Anda di halaman 1dari 10

The response of reinforced

concrete chimneys to vortex


shedding
B. J. V i c k e r y

The University of Western Ontario, Faculty of Engineering Science, London, Ontario,


Canada
R. I. Basu

H.G. Engineering Ltd, Toronto, Ontario, Canada


Predicted vortex induced response of circular cross-section reinforced
concrete structures is compared with observed full-scale behaviour.
On average the predictions agree to within 5% with observations but
have a large scatter attributable to the model's failure to recognize the
dependence of aerodynamic parameters on free stream turbulence. The
high coefficient of variation of the model is, however, not markedly
different from that associated with the prediction of drag loads.
Key words: vortices, turbulence

Predictions of the vortex induced response of chimneys


and towers of circular cross-section are compared with
observed full-scale behaviour. It is demonstrated that the
predictive model produces estimates which are reasonable
on average but which have a large scatter. 1"he full-scale
data gathered in the course of the study strongly suggest
that the prime cause of the scatter was the failure of the
predictive model to recognize the dependency of the basic
aerodynamic parameters on turbulence. The predictive
model is also used to explore the relative importance of
across-wind loads as opposed to drag loads and to define,
in broad terms, the chimney geometries for which acrosswind loads are likely to dominate the design.
The predictive model employed is that suggested by
Vickery and Basu 1'2 and the major features of this model
are now reviewed very briefly. The forces due to vortex
shedding are considered in two uncorrelated parts; those
which exist on a stationary cylinder and those induced by
motion o f the cyfinder. The forces on a stationary body are
modelled, primarily, as a narrow-band random force with a
spectrum centred on the shedding frequency and with a
band-width dependent on the intensity of turbulence but
those forces due to lateral velocity fluctuations are also
modelled. The motion dependent forces are modelled as
amplitude dependent damping forces which are negative
in the vicinity o f the critical velocity, and are dependent on
turbulence and which decrease in absolute magnitude with

324

Eng. Struct., 1984, Vol. 6, October

the r.m.s, motion of the cylinder. This model yields an


expression for the tip motion of a chimney or tower of the
form:
C
{Ks--Ka(1 -- (ff/gL)2)} '/2
where :
t~
C
KS

Ka
t~L
m
d
P

r.m.s, tip displacement


a constant dependent on the aerodynamic and
structural parameters
m~s/Pd 2 the structural mass-damping parameter
_m[3a/pd 2, the aerodynamic mass-damping parameter
a limiting r.m.s, tip displacement
mass per unit length
diameter
air den sity

The equation has approximate solutions:

= C/{Ks--Ka} 1/2

for

K s >>K a

d = aL{1 --Ks/Ka} 1/2

for

Ks < Ka

and :
The first solution corresponds to a forced random vibration
regime with reduced structural damping while the second
corresponds to a 'lock-in' regime where the amplitude is
0141-0296/84/04324-10/$3.00
1984 Butterworth & Co. (Publishers) Ltd

Response of RC chimneys to vortex shedding: B. J. Vickery and R.

determined by the nonlinear aerodynamic damping.


Linking these two regimes is a 'transition' regime near
Ks/K a = 1 in which the amplitude increases very rapidly
with decreases in K s and the response changes in character
from a typical random response with peak values of the
order of three or four times the r.m.s, value to an almost
sinusoidal response with peak values only slightly in excess
of X/2 times the r.m.s, value. This simple model thus
reproduces the major features that have been observed in
experimental programs such as that described by Wootton. 3

1.0

Ca

~v =

{if

~k2(z)dz

},,2

mass/unit length at height z


air density
structural damping as a fraction of critical
correlation length in diameters
aerodynamic damping coefficient (Figures 1

ra

I
30

a peak factor (-- 4)


(2)

{~s--rapa2/me} uz

m(z)

1 I 1 t I I I
8 10
20

1
I
/1 --k-t\21
(B,k) = --~ k3/Zexp[--t----~) ~

(1)

f m(z) s2(z) dz/f 2(z) dz

#s

I
6

Influence of aspect ratio (hlcl) on a e r o d y n a m i c parameters S, C'd, CL and K a

me

19

I
4

Figure 2

(B,k)

(z)

,~1/2

peak tip deflection due to vortex shedding


aspect ratio hid
averag~ diameter of top third
chimney height
r.m.s, lift coefficient (Figures I and 2)
Strouhal number (Figures 1 and 2)
mode shape

CL

h/d

where:

av
X

//Cc,Ka

Response predictions

For chimneys with little or no taper


{d(h)/d(O) > 0.5)
g~L~ll/(h ) ~ 2 ( ~ / ~
87flSu me \ ~ /

. ,

0.~

__f

The equations used to predict the response of the structures


examined are based upon the vortex shedding model described above and derived in an earlier paper. 4 The equations are:

I. Basu

B
k

a spectral bandwidth
a/~ c
mean speed at z = 5/6h

~c

[o~/S
natural frequency of mode

fo

For tapered chimneys (d(h)/d(O) < 0.5)


gt~LP" d4(ze) ff(ze)" ~k(h)(nl/2t) u2
a~, -

(3)

d12~'/2

81r2S2me"f ~k2(z)dzt[Js--Ka ~ e )
0

where:
ze = height at which t~(z) = fod(z)/S

and 2)

t = --

+ ad(z)
dz

(4)
Iz=ze

a = power law exponent; a(z)/a(h) = (z/h) ~


~(z) = mean speed at height z

0.4

1.0

The values of the various parameters used in the prediction


were:

0.3

(a)
,G

"'~ 0.2

35

(b)

/Ie,
0.1
1

10"s

I
10"4

10 -3

k,/d
Figure 1 Variation of aerodynamic parameters C'D, CL, S and K a
with roughness for two-dimensional flow at transcritical Reynolds
numbers (ks = equivalent sand roughness)

(c)

CL, S and K a for two-dimensional flow conditions


were determined from Figure I using an assumed
value of ks of I mm and then adjusted for aspect ratio
as indicated in Figure 2.
Values considered to be invariant were:
/3s

0.01

l
B

1.0
0.25

a
p

3.8
0.15
1.20 kg/m 3

In order to simplify the computations and avoid a


detailed dynamic analysis it was assumed that the
mode shape for the fundamental mode could be
approximated by:

~(z) = (zlh) 2

Eng. Struct., 1984, Vol. 6, October

325

Response of RC chimneys to vortex shedding: B. J. Vickery and R. I. Basu


Table 1 Properties of full-scale structures
Chimney

h (m)
d(h) (m)
d(0) (m)
(m)

h =h/d
ks/cl
CL
S
/<'aT
m e (kg/m)

me/Pd 2
f0 (Hz)
Predicted max. response
Obs. max. response

Tower

No 1

No2

No3

No4

No5

No 1

244
10.6
22.8
12.6
19.4
8x10 5
0.158
0.216
0.48
23 000
121
0.33

201
11.6
17.7
12.8
15.7
8x10-5
0.146
0.207
0.44
23 000
117
0.37

265
12.27
21.85
13.9
19.0
7xl0-S
0,157
0.216
0.47
21 700
94
0.27

180
4.75
9.86
5.6
32.0
1 8 x1 0 -s
0.205
0.226
0.58
10 400
275
0.26

130
6.19
18.12
6.8
19.1
1 5 x1 0 s
0.176
0.214
0.51
7600
137
0.64

66*
4.88
4.88
4.88
13.5
20xl0-S
0.160
0.200
0.45
40 000t
1400
0.42

0.85

0.83

1.39

0.75

1.35

1.17

No 2

b
E
o

g=

3g

0.91
(near mode 1 max.)

A v (predicted/observed) = 1.06; C of V 24%

1.27
(near mode 4 max.)

* Stem height (Figure 10)


t Equivalent (including turret)

The use of these equations to predict the response of


structure No 1, a reinforced concrete chimney with a
height of 244 m and a tip to base diameter ratio of 0.46,
is demonstrated in the Appendix. Since the ratio d(h)/d(O)
is less than 0.50 the equations for a tapered stack should
be employed but, by way of example only, the response is
computed by both methods. The properties of the seven
structures examined are summarized in Table 1. Five of
the structures are chimneys and two are reinforced concrete
towers of circular cross-section. All but tower No 2 were
analysed using either equation (1) or equation (3); the
nature of tower No 2 is such that a number of modes of
vibration need to be considered and a computer analysis
was employed in preference to the simphfied response
equations.
Before proceeding to the comparison of the predicted
and observed responses, some discussion of the accuracy of
the field observations is warranted. Given reasonable care
in the conduct of a field experiment, there is no doubt that
the dominant errors are those associated with the use of
limited sampling periods.
The coefficient of variation, Cv, of estimates of the
r.m.s, response of a lightly damped system to random
excitation is given by:

Comparison of predicted and observed


across-wind tip motions
Chimney No 1
The data for Chimney No 1 were provided by Rumman. 1
The computed frequericy using a modulus of 2.5 x 109
N/m 2 was 0.33 Hz and the measured frequency 0.37 Hz.
The structure was instrumented with two tip accelerometers and observations were made over a period of more
than 12 months. Peak tip displacements observed in sampling periods of 1 minute were obtained from strip cfiarts
and the wind speed noted was the maximum gust (in that
minute) recorded by an aerovane anemometer mounted at a
height of about 10 metres. In preparing the plot in Figure 3,
the recorded speeds have been multiplied by 1.42 in an
endeavour to adjust them to mean speeds at 244 m. The
predictions are based upon the theoretical frequency of
0.33 Hz rather than the observed 0.37 Hz. If the observed
frequency was employed the predictions would move to
the right with the speeds increased by 37/33 but with no
change in amplitude.
Because of the short sampling time the observed data
can be used only to define an envelope. Allowing for the

Cv ~ 1 v ~ f o T
where:
T = sample length (s)
f0 = natural frequency (Hz)
fl = damping as a fraction of critical
The coefficient of variation of peak values determined from
similar samples would be somewhat higher. The sample
lengths employed in the field studies varied from a maximum of 20 minutes to a minimum of 1 minute. At the
longest sample length, the coefficient of variation is roughly
10% while for 1 minute samples it approaches 50%. Of the
seven field studies reported here, five have been described
in detail in the literature. 5-9 The results from the remaining
two were made available by W. Rumman. 1

326

Equation 1
Equation 3

Eng. Struct., 1984, Vol. 6, October

o3k Observotons
| Envelope of
I observations .......

