You are on page 1of 12

Aero-elastic behavior of open-type one-way

tensioned membrane structure models


Cite as: AIP Advances 11, 105201 (2021); https://doi.org/10.1063/5.0062740
Submitted: 07 July 2021 • Accepted: 11 September 2021 • Published Online: 01 October 2021

Zhaoqing Chen, Liang Yin, Lixiang Tang, et al.

ARTICLES YOU MAY BE INTERESTED IN

Modeling and experimental analysis of particle damping phononic crystal plate


AIP Advances 11, 105202 (2021); https://doi.org/10.1063/5.0069603

Uniqueness of parameter estimates obtained from fitting free carrier absorption data of
silicon wafers
AIP Advances 11, 105003 (2021); https://doi.org/10.1063/5.0059258

Structural, electronic, mechanical, thermal, and optical properties of UIr3 under pressure: A
comprehensive DFT study
AIP Advances 11, 105205 (2021); https://doi.org/10.1063/5.0064021

AIP Advances 11, 105201 (2021); https://doi.org/10.1063/5.0062740 11, 105201

© 2021 Author(s).
AIP Advances ARTICLE scitation.org/journal/adv

Aero-elastic behavior of open-type one-way


tensioned membrane structure models
Cite as: AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740
Submitted: 7 July 2021 • Accepted: 11 September 2021 •
Published Online: 1 October 2021

Zhaoqing Chen,1,a) Liang Yin,1 Lixiang Tang,2 and Shuang Wang1

AFFILIATIONS
1
School of Civil Engineering and Architecture, Northeast Electric Power University, Jilin 132012, China
2
Liaoning Provincial Transportation Planning and Design Institute Co., Ltd., Shenyang 110000, China

a)
Author to whom correspondence should be addressed: chenzhq2004@163.com

ABSTRACT
Wind tunnel tests were conducted on an aero-elastic model of an open-type one-way tensioned membrane to investigate the aero-elastic
instability mechanism of membrane structures. The structural response and wind velocities above the membrane were measured and analyzed.
Complementary simulations were carried out to explain the interaction between the wind and the structure. The results indicate that the aero-
elastic instability of the membrane is a kind of vortex induced vibration. This kind of instability is characterized by a sudden increase in the
amplitude, or a sudden change in the dominant vibration mode, when the coming flow velocity exceeds a certain value. The vortices generated
around the time-averaged deformation of the membrane lead to vortex induced vibration when the dominant frequency of the vortices is
close to a certain natural frequency of the membrane. When the aero-elastic instability occurs, the frequency of the vortices is locked in by
the structural vibration within a certain coming flow velocity range, and the total damping ratio of the vibration decreases dramatically and
becomes nearly zero. Finally, the critical wind velocities are defined, and the reduced critical wind velocities corresponding to the first and
second aero-elastic instabilities are 0.84 and 2.3 approximately.
© 2021 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license
(http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1063/5.0062740

I. INTRODUCTION smooth flow and found that a self-excited oscillation in the first
anti-symmetric mode was induced. He concluded that the oscilla-
Membrane structures tend to deform and vibrate significantly tion was induced by the vortices forming and shedding downwind
under the wind load due to their lightweight and flexibility.1 It at a certain velocity. Miyake et al.8 found that the vortices gener-
is believed that wind–structure interaction may make membrane ated on stationary suspended roofs played an important role in the
structures aerodynamically instable and cause a collapse at rela- excitation of the roofs. Minami et al.9 studied the vibration behav-
tively low coming flow velocities.2 However, the mechanism of ior of a two side supported PTFE membrane model and defined the
this wind–structure interaction remains unclear despite extensive state of membrane vibration with progressive waves as flutter and
research. This is probably due to the difficulties resulting from the state that the membrane lifted upward and shifted to another
complex flow separation and geometric nonlinearities.3,4 stationary state as a kind of divergent flutter. Sygulski10 studied the
Some researchers have studied the aero-elastic response of flex- stability of a two edge supported latex rubber membrane and sug-
ible roofs. For example, Kawamura and Kimoto5 applied a modified gested that when the membrane vibrated like a standing wave and
thin airfoil theory to study the aerodynamic stability of a one-way exhibited a three half-wave mode shape, it was a kind of diver-
hanging roof and suggested a criterion for the critical wind velocity gent flutter. Uematsu and Uchiyama11 also studied the response of
at which some aero-elastic instability occurs. Kimoto and Kawa- a hyperbolic–parabolic shaped suspended roof and found that the
mura6 tested the response of a full-scale hanging roof in the natural oscillation for a certain wind direction was quite similar to vortex-
wind and defined the critical velocity as the coming flow velocity excited oscillation; however, no dominant periodicity in the pres-
beyond which the vibration amplitude increases quite rapidly. How- sure fluctuations was found on the rigid model. Wu et al.12 repro-
ever, they did not provide any clear reason for this phenomenon. duced the fluid–structure coupling process of the one-way tensioned
Matsumoto7 studied the oscillation of a tensioned cable roof in a membrane structure on computer and discussed its mechanism.

