133.7 5D
4 Yes 133.7 10D
5 Yes 133.7 20D
6 Yes 211.6 5D
7 Yes 211.6 10D
8 Yes 211.6 20D
Here’s a list of dimensions:

● Inner diameters of main pipe: D = 133.7 mm and D = 211.6 mm;
● Elbows radius of curvature: R = 190 mm (for D = 133,7 mm) and R = 305 mm
(for D = 211.6 mm) (CERN Stores Catalogue - 41.03.07 - 90 GRADES, 3D,
WELDABLE ELBOWS - 316L stainless steel);
● Transducer inter-distance: 660 mm;
● Distance between elbows: 660 mm plus symmetric:
 L = 5D
 L = 10D
 L = 20D
● Outside diameter of the ultrasonic transducers 9: d = 43 mm;
● Length of the transducers = 4.3 mm
● Base Diameter of the deflecting cone = 48.3 mm
● Height of the deflecting cone = 40 mm
● Cylindrical onward projection for transducer attachment and wire attachment = 15
mm
● Total length of the deflecting cone: 40 mm + 15 mm = 55 mm
● Thickness of the fins: 2 mm
9 Transducer type: Polaroid 600 series instrument grade 50 kHz capacitive foil transducer (now marketed by
SensComp, Inc. 36704 Commerce Rd. Livonia, MI 48150 USA http://www.senscomp.com)
2012/010 1180058 3.0 Released
Page 33 of 111
Further data are listed in Table 5-2:

Table 5-2: Geometrical and fluid dynamical parameters
Diameters
0.1337 0.2116 m
Cross section 0.014 0.035 m2
Flow speed 21.92 8.75 m/s
Speed of sound 116 m/s
Flow speed/Speed of sound 0.19 0.08
Reynolds number 930824 588144
Simulation Length L = 5D 0.7 1.1 m
The following results will be provided:
● The velocity field and the streamlines at the symmetry plane.
● The maximum local component of velocity parallel to the wave.
● The volume average of the velocity (considering the cylindrical volume between 2
sensors, with the same diameter of the transducers)
The following simplifications of the model are made:

● The sensors are modelled as disks with cones applied.
● An entrance length of 10 diameters will be considered.
● The fluid is pure C3F8.
● The flow is treated as steady-state.
● The problem is treated as adiabatic.
● The influence of gravity is negligible.
Half a pipe will be simulated, since the plane passing through the centre of the pipe is a
symmetry plane. Also, only the elbow upstream of the transducers will be simulated,
since the elbow downstream doesn’t influence the flow.
2012/010 1180058 3.0 Released
Page 34 of 111
5.2 The CAD modelling of the axial geometry

Initially, the whole axial flow meter, Figure 5-4, was modelled with the software CATIA.
Afterwards, it was decided to model the tube with BlockMesh, and to only model the
transducers with CATIA, in order to achieve a better mesh.
Figure 5-4: First CAD geometry of the axial flow meter, D = 133.7 mm, L = 5D. The two
transducers, with deflection cone and supporting fins, are clearly visible in the image. The
possible choice to import all this STL geometry in SnappyHexMesh was discarded.
In Figure 5-5 are shown the transducers with supporting fins. This is not the definitive
geometry chosen for the simulations either. Indeed, it was observed that the influence
of the fins could be neglected, also to improve the legibility of the velocity field at the
symmetry plane. As a result, the CAD geometry of the transducers used for the set of
simulations is shown in Figure 5-6.
Figure 5-5: Transducers (in yellow) with deflection cones (in red) and supporting fins (in blue).
The rest of the geometry is directly modelled with BlockMesh.
2012/010 1180058 3.0 Released
Page 35 of 111
Figure 5-6: Transducers (in yellow) with deflection cones (in red) without supporting fins (in
blue). The rest of the geometry is directly modelled with BlockMesh.
5.3 The mesh generation

In order to write the BlockMeshDict dictionary file to generate the mesh, it was
necessary to build an excel file which automatically calculated the coordinates of the
vertices. For example, for the coordinates of the elbow, it was implemented the
following system of equations for transformation of points:
{ 5-1
( )
Where is the plane at the outlet of the elbow, is the plane at the inlet of the elbow,
R is the radius of curvature of the elbow, is an angle which goes from 0 at the inlet
to 90 at the outlet, x, y, z are the three coordinates of a generic point of the elbow, in
a orthogonal Cartesian reference system.
Figure 5-7: Zoom of the elbow of the axial flow meter, with a mesh applied. This mesh, which
doesn’t include the transducers, was obtained with BlockMesh.
2012/010 1180058 3.0 Released
Page 36 of 111
In the zoom in Figure 5-7 six blocks are present, but the mesh doesn’t present any
discontinuity since a correct cells number and size was calculated, as explained in
paragraph 4.2.2. A grading towards the tube surface is also evident. For the axial
geometry a grading of 9 was applied.
Figure 5-8: Mesh at the inlet of the axial flow meter. In this cross section the four blocks which
form the tube are clearly visible. It is also evident the grading from the centre to the tube
surface. The square mesh at the centre of the domain was used to avoid too small cells at the
centre. This allowed the most uniform and effective mesh possible.
1
2
Figure 5-9: Mesh near the transducers. A refinement box (1) was applied in the region of the
sound path, in order to locally increase the computational precision. A surface refinement (2)
was applied on the surface of the transducers, in order to improve the geometry and the
computational precision.
2012/010 1180058 3.0 Released
Page 37 of 111
Figure 5-10: Mesh of the tube with transducers, obtained with SnappyHexMesh.
Table 5-3: number of cells and grading used for each simulation
Simulation Grading Number of cells
D133.7 mm RNG k-ϵ No transducers 8 1’400’000
D211.6 mm RNG k-ϵ No transducers 7 1’700’000
D133.7 mm L = 5D RNG k-ϵ 9 2’900’000
D133.7 mm L = 10D RNG k-ϵ 9 3’100’000
D133.7 mm L = 20D RNG k-ϵ 9 2’700’000
D211.6 mm L = 5D RNG k-ϵ 7 2’600’000
D211.6 mm L = 10D RNG k-ϵ 7 2’600’000
D211.6 mm L = 20D RNG k-ϵ 7 2’600’000
2012/010 1180058 3.0 Released
Page 38 of 111
5.4 Results of the simulations

