You are on page 1of 8

ACI JOURNAL TECHNICAL PAPER

Title no. 80-32

Cracking Load for a Wind-Loaded Reinforced Concrete Cooling


Tower

by Kazuya Hayashi and Phillip L. Gould

Since the collapse of several large cooling towers in Ferrybridge, by rapid propagation of cracks in tensile zones fol-
England, a major concern of designers of such shells has been to in- lowed by temporary stiffening and, finally, by yielding
sure adequate safety under wind load. Prior to attaining the classical
buckling state, a failure of a reinforced concrete cooling tower under
of the reinforcement. 1 The methodology required to
wind load may be initiated by rapid propagation of cracks in tensile establish this failure pattern is a very detailed finite ele-
zones. In this sense, estimating the cracking load is very important for ment analysis incorporating geometrical and material
the design of wind-loaded cooling towers. nonlinearities and is beyond the scope of normal design
This paper shows that the cracking load for a wind-loaded rein- procedures.
forced concrete cooling tower may be obtained by a linear elastic fi-
nite element analysis which provides a lower bound to the collapse This paper's objective is to investigate certain aspects
load of such shells as calculated by more sophisticated methods. of the response of a wind-loaded reinforced concrete
cooling tower and estimate the cracking load, which is
Keywords: cooling towers; cracking (fracturing;) finite element method; hyper- a close lower bound to the failure load, by a linear
bolic parabolic shells; loads (forces); reinforced concrete; shells (structural
forms); static loads; stresses; structural analysis; tensile strength; wind pres-
elastic rotational shell finite element analysis. 2 This is
sure. the present level of state-of-the-art design practice.

Hyperbolic natural draft cooling towers are the larg-


est reinforced concrete thin-shell structures being built
today. Use of cooling towers has increased rapidly in
the last several decades in the United States and Eu-
rope, and natural draft cooling towers are perhaps the
REINFORCED CONCRETE
most prominent structures in many large nuclear and HYPERBOLOIDAL SHELL
fossil power plant facilities. An example of a hyper-
bolic natural draft cooling tower is shown in Fig. 1.
Although a natural draft cooling tower consists of sev-
eral structural units (see Fig. 1), the shell or veil re- WATER
DISTRIBUTION
quires the most attention in design and construction,
SYSTEM
especially with the trend toward higher towers during
the past few years.
In 1965, three of the eight cooling towers at a station AIR INLET

in Ferrybridge, England, collapsed during a fairly


moderate wind. Since this catastrophic failure there
has been a significant amount of research devoted to
understanding the structural response of such shells
under wind load. Buckling under wind load has been
regarded as one of the most important factors to con-
sider in the analysis and design of hyperbolic cooling
Fig. }-Hyperbolic natural draft cooling tower
towers and a possible contributing cause of the Ferry-
bridge collapse. Recently, Mang et a!. investigated a
typical hyperbolic cooling tower under wind load and Received June 16, 1982, and reviewed under Institute publication policies.
Copyright © 1983, American Concrete Institute. All rights reserved, including
demonstrated that a tower made of reinforced concrete the making of copies unless permission is obtained from the copyright propri-
etors. Pertinent discussion will be published in the May-June !984 ACI JOUR-
may not buckle in the classical sense but is likely to fail NAL if received by Feb. I, 1984.

318 ACI JOURNAL I July-August 1983


Ka;;uya Hayashi was a graduate student in civil engineering at Washington
Table 1 - Coefficients for determining the
University, St. Louis, from 1980 to 1982 when he earned his MSCE. He has a equation of the meridian curve
BS in civil engineering from the National Defense Academy (Japan) and is a z~O z;;,o
first lieutenant in the Japan Ground Self Defense Force. (Hyperbola) (Ellipse)

Phillip L. Gould, Harold D. Jolley Professor and chairman of the civil engi- a -0.01506 -0.28035
neering department at Washington University earned his PhD at Northwestern b 0.0 0.0
University where his research dealt with thin shell structures. Dr. Gould has c 1.0 1.0
published almost 100 technical papers and wrillen three books. He is the d 0.0 0.0
founding editor of the journal, Engineering Structures, and has been awarded e -224.596 252.22546
a senior U.S. Scientist Award by the Alexander V. Humboldt Foundation in f 12563.3411 -44269.0087
West Germany for his work concerning shell structures.

