Professional Documents
Culture Documents
ABSTRACT: Vibrations of chimneys under the wind action occur in-line with the wind as forced, random vibrations and cross-
wind as vortex-induced vibrations. The former become important at extreme wind speeds, regular vortex shedding causes the
latter and occurs at any wind speed. In the case of vortex resonance, when considerable vibration amplitudes develop, the
structural displacements alter the fluid force, which is known as an aero-elastic coupling of force and structural response.
Regarding the random load caused by turbulent wind, Davenport's pioneering concept of the gust response factor was applied to
develop a simplified equivalent static load. Two different models are available regarding the design for vortex resonance. One is
Ruscheweyh's approach to deal with vortex resonance as a forced vibration and to incorporate the aero-elastic effects in model
parameters derived from experimental data. In the second model, the design for vortex resonance is based on Vickery & Basu's
model of a self-limiting response process. The Eurocode contains both models but relies primarily on the first model. The new
CICIND model code counts solely on the second approach. Experimental observations indicate the relative importance of
parameters such as the Scruton number, the intensity of turbulence and the mode of vibration. Regarding the in-line wind load,
the mechanical model also enters into the discussion. In some cases, the chimney dimensions indicate that a beam behaviour is
neither realistic nor on the safe side. Then, a shell model has to be applied and the wind load cannot be considered as a force.
Rather, pressures have to be utilized. The paper discusses the effect of these parameters and deals with the principal results
concerning the along wind action and the design for vortex shedding.
1385
Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014
151.0 0.709 11.675 0.350
= +
(4)
130.0 0.610 11.675 0.350
110.0 0.516 11.675 0.350 E
E E
90.0 0.423 11.675 0.350 Davenport showed that both components are proportional to
75.0 0.352 11.675 0.350 the intensity of turbulence Iv = v/Vm if the structural
55.0 0.258 11.775 0.550 behaviour is linear and the aerodynamic transmittance is
44.2 0.208 11.800 0.600 quasi-stationary. Introducing factors R and B, the following
30.9 0.145 11.800 0.600 expression for the gust response factor is obtained:
25.0 0.117 11.800 0.600
12.0 0.056 11.825 0.650 G = 1 + 2k I z B + R (5)
0.0 0.000 11.825 0.650
with:
1 1
B= and R =
Table 2. Geometry of sample chimney n.2 (steel) (6)
E 2I E 2I
z [m] z/h r [m] t [m]
40.0 1.000 0.403 0.0063 In equation (5), zs is the height at which the reference flow
17.0 0.425 0.403 0.0063 parameters are utilized for determining G. The gust response
16.0 0.400 0.530 0.0063 factor is a structural, not a load parameter. In it, B takes into
0.0 0.000 0.530 0.0063 account the reduction of the load effect due to the non-
simultaneous occurrence of the load peaks over the structural
surface whereas R accounts for the amplification of the load
2 ALONG WIND ACTION effect due to resonance with wind turbulence. It depends on
one hand on the wind flow parameters e.g. the mean wind
2.1 Design models for gust wind load
profile, the turbulence intensity, and the integral length scales
All gust wind models presently utilized in modern wind of the oncoming flow; it accounts on the other hand for
loading codes go back to the classical Davenport approach. It structural parameters such as the size of the structure and its
amplifies the mean wind force applying the gust response dynamic behaviour, in particular the lowest natural frequency,
factor G. The wind force, given as a line load, is appropriate the related mode shape, and the damping.
for slender, line-like structures such as chimneys: Design wind load models in the Standard Codes are
wz = G w z (1) revisited applications of Davenport's approach.
