Академический Документы
Профессиональный Документы
Культура Документы
Abstract
A review of existing and new considerations for the assessment of wind loads on low-rise
structures from wind tunnel simulation experiments is presented. Attention is given to the
aerodynamics of surface-mounted prisms, followed by a discussion of the atmospheric flow
characteristics near the surface under a variety of atmospheric and upwind terrain conditions.
In the final section, detailed recommendations and requirements are presented for the
appropriate conduct of these experiments.
The content of this article provides detailed justification for wind tunnel simulations based
primarily on the duplication of the turbulence intensities and small-scale turbulence of the
incident flow. In order to satisfy the latter requirement, it is recommended to utilize models of
low-rise buildings that have a scale not smaller than 1:50. Less emphasis should be placed on
duplication of the integral scale of the turbulence, as duplication of both scales requires
equality of the Reynolds number that cannot be achieved in wind tunnels.
Duplication of either the Jensen number ðH=zo Þ or the power-law exponent of the mean
velocity profile is insufficient to guarantee the proper duplication of the necessary turbulence
parameters.
For homogeneous upwind terrain and near-neutral stability, the pertinent turbulence
parameters can be obtained from basic principles without the use of empiricism. For terrain
with scattered upwind obstacles (Dx>15), the turbulence parameters can be obtained above
the blending height with the use of the ‘‘regional’’ roughness length. For a closer spacing of the
roughness obstacles (Dxo15), the displacement height should be included in the analysis to
obtain values of the turbulence parameters. For all situations, the simulation of the turbulence
in the atmospheric surface layer should be based on duplication of the pertinent flow
parameters above the blending height.
0167-6105/$ - see front matter r 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jweia.2003.09.021
ARTICLE IN PRESS
1628 H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649
Nomenclature
Comparison of either flow or pressure measurements requires that both the sampling rate
and record length be properly matched in model and full scale. Comparison of peak pressure
coefficients is only meaningful if they have the same level of probability.
r 2003 Elsevier Ltd. All rights reserved.
Keywords: Bluff body aerodynamics; Vortex formation; Pressure coefficients; Turbulence; Turbulence
scales; Atmospheric surface layer; Ideal, perturbed and complex terrain; Wind tunnel simulations
1. Introduction
2.1. Buffeting
Originally, wind engineers generally assumed that wind loads on structures were
primarily the result of the dynamic interaction of large-scale gusts with the
structures. Nevertheless, Bearman [1] pointed out that the incident flow could affect
the aerodynamic forces on prisms in two ways, by either ‘‘buffeting’’ or
‘‘interaction.’’ The buffeting forces arise directly from low-frequency fluctuations
ARTICLE IN PRESS
1630 H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649
whose wavelengths generally exceed the structure’s major dimensions. The surface
forces respond to these fluctuations as unsteadiness of the mean wind speed and its
direction. To the contrary, velocity fluctuations associated with small wavelengths
are considered less effective in producing peak aerodynamic forces. Experimental
results have demonstrated that the quasi-steady theory (direct consequence of the
buffeting process) can only be applied for the prediction of aerodynamic forces in
stagnation regions ðCpmean > 0Þ; while failing to predict these forces in separated
regions [2–4].
2.2. Interaction
The second way that the incident flow influences the aerodynamic forces is
through a complex interaction of this incident flow with flow separation, shear layer
development, vortex generation and flow reattachment. The experimental results
reported by Gartshore [5] and by Laneville et al. [6] clearly demonstrated the effect of
the small-scale turbulence of the incident flow for normal or bubble separation. An
increased small-scale turbulence content leads to smaller separation bubbles and
early reattachment. The increased deflection of the separated shear layer can easily
be explained with the aid of Euler’s equation:
DV% qV% V2 , 1
¼ þ grad V% curl V% ¼ g grad p: ð1Þ
Dt qt 2 r
The term V% curl V% provides the deflection of the shear layer. Increased vorticity
leads to a shorter separation bubble and earlier roll-up of this layer. This increase of
vorticity in the separated shear layer, in addition to Kelvin–Helmholtz instability,
can only be achieved with the presence of stream-wise vorticity in the incident flow,
because in purely two-dimensional flow the vorticity cannot be maintained and must
dissipate. The increased roll-up of the shear layer produces a strong vortex, which
leads to the development of a low-pressure peak beneath the vortex. Eventually, the
vortex is convected downstream, often preceded by a high-pressure peak [7, 8].
When only small-scale turbulence is present in the incident flow, the vortices are
convected downstream before they reach full maturity, or in other words before the
maximum peak pressure is created. An increase in large-scale turbulence allows the
vortices to attain maturity before being shed downstream away from the leading
edge. This results in peak suctions of larger magnitude and increased duration.
Consequently, both small- and large-scale turbulence play an important part in the
development of the peak pressures. The small-scale turbulence is primarily
responsible for the roll-up of the separated shear layer, while the large-scale
turbulence is important for the vortices to reach full maturity [8,9].
