Professional Documents
Culture Documents
Received 30 December 2004; received in revised form 7 October 2005; accepted 7 October 2005
Available online 28 November 2005
Abstract
In Part I, trajectories of plates, carried by strong winds, were studied experimentally by windtunnel and full-scale tests. The application is to windborne debris occurring in severe windstorms
such as hurricanes. In this paper (Part II), a numerical model of square plate trajectories is described
and compared with experimental data from Tachikawa, and that described in Part I. Generally, good
to excellent agreement is found; lift forces induced by the Magnus effect were found to be signicant
in determining trajectories.
r 2005 Elsevier Ltd. All rights reserved.
Keywords: Debris; Hurricane; Plate; Wind loads
1. Introduction
Windborne debris is known as a major source of damage in strong wind events such as
hurricanes. Using the classication scheme proposed by Wills et al. [1], generic debris types
can be classied as either compact, sheets/plates, or rods. Although the types of test
missile used in impact tests have traditionally been of the compact or rod type [2], there
is also evidence of considerable damage produced by plate-type objectsfor example,
concrete roof tiles during Hurricane Charley in 2004.
Published work on windborne debris was reviewed in a previous paper by Holmes [3].
This included the pioneer work by Tachikawa [4,5] who studied the trajectories of small
Corresponding author.
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
22
at plates, and other objects, in a wind tunnel. Wang and Letchford [6] also studied ight
trajectories of small sheet objects in a wind tunnel.
This paper is a companion to Part I [7], in which new experimental data on the
trajectories of plates of various aspect ratios are described. The present paper describes
numerical solutions of the ight characteristics of plates of square planform, following
initiation of ight. Unlike compact objects, such as spheres [3], which are driven only by
drag forces, the trajectories of plates in strong winds are subject to drag and lift forces, and
also pitching moments, for motion in a vertical plane.
2. Aerodynamic forces and moments on square plates
2.1. Normal force coefficient
The dimensions of a square planform plate, and the angle of attack, a, of the relative
wind are shown in Fig. 1. Flachsbart [8] measured the force coefcient on a square at
plate as a function of the angle of attack. These data were reported by Hoerner [9], and is
also the basis of an ESDU Data Item [10]. Similar measurements were carried out by
Tachikawa [11], and more recently by the present authors at Monash University and Texas
Tech University.
At an angle of attack of about 401 the plate stalls, and the normal force coefcient
shows a sharp peak before falling to a near-constant stalled value of 1.11.2. The assumed
line, as shown in Fig. 2, consisting of linear segments shows quite good agreement with the
experimental data.
The model is as follows:
C N 1:7a=40
for ao40 ,
(1a)
(1b)
(1c)
C N 1:15
centre of pressure
thickness, h
chord, l
span, l
Fig. 1. Dimensions and angle of attack for a square plate.
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
23
2.0
1.6
CN
1.2
0.8
assumed
measured-Monash
measured-TTU
measured-Tachikawa
measured-Flachsbart
0.4
0.0
0
45
90
angle of attack (deg.)
135
180
The normal force is resolved into drag and lift components, but an additional drag
coefcient of 0.1 has been added to allow for the skin friction component. Thus
(2a)
C D 0:1 C N sin a,
C L C N cos a.
(2b)
M
,
1=2rU 2 3
(3)
the relationship between the moment coefcient, the normal force coefcient, and the
centre of pressure position is
C M C N c=.
(4)
Fig. 3 shows the assumed model of the relative centre of pressure position, c/, as a
function of angle of attack. This is a segmented model adjusted to t the available
experimental data, with the following values of c/:
c= 0:3 0:22a=38 for the angle of attack; ap38 ,
c= 0:08 cos 2a 38
c= 0:0
(5b)
(5a)
(5c)
(5d)
(5e)
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
24
0.4
0.3
0.2
c/l
0.1
0.0
0
45
-0.1
90
135
180
assumed
measured-Monash
measured-TTU
measured-Tachikawa
measured-Flachsbart
-0.2
-0.3
-0.4
(6a)
(6b)
0:64U
,
(6c)
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
25
The total lift coefcient is then given by the sum of CL from Eq. (2b), and CLr from Eqs.
(6a) or (b). Magnus effect components for the drag force and pitching moment were not
included in the numerical calculations in this paper.