In the Eurocode formulation, an important difference to the
In equation (1), z is the height above the ground; G is the Davenport approach is that the Eurocode bases the wind force
gust response factor. The mean wind force wm(z) is based on on the peak (or gust) velocity pressure qp. It is defined as the
the mean qm of the wind velocity pressure q(t), fluctuating in short-term maximum occurring within the 10-min interval of
time: the extreme mean wind. The peak velocity pressure is the sum
w z = C
z dz q z (2) of the mean qm and the standard deviation q amplified by the
peak factor kq as in equation (7):
q z = q z + k ! !
in which CD is the aerodynamic drag coefficient, d is the
diameter of the chimney, qm(z) is the 10-min mean of the (7)
& V z
velocity pressure at level z. The mean wind force is a real
q z = "1 + 7I z$
physical quantity, whereas the wind force in equation (1) is an (8)
equivalent static load, intended to reproduce the effects of the 2
stochastic wind loading process on the most important where Iv(z) is the turbulence intensity at height z. The
structural stressing. transition from equation (7) to (8) involves two convenient
To derive the gust response factor, Davenport [1] approximations, namely (q/qm) 2Iv and qm V/2.
considered cantilevered, vertical structures and their response Furthermore, the peak factor has been chosen as kq = 3.5. It
to the wind action, namely the mean (static) and time corresponds to the gust wind speed in a 1-sec gust. It can be
dependent (quasi-static and dynamic) components. He defined shown, that the exact value of equation (7) exceeds the
the gust response factor G as the ratio of the peak wind effect, approximation of equation (8) by a factor of
Ep to the mean response, Em: {1 + (0.7Iv) 1.8}.
E E + k
G= = = 1 + k
(3) The use of the profile of the peak velocity pressure (7) is
E E E one of the merits of this model, because the profile of the
wind force over the building height results in a more realistic
where E is the standard deviation (or rms-value) of the image of the local gust loads at each level.
fluctuating response; kp is the peak factor, which is the ratio of The Eurocode introduces the following formulation to
the peak of the response fluctuation to its standard deviation, express the wind force acting on a structure or a structural
E. The load fluctuations due to wind turbulence provide a component:
wz = c z c) z C
A+,- q z
broad band excitation of the structure. The rms response is
split into resonant and quasi-static (background) components, (9)
ER and EB, so that:
1386
Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014
resonant amplification of the gust effect. The dynamic factor 0.8 CICIND
z/h
smaller than the background contribution B. The size factor cs 0.5
is 1, it decreases as the loaded areas become larger. Their 0.4
0.5
0.5
effect, is not reflected in any of the codes. 0.4
Figure 1 shows the aerodynamic force coefficient 0.3
distribution along the height applied to the chimney sample 0.2
n.1, as derived by wind tunnel tests on a circular cylinder with 0.1
a free-end (Lupi, [7]). It is compared to the distribution 0
0 100000 200000 300000 400000 500000 600000
obtained by applying the CICIND Model Code and the Wind-induced bending moments [kNm]
Eurocode model to both a smooth concrete chimney surface
(ks = 0.2 mm) and to a rough concrete chimney surface (ks =
Figure 3. Bending moments on the sample chimney n.1 due
1.0 mm).
different loading models
Figure 2 shows the differences in the in-line wind loading
profile due to different gust wind load models (Eurocode and Figure 3 shows the resulting bending moment on the sample
CICIND load models, i.e. equations (9) and (11), chimney n.1. Differences in the results are both due to
respectively). The aforementioned superimposition in different loading models and to the different distribution of
equation (11) of the mean load and the triangular gust load aerodynamic coefficients, according to Figure 1. In fact, the
distribution is evident in the CICIND result. The label "EN- accordance between the different models is good, despite the
WT" stands for the Eurocode approach (9) applied to the non- aforementioned differences in the gust load profiles. The
uniform distribution of CD resulting from wind tunnel tests. differences in the response are predominantly due to the
1387
Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014
different values and distributions of the aerodynamic beam distribution, may even exceed 30%. The departure is
coefficients. especially high at the bottom, but it extends over a relatively
On the basis of these and other results (see [8]), it can be high portion of the cylinder height (Figure 5).
concluded that, regarding the along wind action, the CICIND
model code results in a simplified but conservative design. 3 DESIGN FOR VORTEX RESONANCE
Even though the tip-effect is not included in the CICIND Two different models are available regarding the design for
model, this is compensated by the relatively high and constant vortex resonance. One is Ruscheweyh's approach to deal with
force coefficient. However, about 10% of reduction of stresses vortex resonance as a forced vibration and to incorporate the
might be achieved through a more realistic distribution of aero-elastic effects in model parameters derived from
aerodynamic coefficient, e.g. resulting from wind tunnel tests. experimental data. In the second model, the design for vortex
2.2 Structural behaviour: beam versus shell resonance is based on Vickery & Basu's model of a self-
limiting response process.
