You are on page 1of 12

Structural Safety 33 (2011) 367–378

Contents lists available at ScienceDirect

Structural Safety
journal homepage: www.elsevier.com/locate/strusafe

Performance-Based Wind Engineering: Towards a general procedure


M. Ciampoli ⇑, F. Petrini, G. Augusti
Department of Structural and Geotechnical Engineering, Sapienza Università di Roma, Italy

a r t i c l e i n f o a b s t r a c t

Article history: It is widely recognized that the most rational way of assessing and reducing the risks of engineered facil-
Received 8 September 2009 ities and infrastructures subject to natural and man-made phenomena, both in the design of new facilities
Received in revised form 30 June 2011 and in the rehabilitation or retrofitting of existing ones, is Performance-Based Design, usually indicated
Accepted 5 July 2011
by the acronym PBD (but a better term would be ‘‘Performance-Based Engineering’’). The first formal
Available online 17 August 2011
applications of PBD were devoted to seismic engineering and design; later it has been extended to other
engineering fields, like Blast Engineering and Fire Engineering.
Keywords:
Wind Engineering has appeared of great potential interest for further developments of PBD. The expres-
Eolian risk
Probabilistic risk assessment
sion ‘‘Performance-Based Wind Engineering’’ (PBWE) was introduced for the first time in 2004 by an Ital-
Performance-Based Design ian research project. In this paper, the approach proposed by the Pacific Earthquake Engineering Research
Performance-Based Wind Engineering Center (PEER) for Performance-Based Earthquake Engineering is extended to the case of PBWE.
Suspension bridge The general framework of the approach is illustrated and applied to an example case: the assessment of
Out-of-service risk the collapse and out-of-service risks of a long span suspension bridge.
Flutter instability A discussion of the open problems and the relevance of various sources of uncertainty conclude the
paper.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction predict the wind damage to low and medium rise buildings subject
to hurricanes; Rosowski and Ellingwood [18], Ellingwood et al. [19]
A modern approach to structural design should not be limited and van de Lindt and Dao [20] focused on wood structure perfor-
to optimize safety vs. costs, but consider as key objectives the mances under synoptic winds or hurricanes; Khanduri and Morrow
whole range of ‘‘performances’’ in a probabilistic context and [21] studied the vulnerability of buildings to windstorms; Garciano
throughout the whole life-cycle of the structure: this has led to et al. [22] developed a procedure for the assessment of the critical
the recent significant development of ‘‘Performance-Based Design’’ performances for wind turbines subject to typhoons; Bashor et al.
(PBD) procedures (for recent reviews, see e.g. [1] and [2]). [23], Bashor and Kareem [24], Tamura [25] and Kim et al. [26]
The first formal applications of PBD were devoted to seismic focused on comfort of tall building occupants.
engineering and design [3]. More recent, but rapidly expanding, Notwithstanding this large effort in probabilistic assessment of
are extensions of PBD to other problems, and in particular to Wind structural performances, there is not yet consensus on a general
Engineering (the definition ‘‘Performance Based Wind Engineer- Performance-Based approach for the assessment of risk under wind
ing’’, PBWE, appeared for the first time in 2004 [4]). action (‘‘Eolian risk’’). The goal of this paper is a self-contained
The approach is not new. During the last decades, the problem presentation of the probabilistic procedure for the application of
of the assessment of structural safety under wind action has been Performance-Based Design concepts to Wind Engineering that the
studied extensively: advanced probabilistic approaches have been authors have outlined in previous paper [1,27–30]. The procedure
largely adopted in Wind Engineering due the aleatory nature of is then applied to an example case: the assessment of both the fail-
winds (e.g. [5–7]). Great efforts have been devoted to the probabi- ure probability due to flutter instability and the out-of-service risk
listic characterization of both wind actions (e.g. [8–12]) and of a bridge design derived from a preliminary design of the suspen-
response of structures with uncertain characteristics under wind sion bridge over the Messina Strait (Italy) [31]. A discussion of the
(e.g. [13–16]). open problems concludes the paper.
Also the performance assessment of structures subject to wind
actions has been the object of many studies in the past decade.
2. Sources of uncertainty in Wind Engineering
Among the others: Unanwa et al. [17] developed a procedure to
As previously stated PBWE must be tackled in probabilistic
⇑ Corresponding author. Tel.: +39 06 44 585 300;, fax: +39 06 48 84 852. terms, due to the stochastic nature of both resistance and loading
E-mail address: marcello.ciampoli@uniroma1.it (M. Ciampoli). parameters. First, in characterizing the wind field and the corre-

0167-4730/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.strusafe.2011.07.001
368 M. Ciampoli et al. / Structural Safety 33 (2011) 367–378

sponding actions, different sources of uncertainty are to be taken features of the wind effects. Also non environmental actions can
into account: they must be considered differently according to influence the structural response by modifying the aerodynamic
their relation to the structure (Fig. 1). and aeroelastic characteristics of the structure; a relevant example
By definition, in the environment the wind field is considered as is represented by the transit of trains on a railway bridge, as it
if the structure were absent (free-field wind). In the environment determines a change of the dynamic characteristics of a flexible
the basic site-dependent parameters of the wind field (mean value superstructure.
of the velocity in each direction, turbulence intensity, dominant As concern the characterization of the wind field in the environ-
direction of strong winds, etc.) are not influenced by the presence ment, the uncertainty is due to: (i) the intrinsic variability of the
of the structure, while they can be affected by the interaction with basic parameters (the inherent or aleatory uncertainty), arising from
other environmental agents; a typical example is the interaction the unpredictable nature of magnitude and direction of the wind
between wind and waves in offshore sites. velocity and turbulence intensity; (ii) the errors associated to the
The exchange zone is the region around the structure where the experimental measures and the incompleteness of data and infor-
structure and the wind field are strongly correlated, and the effects mation (the epistemic uncertainty); (iii) the modeling of wind ac-
of the interaction between the relevant properties of both the tions and their effects on structural response (the model
structure and the wind field, as well as the presence of nearby uncertainty, often included in the epistemic uncertainty).
structures, cannot be disregarded. Hence aerodynamic and aero- In the exchange zone, the interaction parameters are strongly
elastic phenomena are essential in determining the relevant dependent on the basic parameters that characterize the wind field

ENVIRONMENT EXCHANGE ZONE STRUCTURAL SYSTEM

Aerodynamic and
aeroelastic
phenomena
Wind site basic Structural
parameters systems
Wind action Structural
Site-specific system as
Wind modified by
service loads
Environmental Non
effects (like environmental
waves) actions

Types of uncertainties
1. Aleatory 1. Aleatory 1. Aleatory
2. Epistemic 2. Epistemic 2. Epistemic
3. Model 3. Model 3. Model

Basic parameters ( IM ) Derived parameters ( IP) Independent parameters (SP )

2 Exchange zone

Stru
ctu
re
1 Environment

Vm+ v(t)

Turbulent wind velocity profile


r
rive

Vm

Mean wind velocity profile

Fig. 1. Sources of uncertainty in Wind Engineering.