'~"

/
o ~b

//

/ -.-'/"
o
l

15

Figure 3

.Z'" I
/,,." :

I"~

"

l
I
20
25
5 (h), (m/s)

'~

:"
I
30

Measured and predicted peak tip displacements for


Chimney No 11

Response of RC chimneys to vortex shedding: B. J. Vickery and R. I. Basu

difficulty in obtaining an adequate assessment of the wind


speed, ti,e agreement can be said to be good with the
maximum observed motions being some 18% above the
predictions.

Chimney No 2
The data for Chimney No 2 was again provided by
Rumman. m Unfortunately, the observations are very
limited and only one reliable measure of the peak tip
displacement was obtained. The displacement was the
maximum observed by a ground based transit during an
observation period of about 15 minutes durinB_which the
wind speed at a height of 76 m was 'gusting to 27 m/s'.
Visual observations of the chimney during strong winds
suggest that the one reliable measurement obtained was in
fact near the critical speed. The predictions are based on a
theoretical frequency of 0.37 Hz and no information
concerning the actual frequency was available to the
authors.
Chimney No 3
The properties of Chimney No 3 are fully described by
Melbourne et al. s A well designed experiment was, most
unfortunately, severely handicapped by a lack of wind. At
no time during the period over which the chimney was
instrumented was it subjected to sustained winds in the
vicinity of the critical speed of 20 m/s and the only observations of interest were limited to thunderstorms of comparatively short duration. It was, therefore, necessary to
limit the sampling periods to 5 minutes with a consequent
loss in reiiabifity. In addition to the results presented in
Figure 5 some other observations are worthy of note.

M e a s u r e d peak displacement
(by transit )

Visual estimate

from ground~
"~0.2 ("above 12"m
- motion")-

"

//

15

(equation

1)

,e.-o- \\
I

20

25

~(h),(m/s)
ments for Chimney No 2 ~

Measured rms (5 min average)


for(O u16)~ = 9.0%
0.15

~E0.10
Pred,cted

f'.

005

=I= ~ "

10

15

20
(h), (m/s)

25

3O

Figure 5 Comparison of predicted and observed r.m.s, tip displacements for Chimney No 3 s

00~

~'K:~

Trend of observations

i. Predicted
..x (equot,oo 3)

.~'" "~. \,,

E 0.03

~002
0.01
4

-~ 8 9 1C) 1~ 112 13 14 15 16
0 (h), (m/s)

Figure 6 Predicted and observed peak tip deflections for Chimney


No 4 6

Although the number of observations was limited, the


results obtained show a clear dependence on turbulence
intensity with an increase in response of the order 200%
being noted as the intensity increased from 6.5% to 9%.
An examination of the peak factors (peak/r.m.s.) indicated
values below 2 for the across-wind response near the
critical speed. This observation, however, should be treated
with caution since the sample lengths were of necessity
very short and the r.m.s, value was computed from the
same sample as that used to define the peak; this method
of computation will produce peak factors significantly
lower than the theoretical values for a narrow-band
Gaussian process with peak factors defined as the ratio of
the expected peak in a sample to the r.m.s, value of the
process. Fully reliable peak factors below 2 would be a
clear indication that the response was in the 'transition'
region def'med earlier. The agreement between the predicted and the observed r.m.s, displacement is fair with the
observed values being about 28% below the predictions.

30

Figure 4 Observed and measured peak across-wind tip displace-

(equation 1 ) /

Chimney No 4
The experimental program for Chimney No 4 is fully
described by Muller and Niesser 6 although the results
presented in that publication show an inconsistency which
has been discussed and, it is believed, resolved by Basu and
Vickery. n The corrected measurements are shown in
Figure 6. The observed motions are consistently above the
predictions with the discrepancy at maximum response
being roughly 30%.