In the chimney design, the wind loading may be represented Ruscheweyh's approach is based on the concept of effective
by a wind force or by wind pressures non-uniformly correlation length and accounts for aeroelastic forces by an
distributed over the surface. Which way is appropriate iterative procedure, as the effective correlation length depends
depends on the slenderness of the chimney and the on the amplitude of vibrations. The model parameters are
deformability of the circular cross-section. High slenderness derived from experimental data, therefore the model is
and high ring stiffness of the wall imply a beam-like suitable for typical ranges of turbulence intensity.
behaviour with cosine stress distribution along the Furthermore, as the load depends on the mode shape, the
circumference; then, wind forces are applicable. With thin method is also applicable to non-cantilevered structures.
wall and low aspect ratio the structure behaves as a shell. Ruscheweyh's approach is included in the Eurocode as
Then, the stresses due to wind concentrate at the windward Method 1.
side and pressure distributions are needed for the design The typical feature of the Vickery&Basu model (which is
calculations. included in its full and even extended formulation in the
CICIND and in its simplified version in the Eurocode as
2000 Method 2) is to include the aeroelastic effects into the
modelling of a negative aerodynamic damping. The approach
1500
goes back to Scruton [2], who expressed the aeroelastic force
1000 beam z=0.00
in the most general way as a force with components in-phase
shell z=0.00 and out-of-phase with the motion. The in-phase component is
n22 [kN/m]
0.50
0.40
beam independent model parameters, namely the negative
shell
0.30
aerodynamic damping for small amplitudes of oscillations Ka0
0.20
and the self-limiting amplitude on the rms of the response aL
0.10
(Vickery&Basu, [10], [11]). The negative aerodynamic
0.00 damping shall be able to account for the aeroelastic effects
0 500 1000 1500 2000 due to the fluid-structure interaction by reducing the total
n22 [kN/m]
damping of the structure (see equation (12)). Then, the
spectral formulation of the lift force regards only the force on
Figure 5. Meridional membrane stress n22 compared to beam
the non-vibrating cylinder and allows calculating the variance
distribution on the chimney sample n.1
of the response in resonant conditions.
& d Sc :
Figure 4 shows that in case of low aspect ratio and/or low the
= 1 K &7 81 9 < =>
(12)
thickness-to-radius (the chimney n.1 is taken as example), the
deviation in the maximum membrane stress, compared to the m, 4 a;d
1388
Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014
An important parameter, which influences the size of the threshold is achieved at smaller Sc in case of high Iv. In the
limiting amplitude of the response, is the mode shape. shown example, the Scruton number is varied by varying the
According to the Vickery&Basu approach, the limiting value structural damping of the chimney.
is calibrated for the first mode of the cantilever structure. The Similarly, Figure 7 shows that, at a given level of
CICIND Model Code for steel chimneys ([3], [5]) proposes a mechanical damping (thus Scruton number), the effect of Iv
reduction of such a value for higher modes of vibrations, may considerably reduce the response. The calculation is still
which depends only on the natural frequency of vibration and performed by applying the extended Vickery&Basu
not on the shape of the mode. This issue, however, is not formulation as described in the CICIND code. Within this
further investigated in this paper. formulation, the effect of turbulence is mainly attributed to the
The Scruton number (Sc = 4ms/(D2)) governs the large scales, which are interpreted as a slowly varying mean
amplitude of the response. The latter may lie in the lock-in wind velocity and therefore have a pronounced effect on
range (high level of amplitude), in the transitional range or in vortex-induced vibrations [12]. Besides that, high turbulence
the forced-vibration range (low level of amplitude). In intensity also increases the bandwidth of the lift spectrum.
