Couchesne Laneville1982
Couchesne Laneville1982
Couchesne Laneville1982
net/publication/239400417
CITATIONS READS
47 1,020
2 authors, including:
Andre Laneville
Université de Sherbrooke
85 PUBLICATIONS 1,689 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Andre Laneville on 05 January 2016.
An Experimental Evaluation of
J. Courchesne
Graduate Student.
Drag Coefficient for Rectangular
A. Laneville Cylinders Exposed to Grid
Turbulence
Associate Professor.
Introduction
Turbulence contained in the atmospheric boundary layer, scale of turbulence on the mean drag coefficients (and on the
influences the wind loads experienced by buildings and galloping oscillations [5]) for two-dimensional rectangular
structures in two ways. As approaching the structure, tur- cyhnders at zero incidence with H/D = 0.5, 1 and 2.
bulence induces unsteady buffeting loads and interacts in a Nakamura et al. [6], in a wider range of longitudinal lengths
complex fashion with the flow around the body which leads to (0.1 < H/D < 3), also observed no integral scale effects for a
changes in the mean and unsteady flow patterns. In order to similar range of scale ratios. On the other hand, Nakamura et
estimate the magnitude of these effects, special wind tunnels al. [6] and Laneville et al. [4] observed that the mean drag
are designed to produce thick boundary layers, but the coefficient and the mean base pressure coefficient were largely
resultant simulation is restricted as to the generated scale of influenced by the intensity of turbulence and the "afterbody
turbulence and as to the Reynolds number. This paper reports length" H/D. Petty [7], using Lee's results [8] and Allen and
an investigation of the relative importance of the turbulence Vincenti's blockage correction, confirmed these results. In
characteristics and is concerned with two of the independant more recent tests, Laneville and Williams [13] investigated the
variables describing the neutral boundary layers: the intensity effect of large scale turbulence (6 < Lx/D < 18) on the base
{u'/if) and the macroscale (Lx) of turbulence. The effect of pressure coefficient of 2-D rectangular cylinders exposed to a
these two variables on the dependant variable, the mean (time fixed intensity of turbulence. With small models mounted in a
average sense) drag coefficient, is examined for different large test section (blockage less than 1 percent), they con-
geometries of two-dimensional rectangular cyhnders. cluded that the integral scale of turbulence did not measurably
influence the base pressure coefficient and consequently the
Previous Investigations drag coefficient of the models. In view of the limited data
available, an experimental program was set up to
Schubauer and Dryden [1] were the first to notice that systematically investigate the effect of grid turbulence (with
turbulence increased the mean drag of horizontal flat plates. well defined u'/U and Lx) on the mean drag coefficient of
Vickery [2] measured a thirty percent reduction in the drag two-dimensional rectangular cylinder.
coefficient of a two-dimensional square cylinder exposed to
turbulence superimposed on an otherwise smooth flow.
Bearman [3] observed that the mean base pressure of normal
square and circular plates in turbulent grid flow (u'/U = Experimental Program
0.083 and Lx/D = 0.375) was considerably lower than that All the tests were conducted in a closed jet open circuit type
measured in smooth flow (u'/U = 0.002). Bearman (N.P.L.) wind tunnel with two ft (61 cm) octogonal test
suggested that the principal reason for this was that, com- section. The turbulence intensity in the enipty tunnel was
pared to smooth flow, there was extra entrainment of fluid measured as 0.6 percent.
out of the wake resulting from the mixing of the near wake The models for this investigation had D ranging from 1.25
with the free stream turbulence. In support of this argument, in. (3.175 cm) to 2.5 in. (6.35 cm) and H/D from 0.3 to 3.0.
the base pressure coefficients were shown to correlate well The models were 18 in. (46 cm) long, with end plates, and
with the turbulence parameter {u'/U) (Lx^/A). Laneville et located in the central core of the flow. Care was taken to
al. [4], in the range 0.5 < Lx/D < 5, observed no effect of the machine sharp corners. Dummy spacers with the same cross
section as the models were mounted between the wind tunnel
Contribuled by the Fluids Engineering Division for publication in the walls and the end plates to insure the uniformity of the flow-
JOURNAL OF F L U I D S ENGINEERING. Manuscript receivetJ by the Fluids around the tested models. The clearance between the end
Engineering Division, Septeinber 22, 1980. plates and the dummy spacers was kept to the minimum. The
Journal of Fluids Engineering DECEMBER 1982, Vol. 1 0 4 / 5 2 3
diameter of the end plates was 3 times the model face D. The Table 1
resulting additional drag was less than 0.005 of the total drag
of the model. Grid b M Remarks
The models were attached at their center downstream face no. (in.) (in.)