For normal separation, the vector V% curl V% lies on a plane normal to the
leading edge and normal to the vortex axis. For oblique separation, this vector has
two components. One component lies in the plane normal to the vortex axis,
providing the roll-up of the separated shear layer and vortex formation as discussed
for normal separation. The second component is parallel to the vortex axis giving
rise to the acceleration and velocity in the vortex core [10]. Together, the rotational
ARTICLE IN PRESS
H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649 1631
vortex flow and the axial velocity in the core of the vortex are capable of creating
extremely high suction pressures near upstream corners, particularly for flat- or low-
sloped roofs. Magnitudes of peak suction coefficients in excess of 20 have been
observed in the field and in the laboratory. As the local velocity becomes more
perpendicular to the leading edge (not less than 25 from the normal to the leading
edge), and this vector is directed more upwards (increased pitch), the following
observations were made [11,12]:
* The size of the vortex increases.
* The height and the length of the separated shear layer increases.
* The vortex is the farthest away from the leading edge.
* The peak suction occurs directly beneath the vortex axis.
* Under these conditions, peak suctions exhibit their largest magnitudes.
* Away from the leading edge past the vortex axis, the pressure increases rapidly as
the reattachment region is approached.
* The magnitude of the local velocity at the roof corner and the leading edge
controls the strength of the vortex.
These observational results can also be explained by expanding the V% curl V% term
in Euler’s equation (1).
Based on the above discussions, it is obvious that both the high- and the low-
frequency segments of the velocity spectra, and therefore the turbulence intensity
must be simulated, in order to adequately reproduce the surface pressures under
separated shear layers from laboratory model experiments. For comparison, the
turbulence integral scale represents the large-scale turbulence, while the small-scale
turbulence is evaluated with a parameter initially introduced by Melbourne [13]
S ¼ ½nSu ðnÞ=s2u ðsu =UÞ2 106 evaluated at n ¼ 10U=LB : ð2Þ
Here, LB is the characteristic building dimension, which for a flat roof low-rise
building is its height ðHÞ. The parameter, S; is appropriately based on the content of
the turbulence in the incident flow with a wavelength comparable to the thickness of
1
the separated shear layer. The latter is estimated for low-rise structures at 10 of its
height.
Experiments conducted during the early 1970s by Bearman and Trueman [14],
revealed that the pressure distribution on the side and base surfaces of two-
dimensional rectangular cylinders is controlled by the proximity of the vortex
formation and the strength of these vortices. With this in mind, extensive wind
tunnel experiments of surface-mounted prisms immersed in deep turbulent shear
layers have revealed the following general results:
* Mean and fluctuating base pressure coefficients on square-section cylinders (H>W
or L) are primarily controlled by S and to a lesser degree by the turbulence
intensity. For low values of S the separated shear layer is not reattached which
ARTICLE IN PRESS
1632 H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649
Based on these observations, the ‘‘interaction’’ pressures (mean, rms and peak) on
surface-mounted prisms do not follow a simple universal pattern [20]. Significant
variations must be expected depending on the prism type (H>L,W or HoL,W),
prism geometry (e.g. leading edge discontinuity and rounded leading edge), prism
size (flow attached or not reattached), azimuth angle and the characteristics of the
incident turbulent flow (turbulence intensity and scales). A laboratory model
ARTICLE IN PRESS
H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649 1633
experiment to assess wind loads on scaled models is not simply a matter of placing a
model in a wind tunnel and take measurements. Such an experiment is a multi-
variable problem requiring careful preparation.
To carry out a successful wind tunnel model experiment, it behooves the wind
engineer to have a thorough understanding of the atmospheric flow in the surface
layer below 100 m. The discussion will first concentrate on the relatively simple and
ideal situation of stationary and neutrally stable flow over flat open terrain with a
uniform surface roughness. In reality, this idealized flow is seldom encountered as
turbulence of different origins other than shear-generated, modifies this ideal surface
layer flow. The presence of turbulence generated by buoyancy, large-scale eddies
from the convective boundary layer (CBL), and also those generated over perturbed
and complex terrain, affect the wind loads on low-rise structures appreciably.
Under these ideal conditions the mean velocity profile is well represented by the
classical logarithmic wind profile
UðzÞ ¼ ðU =KÞlnðz=zo Þ: ð3Þ
Here, z is the roughness length, U is the surface friction velocity defined as
o
ðto =rÞ0:5 (to being the surface shear stress) and K=0.4. Also, for surface-layer flow
with purely mechanically produced turbulence Aa ¼ sa =U ¼ 2:4; 1:9 and 1.25 for
a=u,v, and w, respectively [21]. By combining the velocity profile (3) with the
turbulence ratios sa =U expressions for the turbulence intensities are
sa =U ¼ KAa =lnðz=zo Þ: ð4Þ
A general interpolation expression for the logarithmic velocity spectra under ideal
conditions takes the following form:
nSa ðnÞ Af
¼ : ð5Þ
ðU Þ2 ð1 þ Bf a Þb
Here, n is the cyclic frequency and f is the dimensionless frequency representing the
ratio of the height z to the wavelength U/n. In the inertial subrange (high-frequency
range) all velocity spectra plotted in a log–log format, must collapse on a straight
line with a 23 slope, and in the low-frequency range the slope should be +1. With
these conditions and the requirement that the integral of (5) must be equivalent to
ðsa =U Þ2 ; the numerical values of coefficients A and B can be calculated without
resorting to high-frequency velocity measurements [22].
