Vous êtes sur la page 1sur 12

Modelling of Airflow through Wire Mesh Security Screens

Mitchell Cohen1

University of New South Wales at the Australian Defence Force Academy

This study will look to quantify the effect of wire mesh security screens geometry on
airflow. Wire mesh screens are a woven structure, generally used to allow fluid to flow
through while preventing other objects from passing. These are commonly used in the
security industry to prevent against human intrusion. The ventilation aspect of security
screens is an important component in the continuing study to improve security screen
design. By creating a Computational Fluid Dynamic model, validated by literature, these
effects were quantified. It was found that there was a correlation between a length scale
in terms of diameter and spacing with the pressure drop across the screen, measured in
terms of a pressure drop coefficient. When simulated using low velocities there becomes
a significant change in this correlation indicating a larger Reynolds number effect at low
velocities.

Contents
Nomenclature........................................................................................................................................... 2
I. Introduction .......................................................................................................................................... 2
II. Background & Motivation .................................................................................................................. 2
A. Motivation ....................................................................................................................................... 2
B. Standards Compliance ..................................................................................................................... 2
C. Computational Fluid Dynamics ...................................................................................................... 2
III. Theory ................................................................................................................................................ 3
A. Previous Correlation ................................................................................................................. 3
1. Theoretical ................................................................................................................................ 3
2. Experimental ............................................................................................................................. 4
B. Porosity Calculations .................................................................................................................... 4
IV. Computational Approach .................................................................................................................. 4
A. Outline ............................................................................................................................................. 4
B. Model .............................................................................................................................................. 4
C. Meshing Process .............................................................................................................................. 5
D. Finite Volume Method .................................................................................................................... 5
1. Turbulence Models ................................................................................................................... 5
2. Solver Schemes ......................................................................................................................... 6
E. Boundary Conditions .................................................................................................................... 6
F. Verification & Validation ............................................................................................................. 7
3. Grid Independence .................................................................................................................... 7
4. Verification of Results .............................................................................................................. 7
V. Discussion & Results .......................................................................................................................... 8
VI. Conclusions ..................................................................................................................................... 10
VII. Recommendations .......................................................................................................................... 11
Acknowledgments ................................................................................................................................. 11
References ............................................................................................................................................. 11

1
PLTOFF, School of Engineering & Information Technology. ZEIT4500

1
Final Project Report 2015, UNSW Canberra at ADFA
Nomenclature
𝐾 = Pressure Drop Coefficient [-] 𝜌 = Density [kg/m3]
𝐺 = Screen Porosity Function [-] 𝑈 = Free Stream Velocity [m/s]
𝛼 = Porosity [-] d = Diameter [mm]
!
𝑅𝑒! = Reynolds number in terms of diameter [-] = Spacing between Wires [mm]
!
𝑅𝑒!! = Reynolds number in terms of hydraulic diameter I = Turbulence Intensity [-]
P = Static pressure [Pa] 𝐷! = Hydraulic Diameter [m]
a = Length [m] b = Width [m]

I. Introduction
Insect screen doors and windows are a common component in many Australian homes and businesses
to prevent insects from entering while still allowing ventilation and visibility out. Due to these screens weak
construction, they do not provide much resistance from other intruders. The rapid demand for a more secure
alternative has been the catalyst for the development of security screens [1].
Security screen doors and window grilles have a primary intent of preventing human intrusion into both
residential and commercial premises [2]. This has led to the development of a multitude of designs from steel
grate doors combined with mesh fly screens to the woven steel mesh structure this report concerns itself with.
The Australian Standard for Security screen doors and window grilles [1] defines these different designs into
three security screen types. These types are based on the size of object able to be passed through the infill
material and will be discussed in a later section [1].
The security screens considered here are made of a fine metallic infill (usually a 0.8mm diameter
woven mesh steel [2]) fitted to a metallic frame. This provides visibility and airflow while preventing things as
small as insects from entering. It is hypothesised that the obstruction of airflow will depend on the geometrical
characteristics of the screen, such as the diameter and spacing of the strands.

II. Background & Motivation

A. Motivation
The improvement in design and impact resistance of security screens is a primary focus of a research
collaboration program between the University of New South Wales at the Australian Defence Force Academy
and a commercial manufacturer. One of the considerations in the screen design is, how to best maximise the
airflow without compromising safety and security. It is therefore important to determine the airflow
characteristics of security screens with different mesh geometries to contribute to the overall improvement of
design. Since measuring airflow capacities experimentally can be quite time consuming and expensive, creating
a numerical model is an efficient alternative available.

B. Standards Compliance
Australian standard 5039-2008, is concerned with all types of security screen doors and window grilles,
from the way they are attach, whether removable or sliding all the way to the size of objects able to pass through
[1]. The standard defines three specific types of security screen:
Type I: “An infill material that has small enough openings to generally prevent an arm from being
passed through”
Type II: “An infill material that has openings of a size that may allow an arm to be passed through but
which prevents bodily entry”
Type III: “An infill material that has openings small enough to generally prevent insects passing
through” [1]
Therefore this report is concerned with type III screens. While the standard has many performance test such
as dynamic testing of the screen strength, as this study is only concerned with the airflow component of the
design there is only one relevant requirement the standard defines. This is the maximum size of any hole the
geometry of the mesh can leave. To comply with the standard the infill must prevent a probe of 3±0.1mm  
diameter  from  passing  through.  

