Journal of Constructional Steel Research 64 (2008) 896913

www.elsevier.com/locate/jcsr
Behavior and performance of cold-formed steel-framed houses under
seismic action
Dan Dubina
Department of Steel Structures and Structural Mechanics, Politehnica University of Timisoara, 1, Ioan Curea, Timisoara 300224, Romania
Received 28 February 2007; accepted 18 January 2008
Abstract
In recent years important research activity has been undertaken in order to evaluate the earthquake performance of light-gauge steel-
framed house structures. Almost all studies approach the problem of seismic response of these structures by characterising, experimentally and
numerically, the performance of wall panels. Usually, the overall behavior of wall panels is mainly addressed. However, according to experimental
evidences, the performance of the wall panels, as a whole, is governed by the performance of the connectors e.g.: sheeting-to-sheeting connectors,
and sheeting-to-framing connectors. On the other hand the global behavior of the 3D structure of the house is signicantly inuenced by non-
structural elements, traditionally not considered in the design procedures. The present paper summarizes the research activity carried out in the
last few years at the Politehnica University of Timisoara, under the coordination of the author, with the aim to evaluate the performance and to
characterise for design purpose the specic features of these structures. Monotonic and cyclic tests on full-scale shear panels, tests on connection
details, and in situ ambiental vibration tests on a house under construction are reviewed and concluded in the following paper.
c 2008 Elsevier Ltd. All rights reserved.
Keywords: Light-gauge steel houses; Seismic action; Performance-based design
1. Introduction
Steel-framed houses are usually built of light thin-walled
load bearing structures having different solutions for interior
and exterior cladding. This technology is popular and accounts
for an important and increasing market share in US, Japan,
Australia and Europe [1]. The same method is used for
buildings, of small dimensions, of other purposes (ofces,
schools, manufacturing premises, etc.), that are referred to as
small industrial buildings (SIB).
In such structures shear walls are the main structural
elements which act against horizontal loads, e.g. wind and
earthquake. However, the behavior of shear walls subjected to
earthquake is not yet fully understood and, in recent years, an
important effort has been made to clarify certain aspects related
to their shear strength, stiffness and ductility.
Research in USA [25] has been focused mainly towards
experimental testing of shear walls in order to produce practical
racking load values. Load bearing capacities were derived
both from monotonic pushover curves, envelope and stabilized
E-mail address: [email protected].
envelope curves from cyclic tests. Findings of these studies
suggest a conventional elastic stiffness for a wall panel at
0.4 of the ultimate load [6]. Different frame typologies with
various cladding materials were tested, studies were conducted
to determine the inuence of length/height ratios as well as
the effect of openings. Even most of the very detailed studies
avoid addressing an important aspect of shear wall behavior,
e.g. the energy dissipation capacity due to cyclic characteristics.
The effect of gypsum wallboard was also studied, leading to
the conclusion that both strength and stiffness are increased by
the presence of gypsum wallboard, some results suggesting an
increase in terms of ultimate load of up to 30%, compared to
the case of external sheeting only.
Testing and numerical simulation have been combined in
order to account for hysteretic characteristics in an attempt
to provide evidence on the possible values of response
modication factors (q) [7]. Vibration tests of steel-framed
houses were conducted and relatively large damping ratios
were found due to interior and exterior nishes. According
to the tests damping ratio of 6% was accepted for seismic
analysis. A maximum of 1/50 rad story drift angle limit is also
suggested as acceptable during severe earthquakes. In the FE
0143-974X/$ - see front matter c 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2008.01.029
D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913 897
analysis stage, a steel-framed house was subjected to two levels
of seismic waves. The house exhibited good performance,
reaching a maximum drift of 1/300 rad. Even when minimum
required wall length was provided, the maximum drift did not
exceed 1/60 rad.
On the same line, Gad et al. [8,9] proposed a new analytical
approach to evaluate the ductility parameter (R

), for which
they suggest a value between 1.5 and 3.0 to be suitable. Their
research briey assesses inherent structural over-strength and
nds it to be a very important factor as far as earthquake
resistance is concerned. The quantitative evaluation of over-
strength is more difcult, and an empirical evaluation was
performed.
Serrette and co-authors (Serrette, 2006) [13] studied
experimentally, again, cold-formed wall stud panels with OSB
focusing the application of adhesive to bond sheeting to
framing.