Table 2 - Material properties of concrete and


126.31 It steeP
100.07 ft
Concrete

lit. It ft
E 5.836 X 10' lb/ft'
/-' 0.175
')' 152.6 lb/ft'
p 4. 739 lb · sec' /ft'
J: 65858.4 lb/ft'
Steel
ltl.70 ft
1.0 in /, 8.64 X 10' lb/ft'
I lb/ft' = 47.84 N/m'; llb/ft' = 156.95 N/m'; and lib· sec'/ft' = 514.94
10.11 in N · sec 2/m 4 •

In the finite element analysis, bending and twisting


moments as well as in-plane or membrane stress resul-
tant were computed. It is well established that bending
moments have only a minor effect on properly propor-
Fig. 2-Geometry of the analyzed cooling tower (1ft = tioned cooling tower shells, which are thickened at the
30.48 em; 1 in. = 2.54 em) top and base. Therefore, in the interest of simplicity,
they are not considered in the following design calcu-
ANALYSIS OF A HYPERBOLIC COOLING lations. However, the in-plane stress resultants re-
TOWER ported are found from an analysis based on the bend-
In this section, a reinforced concrete cooling tower is ing theory, and they may be somewhat different from
analyzed for dead load and wind load. The dimensions those which would be found from simple membrane
are typical of the largest towers currently under con- theory, especially near the boundaries.
struction. Stress resultants N~, N 0 , and N 0., with their respective
sign conventions are shown in Fig. 3. The subscripts¢
Geometry and assumptions and (} are the curvilinear coordinates, the meridional
The geometry and the thickness profile of the shell is angle and circumferential angle, respectively.
shown in Fig. 2. The minimum shell thickness is 8 in.
(20.3 em) which can easily accommodate two layers of Dead load analysis
reinforcement in each direction and provide adequate For the finite element analysis, the entire shell is
cover. 3 Above the throat the meridian curve is a hyper- modeled by 12 doubly-curved ring elements as shown in
bola while below the throat it is an elliptical segment. Fig. 4. Letting q denote the applied dead load per unit
The equations for these two parts of the meridian curve area of the middle surface of the shell, given by
are obtained by substituting the coefficients a, b, c, d,
e, and fin Table 1, obtained from the exact geometric q = pgh (2)
data, into the equation for a general curve of second
order given as in which g is the acceleration of gravity, the compo-
nents of the load q in the meridional and normal direc-
aZ 2
+ bRZ + cR 2 + dZ + eR + f = 0 (1) tions are

where R = R(Z) is the radius of the shell at vertical co- q., = q sin¢ (3)
ordinate Z as defined on Fig. 2. The simplifying as-
sumption of a circumferentially continuous hinged base Q. = - q cos cf> (4)
was made for this study, although the discrete system
of supporting columns should be considered in actual respectively (see Fig. 3 for sign conventions).
design. 7 The modulus of elasticity E, Poisson's ratio J.t, Results of the dead load analysis are shown in Fig. 5.
;,pecific weight )', mass density p, uniaxial tensile Except for a small region above the throat, the entire
strength of concrete fi, and yield stress of steel f, are shell is in a state of biaxial compression under this
given in Table 2. loading condition.
ACI JOURNAL I July-August 1983 319
HEIGHT
~oo'
...
/

400'

200'

I 100' Net>
I
STRESS RESULTANTS I
I
Ne
I
I
Fig. 3-Sign convention for stress resultants I
I

''
NODAL POINT NUMBERS ~ I KIPS/FT)
-50 -40 -30 -20 -10 0 10 20 30
\ 126.:38"
ELEMENT NUMBERS .. -'o - - - - - -

Fig. 5-Stress resultants for the cooling tower under

~
22.91'
19.29'
dead load (1 kip!ft = 14.58 kN/m; 1ft = 30.48 em)
19.29'