M. Ciampoli et al. / Structural Safety 33 (2011) 367–378 369

and the structural behavior. Examples of derived parameters in to the considered construction (no collapse, occupant safety, acces-
case of suspension bridges are the aerodynamic coefficients sibility, full functionality, limited displacements or accelerations,
(whose plots are sometimes referred to as ‘‘polar lines’’), the aero- etc.) and different ‘‘intensities’’ or mean return periods of the wind
dynamic derivatives (or ‘‘aeroelastic derivatives’’), the Strouhal action. With reference to a specific performance, usually the struc-
number (see [32] Chapters 4 and 6). In order to derive the proba- tural risk is conventionally measured by the probability of exceed-
bility density functions of these parameters, the uncertainty of ing a relevant value of the corresponding DV. This probability of
the basic parameters must be taken into account, while the uncer- exceedance is expressed in terms of a mean annual frequency or
tainty of the relevant parameters in the exchange zone has no a complementary cumulative distribution function (CCDF), that
influence on the basic parameters. are evaluated by taking into account the Eolian hazard (i.e. the fre-
The parameters that characterize the structural behavior, like quency of occurrence of wind actions of specified intensity and
the mechanical and material properties of the structure, are inde- characteristics at the site), the calculated structural response and
pendent of the basic parameters of the wind field, and often their damages, and the correlation between the attained damage level
inherent uncertainty can be considered negligible. In what follows, and the relevant DV. The structural design should be optimized
the uncertain basic parameters in the environment are grouped in by applying a decisional strategy to the risk analyses, with the
the intensity measure vector IM; the uncertain parameters of inter- objective of minimizing the total risk or of maximizing a utility
est in the exchange zone are grouped in the two vectors of derived function. Thus, in general terms, a procedure of PBWE should con-
interaction parameters IP and independent structural parameters sist of the following steps.
SP (Fig. 1). The vectors IM and SP can be assumed not correlated First, the Eolian hazard at the site, in terms of wind intensity
with each other and not affected by the uncertainty of the param- and/or parameters of the wind velocity field shall be defined. The
eters IP, i.e. assessment of the Eolian hazard requires the use of efficient tech-
niques for modeling wind actions on a slender structure, and the
PðIMjIPÞ ¼ PðIMjSPÞ ¼ PðIMÞ ð1Þ
choice of the intensity measure vector IM whose stochastic charac-
teristics are able to describe sufficiently and efficiently the Eolian
PðSPjIPÞ ¼ PðSPjIMÞ ¼ PðSPÞ ð2Þ
hazard at the site. Hence IM must be chosen by considering the
where P(|) is a conditional probability. characteristics of the wind turbulence, direction and arrival pro-
In PBWE procedures, conditional probabilities, conditional den- cesses, the relevant features of the construction site, the structural
sity functions and the total probability theorem are essential. From properties and the interaction phenomena (wind turbulence, vor-
elementary theory of probability, given Eqs. (1), (2), the joint prob- tex shedding, aeroelastic phenomena, aerodynamic effects). How-
ability of IM, IP and SP is: ever, the choice must be done by paying attention to minimizing
the needed information, since today the available data (essentially
PðIM; IP; SPÞ ¼ PðIPjIM; SPÞ  PðIMjSPÞ  PðSPÞ turbulence spectra and maps of expected wind speeds) are still
¼ PðIPjIM; SPÞ  PðIMÞ  PðSPÞ ð3Þ rather limited. Then, the interaction phenomena shall be modeled
in probabilistic terms. This implies the choice and probabilistic
and the complementary cumulative distribution function (CCDF) of characterization of the interaction parameter vector IP, that allow
a continuous random variable X dependent on the basic, derived to take into account the relevant aspects of the interaction be-
and independent parameters is: tween the environment and the structure in the exchange zone.
GðxÞ ¼ PðX > xÞ The following step is the analysis of the structural response,
Z 1 mainly in the context of stochastic dynamics. The probabilistic
¼ GðxjIM; IP; SPÞ  f ðIM; IP; SPÞ dIM dIP dSP modeling of the structural response requires the choice of the rel-
1 evant engineering demand parameter vector EDP (inter-storey
Z 1
¼ GðxjIM; IP; SPÞ  f ðIP jIM; SPÞ  f ðIMÞ drifts, accelerations and velocities of selected points, stresses and
1 displacements, etc.). The evaluation of the structural damage (in-
 f ðSPÞ dIM dIP dSP ð4Þ tended as unacceptable performance) requires the choice (and
probabilistic characterization) of a damage parameter vector DM,
where f() is a probability density function (PDF) and f(|) a condi- that is able to quantify the structural damage due to wind actions
tional probability density function (CPDF). in relation to the considered performances. The choices of EDP and
DM are strongly dependent on the considered structural type and
3. A procedure for Performance-Based Wind Engineering performances. Different parameters can be assumed as DM: they
(PBWE) can be defined by one or a combination of relevant EDPs (e.g. the
inter-storey drift), or by other parameters, representing e.g. the
The procedure for PBWE presented in this paper has been ob- damage to the partition or façade walls in a building as a function
tained by extending the approach proposed for Performance-Based of the inter-storey drift. However, in the latter case it is usually
Seismic Design (PBSD) by the researchers of the Pacific Earthquake rather difficult to establish sound correlations between the evalu-
Engineering Research Center (PEER) (see e.g. [3]). The central ated EDPs and the chosen DMs.
objective of any procedure for Performance-Based Design is the The following step is the definition of the decision variables DVs
assessment of the adequacy of the structure through the probabi- that are appropriate to quantify the performances required for the
listic description of a set of decision variables DVs. Each DV is a structure, in terms of consequences of damage (personal damages,
(quantitative) measure of a specific structural performance that restoration costs, costs due to loss or deterioration of service, alter-
can be defined in terms of interest of the stakeholder or the society ations of users comfort, etc.). The decision variables DVs must dis-
in general. In Wind Engineering, relevant examples of DVs are the tinguish between low and high performances or performance
number of lives lost during windstorms, the economic losses levels: the former (low performances) imply possible consequences
resulting from windstorms, the exceeding of a (collapse or service- on structural and personal safety (e.g. partial or total collapse and
ability) limit state, the discomfort of the occupants, the length of permanent damages); the latter (high performances) affect only ser-
the out-of-service time due to a windstorm, etc. viceability and comfort (e.g. small displacements, limited vibra-
The starting points of the procedure are the relationships, ex- tions, and wind discomfort also in the area around the
pressed in probabilistic terms, between the performances specific construction) [33,34]. For low performances, the significant DV
370 M. Ciampoli et al. / Structural Safety 33 (2011) 367–378