The small-scale turbulence parameter, S, can be obtained by evaluating spectral
functions (5) at the frequency n=10U/H. For a flat-roof structure the characteristic
length scale is H, S and the turbulence intensities must be evaluated at this height,
ARTICLE IN PRESS
1634 H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649
and thus S may be obtained on the evaluating (5) at f ¼ 10: The expected integral
scales can be also derived from the velocity spectra, but now at zero frequency to
obtain
Lax ¼ ðU=4ÞSa ð0Þ=s2a : ð6Þ
Consequently, the flow characteristics for ideal surface layer flow can be obtained
by simply specifying the roughness length, which can be estimated from Wieringa
[23]. However, for terrain whose roughness length exceeds 10 cm, low-rise buildings
will be partially or completely immersed in the roughness layer ðh > 20zo Þ: This layer
is not part of the surface layer, so that expressions (3)–(6) do not apply. Also, the
displacement height, d, can no longer be ignored and must be included in (3) to
obtain a realistic value for the roughness length of the upwind terrain.
layer, whose energy is added at the low-frequency end of the u and v spectra.
Consequently their variances scale primarily with the stability parameter for the
CBL, zi/L. Similarly, the u and v velocity spectra consist of a mixture of velocity
fluctuations whose origin is mechanical at high frequencies, and convective in the
low-frequency range. With the presence of the low-frequency convective turbulence,
the spectral densities in the low-frequency range exceed those of the spectra for ideal
flow without any heat flux [21].
When large-scale turbulence is present, as is the case under these convective
conditions, the integral scales are several times larger than those obtained with (6). In
this case the integral scale must be obtained experimentally from the autocorrelation
function of the velocity fluctuations:
Z N
a
Lx ¼ U Ra ðtÞ dt: ð8Þ
0
Wind engineers have obtained values for the integral scales by matching the observed
velocity spectra with the von Karman spectral models [22]. If this matching is limited
to the high-frequency asymptote of the spectrum, the results will provide low
estimates in the case large-scale convective turbulence is present. Under these
conditions, the S parameter must be evaluated from experimentally obtained velocity
spectra.
Of course, wind engineers have to deal with flow over terrain that is not
homogeneous, but instead exhibits various degrees of upwind roughness. Under
these conditions, the relationships between the mean flow and the turbulence are not
so straightforward. With the presence of only a few isolated obstructions such as
trees, dwellings and small hills, homogeneity should not automatically be assumed.
As the terrain conditions increase in complexity, the turbulence is no longer in
equilibrium with the mean wind profile and parameters derived from the latter, such
as the roughness length and the friction velocity are inadequate to describe the
turbulence.
When air moves from a homogeneous smooth surface to a homogeneous rough
terrain or vice versa, an internal boundary layer develops as the flow adapts itself to
the new homogeneous surface. At the interface of the two layers the wind profile
develops a kink when presented in semi-logarithmic form. The internal boundary
layer grows slowly at a rate of approximately 1:100, while the boundary between the
transition layer and the still unmodified air grows at a rate of about 1:10 [24]. For a
terrain change from rough to smooth, it takes a long time before the turbulence in
the internal boundary layer becomes in equilibrium with the underlying terrain,
while in contrast the mean velocity wind profile adjusts itself much faster [25,26].
Only in very rare situations is the mean wind direction normal to the line delineating
the roughness change.
Frequently, one encounters terrain that is relatively flat with sparsely distributed
roughness elements such as buildings, groups of trees, etc. As long as the average
ARTICLE IN PRESS
1636 H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649
distance between these elements is at least 20 times their height [27], surface wind
profiles outside the wake of the individual elements represent the terrain between
them. Profile measurements near the surface, say below 10 m, give generally rise to
smaller roughness lengths as compared to those obtained from mean velocity
measurements at higher elevations. Profile parameters below the profile ‘‘kink’’ are
then referred to as ‘‘local’’ parameters. In this layer small-scale high-frequency
vertical velocity fluctuations adapt quickly to the new terrain, and parameters
associated with these fluctuations are referred to as ‘‘local’’ parameters, e.g.
sw ; uw; Lwx and the inertial subranges of all velocity components.
In contrast, large-scale low-frequency velocity fluctuations are quite persistent and
do not adapt quickly to the new terrain [26]. Under these conditions the horizontal
turbulence is not in equilibrium with the underlying terrain. Consequently, the
horizontal turbulence intensities su =U and sv =U and their integral scales Lux and Lvx
observed in the surface layer below the profile kink are referred to as ‘‘regional’’
parameters. In order to evaluate this condition, one can obtain the turbulence
intensities once the local roughness length is evaluated with (4). If the observed
values exceed those calculated from (4), the upwind terrain must be classified as
nonhomogeneous. When only observations above the ‘‘kink’’ are available, the local
roughness length may be estimated from sw =U (local parameter) with Aw=1.25 from
z0 ¼ exp½ln z KAw =ðsw =UÞ : ð9Þ
Above the velocity profile ‘‘kink’’ the observed flow parameters are representative
of the terrain over a large upwind distance that include the scattered large surface
obstacles as part of the overall terrain roughness for this long upwind fetch.