particular, with regard to the Vickery&Basu model, the
amplitude of oscillation is dictated by the ratio of the Scruton 0.60
number Sc to the aerodynamic damping 4Ka. The response
0.50
lies in the lock-in range when Sc << 4Ka. Physically, it
means that the mechanical damping is much lower than the
0.40
aerodynamic damping. In any case, the total damping of the
ymax(h)/d
structure will never become negative due to the existence, in 0.30
0
0.08
the modeling, of the self-limiting cycle (see equation (12)).
0.13
Small amplitudes of vibrations are instead predicted when Sc 0.20
0.24
>> 4Ka.
0.10
The core of the Vickery&Basu model is then the modeling
of the negative aerodynamic damping. Little information is
0.00
available on that and the aerodynamic damping at least 0 10 20 30 40 50
depends on the Reynolds number, the turbulence intensity, the Scruton number
in: 9.525
fC d
0.30
d fC d
12.700
V@+ d
(13)
Re = = St = > 5 10H
15.875
0.20
St
19.050
25.400
0.10
The kinematic viscosity of air is equal to 15*10-6 m2/s, 44.451
therefore: 0.00
fC d
15
0 0.05 0.1 0.15 0.2 0.25
>
(14) Iv(h)
St 2
Figure 7. Cross-wind deflections as a function of turbulent
if it assumed St = 0.2, it results:
intensity Iv at different Scruton numbers Sc
fC d > 1.5 (15)
Figure 8 shows the extension of the lock-in range at a given
Figure 6 shows the response curves as function of the Sc (for the chimney sample n.2) and its dependency on the
Scruton number at different levels of turbulence intensity. The turbulence intensity. In the lock-in range, the Strouhal law is
calculation is performed on the sample chimney n.2 by violated and vortex separation is tuned to the structural
applying the Vickery&Basu model in its full (an extended) vibration frequency for a certain interval of wind velocities.
formulation, as described in the CICIND Commentaries [5]. Figure 8 is obtained by applying the CICIND model to the
According to this formulation, the effect of turbulence chimney sample n.2. This model is unique in the sense it
intensity on the response is to reduce the critical Scruton includes the effect of turbulence and the dependency of the
number at which the jump between high and low levels of response on V/Vcr. The Eurocode design methods (Method 1
oscillation occurs. The reason is that the higher is the and Method 2), applied to the same chimney sample, provide
turbulence intensity, the lower is the negative aerodynamic very different results (Figure 9). However, the figure proves
damping, the higher is the ratio Sc/4Ka. Mathematically, that both methods turn to be correct if properly interpreted by
large amplitudes of vibrations occur when Sc/4Ka < 1. This including the effect of turbulence intensity. In fact, as the
Eurocode Method 1 (after Ruscheweyh) is empirical and
1389
Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014
calibrated on the basis of experimental data, it should be and define an equivalent static load which reproduces the
reliable at higher values of turbulence intensity. The opposite effect of critical design stresses (e.g., in the case of a chimney,
holds for the Eurocode Method 2 (a simplified version of the the base bending moment). The design load models (such as
Vickery&Basu model), which only provides results for zero those of the Eurocode and the CICIND) are revisited versions
turbulence intensity and for this reason the response is often of the Davenport approach. They are applied to a represented
overestimated. The figure proves that the design load models real world sample chimney. The effect of non-uniform
for vortex resonance are in principle consistent and physically spanwise distribution of aerodynamic coefficient is also
correct. However, at low values of Sc, as well as in the range investigated.