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-1


© Author(s) 2021
AIP Advances ARTICLE scitation.org/journal/adv

Ding et al.13,14 studied the vibration-induced unsteady aerodynamic at the level of the roof, so it can be regarded as a smooth flow
forces of membrane structures and found that the values of the aero- condition.
dynamic damping coefficients were negative under specific condi- Schematic diagrams of the aero-elastic model and the experi-
tions. Liu et al.15 studied the wind-induced aerodynamic stability of mental setup are shown in Fig. 1. The roof was made from a 0.4 mm
membrane structures through the analytical method and proposed a thick latex sheet. The density ρ and the elastic modulus E of the
formula for evaluating the critical wind velocity of orthotropic sad- latex were 0.4133 kg/m2 and 1.638 × 106 N/m2 , respectively. The
dle membrane structures. Li et al.,16 Zhou et al.,17 and Chen et al.18 height H was 400 mm, the span L was 600 mm, and the width
studied the wind-induced added mass effect of open flat membrane B was 1200 mm. The roof was pre-tensioned by a pair of 12 mm
structures and proposed some methods to evaluate the added mass diameter steel circular tubes that were clamped to two other screw
of membrane structures. rods parallel to the wind direction using two special screw nuts.
From these studies, it can be appreciated that the aero-elastic The 12 mm diameter windward tube might affect the flow around
instability of membrane structures is quite different from that of the roof; however, as the roof will deform downward much more
bridge and high-rise structures. Hence, the traditional theory for than 12 mm under the wind load, the influence of the tube was
bridge and high-rise structures may not be suitable for explaining ignored.
the mechanism of aero-elastic instability of membranes. Some essen- Three laser displacement meters and three hot wires were used
tial questions still need to be answered, such as how to define the to measure the displacements of the roof along the center line and
aero-elastic instability of membranes, which conditions lead to this the wind velocities above the roof surface. The location of the mea-
instability, and how do the wind and structure affect each other. suring points is shown in Fig. 2, where L is the span of the roof,
In this paper, a wind tunnel test on an aero-elastic model of Di represents the displacement measuring point, and Ph-i represents
an open-type one-way tensioned membrane structure is described. the wind velocity measuring point. The lower suffixes h and i rep-
The response of the membrane and the flow field around the mem- resent the height and horizontal position of the measuring point.
brane at different coming flow velocities are analyzed systematically. Three hot wires were fixed at the tips of horizontal rods as shown
Complementary CFD (Computational Fluid Dynamics) simulations in Fig. 1. Therefore, the wind velocities at three points at the same
are carried out to investigate the relationship between wind and the height could be measured simultaneously. The horizontal rods could
membrane. Then, the critical wind velocity is defined and calculated. be moved in the vertical direction so that the wind velocities of all
Finally, the mechanism for the aero-elastic instability of membranes measuring points could be recorded. The laser displacement meters
is discussed. were installed under the wind tunnel floor to measure the vertical
displacements of the roof at Points D1 to D3 . A part of the wind
tunnel floor was removed and covered with a thin transparent plas-
II. AERO-ELASTIC MODEL WIND TUNNEL TEST tic film so that the laser beam could reach the membrane without
The experiments were carried out in a closed circuit wind tun- air leakage. The pretension of the membrane was varied from 20 to
nel, which has a working section that is 4 m wide, 3 m high, and 40 N/m in steps of 10 N/m by changing the location of the tubes
25 m long. A turbulent boundary layer is developed on the wind supporting the membrane with the screw nuts. The wind velocity
tunnel floor and is about 2 cm thick at the model location. As will was changed from 4 to 15 m/s in steps of 1 m/s.
be described, the height of the roof (0.4 m) is much higher than The pre-tension applied on the roof was checked by comparing
this thickness. The turbulence intensity of the flow is about 0.5% the mode frequencies of the roof obtained from the free vibration

FIG. 1. Schematic diagram (a) and photograph (b) of the elastic model.