C. Computational Fluid Dynamics


Computational Fluid Dynamics (CFD) is a widely used simulation tool, which allows physical fluid-flow
and heat transfer scenarios to be modelled. The lack of analytical solutions to the Navier-Stokes equations (the

2
Final Project Report 2015, UNSW Canberra at ADFA
differential equations that govern Newtonian fluid flow) and the arrival of high-speed computers have led to the
development of CFD as a way to obtain approximate numerical solutions to these physical scenarios [3]. CFD
provides the advantages of savings in time and cost, allowing variables to be changed and scenarios to be re-
simulated [4]. In saying this it is limited by a wide range of assumptions and approximations used when solving
the simultaneous equations underlying the model. This in turn means verification and validation of the model
must be conducted from physical test before further analysis can be continued [5]. This report will not recreate
any of the governing equations due to their excessive coverage elsewhere and constraints on report length.
The CFD process involves the substitution of partial differential equations, with what’s known as discretised
algebraic equations. These algebraic equations are developed to approximate the partial differential equations
from the Navier-Stokes equations. This allows the computer to simultaneously solve the equations at discrete
points in the flow and through interpolation a continuous solution is created [3]. Discretisation is therefore a
crucial component in the CFD process and three main techniques are used: finite difference method, finite
element method and boundary element method [4, 3]. The finite difference method splits the flow into grid
points and with the discrete values at these points used to approximate values between these points. The
boundary element method breaks the boundary into discrete segments, which have appropriate singularities
distributed among them. These singularities such as sources, sinks, etc. are measured and used to calculate
values within the flow field. Lastly the finite element method, dissects the flow into small fluid elements. The
conservation equations are than solved for each element and then combined to solve the entire flow [3].
Inherent error in the models is caused by these methods of approximating the governing equations. By
reducing the element sizes through the meshing process these errors get smaller and smaller for each point,
overall improving the accuracy of the model. There is a point where the increase in accuracy will be negligible
compared to the increase in time to simulate the model. Therefore to ensure the important points of the flow are
most accurately modelled, density based meshing is used to increase the accuracy around certain points of
interest in the flow [4].

III. Theory
The pressure drop along a streamline can be calculated by Bernoulli’s equation if, the downstream and
upstream conditions are known. This is assuming that it is inviscid and incompressible flow, an ideal case taking
all viscous losses to be negligible. This theory is the basis for how the pressure drop coefficient k will be
calculated [6].
∆𝑃 = 𝐾0.5𝜌𝑈 !
(1)
Equation 1 shows that the pressure drop is equal to half the density (𝜌) multiplied by the velocity squared.
This is all than multiplied by the pressure drop coefficient (k) to take into account effect of the screen on the
pressure drop. This method makes the assumption that the velocity is constant up and downstream of the screen
and lets k be a measure of the screens effect on the pressure drop [6]. The pressure drop coefficient will be used
by this study to quantify the screens effect on the flow.

A. Previous Correlation
1. Theoretical
In 1967 Pinker and Herbert released a paper concluding a correlation for a pressure drop coefficient
regarding flow through woven screens, could be broken up into two independent functions. One would take into
account the geometry of the screen in terms of porosity and the other would consider the Reynolds number
effect in terms of wire diameter [7].In a study conducted Brundrett in 1993, a rough correlation was found for
the pressure drop in terms of a drop coefficient for incompressible fluid through woven wire and cloth screens
[8]. This correlation was validated against four different studies at low Reynolds numbers and had small errors
no larger than 10%. All of these studies and most of those after who have cited this correlation have been
looking at insect screens involved in agriculture use, such as green houses [9].
𝐾! = 𝑐𝑜𝑠 ! 𝜃  ×𝐺 𝛼 ×𝑓 𝑅𝑒! ×𝑐𝑜𝑠𝜃
(2)
Equation 2 shows Brundretts correlation in terms of the angle the flow approaches (𝜃), the porosity function
(𝐺) and a function, 𝑓 based on Reynolds Number in terms of wire diameter [8]. As we are looking at a simple
model, the flow will be acting perpendicular to the screen and therefore all cosine terms will be equal to one.
The porosity function is an area of study yielding many results. Pinker and Herbert look in depth at four of the
equations used to identify this parameter and conclude that equation 3 gives the best correlation to experimental
data [7]. This is further backed up by Brundrett when deriving his correlation shown by equation 1.
1 − 𝛼!
𝐺 𝛼 =  
𝛼!
(3)