3. Equations of motion
Fig. 4 shows the total angle of attack, a, of the relative wind to a plate. It consists of the
sum of two components: bthe angle of attack of the relative wind induced by the vertical
motion of the sheet, with respect to the horizontal, and ythe angle of the sheet to the
horizontal, i.e. the initial angle of attack at the start of the ight, plus the angle the sheet
has rotated under the action of the rotational moments.
The horizontal acceleration of the plate is given by
d2 x ra C D cos b C L sin bU um 2 v2m
.
dt2
2rm h
(7)
dt2
2rm h
(8)
dt2
2I
(9)
um
vm
vm
U-um
Fig. 4. Relative angle of attack of the wind for a translating, rotating plate. (Note: positive vm is upwardsthe
negative direction is shown above).
ARTICLE IN PRESS
26
where ra is the density of air, rm is the density of the plate material, U is the wind speed, um
is the horizontal velocity of the plate, vm is the vertical velocity of the plate, h is the
thickness of the plate, is the side dimension of the plate (square here), A is the plan area
(2), I is the mass moment of inertia, CD, CL and CM are the drag, lift and moment
coefcients, x is the horizontal distance travelled, z is the vertical distance travelled, g is the
acceleration due to gravity, and t is time.
These equations can be solved numerically using small time steps, for the horizontal,
vertical, and angular velocities and displacements. The normal force coefcient and centre
of pressure positions obtained for static plates at various angles of attack, described in
Sections 2.1 and 2.2, can be used to obtain drag, lift and moment coefcients in the above
equations. Additional lift forces due to the Magnus effect are obtained using the
Tachikawa model, described in Section 2.4.
4. Computed trajectories and comparisons with experimental data
4.1. Comparison with Tachikawas experiments
Tachikawa [4] carried out a series of free-ight trajectory tests on small at plates in a
uniform ow in a wind tunnel. The various downstream positions of the plate in the early
part of the trajectories were photographed using a stroboscope. An important observation
was the differing trajectories and rotational behaviour, depending on the initial angle of
attack, a0, at release.
Numerical calculations were made to compare with the square plate trajectories reported
by Tachikawa. The previous linear segmented models of normal force coefcient and
centre of pressure position were used in these calculations [12], but the results are little
changed if the model shown in Figs. 2 and 3 are used instead. Examples of the comparisons
are given in Figs. 5 and 6. The plate was made of plastic with side length of 40 mm, and
thickness 2 mm. The Tachikawa parameter K (see Appendix A) was equal to 2.3. The
Fig. 5. Comparison of computed trajectories (upper) with Tachikawas recordings [4] (lower). Initial angle of
attack: 151. (Note: vertical scale of Tachikawas measurements is same as horizontal scale).
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
27
Fig. 6. Comparison of computed trajectories (upper) with Tachikawas [4] recordings (lower). Initial angle of
attack: 451. (Note: vertical scale of Tachikawas measurements is same as horizontal scale).
agreement of the rotational positions with the experimental recordings is very good
in both Figs. 5 and 6. In the latter case, the plate rst rotates clockwise, and then anticlockwise in the experiments; this behaviour is also reproduced in the numerical
calculations. The vertical displacements in the calculations, in both cases, are slightly
lower than those recorded in the experiments (note that the vertical displacements are
plotted to a different scale to the horizontal displacements in the numerical calculations in
Figs. 5 and 6).
4.2. Influence of the Magnus effect
The numerical calculations for the Tachikawa plate used in the previous section were
also used to investigate the inuence of the Magnus effect terms on the calculations. The
assumed models for normal force coefcient and centre of pressure position in Figs. 2 and
3, respectively, were used. Some examples are shown in Fig. 7.
In the case of the 151 initial angle, the angular velocities are large and always in the
clockwise directionin that case the Magnus effect generates signicant additional
positive (upwards) lift forces and hence less negative vertical displacements. For the
a0 451, angular velocities are small, and in both clockwise and anti-clockwise directions,
resulting in negligible differences. For the 1501 initial angle, the rotational velocities and
displacements are anti-clockwise for motion from left to right [4], and the Magnus effect
gives negative lift; hence the calculated vertical displacements are more negative when the
Magnus effect is included.
For initial angles between 451 and 1501 (not shown), the angular velocities are small, and
trajectories with and without the Magnus effect forces included differ little from each
other.
