IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS

Numerical modelling of velocity profile parameters of the atmospheric

boundary layer simulated in wind tunnels
To cite this article: A Abubaker et al 2018 IOP Conf. Ser.: Mater. Sci. Eng. 393 012025

View the article online for updates and enhancements.

This content was downloaded from IP address 181.214.2.205 on 11/08/2018 at 02:11

KOD 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 393 (2018) 012025 doi:10.1088/1757-899X/393/1/012025
1234567890‘’“”

Numerical modelling of velocity profile parameters of the

atmospheric boundary layer simulated in wind tunnels

A Abubaker1, I Kostić1 and O Kostić1

1
University of Belgrade, Faculty of Mechanical Engineering, Aeronautical Department,
Kraljice Marije 16, 11120 Belgrade 35, Serbia

E-mail: ikostic@mas.bg.ac.rs

Abstract. Experimental and numerical modeling and simulations of the wind influence within
the atmospheric boundary layer are essential tools in optimum building structural design. Each
of these methods, however, has both advantages and disadvantages. In experimental
investigations performed in wind tunnels, reliable results can be obtained, but detailed
information of the wind profile parameters, such as the surface roughness length z0 or the
friction velocity u*, are difficult to determine. Numerical simulations, on other hand, easily
yield any information of the wind velocity profile. However, the reliability of numerical results
strongly depends on the established and adopted computational model. This paper presents the
computational fluid dynamics (CFD) analysis of the atmospheric boundary layer simulated in
subsonic wind tunnels using appropriate types of obstacles, based on the SST k-ω turbulence
model with optimized unstructured mesh and optimum selection of relevant physical model
parameters, performed in Ansys Fluent software. Results have been compared with the
measurements from the Assiut University wind tunnel with maximum velocity of 4 m ⁄s, and
from subsonic wind tunnel at Belgrade University, with maximum air velocity of 45 m ⁄s.
Detailed comparisons for velocity distributions with these experimental results have shown
very good conformity. Also, the three-parameter fitting methods were successfully established
to define surface roughness length z0 and the friction velocity u*. Obtained results have shown
that the established numerical model is able to substitute a remarkable number of expensive
wind tunnel tests hours within the operational investigations of wind influence on the building
structures.

1. Introduction
The wind flow over different terrains has been investigated for many years in full scale measurements,
numerical simulations and wind tunnel experiments. It is of great importance in many engineering
areas, as for example pollution control, wind turbine sitting and large civil engineering constructions.
Despite the use of numerical methods in the estimation of wind flow in different terrains, the results of
those simulations as every other numerical model need to be verified in real physical experience [1].
Wind tunnels are generally classified into four groups according to flow speed. They are: subsonic or
low-speed, transonic, supersonic, and hypersonic. Subsonic or low-speed wind tunnels are the most
common type used in many applications. Atmospheric boundary layer wind tunnels (ABLWTs) are
usually of the subsonic or low-speed type. Transonic wind tunnels are common in the aircraft industry
since. Supersonic wind tunnels can be used to investigate the behavior of jet engines and military
aircrafts. Hypersonic wind tunnels find their applications in rockets and space vehicles. A further way to
categorize low speed wind tunnels is by dividing them into open loop or closed loop wind tunnels [2].
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
KOD 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 393 (2018) 012025 doi:10.1088/1757-899X/393/1/012025
1234567890‘’“”