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-2


© Author(s) 2021
AIP Advances ARTICLE scitation.org/journal/adv

FIG. 3. Time history of displacements for a coming flow velocity of 10 m/s.

FIG. 2. Measuring points for displacements and wind velocities.

experiment with the results from a finite element analysis to ensure


the accuracy of the pre-tension.
Although there is no full-scale prototype related to the elastic
model, the pre-tension and frequency of such a prototype, calcu-
lated according to the similarity law, should be in a reasonable range. FIG. 4. Time-averaged deformation and rms deformation.
Assume that, for the prototype, the height H p is 8 m, the span Lp is
12 m, and the width Bp is 12 m, the material is Kevlar with a mass
per unit area, and ρp is 2.2 kg/m2 . These are similar to those which down gradually with increasing coming flow velocity. The deformed
Irwin and Wardlaw19 dealt with in their research. The aero-elastic shape of the membrane looks like an arc because its initial shape
scale parameters can be calculated according to the similarity law deforms slightly downward due to its weight.
introduced by Tryggvason,20 which are given as follows: A simple CFD steady-state numerical simulation was applied
The length scale ratio is λL = L/LP = 0.6/12 = 1:20. to the static model with the same deformation shape as the aero-
The mass scale ratio is λm = ρ/ρP = 0.4133/2.2 = 1:5.4. elastic model, and the reference height was the top of the model.
Assume that the wind velocity ratio is λV = 1:2. Figure 6(a) shows the computational domain of the static model,
The elastic force per unit length is λT = λV 2 λL = 1:80. where B, L, and H represent the width, span, and height of the roof
The frequency scale ratio is calculated as λf = λV /λL = 10. model, respectively. Figure 6(b) shows the mesh arrangement for the
curved roof model. The unstructured grid was used to simulate the
The range of the pre-tension in the test model is 20 and 40 N/m, flow field around the structure, and the Reynolds stress model was
which corresponds to 1.6 and 3.2 kN/m at full-scale. The first natural used to simulate the flow motion. Considering that the wind flow
frequency of the test model in still air is 3.9 Hz, which corresponds is an incompressible flow, the inlet boundary condition was defined
to 0.39 Hz at full-scale. as the velocity inlet, the outlet boundary condition was defined as
outflow, and the top, sides, ground, and roof surface were defined as
III. AERO-ELASTIC RESPONSE ANALYSIS non-slip wall conditions.
A. Time-averaged deformation and rms deformation
of the roof
Figure 3 shows the time history of the displacement of three
points on the membrane when the wind velocity was 10 m/s and the
pre-tension was 30 N/m. It is found that the response of the mem-
brane can be divided into two parts: the time-averaged deformation
and the rms deformation. The time-averaged deformation is caused
by the average component of the wind load and corresponds to the
equilibrium position of the roof. The rms deformation is caused by
the fluctuating component of the wind load and corresponds to the
vibration amplitude of the roof. Figure 4 shows the two kinds of
deformations.
Figure 5 shows the time-averaged deformations of the roof for
several wind velocities where D represents the mean value of the dis-
placement and x/L represents the position of the measuring points
FIG. 5. Time-averaged deformations of the roof at different coming flow velocities.
along the span. The equilibrium deformation of the roof moves

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-3


© Author(s) 2021
AIP Advances ARTICLE scitation.org/journal/adv

FIG. 6. (a) Computational domain and (b) mesh arrangement for the curved model.