can be identified with the cost necessary to restore the construc- environment (in terms of IM) and of the structural system (in
tion to the undamaged state (or rebuild it in case of collapse); cor- terms of SP) are made separately, the former depending on the can-
respondingly, DM is the set of damages to be restored, and EDP didate location of the structure, and the latter depending on the
includes the most significant response parameter for the specific candidate design configuration.
case (peak displacement or acceleration at the building top, overall Eq. (5) can be simplified by assuming that the chosen EDP is a
action at the base, local pressure, etc.). The shortcoming of this def- measure of the structural damage (that is, EDP  DM), and express-
inition is the impossibility of including in it the ‘‘intangible’’ losses, ing the performance by a Limit State (LS). The risk assessment is thus
i.e. the losses that cannot be measured in ‘‘monetary’’ terms. High based on the evaluation of the probability of exceedance, given by
performances are related to the users’ comfort/discomfort and, in Z Z Z Z
case of buildings, to inconvenient alterations of the wind field in GðLSÞ ¼ GðLSjEDPÞ  f ðEDPjIM;IP; SPÞ  f ðIP jIM; SPÞ
pedestrian areas around the construction. Appropriate discomfort
criteria, that is, statements specifying the maximum acceptable  f ðIMÞ  f ðSPÞ  dEDP  dIM  dIP  dSP ð6Þ
frequencies of occurrence for various degrees of discomfort can If the limit state is quantified in terms of EDP, the whole procedure
be defined [35]. Using the ‘‘limit states’’ approach (i.e. if the struc- simply requires the evaluation of the probability of exceedance:
tural risk is quantified by the probability of exceeding a limit state), Z 1
ultimate limit states (ULS) are related to low performances (exam- GðEDPÞ ¼ GðEDP jIM;IP; SPÞ  f ðIPjIM; SP Þ  f ðIMÞ  f ðSPÞ
ples are the attainment of the capacity of any significant part of 1

the structure, the fatigue collapse of some elements, the instability  dIM  dIP  dSP ð7Þ
of parts or of the whole structure, etc.), while serviceability limit
states (SLS) are related to high performances (examples are exces- There are several methods for computing the integrals (5–7). In the
sive deformations or vibrations compromising the use of the struc- numerical example below (Section 4), a crude Monte Carlo simula-
ture or its function in service). tion will be used.
The structural risk is then evaluated on the basis of the proba- The final step of the procedure is the optimization of design,
bilistic characterization of the decision variables. Appropriate rela- that is the minimization of risk, by appropriate techniques of deci-
tionships between any DM and the relevant EDP allow evaluating sion analysis. The optimization of design in PBWE would require
the damage states corresponding to given values of the response the development and implementation of a decisional strategy
parameter EDP, and also the resulting losses, taking into consider- aimed at minimizing the total risk or at maximizing an appropri-
ation the relationships between DM and DV. According to the usual ately defined utility. In the high performance case, the relevant
definition of risk as the convolution of hazard, vulnerability and DVs could be identified with the economical losses due to wind-
exposure [36], the relationships between DM and DV take into ac- storms, evaluated taking into account the whole lifetime of the
count the exposure that reflects the consequences of damage. In construction; in the low performance case, casualties and losses
the PBWE procedure adopted in this paper (as in the PEER’s ap- of lives would be involved. The ethical and practical difficulties
proach), the structural risk is defined as the probability of exceed- of taking both aspects into account add to the overwhelming ana-
ing a threshold level of a relevant DV: lytical difficulties and to the lack of reliable data: as suggested in
Z Z Z Z Z [39], it is realistic to leave the intangible and monetary aspects
separate, and open the final decision. These aspects are currently
GðDVÞ ¼ GðDV jDMÞ  f ðDM jEDP Þ
being developed [40]. Moreover, it has to be noted however that
 f ðEDPjIM;IP; SPÞ  f ðIP jIM; SPÞ  f ðIMÞ  f ðSPÞ  dDM the ‘‘complete’’ decisional process is seldom fully pursued in PBSD,
and likewise in PBWE: actual procedures, like those discussed and
 dEDP  dIM  dIP  dSP ð5Þ
presented in this paper, omit the last step, i.e. are limited to eval-
where, as already stated: DM is a damage measure; EDP is the cor- uating the risk, and at most to comparing the risk associated to dif-
responding engineering demand parameter, representing the struc- ferent design alternatives.
tural response; the basic parameters characterizing the Eolian 4. Model of the wind field
hazard (Section 2) are described by a vector IM of intensity mea-
sures; the vector of the interaction parameters IP contains the set The modeling of the wind field assumed in this paper is summa-
of aerodynamic and aeroelastic parameters that allow to take into rized in this section: it is described in detail in several texts, e.g.
account the relevant aspect of the interaction between the environ- [32,41]. A discussion of the influence of the uncertainties in the
ment and the structure; SP is the vector of the parameters charac- considered model and the effect of alternative models is reported
terizing the structural systems and non environmental actions. in Section 6. Indicating by x, y, z respectively the horizontal axis
By means of Eq. (5), the problem of risk assessment is disaggre- orthogonal to the structure (in the example case, a suspension
gated into the following elements: site and structure-specific haz- bridge), the horizontal axis parallel to the structure and the vertical
ard analyses, that is, the assessment of the probability density axis, the three components of the wind velocity field Vx(j), Vy(j),
functions f(IM), f(SP) and f(IP|IM, SP); structural analysis, aimed at Vz(j) in each point j(x, y, z) of the structure (the indication of the
assessing the probability density function of the structural re- dependence on time is omitted for simplicity of notation) can be
sponse f(EDP|IM, IP, SP) conditional on the parameters characteriz- expressed as the sum of a mean value Vm(j), that is time-invariant
ing the wind field and the structural properties; damage analysis, and depends only on the height z above terrain of point j, and tur-
that gives the damage probability density function f(DM|EDP) con- bulent components u(j), v(j), w(j) with zero mean. It is assumed
ditional on EDP; finally, loss analysis, that is the assessment of that Vm(j) = Vm(z) is parallel to the x axis and its value is deter-
G(DV|DM). mined from a database of values recorded at or near the site, and
Under these assumptions, a flowchart similar to the PEER flow- evaluated as the record average over a proper time interval (e.g.
chart can be defined for the proposed PBWE procedure (Fig. 2). 10 min) at a proper height (e.g. 10 m); the probability distribution
With respect to the PEER approach, the step of probabilistic char- function of Vm(z) is assumed as [42]:
acterization of the interaction parameters IP has been introduced.
 " k #
This step requires the assessment of the probabilistic correlation 1 V m ðzÞ
between IP and IM, and can be based on either wind tunnels tests PðV m ðzÞÞ ¼ 1  exp  ð8Þ
2 r
or CFD techniques [37,38]. Moreover, the characterizations of the
M. Ciampoli et al. / Structural Safety 33 (2011) 367–378 371

Hazard analysis

f (IM|O)
Environment Aerodynamic Loss analysis
Structural analysis Damage analysis
info analysis

f (IP|IM,SP) f (EDP|IM,IP,SP) f (DM|EDP) f (DV|DM )


O f (IM )
Decision-
making
IM: intensity measures

O: location Select
f (IP) G(EDP) G (DM ) G (DV )
O, D
D: design
SP: structural system
parameters IP: interaction EDP: engineering DM: damage measures DV: decision variables
parameters demand parameters
D f (SP)

Structural
system info

f (SP|D)

Structural
characterization

Fig. 2. Performance-Based Wind Engineering flowchart.

