You are on page 1of 18

Resilient Cities and Structures 3 (2024) 66–83

Contents lists available at ScienceDirect

Resilient Cities and Structures


journal homepage: www.elsevier.com/locate/rcns

Full Length Article

A robust protocol to compute wind load coefficients of telecommunication


towers and antennas using numerical simulation for risk and resilience
assessment
Mohanad Khazaali a, Liyang Ma b,c, Keivan Rokneddin d, Matteo Mazzotti e, Paolo Bocchini f,g,∗
a
Research Engineer, Extreme Event Solutions at Verisk, Lafayette City Center, Boston, MA 02111, USA
b
Key Lab of Structures Dynamic Behavior and Control of the Ministry of Education, Harbin Institute of Technology, Harbin 150090, PR China
c
Key Lab of Smart Prevention and Mitigation of Civil Engineering Disasters of the Ministry of Industry and Information Technology, Harbin Institute of Technology,
Harbin 150090, PR China
d
Senior Research Scientist, RenaissanceRe Risk Sciences, USA
e
Research Associate, P.M. Rady Department of Mechanical Engineering, University of Colorado Boulder, 1111 Engineering Dr UCB 427, Boulder, CO 80309, USA
f
Professor, Dept. of Civil and Environmental Engineering, Lehigh University, 117 ATLSS Dr, Bethlehem, PA 18015-4729. USA
g
Director, Catastrophe Modeling Center, Lehigh University, Bethlehem, PA, USA

a r t i c l e i n f o a b s t r a c t

Keywords: An accurate estimation of wind loads on telecommunication towers is crucial for design, as well as for perform-
Generalized CFD protocol ing reliability, resilience, and risk assessments. In particular, drag coefficient and interference factor are the most
Lattice tower significant factors for wind load computations. Wind tunnel tests and computational fluid dynamics (CFD) are
Tower panel
the most appropriate methods to estimate these parameters. While wind tunnel tests are generally preferred in
Microwave antenna
practice, they require dedicated facilities and personnel, and can be expensive if multiple configurations of tower
Wind coefficients
Tower-antenna interaction panels and antennas need to be tested under various wind directions (e.g., fragility curve development for system
resilience analysis). This paper provides a simple, robust, and easily accessible CFD protocol with widespread
applicability, offering a practical solution in situations where wind tunnel testing is not feasible, such as complex
tower configurations or cases where the cost of running experiments for all the tower-antennas configurations
is prohibitively high. Different turbulence models, structural and fluid boundary conditions and mesh types are
tested to provide a streamlined CFD modeling strategy that shows good convergence and balances accuracy,
computational time, and robustness. The protocol is calibrated and validated with experimental studies available
in the literature. To demonstrate the capabilities of the protocol, three lattice tower panels and antennas with
different configurations are analyzed as examples. The protocol successfully estimates the drag and lateral wind
loads and their coefficients under different wind directions. Noticeable differences are observed between the esti-
mated wind loads with this protocol and those computed by a simple linear superposition used in most practical
applications, indicating the importance of tower-antenna interaction. Also, as expected, the wind loads recom-
mended by design codes overestimate the simulated results. More importantly, the telecommunication design
codes inadequately identify the most favorable wind directions that are associated with the lowest wind loads,
while the results of the proposed protocol align with observations from experimental studies. This information
may be used to select the tower orientation before construction. The findings of this study are of importance for
the telecommunication industry, which seeks reliable results with minimal computational efforts. In addition,
it enhances the fragility analysis of telecommunication towers under strong winds, and the portfolio risk and
resilience assessment of telecommunication systems.

1. Introduction antennas) that are essential for transmitting and receiving cellular sig-
nals and for radio and television broadcasting. In particular, lattice steel
Telecommunication towers are the main structures of the communi- towers are the most commonly used structures in the telecommunication
cation system. These towers are used to support nonstructural appurte- industry due to their light weight, and are usually designed in two dif-
nances (e.g., vertical panel antennas, fiber-optic cables, and microwave ferent forms: (1) with a square in-plane cross-section (i.e., four-legged)


Corresponding author.
E-mail address: paolo.bocchini@lehigh.edu (P. Bocchini).

https://doi.org/10.1016/j.rcns.2024.02.001
Received 16 August 2023; Received in revised form 12 January 2024; Accepted 6 February 2024
2772-7416/© 2024 The Author(s). Published by Elsevier B.V. on behalf of College of Civil Engineering, Tongji University. This is an open access article under the
CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

where the four main members (i.e., legs) and the secondary members son to numerical analyses. Nonetheless, very few wind tunnels studies
(i.e., bracings) are generally made with angle steel elements; (2) with a have addressed tower-antenna interaction through the use of the aero-
triangular in-plane cross-section (i.e., three-legged), in which the three dynamic interference factor. The engineering science data unit (ESDU)
main members are made by tubular elements, while the secondary mem- proposed an empirical formula to compute the interference factor as
bers consist of angle or tubular elements. In this study, these two tower a function of the drag coefficients and the solidity ratio of the tower
classes with different bracing configurations (i.e., K and X bracings) are [17]. However, this formula underestimated the interference factor val-
considered. ues [15]. Holmes et al. [15] proposed another modified formula that also
The structural design of telecommunication towers is primarily accounts for the wind attack angle and the relative position of the mi-
driven by wind loads. Strong winds or hurricanes can cause partial dam- crowave antenna to the tower, because they have a significant impact on
age or even outright collapse of such structures and lead to service inter- the airflow patterns. Martín et al. [2] performed a broader study consid-
ruptions in many cases, which significantly affects community resilience ering different positions of the microwave antennas and wind directions;
and normal business operations, as well as the activities of the first re- their results were in agreement with the values obtained by Holmes et al.
sponders. For instance, Tapia-Hernández and Cervantes-Castillo [1] in- [15] in most cases. Martín et al. [2] showed that the interference factors
spected several communication towers constructed in Mexico between are sensitive to tower geometry, member sizes, tower drag coefficients,
1990 and 2016; they indicated that accurate and frequent assessments solidity ratio, and tower-antenna relative position. Thus, the experimen-
of tower safety under wind loads are needed to reduce damage, revenue tal results cannot be generalized and more studies are required to have a
losses and high repair costs. Several studies reported complete tower complete parametric characterization of these factors. In general, wind
collapses or partial damage due to strong winds [2–5]. Hurricane Kat- tunnel tests require dedicated facilities and their results are difficult to
rina in 2005 caused serious damage for nonstructural components such generalize to other configurations because of the very limited number
as microwave antennas and fiber-optic links [6]. of cases considered [2]. Also, wind tunnel tests are expensive to per-
Telecommunication system (i.e., telecommunication towers) risk form, in particular for parametric studies which require testing multi-
and resilience assessments usually rely on fragility curves, which rep- ple tower panels under different wind directions (e.g., development of
resent the relationship between the hazard intensity measure (e.g., sus- fragility curves for system resilience analysis).
tained wind speed for hurricane winds) and the probability of reaching In recent years, the practical adoption of computational fluid dynam-
or exceeding a given damage state [7–9]. The tower-antenna(s) inter- ics (CFD) models has become increasingly popular in many engineering
action affects estimates for wind load coefficients and, consequently, applications, due to the widespread availability of sufficient computa-
tower fragility curves. Ignoring those interactions induces errors in the tional resources and improvements in software packages. CFD acts as
calculated wind load values which, in turn, compromises the accuracy a viable solution in scenarios where the implementation of wind tun-
of portfolio risk analyses and community resilience assessment. More- nel testing is not feasible, such as for intricate tower configurations or
over, these issues are of importance for catastrophe modeling and the instances where the cost of running tests for all tower-antennas con-
insurance sector. In fact, an overestimation of the drag coefficient would figurations in a portfolio is excessively high. In general, wind tunnel
lead to an overestimation of the tower vulnerability and, in turn, of the tests are still preferred over CFD analysis and considered more reliable
expected damage, losses, and business interruption. in the professional community. However, when a full parametric study
Considering tower-antenna interactions has become more relevant is needed for reliability, risk or resilience assessments, the ability of
over time since the number of towers and their diversity in terms of CFD to provide results for a vast array of variations and cases at min-
height, geometry, cross-section and structural design has grown substan- imal marginal costs becomes very appealing. The differences in using
tially. Also, telecommunication tower owners allow service providers to the mentioned three methods (design codes, wind tunnel tests, and CFD
install additional microwave antennas at any height and in any direc- models) to compute wind loads and wind load coefficients are summa-
tion. A study by Martín et al. [2] indicated an increase of 20–26 % in rized in Table 1.
the total drag wind loads for specific wind directions due to microwave This paper proposes a general protocol for the use of a parameter-
antennas mounted on the tower. Accordingly, risk and resilience assess- ized and straightforward CFD model in risk and resilience assessments to
ments would greatly benefit from methodologies that can estimate wind compute accurately the wind load coefficients and total wind loads for
load coefficients for any existing tower-antenna configuration. a telecommunication tower. The proposed protocol estimates the aero-
Wind load coefficients and wind loads for telecommunication tower dynamic coefficients for lattice towers and microwave antennas in any
and antennas can be calculated in different ways. Most practical ap- configuration under different wind directions and then adopts the em-
plications and some design codes (e.g., ANSI/TIA [10]) adopt a simple pirical interference factor formula proposed in [15] to model the tower-
superposition to compute the total wind load, by adding the wind force antenna interaction. The protocol requires only generally available com-
acting on the tower structure (FST ) to that acting on the appurtenances putational resources, balancing accuracy, computational time and ro-
(FA ) and neglecting the tower-antenna interaction. This approach un- bustness. In this study, the CFD analysis was performed with COMSOL
derestimates the total drag wind loads under specific wind directions Multiphysics® [18], but other comparable software packages can be
for which the antenna does not shield any portion of the tower. In other used as well. The proposed CFD model and its parameters were cali-
cases, this approach overestimates the total drag wind loads because it brated based on the experimental study performed by Li et al. [16] and
does not take into account the shielding effects. Also, design codes de- validated using an experimental study conducted by Martín et al. [2].
termine the wind load for the case when the wind direction is normal to Scholars in the field who conducted extensive wind tunnel tests ex-
the tower face (i.e., maximum drag forces) and then use trigonometry or amining the tower-antenna interaction were consulted to have an ex-
wind directional coefficients to consider the effect of the wind direction pert opinion on the presented modeling choices and results to confirm
[10–12]. This approach implicitly assumes that the drag coefficient does their validity [19]. The resulting drag forces are compared to those
not depend on the wind direction (i.e., drag coefficients are based only calculated by a standard code [10], as an additional sanity check for
on the solidity ratio of the normal face). This assumption leads to inaccu- the numerical results and to gain insight on the approximations intro-
rate loads, because in reality the drag coefficients change with the wind duced by the design codes. To show the general applicability of the
direction [2,13-16]. In fact, drag coefficients for lattice towers depend proposed protocol, illustrative examples are presented, involving three
on wind direction, tower geometry, members’ cross-section, and solid- panels extracted from a representative lattice tower with microwave
ity ratio. Structural design codes address all involved uncertainties and antennas subjected to different wind directions, with each panel having
approximations by introducing appropriate conservative assumptions. a different geometry and solidity ratio. The Supplementary Material at-
Wind tunnel tests are arguably the preferred way to compute the tached to this paper includes a guide on how to perform the analyses in
wind load coefficients despite requiring copious resources in compari- COMSOL.