Provided the large obstacles are distributed not too inhomogeneously, velocity
observations in this layer give rise to a ‘‘regional’’ roughness length, also referred to
as the ‘‘effective’’ roughness length. In the following sections the flow parameters
from meteorological towers at several locations with relatively open terrain will be
discussed.
transition layer between the internal boundary layer associated with the 250 m
upwind grassland and the undisturbed boundary layer associated with the orchards
further upwind. Consequently, the profile roughness length of 1.5 cm neither
represents the grass nor the orchard.
3.3.3. WERFL
At the Wind Engineering Research Field Laboratory (WERFL) at Lubbock,
Texas, one can study simultaneously flow parameters in the surface layer
(instrumentation at 1.0,2.5,4.0,10.0, and 49.0 m) and the pressures on an
experimental building. The terrain at this facility is flat and partially covered with
grass and cotton during the summer. A single-story residential region is located
about 1 km upwind from the tower location in a westerly direction [30]. Careful
analysis of the available wind profiles resulted in an average local roughness length
of 0.7 cm. Although no direct measurements of the thermal stability were made,
which severely limited the profile analysis, a study of the wind profiles clearly
exhibited the effect of the stability. Also, the horizontal turbulence intensities showed
the effect of the unstable flow.
Two sets of data (20-min averages) were available both starting at midnight
(slightly stable thermal stratification) and terminating the following afternoon. For
ARTICLE IN PRESS
1638 H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649
one set obtained during the winter of 1992, both horizontal turbulence intensities,
observed at z=4 m, exhibit average values during the entire period of su =U ¼ 19:3%
and sv =U ¼13.8%. The second set started with about the same values for these
intensities during the night of June 19, 1992. However both intensities increase
dramatically during the following daytime hours. Especially, sv =U doubled
gradually in magnitude to about 30% during the morning hours, as evidenced by
the presence of convective low-frequency turbulence.
The u; v turbulence intensities based on zo ¼ 0:7 cm; and calculated from (4) are
15.1% and 12%, respectively. Both values are somewhat less than the observed
values under slightly stable conditions (19.3% and 13.8%). The difference of these
results must probably be attributed to the residual low-frequency turbulence
originating from the upwind residential areas. The observed values of the turbulence
integral scales show similar differences between those obtained from records under
convective conditions and those obtained from records with slightly stable
conditions (Table 1). Although the velocity spectra were based on 20-min records,
the effect of the convective turbulence is readily evident even for the spectra at low
heights of 2.5 and 4.0 m.
The next terrain category being discussed is that over terrain with orographic
features such as rolling hills, mountain ridges etc., also referred to as complex
terrain. As was discussed before, the mean wind and turbulence characteristics over
uniform and flat terrain depend solely on the wind speed, the surface roughness and
atmospheric stability. Over complex terrain it is not the Reynolds stresses but the
induced local pressure gradients that control the flow. For instance, the flow in the
wake of a mountain ridge decelerates considerably and requires a distance of more
than 80 times its height for the velocity to recover its undisturbed state. Also the
turbulence is greatly enhanced, and increased variances persist for longer downwind
distances than the recovery distance for the mean velocity profile. Under convective
conditions, the turbulence for complex terrain bears the signature of two low-
frequency sources, one scaling with zi =L (convective turbulence) and the other whose
length scale is associated with the upwind complex terrain features.
Since the vertical velocity fluctuations contain relatively small amounts of low-
frequency energy in comparison with the horizontal ones, they tend to be in
equilibrium with the local terrain including for flow over complex terrain with
Table 1
Observed and theoretical-calculated integral scales (m) at WERFL (z=10 m)
negligible heat flux. The observed vertical velocity spectra under the latter conditions
resemble those over uniform terrain provided the upwind terrain is not too complex.
Its turbulence intensity can then be used to evaluate the local roughness length with
(9). In contrast, the horizontal turbulence intensities depend mostly on the large
‘‘horizontal’’ eddies, generated in the wake of the orographic features, maintaining
their low-frequency fluctuations for much longer time than the high-frequency
fluctuations. The excess of low-frequency turbulence in comparison with that
of ideal terrain with negligible heat flux appears to be the largest for the lateral
component.
With the basics of the aerodynamics of bluff bodies and the detailed characteristics
of the atmospheric surface layer discussed in the previous sections, one can now
approach the wind tunnel simulation process with some confidence. The important
ARTICLE IN PRESS
1640 H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649
element in the section on bluff body aerodynamics, is the roll played by the
turbulence (small scale and large scale) in the formation of vortices under separated
shear layers. In turn, these vortices are responsible for the development of the large
suction peaks near the leading edges of prisms from which the shear layers separate.
Consequently, the turbulence characteristics of the incident flow at the height of
these leading edges require careful simulation. For atmospheric flows at high
Reynolds numbers, the separation of the large and the small wavelengths of the
velocity fluctuations in the frequency domain is farther than for the wind tunnel
flows at much lower Reynolds numbers. This inequality is the most difficult problem
to overcome for the wind tunnel flow simulation of the atmospheric surface layer for
the assessment of wind loads on low-rise structures.