5.4.1 No transducers, D = 133.7 mm
Simulations without transducers were done in order to evaluate the flow development
downstream of the elbow. For this configuration the diameter is 133.7 mm, the total
number of cells is 1’950’000 and the y+ average value is 62. For the axial geometry set
of simulations the k-ϵ turbulence model was used. However, some simulations with the
k-ω model were done in order to compare the results.
5.4.1.1 Velocity contours
Figure 5-11: Velocity magnitude of the flow over the symmetry plane. A clear separation of the
boundary layer occurs at the elbow. Although the stretch downstream of the elbow is 30D long,
the velocity distribution at the outlet is not yet axisymmetric and the wake generated in the
elbow is still clearly visible.
Figure 5-12: Velocity component parallel to the sound path (y axis), over the symmetry plane.
This figure provides more information than the velocity magnitude. Indeed, by looking at the
negative velocity areas, we can see where the vortices occur.
2012/010 1180058 3.0 Released
Page 39 of 111
Figure 5-13: Zoom of the elbow at the symmetry plane, contours of the y component of velocity.
The separation of the boundary layer is evident in this picture. Below the red wake a vortex is
rotating clockwise.
Figure 5-14: Cross section of the pipe, just downstream of the elbow. Note the two major
symmetric vortices rotating from the wall to the centre. This result was expected, and it is
mentioned in [6].
2012/010 1180058 3.0 Released
Page 40 of 111
5.4.1.2 Pressure contours
Figure 5-15: Gauge pressure contours over the symmetry plane. The pressure at the inlet is the
pressure drop. The maximum pressure is achieved on the external side of the elbow. The
minimum, negative pressure is achieved on the internal side of the elbow.
5.4.1.3 Plots
A number of plots were created with Gnuplot software:
Figure 5-16: Distribution of the axial component of velocity in a cross section, at different
distances from the elbow. At the elbow outlet, red line, as expected, the velocity distribution is
not uniform. What was not expected is that, downstream of the elbow, the farther is the cross
section from the elbow the less uniform is the velocity distribution.
2012/010 1180058 3.0 Released
Page 41 of 111
Figure 5-17: Axial velocity downstream of the elbow, measured at the axis. This plot confirms
that at the outlet of the tube the flow didn’t achieve a complete development, since the slope at
the outlet is not zero yet. Simulations with the transducers will show that this phenomenon is of
minor relevance.
Figure 5-18: Axial velocity measured at the axis, in the stretch between the transducers, without
them applied. The aim of this plot is to show the axial velocity distribution where the transducers
will be applied. At L = 5D, the effects of the wake are still strong within the sound path and the
result is the red curve.
2012/010 1180058 3.0 Released
Page 42 of 111
Figure 5-19: Distribution of the gauge pressure, expressed in Pascal, within a cross section, at
different distances from the elbow. At the elbow outlet, red line, the pressure distribution is not
uniform and a higher pressure occurs on the external side of the elbow, where the velocity is
lower. Also, the farther is the cross section, the lower is the pressure: this is the effect of the
pressure drop.
Figure 5-20: Distribution of gauge pressure at the axis, downstream of the elbow. Apart from the
first stretch, which is near the elbow, the pressure drop is linear along the tube.
2012/010 1180058 3.0 Released
Page 43 of 111
5.4.1.4 The SST k-ω turbulence model for this simulation

The simulation of this paragraph was also done by using another turbulence model, the
SST k-ω, in order to check whether there are major differences. The result is that the
SST k-ω model calculates a slightly greater vortex downstream the elbow. For all the
axial geometry simulations the RNG k-ϵ model will be used, whereas for all the angled
geometry simulations the SST k-ω model will be used.
Figure 5-21: Zoom of the elbow at the symmetry plane, contours of the y component of velocity,
for the SST k-ω turbulence model. Compared to the results for the RNG k-ϵ model, the vortex
downstream of the elbow is greater. However, the difference is not significant, nor will be
significant for the rest of simulations.
Figure 5-22: Axial velocity at the axis, downstream of the elbow: comparison between the k-ϵ
and the k-ω turbulence model. The two trends are very similar.
2012/010 1180058 3.0 Released
Page 44 of 111
5.4.2 No transducers, D = 211.6 mm
Figure 5-23: Velocity component parallel to the sound path and to the y axis, over the symmetry
plane, for D = 211.6 mm. The velocity distribution, compared to the simulation for D = 133.7
mm, is very similar. The result is that also for a higher diameter, and therefore a smaller velocity
(since the flow rate is the same) and a longer tube, the flow is not completely developed at the
outlet.
Figure 5-24: Zoom of the elbow at the symmetry plane, contours of the y component of velocity,
for D = 211.6 mm. The velocity distribution, the wake and the vortices are almost the same as
compared to the results of D = 133.7 mm.
2012/010 1180058 3.0 Released
Page 45 of 111
Figure 5-25: Contours of gauge pressure, expressed in Pascal, at the symmetry plane of the
whole geometry, for D = 211.6 mm. The pressure at the inlet is the pressure drop. The pressure
distribution is the same of D = 133.7 mm. The maximum pressure is achieved on the external
side of the elbow. The minimum, negative pressure is achieved on the internal side of the elbow.
The negative pressure provokes high velocity (for Bernoulli's equation) and separation of the
boundary layer.
5.4.2.3 Plots
Figure 5-26: Distribution of the axial component of velocity in a cross section, at different
distances from the elbow. At the elbow outlet, red line, as expected, the velocity distribution is
not uniform. What was not expected is that, downstream of the elbow, the farther is the cross
section from the elbow the less uniform is the velocity distribution.
2012/010 1180058 3.0 Released
Page 46 of 111
Figure 5-27: Axial velocity downstream of the elbow, measured at the axis. This plot confirms
that, as well as for D = 133.7 mm, at the outlet of the tube the flow didn’t achieve a complete
development, since the slope at the outlet is not zero yet. Simulations with the transducers will
show that this phenomenon is of minor relevance, compared to the influence of the transducers
over the flow.
Figure 5-28: Axial velocity measured at the axis, in the stretch between the transducers, without
them applied. The aim of this plot is to show the axial velocity distribution where the transducers
will be applied. At L = 5D, the effects of the wake are still strong within the sound path and the
result is the red curve. However, simulations with transducers applied will show that the effects
over the flow of the elbow are negligible compared to the effects of the transducers.
2012/010 1180058 3.0 Released
Page 47 of 111
5.4.2.4 Outlet at L = 60D

Some further simulations for the “no transducers, D = 211.6 mm” configuration were
done in order to check the reliability of the results. In this simulation the outlet stretch
was lengthened up to 60 diameters, in order to verify whether the velocity distribution
in the outlet cross section would become axisymmetric. The result is that with this
length the velocity doesn’t become axisymmetric either. Anyway, as also remarked in
the previous paragraphs, the influence of the elbow is negligible compared to the
influence of the transducers.
Figure 5-29: Uy contours for D = 211.6 mm and L = 60D. Also after 60 diameters the velocity
distribution in the cross section is not axisymmetric yet, even if the trend is towards the
symmetry, which will be reached with a longer length.
Figure 5-30: Axial velocity downstream of the elbow, measured at the axis. Comparison between
two outlet stretch lengths: L = 30 diameters and L = 60 diameters. The curve for L = 60D is the
continuation of the curve for L = 30D. Also for L = 60D the flow didn’t achieve a complete
development.
2012/010 1180058 3.0 Released
Page 48 of 111
5.4.2.5 Outlet at L = 60D, fluid water and flow velocity 1 m/s