67
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Table 1
Differences in using design codes, wind tunnel tests, and CFD models to compute wind loads.

Features Design codes Wind tunnel tests CFD models

Handles drag coefficient direction No Yes Yes


Considers tower-antenna interaction Majority recommended superposition Yes Yes
Considers tower-antenna relative position No Yes Yes
Considers solidity ratio Yes Yes Yes
Handles wind direction Using wind directionality coefficients and trigonometry Yes Yes
Easy to change structural conditions and repeat analysis No No Yes
Used for design purposes Required Not needed Not needed
Used for reliability, resilience and risk assessments Not preferred Expensive Convenient
Trusted by the professional community N/A Yes Needs calibration and validation
Accuracy in wind load calculation Overestimation of the loads Used as reference∗ Close enough to wind tunnel tests

In reality, field studies should be used as reference, and it is possible that CFD models are able to capture open field conditions better than wind tunnel tests.
Very few datasets are available for comparison.

where 𝑘 is a constant parameter equal to 1.8 for triangular towers and


1.2 for square towers, and b is a parameter that needs to be calibrated.
Holmes et al. [15] calibrated parameter b to be 0.5. 𝜑 is the tower solid-
ity ratio and 𝜃𝑚 is the antenna rotation relative to the normal face of the
tower (measured clock-wise). The parameters Ψ, Ψm , and 𝜃𝑚 are shown
in Fig. 1.

3. Proposed protocol

The proposed protocol to assess the wind load coefficients and the
wind loads is schematically shown in Fig. 2. The input data includes the
Fig. 1. Wind directions and tower to antenna relative position.
properties and parameters needed to create the structural models for
tower, antenna, and the air block, as well as the wind direction and wind
speed parameters. Moreover, the CFD model must specify the turbulence
2. Wind load coefficients for telecommunication tower and
model and mesh properties and adopt a numerical algorithm (solver).
antenna
The CFD analysis provides the wind velocity and pressure fields that can
be integrated over the tower and antenna surfaces to compute the loads.
Tower drag coefficient (CD ), antenna drag coefficient (CDm ), and
The aerodynamic wind load formula yields the wind load coefficients
tower-antenna interaction factor (i.e., interference factor) for different
and Eq. (2) returns the interference factors, which are the main outputs
wind directions are the most critical factors that are needed to accu-
of the analysis. These results can be used to compute the total wind loads
rately compute the total wind loads exerted on the tower. In this study,
in any direction, considering the tower-antenna interaction.
CD and CDm are estimated directly via CFD, considering members’ cross-
section, wind direction, boundary conditions, and variation with tower
3.1. Telecommunication tower, microwave antenna, and air block models
height. For the interference factor, empirical equations are available in
the literature [2,15] and are used in this study. The interference factor
Telecommunication towers come in different types, such as
fa represents the ratio between the incremental drag force produced as
monopole, lattice, and guyed towers. This paper focuses on lattice tow-
a result of attaching antenna(s) to the tower and the drag force of the
ers, but generalizations are straightforward. Since the tower owners
isolated microwave antenna [2]:
(utilities and cell phone carriers) do not publicly release structural de-
𝐹𝑡𝑜𝑤𝑒𝑟+𝑎𝑛𝑡𝑒𝑛𝑛𝑎 (Ψ) − 𝐹𝑡𝑜𝑤𝑒𝑟 (Ψ) tails such as member geometries, this research relied on other resources
fa (Ψ) = ( ) ( ) (1)
1∕2𝜌𝑢2 𝐴Pm Ψm CDm Ψm to set up structural models of the tower for demonstration of the proto-
col. The online database by Cell Reception [21] provides the tower type
where 𝐹𝑡𝑜𝑤𝑒𝑟+𝑎𝑛𝑡𝑒𝑛𝑛𝑎 is the total wind drag force on tower and antenna; and height for all towers in the USA. For a given archetype and height,
𝐹𝑡𝑜𝑤𝑒𝑟 is the wind drag force on the tower alone; 𝜌 is the air density; 𝑢 is the “drawing collection of telecommunication steel tower (DCTST)” col-
the mean wind velocity in the along wind direction; 𝐴Pm is the projected lects details for various towers, which can be used as a proxy for the
area of the microwave antenna in the wind direction; Ψ is the wind specific tower under consideration [22]. In particular, the DCTST col-
direction relative to the main axis of the tower; and Ψm is the wind lection includes information on a group of lattice steel towers that have
direction relative to the main axis of the antenna (Fig. 1). general configurations used in the USA. In this study, DCTST serves as
Carril et al. [20] carried out experimental studies to obtain the inter- a specific illustration, although the protocol can be broadly applied to
ference factor on a square tower with different numbers of microwave various tower configurations. While this approach may not guarantee
antennas. They considered only one configuration of the antennas on accurate results for a specific tower, it is sufficient for regional portfo-
the tower and one wind incidence angle. They recommended an inter- lio analysis, where the structural characteristics of each tower are not
ference factor of 1.0 to be used for a tower panel with a solidity ratio crucial, but their aggregate performance is. Moreover, when studying
equal to 0.2 or less. Holmes et al. [15] showed experimentally that the telecommunication towers, engineers can count on standardization and
interference factor varies significantly with the wind direction and val- repetition of the structural elements and design, which makes towers
ues different than 1.0 are observed in many cases; they proposed an much more similar to each other than, for instance, buildings [7].
empirical formula that fits well the experimental data obtained from Telecommunication towers are formed by assembling several stan-
a single microwave antenna attached to square and triangular lattice dardized panels. Given the similarity among tower panels, it is preferred
towers: to set up the structural model in a parametrized way to enable generat-
[ ( )2 ] { [ ( )]} ing a variety of tower models by adjusting only a few parameters. In par-
fa (Ψ) = exp −𝑘 CD (Ψ) ⋅ 𝜑 ⋅ (1 + 𝑏) + 𝑏 ⋅ cos 2 Ψ − 𝜃𝑚 − 90o (2) ticular, the structural model can be built as a function of the panel width,