3
Final Project Report 2015, UNSW Canberra at ADFA
𝜎! 7.125 0.88
𝑓 𝑅𝑒! = × (𝐼) + (𝐼𝐼) + 0.055log  (𝑅𝑒! 𝑐𝑜𝑠𝜃)(𝐼𝐼𝐼)
𝜎!" 𝑅𝑒! 𝑐𝑜𝑠𝜃 log  (𝑅𝑒! 𝑐𝑜𝑠𝜃 + 1.25)
(4)
Equation 4 is the function relating the pressure drop coefficient to Reynolds number. It is broken into three
components dominant for different Reynolds number ranges. Term (I) is comparable to modelling drag of
cylinders and friction losses in pipes. It is significant for Reynolds numbers less than one. Term (II) is coined a
blending function dominant for Reynolds numbers up to 100. Term (III) creates a nearly constant value for the
entire function when Reynolds numbers are considered high or in excess 200. The porosity and Reynolds
number equations combine in equation 2 to give a theoretical model for pressure drop. This takes into account
the two physical phenomena that cause the pressure drop, separation referred to as the Reynolds number effect
and the increase in velocity caused by change in area due to the screens geometry.
2. Experimental
Back to the future, in 1999 the Lawrence Berkeley National Laboratory conducted an experimental study on
the airflow through woven stainless steel mesh [10]. This paper contained the experimental data this project was
intended to use for validation. The study found a correlation for the screen pressure using regression analysis in
terms of velocity (ft/s) and free hole area or porosity.
𝑈!.!
𝑃=
42.8×10!.!"!
(5)
What this study actually meant by screen pressure is unknown and the values the correlation calculated were
insignificant in size and further were not able to be related to a pressure drop. Unfortunately by the time this
data was received it was too late to conduct our own experimental testing and only trends from literature can be
used to verify any simulated data this studies model produces.

B. Porosity Calculations
Most common equations to calculate porosity of a woven screen
take an orthogonal approach. This assumes that the free area is a
rectangle. This underestimates the free area of the woven geometry as
demonstrated by the shaded area in figure 1. Pinker & Herbert have
developed what they call a sinusoidal porosity which takes into account
this underestimated area, caused by the woven nature of the geometry [7].
2 𝜋
𝛼 = 1 − ! 1 + 2𝑁 ! − cos !! 𝑁 + 3𝑁 1 − 𝑁 !
𝜋 2
(6)
Where,
𝜋𝑚𝑑
𝑁=
2
(7)
This is regarded as the way to calculate “true porosity” [7] and is used
to calculate the porosity of the geometry developed in this study. It is
important to note that the porosity is dependent on the spacing between
wires and the diameter of those wires. These are the parameters which Figure 1 Porosity Estimate [7]
this study considers important in defining the geometry of the screens.

IV. Computational Approach

A. Outline
To create this model an array of software will be used all combined through the ANSYS Workbench.
ANSYS Mechanical APLD and ANSYS ICEM CFD will be used to both build and mesh the complex
geometry. Once the geometry and meshing has been completed the commercial finite volume solver Fluent 15.0
[11] will be used to model the flow through the created geometry. The ANSYS CFD-Post software will be used
to analyses and compare results.

B. Model
The basic model was created after receiving measurement of the standard security screen available. These
measurements were used to plot a sine curve to model the geometry. Once the process of the initial model was
created in Mechanical APLD, MATLAB was used to create a function that would plot the new coordinates
when the wire diameter or spacing was changed. It would than output a command script to create the geometry
in Mechanical APLD for the new dimensions. This allowed the changing of wire diameter and spacing to be
conducted in a timely manner when building the various models. The ANSYS Geometry modeller was then

4
Final Project Report 2015, UNSW Canberra at ADFA
used to import the screen geometry into a wind tunnel model. The geometry was all modelled using quarter
symmetry to cut down on computational time.

C. Meshing Process
ANSYS provides two main meshing software’s, its basic meshing software that strongly limits user control
and its advance software ICEM CFD. This software allows CAD files to be imported, edited and simple or
complex geometry added all within the one software. It provides a mix of meshing techniques such has
tetrahedral and hexahedral meshing and an ability to combine into a hybrid, this allows for complex geometry to
be model efficiently. In regards to meshing it is initially important to understand Eulerian meshing as the
process commonly used for fluids modelling [12]. This mesh is basically a fixed grid that allows the continuum
to move through. This allows for a large distortion in the flow to be model compared to the alternative
Lagrangian meshing, however Eulerian is limited in its precision of the flow solution [12]. In Lagrangian
meshing the nodes of the mesh move with the material and is common in the modelling of structures. It is
accurate in the modelling of large distortions in the flow however is computationally limited.
Initially it was idealised to use a large control volume with a combination of density based tetrahedral and
prism meshing around the complex geometry that would reduce in density as you moved further away, changing
into hexa-core for the large fluid body. The density meshing feature of this software allows for the accuracy of
the model to be increased around points in the flow of significant interest or complex flow characteristics. This
hybrid method is intended to reduce the number of elements and as such reduce computational time.
Unfortunately the computational resources limited the ability to use a large control volume. Therefore only a
small region up and down stream of the screen was modelled. Due to this the entire control volume was meshed
using tetrahedral meshing [13].
This software also allows for extensive editing to the mesh, all the way down to editing single elements if
the user requires. This allows for meshing of complex geometries; as the quality of the mesh can be alter to such
a fine details. The smoothing technique was used to improve the mesh quality in all models up to the required
standard for Fluent to accept the mesh [14].