The wind pressure coefficient of the measurement point i Cpi remain in a slightly convex vibration equilibrium state. With the
on the model can be represented by the following equation: increase in wind speed, the upward deformation will become larger
Pi − PH
and larger. This shows that the vibration equilibrium position of the
Cpi = , (1) membrane is related to the initial deformation shape of the structure
P0 − P∞ and will increase with the increase in wind speed.
where Pi is the wind pressure of measurement point I and PH , P0 , Figure 9 shows the variation of rms displacements of the roof at
and P∞ are the wind pressure, total pressure, and static pressure at Points D1 to D3 for different coming flow velocities. From this figure,
reference height H that is far ahead from the model, respectively. it can be seen that the rms amplitude of the roof increases slowly with
Figure 7 shows the wind pressure coefficient distribution for the coming flow velocity increasing up to 8 m/s, and then, it begins
the two sides of the curved rigid model. It can be seen from this to increase rapidly for coming flow velocities higher than 8 m/s.
figure that for the static model with this shape, the pressure on the
upper surface is positive pressure and the pressure on the lower sur-
face is negative pressure. As a result, the resultant force on the whole B. Analysis of the modal contribution
surface is downward. So, with the increase in wind speed, the resul- The vibration modes of the roof are standing wave modes as
tant force increases gradually, and the average deformation of the shown in Fig. 10. The power spectra of the displacements and the
membrane surface increases gradually. phase differences between the displacement measurement points
As can be seen from Fig. 8, when a lighter material (like rain- can be used to identify the mode shapes.
proof silk) is used as the model of the roof, if an upward force Figure 11 shows an example of the analysis of the vibration
is applied to the membrane surface in uniform flow, the roof will mode for the test case at a coming flow velocity of 8 m/s. Figure 11(a)

FIG. 7. Wind pressure coefficient distribution for the two sides of a curved rigid model: (a) up wall and (b) down wall.

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-4


© Author(s) 2021
AIP Advances ARTICLE scitation.org/journal/adv

mode vibration. Fig. 11(d) shows that, at a frequency of 9.3 Hz, the
time difference between D1 and D3 is about 0.05 s, resulting in a
phase difference of about 0.05 × 2π/(1/9.3) = 0.93π. This feature
indicates that the roof vibration is the second mode, i.e., the first
anti-symmetric mode. The other vibration modes can be identified
using this method. It should be mentioned that the frequencies of
the first three modes of the roof under the wind load are 2.9, 9.3, and
15.9 Hz.

C. Correlation between the wind and structure


The relationship between the wind and the structure can be
evaluated using the correlation coefficients of the displacements and
wind velocities above the roof. The correlation coefficient of any
two random variables x and y can be calculated according to the
following formula:
FIG. 8. Upward equilibrium position of the rain-proof silk model.
∑(xi − x)(yi − y)
rxy = √ , (2)
∑ (xi − x)2 (yi − y)2

where rxy is the correlation coefficient of x and y and xi and yi are


the ith values of x and y. x and y are the mean values of x and y.
The variation of the correlation coefficients at different com-
ing flow velocities is shown in Fig. 12, where rDP are the corre-
lation coefficients of D1 and Pi−1 (i = 40, 70, 100, 130, and 160).
The results show that the displacements of D1 correlate significantly
with the wind velocities of P70-1 and P160-1 at 9–10 and 14–15 m/s,
respectively.
Figure 13 shows the normalized PSD of displacements and
wind velocities at different coming flow velocities, where f i is the
frequency of the ith vibration mode and F i is the peak frequency of
the wind velocity. Usually, the peak frequency of the wind velocity
FIG. 9. rms amplitude vs coming flow velocity. indicates the frequency of the vortex. The result shows that there is
no vortex above the roof when the roof vibrates with the first mode
at a coming flow velocity of 7 m/s. However, when the flow veloc-
shows the time history of displacements at Points D1 , D2 , and D3 . ity increases to 8 m/s, a vortex with a frequency of 16.4 Hz occurs
The normalized Power Spectral Densities (PSDs) of the displace- above the roof and vibration in the third mode at 15.9 Hz occurs
ment at the three points are shown in Fig. 11(b), where S(f ) rep- at the same time. This indicates that the third mode of vibration
resents the PSD of the displacement and σ represents the rms value is a kind of Vortex Induced Vibration (VIV). When the coming
of the displacement. Three peak frequencies at 2.9, 9.3, and 15.9 Hz flow velocity increases to 9 m/s, the third mode vibration becomes
can be seen in this figure. Figures 11(c)–11(e) show the responses of dominant and the peak frequencies of the vibration and vortex are
the three points at three frequencies and were obtained by bandpass nearly the same; which indicates that resonance occurs. At a com-
filtering. Figure 11(c) shows that the three displacement measure- ing flow velocity of 10 m/s, the behavior is similar to that at 9 m/s,
ment points vibrate synchronously at a frequency of 2.9 Hz and the which seems to suggest that a the lock-in phenomenon has occurred.
phase difference between any two points is 0 so that this is the first When the coming flow velocity increases to 11 m/s, the energy of the
third mode becomes weak. However, at a wind velocity of 12 m/s,
another vortex with a frequency close to that of the second mode
vibration occurs and causes another VIV for the following coming
flow velocities.
Based on the above-mentioned research, some preliminary
conclusions can be drawn.
(1) A sudden increase in amplitude or a sudden change in the
vibration mode is a kind of oscillatory instability. This indi-
cates that the structure cannot maintain its equilibrium state
and jumps from one state to another.
(2) The vortices that form around the time-averaged defor-
mation of the vibrating membrane lead to a VIV of the
FIG. 10. First three mode shapes of the roof.
membrane.