Following the results of paragraph 5.4.2.4, a further and last simulation was done,
using this time water instead of gas, and using a flow velocity of 1 m/s instead of 8.75
m\s. The result is that neither the velocity nor the fluid phase influences the flow
development length.
Figure 5-31: Uy contours for D = 211.6 mm, L = 60D, fluid water and v = 1 m/s. Also after 60
diameters the velocity distribution in the cross section is not axisymmetric yet.
Figure 5-32: Axial velocity downstream of the elbow, measured at the axis, for D = 211.6 mm, L
= 60D, fluid water and v = 1 m/s. Even at 60D, the flow didn’t achieve a complete development.
2012/010 1180058 3.0 Released
Page 49 of 111
5.4.3 Transducers at 5 Diameters, D = 133.7 mm

After having studied what happens without transducers, simulations with transducers
at different positions were done. L = 5D is the position closest to the elbow. The results
of this simulation show that the influence of the transducers over the flow distribution
within the sound path is so great that, basically, the distance from the elbow is
negligible.
Figure 5-33: Uy contours for the axial geometry with transducers, L = 5D and D = 133.7 mm. In
this simulation the supporting fins were modelled, and they can be seen above the transducers.
They don’t significantly influence the flow; therefore, they will be eliminated in all the remaining
simulations. A stagnation point can be observed upstream of the second transducer (from right).
Two whirling zones, as a result of the separation of the boundary layer, can be observed
downstream of both transducers. An interesting result is that the deflection cones, designed in
this way, are not sufficient to avoid the separation of the boundary layer and the consequent
vortices. In conclusion, the sound path between the two transducers is characterized by an
extremely non-uniform velocity distribution. This wouldn’t allow an effective calibration of the
device since the vortices can vary with the flow rate.
Figure 5-34: 3D image of the axial flow meter, with Uy contours and streamlines.
2012/010 1180058 3.0 Released
Page 50 of 111
Figure 5-35: Gauge pressure contours. The pressure increases significantly upstream of the
second transducer, at the stagnation point. However, in spite of the non-uniform pressure field,
the consequent variation of density (and therefore of speed of sound) can be considered
negligible.
5.4.3.3 Plots
Figure 5-36: Axial velocity distribution in a cross section set at equal distance between the two
transducers. The velocity distribution, due to the transducers, is completely different from the
parabolic one that characterizes developed flows in tubes.
2012/010 1180058 3.0 Released
Page 51 of 111
Figure 5-37: Axial velocity distribution along the axis, between the two transducers. In the first 2
diameters, the flow is characterized by a negative velocity (whirling area) due to the separation
of the boundary layer.
5.4.3.4 The k-ω turbulence model for this simulation

This configuration was also simulated with the k-ω turbulence model in order to check
if there is any relevant difference. The conclusion is that both the models provide
approximately the same results.
Figure 5-38: Axial velocity distribution along the axis, between the two transducers.
2012/010 1180058 3.0 Released
Page 52 of 111

From this simulation onwards the supports will be removed, since they don’t
significantly influence the flow, in order to improve the legibility of the contours.
Figure 5-39: Axial velocity contours for L = 10D. The first transducer provokes a huge wake,
which is not axisymmetric due to the influence of the elbow over the flow upstream of the
transducers.
5.4.4.2 Plots
2012/010 1180058 3.0 Released
Page 53 of 111
Figure 5-42: Axial velocity contours for L = 20D. The first transducer provokes a huge wake,
which is not axisymmetric due to the influence of the elbow over the flow upstream of the
transducers.
2012/010 1180058 3.0 Released
Page 54 of 111
5.4.5.2 Plots
2012/010 1180058 3.0 Released
Page 55 of 111
Figure 5-45: Axial velocity contours for L = 5D and D = 211.6 mm. Compared to D = 133.7 mm,
the wakes downstream of the transducers are thinner but evident. The wakes are not
axisymmetric due to the influence of the elbow over the flow upstream of the transducers.
5.4.6.2 Plots
2012/010 1180058 3.0 Released
Page 56 of 111
Figure 5-48: Axial velocity contours for L = 10D and D = 211.6 mm.
2012/010 1180058 3.0 Released
Page 57 of 111
5.4.7.2 Plots
2012/010 1180058 3.0 Released
Page 58 of 111
Figure 5-51: Axial velocity contours for L = 20D and D = 211.6 mm.
5.4.8.2 Plots
2012/010 1180058 3.0 Released
Page 59 of 111
5.4.9 Comparison of the velocity results
Figure 5-54: Comparison of the axial velocity distributions along the axis, between the two
transducers, between all the configurations with D = 133.7 mm. The result is that the distance
from the elbow doesn’t significantly influence the flow. On the other hand, the transducers
influence the flow significantly, and this led to the final choice of not adopting this geometry.
2012/010 1180058 3.0 Released
Page 60 of 111
Figure 5-55: Comparison of the axial velocity distributions along the axis, between the two
transducers, between all the configurations with D = 211.6 mm. The result is that the distance
from the elbow doesn’t significantly influence the flow. On the other hand, the transducers
influence the flow significantly, and this led to the final choice of not adopting this geometry.
5.5 Evaluation of the pressure drops
Table 5-4: Comparison of pressure drops calculated with an empirical correlation (page 110 of
[6]) and with the simulations.
Correlation Simulation Delta%
D133.7 mm, No transducers, L =
713.2 629.1 11.8% Pa
30D, Fluorocarbon, w = 21.9m/s
D211.6 mm, No transducers, L =
117.2 105.2 10.3% Pa
30D, Fluorocarbon, w = 8.7m/s
Table 5-5: Evaluation of the pressure drops resulting from the simulations, for the configurations
with transducers.
Pressure
D133.7 mm L = 5D RNG k-ϵ 935.9 Pa
D133.7 mm L = 5D SST k-ω 960.3 Pa
D133.7 mm L = 10D RNG k-ϵ 861.1 Pa
D133.7 mm L = 20D RNG k-ϵ 852.8 Pa
D211.6 mm L = 5D RNG k-ϵ 116.9 Pa
D211.6 mm L = 10D RNG k-ϵ 117.6 Pa
D211.6 mm L = 20D RNG k-ϵ 116.7 Pa
2012/010 1180058 3.0 Released
Page 61 of 111
5.6 Average velocities