68
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Inputs
Problem definition
---------------------------------------
• Tower parameters
• Antenna parameters
• Air block parameters
• Materials CFD results
• Wind speed -----------------------------
• Wind direction parameter CFD analysis Wind velocity fields and
pressure fields on
Analysis parameters
structural surfaces
---------------------------------------
• Turbulence model

Eq. (3)
• Boundary conditions
• Mesh
• Solver

Outputs
Wind load coefficients Wind loads
--------------------------------------- -----------------------------------
• Drag wind coefficients Eq. (4) • Drag wind loads
• Lateral wind coefficients • Lateral wind loads
Eq. (2)

Eq. (7) Compute the total wind loads


considering tower-antenna
Interference factors
interaction

Fig. 2. Main parts of the proposed CFD protocol.

panel height, general topology, and members’ cross-sections. Moreover, choice is that the relative displacements of the elements of a single panel
parameter Ψ is added to describe the rotation of the panel with respect are small, and the resulting effects of fluid-structure interaction are neg-
to the wind direction, which enables investigating the tower response ligible. This conclusion was confirmed by experts in the field [19]. The
to varying angles of wind incidence (Fig. 1). case would be different if the entire tower was examined, because the
It is also preferable to create separate parameterized models for the height of the structure would result in large displacements that neces-
tower panel and the antennas to facilitate the investigation of various sitate a fluid-structure interaction (FSI) analysis. It is well established
configurations, when using the proposed protocol. For instance, if a new and can be shown by FSI studies on flexible structures (e.g., vegetal el-
antenna with different geometry is placed on the tower, there is no need ements, coupled rigid-flexible system in a flowing soap film) that drag
to create a new model of the tower panel, only of the new antenna forces occurring on flexible bodies are different from those occurring on
(which may also be available from another study). The telecommuni- rigid bodies due to deformation and flow patterns [24,25]. Moreover,
cation industry has classified four typical microwave antennas, namely: because of their structural flexibility, the bending of slender structures
microwave antenna without radome, microwave antenna with radome, under strong wind changes the projected area and reduces the drag co-
microwave antenna with cylindrical shroud, and grid microwave an- efficients [24,25]. However, Chapman et al. [23] determined that these
tenna. Once a parametrized microwave antenna model is available, it is effects are effectively negligible when dealing with slender structures
possible for an analyst to simply create an instance of the model with made of metal, like telecommunication towers.
the required geometry by adjusting only a couple of relevant parameters The air block is a box that represents the volume of area relevant to
(e.g., radius and height of the antenna). The wind direction relative to the study. Based on sensitivity analyses performed as part of the present
the antenna(s) must also be considered by using parameter Ψm (Fig 1). study, it is recommended to set an airbox of size Xair equal to 4 to 8
For the sake of computing the wind load coefficients, the structural times the panel width in the X direction and size Yair equal to 5 to
models (i.e., tower panel and microwave antenna) can be defined as 10 times the panel width in the Y direction, while its height Zair must
rigid bodies, so that there is no need to define the material properties be the same as the panel to obtain robust results. The top panel is an
and run a structural analysis. In fact, each panel is studied separately, as exception, because it requires an airbox at least twice as tall as the panel
customary in wind tunnel tests, so that empirical results can be used for itself, as it will be explained in Section 3.3 when discussing boundary
calibration and validation of the numerical models. A side effect of this conditions.

69
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

3.2. Turbulent flow models wall where viscous effects dominate (low Reynolds numbers) and, in the
analyses performed in this work, takes the value 𝜈̃ equal to three times
After extensive investigations, it was concluded that the best turbu- the kinematic viscosity (𝜈) in the far-field (e.g., at the inlet boundary
lence model for this type of analyses is one that augments the Reynolds- of the air volume, where the wind velocity is also imposed). Note that
averaged Navier-Stokes (RANS) formulation. There are several turbu- this choice is compatible with what suggested in the original work of
lence models based on the RANS formulation, which differ in terms Spalart and Allmaras [30], where more details about the formulation
of accuracy, robustness and computational time. The RANS turbulence can be found. The recommended boundary conditions are detailed in
model is recommended because it takes into consideration the turbu- the supplementary material. The main advantage of the Spalart-Allmaras
lent flow oscillations over time and treats them in a time-averaged sense method over the 𝐾 − ε and 𝐾 − ω methods is that the flow in the log
[25]. It is also possible to use other turbulence models, which augment layer (near-wall region) is resolved directly as 𝜈(𝑢
̃ 𝑑 ), where 𝑢𝑑 is the ve-
the Large Eddy Simulation (LES). These models are recommended when locity field tangential to the wall computed at a distance 𝑑 = ℎ∕2 from
a full wind load simulation is necessary (uplift, longitudinal, lateral and the wall itself, being ℎ the size of the mesh. Since 𝜈(𝑢
̃ 𝑑 ) behaves linearly
shear wind loads such as in buildings), but they require a very fine mesh, in the near-wall region, it is easier to solve than the velocity field itself,
which leads to high computational costs and large memory requirements which gives the Spalart-Allmaras method an advantage in terms of effi-
[26,27]. For the sake of determining the wind load coefficients, the ad- ciency and stability over the 𝐾 − ε and 𝐾 − ω methods. To summarize,
vantages of these models were deemed negligible, and their additional this technique allows the analyst to capture the details of the bound-
complexity unjustified. ary layer and near-wall flow physics with good accuracy provided that
In this study, three popular turbulence models based on the RANS the mesh is sufficiently fine, but without the need for an excessively
formulation were considered and tested on several tower panels, namely fine mesh grid as in the 𝐾 − ε and 𝐾 − ω methods. In computational en-
the 𝐾 − ε, 𝐾 − ω, and Spalart-Allmaras models. The 𝐾 − ε model re- vironments like COMSOL, the Spalart-Allmaras model takes advantage
solves the flow field at the boundary layer based on an analytically of the automatic wall treatment that can resolve the entire flow field
defined wall function. The wall function is an empirical function that down to the solid wall, where the viscous sublayer and buffer layer are
is used to model the behavior of the airflow close to the wall (i.e., the located. The benefit of the automatic wall treatment is that it switches
boundary of the air block). For instance, the analytical wall function for between a low Reynolds number formulation and a wall function for-
rough walls by Nikurade is one of the most commonly used [28,29]. mulation depending on the robustness and accuracy of the outcomes
This wall function applies a theoretical lift-off from the wall when sim- [25,30]. The model employs a robust wall treatment formulation when
ulating the flow field. In other words, the fluid flow simulation starts a coarse mesh is used, while a low Reynold number formulation is used
at a fixed distance (lift-off) from the wall. Such approximation led to with dense meshes.
inaccurate results in our investigations. Enhanced wall functions may It should be mentioned that there is active research in this field and
yield better results, but the fact that the accuracy of the model depends other turbulence models could yield even better results. For instance,
on the wall function is an issue in itself. In fact, this requires the an- the shear-stress transport (SST) and improved 𝐾 − ε (i.e., low Reynold
alyst to calibrate additional parameters and make assumptions which 𝐾 − ε model) are promising, even though for this specific application
may or may not apply to the problem at hand. This issue is solved by they would have a large computational cost that at the moment would
using the 𝐾 − ω model. In some environments for CFD analysis, such discourage their application in practice. On the other side of the spec-
as COMSOL, the 𝐾 − ω model implements an automated wall function trum, the Algebraic yPlus and L-VEL turbulence models deliver robust-
that resolves the flow field all the way down to the wall (using a low ness and require low computational time, but their accuracy is too low
Reynolds number formulation), but its performance is particularly sen- for these applications [25].
sitive to the mesh size (i.e., a fine mesh is needed). Also, the 𝐾 − ω
model accuracy is sensitive to the initial guess of K and ω that are es- 3.3. Computational fluid dynamics (CFD) analysis
sential parameters in estimating the turbulent viscosity. Moreover, the
𝐾 − ω model has convergence problems, in particular when dealing with Although the Supplementary Material attached to this paper provides
complex geometries such as lattice towers. In this study, it was found detailed guidance on how to set up and run the CFD analysis in COM-
that the convergence problems could be overcome by properly adjusting SOL, this section highlights some important findings and modeling rec-
mesh sizes, tolerance, and number of iterations. However, even after all ommendations that apply to any CFD software. A critical parameter that
the tuning to reach convergence, the analyses yielded inaccurate results is used to scale the boundary layers of the fluid flow simulation is the
with very long simulation times. For example, a tower panel analysis reference length scale. In our study, it was found out that an ideal value
required five to six hours on regular desktop computer with Core(TM) for the reference length is the average of the tower base dimensions and
i7–2600 CPU @ 3.40 GHz and 16.0 GB installed RAM, considering only its height. Setting the reference length scale to this value or to slightly
one wind direction. This inaccuracy in the results might be related to lower values leads to results that closely match experimental data.
the initial guess of the solution and/or to the refinement in the mesh. Fig. 3 shows the walls of the airboxes used for tower and antenna.
However, a model that is so sensitive to the initial guesses is not ro- In all CFD software packages, wall conditions can be assigned as “slip”
bust enough to be recommended for general practical applications. To or “no slip”. The slip wall condition allows the fluid (i.e., air) to move
avoid the issues of 𝐾 − ε and 𝐾 − ω, the Spalart-Allmaras model is rec- tangentially along the wall, and only prevents the fluid from moving
ommended instead. This model is an improved version of RANS which outside of the domain (i.e., the velocity normal to the wall is zero, while
shows numerical stability, good convergence, and satisfactory results the tangential velocity is nonzero). This condition is appropriate when
for many engineering applications, in particular for airfoil and turbine on the other side of the boundary there is more fluid, moving in a similar
blade applications for which it was originally calibrated. Our analyses way as the one inside the airbox. In contrast, the no slip wall condition
confirmed that the Spalart-Allmaras model satisfactorily balances ac- is used to represent a still, solid object that opposes the fluid’s motion in
curacy, computational time, and robustness. In particular, the model all directions. In other words, both the normal and tangential velocities
proved to be robust with respect to the mesh size and provided reason- of the fluid are zero at a no slip wall. For telecommunication towers, the
able results even with relatively coarse meshes for which other turbulent choice of slip or no slip conditions for the walls depends on the examined
flow models did not converge. Compared to the other available turbu- tower panel location. For the tower panel at the ground level, the base
lent flow models, the Spalart-Allmaras model uses a single additional is fixed to the ground while its top is connected to another panel. In
variable, i.e. the undamped turbulent kinematic viscosity 𝑣̃ to solve the this case, the roughness of the ground and the presence of obstacles can
RANS formulation. This quantity is set to 𝜈̃ = 0 on the boundary of a slow down the air movement, and the velocity approaches zero, which is
wall (realizing a no-slip boundary condition), behaves linearly near the well captured by modeling the bottom wall with a no slip condition. The