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-5


© Author(s) 2021
AIP Advances ARTICLE scitation.org/journal/adv

FIG. 11. Analysis of the vibration mode: (a) time history of zero average displacements at 8 m/s, (b) PSD of the displacements, (c) displacement of 2.9 Hz, (d) displacement
of 9.3 Hz, and (e) displacement of 15.9 Hz.

(3) When the dominant frequency of the vortices is close to The sampling frequency of the response is 500 Hz, and the
that of the membrane, a resonance and lock-in phenomenon number of data points is 300 000. Superimpositions are carried out
occurs. In this case, the membrane absorbs energy from using data extracted outside the peak frequency f 0 ± 0.3f 0 using
the wind, vibrates more severely, and may jump to another a bandpass filter. The data for the first ten cycles in a random
vibration mode. decrement signature are used to evaluate the damping ratio.
Assume that the vibration includes n modes. The contribution
of the ith mode to the total vibration can be evaluated as follows:
D. Damping ratio identification
σi2
αi =
Usually, the aero-elastic instability of structures is considered to
n , (3)
be related to the variation in the total damping, which is the sum of
∑ σj2
structural damping and aerodynamic damping. If the total damping j=1
decreases to zero, it means that aero-elastic instability will occur. The
damping ratio of the membrane is evaluated from the response of where σ i is the rms displacement of the ith mode and can be
the structure by using the half-power bandwidth and the Random determined by the bandpass filter method.
Decrement Technique (RDT). The total damping ratio is evaluated as follows:
n
ξtotal = ∑ αi ξi , (4)
i=1

where ξ i is the ith damping ratio determined by the RDT method.


Figure 14(a) shows the variation of the first three mode damp-
ing ratios for different wind velocities. It can be seen that the higher
mode damping ratios are generally smaller than those of the lower
mode. Figure 14(b) shows the variation of the total damping ratio
with different wind velocities and pre-tensions. It can be seen that
with the increase in wind velocity, the damping ratio increases at
first, reaches a peak value, and then decreases suddenly, nearly
reaching zero. Although the structural damping and aerodynamic
damping cannot be separated accurately, it can be deduced that the
variation in the total damping is mainly caused by the aerodynamic
FIG. 12. Correlation coefficient for displacement and wind velocity.
damping. When the wind velocity increases and a sudden decrease

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-6


© Author(s) 2021
AIP Advances ARTICLE scitation.org/journal/adv

FIG. 13. PSD of displacements and wind velocities at different coming flow velocities: (a) 7–11 m/s coming flow velocities and (b) 12–15 m/s coming flow velocities.

in the damping ratio occurs, this indicates that an aero-elastic insta- A. Rigid model wind tunnel test
bility has been encountered, which coincides with the conclusions of To investigate the influence of vibration on the aerodynamic
former studies. force on the membrane, a wind tunnel test on a rigid model was car-
ried out (Fig. 15). The model was made of acryl glass, and its shape
IV. COMPLEMENTARY WIND TUNNEL TESTS was close to the time-averaged deformation of the membrane with a
AND CFD SIMULATIONS
pre-tension of 30 N/m at a wind velocity of 10 m/s. Wind pressures
To further investigate the interaction mechanism between the along the central line of the model were measured using a pressure
wind and the membrane, complementary wind tunnel tests and CFD transducer. The reference tube of the transducer was connected to
simulations were carried out. a pitot tube placed about 1 m upstream at the height of the rigid

FIG. 14. Total damping ratio of the roof:


(a) damping ratio of the first three mode
vibration of the 30 N/m pre-tensioned
model and (b) total damping ratio of the
three different pre-tensioned models.