Since the sound wave path has a diameter approximately equal to the diameter of the
transducers, the volume we chose to evaluate the average velocity in the sound path is
a cylinder between the two transducers, which has a diameter equal to the transducers
diameter, as shown in Figure 5-56. The velocity so evaluated will best represent the
velocity measured by the transducers.
Figure 5-56: Volume chosen to evaluate the average velocity
The following integrals were numerically solved in order to evaluate the average
velocities:
Average Velocity ∫ ⃗( ) 5-2
Average Absolute Velocity ∫ | ⃗( )| 5-3
Root Mean Square of the

√ ∫⃗ ( ) 5-4
velocity in the volume
Average Velocity of the ̇
5-5
flow
2012/010 1180058 3.0 Released
Page 62 of 111
Table 5-6: Evaluation of the volume averaged velocity in the sound path, and its deviation from
the average velocity in the main tube.
Root
Volume mean
Volume Deviation Deviation Deviation
Average averaged square of
averaged from the from the from the
velocity of absolute the
velocity in average average average
the flow velocity in velocity in
the sound velocity of velocity of velocity of
[m/s] the sound the sound
path [m/s] the flow the flow the flow
path [m/s] path
[m/s]
D133.7 mm
21.92 16.9 -22.8% 17.6 -19.9% 18.8 -14.4%
L = 5D k-ϵ
D133.7 mm
21.92 16.1 -26.5% 16.7 -23.9% 17.8 -19.0%
L = 5D k-ω
D133.7 mm
21.92 17.0 -22.3% 17.8 -18.9% 19.1 -13.0%
L = 10D k- ϵ
D133.7 mm
21.92 16.3 -25.8% 17.0 -22.3% 18.2 -16.9%
L = 20D k- ϵ
D211.6 mm
8.75 7.4 -15.4% 7.6 -13.2% 8.0 -8.8%
L = 5D k- ϵ
D211.6 mm
8.75 7.2 -18.2% 7.4 -15.1% 7.9 -9.5%
L = 10D k- ϵ
D211.6 mm
8.75 6.4 -26.9% 6.7 -23.3% 7.2 -17.9%
L = 20D k- ϵ
D211.6 mm
k- ϵ No 8.75 9.2 5.1% 9.2 5.1% 9.2 5.1%
transducers
2012/010 1180058 3.0 Released
Page 63 of 111
6. Simulations for the angled flow meter

6.1 Problem documentation and simplification of the model
The angled geometry is the second of the two geometries studied. The results of this
study are positive, and show that a flow meter with an angled geometry would be
capable to provide a reliable measurement of velocity and flow rate. CFD studies also
allowed to optimize the angled geometry, and to discard bad configurations.
The set-up of the “Angled Flow Meter” consists of a transversal secondary pipe, of
internal diameter d, intersecting at a certain angle α with the main tube of internal
diameter DMain. The secondary pipe is closed at both ends by a pair of transducers,
wafer-like devices responsible for creating the sound wave. In order for this sound
pressure wave to propagate optimally, static vapour is needed in the transversal
secondary pipe, or, at least, the influence of axial direction velocity components should
be minimized.
Three different sensor positions are taken into account:
1. In the first configuration, Figure 6-1, the transducers are backed-off at a length
L inside their cylinders from the intersection of the main and secondary pipe
wall, where L equals the distance between the vertices of the secondary pipe at
both sides of the main pipe (DMain/sinα). L is also the sound propagation path
length in the main tube. The distance between the transducers is therefore
3*DMain/sinα.
Figure 6-1: Configuration with the maximum transducers inter-distance: 3*DMain/sinα.

2012/010 1180058 3.0 Released
Page 64 of 111
2. In the second configuration, Figure 6-2, the transducers are positioned at the
intersection between the two tubes. The depth of each transducer is fully
accommodated within its secondary tube, but one side of the transducer
touches the flow volume in the main tube. The distance between the
transducers will be (DMain/sinα+d/tgα) where the distance d/ (2*tgα) is the
distance between the centre of the front face of each transducer and the
boundary of the gas flow volume in the main tube.
Figure 6-2: Configuration with the mid transducer inter-distance: DMain/sinα + d/tanα.
3. In the third configuration, Figure 6-3, transducers are positioned with the
maximum inter-distance at which all the sound path is in the main tube. The
centre of each transducer is flush with the inner bore of the main tube. The
distance between the centres of the front faces of the transducers will be L =
DMain/sinα. The sensors are treated as plungers (i.e. s sufficiently long, as shown
in Fig. 1c) to prevent turbulent refrigerants flow behind the transducers in the
traversing tubes.
Figure 6-3: Configuration with the minimum transducers inter-distance: DMain/sinα. In this case
the entire sound path is within the main tube.
2012/010 1180058 3.0 Released
Page 65 of 111
It is intended to check the influence of the sonar transducers on the main tube flow
and to choose the optimum placement.
The goal, for the angled flow meter, is computing the refrigerant flow in order to check
if the fluid is actually static inside the secondary tubes of the “angled flow meter”.
Steady-state simulations will be performed for three different positions of the

transducers, for two different diameters of the main tube (133.7 mm and 211.6 mm)
and three different angles (15°, 30° and 45°)
The total number of simulations will be 2 x 3 x 3 = 18.
The following results will be provided:

● The velocity field and the streamlines at the symmetry plane.
● The maximum local component of velocity parallel to the wave path.
● The volume average of the squared value of the velocity component parallel to the
wave (considering the cylindrical volume between 2 sensors).
Table 6-1: List of all the simulations planned

Simulation No. Main Tube Transducers Secondary Tube
Diameter Position Angle
1 133.7 max 45°
2 133.7 max 30°
3 133.7 max 15°
4 133.7 mid 45°
5 133.7 mid 30°
6 133.7 mid 15°
7 133.7 min 45°
8 133.7 min 30°
9 133.7 min 15°
10 211.6 max 45°
11 211.6 max 30°
12 211.6 max 15°
13 211.6 mid 45°
14 211.6 mid 30°
15 211.6 mid 15°
16 211.6 min 45°
17 211.6 min 30°
18 211.6 min 15°
Dimensions:
1. Inner diameters of main pipe: D = 133.7 mm and 211.6 mm;
2. Inner diameter of the secondary, oblique pipe: d = 50 mm;
3. Intersection angle:
 α = 15°
 α = 30°
 α = 45°
4. Diameter of the transducers = 50 mm
5. Length of the transducers = 2 mm
2012/010 1180058 3.0 Released
Page 66 of 111
6.2 The CAD modelling of the angled geometry
Figure 6-4: CAD modelling of the angled geometry. Configuration with maximum distance
between the transducers, 45°, D = 133.7 mm. The main tube is in light blue, the secondary
tubes are in dark blue and the transducers are in yellow.
Figure 6-5: CAD modelling of the angled geometry. Configuration with mid distance between the
transducers, 45°, D = 133.7 mm.
2012/010 1180058 3.0 Released
Page 67 of 111
Figure 6-6: CAD modelling of the angled geometry. Configuration with minimum distance
between the transducers, 45°, D = 133.7 mm.
6.3 The mesh generation