4.1. General
The simulation experiments are best carried out in long test section wind tunnels
for the generation of deep boundary layers providing scaled facsimiles of the
atmospheric boundary layer (ABL) flow. By varying the surface roughness of the
upstream fetch on the tunnel floor, and with the use of boundary layer trips, vortex
generators and spires placed at the test section entrance, boundary layers with a
depth of 0.5–1.5 m can be developed. Under neutral conditions the ABL depth varies
between 1000 and 2000 m, requiring a model scale of approximately 1:1000.
However, for wind engineering applications the essential depth of the neutral ABL is
assumed to be limited to the height where the wind speed reaches a maximum value
(gradient height), varying between 275 and 550 m [35]. Based on this definition,
model scales of the order of 1:400 would be appropriate, but these scales are too
small for models of low-rise structures, making it difficult to model structural details
and to study their applied aerodynamic forces.
In order to overcome this problem and to better simulate the scale of the high-
frequency turbulence relative to the model size, geometric model scales of the order
of 1:50 are more suitable. With this scale ratio, only the lower part of the surface
layer flow can be simulated. The classical requirements for any model experiment to
simulate the aerodynamic forces are equality of Reynolds number, the Richardson
number, which can only be achieved by proper scaling of the Monin–Obukhov
stability length, and geometric similarity. Most boundary layer wind tunnels are not
equipped to simulate thermal stratification, and certainly could not simulate the
CBL at a scale ratio of 1:50 and simultaneously model the large convective eddies.
Consequently, the flow simulation approach taken here is appropriate for tests in
surface layer flows with neutral thermal stratification. The mean-flow Reynolds
number equality can be relaxed for sharp-edged models, provided it does not fall
below 50,000. Lower Reynolds numbers are not recommended because they lead to
the reduction of the incident small-scale turbulence that would fail to initiate the
proper vortex development associated with the interaction process between flow and
model.
Theoretically, it would be required that the geometric scaling ratio for the model is
equivalent to the ratios of the roughness length and the integral scales. Correct
ARTICLE IN PRESS
H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649 1641
simulation of the integral scale for flow over ideal terrain below 10 m would
require a geometric scaling ratio of the order of 1:300. The model size for a 4.5 m
high single story home would be 1.5 cm. As noted before, these small models are not
of sufficient size to accommodate a large number of pressure taps and are too small
to allow acceptable levels of small-scale turbulence in the incident flow at roof
height.
Some modelers believe that model/full-scale equality of H=zo (Jensen number) is
sufficient to simulate the flow in the surface layer for ideal terrain, and obtain reliable
estimates for the wind loads. However, this condition is not sufficient to guarantee
equality of the pressure coefficients unless the turbulence Reynolds number is
duplicated. For a full-scale Jensen number, H=zo ; there are many combinations of H
and zo that would lead to a model ratio with the same full-scale value. Of course,
only one combination is the correct one providing the same model/full scale
turbulence intensity. Tennekes and Lumley [36] have shown that the ratio of the
large to small turbulence scales, L=l; varies as ðRT Þ3=4 ; where RT ¼ zU =v: Under
strong wind conditions, the ratio L=l for the atmosphere is at least three orders of
magnitude larger than for the simulated flow. Therefore, if the geometric scaling is
based on scaling of the integral scale, the inequality of RT will lead to suppression of
the small-scale turbulence relatively to the large-scale turbulence.
For field application of the small-scale turbulence parameter, S, (2), the
wavelength H/10 will generally fall in the inertial subrange of the velocity spectrum,
while the corresponding model wavelength may well fall in the viscous subrange,
resulting in a much lower value of S. An increase of the model values of this
parameter can be achieved by increasing the model size and/or the upstream floor
roughness. This ad hoc approach has resulted in improved simulation of the
fluctuating pressures under the corner vortexes on the roof of a 1:50 scale model of
the WERFL experimental building [37]. This feat was accomplished with the use of
small spires directly upstream of the model location [15], but has the disadvantage of
breaking up the large-scale eddies. However, Stathopoulos [38] concluded that exact
matching of the integral scale could be relaxed by a factor of two without affecting
the magnitudes of the pressure coefficients too much. Other experiments conducted
by Tieleman et al. [39] showed that the model integral scale should not be smaller
than 20% of the target value. However, both conclusions were reached without
investigating the effects on the duration of the pressure peaks.
With the use of increased floor roughness and small spires, improved duplication
of the turbulence intensities and small-scale turbulence is achieved at the expense of
the integral scale and the surface roughness length, zo : Inversely, concentrating on
the duplication of the latter two parameters does not provide a satisfactory
duplication of the turbulence intensities and S: Consequently, no simple simulation
technique has been devised to simultaneously duplicate the large and the small scales
of the velocity fluctuations together with the mean wind profile. This problem
becomes more acute when trying to simulate large lateral turbulence intensities as
observed over complex terrain and under convective conditions, when the direction
fluctuations are simply too large to be reproduced with the current wind-tunnel
simulation techniques.
