Soil Dynamics and Earthquake Engineering 114 (2018) 505–519

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

Analysis of seismic non-structural damage in single-storey industrial steel T

buildings
⁎
Fabrizio Scozzesea, Giusy Terraccianob, Alessandro Zonaa, , Gaetano Della Corteb,
Andrea Dall’Astaa, Raffaele Landolfob
a
School of Architecture and Design, University of Camerino, Italy
b
Department of Structures for Engineering and Architecture, University of Naples Federico II, Italy

A R T I C LE I N FO A B S T R A C T

Keywords: This study illustrates an approach that allows evaluating the intensity and extent of non-structural damage in
Damage limit state steel buildings. Different levels of cladding panel damage are introduced and a specific cladding panel hysteretic
Non-structural elements model is included in the structural model and calibrated on available data from experimental tests. Numerical
Cladding panels results from nonlinear dynamic analyses are elaborated in the form of fragility curves for the damage limit state
Steel structures
and used to compare the consequences of different cladding behaviours on prototype single-storey industrial
Industrial steel buildings
steel buildings used as case studies. The outcomes are also compared to a damage limit state derived from
Seismic analysis
structural analysis adopting bare-frame models and conventional drift-based damage criteria.

1. Introduction
The seismic design and the structural analysis of steel constructions
are commonly based on models where only the structural elements (e.g.
beams, columns, braces) contribute to the definition of the strength and
stiffness, while non-structural elements (e.g. cladding panels) are con-
sidered in the overall seismic mass but are not included as components
explicitly influencing the structural response, e.g. [1,2]. Exceptions are
few studies published in the past two decades that incorporated clad-
ding panels in the structural model. Earlier investigations made in Italy
[3–5] focused on lightweight sandwich panels in single-storey steel
industrial building and developed hysteretic models for such panels
based on experimental tests. Starting from the evaluation of the effect of
cladding panels on the seismic response, the final goal was to propose a
design approach where non-structural elements are designed to con-
tribute bracing functions [6]. The same goal was pursued in studies
made in Canada [7–11] that investigated the effect of cladding panels
on the behaviour of the roof diaphragms and developed procedures for
their seismic design. Similarly, researches made in Japan [12,13] stu-
died the influence of precast lightweight concrete panels in steel frame
structures and evidenced an enhanced seismic behaviour compensating
the increment of structural mass.
A different perspective is tackled in the study here presented: (1) the
seismic design is based on structural models where panels contribute
only to the definition of permanent loads and mass while their stiffness
and strength are neglected, according to the common practice (bare-
frame models); (2) structural models that explicitly include the non-
linear hysteretic behaviour of panels are afterwards implemented and
used to analyse the seismic response of the designed structures; (3) the
results of the numerical simulations are adopted to evaluate the damage
in the non-structural elements for increasing seismic intensities as a
measure of the performance at the damage limit state (DLS) of steel
buildings, e.g. [1,2], and how the results obtained compare to con-
ventional drift-based code limits for damage verifications using bare-
frame models, e.g. [14,15]. This last aspect is very important to prop-
erly address the seismic performance not only in terms of collapse
under the most severe earthquakes, but also in terms of functionality
and economic loss, as for example examined in recent studies [16,17]
following the 2012 Emilia earthquakes in Italy and highlighting the
important consequences of more frequent low-to-moderate seismic
events. Within this context, numerical applications are presented using
as case studies single-storey industrial steel buildings, a structural so-
lution very common in many regions of Southern Europe with moderate
to high seismic hazard. The case studies are designed according to the
current version of the Eurocodes [14,15], neglecting the stiffness and
strength of the claddings. Afterwards, a geometric and material non-
linear model that includes the hysteretic behaviour of vertical and roof
claddings is described and adopted to perform nonlinear static and
dynamic analyses, aiming at an examination of the damage level at-
tained in each panel and the damage extent in the buildings.

⁎
Correspondence to: University of Camerino, Viale della Rimembranza, Ascoli Piceno 63100, Italy.
E-mail address: alessandro.zona@unicam.it (A. Zona).