Unlike the axial configuration in which only the transducers were imported in
SnappyHexMesh through the STL format, for the angled configuration the whole CAD
geometry was imported and processed with SnappyHexMesh.
In order to achieve a correct mesh for the main pipe, with a square mesh near the axis
and cells parallel to the wall near the wall, a “base” mesh was done with BlockMesh.
Figure 6-7: Mesh at the symmetry plane of the angled geometry, D = 211.6 mm, angle 45° and
max transducers distance. It can be noticed the grading towards the outlet, the grading towards
the inlet and the grading towards the wall. The gradings towards outlet and inlet allowed
reducing the total amount of cells, whereas the grading towards the wall allowed obtaining a
correct y+ value.
2012/010 1180058 3.0 Released
Page 68 of 111
Figure 6-8: Mesh on the tube surface and the secondary tubes. For the secondary tubes, a
surface refinement of value “1” has been applied, which means that the cell has been divided
into four on their surface.
Table 6-2: Number of cells for every simulation. A higher number of cells correspond to a smaller
grading towards the wall. The grading towards the wall is indeed applied in order to both achieve
the correct value of y+ and reduce the total number of cells.
Grading
Simulation Number of cells
(towards the wall)
D133.7 mm, α = 15°, distance max 1.1 4’400’000
D133.7 mm, α = 45°, distance max 2 1’300’000
D133.7 mm, α = 15°, distance mid 2 7’400’000
D133.7 mm, α = 15°, distance min 2 5’100’000
D133.7 mm, α = 30°, distance min 1.1 8’100’000
D211.6 mm, α = 15°, distance max 2 2’600’000
D211.6 mm, α = 15°, distance mid 1.8 3’100’000
2012/010 1180058 3.0 Released
Page 69 of 111
6.4 Results of the simulations

For all the simulation of the angled geometry the SST k-ω turbulence model was
chosen, being this model more recent than the RNG k-ϵ (see 4.3.2.2). However, a
simulation with the k-ϵ model was done in order to cross check the differences, and the
result is that the difference between the two models is negligible.
6.4.1 Transducers at the maximum distance, α = 15°, D = 133.7 mm
Figure 6-9: Contours of the axial component of velocity.
The negative velocity on one side of both secondary tubes is a clear sign of the
presence of a vortex. The main stream drags the fluid in the secondary pipes, and the
result is a clockwise-rotating vortex in the upper tube and a counter clockwise-rotating
vortex in the lower tube.
The main stream is influenced by the upper secondary tube, because of the edge facing
towards the main stream which causes a wake. However, this wake doesn’t influence
the measurement as it is located outside the sound path. The greater is the angle the
smaller is the wake, as the elliptic intersection between main and secondary tubes that
causes the wake decreases when the angle increases.
Figure 6-10: Contours of gauge pressure, expressed in Pascal
The gauge pressure in the upper secondary pipe is greater than the one in the lower
secondary pipe. This happens because the fluid in the main pipe acts as a plunger
2012/010 1180058 3.0 Released
Page 70 of 111
towards the still fluid in the secondary pipe, and transfers dynamic pressure to it. The
greater is the angle the lower is the pressure in the secondary pipe. The theoretical
upper limit for the pressure in the secondary tube is the total pressure of the main
stream, which would be reached for an angle of 0° (which makes the secondary tube
like a Pitot tube). The lower limit for the pressure in the secondary tube is the static
pressure of the main stream, which would be reached for an angle of 90° (which makes
the secondary tube like a static pressure gauge). However, this pressure difference can
be considered negligible in terms of consequent variation of density, and speed of
sound. Here’s some calculation made with RefProp:
Table 6-3: C3F8 Specified state points

Temperature (°C) Pressure (Pa) Density (kg/m3) Speed of sound (m/s)
20 50000 3.9001 116.11
20 50500 3.9396 116.10
20 51000 3.979 116.09
6.4.1.3 Plots
Figure 6-11: Axial velocity measured in the sound path, between the two transducers. The
direction z is vertical and perpendicular to the tube axis, and the dimensionless z/D range [-0.5,
0.5] is within the main pipe. The total dimensionless covered length is 3, since the secondary
tube length is 3 DMain/sinα.
2012/010 1180058 3.0 Released
Page 71 of 111
Figure 6-12: Gauge pressure measured in the sound path, between the two transducers. The
6.4.1.4 The k-ϵ turbulence model for this simulation
Figure 6-13: Contours of the axial component of velocity. The velocity distribution calculated with
the k-ϵ turbulence model is practically identical to the one calculated with SST k-ω.
2012/010 1180058 3.0 Released
Page 72 of 111
Figure 6-14: Contours of the axial component of velocity. The wake downstream of the upper
secondary pipe is smaller than the wake in the configuration with α = 15°.
Figure 6-15: Contours of gauge pressure, expressed in Pascal. The pressure in the upper
secondary tube is smaller than the pressure in the configuration with α = 15°.
2012/010 1180058 3.0 Released
Page 73 of 111
6.4.2.3 Plots
2012/010 1180058 3.0 Released
Page 74 of 111
Figure 6-19: Contours of the axial component of velocity with streamlines, 3D view.
2012/010 1180058 3.0 Released
Page 75 of 111
Figure 6-20: Contours of gauge pressure, expressed in Pascal. The pressure in the upper
secondary tube is smaller than the pressure in the configuration with α = 30°.
6.4.3.3 Plots
2012/010 1180058 3.0 Released
Page 76 of 111
6.4.4 Transducers at the mid distance, α = 15°, D = 133.7 mm
secondary pipe varies with the angle but not with the secondary tube length.
2012/010 1180058 3.0 Released
Page 77 of 111
6.4.4.2 Plots
0.5] is within the main pipe.
2012/010 1180058 3.0 Released
Page 78 of 111
6.4.5.2 Plots
2012/010 1180058 3.0 Released
Page 79 of 111