to a circular tube. This tube was parallel to the flow and 1 0.08 0.5 uniplanar (round bars)
contained a small force balance. The drag force was measured 2 0.25 1.25 biplanar (rectangular bars)
using an Interface Mini Beam strain gauge load cell. The base 3 0.50 2.00 biplanar (rectangular bars)
pressure was measured on the centerline (2 in. away from 4 0.75 4.00 biplanar (rectangular bars)
absolute center) of the model's downstream face via a Tygon
tube that was connected to a Barocell type 572D pressure model with a given H/D exposed to an increased turbulence
transducer. The pressure readout was achieved with a intensity behaves as a model with increased H/D in smooth
Datametrics Electronic Manometer type 1018. flow. Correction procedure developed in smooth flow could
Turbulence was generated using biplanar square mesh grids be used if the equivalent H/D was known. The determination
installed at the wind tunnel test section entrance. Four grids of the equivalent H/D is not required for models with H/D >
were constructed allowing for a variation of the intensity of 1 since f is not a function of H/D in this range [10]. Equation
turbulence from 1.2 percent to 13 percent and a variation of (1) can be used directly in this case with f = 1.19. For models
the non-dimensional integral scale of turbulence, Lx/D, from with H/D < 1.0 where f is a function of H/D, the procedure
0.3 to 1.6. The model was located at a distance of over 30 bar used to correct for data was the one suggested by Maskell [12]
sizes downstream of the grid. Table 1 gives a summary of the for smooth flow exposition. From [10], the overestimate of
physical characteristics of the grids. All the turbulence the correction by the method suggested by Maskell is not
properties were measured with hot wire anemometry (DISA D severe for blockage ratio less than 10 percent for models with
generation system with linearizer). The linearity was main- H/D < 1.0. In view of the equivalent elongation of the
tained within 1 percent and the precision on u'/U was 5 models' afterbody by turbulence, this method was applied as
percent. The longitudinal scale of turbulence was determined long as the correction of the drag coefficient was less than the
via a real time spectral analysis (Spectral Dynamics Corps correction for the H/D = 1 model in smooth flow. When the
SD300A R.T.A.) of the linearized output. correction on by Maskell's method was larger than the one
In smooth flow, the Reynolds number (based on U and D) for square section models in smooth flow, equation (1) was
varied from 2 • 10" to 1 * 10^ while in turbulent flow, the used with f = 1.19.
maximum for the Reynolds number was reduced to 7 ' lO''.
Since a Reynolds number effect was observed on the mean Results
drag coefficient up to 4 * 10'', only the results with larger
Reynolds numbers were considered in the present analysis. In In order to discuss the effect of turbulence on models
the case of models with H/D in the vicinity of 0.6, the subjected to blockage, care must be taken to vahdate the
Reynolds number had to be larger than 6 ' 10*. A similar correction procedure when the correction itself is function of
effect was observed by Bearman and Trueman [9]. turbulence. The previous work by [4, 5 and 13] has indicated
the secondary role of the turbulence integral scale and at the
Blockage Correction Procedures same time uncoupled the blockage effect from the scale effect.
Thus it is admitted here that the intensity of turbulence is the
Previous work [10] in smooth flow has shown that essential variable affecting the blockage correction along with
Maskell's method [12] underestimafes the corrected drag H/D. Figures 1 and 2 compare the blockage correction
coefficient of models with H/D > 1. Moreover, because it technique used for low and high levels of turbulence with the
includes the ratio of the mean base pressure coefficient to the extrapolation of the data for several blockage ratios. Most of
mean drag coefficient, this method was inapplicable to the predicted C^c by the procedure are within 3 percent of the
models with H/D > 2. In reference [10], it was also shown Coc obtained by the extrapolation of the €„. This com-
that the drag correction coefficient for models with H/D > 1 parison gave sufficient confidence in the procedure and in its
was independent of H/D and that a simple empirical method application to all the data. Figure 3 shows the resulting drag
gives accurate results for blockage ratios less than 13 percent. curve for the H/D variable when the turbulence level was 11.2
The equation used for correction was percent.