The variation of Vm(z) with the height z above a horizontal terrain of 5. Application of PBWE to the design of a long span suspension
homogeneous roughness is well described by the logarithmic law bridge
[11,42]:
  In order to check the validity of the proposed procedure for
u z
V m ðzÞ ¼  ln ð9Þ PBWE, a bridge design derived from the 2005 preliminary design
Ka z0
of the suspension bridge over the Messina Strait [31] has been con-
where Ka is the von Kármán’s constant, usually set equal to 0.41; z0 sidered as a case example (Fig. 3). To develop this example, a num-
is defined as the ‘‘roughness length’’; u is the friction or shear ber of simplified assumptions are introduced into the general
velocity (in m/s), given by: [(F)1/2  V10], where F is a coefficient procedure described so far. For details of the computations, see
dependent on the roughness length z0. [45]. The main span of the bridge is 3300 m; the total length,
The turbulent components of the wind velocity are modeled as including the side spans, is 3666 m. The cross section of the deck
Gaussian ergodic independent processes [11]. By considering the (61 m wide) is composed of three box elements, the external ones
wind acting on N points, and neglecting the component v, the tur- for the roadways and the central one for the railway. Each roadway
bulent components u and w are completely characterized by the is composed of three lanes 3.75 m wide (two driving lanes and one
power spectral density matrices [S]i (i = u, w). The diagonal terms emergency lane), and the railway has two tracks. The towers are
(auto-spectra) Sij ij ðnÞ of [S]i (j = 1, 2,. . ., N) can be expressed in 383 m high; the bridge suspension system is made by two steel
terms of normalized one-side power spectral density ([11,43]) as cables, each with a diameter of 1.24 m and a total length, between
nSuj uj ðnÞ 6:868  nu the anchor blocks, of approximately 4700 m, and by 121 pairs of
¼ ð10Þ rope hangers.
r2u ½1 þ 10:302 n2u ðzj Þ5=3
As indicated in Section 3, the structural performances are dis-
nSwj wj ðnÞ tinguished between high and low performances. In what follows,
6:103 nw
¼ ð11Þ the serviceability of the bridge under wind actions (high perfor-
r2w ½1 þ 9:1545 n2w ðzj Þ5=3 mances) is investigated by considering the behavior of the deck
where n is the current frequency (in Hz); z is the height above ter- in service, while, with regard to structural safety (low perfor-
rain (in m), r2u and r2w are the variances of the velocity fluctuations, mances), the flutter stability is considered. A more detailed descrip-
given by tion of the required performances for this complex structure can be
found in [46]. Adopting the Limit State format, hence the simplified
r2u ¼ ½6  1:1 arctanðlogðz0 Þ þ 1:75Þ  u2 ð12Þ form of PBWE [Eq. (7)], the first step of the procedure requires the
choice of appropriate EDPs (i.e. DMs), that are able to quantify the
rw considered performances. For high performances, relevant EDPs
¼ 0:5 ð13Þ
ru could be the deck rotational velocity, the deck translational and
ni(z) is a non-dimensional height-dependent frequency given by rotational accelerations, the deck twist that could generate a mis-
alignment of the rails, etc. In the example case, three EDPs have
nLi ðzÞ
ni ðzÞ ¼ ð14Þ been considered: the rotational velocity (Vrot) and the components
V m ðzÞ of the acceleration in the longitudinal Al and vertical Av directions,
In the example case, the integral scale Li(z) of the turbulent compo- all referred to the centroid (center of mass) of the cross section.
nents for i = u, w have been derived according to the procedure Two high performance levels (SLS-1 and SLS-2) have been identi-
given in [44]. fied, that correspond to full serviceability (SLS-1: both train and
372 M. Ciampoli et al. / Structural Safety 33 (2011) 367–378

+383 m +383 m
+63 m
+52 m
+118 m
+53 m +77 m

183 m 183 m
S C
777 m 3300 m 627 m

Main suspension cables (twin)

50.63 m
variable

Wind
2.78 m 12.00 m Hangers barriers

7.18 m
Service
2.69 m

lane
Railway grider Roadway grider
3.67 m

14.74 m 8.48 m

53.08 m

61.00 m

Fig. 3. Characteristics of the long-span suspension bridge considered as an example case (2005 preliminary design).

vehicle transits are allowed) and partial serviceability (SLS-2: only evaluated by aeroelastic theories [47,48]. With regard to low per-
train transit is allowed). The two levels are identified by different formance (ULS-FL) and the determination of the flutter velocity,
thresholds. In the low performance criterion (ULS-FL: avoidance the wind flow has been modeled as non turbulent (laminar).
of flutter instability), the motion must not diverge: the EDP is iden- The PBWE procedure requires, for each considered perfor-
tified with the total damping d of a relevant displacement time-his- mance, the evaluation of the annual probability of exceeding the
tory, i.e. the vertical displacement of the mid-span section, that threshold value of each response parameter [Eq. (7)]. In the numer-
must be positive. For each performance, failure is attained when ical example, the IP and SP parameters have been considered deter-
the peak value (with reference to the whole bridge deck) of the rel- ministic, and IM is reduced to one parameter, namely the mean
evant EDP overcomes the corresponding threshold (that defines the wind velocity evaluated at the height of 10 m, that is assumed as
‘‘failure criterion’’) during the considered windstorm. Relevant a random variable, with the Weibull CDF given by Eq. (8). In order
threshold values are summarized in Table 1. to describe the 10-min average wind velocity, the parameters r
The analyses have been carried out in time domain on a finite and k have been set equal to 6.02 m/s and 2.02 according to a data-
element model of the bridge; the total number of elements (beams, base available for the site of the bridge. The roughness length z0
no compression cable elements and gaps) is 1614, and the number has been set equal to 0.05 m. This value, that corresponds to a ter-
of nodes is 1140. A Newmark’s time integration scheme has been rain of second category according to Structural Eurocode EN 1991
adopted, to take into account nonlinear and second order effects. [49], is higher than the currently adopted value for over-water
The incident turbulent wind velocity time series have been gener- roughness length: however, it has been specifically indicated in
ated as components of a multivariate, multidimensional Gaussian the guidelines for the design of the bridge (see Section 6.1 for an
stationary stochastic process; the Weighted Amplitude Wave investigation of the effect of this value). Starting from the mean
Superposition method (WAWS) and a Proper Orthogonal Decom- wind velocity evaluated at the height of 10 meters, the mean wind
position (POD) of the PSD matrix [43] have been adopted. Starting velocity Vm(zdeck) at the height of the deck zdeck (equal at midspan
from the wind velocity time series, the wind actions have been to 77 m) is evaluated by Eq. (9).