https://doi.org/10.1016/j.soildyn.2018.07.047
Received 26 February 2018; Received in revised form 4 July 2018; Accepted 28 July 2018
0267-7261/ © 2018 Elsevier Ltd. All rights reserved.
F. Scozzese et al.
2. Criteria for non-structural damage evaluation
The seismic performance of a building in terms of non-structural
damage can be based on either local response quantities, describing the
behaviour of the non-structural elements and attained damage, or
global response quantities such as the interstorey drift. In general, at-
tention should be paid to the relations between the accuracy of eva-
luation criteria based on global response parameters with respect to
those based on local response parameters, as for examples investigated
in [18]. This makes comparisons between the two approaches (local-
and global-based) very useful to gain more insight into the possible
limitation of damage criteria based solely on global response para-
meters.
In this study the damage evaluation of the non-structural elements is
pursued through their inclusion in the structural model. Being the en-
gagement of the non-structural elements in the seismic response ex-
plicitly described, two outcomes are achieved: (1) estimation of the
local seismic demand on the non-structural elements; (2) estimation of
the influence of the non-structural elements on the global and local
seismic response of the building as compared to the bare-frame model.
Since non-structural elements are included in the structural model, a
performance-based approach for DLS assessment can be derived mon-
itoring different damage conditions attained by the cladding panels
under seismic events. Specifically, the local panel-related damage cri-
teria proposed to characterize the attainment of the DLS for a structural
system are listed and described in Table 1. Given the lack of sufficiently
large studies on this topic, the adopted damage conditions are based on
engineering judgment while the implemented approach is inspired by
recent studies made for masonry infill walls used in reinforced concrete
buildings [19].
3. Nonlinear modelling of cladding panels
3.1. Hysteretic model
The inclusion of the non-structural elements in the nonlinear model
of the building requires that the nonlinear cyclic behaviour of each
panel (or assembly of panels) is simulated through properly calibrated
hysteretic models. The availability of a suitable hysteretic model is
fundamental to realistically simulate the cyclic response of the cladding
panels. To this aim, the Pinching4 [20] one-dimensional model im-
plemented in OpenSees [21] is adopted to catch the main features of the
shear-displacement experimental response observed for non-structural
elements as used in steel constructions in Europe [4,5], i.e. the pinched
shape of the cycles and the strength and stiffness deterioration under
cyclic loading.
In the Pinching4 implementation, the hysteretic model is governed
by 34 parameters, 16 control the multilinear envelope curve (response
backbone curve), 6 control the trilinear unload-reload path, and 12 set
the properties of three possible hysteretic damage rules, i.e. unloading
stiffness degradation, reloading stiffness degradation and envelope
strength degradation. For the purposes of the present investigation, the
aforementioned parameters are calibrated in order to fit the cyclic
Table 1
Panel-related damage conditions.
Damage level Description
Low-damage An assigned percentage of panels exceed their elastic response
limit (in this study investigations are made using 30% and
50% values);
Medium-damage The totality of panels exceed their elastic response limit;
Strong-damage At least one panel exceeds its maximum shear resistance, thus,
providing negligible contribution in terms of both stiffness and
strength.
506
Soil Dynamics and Earthquake Engineering 114 (2018) 505–519
response of the cladding panels considered in this study for which ex-
perimental tests are available in the scientific literature, e.g. [4,5].
3.2. Example of calibration based on experimental tests
Two different cladding panels are considered in this work, hereafter
referred to as type A and C according to the terminology adopted in De
Matteis and Landolfo [5]. These panels belong to a typology of light-
weight sandwich panels which are widely used as enclosure elements in
both industrial and civil buildings [4,5]. Both typologies consist of
sandwich panels made of external steel sheets (thickness of 0.6 mm)
and insulating polyurethane core for a total thickness of 40 mm. Em-
bossed sheets with slight stiffening ribs are provided to the panel type
A, while completely flat sheets are employed for the type C panel.
Furthermore, the type C panel edges are reinforced with internal cold
formed channel profiles connected to the sheets through aluminium
rivets of 6.3 mm diameter. Panels are usually connected to the main
structural frame through proper cladding rails (vertical cladding pa-
nels) or purlins (roof panels) by means of bolts (generally of 8 mm in
diameter and 110–120 mm spacing) [5].
Each panel is modelled in OpenSees by means of a couple of diag-
onal truss elements with nonlinear axial behaviour following the
Pinching4 model presented above. For both panel typologies the
parameters governing the shear-displacement relationship are cali-
brated based on the experimental results available in [4,5], consisting
of cladding sandwich panels type A and C (dimension
1000 ×2500 ×40 mm) subjected to cyclic shear loading.
The backbone curve for panel A is schematically represented in
Fig. 1a, superimposed to the envelope of the experimental cyclic re-
sponse. It consists of a trilinear model characterized by: a yielding point
at 12 kN (d=20 mm), a hardening behaviour until the maximum shear
of 17 kN (d=80 mm), and a gradual softening branch where the force
drops to zero at d = 320 mm. The simulated quasi-static nonlinear
response for panel A is shown in Fig. 1b, and it is worth noticing its
satisfactory match with the pinched experimental shear-displacement
response (Fig. 1a).
The backbone curve for the panel C, shown in Fig. 2a, consists of a
quadrilinear model having the following features: a first yielding point
at 11 kN (d=5 mm), followed by a branch with slightly reduced stiff-
ness until the maximum strength of 23 kN (d=25 mm), and a perfectly
plastic response extended until the 80 mm of displacement, followed by
a gradual softening behaviour with the null shear force attained at
d= 320 mm. The simulated quasi-static nonlinear response for panel C
is reported in Fig. 2b.
It is important to clarify that the latter part of the backbone curves
(softening trends) for both A and C panels were extrapolated from the
experimental results, trying to realistically describe the complex dete-
riorating behaviour (panels buckling, connections failure, etc.), al-
though the available test results do not characterize the response be-
yond the 80 mm displacement threshold (Fig. 1a and Fig. 2a).
4. Numerical application
4.1. Case studies
The structural system investigated in this paper consists of single-
storey industrial steel buildings composed of five single span duo-pitch
portal frames equally spaced (Fig. 3). The portal frames are connected
in the longitudinal direction by hot-rolled beams at the apex, at the
eaves and at the crane-supporting bracket level. Horizontal forces are
withstood by two different seismic-resistant mechanisms: in the trans-
verse direction (X-direction) the resistance is due to continuous and
rigid frame action, which exploits the moment resisting connections
between beams and columns; along the longitudinal direction (Y-di-
rection) the resistance to horizontal forces is provided by vertical bra-
cings placed in the outer spans of the building. Purlins, used to support
F. Scozzese et al.
(a)
Fig. 1. Nonlinear behaviour of cladding panels: (a) Experimental cyclic response of Type A panel system (1000 × 2500 × 40 mm) under shear dynamic loads with
envelope backbone curve superimposed; b) experimental vs numerical cycle responses.
the roof cladding and positioned every 2.5 m, transfer the forces from
the roof cladding to the rafters. Roof bracings are arranged in the outer
bays (Fig. 3) to transfer horizontal forces to the vertical bracings.
Full-strength bolted end-plate connections at eaves and apex are as-
sumed. As typical for this structural type, the rafter connection to the
column is equipped with a haunch (10% of the span long) used to improve
the performance of the rafter and facilitate a bolted connection to the
column. For the same reason, small haunches are considered at the apex as
well. The roof purlins are connected to the rafters by pinned connections
and assumed as simple spanned between two consecutive frames. All the
columns are assumed as hinged at their base. Hot-rolled I or H sections are
used for beams, columns, and purlins. Cold-formed steel square hollow
sections were chosen for vertical braces, while hot-rolled L-shaped profiles
are used for roof bracings. The selected steel grade is S275, with char-
acteristic yield strength equal to 275 MPa and elastic modulus equal to
210,000 MPa.
The transverse bay width (Lx), the longitudinal bay width (Ly), the
height at the eaves (H) and the height of the crane-supporting bracket
(measured at top surface of the bracket) are assumed as design para-
meters (Fig. 3). Four geometrical configurations are selected (Table 2).
In all cases, the roof pitch is equal to 6°.
4.2. Structural design
The four structural systems, assumed to be located in L’Aquila
(a)
Fig. 2. Nonlinear behaviour of cladding panels: (a) experimental cyclic response of Type C panel system (1000 ×2500 ×40 mm) under shear dynamic loads with
envelope backbone curve superimposed; b) experimental vs numerical cycle responses.
507
Soil Dynamics and Earthquake Engineering 114 (2018) 505–519
(b)
(Italy), are designed according to the Eurocodes [14,15] in order to
fulfil the performance requirements at the ultimate limit state (ULS)
and the damage limit state (DLS), to ensure a safe and functional
structure, respectively. The global behaviour of the systems for the
design purposes is analysed through a 3D elastic model developed with
the commercial software MIDAS Gen [20]. The considered static loads
are: permanent loads, wind actions, snow loads, crane loads (crane self-
weight plus equivalent static loads for the vertical variable hoist load
and the horizontal variable forces caused by acceleration, deceleration
and skewing of the crane during its movement along the runway
beams). Two positions of the crane, along the longitudinal direction of
the building, are considered in order to obtain the worst load ar-
rangements for ULS and DLS conditions, and these correspond to the
second and third portal frame alignment respectively. Frame im-
perfections are modelled by means of a system of equivalent horizontal
forces, according to Eurocode 3 [14]. Seismic design is based on the
evaluation of the seismic demand through modal response spectrum
analysis on the global three-dimensional elastic structural models, by
considering both the horizontal and vertical seismic components, the
latter because of the large transverse span. The seismic hazard is de-
scribed by means of the parameters (ag, F0, Tc*) defined in [15] and
reported in Table 3. The soil category at the site is type C according to
[15], i.e. deep deposits of dense or medium-dense sand, gravel or stiff
clay with thickness from several tens to many hundreds of metres. Each
case study is designed by assuming a behaviour factor q = 4 for both
(b)
F. Scozzese et al. Soil Dynamics and Earthquake Engineering 114 (2018) 505–519