~-Rlt
19. 29'
19. 29'
on specialized wind tunnel studies or full-scale mea-
119.19" surements.3
62. 14' ® I In this case, the loading takes the form
I

62.14"
q~ = 0 (a)
q9 = 0 (b)
62.14' ,..,,, Qn = Qnz (Z)H9({)) (c) (5)
I

62.14' The vertical variation qnz is taken from Reference 3


using the 100 year return period isotach chart and Ex-
62.14' posure C.
The circumferential distribution H 0 is usually defined
by normalized values at equal angle increments from
_j the windward direction as shown on Fig. 6. Such a cir-
cumferential distribution of wind pressure may also be
195.80' represented by a Fourier series in the circumferential
angle 0, i.e.
z

Fig. 4-Discretization of the analyzed cooling tower (1


ft = 30.48 em) E d cosjfJ (6)
j =0

Wind pressure coefficients and the corresponding


Wind load analysis Fourier coefficients used for this wind load analysis are
The present method of designing cooling towers for given in Table 3 (including internal suction). 1
wind loading is basically pseudostatic, with the magni- The results of the wind load analysis under the de-
tude and vertical variation of the wind forces taken sign wind velocity of V130 = 90 miles per hr (40.2 me-
from general sources that are documented in codes and ters per sec) are shown in Fig. 7, 8, 9, and 10. Fig. 7
standards and the circumferential distribution based shows the variation of the membrane stress resultants
320 ACI JOURNAL I July-August 1983
e-11 o·
-- -)
,
(
I
I
\

...
WINO

O.IOPr
\
I
I
I
I
I
I
400

I
I 300'
I
I

Fig. 6-Circumferential wind pressure distribution


zoo'
along the windward meridian while Fig. 8, 9, and 10
show the variation of the membrane stress resultants
N 0 , N,,, and N 66 , respectively, in the circumferential di- e= o·
rection at selected nodal points as denoted on Fig. 4.
100'
The maximum tensile and compressive meridional
forces tend to coincide with the locations of the peak
Ne
pressure and suctions as shown on Fig. 6.

DETERMINATION OF THE CRACKING LOAD


Principal stress resultants and principal I KIPS/FT) -20 -10 0 10 20 30 40 50 60
directions
Considering a varying wind load in combination with Fig. 7-Stress resultants for the cooling tower under
a constant dead load, the stress resultants may be writ- wind load (I kip/ft = 14.58 kN/m; I ft. = 30.48 em)
ten in the form
Table 3 - Wind pressure coefficients 1
N 0 v + AN6 ,..

+.:=
(a) () (deg)
Angle from
Nuo + ANnw (b) windward H,
ANu.,w (c) (7) meridian coefficients j

0 -1.56 0
in which, A is the wind load intensity factor and N"", 15 -1.24 I -0.27918
30 -0.74 2 I -0.61978
Now• and Nrlow are meridional, circumferential, and in- 45 -0.03 3 -0.50927
plane shearing stress resultants due to design load wind 60 0.68 4 I -0.09167
75 0.82 5 0.11794
load. Similarly, N"'v and N 110 are meridional and cir- 90 0.37 6 I 0.03333
cumferential stress resultants due to dead load. 105 -0.06 7 -0.04474
Principal stress resultants in a shell element subject to 120 -0.15 8 I -0.00833
135 -0.06 9 I -o.oo9n
the aforementioned membrane stress resultants may be
computed from the theory of elasticity as follows'
\50
165
180
------------~---
'
-0.12
-0.09
-0.10
---~---
IU_
10
II
-0.01356
-0.11597
-0.01~67