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-7


© Author(s) 2021
AIP Advances ARTICLE scitation.org/journal/adv

transient method, was used for the calculation. The time step and
maximum number of iterations per time step were 0.001 s and 20,
respectively. The boundary condition for the roof surface is the wall,
and the inlet and outlet boundary conditions are velocity-inlet and
outflow, respectively.
The mean wind pressure coefficients for the rigid model from
the wind tunnel tests and CFD simulations are compared in Fig. 17,
where Cp is the wind pressure coefficient. This figure shows that the
two results agree well in both values and trends.
Figure 18 shows the vortex contours at different wind veloci-
ties around the deformed rigid models whose shape is the same as
the elastic model for the same wind velocities. It can be seen that
there is almost no flow separation occurring at a coming flow veloc-
ity of 7 m/s. However, some significant vortices form at a coming
FIG. 15. Rigid model. flow velocity of 8 m/s. This change indicates that the increasing com-
ing flow velocity and deformation are the reasons for the formation
model. The wind air velocities above the model were also measured of the large scale vortex, which may make the roof unstable.
by three hot wires to compare the flow field with that of the elastic Figure 19 shows the normalized PSD of the lift coefficient Cl
model. for the rigid models. Two significant peaks can be observed in the
PSD of the rigid model at a coming flow velocity of 8 m/s: one
is 0.1 Hz and another is 15.3 Hz. The latter is quite close to the
B. CFD simulations
frequency of the vortex (F3 = 16.4 Hz) observed above the elastic
CFD simulations of the rigid models with the same shape of model (in Fig. 13) at the corresponding wind velocity. This indi-
time-averaged deformation as the elastic membrane at each coming cates that the vortex observed above the elastic roof forms because
flow velocity were carried out to study the relationship between the of the deformed shape of the structure and the velocity of the com-
vortices and deformation. ing flow rather than the vibration of the roof. With an increase in
The calculation was carried out using FLUENT software for 2D the coming flow velocity, the frequency of the vortex increases to
conditions. In total, about 900 K structural grids were used in the cal- 16.4 Hz at 9 m/s and disappears at 10 m/s. Another vortex, with a
culation, and the meshing strategy is shown in Fig. 16. The thickness frequency around 2 Hz, develops for coming flow velocities between
of the roof is 0.005 mm. 9 and 12 m/s. However, the vortex with a frequency close to 16 Hz,
The Reynolds number (Re ) can be calculated using the follow- which was observed above the elastic model, did not disappear for
ing equation: coming flow velocities between 9 and 11 m/s. This phenomenon is
Re = 6.9 × 104 UDmax , (5) considered to be caused by the lock-in phenomenon of VIV where
the vortex will be captured by the membrane vibration within a wind
where U is the coming flow velocity that changed from 7 to 15 m/s speed range once VIV happens. Figure 18(b) shows a similar situ-
and Dmax is the max value of the vertical deformation that changed ation for the formation of a vortex, which causes a second VIV: a
from 0.0125 to 0.0861 m. vortex of 8.2 Hz, whose frequency is close to the frequency of the
The Reynolds number for the model changed from 6.0 × 103 vortex observed above the elastic model, forms (F2 = 9.0 Hz) above
to 8.9 × 104 . As a result, the flow separation mode of the roof was the rigid model at a coming flow velocity of 12 m/s and disappears
the laminar separation mode. Hence, the laminar model, a kind of quickly at 14 m/s.

FIG. 16. CFD calculation domain. FIG. 17. Comparison of results obtained by simulation and experiment.

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-8


© Author(s) 2021
AIP Advances ARTICLE scitation.org/journal/adv

FIG. 18. Instantaneous vortex contours around the rigid model for different coming flow velocities: (a) vortex contours at 7 m/s and (b) vortex contours at 8 m/s.

FIG. 19. PSD of Cl on the rigid model


at different coming flow velocities: (a)
PSD of Cl among 8–11 m/s and (b)
PSD of Cl among 12–15 m/s.

C. Influence of the vibration on the vortex that the vortices above the roof were locked onto the roof surface
Forced vibration simulations were conducted to study the influ- and moved downward along the roof surface.
ence of the vibration on the vortex above the structure. The dis- Figure 21 shows the normalized PSD curves of the wind veloc-
placements of the 30 N/m pre-tensioned elastic model when tested ity at different heights above the roof. This figure shows that the
at a coming flow velocity of 10 m/s were loaded on the nodes of the
structure in the CFD code. A dynamic mesh was used to simulate the
movement of the roof. The calculation parameters were the same as
those used in the rigid model simulation.
Figure 20 shows the instantaneous vortex contours around the
FIG. 20. Instantaneous vortex contours around a vibrating roof.
vibrating roof at a coming flow velocity of 10 m/s. The result shows