Fig. 3. Geometry of the considered case studies.

Table 2 imperfections of the lateral bracing systems, as later described. Material

Geometry parameters for the considered case studies. nonlinearity is included using a diffused plasticity approach, i.e. all struc-
Case study Lx (m) H (m) Hc (m) Ly (m)
tural elements are modelled using force-based frame elements with non-
linear fibre sections, with each fibre following the Steel02 constitutive law
X20Y6 20.00 6.00 4.50 6.00 (Giuffré-Menegotto-Pinto model) as implemented in OpenSees [21]. In
X20Y8 20.00 6.00 4.50 8.00 addition, concentrated plasticity models are adopted for the description of
X30Y6 30.00 9.00 6.00 6.00
the behaviour of the connections with the braces, whose behaviour is de-
X30Y8 30.00 9.00 6.00 8.00
scribed following the approach proposed by Hsiao et al. [22,23] in order to
account for the buckling phenomenon in compression and properly mod-
Table 3 elling the gusset plate connections. Specifically, the nonlinear out-of-plane
Seismic hazard parameters. rotational behaviour of the gusset plate connection is simulated through a
rotational nonlinear spring located at the physical end of the brace. The
Site Latitude Longitude Limit state ag Fo Tc*
(g) (-) (s)
elastic stiffness kcon of the spring is calibrated based on the actual geometry
of the connection through the following Eq. (1):
L’Aquila 13.396 42.356 DLS 0.104 2.330 0.280
ULS 0.261 2.360 0.350 Es ⎛ WW t 3 ⎞
kcon = ⎜ ⎟

La ⎝ 12 ⎠ (1)

the moment resisting frames and frames with concentric bracings
(Ductility Class Medium).
In the transverse direction, the design of the frame structures is
governed by the limits of the DLS, in terms of horizontal deflection or
difference of deflection between two consecutive portal frames. For
DLS, the horizontal drift is limited to 1/200 (i.e., the 0.5% of the
height), following the most conservative assumption that claddings are
not designed to be damage-tolerant to seismic actions [15]. In the
longitudinal direction, the cross section of the beams and diagonals is
controlled by the slenderness limits, i.e. the diagonal members of X-
braced frames are designed to enforce their normalized slenderness in
the range 1.3–2, while for single bay diagonal braces the normalized
slenderness is limited to 2.0 as per Eurocode 8 [15].
A list of the designed cross-sections of all the structural members is
reported in Table 4, for each case study. The crane runway beams were
assumed to be the same for all cases and made of a HEA400 steel beam.
where Es is Young's modulus, Ww is the Whitmore width defined by a 45°
projection angle, La is the average of L1, L2 and L3 as shown in Fig. 4, and t
is the thickness of the gusset plate. The post elastic stiffness is assumed as
the 1.0% of kcon. In order to catch the buckling phenomenon of the lateral
braces during the compression phases, each brace is discretized into a
proper number of nonlinear sub-elements and a sinusoidal curvature is
assigned by modifying parametrically the coordinates of the nodes of the
intermediate sub-elements (Fig. 5). This initial curvature, representing the
local imperfection of the diagonal brace, has the role of triggering the
buckling by furnishing a preferential buckling shape to the element. The
value of the initial imperfection is chosen in such a way to furnish a
buckling axial force consistent with the ultimate value of the axial com-
pressive force Nb,Rd provided according to Eurocode 3 [14] and given by
Eq. (2):
fy
Nb, Rd = χA
γM1 (2)

4.3. Nonlinear structural model
The geometric and material nonlinear finite element model adopted for
studying the seismic response of the considered case studies including the
vertical and roof cladding panels is developed in OpenSees [21]. The cor-
otational approach is adopted in order to take into account the nonlinear
geometric effects due to both the large displacements and the local
being γM1 = 1, A the cross-section area, fy the nominal yielding stress and χ
the buckling reduction factor depending on the type of profile and the
slenderness of the element. The calibration of the values of the initial sway
δ of the lateral braces is achieved by performing nonlinear displacement-
controlled static analyses on each geometric configuration of the braces:
the single brace configuration and the X-configuration. The latter is made
by two braces, one continuous and one cut, intersecting each other with a

Table 4
Sections of the designed structural elements of the considered case studies.
Case Study Column Rafter Vertical X-brace Vertical single brace Long. beam Purlins Roof bracing

X20Y6 HE600M HE500A RHS-CF-60 × 60 × 2 RHS-CF-90 × 90 × 2.6 IPE270 HE220A L20 × 3

X20Y8 HE600M HE500A RHS-CF-70 × 70 × 2.6 RHS-CF-100 × 100 × 2.6 HE300B HE220A L20 × 3
X30Y6 HE900M HE800A RHS-CF-80 × 80 × 2.6 RHS-CF-100 × 100 × 2.6 IPE270 HE220A L50 × 4
X30Y8 HE900M HE800A RHS-CF-90 × 90 × 2.6 RHS-CF-120 × 120 × 3.2 HE300B HE220A L50 × 4