The computational fluid dynamics (CFD) is a tool which has increasingly been used to study a wide
variety of processes in the atmospheric boundary layer (ABL). There are several numerical models
that can be employed in the simulation of the ABL due to their availability of appropriate boundary
conditions and meteorological data [3, 4]. One of them is a shear stress transport (SST) k-ω model
which employs the k-ω model near the surface, and k- model in free flow. Good performance of the
SST k-ω model for ABL flow around blunt bodies has been shown, for example, in [5, 6].
This paper presents the computational fluid dynamics (CFD) analysis of the atmospheric boundary
layer, simulated in wind tunnels. CFD results have been compared with the available experimental
data from two subsonic wind tunnels, the Assiut University wind tunnel with maximum velocity of 4
m/s [7], and the Belgrade University (Faculty of Mechanical Engineering) wind tunnel, with
maximum air velocity of 45 m/s [8]. Experimental data and CFD analyses have been used to estimate
the mean velocity profiles in wind tunnel test sections for different flow speeds passed a variety of
different obstacles, roughness elements and other devices, in order to validate the applied
methodology for proper simulation of complex terrain influence on the ABL.
Also, mean velocity profiles in the surface layer can be characterized by determining the zero-plane
displacement d, the roughness length z0 for the given roughness and boundary layer height. Accurate
evaluation of these parameters requires knowledge of the shear velocity u*, which may be measured
directly using a skin-friction balance, inferred from Reynolds stress measurements, deduced using the
momentum integral equation, or evaluated indirectly from the mean velocity profile [9, 10, 11, 12].
The indirect procedures are less time consuming, and they have been used for the analyses presented
in this paper.

2. Wind velocity profiles in atmospheric boundary layer

There are two main methods for describing the mean horizontal wind profile in the ABL [13], the
logarithmic, or log law and the power law (see Figure 1).

Figure 1. Log-law and power-law region

in mean wind speed profile.

2.1. The log law model

According to the log law model, which is applied for inner ABL domain, the mean wind speed U(z) at
height z above the ground can be mathematically expressed as:
u z-d
U ( z )= * ln( ) (1)
k z0
It can be rearranged as:

2
KOD 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 393 (2018) 012025 doi:10.1088/1757-899X/393/1/012025
1234567890‘’“”

k U (z )
z=z 0 exp ( )+d (2)
u*

where u* is shear velocity, k is Von Karman’s constant whose typical value is 0.4, d is zero-plane
displacement which is the depth of still air trapped among the roughness elements, and z0 is roughness
length which represents the size of the eddies produced from the wind moving over a rough surface.
Table 1 gives the appropriate value of roughness length, for various types of terrain types (adapted
from the Australian Standard for Wind Loads, AS1170.2, 1989) [14].

Table 1. Terrain types, roughness length

Terrain types roughness length (m)
Very flat terrain (snow, desert) 0.001-0.005
Open terrain (grassland, few trees) 0.01-0.05
Suburban terrain (buildings 3-5 m) 0.1-0.5
Dense urban (buildings 10-30 m) 1-5

2.2. The power law model

According to the power law, which is applied for outer ABL domain, the mean wind speed U(z) at
height z above the ground can be expressed by equation:
z-d α
U ( z )=U s ( ) (3)
zs

where Us is the mean wind speed at a chosen reference height, zs is the reference ABL height, α is
power index or exponent which changes with surface roughness (terrain type).
Figure 2 shows an example of different power law profiles (and corresponding  values) for
different types of terrain, obtaind in [15].

Figure 2. The power law profiles for the velocity distrubutions in boundary
layer over different terrains.

3. Experimental facilities
Simulation of the ABL in a wind tunnel requires designing and experimenting with the different
passive device configurations to be added at the entrance of the test section in order to create the
desired boundary layer properties upstream of the test object (building, bridge, etc.). In this study, two
wind tunnels have been chosen, and their experiments aimed to simulate the atmospheric boundary
layer within the test section, at quite different flow speeds, have been presented.

3
KOD 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 393 (2018) 012025 doi:10.1088/1757-899X/393/1/012025
1234567890‘’“”

First experiments were carried out in an open-loop, low-speed (up to 4 m/s), atmospheric boundary
layer wind at the Mechanical Engineering Department of the Assiut University, as described in [7].
This wind tunnel had the test section with length of 1.7 m, and had 1×1 m square cross section area.
The simulated ABL was generated along the 3.5 m boundary layer development section, by the use of
three triangular flat spires combined with 710 cubes (roughness elements), as shown in Figure 3.