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
28
-0.5
-1.0
-1.5
No Magnus
Effect
With Magnus
Effect
-2.0
-2.5
-3.0
0.5
0.0
-0.5
No Magnus
Effect
With Magnus
Effect
-1.0
-1.5
-2.0
-2.5
-3.0
0.0
0
-0.5
No Magnus
Effect
With Magnus
Effect
-1.0
-1.5
-2.0
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
29
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
30
0.2
1
2
3
Calculated
no Magnus Effect
-0.2
-0.4
-0.6
1
12
1
2
3
Calculated
no Magnus Effect
10
8
6
4
2
0
0
0.2
0.4
0.6
0.8
time (secs)
Trajectory of basswood plate (plate10 - 0deg- 15.6 m/s)
4
1
2
3
Calculated
no Magnus Effect
0
0
0.2
0.4
0.6
0.8
time (secs)
Fig. 8. Comparison of computed trajectories with Texas Tech data. Plate 10, basswood, 75 mm 75 mm 9 mm.
U 15:6 m=s, K 3:0.
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
31
0.2
-0.2
1
2
3
-0.4
Calculated
no Magnus
Effect
-0.6
0
20
15
10
1
2
3
Calculated
no Magnus
Effect
0
0
0.5
1.5
time (secs)
Trajectory of basswood plate (plate 10 - 0deg- 21.4 m/s)
6
5
4
3
2
3
Calculated
no Magnus
Effect
0
0
0.5
time (secs)
Fig. 9. Comparison of computed trajectories with Texas Tech data. Plate 10, basswood, 75 mm 75 mm 9 mm.
U 21:4 m=s, K 5:6.
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
32
0.6
0.4
0.2
1
2
3
Calculated
no Magnus
Effect
0
-0.2
-0.4
-0.6
2
25
1
2
3
Calculated
no Magnus
Effect
20
15
10
5
0
0
0.2
0.4
0.6
time (secs)
Trajectory of basswood plate (plate 10 - 0deg- 25.6m/s)
6
5
1
2
3
Calculated
no Magnus
Effect
4
3
2
1
0
0
0.2
0.4
0.6
time (secs)
Fig. 10. Comparison of computed trajectories
75 mm 75 mm 9 mm. U 25:6 m=s, K 8:0.
with
Texas
Tech
data.
Plate
10,
basswood,
other generic forms of debris (e.g. [3]). Its reciprocal has also been suggested by Baker [13]
as a suitable parameter. The effect of varying K can be seen in Figs. 810, in which the
Tachikawa parameter varies between 3 and 8.
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
33
0.5
2-15-1
2-15-2
2-15-3
Calculated
0.4
0.3
0.2
0.1
0
0
14
12
10
8
2-15-1
2-15-2
2-15-3
Calculated
6
4
2
0
0
0.2
0.4
0.6
0.8
time (secs)
horizontal displacement (m)
2-15-1
2-15-2
2-15-3
Calculated
2
1
0
0
0.2
0.4
0.6
0.8
time (secs)
Fig. 11. Comparison of computed trajectories with Texas Tech data. Plate 8, plywood, 75 mm 75 mm 3 mm,
a0 151, U 15:4 m=s, K 5:9.
Figs. 13 and 14 present some results for plates with zero initial angle of attack, in a nondimensional form. In Fig. 13, numerical calculations of non-dimensional horizontal plate
velocity (um/U) are plotted as a function of the non-dimensional horizontal displacement
(xg/U2), for two values of the Tachikawa parameter, K. The weak dependency on the
Froude number, Fr, (dened in Eq. (A.4)), for the higher value of K, is shown in this
gure. A higher value of K corresponds to a lighter plate, and, not surprisingly, to a more
rapid increase in velocity with increasing distance.
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
34
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
0
0.5
1.5
10
1-135-1
1-135-2
1-135-3
Calculated
8
6
4
2
0
0
0.1
0.2
0.3
0.4
time (sces)
Trajectory of plywood plate (Plate 8- 135 deg)
1.5
1-135-1
1-135-2
0.5
1-135-3
Calculated
0
0
0.1
0.2
0.3
0.4
time (secs)
Fig. 12. Comparison of computed trajectories with Texas Tech data. Plate 8, plywood, 75 mm 75 mm 3 mm,
a0 1351, U 14:6 m=s, K 5:3.
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
35
um/U
0.6
0.4
Fr2= 100
Fr2= 200
Fr2= 300
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
x
Square Flatplate 0 degrees K=17.6
1.0
0.8
um/U
0.6
0.4
Fr2= 100
Fr2= 200
Fr2= 300
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
x
Fig. 13. Non-dimensional horizontal plate velocity versus non-dimensional horizontal displacement. K 6:7,
upper; K 17:6, lower. Varying values of Fr2 .