D. Finite Volume Method


The finite volume method is the most common CFD technique used today. It is an integral method used by
most commercial solvers and has been thoroughly validated as a general purpose CFD process [14]. There are
three steps that are the basis to the method, firstly the formal integration of the Navier-Stokes equation across
the finite control volumes that make up the flow domain. Second is the discretisation processes involving the
replacement of the integral equations solving for the flow characteristics, with a system of algebraic equations.
Lastly this system is solved via a method of iteration [5]. The step that sets this method apart from others is the
integration of the control volume. The reasons for this are because of the equations that are obtained from this
step, describing the conservation of relevant characteristics for each of the finite control volume elements. This
provides what can be considered as a clear relationship between the numerical systems and the physical
conservation principles [5].
1. Turbulence Models
There is a large array of turbulence models to choose from in CFD alone, specifically looking at Fluent as it is
the software being used in this study, and there is still a long list of viable schemes available. A turbulence
model is a computational procedure that allows a wide variety of flow scenarios to be solved. In most cases it is
not important for the details and characteristics of the turbulence itself but rather what it contributes as an effect
on the mean flow [5]. Two different turbulence models were used for this project, the simple two equation k-ω
model combined with a Shear-Stress Transport (SST) model and a three equation k-kl-ω Transition model.
These were selected based on both complexity and relation to this projects scenario.
The SST k-ω model was developed to combine the robust and accurate k-ω model for the near-wall region of
the flow with the freestream independence of the k-ε model. By combining these two models the complexities in
the flow that change from the inner region of the boundary layer to the higher Reynolds number flow of the
freestream can be blended. This allows for the peak in turbulence kinetic energy near the wall to be modelled
without compromising the accuracy in the freestream [15].
The k-kl-ω Transition model is mainly used where the development and transition onset of a boundary layer
is important. It calculates the transition of boundary layer from laminar to turbulent taking into account;
turbulent kinetic energy, laminar kinetic energy and an inverse turbulent time scale. It was therefore selected as
the transition and its effects around the complex geometry of this model are believed to be of importance [15].
While more complex models and even user specific models could have been created, for both time of
development and the computational power available for solving these models, it was decided that these two and
three equation models would suffice.

5
Final Project Report 2015, UNSW Canberra at ADFA
2. Solver Schemes
The first selection available for the solvers is whether it is a pressure or density-based solver. As we are not
interested in density based problems or compressible flow, our focus remains on the pressure-based solver. The
first choice when using a pressure-based solver is the method in which the pressure and velocity will be
coupled. There are four segregated methods available and a fractional step method (FSM) [14]. For simple
laminar flows with no extra models active the SIMPLE or SIMPLEC methods should be used as they are time
saving methods. For transient flow calculations it is recommended to use a PISO algorithm. One of the limits of
the SIMPLE methods is that new velocities and fluxes do not satisfy the momentum balance once the pressure
correction equation has been solved. In light of this the PISO method solves to additional correction equations to
improve the solution [5]. The most comprehensive method is the coupled scheme. It has greatly improved
performance than the segregated schemes. This is due to the segregated schemes solving the conservation of
momentum equations and the pressure correction equation separately, whereas the coupled scheme solves the
momentum equations together with the pressure-based continuity equations [15].
The discretization scheme is the next important choice and for Fluent there are three choices in regards to the
gradient scheme, they are: Green-Gauss Cell Based, Green-Gauss Node Based or Least Squares Cell Based. The
Green-Gauss approach computes the gradients at the center of the cell. The node-based method is more accurate
than the cell-based method since it takes into account the values of the neighboring cells. The disadvantage to
this is the increase in computational expansiveness that comes with the increase in accuracy. The Least Squares
method uses linear approximation to calculate the gradient. On irregular unstructured meshes, it has been
observed to have comparable accuracy to the node-based method. It is however much less expansive [14].
There is also available the choice of pressure interpolation scheme. There is a linear and standard scheme
that is the simplest of the schemes and not accurate when the flow becomes complex. A second order scheme
reconstructs face pressure using a central differencing equation. There is a Body Force Weighted scheme that
solves the pressure by assuming a normal gradient of difference from the pressure and body forces constants.
Lastly there is a PRESTO! Scheme, which main feature is its availability on all mesh types. It has comparable
accuracy to the other schemes but is more time consuming to use [15].
The solver also has a pseudo-transient relaxation setting, which can be used for steady flow to allow the
balance of momentum to remain accurate [16]. In terms of unsteady flow the transient time step will need to be
enabled.
Once the discretization schemes for the gradient and pressure are defined, the accuracy of the governing
equations is selected. For a problem such as this one where the flow is not aligned with the mesh, second-order
upwind discretization is selected for each of the governing equations [15]. Table 1 list the solver methods
selected for this study.
Solver Pressure Based
Coupling Scheme Coupled
Discretization Gradient Scheme Least Squares Cell Based
Discretization Pressure Scheme Second Order
Governing Equation Schemes Second Order Upwind
Table 1 Solution Methods