Figure 3. Photograph of the test section at Figure 4. Photograph of the test section at
the Assiut University wind tunnel. the Belgrade University wind tunnel.

Mean vertical velocity distribution was measured at different heights in the middle of test section,
at the distance 3.6 m from the inlet of the boundary layer development section, and experimental
results have been obtained in the empty wind tunnel, wind tunnel with spires only, and wind tunnel
with the combination of spires and arrays of roughness elements.
Second experiment was carried out in the wind tunnel built at the Belgrade University, Faculty of
Mechanical Engineering (Aeronautical Engineering Department), which is of a closed-loop type. Its
test section is 6 m long, with 2.8×2.2 m octagonal cross section, and maximum speed was up to 60 m/s
with empty test section, at the time when tests were performed (more details described in [8]).
As shown in Figure 4, ABL was simulated by four small wall barriers, seven elliptic spires and
1156 small pyramids (roughness elements) placed on the floor, in the front domain of the test section.
Mean velocity distribution was measured at 3 m distance from the test section inlet, at different
heights from the floor.

4. Calculation procedure
The CFD simulations, presented in this paper, were performed in ANSYS Fluent using 3D steady state,
density-based, RANS approach with the SST k-ω model which was adopted by [16]. Fluid was air, and
its viscosity was described by Sutherland law, using three coefficient method. Second order
discretization schemes were used for the convective and viscous terms of the governing equations.
Numerical convergence was achieved when the solution monitor for mass flow rate through the control
volume outlet showed no change, and remained constant observing significant number of digits.
Due to the vertical symmetry of the flow, for both wind tunnels the half-models for the effuser and
test section were modeled, in order to minimize the number of mesh elements within the control
volume. The unstructured meshes were used for both cases, and attention was paid to appropriately
increase the number of elements on the lower wall of test sections, but still keep the total number of
elements at reasonably low values, with satisfactory mesh quality, see Figures 5 and 6.
Numerical analyses were applied for two cases for Assuit University wind tunnel - in the first case
spires only were mounted on the lower wall of boundary layer development section, and in the second
case both spires and surface roughness elements were used to simulate obstacles which generate
atmospheric boundary layer. In Belgrade University wind tunnel, tests were performed with all elements
included (wall barriers, spires and roughness elements mounted on the lower wall of test section).

4
KOD 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 393 (2018) 012025 doi:10.1088/1757-899X/393/1/012025
1234567890‘’“”

Figure 5. Unstructured surfaces mesh for the Figure 6. Unstructured surfaces mesh for the
calculation of airflow at Assuit University. calculation of airflow at Belgrade University.

Also, the indirect methods used to estimate u*, z0 and d based either on experimental data, or
numerical simulation, can be classified into several ways. One of the approaches uses the generalized
reduced gradient (GRG) non-linear least squares method built in the Microsoft Office Excel Solver
add-in [17], and it calculates the values of the three parameters u*, z0, and d by minimizing the sum of
squared differences between measured and predicted z values on the logarithmic law. The parameter
values are selected from the results that best fit the considered velocity profile (the so-called three
parameter fitting). The fitting procedure is also applied for the power law profiles.

5. Results and discussion

5.1. Comparison of CFD and wind tunnel results

All CFD simulations presented in this paper have been performed at 1:1 scale (i.e. using the actual
wind tunnel model dimensions). The presentation of the CFD and wind tunnel results and their
comparison has been done for three general cases, in order to validate here established computational
method over a wide range of simulated wind speeds.
The first two cases consider the Assuit University wind tunnel tests and their CFD simulations. The
flow velocities were obtained by three different fan speeds of 500, 1000, 1440 rpm during the tests,
and the maximum flow speed achieved was 4 m/s at 1440 rpm.
Figure 7 shows comparisons between the numerically simulated and measured mean velocity
profiles for different fan speeds, with only spires used for ABL generation. The CFD results and the

Figure 7. Mean velocity distribution Figure 8. Mean velocity distribution

comparisons for Assuit University wind comparisons for Assuit University wind
tunnel, case with spires only. tunnel, case of spires and roughness elements.