6. Conclusions
Numerical calculations of the trajectories of plates of square planform have been made,
and compared with experimental data from two wind tunnels. Generally, good to
excellent, agreement has been shown.
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
36
0.8
0.6
0.8
#2 Vw=6.1m/s
#5 Vw=15.6m/s
#7 Vw=5.6m/s
#8 Vw=10.9m/s
#9 Vw=14.6m/s
#10 Vw=15.6m/s
#13 Vw=10.0m/s
Calculated
0.4
0.2
0
0.0
um/U
um/U
0.6
#1 Vw=26.0m/s
#3 Vw=18.8m/s
#4 Vw=10.2m/s
#5 Vw=23.1m/s
#7 Vw= 8.3m/s
#8 Vw=16.4m/s
0.4
0.2
0
0.2
0.4
xg/ (U2)
0.6
0.0
0.8
0.4
xg/U2
0.6
0.8
0.8
0.6
um/U
0.8
um/U
0.2
1.2
#4 Vw=13.5m/s
#8 Vw=21.5m/s
#11 Vw=13.6m/s
#13 Vw=19.9m/s
Calculated
0.4
0.2
0
0.0
#9 Vw=22.0m/s
#10 Vw=23.1m/s
#11 Vw=10.2m/s
#12 Vw=23.0m/s
#13 Vw=15.1m/s
Calculated
0.2
0.4
xg/ (U 2)
0.6
#2 Vw=14.8m/s
#4 Vw=16.5m/s
#7 Vw=13.6m/s
#11 Vw=16.5m/s
Calculated
0.6
0.4
0.2
0.8
0
0.0
0.2
0.4
xg/U2
0.6
0.8
Fig. 14. Non-dimensional horizontal plate velocity versus non-dimensional horizontal displacementcomparison of numerical calculations with experiments, for varying values of K.
The numerical models incorporate the angular displacements and velocities, and
Magnus effect lift forces. Simple segmented models of normal force coefcient, centre of
pressure position, and Magnus effect lift seem to be quite adequate for engineering
predictions of trajectories.
The Tachikawa parameter, K, is a governing non-dimensional parameter for
determining trajectories of windborne debris of all generic types, including the plate type.
Non-dimensional presentations of horizontal plate velocity against horizontal displacement (e.g. Figs. 13 and 14), have practical value to allow prediction of horizontal impacts
speeds of full-scale plate-type debris on walls of buildings of xed distance, x, downstream
from a debris source.
Acknowledgements
The research in this paper was supported through the John P. Laborde endowed Chair
at Louisiana State University held by the rst author, and by the L.S.U. Sea Grant
Program. The rst author also acknowledges Professor Bill Melbourne and Dr. John
Cheung of Monash University, Australia, for allowing access to the 450 kW wind tunnel
for force measurements on a at plate. Useful discussions with Professor Chris Baker
(University of Birmingham) and Associate Professor Elizabeth English (LSU) are also
acknowledged.
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
37
(A.1)
where, ra is the density of air, U is the wind speed, m is the mass of the plate, is the side
dimension of the plate, I is the mass moment of inertia, CD, CL and CM are the drag, lift
and moment coefcients, x is the horizontal distance travelled (or the distance to the wall
of a building at impact), g is the acceleration due to gravity, a0 is the initial angle of attack.
There are a number of different ways of reducing the above variables into nondimensional quantities.
Tachikawa [4] dened a non-dimensional number dening the trajectories of all types of
windborne objects. For a plate, this can be written as
K
ra U 2 2
r U2
a
,
2mg
2rm hg
(A.2)
where h is the thickness of the plate, and rm is the density of the material of the plate. This
parameter represents a ratio of aerodynamic forces to gravity forces on the plate.
K can also be written as follows:
2
1 ra
U
K
,
2 rm
h
g
i.e. K is the product of a density ratio, a ratio of dimensions, and the square of a Froude
number, Fr, based on .
The ratio In was also dened by Tachikawa:
I
.
(A.3)
m2
For a square plate, In is xed and equal to 1/12.
Two Froude numbers, based on the characteristic plate length, , and the horizontal
distance travelled, x, can be obtained as
In
U
Fr p ,
g
(A.4)
U
Frx p .
gx
(A.5)
The drag, lift and moment coefcients are themselves functions of the angle of attack
during a trajectory.
Hence, for a square plate, Eq. (A.1) can be written as
u
m
(A.6)
F K; Fr ; Frx ; a0 .