E. Boundary Conditions
The boundary conditions had to be determined very
specifically due to the significant effect on the
simulations, and the calculated results. Due to the
solver being pressure and velocity coupled, these
parameters were the first input of significance. The
pressure was defined only at the outlet of the control
volume, being one atmosphere while the velocity was
defined only at the inlet using a range of velocities
from 2.5m/s-20m/s. For models of changing geometry
a constant inlet velocity of 10m/s was used. Two out of
the six control volume surfaces were defined as
symmetry to represent the quarter symmetry being
modelled. The working fluid was defined as air at sea
level and the wire screen defined as a solid, not
changing in dimension when subjected to force. The
turbulence models boundary conditions were one of Figure 2 Turbulence Intensity Experimental Results [17]
vague description and very little direction in other
literature. The inputs for these conditions were therefore selected to be the hydrodynamic diameter, the

6
Final Project Report 2015, UNSW Canberra at ADFA
turbulence intensity and laminar kinetic energy. The hydraulic diameter was calculated using equation 8 [6]
while the laminar kinetic energy was selected from the Fluent library [15].
2𝑎𝑏
𝐷! =  
𝑎+𝑏
(8)
Where a and b are the length and width defining the cross sectional area of the duct. In regards to turbulence
intensity a previous study measuring the characteristics of flow up and downstream of a wire mesh screen
measured the turbulence intensity and is plotted in figure 2 [17]. The Fluent user guide provides equation 9 to
calculate this intensity and comparing these two sources along with a parametric study of our own have
determined the turbulence intensity to be 4% for all models.
𝐼 = 0.16(𝑅𝑒!! )!!/!
(9)
The Equation shows that the turbulence intensity (I) is a function of Reynolds number in terms of hydraulic
diameter. Table 2 outlines the boundary conditions defined and any constants used throughout the models.
Condition Inlet Outlet
Pressure Undefined 101325 Pa
Velocity Defined Undefined
Turbulence Intensity 4% 4%
Hydraulic Diameter Defined Defined
Laminar Kinetic Energy 1e-06 m2/s2 1e-06 m2/s2
Table 2 Boundary conditions

F. Verification & Validation


To verify the CFD model one must compare results to exact analytical solutions or to previously verified
models [18]. This project intended to also verify the model against experimental results, which as mentioned
early was no longer viable given the form the results were received in. The model will be validated by doing a
grid independence study to determine the models
dependence on mesh size [19].
3. Grid Independence
The grid independence study was conducted on each
different geometrical model using three different
densities, increasing the amount of elements in each
model by an approximate multiple of 1.42. As the
element number increased the computing time
increased exponentially causing the fine mesh to take
weeks to compute. Figure 2 shows a graph plotting
the results of the model with the largest percentage
error. This model was of wire diameter 0.9mm and a
wire spacing of 2.5mm. Its largest error was
approximately 5% between the coarse and medium
densities. 5% was the maximum error this study Figure 3 Grid Independence for 0.9mm x2.5mm Model
would allow, constrained by the fact that the grid was
approaching independence. This approach towards
Mesh Density Coarse Medium Fine
independence is seen as the number of elements
increases the difference between the results reduces. Pressure Drop (Pa) 122 115 112
In an ideal world, the grid would be refined until no Difference (Pa) 6 3 N/A
difference was seen. This is a diminishing returns Error 4.918% 2.609% N/A
situation and due to computational availability was Table 3 Grid Independence for 0.9mm x 2.5mm Model
unable to be achieved in this project [20]. The results
seen by this study show that the grid refinement is not high enough for independence but it is approaching it and
due to time constraints we can be confident that the error due to grid dependence is small.
4. Verification of Results
To verify the models, Brundetts method has been used to find a pressure drop coefficient explained in the
theory section. This analytical method is compared to the simulated results in figures 4 &5. The simulated
results are broken up into the parameters that have been changed for each test e.g. the purple diamond represents
the results from changing the spacing between the wires.