2012/010 1180058 3.0 Released
Page 80 of 111
6.4.6.2 Plots
2012/010 1180058 3.0 Released
Page 81 of 111
6.4.7 Transducers at the minimum distance, α = 15°, D = 133.7 mm
Figure 6-32: Contours of the axial component of velocity. A counter clockwise vortex occurs
downstream of the lower transducer, within the sound path. This might affect the measurement
because the velocity distribution is not uniform. What was expected is that the shape of the
vortex, and consequently the velocity distribution, varies with the flow rate. This is not true as
confirmed by simulations with 0.6 m/s (half flow), in which the velocity distribution is almost
identical.
6.4.7.2 Plots
0.5] is within the main pipe. The velocity distribution, unlike configurations in which the
transducers are not impinging on the flow, is not axisymmetric and a negative velocity (clear
sign of a vortex) can be observed in the first 0.5 diameters.
2012/010 1180058 3.0 Released
Page 82 of 111
confirmed by the simulations with half flow.
2012/010 1180058 3.0 Released
Page 83 of 111
6.4.8.2 Plots
2012/010 1180058 3.0 Released
Page 84 of 111
6.4.9.2 Plots
2012/010 1180058 3.0 Released
Page 85 of 111
In the configurations with a diameter of 211.6 mm, compared to D = 133.7 mm, the
velocity distribution remains approximately the same, with the same vortices and
wakes. For a detailed description, refer to 6.4.1.
2012/010 1180058 3.0 Released
Page 86 of 111
6.4.10.2 Plots
tube length is 3 DMain/sinα. For the configurations with D = 211.6 mm, the velocity distribution is
even more uniform, as the velocity is smaller.
2012/010 1180058 3.0 Released
Page 87 of 111
6.4.11.2 Plots
2012/010 1180058 3.0 Released
Page 88 of 111
2012/010 1180058 3.0 Released
Page 89 of 111
6.4.12.2 Plots
2012/010 1180058 3.0 Released
Page 90 of 111
secondary pipe varies with the angle but not with the secondary tube length.
6.4.13.2 Plots
2012/010 1180058 3.0 Released
Page 91 of 111

2012/010 1180058 3.0 Released
Page 92 of 111
6.4.14.2 Plots
2012/010 1180058 3.0 Released
Page 93 of 111
6.4.15.2 Plots
2012/010 1180058 3.0 Released
Page 94 of 111
vortex, and consequently the velocity distribution, varies with the flow rate. This is not true (as
confirmed by simulations with 0.6 m/s half flow), in which the velocity distribution is almost
identical.
2012/010 1180058 3.0 Released
Page 95 of 111
6.4.16.2 Plots
2012/010 1180058 3.0 Released
Page 96 of 111
6.4.17.2 Plots
2012/010 1180058 3.0 Released
Page 97 of 111
2012/010 1180058 3.0 Released
Page 98 of 111
6.4.18.2 Plots
2012/010 1180058 3.0 Released
Page 99 of 111
6.4.19 Comparison of the velocity results
Figure 6-68: Axial velocity measured in the sound path, between the two transducers, for the
maximum distance and three different angles. The direction z is vertical and perpendicular to the
tube axis, and the dimensionless z/D range [-0.5, 0.5] is within the main pipe. This plot shows
that the most axisymmetric configuration is with an angle of α = 45°. This can be explained
considering that for an angle of 45° the intersecting area between main and secondary tube is
smaller. Therefore, for α = 45° the flow doesn’t widen in correspondence of the intersection.
mid distance and three different angles. As well as for the maximum distance, for the mid
distance the most axisymmetric configuration is with an angle of α = 45°. This can be explained
considering that for an angle of 45° the intersecting area between main and secondary tube is
smaller. Therefore, for α = 45° the flow doesn’t widen in correspondence of the intersection.
2012/010 1180058 3.0 Released
Page 100 of 111
minimum distance and three different angles.
Figure 6-71: Axial velocity measured in the sound path, between the two transducers, for D =
211.6 mm, maximum distance and three different angles. In D = 211.6 mm, the lower velocity
in the pipe makes the velocity distribution more uniform than in D = 133.7 mm.
2012/010 1180058 3.0 Released
Page 101 of 111
mid distance and three different angles.
minimum distance and three different angles.
2012/010 1180058 3.0 Released
Page 102 of 111
6.4.20 Comparison of the pressure results
Figure 6-74: Gauge pressure measured in the sound path, between the two transducers, for
three different angles, D = 133.7 mm and maximum distance. The lower is the angle, the higher
is the pressure in the upper secondary pipe (which is the pipe facing upriver). Indeed, the fluid in
the main pipe transfers kinetic energy to the fluid in the secondary pipe. The limit case is when
the secondary pipe is perpendicular (α = 90°): in this case no kinetic energy is transferred and
no increase of pressure in the secondary pipe is registered. For further information refer to
paragraph 6.4.1.
Figure 6-75: Gauge pressure measured in the sound path, between the two transducers for three
different angles, D = 133.7 mm and mid distance.
2012/010 1180058 3.0 Released
Page 103 of 111
different angles, D = 133.7 mm and minimum distance.
different angles, D = 211.6 mm and maximum distance.
2012/010 1180058 3.0 Released
Page 104 of 111
different angles, D = 211.6 mm and mid distance.
different angles, D = 211.6 mm and minimum distance.
2012/010 1180058 3.0 Released
Page 105 of 111
6.4.21 Half flow rate simulations

In order to check whether the velocity distribution and the size of vortices and wakes
vary with the flow rate, a set of simulations with half flow rate (0.6 kg/s) has been
done for 3 configurations: α = 30°, D = 133.7 mm, and all three distances between the
transducers.
Full flow rate

1.2kg/s
Figure 6-80: Axial velocity measured in the sound path, between the two transducers, for full
flow rate (1.2kg/s), α = 30°, D = 133.7 mm and three different distances.
Half flow rate

0.6kg/s
Figure 6-81: Axial velocity measured in the sound path, between the two transducers, for half
flow rate (0.6kg/s), α = 30°, D = 133.7 mm and three different distances. The velocity profile for
half flow rate is practically identic to the full flow rate profile. Surprisingly, the velocity
distribution remains the same also for the minimum distance configuration, in which a major
eddy occurs within the sound path. These results are very positive as far as the calibration of the
instrument is concerned.
2012/010 1180058 3.0 Released
Page 106 of 111
6.5 Average velocities