508
F. Scozzese et al. Soil Dynamics and Earthquake Engineering 114 (2018) 505–519

Table 5
First four vibration periods for the bare-frame models.
Case study T1 [s] T2 [s] T3 [s] T4 [s]

X20Y8 0.67 0.4799 0.4797 0.35

X20Y6 0.60 0.4109 0.4107 0.32
X30Y8 0.86 0.6153 0.6150 0.38
X30Y6 0.80 0.5562 0.5561 0.39

4.4. Nonlinear model for cladding panels

Two different hypotheses of non-structural cladding elements are

Fig. 4. Geometrical quantities involved into the gusset plate modelling.
gusset plate connection. A schematic representation of the modelling cri-
teria adopted for the different brace configurations is given in Fig. 5 where
the local imperfections δ are also highlighted.
All columns are pinned at their bases. Loads are assigned as either
point (i.e., crane loads) or uniformly distributed forces (i.e., gravity
loads acting on beams and columns), while the masses are lumped at
the nodes according to their real spatial distribution and influence area.
The Rayleigh model is adopted for the system inner (elastic)
damping by assuming a damping rate of ξ = 5% on the first and third
vibration modes (modal shapes and periods are reported in Fig. 8 and
Table 5, respectively).
considered, consisting of the sandwich panel type A and panel type C
already introduced in the previous Section 3, where the properties of a
pair of Pinching4 truss elements were calibrated in order to reproduce
the cyclic experimental response of the 1000 × 2500 × 40 mm panels
under shear loading. The number of panels (1000 × 2500 × 40) em-
ployed in each case study varies based on the specific geometric con-
figuration, and a schematic representation of the vertical and roof panel
distribution is given in Fig. 6 for case study X20Y8, in which the
opening locations on the cladding walls are also depicted. It is worth
noticing that the vertical cladding panels in the longitudinal direction
do not cover the entire façade of the building, but they extend up to the
crane support level, leaving the whole upper field free for ribbon
windows. The equivalent truss model calibrated for the single panel is
now extended to simulate the nonlinear contribution of whole panel
assemblies of both vertical and roof cladding components belonging to
different structural fields, as shown in Fig. 7 (the truss elements are
depicted for the faces in-view only, but they are inserted in the shadow
sides as well).
According to the modelling strategy described, each couple of truss
elements is equivalent, in terms of cyclic dynamic response, to a single

Fig. 5. Brace modelling with assigned geometric imperfection: (a) single brace case; (b) X-brace case.

509
F. Scozzese et al.
Fig. 6. Panel distribution for the case study X20Y8.
Fig. 7. Equivalent truss elements location (shown only in the visible faces)
within the structural model.
panel, and hence in order to account for the real number of panels
placed within the different structural fields the following hypothesis are
made: (i) the presence of several adjacent panels is accounted by in-
creasing the strength and the stiffness of each pair of truss elements by a
factor equal to the number of panels, i.e. the actual number of panels,
which can also be a fractional number (this assumption is consistent
with the condition of panels working in parallel); (ii) the presence of
510
Soil Dynamics and Earthquake Engineering 114 (2018) 505–519
multiple rows of panels along the height is accounted by following the
analogy of a system working in series, according to which the dis-
placements of Ns rows of panels are Ns times larger than those of a
single row, while the stiffness of the whole system is reduced by a factor
equal to 1/Ns; (iii) the presence of concentrated openings is accounted
by introducing a reduction factor equal to 0.5 for both stiffness and
strength of the truss elements; (iv) the cladding panels are assumed not
connected side by side.
4.5. Modal analyses
Results of the modal analyses made using the bare-frame models as
well as the models including cladding panels are presented. For what
concerns the systems without panels, the first four vibration periods are
summarised in Table 5, whereas the corresponding modal shapes
(which are common for all the buildings) are schematically shown in
Fig. 8. It can be observed what follows: the first mode is a local
translational mode in X direction of the central portal frame only, being
such frame more deformable (no roof restraints) and with a higher mass
due to the presence of the crane loads; the second and third modes
(periods very close to each other) also interest the portals’ deformability
along X; the fourth mode is a global longitudinal (along Y) mode. The
vibration periods for the buildings with cladding panels are summarised
in Table 6, for panel typologies A and C. The corresponding modal
shapes (which are common for both panel typologies) are shown in
Fig. 9. It can be observed that the cladding panels notably reduce the
vibration periods and slightly modify the modal shapes. Furthermore,
the elastic stiffening action exerted by panels type C is higher than that
of panels type A.
4.6. Pushover analyses
A preliminary investigation on the influence in the lateral response
of the cladding and roof panels is carried out by performing nonlinear
static (pushover) analysis. As shown in Fig. 10, the control nodes belong
to the extreme portal frame and they are chosen at different heights for
the two orthogonal directions: the top of the column in X direction, and
the crane level in Y direction, the latter corresponding to the head of the
panels along Y. The analyses are performed in control of displacement,
by pushing the building in each direction by means of sets of horizontal
forces applied at the nodes of the roof level. Selected results are pre-
sented for case study X20Y8 (Table 4) only, nevertheless, very similar
results are observed for the other considered geometries. Thus, the
following comments apply to all the considered case studies.
Fig. 11 compares the capacity curves related to the bare frame
structure (black bold line) and those for the systems with infill panels
type A (red dashed line) and type C (blue dotted line). It is noted that
the contribution provided by the panels consists of a slight elastic
stiffness increment and a rise of the maximum shear capacity in both
the X- and Y-directions. The rise in resistance is particularly significant
in the longitudinal direction (Y), mainly because of the higher number
of panels contributing to withstand the lateral loads.
With regard to the bare system response, it is worth noticing that,
being the central portal frame not restrained by the roof braces (only
located at the outer bays, as shown in Fig. 3) it experiences differential
displacements with respect to the lateral portal frames, by inducing a
state of tension on the longitudinal purlins and beams (catenary effect).
The lack of a proper rigid diaphragm at the roof level makes the re-
sistance mechanism along the X direction particularly sensitive to the
second order effects. When the cladding panels are instead included,
because of their location within the extreme portal frames only, a fur-
ther amplification of the aforesaid imbalance between the portal lateral
displacements is observed and the response under lateral loads changes
as witnessed by the modified trends of the internal stresses for the
beams of the extreme portal frames, reported in Fig. 12 in the cases
without panels and with panels types A and C. It can be observed that
F. Scozzese et al. Soil Dynamics and Earthquake Engineering 114 (2018) 505–519

Mode 1 Mode 2 Mode 3 Mode 4

Fig. 8. Modal analysis: first four modal shapes for bare buildings.