(8) in which

in which F Nlloll
c 1
/z (N11 o - Noo)
A Y2 (Non + Nuv) D liz (Nuu - N¢w)
B 1;2 (Now + N,, .. )
J 14 (NOD - N 6 v) 2 and {3 is defined on Fig. 3. Fig. 11 depicts the variation
H Vz (Nov - N"'v) (Now of the maximum principal stress resultant N, in the cir-
N.,w) cumferential direction at several nodal points, while
I '!4 (N, .. - NouV + (N"""V Fig. 12 shows the principal directions for the case of A
= 1.0 (design wind load). Fig. 12 also indicates the
These forces act in two mutually perpendicular direc- load-carrying mechanism of the wind-loaded shell
tions given by without regard to the magnitude of the stress resul-
tants, assuming that the shell is uncracked. In the lower
regions of the shell, the maximum tensile values occur
tan 2{3 = (9)
C + DA at 0 = 0 deg (along the windward meridian).
ACI JOURNAL I July-August 1983 321
60 HP. 2

H.P. 4
50
N.P. 6

-40 N.P. 8
N.P. 10

10

0
I IOPS/FT I
-10
9 I DEGREE I
-20

-30

-40

Fig. 8-Meridional stress resultants under wind load (1 kiplft = 14.58 kN/m)

N.P. 2

N.P. 3
6.0
N.P. 6
<4.0
N.P. 9

2.0

Ne
0
I kiPS/FT I 12.0.: 1~::-;--150- ~-
...... _____ _ 180
-2.0
8 I DEGREE I

-4.0

-6.0

-8.0

Fig. 9-Circumferential stress resultants under wind load (1 kip/ft = 14.58 kN/m)
Cracking load the load intensity factor for the cracking load for a
Failure of the reinforced concrete cooling tower may varying wind load can be estimated at any level Z by
be initiated by rapid propagation of cracks in tensile
zones and, finally, by yielding of the reinforcement.' N~n (Z) + 'Aa (Z) N,.,. (Z, 0) = N,, (Z) (10)
Consequently, for design purposes it seems to be ade-
quate to predict the cracking load, which is thought to Now (Z, 0) indicates the meridional stress resultant due
represent a lower bound to the ultimate load. As noted to the design wind load along the windward meridian,
previously, Fig. 11 indicates that the maximum princi- 0 = 0 deg; 'A,, (Z) is the unknown load intensity factor
pal stress resultant occurs along the windward merid- for the cracking load; and N". (Z) is the tensile strength
ian, (} = 0 deg. Furthermore, Fig. 12 illustrates not per unit length of the concrete, calculated as the ulti-
only that cracks along the windward meridian would be mate uniaxial tensile stress of the concrete times the lo-
normal to this meridian (horizontal) but also that the cal thickness of the shell. Fig. 13 shows the variation of
windward meridian represents a trajectory of the prin- N""' N¢.,., N., 0 + N""' and N" along the windward me-
cipal stress resultants. In other words, as far as the ridian. Also indicated is the capacity of the vertical re-
meridional stress resultant is concerned, the windward inforcement A,f.. calculated based upon ACl Code
direction governs the critical loading condition. Now provisions•
322 ACI JOURNAL I July-August 1983
6.0

<4.0

2.0

Ne~
0
I KIPSIFT I 165 180
-2.0
9 I DEGREE I
-4.0

-6.0 H.P. 2

N.P. 4
-8.0
N.P. I
-10.0 N.P. 8

N.P. 10
-12.0

-14.0

Fig. JO-in-plane shearing stress resultants under wind load (1 kiplft kN/m)

30 H.P. 2

N.P. 4

N.P. I
20 N.P .

NP. 10

10

9I DEGREE I
0
I KIPS/FTI

-10

Fig. 11-Maximum principal stress resultants, ,\ 1.0 (1 kip/ft 14.58 kN/m)