5
KOD 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 393 (2018) 012025 doi:10.1088/1757-899X/393/1/012025
1234567890‘’“”

wind tunnel results are in fairly good agreement, except in the inner layer of ABL. In actual
experiments, although no roughness elements were applied, the wind tunnel wall had its inherent
surface roughness. Since this value was not published, the CFD calculations were performed assuming
a smooth wall which, at such small speeds, has caused small inevitable differences between the
measured and calculated speeds near the wall surface.
Figure 8 shows measured mean velocity profiles for different fan speeds compared with the CFD
results. In this case, both spires and surface roughness elements were used, and comparison shows
good agreements for velocity profiles across the whole relevant height.
The third case considers the Belgrade University wind tunnel, where the experiment and here
presented CFD calculations were performed for flow speed of 45 m/s, about ten times higher than the
maximum flow speed achieved in the Assuit University wind tunnel. Relative velocity profiles U/U0
obtained along relative height z/zs at the middle of test section are used for the presentation of both
experimental and numerical results.
Figure 9 shows comparison between the experimental and the numerically obtained values of the
relative velocity. The CFD results for this case also show good agreements with the experiment, for
the practical engineering purposes.

Figure 9. Relative velocity profile comparisons

for Belgrade University wind tunnel.

Figure 10 shows the spatial contour of eddy viscosity calculated for the Belgrade University wind
tunnel test, the region of high eddy viscosity grows massively large behind the spires and then it fades
out along the test section.

Figure 10. Eddy viscosity in Belgrade University wind tunnel.

6
KOD 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 393 (2018) 012025 doi:10.1088/1757-899X/393/1/012025
1234567890‘’“”

5.2. Determination of wind velocity profile parameters

Indirect method used the validation of the logarithmic law in inner layer, from equation (2) predicted
height z was plotted against measured height z at same value of measured velocity. Values of u*, z0 and
d were determined from the height curve (z) fitting, using the (GRG) non-linear least squares method.
Table 2 shows the estimated values of u*, z0 and d for three mentioned cases of atmospheric
boundary layers simulations. The sum of squared residuals is computed and minimized using the
Solver add-in to obtain the set of parameter values that best describes the experimental data. The
confidence of best-fit values is then visualized and assessed in a generally applicable and easily
comprehensible way.
Table 2. Values of u*, z0 and d
Shear velocity Roughness length Zero-plane
u* (m/s) z0 (m) displacement d (m)
Case 1 (Assuit, spire only) 0.108 0.0000285 0
Case 2 (Assuit, spire and roughness) 0.239 0.0062 0
Case 3 (Belgrade) 4.753 0.0101 0

Figures 11 and 12 show the best fits for cases 1 and 2, for the results from Assuit University wind
tunnel, obtained at 1000 rpm fan speed. Figure 13 shows the fit for case 3, where the predicted height z
values have been compared with the measured height z values taken from the experimental results
from Belgrade University wind tunnel.

Figure 11. Best fit of predicted height z to Figure 12. Best fit of predicted height z to
measured height z for case 1. measured height z for case 2.

Figure 13. Best fit of predicted height z to

measured height z for case 3.