Table 6 Fig. 14 refers to the building X20Y8 equipped with panels type A,
First four vibration periods for the models with panel type A and panel type C. whereas Fig. 15 concerns the building X20Y8 with panels type C. It can
T1 [s] T2 [s] T3 [s] T4 [s]
be observed (Fig. 14), that in the portal frame (X-direction) the con-
sidered panels’ damage states (i.e., by using the acronyms introduced in
Case study A C A C A C A C paragraph 4.8: DLS1 and the attainment of the peak strength shear for
both panels type A and C, and also DLS2 for panels type C only) always
X20Y8 0.45 0.36 0.29 0.23 0.27 0.21 0.24 0.19
X20Y6 0.38 0.31 0.27 0.23 0.24 0.18 0.19 0.17
occur before the yielding of the beams. Conversely, along the long-
X30Y8 0.55 0.43 0.35 0.30 0.34 0.26 0.26 0.24 itudinal direction (Fig. 15) an earlier activation of the nonlinear re-
X30Y6 0.50 0.39 0.35 0.31 0.30 0.23 0.23 0.22 sponse of the concentric braces can be observed: cladding panels start
damaging after the yielding of the tensioned braces and the buckling of
the compressed braces, with the only exception of the DLS1 condition
the flexural demand on the beam is initially the same for all cases (with on panels type C which is attained slightly before the braces’ yielding
and without panels); afterwards, the response with panels deviates from (almost concurrently with the buckling of compressed braces). It is
the reference one (bare structure) and the system with panels (both worth noticing, that the damage states plotted in the figures above are
type A and C) shows an inverted trend in terms of out-of-plane bending attained by all the cladding panels simultaneously, i.e., the 100% of
moment (Fig. 12a), while the in-plane bending moment (Fig. 12b) panels in each direction experiences that specific damage level for a
follows the same path as the reference case but without the softening given displacement dPC of the control node.
branch, and this implies a reduction of the second order effects on the
portal frames due to the presence of the panels. In terms of axial force
on the portal beams (Fig. 12c), the differences are visible since the early 4.7. Multiple-stripe analyses
stages of the analysis: both the A and C panels produce an increment of
the axial loads on the beams, but such contribution is higher with pa- In order to evaluate the dynamic response of the structure under
nels type C, being the latter stiffer and more resistant than the other increasing seismic intensity, multiple-stripe analysis (MSA) [24] is
typology. Furthermore, the presence of the roof panels leads to a new adopted. The seismic hazard characterization proposed within the fra-
strut-and-tie system involving the whole structure, middle portal frames mework of the RELUIS-RINTC project [25] funded by the Italian Civil
included. Protection Department, is adopted. The seismic hazard curves for L’A-
The combination of the behavioural aspects illustrated above is quila city are discretized in 10 intensity levels corresponding to the
responsible for the changes in the pushover response in X-direction, following return periods (expressed in years): 10, 50, 100, 250, 500,
despite the symmetrical location of the panels. The aforementioned 1000, 2500, 5000, 10000, 100000. The Intensity Measure (IM) is re-
phenomenon is not observed along the Y-direction, as shown in Fig. 13 presented by the spectral pseudo-acceleration Sa(T1) at the system
in terms of axial stress on the concentric lateral braces. In this case the period T1. The most of the ground motion records used to perform
panels do not modify the load path within the structural elements but nonlinear dynamic analysis are extracted from the Italian accelero-
simply furnish a parallel contribution in terms of both stiffness and metric database [26], and just in few cases the records from the
strength, hence resulting in a global contribution on the capacity curves NGAwest2 database [27] are selected in order to overcome the scarcity
without locally affecting the elements’ response. The onset of the of Italian records with compatible spectra. Further details on the
buckling on the compressed braces is not influenced either (Fig. 13b). adopted seismic hazard representation can be found in [25]. MSA
Pushover analyses can be conveniently used to investigate the da- analyses are carried out by using a set of 20 pairs (two horizontal
mage state of the structural components when different damage levels seismic components) of ground motion time-histories for each of the 10
are attained by the cladding panels, as reported in Fig. 1 and Fig. 2 IMs. Each accelerogram is scaled at the current IM level in order to
where the response of the following structural elements is monitored: achieve the target spectrum compatibility in terms of pseudo-accel-
beams of the portal frames along the X-direction (Fig. 14a and Fig. 15a), eration ordinates. It is worth noticing that the accurate selection of the
and concentric braces along the Y-direction (Fig. 14b and Fig. 15b). records allowed to minimise the scale factors required to ensure the

Mode 1 Mode 2 Mode 3 Mode 4

Fig. 9. Modal analysis: first four modal shapes for buildings with cladding panels.