70
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Fig. 3. Wall definitions a) Tower body and the air block boundary conditions and b) Microwave antenna and the air block boundary conditions.

other faces of the air block (left, right and top) are actually immersed which has the benefit of being compatible with the air tetrahedrons.
in more air, so they can be modeled with the slip condition. For panels The lateral surface of the microwave antenna uses a mesh of quadrilat-
located in the middle of the tower, all four lateral walls are submerged in eral elements, while the front and back sides use unstructured triangular
more fluid that moves in the direction of the wind, so the slip condition elements. These mesh options are available in all software packages.
is appropriate for the left, right, top, and bottom walls (Fig. 3). Similar A careful choice of solver is also essential to run the CFD analysis.
considerations apply to the bottom, left, and right walls of the top-most Since winds can be considered stationary [33] and the output of the
panel in the tower, but special consideration needs to be given to its top analysis (i.e., lateral and drag coefficients) is not changing over time in
wall. In this case, the top wall of the air block needs to be placed far an average sense [34], a steady-state analysis is sufficient. The stationary
from the top face of the tower (at least twice the panel height) to allow solver Parallel Sparse Direct Solver (PARDISO) is used as liner system
the fluid to move smoothly over the tower without interfering with the solver and a segregated solver is used to handle the fluid parameters
(slip) wall which would preclude vertical displacements (i.e., the top (velocity, pressure, and turbulence variables) for a segregated solution
wall of the air block is not flush with the top face of the tower panel, approach (dividing the solution process to sub-steps). Finally, to account
while the bottom wall of the air block is flush with the bottom face of for different wind directions (Ψ), a parametric study done by setting a
the tower panel). parametric sweep can be carried out. In this way, the analysis is repeated
Even though the boundary conditions just described are the most for various rotations of the structure around its vertical axis over the
realistic, we performed several tests on different tower panels by chang- specified range of wind incidence angles.
ing the wall boundary conditions and observed very small differences For each wind direction, the CFD model computes the velocity field
in the estimated drag forces using these slip or no slip conditions (ap- and the pressure field. Then, the total stresses can be simply found by
proximately 0.05 %). This is due to the fact that the size of the airbox summing pressure stresses and viscous stresses. COMSOL simplifies this
compared to the tower is such that the boundary conditions have small process and calculates all values of total stresses, viscous stresses and
effect on the pressure against the tower, and the viscous forces devel- pressure stresses on each element of the structure. The forces vector 𝑭
oped at the no slip walls are negligible. Given that the slip condition at is computed by integrating the total stresses tensor (𝝅) projected onto
all lateral walls reduces the computational cost of the analysis and does the normal to the surface (𝒏) over the surface (𝑆) of the tower:
not impact the results, it is recommended to use the slip condition also
for the walls of the panels connected to the ground. This approximation 𝑭 (Ψ) = [𝒏 ⋅ 𝝅(Ψ) ] 𝑑𝑆 (3)
∫∫
is common in practice [31,32]. Thus, slip wall is used to model all men- 𝑆
tioned walls of the air block in case of the tower panel and as well for Eq. (3) is also used to estimate the forces vector on the antenna by
the antenna. On the other hand, the no slip condition needs to be used to replacing Ψ with Ψm . In this study, Ψ and Ψm are measured in a clock-
model the fluid flow boundary with the structural elements (i.e., tower wise direction (Fig. 1). Then, the drag and lateral wind load coefficients
panel and the microwave antenna). Yang et al. [32] have used similar are simply estimated using the aerodynamic wind load formula and the
boundary conditions in performing CFD analysis and managed to elimi- forces computed from Eq. (3). For instance, the tower panel drag coef-
nate the effect of boundary layers in the wind tunnel to replicate similar ficients (CD ) are computed as:
conditions in their experiments. In this analysis, the antenna is placed in 2 ⋅ 𝐹𝑦 (Ψ)
the middle of the air block (i.e., does not touch any wall) where the fluid CD (Ψ) = (4)
𝜌 ⋅ 𝑢2 ⋅ 𝐴P (Ψ)
can move freely around it and not interfering with the air block walls.
The fluid velocity used for the test is set as boundary condition for the where 𝐹𝑦 is the along wind load component (i.e., drag force 𝐹𝐷 ) of 𝑭 ; 𝜌
inlet wall. The outlet wall, on the other hand, is set with the boundary is assumed to be equal to 1.225 kg/m3 in this study; and 𝐴P is the pro-
condition in which the air pressure is equal to the atmospheric pressure. jected area associated with each wind direction. It should be noted that
Different mesh types have been investigated to determine the best the computation of 𝐴P is laborious because in lattice structures members
modeling option. The domain occupied by the fluid requires a three shield each other. In other words, the calculation of 𝐴P is not straight-
dimensional representation and the best results were obtained with un- forward and is based on the modeler judgment to select which member’s
structured tetrahedral elements. The tower and antenna are modeled surface is directly prone to wind. To avoid the complexity involved in
as rigid bodies, so only their lateral surfaces are relevant and can be the computation of the projected area, the reference area is frequently
mapped with two dimensional elements. Unstructured triangular ele- used as alternative in the analysis and design of lattice towers [2,10-
ments are employed to mesh the external surface of the tower panel, 13,15,35,36]. As a result, it is customary to use only one reference area