Sc 4Ka, the model parameters as well as the influence of Gust wind load models to calculate the along wind response
turbulence intensity should be carefully investigated, as the are applicable to line-like structures. In case of a line-like
two methods differ considerably even at relatively high values structure, force coefficients and not pressure coefficients, are
of turbulence intensity. Figure 7 shows the impressively high required. However, depending on the joint combination of
amplitudes of vibrations at small values of Sc. aspect ratio and ratio of wall thickness-to-radius, ovalization
Furthermore, the scale of turbulence would be another of the cross section might become significant. As a result, the
parameter to be investigated, in addition to the intensity of beam-like model might be neither realistic nor on the safe side
turbulence. The scale of turbulence alone is not, however, a and tensile stresses at the windward side become the leading
representative parameter, as it must be related to the diameter stresses. Depending on the aspect ratio, they even excess of
of the chimney. In general, whereas the effect of large 30% the corresponding stresses calculated by using a beam
turbulence scales is usually interpreted as a slowly varying model. When a shell model of the chimney has to be used,
wind speed, the small scales have an influence especially on wind pressures and not forces have to be provided.
the coherence along the chimney axis, because they destroy The second issue which is addressed by the paper concerns
the vortex cells. the design for vortex resonance. The two approaches, after
Ruscheweyh and after Vickery&Basu - as well their
Sc = 15.88 application in the Standard Codes - are considered. The core
0.60 of the Vickery&Basu-based model codes is the modelling of
the negative aerodynamic damping, which includes aeroelastic
0.50
effects. Little information is available on that and the
0.40 0.00
aerodynamic damping depends at least on the Reynolds
number, the turbulence intensity, the mean wind velocity and
y(h)/d
0.05
0.30
0.10 the aspect ratio. Among Standard Codes, the CICIND
0.20 0.15 modelling is unique in the sense that it includes the effect of
0.20 turbulence. High turbulence intensity reduces the negative
0.10 0.25
aerodynamic damping and increases the spectral bandwidth of
0.00 the lift force. The comparison of results between CICIND and
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 Eurocode reveals that the full formulation in the CICIND
V/Vcr
model is, in principle, theoretically correct and consistent.
However, design models for vortex resonance, especially at
Figure 8. Extension of the lock-in range as a function of Iv relatively low Scruton numbers, often provide very different
Sc = 15.88
results. Therefore, further research would be advisable in the
0.60 investigation of the model independent variables (like self-
EN - Method 2 limiting amplitude of vibration and aerodynamic damping)
0.50 EN - Method 1 and of the parameters they depend on (like the mode shape,
0.40
CICIND the intensity and scale of turbulence, the latter related to the
cylinder diameter).
ymax(h)/d
0.30
REFERENCES
0.20
[1] A.G. Davenport, The response of slender line-like structures to a gusty
0.10 wind. Proc. Soc. Civ. Engineers, vol. 23, pp. 389, 1962.
[2] C. Scruton, On the wind-excited oscillations of stacks, towers and
0.00 masts. Proc. of Symposium no.6. Wind effects on Building and
0 0.05 0.1 0.15 0.2 0.25 Structures, Teddington, UK, 26-28 June 1963.
Iv [3] CICIND Model Code for Steel Chimneys, The CICIND Chimney
Standard, 2010.
Figure 9. Comparison of model codes for vortex resonance [4] CICIND Model Code for Concrete Chimneys, Part A: The Shell, 2011.
[5] CICIND Commentaries for Steel Chimney Code, 2011.
[6] CICIND Commentaries for Model Code for Concrete Chimnes, 2011.
4 CONCLUSIONS AND FUTURE OUTLOOKS [7] F. Lupi, Eds., A new aerodynamic phenomenon and its effects on the
The paper investigates the modeling of vibrations of design of ultra-high cylindrical towers. Phd Dissertation, University of
Florence / Technical University Braunschweig, Shaker Verlag.
chimneys under the wind action. [8] H.J. Niemann, F. Lupi, International Standardizazion of Wind Actions
All design load models for the gust wind load are based on on Chimeys, CICIND Report, 2013.
the Davenport approach of the gust response factor. The latter [9] A.W. Marris, A review of vortex streets, periodic wakes and induced
is a structural parameter, able to amplify the mean wind load vibration phenomena. Journal Basic Eng., Trans. Am. Soc. Mech. Eng.,
86, 185-193, 1964.
1390
Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014
1391