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-9


© Author(s) 2021
AIP Advances ARTICLE scitation.org/journal/adv

FIG. 21. PSD of the wind velocities at different heights


above the model: (a) forced vibration CFD simulation result,
(b) aero-elastic model experiment result, (c) rigid model
CFD simulation result, and (d) rigid model experiment
result.

forced vibration simulation results agree well with the elastic model Table I shows the reduced critical wind velocities of the roof. It
experimental results but are quite different from the rigid CFD sim- is found that the reduced critical wind speeds U c1 and U c2 are ∼0.8
ulation results. This phenomenon proves that vibration can lock in and 2.3, respectively.
the frequency of the vortex that forms above the membrane roof in
a certain range of coming flow velocities after VIV occurs.
VI. CONCLUSION
V. CRITICAL WIND VELOCITY FOR INSTABILITY The response of an open-type one-way tensioned membrane
Considering that the resonance is caused by a vortex whose fre- and the wind velocities above the roof have been investigated in
quency is close to a natural frequency of the roof, the first critical uniform flow to study the instability characteristics of the mem-
wind speed of instability (U 1 ) is defined as the wind speed that a brane and derive a definition of aero-elastic instability for membrane
vortex, whose shedding frequency is close to the third mode natural structures. Complementary wind tunnel tests and CFD simulations
frequency of the structure, forms. Similarly, the second critical wind were conducted to study how the vortex above the roof forms and
speed of instability (U 2 ) is defined as the wind speed that a vortex, the influence of the roof vibration on the vortex. The conclusions
whose shedding frequency is close to the second natural frequency of are as follows:
the structure, forms. A reduced critical wind velocity can be defined
(1) The aero-elastic instability of membrane structures is caused
as follows:
Ui by vortex induced vibration and can be characterized as a
Uci = , (6) sudden increase in amplitude or a sudden change in the
f nL
vibration mode.
where U ci is the ith reduced critical wind velocity, U i is the ith critical
(2) The vortices generated around the time-averaged deforma-
wind velocity, f n is the corresponding resonance frequency, and L is
tion of a membrane lead to a vortex induced vibration when
the span of the roof.
the dominant frequency of the vortices is close to a certain
natural frequency of the membrane.
TABLE I. Critical wind velocities for one-way type membranes at different (3) Once the aero-elastic instability of membrane structures
pre-tensions. occurs, the vibration can lock in the frequency of the vortex
above the roof within a certain wind velocity range, and the
Pre-tension
total damping ratios of the structure can decrease to nearly
(N/m) U 1 (m/s) f3 (Hz) U 2 f2 (Hz) L (m) U c1 U c2
zero.
20 7 13.8 11 8.2 0.6 0.84 2.2 (4) With an increase in coming flow velocity, the membrane can
30 8 15.9 12 8.8 0.6 0.84 2.3 experience numerous aero-elastic instabilities. According to
40 9 17.7 13 9.6 0.6 0.85 2.3 the current study, the reduced critical wind velocities for the
first and second instabilities are ∼0.84 and 2.3.