71
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

(𝐴𝑟 ) for all wind directions. The reference area is defined as the net area block dimensions was conducted. The test stopped when the results of
of the face directly invested by the wind when Ψ = 0°: the drag forces reached stability, meaning that the relative difference
( ) | | between the precedent and subsequent analyses was less than 1.5 %.
𝐴𝑟 = 𝐴 P Ψ = 0 o = |𝑛𝑦 | 𝑑𝑆 (5) In this analysis, the drag forces and drag coefficients as a function of
∫ ∫ | |
𝑆̃ (Ψ=0o ) the wind direction were calculated to calibrate these parameters. It is
|𝑛𝑦 | is the absolute value of the y component of the vector normal noted that in most cases a fine mesh is needed to accurately estimate
to the surface; 𝑆̃ (Ψ) is the surface exposed to the wind for direction Ψ. the results (e.g., 2.5 million elements for a single panel). Also, large
The procedure is adapted to compute the lateral coefficients (CL ) using air block dimensions are required in most cases, as recommended in
the lateral wind load component (i.e., 𝐹𝑥 ) computed by Eq. (3). In this Section 3.1 to obtain stability in the results. When the modeling pa-
analysis, the lateral wind component is called 𝐹𝐿 . rameters recommended in the previous section were used (e.g., using
Similarly, the drag coefficients of the microwave antenna (CDm ) are fine unstructured tetrahedral elements to model the fluid and unstruc-
calculated using Eq. (4), by replacing 𝐴P with 𝐴Pm . The projected area tured triangular elements to mesh the external surface of the tower),
of the microwave antenna (i.e., 𝐴Pm ) is computed numerically: the results of the analysis were very close to the experimental results, as
shown in Table 2. It can be seen that the maximum absolute deviation s
( ) | |
𝐴Pm Ψm = |𝑛𝑦 | 𝑑𝑆 (6) is 5.7 % at Ψ= 0° and a deviation less than 2 % is observed for all other
∫ ∫ | |
𝑆̃ (Ψm ) wind directions. The average error in the drag coefficient computations
under different wind directions is 2.43 %, which is consistent with the
Here, the projected area is used because the antenna has a simple
goal of keeping the error lower than 5 %.
geometry. It is also possible to use the reference area instead.
Another important parameter that needs to be calibrated carefully
Once, the wind load coefficients for tower panels and antenna are
is the tower location inside the air block. For this purpose, a triangular
computed, Eq. (2) is used to compute the interference factor. Finally,
lattice tower panel was examined. This second example served also as a
the total drag forces (𝐹𝐷𝑇 ) resulting from the tower and the antenna
way to test the application of the protocol to a different type of tower.
interaction are computed using Eq. (7):
The tower legs have a circular cross-section, while the horizontal and di-
1 [ ( ) ( )] agonal bracings are made with angle cross-section. In this analysis, four
𝐹DT (Ψ) = ⋅ 𝜌 ⋅ 𝑢2 ⋅ CD (Ψ) ⋅ 𝐴𝑟 + fa (Ψ) ⋅ CDm Ψm ⋅ 𝐴Pm Ψm (7)
2 cases are considered. First, a tower panel with 5 cells, as shown in Fig. 4.
The focus of the analysis is on the three central cells (i.e., cells 2, 3, and
4. CFD model calibration and validation 4), but the analysis region is expanded to include one cell above and one
below, to capture well the fluid motion at the interface between cells 1
The accuracy of the CFD analysis is highly dependent on the assump- and 2 as well as cells 4 and 5. This case is used as reference and compared
tions and approximations introduced in the models, such as the selected against the results obtained with cases including only the three central
turbulence model, mesh size, wall functions, boundary conditions, and cells. Second, the tower panel with the three central cells is modeled by
the structural model (i.e., rigid body). Previous research indicates that leaving a large space between the top face of the tower panel and the top
CFD results with an average error up to 10 % are still acceptable for wall of the air block. In this case, the airflow can move freely at the top
some applications [37,38]. To demonstrate the sensitivity of fragility without interacting with the structure (i.e., freestream). Third, the top
curves to wind load coefficients and determine the necessary level of ac- and bottom walls of the air block are flush with the top and the bottom
curacy in the CFD results, we studied as example an archetype 35 m high faces of the tower panel. Fourth, the tower panel is positioned in the
steel monopole tower subjected to wind loads. The tip displacement is middle of the air block. The wind fields resulting from the CFD analyses
a common “engineering demand parameter” used in structural fragility for these four cases are presented in Fig. 5. Based on the simulated drag
analyses as a proxy of the global structural response, and a way to assess forces at Ψ= 0°, there is a small discrepancy of 2.2 %, 0.14 %, and 3.9 %
the level of damage. For demonstration purposes, it was assumed that for cases 2, 3, and 4 compared to the reference case 1. Therefore, case 3
the tower is located in open terrain and under a design wind speed set that is associated with minimum discrepancy is confirmed to be the best
equal to 40 m/s. We used the aerodynamic wind load formula [35] to modeling strategy, which provides accurate results while enabling the
compute the load applied at the tip of the tower. The drag coefficient study of separate tower panels and avoiding unnecessary computational
was approximated by a random variable having a lognormal probability costs due to using a large air block. Case 3 captures well the wind field
distribution with mean equal to 1.2 and coefficient of variation equal velocity/stresses and the continuity in the boundary conditions. In all
to 0.42, based on the study performed by Giaccu and Caracoglia [39]. analyses, we also realized that placing the tower close to the inlet wall
Failure was defined as the tip displacement exceeding 1 % of the tower (i.e., Xair/2 and Yair/5) leads to accurate results. This is consistent to
height [39]. A Monte Carlo simulation with one million samples was the practice of placing the structural model close to the region where
conducted and the statistics of the tip displacement were calculated. the fans are installed, which is customary in wind tunnel tests [2,14,20].
We observed that increasing the mean value of the drag coefficient by For validation, the structures investigated in the experimental work
10 % can increase the probability of failure in the structure by 31 %, performed by Martín et al. [2] were modeled to assess the drag coeffi-
suggesting that structural reliability estimates can quite be sensitive to cient results using the proposed protocol and compare them with the ex-
the drag coefficient. In particular, it was determined that changes in perimental results. The experimental work studied different tower con-
the drag coefficient larger than 6.5 % can lead to significant changes in figurations, but this analysis was performed on the square tower panel
the tower reliability index. On the other hand, our preliminary studies with 6 m height and 1.17 m width and a tower panel with two antennas
show that if the error on the wind load parameters is capped at 5 %, under normal wind (Ψ= 0°). The two configurations are shown in Fig. 6.
then the potential error on the reliability index is substantially smaller In both cases, our analyses show general agreement and a discrepancy
than 5 %. Thus, our goal in this study is to calibrate a CFD model with much smaller than 5 % on the drag coefficients, as shown in Table 3.
errors smaller than 5 %. As an additional qualitative check on the results, we studied the trend
The protocol was calibrated using the experimental works by Li et al. of the drag coefficients for the isolated antenna with respect to the wind
[16] on a square tower panel with round members. The accuracy of incidence angle and compared our findings with experimental results
the results is very sensitive to the mesh size, air block dimensions, and [2,15]. In this analysis, we did not perform a quantitative comparison
tower location inside the air block. For example, very large discrep- because the antennas used in the experiments and those considered in
ancies with the experimental results (i.e., up to 40 %) were observed our study have different geometry. However, both the experimental and
when using inappropriate mesh types and sizes and small air block di- numerical results show maximum drag coefficients at Ψ= 0° and mini-
mensions. A convergence test examining different mesh sizes and air mum drag coefficients at Ψ= 90°, and they follow a very similar trend.

72
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Table 2
Experimentally and numerically estimated drag coefficients (calibration analysis).

Wind angle (Ψ) Experimental drag coefficient Simulated drag coefficient Difference [%]

0° 1.401 1.481 −5.7


15° 1.385 1.409 −1.7
30° 1.430 1.412 1.2
45° 1.317 1.302 1.1

Fig. 4. Tower panel consists of five cells (the reference case).

73
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Fig. 5. Velocity field of the CFD analysis for a) tower model with compensating segments (reference case), b) tower model with large air domain space from the top
(case 2), c) tower model flush with the air block faces (case 3), and d) floating tower (case 4).

Fig. 6. Structural models of the tower and antenna simulated based on experimental study by Martín et al. (2016).

74
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Table 3
Experimentally and numerically estimated drag coefficients (validation analysis). For the
experimental results, the case with turbulence intensity equal to 0.2 % is considered.

Drag coefficient computations at Ψ = 0°


Model Experimental results Numerical results Difference [%]

Tower panel 2.15 2.21 −2.8


Tower panel with two antennas 4.20 4.14 1.4

Table 4
Geometry of the three examined tower panels in the numerical example.