511
F. Scozzese et al.
Fig. 10. Location of the X and Y pushover control nodes: (a) frontal view; (b) roof view; (c) lateral view.
ground motions spectrum compatibility, as can be seen in Table 7, in
which both the mean (SFmean) and maximum (SFmax) values of the scale
factors of each IM level are summarised.
The results obtained from MSA are processed in terms on fragility
curves and presented in details in the following paragraph.
4.8. Fragility curves
This section aims at estimating the probability of exceeding the
performance level corresponding to DLS at different intensity levels.
This task is achieved through the application of the local damage cri-
teria introduced in Section 2 and discussed in Table 1, according to
which the attainment of the DLS is defined by monitoring the onset of
the damage levels of the cladding panels. The aforementioned criteria
apply to both the cases with panel type A and panel type C. However,
for the cases with panels type C only, given the early first yielding
followed by minor stiffness reduction shown by this cladding element
(the first yielding of a single panel type C is observed at only 5 mm of
horizontal displacement), an additional performance condition, i.e. first
point of the perfectly plastic range of deformation, is used for the low-
and medium-damage conditions. Accordingly, yielding on panels type A
as well as first yielding on panels type C is indicated with DLS1 while
DLS2 denotes the first point of the perfectly plastic range of deforma-
tion for panels type C.
Based on MSA obtained results, a set of fragility functions P[D|IM]
is computed, each expressing the probability of exceeding the different
damage states described above (D) conditional on the considered levels
of IM. The values of P[D|IM] are first obtained via empirical method,
i.e. counting the fraction of the damaged panels at each stripe of IM,
and afterwards a set of cumulative distribution functions are fitted to
the aforesaid empirical values by using a maximum likelihood-based
linear regression procedure exploiting the Probit link function [28], as
also proposed in Baker (2015) [29].
Rather than monitoring the state of each single panel, the damage
(a)
Fig. 11. Pushover analysis of case study X20Y8 with and without panels: capacity curves in the (a) X-direction and (b) Y-direction.
512
Soil Dynamics and Earthquake Engineering 114 (2018) 505–519
conditions are evaluated by monitoring the panel assemblies composing
the different structural fields, each of which is accounted in the finite
element model by using a couple of equivalent nonlinear truss ele-
ments, as previously described. Given that the yielding threshold is
never attained in the roof panels, only the vertical panels are considered
in the non-structural damage evaluation. In this way, a total of ten
panel assemblies are present: two in the transverse direction and eight
along the longitudinal direction.
A comparison between the fragility curves obtained using the local
damage criteria and the fragility curves derived from the conventional
drift-based criteria using the bare-frame models is made from
Figs. 16–19. For the local damage criteria, each of the four case studies
is analysed twice, the first time supposing cladding panels type A and
the second time supposing cladding panels type C. It is worth noticing
that the low-damage condition with 30% panels damaged is not re-
ported in X-direction because of the presence of just two panel assem-
blies, which only allows to assess the condition with 50% (one panel
out of two) or 100% (two panels) damaged. According to the trends of
the fragility curves reported, it is possible to observe what follows.
For what concerns the case studies assumed as having panels type A:
• the probability of exceeding the DLS performance level as measured
by the conventional drift-based criteria is generally higher than the
probability measured by the adopted medium-damage local criteria,
for any IM level and for both the X- and Y-directions;
• in X-direction, because of the very limited number of panel assem-
blies, the low-damage (≤50% panels at DLS1) and the medium-
damage (100% panels at DLS1) conditions are equivalent and lead
to the same (or very similar) fragility functions;
• in Y-direction, the fragility functions related to the conditions 50%
and 100% panels at DLS1 are very similar to each other, while the
curves for the 30% panels at DLS1 are significantly more probable
than the previous ones and approach the drift-based fragility curves
(the low-damage curve approximates the drift-based one reasonably
(b)
F. Scozzese et al. Soil Dynamics and Earthquake Engineering 114 (2018) 505–519

(a) (b)

(c)
Fig. 12. Pushover analysis in the X-direction of case study X20Y8 with and without panels: evolution in the eaves beam in the extreme portal frame of (a) out-of-
plane bending moment; (b) in-plane bending moment; (c) axial force.

well for the case studies X20Y8 and X30Y6).
For what concerns the case studies assumed as having panels type C:
• the probability of exceeding the DLS1 by using the local-damage
criteria is always notably higher than that corresponding to the
conventional drift-based approach on the bare systems, and this is
true for any IM level and for both the X- and Y-directions;
• conversely, the probability of exceeding the DLS2 by using the local-
damage criteria is always lower than the one corresponding to the
conventional drift-based criteria, and the trends of the fragility
functions, in this case, are very close to those observed with panels
type A at DLS1;
• in X-direction, because of the very limited number of panel

(a) (b)
Fig. 13. Pushover analysis in the Y-direction of case study X20Y8 with and without panels: axial force evolution in the a) tensed diagonal; b) compressed (buckled)
diagonal.

513
F. Scozzese et al. Soil Dynamics and Earthquake Engineering 114 (2018) 505–519

(a) (b)
Fig. 14. Evolution of the portal beams response (in-plane bending moment) under nonlinear static analysis and comparison with the attainment of the cladding
panels damage states. Case study X20Y8 with panels type A (a) and type C (b).

(a) (b)
Fig. 15. Evolution of the concentric braces response (axial force) under nonlinear static analysis and comparison with the attainment of the cladding panels damage
states. Case study X20Y8 with panels type A (a) and type C (b).

Table 7 DLS1 performance threshold conditional on IM= 2 (i.e., P[DLS1|

Mean and maximum scale factors applied on recorded ground motions to per- IM=2]) are always lower than 10%. Furthermore, the aforesaid con-
form MSA. ditional probabilities (P[DLS1| IM=2]) are always null when evaluated
IM 1 2 3 4 5 6 7 8 9 10 by means of the conventional drift-based criteria. It can therefore be
concluded that the code prescription on the structural behaviour at the
SFmean 4.0 4.0 3.0 2.4 2.2 2.6 2.8 3.3 4.1 5.6 serviceability limit state are well satisfied, even when the contribution
SFmax 6.6 6.6 6.3 6.7 6.8 6.8 9.3 9.0 11.3 9.9
of non-structural cladding panels is explicitly accounted for in the
analysis.
One more aspect which deserves to be highlighted consists of the
assemblies, the low-damage (≤50% panels at DLS1 or DLS2) and
different sensitivity to the onset of the serviceability limit state shown
the medium-damage (100% panels at DLS1 or DLS2) conditions are
by the two structural directions (X and Y). In fact, according to the drift-
equivalent and lead to the same (or very similar) fragility functions,
based criteria, the seismic intensities at which the DLS starts to be
as also observed before for panels type A;
• in Y-direction, the fragility functions related to the conditions 50%
probable are those corresponding to IM 4–5 (seismic events occurring in
average every 250–500 years) along X-direction, with a non-negligible
and 100% panels at DLS1 (or DLS2) are very similar to each other,
rise in probability of exceedance at IM 6 (1000 years return period).
while the curves for the 30% panels at DLS1 (or DLS2) are sig-
Along Y-direction, because of the stiffer seismic resistant mechanism
nificantly more probable than the previous ones.
available (with concentric lateral braces), the triggering seismic in-
tensities for the DLS evaluated via drift-based approach are all shifted
It is worth noticing that, in all cases, the probability of attaining the
toward IM 6. On the other hand, adopting the local-damage criteria, the
DLS is null for IM ≤ 2, i.e., for seismic intensities lower or equal to the
probability of exceeding the DLS conditional on IM is strongly influ-
reference one (with 50 years return period) provided by the seismic
enced by the panel typology and the choice of the performance con-
code [15] and used for the design of the present case studies at the
dition (DLS1 or DLS2): assuming the DLS1 with panels type C the onset
serviceability limit state. The only exceptions are represented by the
of the limit state is strongly anticipated, with non-negligible prob-
systems X30Y6 (along Y-direction) and X30Y8 (along X-direction) with
abilities of exceedance (higher than 30%) starting from IM 3–4
panels type C, for which, however, the probabilities of exceeding the
(100–250 years return period) in X-direction, and from IM 5–6