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-10


© Author(s) 2021
AIP Advances ARTICLE scitation.org/journal/adv

6
Therefore, for the tensioned membrane structures, the vor- E. Kimoto and S. Kawamura, “Aerodynamic behavior of one-way type hanging
tex close to the fundamental frequency of the structure should be roof,” J. Wind Eng. Ind. Aerodyn. 13, 395–405 (1983).
7
avoided by optimizing the structure shape and installing the vor- T. Matsumoto, “Self-excited oscillation of a pretensioned cable roof with single
tex spoiler so as to prevent the large amplitude vibration of the curvature in smooth flow,” J. Wind Eng. Ind. Aerodyn. 34, 303–318 (1990).
8
structures caused by aero-elastic instability. A. Miyake, T. Yoshimura, and M. Makino, “Aerodynamic instability of sus-
pended roof models,” J. Wind Eng. Ind. Aerodyn. 42, 1471–1482 (1992).
9
H. Minami, Y. Okuda, and S. Kawamura, “Experimental studies on the flutter
ACKNOWLEDGMENTS behavior of membranes in a wind tunnel,” in Proceedings of the 4th Conference on
Space Structures (Thomas Telford, London, 1993), Vol. 4, pp. 935–945.
The authors acknowledge the financial support from the 10
R. Sygulski, “Stability of membrane in low subsonic flow,” Int. J. Non-Linear
National Natural Science Foundation of China (Grant No. Mech. 42, 196–202 (2007).
51878129), the Department of Finance of Jilin Province (Grant No. 11
Y. Uematsu and K. Uchiyama, “Aeroelastic behavior of an H.P.shaped sus-
JJKH20180424KJ), and the Development Program of Science and pended roof,” in Proceedings of IASS Symposium: Shells, Membrane and Space
Technology Bureau in Jilin City (Grant No. 201831787). Frames, Osaka (Elsevier Science Publishers BV, 1986), Vol. 2, pp. 241–248.
12
Y. Wu, X. Y. Sun, and S. Z. Shen, “Computation of wind–structure interaction
AUTHORS’ CONTRIBUTIONS on tension structures,” J. Wind Eng. Ind. Aerodyn. 96, 2019–2032 (2008).
13
W. Ding, Y. Uematsu, M. Nakamura, and S. Tanaka, “Unsteady aerodynamic
All authors contributed equally to this work. forces on a vibrating long-span curved roof,” Wind Struct. 19(6), 649–663 (2014).
The authors have no conflicts to disclose. 14
W. Ding and Y. Uematsu, “Large eddy simulation of unsteady aerodynamic
behavior of long-span vaulted roofs,” J. Zhejiang Univ., Sci., A 18(10), 793–806
DATA AVAILABILITY (2017).
15
C. J. Liu, X. W. Deng, and Z. L. Zheng, “Nonlinear wind-induced aerodynamic
The data that support the findings of this study are available stability of orthotropic saddle membrane structures,” J. Wind Eng. Ind. Aerodyn.
from the corresponding author upon reasonable request. 164, 119–127 (2017).
16
Y. Q. Li, L. Wang, Z. Y. Shen, and Y. Tamura, “Added-mass estimation of flat
REFERENCES membranes vibrating in still air,” J. Wind Eng. Ind. Aerodyn. 99(8), 815–824
(2011).
1 17
Q.-S. Yang and R.-X. Liu, “On aerodynamic stability of membrane structures,” Y. Zhou, Y. Q. Li, Z. Y. Shen, L. Wang, and Y. Tamura, “Numerical analysis
Int. J. Space Struct. 20(3), 181–188 (2005). of added mass for open flat membrane vibrating in still air using the boundary
2
X. Y. Sun, K. Arjun, and Y. Wu, “Investigation on wind tunnel experiment of element method,” J. Wind Eng. Ind. Aerodyn. 131, 100–111 (2014).
oval-shaped arch-supported membrane structures,” J. Wind Eng. Ind. Aerodyn. 18
Z. Q. Chen, Y. Wu, and X. Y. Sun, “Research on the added mass of open-
206, 104371 (2020). type one-way tensioned membrane structure in uniform flow,” J. Wind Eng. Ind.
3
X. Y. Sun, R. Yu, and Y. Wu, “Investigation on wind tunnel experiments of ridge- Aerodyn. 137, 69–77 (2015).
valley tensile membrane structures,” Eng. Struct. 187, 280–298 (2019). 19
P. A. Irwin and R. L. Wardlaw, “A wind tunnel investigation of a retractable
4
A. Kandel, X. Y. Sun, and Y. Wu, “Wind-induced responses and equivalent static fabric roof for the Montreal Olympic Stadium,” in Proceedings of 5th Interna-
design method of oval-shaped arch-supported membrane structure,” J. Wind Eng. tional Conference on Wind Engineering (Pergamon Press, CO, 1979), Vol. 1,
Ind. Aerodyn. 213, 104620 (2021). pp. 925–938.
5 20
S. Kawamura and E. Kimoto, “Aerodynamic stability criteria of one-way types B. V. Tryggvason, “Aeroelastic modeling of pneumatic and tensioned fabric
of hanging roofs in smooth uniform flow,” in Proceedings of 5th International structures,” in Proceedings of 5th International Conference on Wind Engineering
Conference on Wind Engineering (Pergamon Press, CO, 1979), Vol. 1, pp. 939–948. (Pergamon Press, CO, 1979), Vol. 1, pp. 1061–1072.

AIP Advances 11, 105201 (2021); doi: 10.1063/5.0062740 11, 105201-11


© Author(s) 2021

You might also like