¢(AJJ == 0.9D + 1.3W (II) concrete shell. This is most notable in the upper re-
gions which are controlled by the minimum steel pro-
and provided by uniformly spaced bars. Practically, visions. To fully compensate for the tensile capacity of
however, for about the upper one-third of the shell, the the concrete lost during cracking, the reinforcement
vertical reinforcement is controlled by the 0.35 percent would practically have to be doubled in this region of
minimum reinforcement requirement.' the shell, as indicated on Fig. 13.
Referring to Fig. 13 at nodal point 9, Z = 207.28 ft On the other hand, the circumferential stress resul-
(63.18m), the stress resultants were computed as No" tants shown in Fig. 7 are much smaller than the tensile
== -31.21 kips/ft (-455.0 kN/m), N¢" = 56.15 kips/ strength of the concrete, and due to these stresses, the
ft (818.7 kN/m), and N,,. = 43.87 kips/ft (639.6 kN/ concrete would not be expected to crack prior to the
m). Substituting into Eq. (10) gives the minimum value formation of meridional cracks sufficient to cause fail-
of A,., (Z), A,., = 1.34. It is also of interest to note that ure. On the other hand, there is evidence that thermally
the capacity of the steel reinforcement is not always induced horizontal cracks may appear once the tower is
even as much as the tensile capcity of the uncracked in service. 7 These were not considered.
ACI JOURNAL I July-August 1983 323
t ll i ~ I ~
I
400'
J ~ ~ ~ I
I

....
/I II\\\ \ I I
\//I I l \\\ r I
X
C>

""X
200'
\II I I I\\\ \ I
\Ill I I\\\ \ I
100'

\Ill I II\\ \ I
\ I I I II I I\ \
Fig. 12-Principaf directions(-: tension,-: compression), ,\ = 1.0 (1ft =
30.48 em)

500
CAPA~~T~~-A~f~ I Table 4 - Comparison of load intensity factors
r-----
/STEEL

~ Load intensity factors'

v
,
'
CONCRETE
CAPACITY
TENSILE
Ntc
No reinforcement
Regular
Cracking
1.28 ( 1.34)*
Ultimate
1.28
I
I:
reinforcement 1.34 ( 1.38)* 1.49
:~
I

L_,,,
~
DOUBLED STEEL
CAPACITY I 2A•fy I
Doubled
reinforcement 1.40 2.26
*The value in the parenthesis was obtained in this investigation.

300' area with the concrete area to compute N,,. It can be

'l Ll seen that this relatively simple approach provides a


close estimate of the cracking load and a lower bound
to the ultimate load of the shelL Regarding the contri-
bution of the reinforcement, cracking load is only
\ I moderately affected in both solutions. However, ul-
\I \. timate load may be substantially raised by increasing
l ,\
the reinforcement, as reported by Mang et a!., 1 and
cannot be replicated using the simplified approach of
I\ \
this paper.
i
I
I
\
\
\ Change of wind velocity and shell geometry
N<t>o N<t>D
+
N<t>w\
\ Since the thickness of a cooling tower has been ob-
\
served to have a significant influence on the cracking
load, it may be desirable to thicken the cooling tower
-40 -20 0 20 40 60 I KIPS/F'T I to insure an adequate load intensity factor against
cracking. The cracking load intensity factor was inves-
Fig. 13-Comparison of capacities (1 kip!ft = 14.58 tigated under changes in the shell thickness and design
kN!m; 1ft = 30.48 em) wind velocity. Design wind velocity Y;- 30 was taken as 80
(35.76), 90 (40.23), and 100 (44.69) miles per hr (meters
Table 4 lists the load intensity factors obtained by per sec) and the minimum shell thickness as 7 (17.78),
Mang et a\. 1 for cracking load and ultimate load, 8 (20.32), and 9 (22.86) in. (em), respectively. Fig. 14
with and without regarding reinforcement, along with shows the variation of the critical load intensity factor
the cracking load intensity factors obtained in this in- due to the changes in design wind velocity and shell
vestigation. Here, the cracking load with reinforcement thickness and confirms that thickening is indeed effec-
'A" = I .38 was obtained by using the transformed steel tive in increasing the cracking load.
324 ACI JOURNAL I July-August 1983
CONCLUSIONS
For the cooling tower considered, it was shown that
d"ecklnq ~o•d intenaity fotc:tor
the critical load intensity factor for cracking may bees- 2.0
I no r"eintorc.ment l oOtairwd ~ tr.t,enq et dl ( ll
timated by a linear elastic finite element analysis. This Cr"~kinq loU intensity factor"! ,.._,.rainq
factor agrees closely with the values obtained by a far ,..lnfo,.cemanf ) obf.iMd b~ M•n9 f't cHI I I
1.8
more sophisticated calculation (see Fig. 14 and Table uttlm•N lo.a intensity f~tor