514
F. Scozzese et al.
A-type panels – X-direction
C-type panels – X-direction
C-type panels – X-direction
Fig. 16. Fragility curves for case study X20Y6.
(500–1000 years return period) in Y-direction; assuming the low-da-
mage condition with ≤ 50% panels at DLS1 or DLS2, for respectively
panels type A and C, the probabilities of exceeding the limit state be-
come higher than 20–30% starting from IM 7 in X-direction, and from
IM 8 (5000 years return period) in Y-direction.
Based on the specific case studies examined in this work and in light
of the results discussed above, it appears that the conventional drift-
based criteria might not catch the actual damage state experienced by
515
Soil Dynamics and Earthquake Engineering 114 (2018) 505–519
A-type panels – Y-direction
C-type panels – Y-direction
C-type panels – Y-direction
the non-structural cladding panels, hence leading to the overestimation
of the DLS|IM exceedance probability for a wide range of IM levels.
It is necessary to comment that the strong-damage condition (at
least one panel exceeding the maximum shear resistance) foreseen by
the proposed local damage criteria is not predominant within the in-
vestigated case studies and for this reason the corresponding fragility
curves are not reported in the charts, being limited to the higher IM
levels only, i.e. IM 9–10.
F. Scozzese et al.
A-type panels – X-direction
C-type panels – X-direction
C-type panels – X-direction
Fig. 17. Fragility curves for case study X20Y8.
5. Conclusions
In this work the seismic response of single-story industrial steel
buildings was studied with a structural model that includes the non-
linear behaviour of cladding panels in addition to the nonlinear beha-
viour of the steel bare-frame (beams, columns, braces). This is achieved
by adding to the bare-frame model equivalent trusses with nonlinear
hysteretic behaviour describing the cladding panels and calibrated
using experimental test results available in the literature. In this way it
becomes possible pursuing the primary objective of this work: predict
516
Soil Dynamics and Earthquake Engineering 114 (2018) 505–519
A-type panels – Y-direction
C-type panels – Y-direction
C-type panels – Y-direction
the non-structural damage in terms of intensity and extent, based on the
cladding panel performance. In addition, the results include the influ-
ence that the non-structural components have on the local and global
seismic response. Three levels of damage, combining quality and
quantity, are considered: (i) low-damage condition when an assigned
percentage of panels exceed yielding (30% and 50% values were con-
sidered in this study); (ii) medium-damage condition when 100% of
panels exceed yielding; (iii) strong-damage condition when at least one
panel exceeds its maximum shear resistance. The application to a set of
four single-storey industrial steel buildings, designed according to the
F. Scozzese et al.
A-type panels – X-direction
C-type panels – X-direction
C-type panels – X-direction
Fig. 18. Fragility curves for case study X30Y6.
Eurocodes using bare-frame models, show that the low- and medium-
damage conditions have relations with conventional checks based on
drift as derived from bare-frame models, while the strong-damage
condition is achieved only for earthquake intensities relevant to very
long return periods. The results obtained with the adopted approach
including non-structural elements allows a number of benefits over
conventional drift-based check at a minor additional computational
cost, i.e. clear information on the prominence of the damage and its
distribution, inclusion in the results of the effect of the interactions
between structural elements and cladding panels, possibility to assess
517
Soil Dynamics and Earthquake Engineering 114 (2018) 505–519
A-type panels – Y-direction
C-type panels – Y-direction
C-type panels – Y-direction
the expected performance of different cladding panels as a support to
the design of the optimal solution related to seismic performance and
costs (cladding panels initial costs and cost that might be sustained after
an earthquake). Even though the presented approach is rather simple to
be integrated in bare-frame models of steel structures, a major limita-
tion currently lies with the lack of experimental tests on both panels
and integrated systems (panels as mounted on the steel frame), data
that are indispensable to properly calibrate the hysteretic behaviour of
the non-structural elements. Accordingly, it is desirable that tests under
cyclic loading histories on claddings and integrated systems will be
F. Scozzese et al.
A-type panels – X-direction
C-type panels – X-direction
C-type panels – X-direction
Fig. 19. Fragility curves for case study X30Y8.
made available in the future so that more accurate evaluations of the
damage limit state can be made by structural engineers.
Acknowledgments
This study was developed within the RELUIS-RINTC 2015–2018
research project funded by the Italian Department of Civil Protection
and coordinated by Professor Iunio Iervolino, University of Naples
Federico II, and Professor Paolo Bazzurro, Istituto Universitario Studi
518
Soil Dynamics and Earthquake Engineering 114 (2018) 505–519
A-type panels – Y-direction
C-type panels – Y-direction
C-type panels – Y-direction
Superiori, Pavia, Italy. The presented contents benefited from con-
structive discussions during the plenary meetings as well as from the
meetings on the damage limit state supervised by Professor Donatello
Cardone, University of Basilicata, Italy. The above cited support and
contributions are gratefully acknowledged.
References