ARTICLE IN PRESS
1642 H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649
Table 2
Roughness lengths and turbulence intensities at z=4 m for homogeneous surfacesa
(4) the standard deviations of the velocity components upwind of the roughness
change can be calculated.
Together with the assumption that these standard deviations remain unchanged
downstream of the roughness change, and with the evaluation of the mean velocity
over the new terrain at the same height, conservative estimates of the turbulence
intensities in the internal boundary layer can be calculated.
For perturbed terrain with widely separated obstacles ðDx > 15 HÞ; the logarithmic
mean wind profile exhibits a ‘‘kink’’ at about twice the average obstacle height, H
[40]. Here, Dx is the average horizontal distance between the obstacles. Below the
‘‘kink,’’ and outside the wake of any of the upwind obstacles, the profile is
representative of the local terrain, while above the ‘‘kink’’ it represents the regional
terrain, constituting the integral effects of these obstacles [26]. Low-rise structures of
height, H, of 10 m or less are located in the region below the profile kink. The ‘‘local’’
flow parameters in this layer can be obtained as before by estimation of the local
roughness length [23], while the ‘‘regional’’ roughness length should be used to
obtain the horizontal turbulence intensities. For this type of terrain, the ‘‘regional’’
roughness applies to profiles above 2 H, with H being the average height of the large
roughness elements. The ‘‘regional’’ or ‘‘effective’’ roughness length for terrain with
overlapping wakes ðDxo15 HÞ can be obtained from Ref. [41, Table 2], and applies
to profiles above H þ 1:5Dx:
For the first eight terrain categories of Table 2, the blending height ð20zo Þ is small
in comparison with the 4-m height of a one-story low-rise building. For the rougher
terrain categories, 10–14 of this table and categories 5–8 of the Davenport roughness
classification ([41, Table 3]) a low-rise building is completely embedded in the
roughness layer. Flow parameters above the blending height for these situations can
be obtained with the appropriate value of zo and the ideal-terrain expressions (2)–(6).
The flow regime below the blending height can only be established by a wind tunnel
experiment with the surface roughness around the model carefully modeled, and the
flow above the blending height obtained with the properly selected roughness length.
The upper limit of the roughness length of the eighth roughness class of Table 2 is
zo ¼ 0:09 m. The corresponding turbulence intensities at z=4 m are su =U=25.3%
and sv =U ¼ 20%: These are very high values to achieve with current wind tunnel
simulation techniques at a 1:50 scaled height of 8 cm above the tunnel floor.
Similarly, for complex terrain the stream wise and lateral turbulence intensities and
their integral scales are simply too large to be reproduced in currently available wind
tunnels.
modification also tends to reduce the size of the model integral scales. To increase the
size of these scales, eliminate the small spires and increase the size of the spires at the
tunnel’s test section entrance. Once the turbulence intensities of the simulated flow
duplicate the field data as close as possible, the observed wind loads should be close
facsimiles of what to expect in full scale.
For a low-rise building located in a rough terrain for which wake-interference flow
exists (Dxo15 H), the roughness of height H in the immediate neighborhood should
be carefully modeled to a distance of several hundred meters, depending on the
model height. Beyond that distance, uniform roughness must be employed to
simulate the flow above this roughness at height 2H and higher. The parameters for
this flow must be obtained with the use of the proper roughness length, obtained
from the Wieringa [23,41] or the Davenport [35] roughness classifications. If larger
buildings and obstacles are present, their modeling should extend to larger upstream
distances, and it may be next to impossible to simulate the high turbulence–intensity
flows above these roughness elements. Similarly, if the upwind terrain is complex but
the local terrain is open, the horizontal turbulence intensities are exceptionally high
and would be difficult to simulate with current wind tunnel simulation techniques.
In Fig. 1 wind load coefficients are shown for seven wind tunnel roughness
configurations of continually increased roughness, starting with an open tunnel
without any floor roughness or spires. The flow parameters at z=8 cm, correspond-
ing to the scaled (1:50) roof height of the WERFL experimental building [44], are
presented in [45, Table 1]. The upper limit of the turbulence intensity is
approximately 20%. The lower limit of 7% corresponds to category D terrain
[42]. The remaining roughness configurations with su =U varying from 14.3% to
19.3% are representative of open terrain for the Wieringa [23] open-terrain classes of
#3 (flat snow field) through #8 (low mature agricultural crops). The results presented
in Fig. 1 show clearly that no single pressure coefficient can represent the ASCE 7-95
20.0
Azimuth: 25degr.
8
-(Cpmean)avg.
15.0 (Cprms)avg 4
Load Coefficients
-(Cppeak)avg
2 3
-(Cppeak)max
7
10.0
1
6
5.0
0.0
5.0 10.0 15.0 20.0
Su/U, %
Fig. 1. Variations of the load coefficients under the corner vortex with turbulence intensity. Numbers refer
to the different roughness configurations.
ARTICLE IN PRESS
1646 H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649
5. Conclusions
* For flow over ideal terrain (flat, smooth and uniform) without heat flux, all flow
parameters can be obtained from the basic expressions by prescribing the surface
roughness length, zo :
* For perturbed terrain the flow parameters above the blending height can be
obtained once the roughness length for this layer has been established from
roughness classification tables.