At the end of the simulation process, the average velocities within the sound path
(Figure 6-82) were calculated by numerically solving the integrals showed in paragraph
5.6.
Figure 6-82: Volume (the coloured area) in which the average velocity were calculated.
In order to isolate the cylinder (coloured in Figure 6-82) from the rest of the geometry
and calculate the average velocity over it, the following formula (a cylinder in the 3D
space) was implemented in Paraview.
( )
( ( )
) 6-1
 Where: x, y and z are the coordinates of the reference system

 y0 is the y coordinate of the centre of the secondary pipe
 α is the intersection angle
 r is the radius of the secondary tube
Results are shown in Table 6-4. The theoretical average velocity deviation expected in,
for example, the maximum distance configurations, is -66.67%, since 2/3 of the
secondary pipe is characterized by still flow, or, at least, the volume averaged velocity
in 2/3 of the pipe is zero.
2012/010 1180058 3.0 Released
Page 107 of 111
Table 6-4: Evaluation of the average axial velocity calculated in a cylindrical volume defined by
the transducer diameter and spacing with comparison with the average flow in the main tube.
Average Volume- Vc Interface
Signed % Vs
DMain [mm],  [°] axial flow averaged axial correction aperture
deficit, and
and transducers velocity v; velocity Vs, factor [cm2]
expected deficit
distance main tube [m/s] between from (5cm ID
from sound path
[m/s] transducers sound path sound tube)
133.7, 15°, max 21.92 7.81 -64.4% (-66.7%) 3 75.9
133.7, 30°, max 21.92 7.50 -65.8% (-66.7%) 3 39.3
133.7, 45°, max 21.92 8.05 -63.3% (-66.7%) 3 27.8
133.7, 15°, mid 21.92 17.71 -19.2% (-24.1%) 1.32 75.9
133.7, 30°, mid 21.92 17.68 -19.3% (-22.2%) 1.29 39.3
133.7, 45°, mid 21.92 17.43 -20.5% (-18.9%) 1.23 27.8
133.7, 15°, min 21.92 19.29 -12.0% (0%) 1 75.9
133.7, 30°, min 21.92 18.17 -17.1% (0%) 1 39.3
133.7, 45°, min 21.92 19.00 -13.3% (0%) 1 27.8
133.7, 30°, max 10.96 3.76 -65.7%(-66.7%) 3 39.3
133.7, 30°, mid 10.96 8.80 -19.7% (-24.1%) 1.29 39.3
133.7, 30°, min 10.96 9.20 -16.0% (0%) 1 39.3
211.6, 15°, max 8.75 2.95 -66.3% (-66.7%) 3 75.9
211.6, 30°, max 8.75 2.84 -67.5% (-66.7%) 3 39.3
211.6, 45°, max 8.75 2.85 -67.4% (-66.7%) 3 27.8
211.6, 15°, mid 8.75 7.72 -11.8% (-16.7%) 1.20 75.9
211.6, 30°, mid 8.75 7.64 -12.7% (-15.3%) 1.18 39.3
211.6, 45°, mid 8.75 7.71 -11.8% (-12.8%) 1.15 27.8
211.6, 15°, min 8.75 8.20 -6.3% (0%) 1 75.9
211.6, 30°, min 8.75 8.03 -8.3% (0%) 1 39.3
211.6, 45°, min 8.75 8.16 -6.8% (0%) 1 27.8
Column 3 of Table 6-4 shows the calculated velocity component, vc, in the direction of
flow in the main tube. This is averaged in a cylindrical volume defined by the diameter
and spacing of the two transducers, using axial velocity values at points along the
whole sound path, including, where applicable, the regions beyond the confines of the
main tube. The signed percentage deficit, SD, of the calculated flow, , from the
average axial flow, v, in the main tube, given by
is shown in column 4 of Table 6-4, where it is compared with the expected deficit
based on the ratio of the sound path crossing the flowing vapour, , to the total
length, L, including the path length, L’, in quasi-static vapour. The expected per cent
deficit, ED, is defined as:
6-2
2012/010 1180058 3.0 Released
Page 108 of 111
Taking an example for the configuration of Figure 6-1, with 15° crossing angle in a
133.7 mm diameter main tube, a value of -64.4% for SD is calculated. In this
geometry 2/3 of the sound path is characterized by quasi-static flow. The expected
deficit from sound path arguments ED would be -66.7%, meaning that the flow velocity
measured in such an instrument would need to be multiplied by a correction factor, CF,
to find the velocity in the main tube, where (maintaining the sign convention of the
expected deficit) the correction factor is expressed in terms of expected deficit as
( ) 6-3
In the configuration of Figure 6-1, CF = 3, while in the configuration of Figure 6-2 the
corresponding correction factor depends explicitly on the crossing angle and main tube
diameter, as shown in column 5 of Table 6-4. In the configuration of Figure 6-3, where
the entire sound path is in the flowing vapour, the correction factor is unity.
The difference between the axial velocity deviation and the deviation expected from
length arguments is a quality estimator for the particular geometry of ultrasonic flow
meter, taking into account the effects of turbulence and vortices in the sound path. It
can be seen from Table 6-4 that the differences are minimised in the geometry of
Figure 6-1 where the transducers are backed off a significant distance from the main
tube.
2012/010 1180058 3.0 Released
Page 109 of 111
7. Conclusions
The CFD analysis has shown that no one of the 8 axial geometry configurations studied
is capable to provide a reliable measurement of the volumetric flow in the pipe,
whereas most of the angled geometry configurations would be instead good solutions.
The reason why the axial geometry is to discard is twofold:
1. The transducers have a strong influence over the flow, and therefore the
average velocity is underestimated by approximately 20%.
2. Even if there weren’t any influence of the transducers, the velocity would be
overestimated by approximately 5%, because only the flow in the centre of the
tube is measured.
On the other hand, the angled flow meter is capable to work properly, as the sound
wave would in both cases passes through the whole cross section. The 18 simulations
helped to discard the following configurations:
1. Configurations with D = 211.6 mm, which although allow a considerably lower

pressure drop than configurations with D = 133.7 mm, are characterized by a
too slow flow. Therefore, the velocity component parallel to the sound wave
would be too small and the measurement uncertainty would be unacceptable.
2. Configurations with α = 15°, which although would increase the velocity

component parallel to the sound wave, would not be easy to manufacture at an
acceptable cost. However, a detailed cost analysis wasn’t done yet.
3. Configurations with a minimum distance between the transducers, which would