03-

0.1

/ ~.~.~

Chimney No 5
The experimental program for Chimney No 5 has been
fully described by Christensen and Askegaard, 7 the detailed
observations were kindly made available to the authors
and have been re-analysed by Basu 12 in considerable detail.
The results presented in Figure 7 show a marked discrepancy between the predicted and observed tip displacements.
The maxima are in fair agreement but the variations with
speed are completely different. This major discrepancy
prompted a more detailed study and use was made of
measured values of ~L derived from pressure tappings at
a height of 118 m above ground. It was found that if the
quantity, )~(h)/~L, was examined then the variation with
wind speed did indeed follow the predicted form although
the observed values were roughly 25% below the predictions as shown in Figure 8. A further analysis of the data
indicated a strong dependence of CL on the intensity of
turbulence as shown in Figure 9. Figure 9 shows the variation of CL (total), the full spectrum value of CL, as a

Eng. Struct., 1984, Vol. 6, October

327

Response of RC chimneys to vortex shedding." B. J. Vickery and R. I. Basu


Observed
015

"E 0 1 0

0.05

/ .l L.--'"
I

15

Figure 7 Predicted
NO

20
25
~(h), ( m / s )
and

30

peak t i p deflections f o r Chimney

observed

57

function of the intensity of the longitudinal component


of turbulence. Spectral measurements indicated that about
one-half of the variance ( C [ ) was associated with a 'peak'
at the shedding frequency and hence, C L (vortex), the lift
coefficient associated with shedding is aboul 70(/, of
CL (total). Also shown in Figure 9 is the value o f Ct.
derived from quasi-steady assumptions and the fact thal
this is small is an indication that buffeting due to lateral
velocity fluctuations cannot alone explain the marked
increase in C L with turbulence.
Values of CL (vortex) derived from the observations of
Chimney No 3 (using the assumption that the sole cause of
the discrepancies between the predicted and observed
displacements was due to errors in CL) are plotted in
Figure 9. These cannot be given great weight but they do fit
the general pattern of the results from Chimney No 5.

Observed
A [Predicted]
v . [ ~ J
: 1,35

08

Tower No 1

Predicted
(equation I)

The limited experimental program conducted for Tower


No 1 (Figure 10) has been described by Vickery. a Field
observations were limited to one set of observations
although motions of a similar order (as gauged by occupant
response) had been noted on numerous occasions with wind
speeds in the vicinity of 12 m/s. The predicted and observed
r.m.s, amplitudes are in fair agreement as are the peak
values derived from samples of varying length.

C of V = 24%

~Q6

./f-.-.-aQ= 7-:-

~..~J l ) . /

0.2

16

17

" - - . . . . ' 0 8 ~ Predicted


~06
Pred icted

..,~'"

18

20

19

22 213 214 25

21

d (h),(m/s)
Figure 8 Comparison o f adjusted peak deflections f o r C h i m n e y

Tower No 2

No 5

The experimental program for Tower No 2 has been fully


described by Shears. 9 Since a number of modes contribute
to the tip motion (Figure 11), the simple expressions
(equations (1) and (3)) could not be employed. The predicted and observed peak tip displacements are shown in
Figure 12 and, in the vicinity of tile predicted maximum
responses in the dominant modes (modes 1 and 4), there is
fair agreement with the observations. These results are
discussed in greater detail elsewhere by Basu and Vickery 2
and by Basu. 12

Stigsnaes observations ((CL(total)), ( Ref 7)


o Inferred from MT ISA ((~L(vortex)),(Ref 5)
0 25
Trend CL (total) vs u / u ~ /
(Stigsnoes data)
/
~

02C

,k~ 0.15
/
010

Trend CL(vortex) vs u/u


(Stigsnees date)

/"

/"

005
o

~L (Quasi-steady with Cd=0.7, ~/5 = 0 8 )

~)
0 6

008

0.10

012

014

Figure 9

Variation o f r.m.s, lift coefficient (L?L) w i t h intensity


of turbulence

200 m chimney
Tip displacements of a 200 m, 15 m diameter reinforced
concrete chimney have been reported by Sanada and Nakamura ~3 but since no information was presented in regard to
the structural properties no comparison with predictions
was possible. The results obtained, however, include valuable
observations of the mean drag and the r.m.s, lift and drag
coefficients measured at 140 m above ground. Selected
results from the work of Sanada and Nakamura are shown
in Figure 13 and of particular interest is the variation of

( ~ N J o v _ rail h g
z

<15

-~

Glas~ - rein f o r ~ '


Fplast ic shroud

O~

15

_:

~
305m I

274 3 m

"

~1.0

Meesured

g~ell d i a m e t e r = 6 5 m

17" %1'4~ I, m,',,

10

rr
/
l
5

(stem
L
10

o'

1 0
1000
NO o f c y c l e s , ( N )

co~
u!

only)
l
15

~(h),(m/s)
Figure 10 Predicted and observed r.m.s, and peak accelerations
f o r Tower No 18

328

D,ameter. 3 66 m
'

peek

Eng. Struct., 1984, Vol. 6, October

Freq&.~ncy
Ob~rv~
0 26 ~;z
Calculsted 0 26

Figure 11

|
053
0 51

096
10

Shell d i a m e t e r 24 4 m

1 54
1 69

Mode shapes and frequencies f o r Tower No 2 ~

Response o f RC chimneys to vortex shedding: B J Vickery and R I Basu


Emley Moor T.M Tower
= 322m)
( ( ~ v e d peak
deflection (allmodes)
EI

0.2

~'DCL and, CD with Reynolds number. Noting that the


spectra of CD and ~, the r.m.s, longitudinal component of
turbulence, are very similar it is reasonable to assume that:

/
J ~--Predtcted
L~
(all modes)

CD/CD ~ {~/U = i u

where iu is the intensity of the longitudinal component of


turbulence. This assumption permits CL and C'D to be replotted as functions of iu as shown in Figure 14. While the
level o f scatter is large, it is clear that these results are
consistent with those in Figure 9. The spectra of Cr shown
in Figure 13 also confirm the observations from Chimney
No 5 that the spectral peak near the shedding frequency
contributes a little more than half the total variance. The
drag coefficient, Co, exhibits a dependency on turbulence
and, although this cannot be separated from a dependency

'~01

10

20

30

40

Et (h ) , ( m / s )

Figure 12 Comparison of predicted and observed peak tip deflections for Tower No 29

0.3

03

0.7
~o
0,2

OOoO

.:

.-

05

ii o
0

0.3

0.1--

02

Ol

0.1
I
2

I
3

Re (= 107)

Re ( =I0 7)

15

C1

Su (f) / ~
Rec 10- 2

SL(f)/oL2
C2

C3

10

c610 5
15

Rec 15- 1
C4

C5

2
0
0

05
Hz

10

2<
0
0

Figure 13 Observations by Sanada and Nakamura at level 140 m on


a 200 m high, 15 m diameter, RC chimney:
(a) Strouhal number as a function of Re where R0 ~ u(140) x 106
(b) Mean drag coefficient (~'D), and r.m.s, drag (CD) and r.m.s, lift
(L~L) coefficients
(c) Spectra of longitudinal turbulence [Su(f)], drag force [SD(f)]
and lift force [S L (f) ]
C1,C2, C3 Wind o f f l a n d , O ( 1 4 0 m ) ~- 17m/s, ou/O ~- 14%,
CL ~- 0.19 and CD ~--0.18
C4, C5, C6 Wind o f f water, D (140 m) ~- 37 m/s, au/O ~- 7%,
CL ~ 0.07 and CD ~- 0.07.

i
05
Hz

10

05
Hz

10

The spectral peaks in C3 and C6 can be represented by:

fSL(f)
where
f o r C3,
B = 0 2 4 , and

k2

d( f - - fs)~2/

k = 0.72 and B = 0.36 and for C6, k = 0.79 and

CL (vortex)
C L (total)

Note that the spectra shown are such that f

Eng Struct, 1984, Vol

S ( f ) d f = o -2

6, O c t o b e r

329

Response of RC chimneys to vortex shedding: B. J. Vickery and R. I. Basu

on Reynolds number, the former appears, in the light of


the (limited) data at transcritical Reynolds numbers, the
more likely. The dependency of CL on turbulence is discussed further elsewhere. 14

an 'expected' wind speed even though tile assessed probabilities of the two events are comparable.
Relative significance of drag loads

Discussion of predicted and observed responses


In the group of structures examined the ratio of the predicted to the observed maximum tip displacements averaged
about 1.05 with a coefficient of variation of about 25%.
The use of detailed computer analyses rather than equations (1) and (3) would result in a decrease of the average
to about 1.0 or a little less, but the coefficient of variation
would remain much the same. Two questions that can be
usefully addressed are:
1. What is the prime cause of the comparatively high
level of scatter?
2. What level of scatter (reliabifity) can be considered
acceptable?
It would be somewhat premature to answer the first
question conclusively but, at this stage, the prime source of
error appears to be the failure of the approach adopted to
recognize the dependency of the aerodynamic parameters
on turbulence. The predictions will be reviewed in detail
at a later date using the Sanada and Nakamura data but a
preliminary examination indicates that the discrepancies
for Chimneys 3 , 4 and 5 and Tower 2, for which turbulence
data is available or can be estimated from site descriptions,
would be significantly reduced. There is insufficient turbulence data available at present to review the predictions for
Chimneys 1 and 2 and Tower 1.
In regard to the second question there is no doubt that
an improved level of reliability would be most welcome
but, in comparison to the reliability of predictions of drag
loads, the present estimate of 25% is perhaps acceptable.
For a given wind speed there is no doubt that drag loads
can be predicted with a coefficient of variation (reliability)
significantly better than 25% but the reliability of the
predictive model is only one part of the overall problem.
The design load can be considered as an estimate of the
maximum load expected in the life of the structure and the
load factor as a multiplier required to account for uncertainties arising from inadequacies in the predictive model
and uncertainties arising from the variability in wind speed.
In the case of drag loads which increase roughly as the
square of wind speed, it is the latter source which is dominant. Assuming a Type 1 extreme value distribution for
wind speed with 'typical' mode/slope ratio of 10 the coefficient of variation of the wind load due to wind speed
uncertainties only is roughly 18% and the overall level of
uncertainty is closer to 20%. In the case of the across-wind
response (which, for the most part, will attain a maximum
at speeds below the design speed) the overall level of
uncertainty is determined almost entirely by the inadequacies of the predictive model and is, therefore, about
25%. It would appear then, that while the level of uncertainty in the prediction of the across-wind or vortexinduced response of chimneys is high it is comparable with
that which, in conjunction with present-day load-factors,
is considered acceptable for drag loads. Notwithstanding
the above discussion, there is a distinct practical difference
between a failure due to the occurrence of a wind speed
with a return period of perhaps 10000 years or more and
a failure, due to the inadequacy of a predictive model, at