* Duplication of the high turbulence intensities over very rough and complex
terrain is difficult to achieve with current simulation techniques.
* Concentrating on the duplication of the horizontal turbulence intensities and the
small-scale turbulence parameter at those heights where the wind loads are being
measured on the model can attain the best wind tunnel simulation of the
fluctuating pressures.
References
[1] P.W. Bearman, Turbulence effects on bluff body mean flow, Proceedings of the Third National
Conference on Wind Engineering Research, Gainesville, FL, 1978, pp. 265–272.
[2] H. Kawai, Pressure fluctuations on square prisms—applicability of strip and quasi-steady theories,
J. Wind Eng. Ind. Aerodyn. 13 (1983) 197–208.
[3] C.W. Letchford, R.E. Iverson, J.R. McDonald, The application of the quasi-steady theory
to full scale measurements on the Texas Tech Building, J. Wind Eng. Ind. Aerodyn. 48 (1993)
111–132.
ARTICLE IN PRESS
1648 H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649
[4] H.W. Tieleman, M.R. Hajj, Pressures on a flat roof—application of quasi-steady theory, Proceedings
of the 10th ASCE Conference on Engineering Mechanics, University of Colorado, 21–24 May 1995,
pp. 557–560.
[5] I.S. Gartshore, The effect of free stream turbulence on the drag of rectangular two-dimensional
prisms, Boundary Layer Tunnel Laboratory Report, BLWT-4-73, University of Western Ontario,
London, Canada.
[6] A. Laneville, I.S. Gartshore, G.V. Parkinson, An explanation of some effects of turbulence on bluff
bodies, Proceedings of the Fourth International Conference of Wind Effects on Structures, London,
UK, pp. 333–341.
[7] C.W. Letchford, Simultaneous flow visualization and pressure measurements on the Texas Tech
building, Proceedings of the Ninth International Conference on Wing Engineering, New Delhi, India,
pp. 524–535.
[8] P.J. Saathoff, W.H. Melbourne, Effects of free-stream turbulence on surface pressure fluctuations in a
separation bubble, J. Fluid Mech. 337 (1997) 1–24.
[9] Q.S. Li, W.H. Melbourne, An experimental investigation of the effects of free-stream turbulence on
stream wise surface pressures in separated and reattaching flows, J. Wind Eng. Ind. Aerodyn. 54/55
(1995) 313–323.
[10] O.L. Rediniotis, N.T. Hoang, D.P. Telionis, Multi-sensor investigations of delta-wind high-alpha
aerodynamics, Proceedings of the 29th Aerospace Sciences Meeting, Reno, Nevada, AIAA-91-0735,
1991.
[11] R. Marwood, C.J. Wood, Conical vortex movement and its effect on roof pressures, J. Wind Eng.
Ind. Aerodyn. 69–71 (1997) 589–595.
[12] F. Wu, P.P. Sarkar, K.C. Mehta, Z. Zhao, Influence of incident wind turbulence on pressure
fluctuations near flat-roof corners, J. Wind Eng. Ind. Aerodyn. 89 (2001) 403–420.
[13] W.H. Melbourne, Turbulence effects on maximum surface pressures—a mechanism and possibility of
reduction, Wind Engineering, Vol. 1, Pergamon Press, Oxford, 1980, pp. 541–551.
[14] P.W. Bearman, D.M. Trueman, An investigation of the flow around rectangular cylinders, Aeronaut.
Q. 23 (1972) 229–237.
[15] H.W. Tieleman, Pressures on surface-mounted prisms: the effects of incident turbulence, J. Wind Eng.
Ind. Aerodyn. 49 (1993) 289–300.
[16] H.W. Tieleman, D. Surry, J.-X. Lin, Characteristics of mean and fluctuating pressure coefficients
under corner (Delta wing) vortices, J. Wind Eng. Ind. Aerodyn. 52 (1994) 263–275.
[17] J.-X. Lin, D. Surry, H.W. Tieleman, The distribution of pressure near roof corners of flat roof
buildings, J. Wind Eng. Ind. Aerodyn. 56 (1995) 235–265.
[18] W.H. Melbourne, Turbulence and leading edge phenomenon, J. Wind Eng. Ind. Aerodyn. 49 (1993)
45–64.
[19] H.W. Tieleman, D. Surry, K.C. Mehta, Full/model-scale comparison of surface pressures on the
Texas Tech experimental building, J. Wind Eng. Ind. Aerodyn. 61 (1996) 1–23.
[20] H.W. Tieleman, R.E. Akins, The effect of incident turbulence on the surface pressures of surface-
mounted prisms, J. Fluids Struct. 10 (1996) 367–393.
[21] A.H. Panofsky, J.A. Dutton, Atmospheric Turbulence, Wiley-Interscience, New York, 397pp.
[22] H.W. Tieleman, Universality of velocity spectra, J. Wind Eng. Ind. Aerodyn. 56 (1995)
55–69.