Table 1
Considered performances and failure criteria.

Limit state Performance EDP Threshold


SLS SLS-1 Full serviceability Vrot Al Av max {Vrot} = 0.040 rad/s max {Al} = 2.5 m/s2 max {Av} = 0.7 m/s2
SLS-2 Partial serviceability max {Vrot} = 0.043 rad/s max {Al} = 2.5 m/s2 max {Av} = 1.0 m/s2
(only railway traffic is allowed)
ULS ULS-FL Preservation of structural integrity Total (structural + aerodynamic) min {d} = 0
damping d
M. Ciampoli et al. / Structural Safety 33 (2011) 367–378 373

Obviously, considering just one parameter as the only source of 5.2. Results for low performance (avoidance flutter instability)
uncertainty in characterizing the wind field represents a drastic
simplification: however, in the Authors’ opinion this choice does In evaluating the flutter condition, the aerodynamic coefficients
not affect the validity of the proposed procedure. The developed in Fig. 9a have been considered. By means of a time-domain anal-
numerical calculations shall be considered as a check of its feasibil- ysis based on the Quasi-Steady (QS) theory for the evaluation of the
ity; the role of different sources of uncertainty will be considered if aeroelastic forces [47,51], the time series of the vertical displace-
the real problem of the optimal design of the considered bridge ment at bridge midspan for several values of Vm(zdeck) have been
should be faced with. Moreover, in parallel with the development obtained [Fig. 10a]. Damping has been evaluated as the parameter
of the procedure, a huge effort is currently devoted to collect data d of the envelope function:
about the probabilistic characterization of the relevant parameters,
together with their correlation. In the example case, only the alea-
qðtÞ ¼ q  q0  edt ð17Þ
tory uncertainty has been considered; however in the following, of the time series. d represents the total damping of the structural
some remarks about the relevance of the epistemic and model system, which is the sum of the structural component (assumed
uncertainties will be illustrated. to be constant and equal to 0.5%) and the aerodynamic component.
In Fig. 10(b), the total damping d is shown as a function of Vm(zdeck).
The low performance criterion: d > 0 is satisfied for
5.1. Results for high performances (serviceability) Vm(zdeck) < 70 m/s. Hence the critical flutter velocity is:
Vcrit = 70 m/s. The probability of loss of the structural integrity as
The annual probability densities and distribution functions of a consequence of flutter instability is equal to the probability that
the three considered EDPs, evaluated by Monte Carlo simulation the mean wind velocity exceeds 70 m/s. Note that the design
(500 runs), are reported in Fig. 4. From Fig. 4c it is evident that guidelines for this bridge [31] indicate 57 m/s as the limit flutter
the longitudinal acceleration Al is always lower than the given wind velocity; therefore the corresponding performance criterion
threshold; hence the corresponding serviceability criterion is auto- would be satisfied in a deterministic framework, since
matically satisfied in the considered example. This would suggest Vcrit > 57 m/s. In Section 6, the role of the uncertainties of the mod-
that the criterion does not appear significant for the design: there- el assumed for evaluating Vcrit and of the aerodynamic coefficients
fore, it has been neglected in the successive elaborations. will be discussed.
In Figs. 5 and 6, the fragility curves of the bridge deck, evaluated
considering three different threshold values of the two relevant
6. Investigation on further sources of uncertainty
EDPs (the rotational velocity Vrot and the vertical acceleration Av)
are shown. By definition, the fragility functions P(EDP|IM) define
6.1. Uncertainty of the roughness length
the probability of violation of a threshold: those shown in Figs 5
and 6 have been assessed by following a procedure analogous to
In the analyses illustrated in Section 5.1, only the randomness
the procedure described in [50]. For each EDP, a performance indi-
of the mean value of the wind velocity has been considered. In Sec-
cator Yi has been introduced, defined as
tion 2, it has been recognized that also the uncertainty of the
roughness length z0 should be taken into account; however, very
EDPi little research has been so far performed to characterize the vari-
Yi ¼ ð15Þ
EDPi ability of z0. As a preliminary check of its relevance, Monte Carlo
simulations have been repeated assuming three different values
where EDPi is the threshold value corresponding to the relevant of the roughness length: z0 = 0.05 m (the value assumed in
performance criterion. Under these assumptions it is possible to Section 5.1); z0 = 0.10 m; z0 = 0.20 m. In Fig. 11, the three resulting
write the fragility function as follows: functions G(Vrot) are reported: the large effect of variations in the
roughness length on the risk of the bridge appears evident (the risk
PðEDPi P EDPi jIMÞ ¼ PðY i P 1jIMÞ ¼ PðIM P IMðY i ¼ 1ÞÞ ð16Þ decreases when z0 is increased).

where IM(Yi = 1) is the value of IM for which Yi = 1. IM(Yi = 1) is as- 6.2. Uncertainty of the aerodynamic coefficients
sumed as a stochastic variable described by a Lognormal distribu-
tion whose median and fractional standard deviation have been In Section 5.2, only the variation of the mean wind velocity Vm
computed by using the results of Monte Carlo simulation. has been considered in deriving the critical (flutter) velocity Vcrit,
In Fig. 7, the fragility curves evaluated for the threshold values and the uncertainty of the interaction parameters IP (in this case,
corresponding to each full serviceability criterion (Table 1, SLS-1) the aerodynamic coefficients) has been disregarded. The relevance
are shown (thin lines): it can be noted that for small wind veloci- of the uncertainty of the aerodynamic coefficients has been inves-
ties, the Av performance criterion (denoted by +) prevails, while for tigated by a parametric analysis, adopting a linear combination of
large velocities the Vrot criterion (denoted by ) prevails. The ser- the two sets of aerodynamic coefficients A and B (Fig. 9), that were
viceability of the bridge should be evaluated with respect to viola- derived in experimental tests on models at different scales [29,52]:
tion of either performance criterion: the corresponding fragility cL ðaÞ ¼ cL B ðaÞ  PL þ cL A ðaÞ  ð1  PL Þ
curve, derived directly from results of Monte Carlo simulation, is ð18Þ
cM ðaÞ ¼ cM B ðaÞ  PM þ cM A ðaÞ  ð1  PM Þ
plotted as a bold line in the same Fig. 7. In Figs 8 (a), (b), (c), the
complementary cumulative probability distribution functions where: ci_j (i = L, M and j = A, B) is the aerodynamic coefficient i, cor-
G(EDP), evaluated by Eq. (7) for each EDP, are reported. From these responding to the set j; a is the generic angle of attack; PL and PM are
functions, the exceedance probabilities corresponding to the combination parameters, which vary between 0 and 1. The resulting
threshold values reported in Table 1 (i.e. the probabilities of not flutter velocities are shown in Fig. 12; the points marked by a cross
satisfying the corresponding performances) are derived. These val- have been evaluated by the FE analysis, the others have been de-
ues are reported in Table 2, together with the values of the proba- rived by linear interpolation. The response surface in Fig. 12 shows
bility of exceeding any performance criterion for the same limit that the uncertainty of the interaction parameters should not be
state, evaluated for both SLS-1 and SLS-2. disregarded; in particular, the influence of the uncertainty of the
374 M. Ciampoli et al. / Structural Safety 33 (2011) 367–378