7
KOD 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 393 (2018) 012025 doi:10.1088/1757-899X/393/1/012025
1234567890‘’“”

Assuming an object model in scale 1:100 within the simulated ABL in wind tunnels, comparisons
with Table 1 show that the estimated z0 for case 1 would correspond to very flat area (snow, desert),
value z0 for case 2 would correspond to the urban and suburban area, while z0 for case 3 would
correspond to the ABL generated by the dense urban area, with buildings 10-30 m. On the other hand,
all determined zero-plane displacements are equal to zero, because measurements in wind tunnels
were made dowstream form the simulated surface roughnesses.
Also, the power law representation of the mean velocity profile in the outer ABL layer is possible
and is often used in wind engineering applications. Values of α were determined for power law
expression in equation (3), after estimating the zero-plane displacement d.
Figures 14 and 15 show power law profiles for cases 1 and 2, compared with the measured relative
velocity profiles from Assuit University wind tunnel, at 1000 rpm fan speed. For case 1, the power law
gave computed value of α = 0.12, corresponding to a flat area (also see figure 2), while for case 2 the
computed value α = 0.24 would correspond to the suburban area.

Figure 14. Best fit of the power law to Figure 15. Best fit of the power law to
measured relative velocity profile for case 1. measured relative velocity profile for case 2.

Figure 16 shows power law profile for case 3, i.e. the Belgrade University wind tunnel. Power law
in this case gave the computed value of α = 0.37 which, compared with figure 2, would correspond to
a large city area. These computations show that the obtained log law curves for the inner ABL layer,
and power law dependences for the outer ABL layer, show very good match by categories. Namely,
the first wind tunnel test case has obviously modeled the ABL that would correspond to a very smooth
and flat area, the second would resemble boundary layer generated by a suburban domain,
characterized by mixture of small houses and trees, while the third ABL would be generated by a big
city building structures.

Figure 16. Best fit of the power law to

measured relative velocity profile for case 3.

8
KOD 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 393 (2018) 012025 doi:10.1088/1757-899X/393/1/012025
1234567890‘’“”

Previous section has shown that here presented CFD calculation model can readily be used as a
“virtual” wind tunnel and substitute many hours of expensive real wind tunnel tests, spent in the
preparations of optimum obstacle setups for ABL simulations, for wind velocities in the range from 1
to 45 m/s. In that case, the log and power law fitting methods shown in this section should also be used
to verify if the wind tunnel obstacle setups, defined through the CFD modeling and analyses, would
give the desired type of atmospheric boundary layer velocity profile. This is mandatory for proper
testing of the wind influence on any desired building structure in its natural environment, either if we
use the real, or here presented “virtual” CFD wind tunnel for that purpose.

6. Conclusion
The atmospheric boundary layer (ABL) is the region of air which is greatly influenced by the local
terrain, i.e. the Earth's surface. For structural design purposes, it is important to understand the
atmospheric boundary layer and the flow characteristics associated with it. Although wind tunnels
have been extensively used to simulate such ABL in their test sections, there are many aspects of
information considering the wind profile parameters that still difficult to measure. Therefore,
numerical simulations represent an important additional investigative tool.
In this paper, the CFD calculation model has been established, with an aim to properly simulate
the exact conditions that exist in wind tunnels during the ABL tests. CFD analyses were performed in
order to validate the adopted calculation algorithm and setups for wide speed range and different
terrain types. Experimental results for three test cases, from two different wind tunnels, have been
used for verifications. In the first wind tunnel, small speeds were used, in the range 1 ÷ 4 m/s, while in
the second tunnel, the speed of 45 m/s was applied. Depending on the obstacle types applied in wind
tunnels for ABL simulations, tests in the first tunnel were divided to two cases, while in the second
tunnel one test setup has been investigated.
The velocity profiles obtained by CFD analyses, using here presented calculation model, were
compared with the experimental results, and good agreements have been obtained for operational
engineering purposes. By this, here presented CFD calculation model can successfully be used to
substitute a certain part of the expensive wind tunnel tests, both in the definition of optimum obstacle
setups for ABL simulations, and for further tests of building structure models.
Another important aspect in obstacle setup verifications is proper fitting of the ABL velocity
profile by log law in inner, and power law in outer domain, using indirect approach. The three
experimentally obtained velocity profiles that were used for the validation of CFD calculations, were
first fitted by the log law in the inner domains, and later by power law in outer domains. The two
independent fitting procedures have achieved very good correlation, showing that the first experiment
would correspond to ABL generated over flat and smooth terrain, second to a suburban area ABL, and
third by a large city building structures. Such simple but efficient fitting procedure can be used both in
wind tunnel tests and in CFD analyses in order to verify if the selected obstacle setup gives the desired
type of ABL velocity profile, corresponding to the actual environment that should be simulated.