Tower panel number Leg section (mm) Bracing section (mm) Height (mm) Base width (mm) Top width (mm)

9 ∅114×5 L63×5 5000 1680 1680


6 ∅180×10 L70×5 6000 2670 2130
2 ∅219×12 L80×6 5000 4650 4200

5. Numerical examples tower panel. In this analysis a uniform flow of wind speed (i.e., 𝑢) equal
to 50 m/s is used. A uniform flow assumption is adopted, similar to the
The proposed protocol is applied to a tower with structural char- conditions often used in wind tunnel tests [2,32]. While for a correct
acteristics inspired by the DCTST collection [22]. The tower contains simulation of the structural behavior under strong wind it is essential
ten panels with equilateral triangular cross-section, as shown in Fig. 7. to capture the complexities and variations present in a real-world At-
The total height of the tower is 50 m and its width at the base is 5.10 m. mospheric Boundary Layer (ABL), for the sake of computing the wind
Each panel has the main members made with a tubular cross-section and load coefficients these complexities are typically neglected. Indeed, our
the secondary members made with an equal-legged angle cross-section. study showed that these coefficients do not even depend on the wind
Three tower panels located at different heights along the tower (2, 6, intensity, let alone the wind intensity variation along the height. Once
and 9 which are highlighted in red in Fig.7) are examined to show the the coefficients are computed in this way, they can then be combined
generality and robustness of the proposed protocol. Panel 2 features K- with an appropriate wind load profile for finite element analysis, and
bracings and is located at 5 m from ground. Panel 6 features X-bracings contribute to capturing the structural behavior. Table 5 shows that the
and is located at 27 m from the ground. Panel 9 is located at 43 m from maximum absolute lateral coefficients are 0.344, 0.186, and 0.161 for
the ground and has X-bracings like panel 6, but while panels 2 and 6 tower panels 2, 6, and 9, respectively. These values confirm that the lat-
are tapered, panel 9 is straight. Table 4 provides the structural details eral coefficients are very small compared to the drag coefficients. The
of each tower panel, which are used to build a specific instance of the sign of the lateral coefficients depends on the structure geometry and
parametric model described in Section 3.1. wind attack angle. A similar conclusion was reached by previous stud-
A typical cylindrical microwave dish antenna of 1.80 m in diam- ies [13,16,20]. The table indicates a small discrepancy between the drag
eter is considered, with geometric details taken from the RFS catalog coefficient values computed for Ψ = 0° and Ψ = 120°. This little dis-
[40]. The characteristics of the selected microwave dish are portrayed crepancy, in particular for panels 2 and 6 (i.e., −3.56 % and 2.12 %,
in Fig. 8. The radius and the height of the cylinder (dimensions ∅/2, B, respectively) is due to the fact that the tower is not perfectly symmetric,
and C in Fig. 8) are used in the parametrized model to create an instance due to the orientation of the cross-section of the angle bracings. Such
of this antenna. Once the structural models are created, the procedure small discrepancies are observed also in wind tunnel tests, and high-
described in Secs. 3.2 and 3.3 is applied to perform the CFD analysis. light the sensitivity of the load coefficients to relatively minor details
The meshes for the tower body and the microwave antenna are shown in [16,32]. Based on Table 5, the maximum drag forces and coefficients
Figs. 9(a) and (b), respectively. In this analysis, panel 2 is placed within for the tower panels are associated with Ψ = 0°, whereas the minimum
an air block of size 16 m x 16 m x 5 m and modeled with 2.1 million el- values are associated with wind directions ranging between Ψ = 30° and
ements. Panel 6 is placed within air block of size 11 m x 13 m x 6 m and Ψ = 45°. Moreover, as the tower bends under wind loads at Ψ = 0°, only
modeled with 1.3 million elements. Finally, panel 9 is placed within air one leg takes the entire compressive force which is particularly trouble-
block of size 8 m x 10 m x 5 m and modeled with 3.3 million elements. some for buckling, whereas two tower legs share the compressive force
The microwave antenna is placed within air block of size 11 m x 11 m for wind directions between Ψ = 30° to Ψ = 45° (see Fig. 10). Thus,
x 11 m and modeled with 1.0 million elements. for wind in the direction Ψ = 0° two issues compound each other (i.e.,
largest drag wind force and only one leg taking all the compression).
6. Results and discussion This observation has important ramifications for tower installations as
towers benefit from placement in the position most likely to face strong
6.1. Drag and lateral coefficients winds with Ψ = 30° to 45° based on the historical wind records at the
site.
The drag and lateral wind forces and coefficients of each tower panel The same procedure is applied to the microwave antenna, but using
under wind attack angles ranging from 0° to 120° with an increment of different wind directions that account for the different symmetry (i.e.,
15° are numerically assessed. Fig. 10 shows the drag force (𝐹𝐷 ) and lat- Ψm = 0° to 180° with an increment of 10°). In this analysis, the survival
eral force (𝐹𝐿 ) directions and the tower position at Ψ= 0°, 60°, and 90°. wind speed (i.e., 𝑢 = 55.56 m/s) is used [40]. The wind velocity fields
Due to the symmetry of the equilateral triangle, the CFD results are ex- resulting from the CFD analysis of the antenna for Ψm = 0° and 90°
pected to be identical for two tower positions shifted by Ψ= 120° and, are presented in Fig. 12. The surfaces exposed to wind (𝑆̃ ) used for
therefore, there is no need to examine wind directions for Ψ > 120°. The the calculation of the projected area for Ψm = 0°, 90°, and 180° are
domains of analysis for the three examined panels at Ψ= 0° are shown in highlighted in blue in Fig. 13.
Fig. 11. Following the computational procedure described in Section 3.3, Different inlet wind velocities are investigated for all models (i.e.,
the drag and lateral forces and their coefficients for different wind di- the tower panels and the antenna). In this case, no significant differ-
rections are estimated and the results are shown in Table 5. The table ences are observed in the assessed drag coefficients. Researchers have
also reports the reference area and the Reynolds number (Re) for each confirmed experimentally that the drag coefficients do not show any sig-

75
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Fig. 7. A 3D model of a representative telecommunication tower and the examined panels (panels 2, 6, and 9).

nificant difference when various wind speeds are used, and they have cients and the projected areas when the mean wind speed changes, as
also proved that the drag coefficients are independent of the Reynolds shown in Figs. 15(b) and (c), respectively.
number, in particular when the Reynolds number is larger than 104 The results are also compared to what would be obtained by us-
[2,13,16]. Fig. 14 shows that the Reynolds number for the antenna is ing design standards to assess the wind loads. Design codes provide the
indeed larger than 104 . The computed drag forces, drag coefficients, and drag coefficient as a function of the solidity ratio and show that the
projected areas for the antenna model under two inlet wind speeds (i.e., drag coefficients decrease as the solidity ratio increases. In this study,
𝑢 = 55.56 m/s and 𝑢 = 30 m/s) as a function of the wind direction are the solidity ratio (𝜑) represents the ratio between the reference area and
plotted in Fig. 15. Clearly, no changes are observed in the drag coeffi- the enclosed (i.e., solid) area of the tower face for Ψ = 0°. Since each

76
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Fig. 8. Cylindrical microwave dish antenna model.

Fig. 9. Mesh of the air block and the structure of a) tower body and b) microwave antenna.

side of the examined tower panel has the same solidity (i.e., same ge- putes the drag forces for Ψ = 60° and 90° by multiplying the drag forces
ometry configuration), it is expected that the drag coefficients be nearly for Ψ = 0° by wind directional coefficients equal to 0.80 and 0.85 for
identical for rotations of 60°. A small difference is attributed to the dif- Ψ = 60°, and 90°, respectively. As a result, the drag force associated with
ference in the orientation of horizontal and diagonal bracings. Fig. 16 Ψ = 90° is larger than the drag force associated with Ψ = 60°. However,
shows that drag coefficients are almost identical for rotations of 60°, the CFD results show that the drag forces for Ψ = 90° is lower than for
and they decrease as the solidity ratio increases. Table 6 reports the Ψ = 60°.
simulated drag forces applied to the tower panels and the values cal- The CFD analyses were performed within a reasonable simulation
culated based on the telecommunication standard code [10]. Note that time using a high-performance Linux workstation with 128 GB of RAM
this code provides values only for Ψ = 0°, 60°, and 90°. As expected, the and 16 CPU cores plus 768 GPU cores. The average simulation time of
results show that the drag forces computed by the code are larger (i.e., tower panel with dense mesh size was approximately 40 to 61 h for all
conservative) than those computed by CFD or observed in wind tunnel considered wind directions (i.e., nine CFD analyses, each taking 4 to 7
tests. Another interesting observation is that the standard code com- h).

77
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Fig. 10. Wind force components and tower position under three wind directions.

Fig. 11. Domain of analysis for a) tower panel 2, b) tower panel 6, and c) tower panel 9.

78
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Table 5
Drag and lateral forces and coefficients of the selected tower panels.