7
Final Project Report 2015, UNSW Canberra at ADFA
Figure 4 Analytical &Simulated Pressure Figure 5 Analytical & Simulated Drop coefficient vs
drop vs Re Porosity

When pressure drop is plotted against Reynolds number in figure 4 it can be observed that the simulated
solutions fits the analytical one at low Reynolds numbers but diverges along with the increase in Reynolds
number. This sort of divergence is seen again in figure 5 when the drop coefficient is plotted against porosity
with the results diverging as the porosity decreases. The overall
trend of the results matches as seen in figure 4 as the results are
scattered in a similar matter for both the simulated results and the
analytical. As stated when first looking at the analytical model, it
seems a crude approximation with the only extensive verification at
lower Reynolds numbers. Another cause for this divergence could
be due to the small cross-sectional area this study has modeled.
This would mean that the pressure loss due to the no-slip condition
against the wall would have a larger effect on the average pressure
loss compared to that of a large cross-sectional area. This is
observed to have an effect on the pressure drop when the pressures
along streamlines at different distances from the wall are plotted in
figure 6. The streamline closet to the wall has the largest pressure Figure 6 Pressures for 0.9mm x 2.5mm
drop, therefore a small cross-sectional are will have a larger impact Model
on the overall pressure and is expected to be the cause of this divergence. Unfortunately without experimental
results to confirm this study is unable to fully verify the models, but can conclude that the trends are in line with
what is expected from analytical model and similar experimental studies [8, 7].

V. Discussion & Results


While the model has been validated and verified in terms of pressure and velocity results, the physical
representation of the flow has not been observed. In figure 7 we first look at the velocity contour for a XZ plane
half way along the y-axis. It is important to note that the right hand edge is a wall while the left hand edge is a
surface of symmetry, the flow is moving from top to bottom. It can be
observed that a boundary layer has formed along the right hand edge and
grows as the flow travels further through the duct as expected in pipe
flow. The boundary layer is affected just before the screen as the flow
speeds up around the boundary layer in turn reducing its thickness [3]. In
terms of the effect due to the screen, there is an increase in velocity
through the screen that is expected due to it being incompressible flow
and follows the conservation of mass law that as the area decreases the
velocity increases. Therefore downstream of the screen there is a
decrease in velocity as the area increases back to the full duct [3]. A
sanity check to ensure the aerodynamics through the screen is accurate in
a physical sense would be looking for separation of flow just downstream
Figure 7 Velocity Contour 0.9mm x of the screen. This is due to the screen in a simple sense being an array of
2.5mm Model cylinders, therefore it is expect the flow separates on the trailing side of
the cylinders [3]. Due to the interaction between the screens wires it is
hard to display a clear image of this separation but in figure 7 the areas of zero velocity downstream of the
screen is the best example of separation this study could display in an image format. The lower velocity
contours seen further downstream show that there is an interaction effect between the wire causing the
turbulence to have a non-uniform effect further downstream.

8
Final Project Report 2015, UNSW Canberra at ADFA
An example of why this interaction is causing areas of non-uniform
flow can be seen in figure 8.This figure shows the turbulence kinetic
energy of the flow on the same plane as figure 7. As you can see near
the wall the turbulence kinetic energy is lower than in the middle of the
flow where there are areas of higher turbulence kinetic energy. What is
interesting about this is what seem to be multiple areas occurring across
the x-axis. This is in contrast to a continuous region across the x-axis.
These isolated concentrations suggest that there is an effect caused by
the close proximity of the wires to one other. This indicates the flow
downstream acts in a manner not expected of a single cylinder in flow. It
is therefore implying that the pressure drop is not solely due to a pure
Reynolds number effect but additionally caused by the interaction
between the wire strands. After looking at these measurements it is
beneficial to look at the actual flow paths. Figure 9 shows the Figure 8 Turbulence Kinetic Energy
Contour 0.9mm x 2.5mm Model
streamlines of the same model as the previous two figures. It is
interesting to observe the regions of circulation seen throughout the cross-section. The streamlines seem to
interact in a way where they swirl together in the same direction next to another region that swirls in the
opposite direction. This demonstrates the complex interactions occurring when the flow separates after the
screen. This suggest that the vortices formed after separation will
influence the adjacent streamlines affecting the over uniformity of the
flow.
To find a correlation it is important to have it in dimensionless form
to allow ease of use across different models. As mentioned in the theory
section a pressure loss coefficient will be used to quantify the effect of
the screen on the flow. Reynolds number is a very common and useful
characteristic to use when analysing flow and will be used as a variable
to compare. Another important characteristic of the screen is its
geometry. We have already demonstrated the method to calculate the
porosity of the screen and its use in our analytical model, however since
there are many debated methods to correctly calculate the porosity a
more simple solution has been found. Since porosity is just a function of
Figure 9 Streamlines 0.9mmx2.5mm
wire spacing and diameter, a length scale formed out of these two
Model
parameters can be used to quantify the screens geometrical
characteristics. It was decided that the length scale for this study would be in the form of diameter over spacing
(D/S).
The parameters changed for testing were intended to replicate the current existing screens the university has
access to. For these reasons four diameters were simulated (0.8mm, 0.9mm, 1mm and 1.2mm) with a constant
spacing of 2.5mm. The spacing was varied three times for the simple 0.9mm diameter model. The only restraint
place on the spacing is the 3mm test probe that the Australian standard prescribes. Therefore it was decided that
one variable spacing would be the largest possibly spacing being 3mm. To maintain consistency it was then
decided to model the same change in distance, for a smaller spacing being 2mm.
Figures 10 & 11 show the loss coefficient plotted against Reynolds number in terms of diameter and the
length (D/S) for the transitional turbulence model.