[23] J. Wieringa, Representative roughness parameters for homogeneous terrain, Boundary-Layer
Meteorol. 63 (1993) 323–363.
[24] E.W. Peterson, N.E. Busch, N.O. Jensen, J. Hfjstrup, L. Kristensen, E.L. Peterson, The
effect of local terrain irregularities on the mean wind and turbulence characteristics near the ground,
WMO-No. 510, Symposium on BL Physics Applied to Air Pollution, Norrk . oping,
. 1978,
pp. 45–50.
[25] H.A. Panofsky, D. Larko, R. Lipschutz, G. Stone, E.G. Bradley, A.J. Bowen, J. Hfjstrup, Spectra of
velocity components over complex terrain, Q. J. R. Meteorol. Soc. 108 (1982) 215–230.
[26] A.C.M. Beljaars, P. Schotanus, F.T.M. Nieuwstadt, Surface layer similarity under nonuniform fetch
conditions, J. Climate Appl. Meteorol. 21 (1983) 1800–1810.
ARTICLE IN PRESS
H.W. Tieleman / J. Wind Eng. Ind. Aerodyn. 91 (2003) 1627–1649 1649
[27] J. Wieringa, Estimation of meso scale and local-scale roughness for atmospheric transport modeling,
in: C. de Wispelaere (Ed.), Air Pollution Modeling and its Application, Plenum, New York, 1981,
pp. 279–295.
[28] A. Korrell, H.A. Panofsky, R.J. Rossi, Wind profiles at the Boulder Tower, Boundary-Layer
Meteorol. 22 (1982) 295–312.
[29] H.A. Panofsky, Vertical variations of roughness length at the Boulder atmospheric observatory,
Boundary-Layer Meteorol. 28 (1984) 305–308.
[30] M.L. Levitan, K.C. Mehta, Texas Tech field experiments for wind loads part II: meteorological
instrumentation and terrain parameters, J. Wind Eng. Ind. Aerodyn. 43 (1992) 1577–1588.
[31] ESDU 85020, Characteristics of the atmospheric turbulence near the ground, ESDU International,
London, UK, 1993.
[32] J.A. Dutton, H.A. Panofsky, D. Larko, H.N. Shirer, G. Stone, M. Vilardo, Statistics of wind
fluctuations over complex terrain, US Department of Energy Report DOE/ET/20560-1 UC-60, 1970,
142pp.
[33] H.A. Panofsky, C.A. Egolf, R. Lipschutz, On characteristics of wind direction fluctuations in the
surface layer, Boundary-Layer Meteorol. 15 (1978) 439–446.
[34] H.W. Tieleman, R.E. Akins, P.R. Sparks, A comparison of wind tunnel and full-scale wind pressure
measurements on low-rise structures, J. Wind Eng. Ind. Aerodyn. 8 (1981) 3–19.
[35] A.G. Davenport, A rationale for determination of design wind velocities, Proc. ASCE 86 (1960)
39–66.
[36] H. Tennekes, J.L. Lumley, A First Course in Turbulence, MIT Press, Cambridge, MA, USA, 1972.
[37] H.W. Tieleman, T.A. Reinhold, M.R. Hajj, Importance of turbulence for the production of surface
pressures on low-rise structures, J. Wind Eng. Ind. Aerodyn. 69–71 (1997) 519–528.
[38] T. Stathopoulos, D. Surry, Scale effects in wind tunnel testing of low buildings, J. Wind Eng. Ind.
Aerodyn. 13 (1983) 313–326.
[39] H.W. Tieleman, M.R. Hajj, T.A. Reinhold, Wind tunnel simulation requirements to assess wind
loads on low-rise buildings, J. Wind Eng. Ind. Aerodyn. 74–76 (1998) 675–685.
[40] A.C.M. Beljaars, The derivation of fluxes from profiles in perturbed areas, Boundary-Layer
Meteorol. 24 (1982) 35–55.
[41] J. Wieringa, Updating the Davenport terrain roughness classification, J. Wind Eng. Ind. Aerodyn. 41
(1992) 357–368.
[42] ASCE 7-95, Standard, Minimum design loads for buildings and other structures, American Society of
Civil Engineers, New York, 1995.
[43] J. Lieblein, Efficient methods of extreme-value methodology, NBSIR Report, US Department of
Commerce, National Institute of Standards and Technology, 1974.
[44] M.L. Levitan, K.C. Mehta, Texas Tech field experiments for wind loads Part I: Building and pressure
measuring system, J. Wind Eng. Ind. Aerodyn. 41–44 (1992) 1565–1576.
[45] H.W. Tieleman, M.R. Hajj, T.A. Reinhold, Pressure characteristics for separated flows, in: A. Larsen,
G.L. Larose, F.M. Livesey (Eds.), Wind Engineering into the 21st Century, Balkema, Rotterdam,
1999.
[46] L. Tian, Analysis of time-varying characteristics of simulated turbulence in wind tunnel, Masters
Thesis, Engineering Mechanics Department, Virginia Technology, available: http://www.lib.vt.edu
On Line Resources, VT electronic theses and dissertations, ‘‘Search VT ETD’s, enter ‘‘Tian’’ and
press seek.