70
(a) 0.12 (b)
Vrot [rad/s] SLS-1
SLS-1 SLS-2
SLS- 2
0.10 100%
0.1

0.08
0.08

0.06

0.04

0.02
Vm (zdeck ) [m/s]
0
0.00 0 0%

Other
0.000
0.008
0.017
0.025
0.034
0.000
0.008
0.017
0.025
0.034
0.042
0.051
0.059
0.067
0.076
0.084
0.093
0.101
Other
0 5 10 15 20 25 30 35

Other
0.000
0.008
0.017
0.025
0.034
0.042
0.051
0.059
0.067
0.076
0.084
0.093
0.101
210

(c) 3.00 (d)


Al [m/s2] 100%
100%
2.50
SLS-1 SLS-2

2.00

1.50

1.00

0.50
Vm (zdeck ) [m/s]
0.00 0 0%

Other
0.000
0.036
0.071
0.107
0.134
0.178
0.214
0.250
0.285
0.321
0.357
0.392
0.428
0.000
0.036
0.071
0.107
0.143
0.178
0.214
0.250
0.285
0.321
0.357
0.392
0.428
Other
0 5 10 15 20 25 30 35

70
4.00
(e) Av [m/s2] SLS-1 SLS-2
(f)
3.50 100%
100%
3.00

2.50

2.00

1.50

1.00

0.50
Vm (z deck ) [m/s]
0.00 0 0%
0%
0 5 10 15 20 25 30 35
Other
0.000
0.205
0.410
0.614
0.819
1.024
1.228
1.433
1.637
1.842
2.047
2.251
2.456
0.614
0.819

2.251

Other
0.000
0.205
0.410

1.024
1.228
1.433
1.637
1.842
2.047

2.456
Other

Fig. 4. High performances: results of Monte Carlo simulations (500 runs): (a), (c), (e): maximum values of Vrot, Al and Av; (b), (d), (f): corresponding histograms and
cumulative distribution functions.

moment coefficient cM is more relevant than the uncertainty of the the quasi-steady (QS) theory (the instantaneous aeroelastic forces
lift coefficient cL. acting on the structure are the same that act on the structure when
it moves with constant translational and rotational velocities, with
6.3. Uncertainty of the wind-action model values equal to the actual instantaneous ones: the angle of inci-
dence changes with time due to both the incident wind turbulence
Also the uncertainties in modeling the aeroelastic actions play a and the torsional rotation – twist – of the deck, but is computed by
major role: this question has been investigated with some detail in considering the relative velocity between the wind and the deck);
[28] and [47]. Lacking a quantitative treatment, four simplified the modified quasi-steady (QSM) theory (the aerodynamic lift and
models of increasing complexity have been successively consid- moment coefficients are considered variable with time and mea-
ered, namely: the non-aeroelastic (NO) theory (aeroelastic effects sured by wind tunnel tests [52]). The analyses have been per-
are disregarded, and the angle of incidence of the wind changes formed for mean incident wind speeds equal respectively to 21,
with time only due to wind turbulence); the steady (ST) theory 45 and 57 m/s. It has been noted (see [47]) that, when the action
(the angle of incidence changes with time due to both the incident model complexity increases (from NO to QSM), both the mean val-
wind turbulence and the torsional rotation – twist – of the deck); ues and the variances of the deck response parameters (displace-
M. Ciampoli et al. / Structural Safety 33 (2011) 367–378 375

1
P( Vrot > v rot*|V m(zdeck))
(a) 1.0
0.9
0.8
0.8

0.7

G(EDP)
0.6
0.6

0.5 0.4

0.4
0.2
0.3

0.2 0
0.1 0 0.02 0.04 0.06 0.08 0.10 0.12
V m(zdeck) [m/s] EDP = Vrot - DM = max (vrot) [rad/s]
0
0 5 10 15 20 25 30 35

Fig. 5. Fragility curves for EDP = Vrot and three different threshold values – D:
(b) 1.0
mrot = 0.02 rad/s; s: mrot = 0.03 rad/s; h: mrot = 0.04 rad/s (full serviceability).
0.8

G(EDP)
0.6

1
0.4
0 .9 P( A v > a v*|V m(z deck ))

0 .8 0.2

0 .7

0 .6
0 0.10 0.20 0.30 0.40 0.50
0 .5
EDP = Al - DM= max (Al ) [m/s 2 ]
0 .4

0 .3 (c) 1.0
0 .2
0.8
0 .1 Vm(z deck ) [m/s]
G(EDP)

0 0.6
0 5 10 15 20 25 30 35

Fig. 6. Fragility curves for EDP = Av and three different threshold values – D: 0.4
am = 0.45 m/s2; s: am = 0.63 m/s2; h: am = 0.90 m/s2 (full serviceability).

0.2

0
1 0 1 2 3

0.9 * + EDP = A v - DM= max (a v) [m/s2 ]

0.8 Fig. 8. Exceedance probabilities (a) G(Vrot), (b) G(Al), (c) G(Av).

0.7
+*
0.6

0.5
Table 2
0.4 ‘‘Failure’’ probabilities.

0.3 Limit Failure Probability of Probability of not


state criterion not satisfying satisfying either
0.2 each performance performance
criterion criterion
0.1
Vm(zdeck ) [m/s] SLS SLS-1 Vrot P 0.04 rad/s 0.0720 0.1541
0 Al P 2.5 m/s2 ffi0
0 5 10 15 20 25 30 35 Av P 0.7 m/s2 0.150
SLS-2 Vrot P 0.043 rad/s 0.0620 0.0980
Fig. 7. Comparison between the fragility curves corresponding to each full Al P 2.5 m/s2 ffi0
serviceability criterion (+: Av; : Vrot) and (in bold line) to the violation of either Av P 1.0 m/s2 0.0920
performance criterion.
376 M. Ciampoli et al. / Structural Safety 33 (2011) 367–378

0.3 1.0
(a) z0=0.2
0.2 z0=0.1
0.8 z0=0.05
0.1

0
0.6

G(EDP )
-10 -8 -6 -4 -2 0 2 4 6 8 10
-0.1
[deg]
-0.2 0.4

-0.3

-0.4 0.2

-0.5
Drag Lift Moment 0
0 0.02 0.04 0.06 0.08 0.10 0.12

0.3 EDP = Vrot - DM = max ( vrot) [rad/s]


(b)
0.2 Fig. 11. Exceedance probability functions G(Vrot) for three values of the roughness
length.
0.1