Figure 11 Drop Coefficient vs Length Scale


Figure 10 Pressure Loss Coefficient vs Reynolds
Number
When looking at figure 10, it is seen that the pressure drop coefficient when the velocity is changed (for the
same geometry) trends in a different direction to that of the geometry changing. This shows that there is a
correlation with the velocity and the pressure drop due to the screen (not taken into account by Bernoulli’s). The
difference however, between the coefficient values is insignificant at higher Reynolds numbers (~800) and due

9
Final Project Report 2015, UNSW Canberra at ADFA
to insufficient resources and time this relation was not perused since our test range is within the region of
insignificant discrepancy. There is no clear trend however
between the Reynolds number and the pressure drop
coefficient.
Figure 11 shows a strong correlation between the pressure
drop due to the screen and the length scale created to represent
the screens geometrical characteristics. While slight divergence
with the analytical model is seen as the length scale increases, it
supports the previous hypothesis that the major cause of
pressure drop is that of the geometry and its interacting effects
downstream of the screen rather than a Reynolds number effect.
This in turn can explain the larger loss coefficient on the same
screen when the velocity is reduced to low values in the
previous figure 10. Since a correlation was observed between Figure 12 MATLAB Modelled Line
length scale and pressure drop coefficient, MATLAB was used
to find an equation to fit the data and extrapolate it. A non-linear line fit function was used to find the equation.
This process involved approximating the line by transferring the plot to linearity. This was done by taking the
natural log of both sides. The non-linear line fit function than used that approximation to find a more exact
solution. Figure 12 shows the results as well as the trend for the analytical model. The equation for this studies
model gives a resultant equation of:
𝑦 = 0.1014𝑒 !.!"!#!
(10)
Where y is the pressure drop coefficient and x is the length scale.
This correlation is only feasible for velocities from 5m/s and higher.
While this study has unfortunately not been able to compare to
experimental data it seems an adequate trend when comparing to the
analytical solution and gives purpose for further research.
The two turbulence models were compared in figure 13. Using
the same method to find a line of best fit the two turbulence models
were plotted against each other for the range of length scales
simulated. It shows negligible difference between the models
indicating that for the boundary conditions tested, both models were
accurate in accounting of the effect of turbulence has on the flow.
The turbulence effect on the flow has been clearly indicated has
Figure 13 Comparing Turbulence important when previously looking at the physical measurements of
Models the flow. Therefore it is an area that can be further developed as
Fluent offers the ability to implement a user-defined turbulence
model.

VI. Conclusions
A range of screen geometries and incoming flow velocities have been simulated in this report with two
conclusions being found. It was observed that when changing the geometry of the screen there was no
correlation of its effect (quantified by pressure drop coefficient) on the flow with Reynolds number, apart from a
trend of increasing pressure drop with increasing Reynolds number(accounted for by Bernoulli’s). A correlation
could be found however when the same effect is plotted against a length scale of the screens wire diameter on
wire spacing. It was than concluded that the effect of the screen on the flow was largely due to the geometry of
the screen rather than the Reynolds number of the flow. It was further observed that the streamlines of the flow
interacted with each other downstream of the screen, which caused swirling in the flow and regions of large
turbulence, all factors contributing to the pressure drop observed. This meant that the pressure drop was not due
solely to a Reynolds number effect but largely due to the interaction of the flow caused by the geometry.
When looking at the effect due to change in velocity, there was a clear increase in pressure drop with the
increase of velocity. This can be afforded to the velocity squared term in Bernoulli’s equation. When the drop
coefficient was plotted against Reynolds number in terms of velocity, it was seen that for velocities higher than
5m/s that the discrepancy between coefficients was insignificant. At velocities less than 5m/s the coefficient
increased meaning the correlation this study has provided is only valid for a small range of velocities.
Without experimental results to verify this model it is not appropriate to conclude on the overall accuracy of
this model. It can be concluded however that there is a correlation between the geometry of the screen and the
pressure drop caused by the screen. This effect is not caused just by the restriction of area the screen causes but
also the interaction of the separated flow downstream of it.

10
Final Project Report 2015, UNSW Canberra at ADFA
VII. Recommendations
The first recommendation from this study is for further work conducting of experimental tests to verify this
model. This model will have a difference to any experimental data taken, due to the small size of the geometry
modelled. This leads to another recommendation to expand the cross-sectional area of the screen being
modelled, along with the range of parameters controlling the screens geometry, being diameter and spacing. A
more comprehensive look at the effect of velocity is also required to find the correlation it has with the effect of
the screen at lower values.

Acknowledgments
My deepest gratitude goes to my thesis supervisors Dr Krishakumar Skankar and Mr Alan Fien. Their
guidance and motivation has allowed me to reach my projects goal. My constant search to find information has
given me the skill set to avoid ever being lost in a sandstorm. Last of all to my family and friends who
encouraged me to give up sunlight in the quest to complete this project.