4). It is suggested that the linear elastic approach is an u


obt•ined l)~ t.t•nll •' dl Ill

appropriate method for the design of the cooling tower ""


1.6
under wind load because, up to the cracking load, rein- "'t;
0

forced concrete cooling towers essentially behave line- ~ ft'-457psi


arly, and the cracking load therefore represents a lower ...,..
;jj 1.4
bound to the ultimate load. This is valid, provided that ~
...
the shell does not buckle at a lower load, as confirmed ~
c
by Mang et a!.' A further qualification is that imper- c-<
-' 1.2
fections in construction have been neglected in these -'
~
calculations. While this is beyond the scope of the ;::
t•8" !no reinfof"'C~)
~
present study, it is possible that such imperfections 1.0
could alter the stress resultants significantly and thereby t•7"!no ,-einforc~enT I

affect the calculated minimum Aa.


Fig. 13 shows that the designed steel capacity of only
the middle region of the shell exceeds the concrete ten- 80 90 100
sile capacity of the shell. The steel capacity would have DESIGN WIND VEJ.OCITY I MPH I
to be nearly doubled in some regions, as indicated in
Fig. 13, to exceed the concrete tensile capacity of the Fig. 14-Variation of the critical load intensity factor
(1 psi = 6896.55 Pa; 1 in.= 2.54 em; and 1 MPS
shell over the entire shell, which is generally considered 0.447 mps)
to be desirable in guarding against a strength collapse
failure in reinforced concrete beams. 8
Tower Shells-Practice and Commentary," (ACI 334.2R-77)(Revised
ACKNOWLEDGMENTS 1982), American Concrete Institute, Detroit, 1982, I 0 pp.
The many contributions of Professor H. Mang and Mr. C. Gue-
4. Thorn, H. C. S., "New Distribution of Extreme Winds in the
delhoefer to this study are deeply appreciated. Also, the first writer
United States," Proceedings, ASCE, V. 94, ST7, July 1968, pp. 1787-
wishes to express his sincere appreciation to Japan Ground Self De-
1801.
fense Force whose support enabled him to pursue this study. 5. Hayashi, K ., and Gould, P. L, "Some Aspects of the Response
of a Wind-Loaded Reinforced Concrete Cooling Tower," Structural
REFERENCES Engineering Research Report No. 63, Department of Civil Engineer-
I. Mang, H. A.; Floegl, H.; Trappe!, F.; and Walter, H., "Wind- ing, Washington University, St. Louis, May 1982, 108 pp.
Loaded Reinforced Concrete Cooling Towers: Buckling or Ultimate 6. ACI Committee 318, "Building Code Requirements for Rein-
Load?," Engineering Structures (Sussex), V. 5, No.3, July 1983, pp. forced Concrete (ACI 318-77)," American Concrete Institute, De-
163-180. troit, 1977, 102 pp.
2. Basu, P. K., and Gould, P. L., "SHORE-Ill Shell of Revolu- 7. Larrabee, R. D.; Billington, D.P.; and Abel, J. F., "Thermal
tion Finite Element Program-Theoretical Manual," Structural En- Loading of Thin-Shell Concrete Cooling Towers," Proceedings,
gineering Research Report No. 48, Department of Civil Engineering, ASCE, V. 100, ST12, Dec. 1974, pp. 2367-2383.
Washington University, St. Louis, Sept. 1977, 70 pp. 8. Large, George E., and Chen, T. Y., Reinforced Concrete De-
3. ACI-ASCE Committee 334, "Reinforced Concrete Cooling sign, 2nd Edition, Ronald Press, New York, 1957, 527 pp.

ACI JOURNAL I July-August 1983 325

You might also like