330

Eng. Struct., 1984, Vol. 6, October

One advantage of working with the approximate relationships (equations (1) and (3)) is that it permits a more
general exploration of trends than is possible with a more
exact computer solution. The mean modal coefficient
induced by drag loads (do) can be computed for a chimney
with a straight taper if it is assmned, as before, that the
mode shape is of the form, if(z) = (z/h) 2, and that the
mean wind profile follows a power law with an exponent
a. A simple analysis yields the result :

CDPdohft~t
1
dD -- 8rflf2om~e ~ 3 + 2 ~ - -

1-0 }
4+2o~

The maximum r.m.s, lateral response due to shedding can


be expressed in a similar form and for a = 0.15 and/3 = 0.25
it follows that :

CLPdoh~ ~
l
~1/2
(tZv)max -- 87r2f2me {)tO(/3s + 13a)I f(O)
where:
)to = h/do
0

= dh/d o

dh = tip diameter
do = base diameter
Uh = mean speed (at height h) at which the lateral
response attains a maximum.

f(O) = O.115

(1 + 50) 3

t ''2

6 + 2(1 + 50)/)t01
for 0-+ I
(0 > approx.)

and:
0.039

f(O) - - -

for small 0
(0 < approx.)

( 1 - 0 ) 2.8

The ratio of the peak lateral response to the peak drag


response at the speed (Uh) at which the lateral response
attains a maximum is then:

{lv _ gfL [
~iD

} '/2

h(O)

GCD )to(~s + ~a)

where:
0.67(1 + 50)3/2(1 + (1 + 50)/3)to) in

h(O) =

1+3.30
for 0 -~ 1

and:

h(O) =

0.55
( 1 + 3 . 3 0 ) ( 1 - - 0 ) 2.8

for small 0

g = the peak factor (typically about 4)


G = the along-wind gust factor (typically about 2).
For the two limiting cases of a uniform chimney (0 = 1)
and a strongly tapered chimney (0 -+0) we have:

h(O = 1) = 2.29(1
h(O = 0 ) = 0 . 5 5

+ 2 / ) t o ) 1/2 ~ 2.4

Response of RC chimneys to vortex shedding: B. d. Vickery and R. I. Basu


Two questions of interest can be investigated with the
use of the foregoing results. The first is whether the drag
loads add significantly to the maximum lateral loads and
the second is whether the overall design will be dominated
by drag or by lateral loads. Recognizing that the first question is only of interest from a design point of view if the
lateral loads exceed the drag loads and then assuming, with
justification, that the dynamic drag loads are uncorrelated
with the dynamic lateral loads, the response in two dimensions will be as illustrated in Figure 15. For this situation
the peak combined response, ac, is given by:
~2

^2 + ~i~

1+

ac = av

where E is the Young's modulus of concrete and p the


density o f concrete.
In view of the weak dependence of the frequency on
the ratios of tip diameter to base diameter and tip thickness
to base thickness (th/to) these will be omitted as variables
and the approximation:
0.25 ~_E ~ 850 m/s
will be employed. If we now accept a nominal design speed
of 45 m/s and note that:
1

(av/(G -- 1) a D)2 __ 1

fth =sfodox3{3(1--O)I'ls

For a typical gust factor of 2 the ratio ac/av will not exceed
(4/3) v : and will approach (1 + (8D/#v)~) u2 as the lateral
loads become more dominant. It can be concluded that if
the peak lateral loads exceed the peak drag loads then the
maximum combined load will be no greater than 115% of
the computed peak lateral load and it is adequate to combine the across-wind (lateral) and drag responses using the
approximation:

for small 0

C7)

fth = S lode - -

for 0 -+ 1

we have:

(lv

1 ( g(]L I (

IU2

r(O)

where :

ac ~-- av{l .-1-UDr.2lUvl'~~.1/2


2
In approaching the second question as to the critical load
case for a given chimney, it is not intended to obtain
accurate assessments but to examine trends. The ratio of the
peak across-wind response to the drag response at the design
speed UD (the mean design speed at height h) is:

r(O) =

121
for small 0

(1 + 3.38) ( 1 - - 0 ) zs

9.56(1 + 50)7/2(1 + (1 + 50)/3Xo) 1'2


1+3.38
for0~l

(riD)design

h(O)

\UD ! " ~ D "

~-0(fls + fla)

Since the ratio/~h/~O is obviously critical, further progress


can only be made if an expression for the frequency of the
chimney is introduced and this requires the use of an
empirical relationship. Such a relationship has been suggested by Rumman lO after examining some 12 existing
chimneys and is:

Considering the two limiting cases of 0 = 1 and 0 ~ 0 we


have for 0 = 1 :

av

gel ] [ l _ ~ L ~ ' / 2
X4"s \GCD $2] ~ s + l~a]

_ 1200(

(aD)design
and for 0 = O:

av

_121./

(aD)desig
n

o-

o. o

~(

1'/2

\G----~-~DS2I
\~s--~al

It is clear that the aspect ratio, X0, is the most significant


variable and it is of interest to find the value of Xo at which
shedding forces dominate the design. For a strongly tapered
chimney typical values would be:

or:

Io

~.4.5

gel

oT.,

g = 4, G = 2, CL = 0.15, ~ D = 0.6, S = 0.20, l = 1,


3s + 3a = 0.008 a n d av/(~D)design ~ 16 900/X 4"s

0.25

E-~~

15.0 m diometer
[(IMeasuring position

0.20

0.15

I ...~;!L; i - ~,'~o o.
X C------~ .~,~ "-,,..,.~..l-:-a'. -c~

,G

1.2

1.0

o8
0.6

OlO

elo,...~.~Jooo
~ o

o.o~

-./"

~"

~ . ~ . i

0.2

CL (Ouosi-steody with Co=0.6, i, = 0.8iu )


,

0.4

~ . ~ . ~ "

10

15

20

( " ~ o / ~ o -- ,~)0/,
Figure 14 Variation o f t~ L and CD with turbulence intensity

and hence shedding will dominate for k o < 8.7. For a


uniform chimney C L is strongly dependent on Xo and
cannot be regarded as constant, a better approximation
for k < 25 would be CL ~ 0.03 ~/X. For the same values
of the other variables we have:

t~v/(t~D)design ~ 33 500/X~
and hence shedding will dominate for X < 13.5. Similar
calculations yield:
for0
for0
for0
for 0
for 0

=0.2;
=0.4;
=0.6;
= 0.8;
= 1.0;

Xo<
Xo<
ko<
Xo <
ko<

8.7
9.6
11.5
12.0
13.5

or
or
or
or
or

h/dh<44
h/dh<24
h/dh<19
hid h < 15
h/dh < 13.5

The above results apply for concrete chimneys similar


in proportions to existing structures. If, with higher strength
concretes and perhaps prestressing, there was a move to

Eng. Struct., 1984, Vol. 6, October

331

Response o f RC chimneys to vortex shedding: B. J. Vickery and R. I. Basu

L_
lr/ / W'-:v

tO

drag loads. An examination o f the relative significance o f


along-wind (drag) and the loads due to vortex excitation o f
the fundamental mode shows that the latter will probably
dominate the design if the height to base diameter is less
than about 13 or 14 for structures with no taper and less
than about 9 for strongly tapered structures, unless the
above ratio is sufficiently small to raise the critical speed
well beyond the design speed.

References
1

Q;
._1

Drag response, ao
=

-r 6o

3
4

6b

5
--I

Figure 15 Combination of across-wind (lateral) and drag response

6
7

thinner walls (smaller values o f m / p d 2) and/or lower values


of structural damping, values of (/3s + ~a) may approach
zero (or be negative) and severe motions, such as those
that have been observed in the case of steel chimneys, will
result. In addition, the analysis was directed only at maximum responses and the question o f concrete (or steel)
fatigue due to repeated loads below the design values was
not addressed.
Vortex excitation will cease to be a consideration if the
critical speed is well beyond the design speed. The value
Vc/Vdeagn above which shedding can be ignored is debatable but, if it is accepted that the load factors used for
drag loads are not excessive and that these loads increase
roughly in proportion to velocity squared, then it can be
argued that the response due to shedding should be examined for ratios o f Vc/Vdesig n up to, at least, Lv/L-Fwhere
L F is the load factor used in conjunction with the drag
loads.

9
10
11

12

13
14

Conclusions
Comparisons between the observed full-scale behaviour o f
seven reinforced concrete structures o f circular crosssection and predictions using a model suggested by the
authors indicate that:

on average, the predictions agree, to within about 5%,


with the full-scale observations;
the predictions show a high level of scatter with a
coefficient o f variation o f about 25%;
the prime cause o f the scatter appears to be the failure
o f the model to recognize the dependence of the aerodynamic parameters on free stream turbulence.