not allow a reliable measurement since a major eddy is located within the sound
path downstream of the first transducer and, therefore, the measured velocity
would be different from the average one.
The best configurations are α = 30° and α = 45°, with the maximum or the mid
distance between the transducers.
In order to choose the best geometry, other considerations involving cost and
manufacturing facility have to be done.
2012/010 1180058 3.0 Released
Page 110 of 111
8. Bibliography
[1] J. Botelho Direito, E. Perez Rodriguez, K. Egorov, L. Zwalinski, A. Bitadze, “EDMS Document
No.1083852 - General description of the full scale thermosiphon cooling system for ATLAS
SCT and Pixel,” CERN, 2010.
[2] OpenCFD, “OpenFOAM User Guide,” 2011. [Online]. Available: www.openfoam.com.
[3] F. R. Menter, “Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications,”
AIAA Journal, vol. 32, no. 8, pp. 1598-1605, 1994.
[4] F. R. Menter, “Zonal Two Equation k-ω Turbulence Models for Aerodynamic Flows,” AIAA
Paper , pp. 93-2906, 1993.
[5] S. B. Pope, Turbulent Flows, Cambridge University Press, 2000.
[6] I. E. Idelchik, M. O. Steinberg, Handbook of Hydraulic Resistance, Jaico Publishing House,
2003.
[7] S. V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere Publishing Corporation,
1980.
[8] J. John D. Anderson, Computational Fluid Dynamics - The Basics with Applications, McGraw-
Hill, 1995.
[9] Gennaro Bozza, Enrico Da Riva, “Project Request, EDMS No. 1157219, Sonar Flow Meter
Device,” CERN, 2011.
[10]Jose Botelho Direito, “EDMS Document No. 1074962 - Sonar Flow Meter Gas Analyzer,
Mechanical Drawings,” CERN, 2010.
2012/010 1180058 3.0 Released
Page 111 of 111
9. Index
ATLAS, 6, 8, 9, 28 OpenFOAM, 16, 17, 21, 27

Bernoulli, 45 Paraview, 28
block, 17, 18, 19, 20 Pixel, 6
BlockMesh, 16, 17, 18, 19, 20, 21, 34, 35, Post-processing, 28
67 Prandtl, 27
boundary condition, 24 Pre-processing, 16
boundary layer, 23, 27 Processing, 25
C2F6, 6, 7, 11 RANS, 26
C3F8, 6, 7, 11, 29 refinement box, 36
CAD, 16 RefProp, 11, 70
CATIA, 16 Reynolds, 26
cells, 37, 68 RNG k-ϵ, 26
CFD, 5, 8, 10, 16 SCT, 6
CheckMesh, 21 SIMPLE, 27
Conservation, 25 skewness, 21
contours, 5, 10 SnappyHexMesh, 16, 21, 22, 24, 34, 37, 67
deflection cones, 30 SST k-ω, 26, 27
detector, 7 steady, 27, 33
eddy, 26, 27, 105 Stereolithography, 16, 21
energy, 11, 25, 26, 27 STL, 16, 21, 34, 67
Excel, 19 supports, 30, 31
Finite Volume, 16 surface refinement, 36
GnuPlot, 28, 40 symmetry, 18, 31, 33
grading, 17, 37, 68 thermosiphon, 6, 28
gravity, 33 transducers, 5, 7, 8, 9, 10, 13
hexahedral, 16 TRT, 6
law of the wall, 23 turbulence, 26, 27
Mach, 27 turbulent dissipation rate, 26, 27
mass, 6, 25, 26 turbulent kinetic energy, 26, 27
mixture ratio, 5, 7 ultrasonic, 5, 7, 8, 12, 13
momentum, 11, 25, 26, 28 under-relaxation, 24
Navier Stokes, 25, 26 vortex, 39, 43, 69, 81, 82, 83, 84, 94, 95,
NIST, 11 96, 97, 98
non-orthogonality, 21 y+, 23, 38, 67, 68

A Combined UFM and Binary Vapour Mixture Analyser For The Atlas Silicon Tracker PDF

Uploaded by

Copyright:

Available Formats

A Combined UFM and Binary Vapour Mixture Analyser For The Atlas Silicon Tracker PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Combined UFM and Binary Vapour Mixture Analyser For The Atlas Silicon Tracker PDF

Uploaded by

Copyright:

Available Formats

CERN EDMS NO. REV.

CH1211 Geneva 23 1180058 3.0 Released

A COMBINED ULTRASONIC FLOW METER AND

DOCUMENT PREPARED BY: DOCUMENT CHECKED BY: DOCUMENT APPROVED BY:

Gennaro Bozza EN/CV/DC Greg Hallewell /CPPM Greg Hallewell /CPPM

2012/010 1180058 3.0 Released

1.0 2011-12-06 111

2012/010 1180058 3.0 Released

3. Principle of operation of the Acoustic Flow Meter ........................ 11

5. Simulations for the axial flow meter ............................................ 30

6. Simulations for the angled flow meter ......................................... 63

2012/010 1180058 3.0 Released

6.4 Results of the simulations ................................................................................. 69

7. Conclusions ................................................................................ 109

2012/010 1180058 3.0 Released

2012/010 1180058 3.0 Released

The upgrade may require a change from C3F8 (octafluoro-propane) to a blend

2012/010 1180058 3.0 Released

The instrument, illustrated in different possible geometries in the following chapters, is

2012/010 1180058 3.0 Released

2.1 The flow meter for lower flow rates applications

2012/010 1180058 3.0 Released

2.2 The flow meter for higher flow rates applications

Figure 2-4: Layout of an in-line flow meter.

Figure 2-5: Configuration with transducer inter-distance = 3*Dmain/sinα.

2012/010 1180058 3.0 Released

Chapter 4 is dedicated to Computational Fluid Dynamics. The models used, the

2012/010 1180058 3.0 Released

3. Principle of operation of the Acoustic Flow Meter

2012/010 1180058 3.0 Released

measured by the microcontroller, which also handles communication via a USB/RS232

3.3 The Transducer

2012/010 1180058 3.0 Released

3.4 The algorithms for calculating the sound transit times

Gas mixture is continuously analysed using SCADA 6 software, by comparing the

2012/010 1180058 3.0 Released

3.4.2 Algorithms for the angled geometry

In configurations where the transducer spacing includes an element L’ containing

From either equations above the velocity v can be expressed as

2012/010 1180058 3.0 Released

2012/010 1180058 3.0 Released

4. Procedure for a CFD analysis

OpenFOAM uses the Finite Volume discretization method on unstructured meshes. It

BlockMesh is a simple meshing tool which allows to generate simple meshed

SnappyHexMesh doesn’t generate geometries, but allows to generate a mesh on

2012/010 1180058 3.0 Released

4.2.2 Mesh generation with BlockMesh

2012/010 1180058 3.0 Released

2012/010 1180058 3.0 Released

4.2.2.1 BlockMeshDict with Excel

2012/010 1180058 3.0 Released

2012/010 1180058 3.0 Released

4.2.3 Mesh generation with SnappyHexMesh

● Surface data files in STL format, either binary or ASCII, located in a

The three stages of the meshing process are:

2012/010 1180058 3.0 Released

2012/010 1180058 3.0 Released

4.2.4 The y+ parameter

If no wall functions are used, the y+ is generally < 1.