References
[1] Zarraonandia G 2010 Influence on Wind Shear and Turbulence in Flow Over, University of
Norwegian, MSc Thesis
[2] Lindgren B and Johansson A 2002 Design and Evaluation of a low speed Wind Tunnel with
Expanding Corners, Technical report TRITA-MEK 14
[3] Stangroom P 2004 CFD Modelling of Wind Flow Over Terrain, University of Nottingham, PhD
Thesis
[4] Yassen Y E and Abdelhamed A S 2015 CFD Modeling of the Atmospheric Boundary Layer in
Short Test Section Wind Tunnel, American Journal of Aerospace Engineering 2(1) 38-46
[5] Yang W, Quan Y, Jin, X, Tamura Y and Gu M 2008 Influences of Equilibrium Atmosphere
Boundary Layer and Turbulence Parameter on Wind Loads of Low-rise Buildings, Journal

9
KOD 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 393 (2018) 012025 doi:10.1088/1757-899X/393/1/012025
1234567890‘’“”

of Wind Engineering and Industrial Aerodynamics 96(10) 2080-2092

[6] Karim M M, Rahman M M and Alim M A 2011 Performance of SST k-ω Turbulence Model for
Computation of Viscous Drag of Axisymmetric Underwater Bodies, IJE Transactions B:
Applications 24(2)139-146
[7] Al-Nehari H, Abdel-Rahman A K, Nassib A and Shafey H M 2010 Design and Construction of
A Wind Tunnel for Environmental Flow Studies, Journal of Engineering Sciences, Assiut
University 38(1) 177-193
[8] Stefanović Z and Pešić S 1992 Simulation of Atmospheric Boundary Layer in Laboratory
Conditions, International Symposium on Contemporary Problems in Fluid Mechanics,
Belgrade, Serbia
[9] Wieringa J 1993 Representative Roughness Parameters for Homogeneous Terrain, Journal of
Boundary Layer Meteorology 63 323-363
[10] Zaki A S, Hagishima A, Tanimoto J, Mohammad F A and Razak A A 2014 Estimation of
Aerodynamic Parameters of Urban Building Arrays Using Wind Tunnel Measurements,
Journal of Engineering Science and Technology 9(2) 176-190
[11] Farell C and Iyengar S K A 1999 Experiments on the Wind Tunnel Simulation of Atmospheric
Boundary, Journal of Wind Engineering and Industrial Aerodynamics 79 11-35
[12] Farell C and Iyengar S K A 2001 Experimental Issues in Atmospheric Boundary Layer Simulations:
Roughness Length and Integral Length Scale Determination, Journal of Wind Engineering and
Industrial Aerodynamics 789 1059-1080
[13] Dutta R S 2004 Boundary Layer Characteristics Just Above Sub-Urban Roughness, University
of Texas Tech, MSc Thesis
[14] Holmes D J 2001 Wind Loading of Structures, Spon Press
[15] Ikhwan M 2005 Investigation of Flow and Pressure Characteristics around Pyramidal Buildings,
University of Karlsruhe, PhD Thesis
[16] ANSYS FLUENT 16.0, Theory Guide 2015, ANSYS Inc
[17] Kemmer G and Keller S 2010 Nonlinear Least- Squares Data Fitting in Excel Spreadsheets,
Article in Nature Protocol 5(2) 267-281

10