While the predictive model exhibits a high coefficient o f


variation, it is demonstrated that this is not markedly
different from that associated with the prediction o f the

332

Eng. Struct., 1984, Vol. 6, October

Vickery, B. J. and Basu, R. 1. 'Across-wind vibrations of


structures of circular cross-section - Part I - Development
of a mathematical model for two-dimensional conditions',
J. Wind Engineering and ln.dust. Aerodynamics, 1983, 12
(1), 49-73
Basu, R. I. and Vickery, B. J. 'Across-wind vibrations of structures of circular cross-section - Part II - Development of a
mathematical model for full-scale application', J. Wind
Engineering and Indust. Aerodynamics, 1983, 12 (1), 75-97
Wootton, L R. 'The oscillations of large circular stacks in the
wind', Proc. Inst. Or. Eng. 1969, 43, 573-598
Vickery, B. J. and Basu, R. I. 'Simplified approaches to the
evaluation of the across-wind response of chimneys', Proc. 6th
Int. Conf. on Wind Engineering, Gold Coast, Australia, March
1983
Melbourne, W. H., Cheung, J. C. K. and Goddard, C. R.
'Response to wind action of the 265 m Mt Isa Stack', Proc.
ASCE, J. Struct. Div. 1983, 109 (11), 2561-2577
Muller,F. P. and Niesser, H. 'Measurements of wind induced
vibrations on a concrete chimney', Z Indust. Aerodynamics,
1976, 1 (3), 239-248
Christensen, O. and Askegaard, V. 'Wind forces on and excitation of a 130 m concrete chimney', J. Industrial Aerodynamics,
1978, 3 (1),61
Vickery, B. J. 'Three studies involving a comparison of fullscale behaviour with that determined theoretically and/or
experimentally', Symposium on Full-Scale Measurements of
Wind Effects on Tall Buildings and Other Structures, The
University of Western Ontario, London, Ontario, Canada,
June 1974
Shears, M. 'Report on wind and vibration measurements taken
at the Emley Moor Television Tower', J. Industrial Aerodynamics, 1976, 1 (2), 113-124
Rumman, W. Private Communication, Dept. of Civil Engineering, University of Michigan, Ann Arbor, Michigan, USA, 1983
Basu, R. I. and Vickery, B. J. 'A comparison of model and fullscale behaviour in wind of towers and chimneys', Wind Tunnel
Modelling for Civil Engineering Applications, Cambridge
University Press, 1982, pp 654-668
Basu, R. I. 'Across-wind response of slender structures of
circular cross-section to atmospheric turbulence', PhD Thesis,
t:ac. Eng. Science, The University of Western Ontario, London,
Ontario, Canada, 1982
Sanada, S. and Nakamura, O. 'Full-scale measurements of
wind forces acting on a 200 m concrete chimney', Kajima Inst.
of Construction Technology, Chofu City, Tokyo, Japan, 1983
Vickery, B. J. and Daly, A. 'Wind tunnel modelling as a means
of predicting the response of chimneys to vortex shedding',
Int. Conf. on Design Against Wind-Induced Failure, Bristol,
England, January, 1984, Engng Struct. 1984, 6 (4), 363-368

Appendix

Sample calculations for Chimney No 1


h = 244m;

a = 12.6 m;

me/pd z = 121"

f0=0.33Hz;

d(0)=22.8m;

d(h) = 10.6 m;

g = 3.8;

KaT = 0.48 ;

S = 0.216;

me = 23 000 kg/m;
CL = 0.158;
X = 19.4;

Response o f RC chimneys to vortex shedding: B.

8rr 2 x 0.2162
~(h)

1
x

121

V~

{
x

! 1/2

2 ( 1 ~ - + 2)J

~B,k)

1.32 x 106 X x2(l -- 0.535X) 4

~d'lx/t

(1 + 0.33/x) u2

=h

X2(1 --0.535X) 4

{ l f (z/h)2dz}l/2{O.O1

0481"2~1

.9 = 3.71

=/1-k-1~2|

34.8(1 --0.535x)
Uh =

k = fi(h)/19.8
a(h) a/2

y(h) = 0.199 ~(h)


19.8

1
0.216 x 0.33(22.8 -- 12.2x) = ~(h)-x ~

g = a(zc)/ac = a ( h ) / 1 . 0 3 a c

~(B,k) = 2 19.8

(1 + 0.33]x) 1/2

x = Ze/h; ze = height at which (r~/Ze) = fo(Ze)/S

w2

~(8,k)=~k exp[-~t~)~; B=0.25


ao = f o a / S ;

Vickery and R. I. Basu

Putting.x = ze/h

Treating the chimney as slightly tapered


3.8xO.158x12.6xl

,I.

{ (19"8121
exp --8 1 - - u _ ~ ! !

3/2
exp

{ (19.812
--8 1
----~]

~
! m

xd4(ze) J(ze)/x/t

Y = 81r2 x 0.2162 x 23 000 x 244/5 x (0.01 --0.48/121) 1/2

da(ze) = (22.8 - 12.2ze/h) 4


~(ze) = (ze/h) 2
t = 0.05 + 0.15(22.8 -- 12.2ze/h)/ze

m/s

Hence:

Treating the chimney as significantly tapered


3.8 x0.158 x 1.0 x 1.2 x ~

X0.15

x = ze/h

.~', m

~(h), m/s

1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2

0.150
O. 186
0.214
0.229
0.228
0.207
0.168
0.114
0.058

16.2
18.3
20.6
23.0
25.5
28.3
31.4
35.0
39.6

Eng. Struct., 1984, V o l . 6, O c t o b e r

333

Anda mungkin juga menyukai