Wind angle (Ψ) 0° 15° 30° 45° 60° 75° 90° 105° 120°

Tower Panel 2 (𝐴𝑟 = 4.60 m , Re = 2.71×10 )


2 5

Drag force (𝐹𝐷 ) [kN] 14.818 13.727 12.200 13.022 13.630 13.556 13.120 14.883 15.349
Drag Coefficient (CD ) 2.104 1.949 1.732 1.849 1.935 1.925 1.863 2.113 2.179
Lateral force (𝐹𝐿 ) [kN] −1.141 −1.384 −1.125 −0.564 1.028 2.365 2.419 2195 1.130
Lateral Coefficient (CL ) −0.162 −0.197 −0.160 −0.080 0.146 0.336 0.344 0.312 0.160
Tower Panel 6 (𝐴𝑟 = 3.76 m2 , Re = 2.37×105 )
Drag force (𝐹𝐷 ) [kN] 8.938 8.642 7.919 8.306 8.542 8.422 8.034 8.540 8.857
Drag Coefficient (CD ) 1.553 1.501 1.375 1.443 1.484 1.463 1.395 1.483 1.520
Lateral force (𝐹𝐿 ) [kN] −0.550 −1.072 −1.067 −0.927 −0.272 0.527 0.345 0.122 −0.489
Lateral Coefficient (CL ) −0.096 −0.186 −0.185 −0.161 −0.047 0.092 0.060 0.021 −0.085
Tower Panel 9 (𝐴𝑟 = 2.05 m2 , Re = 2.13×105 )
Drag force (𝐹𝐷 ) [kN] 5.584 5.468 4.821 5.317 5.570 5.410 4.887 5.412 5.581
Drag Coefficient (CD ) 1.779 1.742 1.536 1.694 1.774 1.723 1.557 1.724 1.778
Lateral force (𝐹𝐿 ) [kN] −0.342 −0.451 −0.506 −0.360 0.128 0.135 0.122 −0.044 −0.315
Lateral Coefficient (CL ) −0.109 −0.144 −0.161 −0.115 0.041 0.043 0.039 −0.014 −0.100

Table 6
Drag forces based on CFD analysis and TIA standard code.

Wind angle (Ψ) Panel 2 Panel 6 Panel 9

Drag Forces [kN] Drag Forces [kN] Drag Forces [kN]

CFD model TIA standard CFD model TIA standard CFD model TIA standard

0° 14.818 16.096 8.938 10.842 5.584 6.492


60° 13.630 13.875 8.542 9.471 5.570 5.676
90° 13.120 14.430 8.034 9.814 4.887 5.880

Fig. 12. Top view of the velocity wind field on the microwave antenna a) at Ψ = 0° and b) at Ψ = 90°.

Fig. 13. Projected area of the microwave antenna for a given wind direction.

Fig. 14. Simulated cell Reynolds number based on the CFD model.

79
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Fig. 15. CFD analysis of the microwave antenna under two mean wind speeds a) drag forces, b) drag coefficients, and c) projected areas.

80
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

6.2. Total wind loads considering tower-antenna interaction

In this analysis, it is assumed that a single cylindrical microwave an-


tenna is mounted on each tower panel, attached to one corner, as indi-
cated in Fig. 17 and often done in practice. Eq. (2) is used to account for
the tower-antenna interaction. It should be noted that Eq. (2) is based
on the convention that positive values of 𝜃𝑚 correspond to clock-wise
rotations, so in this example 𝜃𝑚 is 60° (or −300°). The interference fac-
tors for all three panels are presented in Fig. 18. It can be seen that the
maximum values of the interference factor are observed at Ψ = 150° and
Ψ = 330°, when there is no overlap between the tower and the antenna.
In this case, the microwave antenna accelerates the airflow over the
tower and increases the total drag forces. Instead, the minimum inter-
ference factors occur at Ψ = 60° and at Ψ = 240°, due to tower-antenna
shielding effects. The values of the interference factor show a substan-
tial variation with the wind direction. Thus, the interaction between the
tower and the microwave antenna has a significant effect on the calcu-
Fig. 16. Drag coefficients in wind directional axis of the tower panels.
lation of the total drag forces and it should not be ignored. In contrast,

Fig. 17. Position of the microwave antenna a) panel 2, b) panel 6, and c) panel 9.

Fig. 18. Interference factors as a function of wind direction of a single cylindrical microwave antenna added to a tower panel.

81
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

Fig. 19. Total drag forces of tower panel 6 and the microwave antenna.

the results show small sensitivity to the panel topology and solidity ra- at Ψm = 90°. These observations can be useful for tower owners to al-
tio (the values for the three investigated panels are almost the same), leviate tower damage when installing a new tower in a site, based on
suggesting that these results can be a good approximation for a wide historical strong wind records. The examined tower panels experience
ranges of panels and antennas with similar configurations. Then, the to- very small lateral forces compared to the drag forces, which confirms
tal wind loads for the three panels can be computed based on Eq. (7). that the drag coefficients are the most important parameters in wind
In Fig. 19, the wind loads for panel 6 are compared with those obtained load calculation.
by linear superposition. The figure shows that the total drag forces com- The calculated drag forces are compared to those computed by a
puted based on superposition are overestimated for Ψ = 30° to Ψ = 90° standard code. As expected, the standard code overestimates the drag
and for Ψ = 210° to Ψ = 270° because they neglect the overlap (shielding forces and drag coefficients. What is more, the design code determines
effects) between the tower and the antenna (i.e., fa < 1). With simple su- that the minimum drag forces are at Ψ = 60°, while the protocol shows
perposition, a large overestimation occurs at Ψ = 60° when the antenna that for the investigated examples the most favorable directions are
completely blocks the airflow on the tower panel. In contrast, the forces Ψ = 30° and 90°. Finally, the total drag forces on the tower and the an-
are underestimated for the remaining wind directions because they do tenna computed by simple linear superposition, which is the approach
not account for the airflow acceleration when there is no overlap be- used in many engineering applications, are substantially inaccurate be-
tween the tower and the antenna (i.e., fa > 1). cause they disregard the tower-antenna interaction, and may lead to an
It should also be noted that the microwave antenna orientation (i.e., erroneous overestimation of the tower safety.
𝜃𝑚 ) may change over the course of a hurricane event, which leads to dif- The innovation of the presented protocol lies primarily in the devel-
ferent interference between the tower and the antenna. In this case, the opment of guidelines for systematic assessment of the load coefficients.
computation of the interference factor should be updated, and analysts By means of the provided guidelines, the proposed approach will be ap-
must make sure to take this point into consideration. These observations plicable in practice to various problems that involve portfolio analysis,
are important in performing structural dynamic time history analyses regional resilience assessment, and catastrophe modeling. In fact, for all
for damage and resilience assessments, reliability analysis and decision these analyses it is essential to have accurate load parameters for a very
making. large number of wind-antenna configurations, which would otherwise
require an impractical number of wind tunnel tests. For these reasons,
7. Conclusions the proposed approach is not meant to affect the design of an individual
tower, but it can have a transformative impact on the resilience assess-
This paper presents an accessible computational protocol to deter- ment of telecommunication infrastructure portfolios.
mine wind load coefficients for telecommunication towers, including
the effect of the wind direction and the tower-antenna interaction. The Relevance to Resilience
analyses show that the proposed protocol is robust, cost effective, and
yields a level of accuracy that is appropriate for practical applications This article proposed a robust and straightforward numerical
(error less than 5 % on the total loads). The protocol can be applied to simulation-based protocol for precisely estimating the wind load coef-
any tower-antenna configurations to estimate the wind load coefficients ficients and interference factor of telecommunication towers. These es-
and wind loads in a fast and easy way. It is calibrated and validated by timated parameters hold significant value for performing fragility and
means of experimental results available in the literature. It is found out reliability analyses, which are a building block of as resilience and risk
that the mesh size, wall boundary conditions, relative location of the assessments.
tower inside the air block, and air block dimensions affect the results
significantly, and thus recommendations are provided to tune these pa- Declaration of competing interests
rameters.
Triangular cross-section tower panels with different topology and so- The authors declare that they have no known competing financial
lidity ratios and a cylindrical microwave antenna are examined as illus- interests or personal relationships that could have appeared to influence
trative examples. Maximum drag coefficients are obtained at Ψ = 0° (i.e., the work reported in this paper.
the design forces), while the minimum values are observed at Ψ = 30°
for all three tower panels. The results also show that drag coefficients Acknowledgments
are almost identical for rotations of 60°, and they decrease as the solid-
ity ratio increases. For the microwave antenna, maximum drag coeffi- This research is part of the activities of the Catastrophe Modeling
cients are found at Ψm = 0° and minimum drag coefficients are noticed Center at Lehigh University (www.catmodeling.org). The support from