Coc/Co = i-i(AM/AS), (1) Figure 4 shows the effect of the longitudinal scale of tur-
bulence on the mean drag coefficient in the cases of the
where f is the drag correction coefficient which was function models with H/D = 0.75 and 1.0 exposed to grid turbulence
of H/D for models with H/D < 1. In the absence of validated with 7 = 3 . 1 and 7.5 percent. The lines enveloping the results
blockage correction procedures for turbulent flows, a method are separated by a total deviation of 5 percent. Similar results
based on several observations was developed to take into were obtained for models with different H/D (0.5, 0.63, 0.75,
account the blockage produced by the models. The first 1.0, 1.5, and 3.0) and different levels of turbulence (0.006,
observations by [4 and 5] in turbulent flows indicate that a 0.083, and 0.112). In the case of 2-D rectangular cylinders
Nomenclature
A = plate area
AM = model area D = frontal dimension of the
AS = wind tunnel test section cross- rectangular cylinder per- Lx =
longitudinal integral scale
sectional area pendicular to the flow M =
grid mesh size
b - grid bar size direction q =
freestream dynamic pressure
Co = measured mean drag coef- H = lateral dimension of the u' =
rms value of the longitudinal
ficient = mean drag iorct/(q rectangular cylinder parallel velocity component
X AM) to the flow direction U = mean value of the
Cue = corrected mean drag coef- / = intensity of turbulence longitudinal velocity com-
ficient = u'/U ponent
524 Vol. 104, DECEMBER 1982 Transactions of the A S M E
-r- T r- I r - i r-
1 r
-H/D= .75
I • 3.1%
. • 2.32 t.06 .
""J i ^ H/0 = ,75
I = 75%
-2.20
H/D = 1.0
.02 .04 .06 .08 1=3.1 %
AM / AS
-2,00
F i g . 1 C o m p a r i s o n o l b l o c k a g e c o r r e c t i o n t e c h n i q u e w i t h d a t a ex-
t r a p o l a t i o n {I = 0.6 p e r c e n t ) H/D • 1.00
I =75%
3.0 - C
2.6 - F i g . 4 E f f e c t of t h e m a c r o s c a l e of t u r b u l e n c e o n t h e d r a g c o e f f i c i e n t
f o r t h e m o d e l s w i t h H/D = 0.75 a n d 1.0 e x p o s e d t o / = 3.1 a n d 7.5
percent
2.4 -
drawn on these figures are fitted lines through the data and
represent the mean values of the corrected drag coefficient for
cyhnders with different afterbody length, H/D, as the in-
tensity of turbulence is varied. Rectangular cylinders with
2.0 - H/D larger than 0.63 see their C^,^ reduced as the turbulence
level is increased. In smooth flow, the maximum C^c of 2.90
occurs for the cylinder with H/D = 0.63, which agrees with
[9, 10, 6]. In grid turbulence, models with 0.5 < H/D < 0.63
see their drag coefficient reach the smooth flow maximum
Cpc and decrease as the level of turbulence is increased.
Models with 0.3 < H/D < 0.5 show a similar trend but the
peak is much less pronounced and the smooth flow maximum
C[,c is not reached.
Figure 9 gives a summary of all the results and includes the
results obtained by Nakamura and Tomonari [6]. Again here
the effect of turbulence intensity is clearly to shift the
versus H/D curve to the left and to reduce the maximum CQC
progressively as the level of turbulence is increased.
Nakamura and Tomonari results are not corrected for
F i g . 2 C o m p a r i s o n of b l o c k a g e c o r r e c t i o n t e c h n i q u e w i t h d a t a ex- blockage effects that were small in their test. The agreement
t r a p o l a t i o n (f = 11.2 p e r c e n t )
as seen is good.
with H/D > 0.31, it confirms that the scale of turbulence
ratio, Lx/D, plays a secondary role in the mean flow about Discussion
rectangular cyhnders. This was also observed by [4, 5 and 13]. As indicated by Figs. 5 to 9, and 9 principally, the behavior
Figures 5, 6, 7, and 8 summarize the effect of the intensity of of the mean flow around rectangular bluff bodies is a strong
turbulence on the C^c for different H/D models. The lines function of the afterbody length H/D and the level of tur-
Journal of Fluids Engineering DECEMBER 1982, V o l . 1 0 4 / 5 2 5
2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0
ICA) 1.50
Fig. 5 Effect o f t h e i n t e n s i t y o f t u r b u i e n c e o n t h e d r a g c o e f f i c i e n t . I I L I I I I
2.0 4 0 6.0 80 10 0 12.0 14 0
I (%)
Fig. 7 E f f e c t of t h e i n t e n s i t y of t u r b u i e n c e o n t h e d r a g c o e f f i c i e n t
1 1 1 r
turbulence. In smooth flow with very low levels of turbuletice, Fig. 8 Effect of t h e i n t e n s i t y of t u r b u i e n c e o n t h e d r a g c o e f f i c i e n t
DISCUSSION
B.E.Lee' Authors' Closure
The authors of this paper are to be encouraged in their The authors thank Dr. B. E. Lee for his words of en-
couragement and his interest for this work. We will try to
' Lecturer, Department of Building Science, University of Sheffield, Shef- answer the questions relative to the experimental set up and
field, S10 2TN, England. then discuss the question of integral scales.