[1] Ballio G, Mazzolani FM. Theory and design of steel structures. London: Chapman
F. Scozzese et al.
and Hall ltd; 1983.
[2] Landolfo R, Mazzolani FM, Dubina D, Da Silva LS, D’Aniello M. Design of steel
structures for buildings in seismic areas. ECCS Eurocode Design Manuals. Berlin:
Ernst & Sohn; 2015.
[3] De Matteis G, Landolfo L, Mazzolani FM. Diaphragm effect for industrial steel
buildings under earthquake loading. J Constr Steel Res 1998;46(1–3):357–8.
https://doi.org/10.1016/S0143-974X(98)00071-6.
[4] De Matteis G, Landolfo R. Structural behaviour of sandwich panel shear walls: an
experimental analysis. Mater Struct 1999;32(5):331–41. https://doi.org/10.1007/
BF02479624.
[5] De Matteis G, Landolfo R. Modelling of lightweight sandwich shear diaphragms for
dynamic analyses. J Constr Steel Res 2000;53(1):33–61. https://doi.org/10.1016/
S0143-974X(99)00038-3.
[6] De Matteis G. Effect of lightweight cladding panels on the seismic performance of
moment resisting steel frames. Eng Struct 2005;27(11):1662–76. https://doi.org/
10.1016/j.engstruct.2005.06.004.
[7] Rogers CA, Tremblay R. Inelastic seismic response of frame fasteners for steel roof
deck diaphragms. J Struct Eng 2003;129(12):1647–57. https://doi.org/10.1061/
(ASCE)0733-9445(2003)129:12(1647).
[8] Rogers CA, Tremblay R. Inelastic seismic response of side lap fasteners for steel roof
deck diaphragms. J Struct Eng 2003;129(12):1637–46. https://doi.org/10.1061/
(ASCE)0733-9445(2003)129:12(1637).
[9] Essa HS, Tremblay R, Rogers CA. Behavior of roof deck diaphragms under quasi-
static cyclic loading. J Struct Eng 2003;129(12):1658–66. https://doi.org/10.1061/
(ASCE)0733-9445(2003)129:12(1658).
[10] Mastrogiuseppe S, Rogers CA, Tremblay R, Nedisan CD. Influence of nonstructural
components on roof diaphragm stiffness and fundamental periods of single-storey
steel buildings. J Constr Steel Res 2008;64(2):214–27. https://doi.org/10.1016/j.
jcsr.2007.06.003.
[11] Rogers CA, Tremblay R. Impact of diaphragm behavior on the seismic design of low-
rise steel buildings. Eng J 2010;47(1):21–36.
[12] Okazaki T, Nakashima M, Suita K, Matusmiya T. Interaction between cladding and
structural frame observed in a full-scale steel building test. Earthq Eng Struct Dyn
2007;36(1):35–53. https://doi.org/10.1002/eqe.618.
[13] Ma J, Nakata S, Yoshida A, Tamura Y. Effects of precast cladding systems on dy-
namic characteristics of steel frame buildings by ambient and free vibration tests.
[Article ID5231727]. Shock Vib2017. https://doi.org/10.1155/2017/5231727.
[14] European Committee for Standardization. Eurocode 3: Design of steel structures -
Part 1-1: General rules and rules for buildings. EN1993-1-1; August 2005.
[15] European Committee for Standardization, Eurocode 8: Design of structures for
earthquake resistance - Part 1: General rules, seismic actions and rules for buildings.
519
Soil Dynamics and Earthquake Engineering 114 (2018) 505–519
EN 1998-1; December 2004.
[16] Braga F, Gigliotti R, Monti G, Morelli F, Nuti C, Vanzi I, Salvatore W. Post-seismic
assessment of existing constructions: evaluation of the shakemaps for identifying
exclusion zones in Emilia. Earthq Struct 2014;8(1):37–56. https://doi.org/10.
12989/eas.2015.8.1.037.
[17] Braga F, Gigliotti R, Monti G, Morelli F, Nuti C, Vanzi I, Salvatore W. Speedup of
post earthquake community recovery: the case of precast industrial buildings after
the Emilia 2012 earthquake. Bull Earthq Eng 2014;12:2405–18. https://doi.org/10.
1007/s10518-014-9583-3.
[18] Freddi F, Tubaldi E, Ragni L, Dall’Asta A. Probabilistic performance assessment of
low-ductility reinforced concrete frames retrofitted with dissipative braces. Earthq
Eng Struct Dyn 2013;42(7):993–1011. https://doi.org/10.1002/eqe.2255.
[19] Cardone D, Perrone G. Developing fragility curves and loss functions for masonry
infill walls. Earthq Struct 2015;9(1):257–79. https://doi.org/10.12989/eas.2015.9.
1.257.
[20] Midas GEN. Integrated solution system for building and general structures v2.1.
Midas Information Technology Co. Ltd; 2016.
[21] McKenna F. OpenSees: a framework for earthquake engineering simulation. Comput
Sci Eng 2011;13(4):58–66. https://doi.org/10.1109/MCSE.2011.66.
[22] Hsiao PC, Lehman DE, Roeder CW. Improved analytical model for special con-
centrically braced frames. J Constr Steel Res 2012;73(1):80–94. https://doi.org/10.
1016/j.jcsr.2012.01.010.
[23] Hsiao PC, Lehman DE, Roeder CW. Evaluation of the response modification coef-
ficient and collapse potential of special concentrically braced frames. Earthq Eng
Struct Dyn 2013;42(10):1547–64. https://doi.org/10.1002/eqe.2286.
[24] Jalayer F, Cornell CA. Alternative non-linear demand estimation methods for
probability-based seismic assessments. Earthq Eng Struct Dyn 2009;38(8):951–72.
[25] Iervolino I, Spillatura A, Bazzurro P. RINTC Project - Assessing the (implicit) seismic
risk of code-conforming structures in Italy. COMPDYN 2017, Papadrakakis M.,
Fragiadakis M. (editors) Rhodes Island, Greece, 15-17 June; 2017.
[26] Luzi L, Hailemikael S, Bindi D, Pacor F, Mele F, Sabetta F. ITACA (ITalian
ACcelerometric Archive): a web portal for the dissemination of Italian strong-mo-
tion data. Seismol Res Lett 2008;79(5):716–22. https://doi.org/10.1785/gssrl.79.5.
716.
[27] Ancheta TD, Darragh RB, Stewart JP, Seyhan E, Silva WJ, Chiou BSJ, Wooddell KE,
Graves RW, Kottke AR, Boore DM, Kishida T, Donahue JL. NGA-West2 database.
Earthq Spectra 2014;30(3):989–1005. https://doi.org/10.1193/070913EQS.197M.
[28] Agresti A. Categorical data analysis. John Wiley & Sons; 2013.
[29] Baker JW. Efficient analytical fragility function fitting using dynamic structural
analysis. Earthq Spectra 2015;31(1):579–99.