U
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
38
1=2ra A
,
m
(A.7)
(A.8)
m2
,
(A.9)
I
where A is the planform area, equal to 2 for a square plate.
O is the reciprocal of the Tachikawa parameter K. F was described as a buoyancy
parameter [13]it is effectively a density ratio. D is the reciprocal of In used by Tachikawa
[4], and is equal to 12 for a square plate.
For a rectangular plate with span B, and chord D, in the above relations should be
replaced by B, and an additional non-dimensional parameter, the aspect ratio, B=D
included.
D
x
F; z
F; y yF; t
F:
(A.10)
Alternative variables, suggested by Tachikawa [4], have been used in the present paper:
tg
xg
zg
x
.
(A.11)
;
z
;
y;
t
U
U2
U2
For both schemes, the horizontal and vertical non-dimensional missile velocities are
dened as
u
v
m
m
u
; v
.
U
U
Using the variables dened in Eq. (A.11), the equations of motion (7), (8) and (9) can be
re-written in non-dimensional form as follows [4]:
d2 x
KC D cos b C L sin b1 u 2 v 2 ,
dt2
(A.12)
d2 z
KC D sin b C L cos b1 u 2 v 2 1,
dt2
(A.13)
d2 y
DKF r2 1 u 2 v 2 ,
dt2
q
(A.14)
where cos b 1 u =
q
1 u 2 v 2 and sin b v = 1 u 2 v 2 .
ARTICLE IN PRESS
J.D. Holmes et al. / J. Wind Eng. Ind. Aerodyn. 94 (2006) 2139
39
The solutions to these equations are consistent with the non-dimensional relationship of
Eq. (A.6). The non-dimensional horizontal missile speed at impact depends primarily on
the Tachikawa parameter, K, and the non-dimensional displacement, x xg=U 2 . It also
depends on the initial angle of attack of the plate, a0, since this will determine the values of
the drag and lift coefcients during the trajectories. Noting Eq. (A.14), there is also a weak
dependency of the non-dimensional horizontal missile speed on the Froude number, Fr,
due to the effect of the rotational displacement on the force coefcients. Note that the nondimensional horizontal displacement x xg=U 2 is also the reciprocal of the square of a
Froude number based on the displacement x, as dened in Eq. (A.5).
References
[1] J.A.B. Wills, B.E. Lee, T.A. Wyatt, A model of wind-borne debris damage, J. Wind Eng. Ind. Aerodyn. 90
(2002) 555565.
[2] American Society for Testing Materials, Standard specications for performance of exterior windows,
curtain walls, doors and storm shutters impacted by windborne debris in hurricanes, ASTM E 1996-02, 2002.
[3] J.D. Holmes, Trajectories of spheres in strong winds with application to windborne debris, J. Wind Eng. Ind.
Aerodyn. 92 (2004) 922.
[4] M. Tachikawa, Trajectories of at plates in uniform ow with applications to wind-generated missiles,
J. Wind Eng. Ind. Aerodyn. 14 (1983) 443453.
[5] M. Tachikawa, A method for estimating the distribution range of trajectories of windborne missiles, J. Wind
Eng. Ind. Aerodyn. 29 (1988) 175184.
[6] K. Wang, C.W. Letchford, Flying debris behaviour, 11th International Conference on Wind Engineering,
Lubbock, Texas, June 25, 2003.
[7] N. Lin, C.W. Letchford, J.D. Holmes, Investigations of plate-type windborne debris, Part I, Experiments in
full scale and wind tunnel, submitted for publication.
[8] O. Flachsbart, Messungen an ebenen und gewolbten Platten, Ergebnisse der AVA. IV, 1932.
[9] S.F. Hoerner, Fluid-dynamic drag, Hoerner Fluid Dynamics (1965).
[10] ESDU, Fluid Forces and Moments on Flat Plates. Data Item 70015, Engineering Science Data Unit,
London, 1970.
[11] M. Tachikawa, Trajectories and velocities of typhoon-generated missiles, Part I, Aerodynamic characteristics
of at plates and equations of motion, Trans Architect Inst Japan 314 (1981) 110.
[12] J.D. Holmes, E.C. English, C.W. Letchford, Aerodynamic forces and moments on cubes and at plates, with
application to windborne debris, Fifth International Colloquium on Bluff-body Aerodynamics and
Applications, Ottawa, Canada, July 1115, 2004.
[13] C.J. Baker, Solutions of the debris equations, Sixth UK Conference on Wind Engineering, Craneld,
September 1517, 2004.