82
M. Khazaali, L. Ma, K. Rokneddin et al. Resilient Cities and Structures 3 (2024) 66–83

the Pennsylvania Department of Community & Economic Development [19] P. Martín, V.B. Elena, Personal communication, (2021).
(DCED) through grant PIT-19–02 and of Lehigh University through the [20] Carril CF, Isyumov N, Brasil RMLRF. Experimental study of the wind forces on rect-
angular latticed communication towers with antennas. J Wind Eng Ind Aerodyn
“Research Futures: Major Program Development” and the “Research Fu- 2003;91:1007–22. doi:10.1016/S0167-6105(03)00049-7.
tures: Special Seed Funding Opportunity” grants are gratefully acknowl- [21] Cell Reception. CellReception Find a better signal 2021.
edged. http://www.cellreception.com/ accessed September 2, 2021.
[22] China Tower Co., Drawing collection of telecommunication steel tower, (2014).
http://www.jianbiaoku.com/webarbs/book/86043/2508281.shtml (accessed
Supplementary materials September 2, 2021).
[23] Chapman JA, Wilson BN, Gulliver JS. Drag force parameters of rigid and flexible veg-
etal elements. Water Resour Res 2015;51:3292–302. doi:10.1002/2014WR015436.
Supplementary material associated with this article can be found, in
[24] Gao S, Pan S, Wang H, Tian X. Shape deformation and drag variation of a cou-
the online version, at doi:10.1016/j.rcns.2024.02.001. pled rigid-flexible system in a flowing soap film. Phys Rev Lett 2020;125:034502.
doi:10.1103/PhysRevLett.125.034502.
References [25] W. Frei, Which turbulence model should I choose for my CFD appli-
cation?, COMSOL multiphysics. (2017). https://www.comsol.com/blogs/
which-turbulence-model-should-choose-cfd-application/ (accessed September
[1] Tapia-Hernández E, Cervantes-Castillo JA. Influence of the drag co-
1, 2021).
efficient on communication towers. Int J Civ Eng 2018;16:499–511.
[26] Clannachan GH, Lim JB, Bicanic N, Taylor IJ, Scanlon TJ. Practical application
doi:10.1007/s40999-017-0157-z.
of CFD for wind loading on tall buildings. In: in: 7th Int. Conf. Tall Build; 2009.
[2] Martín P, Elena VB, Loredo-Souza AM, Camaño EB. Experimental study of the effects
p. 767–76.
of dish antennas on the wind loading of telecommunication towers. J Wind Eng Ind
[27] Daemei AB, Darvish A, Aeinehvand R, Razzaghipour A. Large-eddy simulation (LES)
Aerodyn 2016;149:40–7. doi:10.1016/j.jweia.2015.11.010.
on the square and triangular tall buildings to measure drag force. Adv Civ Eng
[3] SMITH B. 50 years in the design of towers and masts. From IASS recommendations to
2021;2021:1–11. doi:10.1155/2021/6666895.
current procedures. in: Symp. int. assoc. shell spat. struct. 50th 2009 valencia evol.
[28] Katopodes ND. Chapter 8 - channel bed resistance. In: Katopodes ND,
trends des. anal. constr. shell spat. struct. proc.. Editorial Universitat Politècnica de
editor. Free-Surf. flow. Butterworth-Heinemann; 2019. p. 416–63.
València; 2009.
doi:10.1016/B978-0-12-815487-8.00008-1.
[4] Travanca R, Varum H, Real PV. The past 20 years of telecommunication structures
[29] J. Nikuradse, Strömungsgesetze in rauhen Rohren. VDI-Forsch, (1933).
in Portugal. Eng Struct 2013;48:472–85.
[30] Spalart P, Allmaras S. A one-equation turbulence model for aerodynamic flows. in:
[5] Yanda M, Abejide OS, Ocholi A. Evaluation of collapse mechanism of telecommuni-
30th Aerosp. sci. meet. exhib.. Reno, NV, USA: American Institute of Aeronautics
cation tower. Niger J Technol 2020;39:1035–42.
and Astronautics; 1992. doi:102514/61992-439.
[6] Banipal K. Strategic approach to disaster management: lessons learned from Hurri-
[31] B.S. Paul, Numerical Analysis of the LCS 2 airwake to understand po-
cane Katrina, Disaster Prev. Manag. Int J 2006.
tential interactions during helipcopter operations, in: Day 2 thu novemb.
[7] Gao S, Wang S. Progressive collapse analysis of latticed telecommunication towers
05 2015, SNAME, Providence, Rhode Island, USA, 2015: p. D021S007R012.
under wind loads. Adv Civ Eng 2018. doi:10.1155/2018/3293506.
https://doi.org/10.5957/WMTC-2015-104.
[8] Tian L, Zhang X, Fu X. Collapse simulations of communication tower sub-
[32] Yang F, Dang H, Niu H, Zhang H, Zhu B. Wind tunnel tests on wind loads acting on an
jected to wind loads using dynamic explicit method. J Perform Constr Facil
angled steel triangular transmission tower. J Wind Eng Ind Aerodyn 2016;156:93–
2020;34:04020024. doi:10.1061/(ASCE)CF.1943-5509.0001434.
103. doi:10.1016/j.jweia.2016.07.016.
[9] Ma L, Khazaali M, Bocchini P. Component-based fragility analysis of transmis-
[33] Deodatis G. Simulation of ergodic multivariate stochastic processes. J Eng Mech
sion towers subjected to hurricane wind load. Eng Struct 2021;242:112586.
1996;122:778–87.
doi:10.1016/j.engstruct.2021.112586.
[34] Liu X, Gui N, Wu H, Yang X, Tu J, Jiang S. Numerical simulation of flow past
[10] ANSI/TIA-222-G, Structural standard for antenna supporting structures and anten-
a triangular prism with fluid–structure interaction. Eng Appl Comput Fluid Mech
nas, (2005).
2020;14:462–76. doi:10.1080/19942060.2020.1721332.
[11] ASCE07, Minimum design loads for buildings and other structures., Standard
[35] T.G. Mara, R.H. Behncke, Examination of yawed wind loading on transmission tow-
asce/sei 7-10., 2010.
ers, (2015) 522–37. https://doi.org/10.1061/9780784479414.042.
[12] ASCE-74 Guidelines for electrical transmission line structural loading,
[36] T.G. Mara, R.H. Behncke, Updating ASCE manual No. 74: guide-
fourth edition. Reston, VA: American Society of Civil Engineers; 2020.
lines for electrical transmission line structural loading, (2015) 154–65.
doi:101061/9780784415566.
https://doi.org/10.1061/9780784479414.013.
[13] Bayar DC. Drag coefficients of latticed towers. J Struct Eng 1986;112:417–30.
[37] Calis HPA, Nijenhuis J, Paikert BC, Dautzenberg FM, Van Den Bleek CM. CFD mod-
doi:10.1061/(ASCE)0733-9445(1986)112:2(417).
elling and experimental validation of pressure drop and flow profile in a novel struc-
[14] Calotescu I, Torre S, Freda A, Solari G. Wind tunnel testing of telecommuni-
tured catalytic reactor packing. Chem Eng Sci 2001;56:1713–20.
cation lattice towers equipped with ancillaries. Eng Struct 2021;241:112526.
[38] Motlagh AA, Hashemabadi SH. 3D CFD simulation and experimental validation of
doi:10.1016/j.engstruct.2021.112526.
particle-to-fluid heat transfer in a randomly packed bed of cylindrical particles. Int.
[15] Holmes JD, Banks RW, Roberts G. Drag and aerodynamic interference on microwave
Commun. Heat Mass Transf. 2008;35:1183–9.
dish antennas and their supporting towers. J Wind Eng Ind Aerodyn 1993;50:263–9.
[39] Giaccu GF, Caracoglia L. Wind-load fragility analysis of monopole towers by Lay-
doi:10.1016/0167-6105(93)90081-X.
ered Stochastic-Approximation-Monte-Carlo method. Eng Struct 2018;174:462–77.
[16] Li L, He S, He X, Jing H. Experimental study on wind force coefficient of a
doi:10.1016/j.engstruct.2018.07.081.
truss arch tower with multiple skewbacks. Adv Struct Eng 2020;23:2614–25.
[40] RFS, A.W.S. Microwave antenna system relocation kit - key products for
doi:10.1177/1369433220916935.
18GHz point-to-pointapplications, (2006). https://docplayer.net/21041363-
[17] ESDU, ESDU 81028: lattice structures. Part 2: mean fluid forces on tower-
Aws-microwave-antenna-system-relocation-kit.html (accessed September 6, 2021).
like space frames., (1993). https://www.esdu.com/cgi-bin/ps.pl?sess=unlicensed_
1210825160241lfc&t=doc&p=esdu_81028d (accessed August 25, 2021).
[18] COMSOL v.5.4, COMSOL Multiphysics® software, (2021). www.comsol.com. COM-
SOL AB, Stockholm, Sweden.

83

You might also like