0
-10 -8 -6 -4 -2 0 2 4 6 8 10
-0.1
[deg]
-0.2 Vcrit
400
-0.3

-0.4 300

-0.5
Drag Lift Moment 200

Fig. 9. Polar lines (plots of aerodynamic coefficients): (a) basic; (b) alternative.
100

0
(a) 0.525
1
q [m] 1
0
0.520 0.5
q=q+q0· e-δt PPMM 0.5
P
PL
L
0.515 0 0
= computed by the = obtained by minimum
0.510 FE model squares method

Fig. 12. Critical flutter velocities: response surface obtained by varying the polar
0.505 lines.

t t[sec]
[sec]
0.500 ments, rotations, velocities) decrease. The same trend appears also
600
600 700 800 900
900 1000
1000 in the distributions along the whole length of the bridge of the
maximum transversal and vertical displacements.
(b) 1.5
Damp (%)
1.0 7. Concluding remarks

0.5 This paper essentially aims at setting up a roadmap towards a


general procedure for the assessment of the performances of struc-
0.0 tures subject to wind actions. The approach proposed by PEER for
0 10 20 30 40 50 60 70 80 Performance-Based Seismic Engineering has been extended to Per-
-0.5 formance-Based Wind Engineering (PBWE): the structural risk is
Wind Velocity [m/s]
identified with the mean annual probability of exceeding a thresh-
-1.0
old level of a Decision Variable. The general aspects of the proposed
approach have been outlined and exemplified with reference to a
-1.5
specific case: the assessment of the out-of-service and collapse
Total Structural Aerodynamic
risks of a long span suspension bridge. The main results of the
Fig. 10. (a) Typical time series of the vertical displacement at bridge midspan for a numerical calculations have been illustrated and discussed. In
given value of Vm; (b) damping as functions of Vm. the final Section 6, a few qualitative results on the effects of some
M. Ciampoli et al. / Structural Safety 33 (2011) 367–378 377