References

[1] Committee CS-023, "Security screen doors and security window grilles," Council of Australian Standards,
2008.
[2] CrimSafe Security Systems Pty Ltd, "CrimSafe," [Online]. Available: http://crimsafe.com.au. [Accessed
April 2015].
[3] B. R. Munson, T. H. Okiishi, W. W. Huebsch and A. P. Rothmayer, Fluid Mechanics, 7th Edition ed.,
Wiley, 2012.
[4] Y. Nakayama and R. Boucher, Introduction to Fluid Mechanics, Butterworth-Heineman ed., Elsevier, 2000.
[5] H. K. Versteeg and W. Malalasekera, An Introduction to Computational Fluid Dynamics: The Finite
Volume Method, 2nd Edition ed., Pearson Education Limited , 2007.
[6] American Society of Heating, Refrigerating and Air-Conditioning Engineers, ASHRAE Handbook of
Fundamentals, New York, 1967.
[7] M. H. R.A Pinker, "Pressure Loss Associated with Compressible Flow through Square-Mesh Wire Gauzes,"
Journal of Mechanical Engineering Science, vol. 9, no. 1, pp. 11-24, 1967.
[8] E. Brundrett, "Prediction of Pressure Drop for Incompressible Flow through Screens," Journal of Fluids
Engineering, vol. 115, no. June, pp. 239-242, 1993.
[9] A. S. M. Teitel, "PRESSURE DROP ACROSS INSECT-PROOF SCREENS," Transactions of the ASAE,
vol. 41, no. 6, pp. 1829-1834, 1998.
[10 M. Roberts, "Air Flow Through Woven Stainless Steel Mesh," Lawrence Berkeley National Laboratory,
] 1999.
[11 R. Rahimi and D. Abbaspour, "Determination of pressure drop in wire mesh mist eliminator by CFD,"
] Chemical Engineering and Processing , vol. 47, pp. 1504-1508, 2008.
[12 J. Donea, A. Huerta, J.-P. Ponthot and A. Rodr´ıguez-Ferran, Encyclopedia of Computational Mechanics,
] Volume 1: Fundamentals ed., Wiley, 2004.
[13 ANSYS, "ICEM CFD User Manual," 2014. [Online]. Available:
] https://support.ansys.com/AnsysCustomerPortal/en_us/Knowledge+Resources/Online+Documentation/Prev
ious+Releases/15.0.
[14 ANSYS, "Fluent 15.0 User Guide," 2014. [Online]. Available:
] https://support.ansys.com/AnsysCustomerPortal/en_us/Knowledge+Resources/Online+Documentation/Prev
ious+Releases/15.0.
[15 ANSYS, "Fluent 15.0 Theory Guide," 2014. [Online]. Available:
] https://support.ansys.com/AnsysCustomerPortal/en_us/Knowledge+Resources/Online+Documentation/Prev
ious+Releases/15.0.
[16 C. T. Kelly and D. E. Keyes, "Convergence Analysis Of Pseudo-Transient Continuation," Society for
] Industrial and Applied Mathematics J. Numer, Anal, vol. 35, no. 2, pp. 508-523, April 1998.
[17 L. Frantisek, T. Jan and J. Miroslav, "Measurement of the airflow velocity upstream and downstream a wire
] mesh using constant temperature anemometry," EPJ Web of Conferences, vol. 67, 2014.
[18 D. P. Aeschliman and W. L. Oberkampf, "Experimental Methodology for Computational Fluid Dynamics
] Code Validation," AIAA Journal, vol. 36, no. 5, pp. 733-741, May 1998.

11
Final Project Report 2015, UNSW Canberra at ADFA
[19 American Institute of Aeronautics and Astronautics, Guide for the verification and validation of
] computational fluid dynamics simulations, American Institute of Aeronautics and Astronautics, 2002.
[20 C. J. D. V. W. Mohamed Sukri Mat ALI, "GRID CONVERGENCE STUDY FOR A TWO-
] DIMENSIONAL SIMULATION OF," Seventh International Conference on CFD in the Minerals and
Process Industries, Melbourne, 2009.
[21 J. Tu, G. H. Yeoh and C. Liu, Computational Fluid Dynamics - A Practical Approach, Elsevier, 2008.
]
[22 A. S, A. B and R. A, "Computational Fluid Dynamic Modeling of Pressure Drop through Wire Mesh Screen
] Regenerators," AIP Conference Proceedings , vol. 710, 2004.
[23 G. Doig, T. Barber, A. Neely and D. Myre, "Experimental validation as an integral component of
] computational fluid dynamics research," J. ANZIAM, vol. 51, pp. 265-279, May 2012.
[24 J. D. Anderson Jr, Fundamentals of Aerodynamics, 5th Edition ed., New York: McGraw-Hill, 2011.
]

12
Final Project Report 2015, UNSW Canberra at ADFA

Vous aimerez peut-être aussi