sources of uncertainty have been presented: their great importance [23] Bashor R, Kijewski-Correa T, Kareem A. On the wind-induced response of tall
buildings: the effect of uncertainties in dynamic properties and human
is evident. It can be concluded that PBWE is feasible, but to make it
comfort thresholds. In: Proceedings of the 10th Americas conference on wind
more reliable it is essential to improve the probabilistic description engineering, Baton Rouge, LA; 2005. <http://www.nd.edu/~nathaz/>.
of the parameters of the wind field at the site and the phenomena [24] Bashor R, Kareem A. Probabilistic performance evaluation of buildings: an
that represent the interaction between the wind actions and the occupant comfort perspective. In: Proceedings 12th international conference
on wind engineering, Cairns, Australia; 2007. <http://www.nd.edu/~nathaz/>.
structure. This will require collecting and elaborating many more [25] Tamura Y. Wind and tall buildings. In: Borri C, Augusti G, Bartoli G, Facchini L,
experimental data, and much further research work also to improve editors. Proceedings of the 5th European and African conference on wind
the framework proposed in this paper, including optimization. engineering; Florence, Italy. Florence: Firenze University Press; 2009. p. k3–
k28.
[26] Kim JY, Yu E, Kim DY, Kim SD. Calibration of analytical models to assess wind-
Acknowledgements induced acceleration responses of tall buildings in serviceability level. Eng
Struct 2009;31(9):2086–96.
[27] Ciampoli M, Petrini F, Augusti G. A procedure for performance-based wind
The researches presented in this paper have been developed engineering. In: Furuta I, Frangopol D, Shinozuka M, editors. Safety, reliability
within the Wi-POD Project (2008–2010) and other research pro- and risk of structures, infrastructures and engineering systems – proceedings
of the 10th international conference on structural safety and reliability,
jects in wind engineering, partially financed by the Italian Ministry
ICOSSAR09; Osaka, Japan, AK. Leiden: CRC Press; 2009 (Paper in CD-ROM).
for Education, University and Research (MIUR). This paper is an ex- [28] Petrini F, Ciampoli M, Augusti G. The role of uncertainties in Aeolian Risk
tended and updated version of the papers presented at ICOSSAR’09 assessment. In: Papadrakakis M, Stefanou G, Papadopoulos V, editors.
Computational methods in stochastic dynamics, computational methods in
[27], COMPDYN’09 [28] and IMECE’09 [30], and is a synthesis of the
applied sciences, vol. 22. Springer; 2011. p. 187–208.
results illustrated in [45]. [29] Petrini F, Ciampoli M, Augusti G. A probabilistic framework for performance-
based wind engineering. In: Borri C, Augusti G, Bartoli G, Facchini L, editors.
Proceedings of the 5th European and African conference on wind engineering;
References Florence, Italy. Florence: Firenze University Press; 2009 [abstract 119, paper in
CD-ROM].
[1] Augusti G, Ciampoli M. Performance-based design in risk assessment and [30] Augusti G, Ciampoli M, Petrini F. Reliability of structural systems under wind
reduction. Probabilist Eng Mech 2008;23(4):496–508. action. In: Proceedings of the ASME 2009 international mechanical
[2] Augusti G, Ciampoli M. Performance-based design as a strategy for risk engineering congress and exposition IMECE2009: mechanical systems and
reduction: application to seismic risk assessment of composite steel-concrete control, Parts A and B, vol. 10, November 13–19, Lake Buena Vista, Florida,
frames. In: Korytowski A, Malanowski K, Mitkowski W, Szymkat M, editors. USA; 2009 [Paper No. 12357 in CD-ROM].
System modeling and optimization – 23rd IFIP TC7 Conference. Springer [31] Bontempi F. Basis of design and expected performances for the Messina Strait
Verlag; 2009. p. 3–22. Bridge. In: Proceedings of the international conference on bridge engineering –
[3] Porter KA. An overview of PEER’s performance-based engineering methodology. In: challenges in the 21st Century, November 1–3, Hong Kong, China; 2006 [Paper
Armen Der Kiureghian A, Madanat S, Pestana JM editors. Applications of statistics in CD-ROM].
and probability in civil engineering – proceedings of the ninth international [32] Simiu E, Scanlan RH. Wind effects on structures, 3rd ed. New York: John Wiley
conference on applications of statistics and probability in civil engineering ICASP9; & Sons Inc; 1996.
San Francisco, CA, USA. Rotterdam: Millpress; 2003. p. 973–80. [33] Augusti G, Ciampoli M. First steps towards Performance-based wind
[4] Paulotto C, Ciampoli M, Augusti G. Some proposals for a first step towards a engineering. In: Bartoli G, Ricciardelli F, Saetta A, Sepe V, editors.
performance based wind engineering. In: Proceedings of the IFED- Performance of wind exposed structures: results of the PERBACCO Project,
international forum in engineering decision making; First Forum, December, Florence: Firenze University Press; 2006. p. 13–20.
5–9, Stoos, Switzerland; 2004. <http://www.ifed.ethz.ch>. [34] Stathopoulos T. Wind and Comfort. In: Borri C, Augusti G, Bartoli G, Facchini L,
[5] Kareem A. Wind effects on structures: a probabilistic viewpoint. Probabilist editors. Proceedings of the 5th European and African conference on wind
Eng Mech 1987;4(2):166–200. engineering. Florence: Firenze University Press; 2009. p. k67–k82.
[6] Schuëller GI, editor. A state-of-the-art report on computational stochastic [35] Kwok KCS, Burton MD, Hitchcock PA. Human perception of tall building
mechanics. Probabilist Eng Mech 1997;12(4):197–321. motions in strong wind environments. In: The 5th international
[7] Kareem A. Numerical simulation of wind effects: a probabilistic perspective. J advanced school on wind engineering – GCOE program new frontier of
Wind Eng Ind Aerodyn 2008;96:1472–97. education and research in wind engineering; Opole, Poland, March 23–25,
[8] Davenport AG. On the assessment of the reliability of wind loading on low 2009. p. 159–74.
buildings. J Wind Eng Ind Aerodyn 1983;11(1–3):21–37. [36] Augusti G, Ciampoli M. Heritage buildings and seismic reliability. Prog Struct
[9] Kasperski M, Niemann HJ. A general method of estimating unfavorable load Eng Mater 2000;2(2):225–37.
distributions for linear and non-linear structural behavior. J Wind Eng Ind [37] Morgenthal G. Aerodynamic Analysis of Structures Using High-resolution
Aerodyn 1992;43(1–3). Vortex Particle Methods. Ph.D. Thesis, University of Cambridge, Department of
[10] Holmes JD. Wind loading of structures, application of probabilistic methods. Engineering, Magdalene College; 2002.
Prog Struct Eng Mater 1998;1(2):193–9. [38] Bruno L, Khris S. The validity of 2D numerical simulations of vortical structures
[11] Solari G, Piccardo G. Probabilistic 3-D turbulence modeling for gust buffeting around a bridge deck. Math Comput Model 2003;37:795–828.
of structures. Probabilist Eng Mech 2001;16:73–86. [39] Augusti G, Borri A, Ciampoli M. Seismic protection of constructed facilities:
[12] Minciarelli F, Gioffrè M, Grigoru M, Simiu E. Estimates of extreme wind effects optimal use of resources. Struct Safety 1994;16(1,2):91–109.
and wind load factors: influence of knowledge uncertainties. Probabilist Eng [40] Spence SMJ. Time Domain Non-Gaussian Optimization of Wind Excited Tall
Mech 2001;16:331–40. Buildings under Vulnerability Constraints. PhD Dissertation, University of
[13] Kareem A. Aerodynamic response of structures with parametric uncertainties. Florence and Technical University of Braunschweig (joint PhD course), Italy
Struct Safety 1988;5:205–25. and Germany; 2010.
[14] Solari G. Wind-excited response of structures with uncertain parameters. [41] Holmes JD. Wind loading of structures. Spoon Press; 2001.
Probabilist Eng Mech 1997;12(2):75–87. [42] Lungu D, Rackwitz R. Joint Committee on Structural Safety – Probabilistic
[15] Pagnini LC, Solari G. Serviceability criteria for wind-acceleration and damping Model Code, Part 2: Loads; 2001. <http://www.jcss.ethz.ch/>.
uncertainties. J Wind Eng Ind Aerodyn 1998;74–76:1067–78. [43] Carassale L, Solari G. Monte Carlo simulation of wind velocity field on complex
[16] Zhang L, Jie L, Peng Y. Dynamic response and reliability analysis of tall structures. J Wind Eng Ind Aerodyn 2006;94:323–39.
buildings subject to wind loading. J Wind Eng Ind Aerodyn 2008;96:25–40. [44] ESDU (Engineering Sciences Data Unit) Report N. 86010. Characteristic of
[17] Unanwa CO, McDonald JR, Mehta KC, Smith DA. The development of wind atmospheric turbulence near the ground. Part III: variations in space and time
damage bands for buildings. J Wind Eng Ind Aerodyn 2000;84:119–49. for strong winds (neutral atmosphere); 2001. <http://www.esdu.com>.
[18] Rosowski DV, Ellingwood BR. Performance-based engineering of wood frame [45] Petrini F. A probabilistic approach to Performance-Based Wind Engineering
housing: fragility analysis methodology. J Struct Eng 2002;128(1):32–8. (PBWE). PhD Dissertation, University of Rome ‘‘La Sapienza’’, Rome, Italy;
[19] Ellingwood BR, Rosowsky DV, Li Y, Kim JH. Fragility assessment of light-frame 2009.
wood construction subjected to wind and earthquake hazards. J Struct Eng [46] Petrini F, Bontempi F. Estimation of fatigue life for long span suspension bridge
2004;130(12):1921–30. hangers under wind action and train transit. Struct Infrastruct Eng
[20] van der Lindt JW, Dao TN. Performance-based wind engineering for wood- 2010;7(7):491–507.
frame buildings. J Struct Eng 2009;135(2):169–77. [47] Petrini F, Giuliano F, Bontempi F. Comparison of time domain techniques for
[21] Khanduri AC, Morrow GC. Vulnerability of buildings to windstorms and the evaluation of the response and the stability in long span suspension
insurance loss estimation. J Wind Eng Ind Aerodyn 2003;91:455–67. bridges. Comput Struct 2007;85:1032–48.
[22] Garciano LE, Maruyama O, Koike T. Performance-based design of wind [48] Salvatori L, Borri C. Frequency and time-domain methods for the
turbines for typhoons. In: Augusti G, Schuëller GI, Ciampoli M, editors. numerical modeling of full-bridge aeroelasticity. Comput Struct
Safety and reliability of engineering systems and structures – proceedings of 2007;85:675–87.
the 9th international conference on structural safety and reliability [49] CEN 2004. EN 1991 1-4: actions on structures: Part 1–4: wind actions.
ICOSSAR05; Rome, Italy. Rotterdam: Millpress; 2005 (Paper in CD-ROM). Brussels, BE.
378 M. Ciampoli et al. / Structural Safety 33 (2011) 367–378

[50] Jalayer F, Franchin P, Pinto PE. A scalar damage measure for seismic reliability [52] Diana G, Falco M, Bruni S, Cigada A, Larose GL, Damsgaard A, et al. Comparison
analysis of RC frames. Earthquake Eng Struct Dyn 2007;36:2059–79. between wind tunnel tests on a full aeroelastic model of the proposed bridge
[51] Borri C, Costa C. Quasi-steady analysis of a two-dimensional bridge deck over Stretto di Messina and numerical results. J Wind Eng Ind Aerodyn
element. Comput Struct 2004;82:993–1006. 1995;54/55:101–13.

You might also like