Experimental tests and FE modeling was employed by De
Matteis [10] to assess shear behavior of sandwich panels both
in single story and multi-story buildings. A number of six
monotonic and six cyclic tests were performed on full-scale
sandwich panel specimens of different congurations. In the
nal stage of the study, dynamic modeling on panels integrated
in building structures, under real earthquake records was
performed. According to the conclusions diaphragm action can
replace classical bracing solutions only in low-rise buildings,
and in areas of low seismicity.
Both experimental and numerical studies on wall studs shear
panels sheeted with OSB has been realised at the University
Federico II in Naples by Landolfo et al. [11,12]. The authors
calibrated a numerical model based on Richard & Abbot
formulation, adopted to account for the pinching effect, which
usually characterises the cyclic response of such a type of panel.
One the largest experimental programs carried out in Europe
on light-gauge steel wall panels, aiming to characterise their
cyclic response has been undertaken at the Politehnica
University of Timisoara. Both wall panels sheeted with
corrugated steel sheeting and OSB has been tested [14]. On the
basis of experimental results a simplied numerical model has
been proposed by authors with the aim to obtain a 3D analysis
of house framing [15]. Later, this research was completed with
tests on seam and sheeting-to-framing connections [16].
Recently, in situ ambiental vibrations tests on a real house
structure under construction, have been performed by Dubina
et al. [17] in order to observe the overall response of the
building in subsequent steps of erection.
The main results of this extensive investigation are
summarized in the present paper.
2. Experimental investigation of wall stud shear walls
2.1. Description of experimental program [14]
The experimental program was based on six series of full-
scale wall tests with different cladding arrangements based on
common practical solutions in housing and SIB (Table 1).
Fig. 1. Experimental arrangement.
Each series consisted of 3600 mm 2440 mm identical
wall panels tested statically, both monotonic and cyclic. The
main frame of the wall panels were made of cold-formed steel
elements, top and bottom tracks were U154/1.5, while studs
were C150/1.5 proles, xed at each end to tracks with two
pair SPEDEC SL4-F-4.8 16 (d = 4.8 mm) self-drilling
self-taping screws. In specimens using corrugated sheet as
cladding the sheets were placed in a horizontal position with
a useful width of 1035 mm and one corrugation overlapping
and tightened with seam fasteners SL2-T-A14-4.8 20 (d =
4.8 mm) at 200 mm intervals (Series I and II). Corrugated
sheet was xed to the wall frame using SD3-T15-4.8-22
(d = 4.8 mm) self-tapping screws, sheet ends being xed in
every corrugation, while on intermediate studs at every second
corrugation. Additionally on the interior side of specimens in
Series II, 12.5 mm thick gypsum panels (1200 2440 mm)
were placed vertically and xed at 250 mm intervals on each
vertical stud.
Bracing was used in Series III specimens, by means of
110 1.5 mm straps on both sides of the frame. Steel straps
were xed to the wall structure using SPEDEC SL4-F-4.8 16
(d = 4.8 mm) and SD6-T16-6.3 25 (d = 6.3 mm) self-
drilling screws, the number of screws being determined to avoid
failure at strap end xings and facilitate yielding.
10 mmOSBpanels (12002440 mm) were placed in similar
way as the gypsum panels in earlier specimens (Series OSB I
and OSB II), only on the external side of the panel and xed to
the frame using bugle head self-drilling screws of d = 4.2 mm
diameter at 105 mm intervals.
The full-scale testing program was completed with tensile
tests to determine both material properties for components and
behavior of connections.
Experiments, under horizontal monotonic and cyclic
loading, were conducted using displacement control, at the
same time measuring the corresponding load with load cell
(Fig. 1).
A monotonic test using a loading velocity of 1 cm/min
was performed for each type of panel. Based on the results,
initial stiffness (K
o
) and conventional elastic limit (
el
) was
determined using the methodology presented in Fig. 2.
The conventional elastic limit displacement (
el
) was used
to determine the displacement amplitudes for the cyclic tests.
898 D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913
Table 1
Description of wall specimens
Series Opening Bracing Exterior cladding Interior cladding Testing method Loading velocity No. test
O Monotonic 1 1
I Corrugated Sheet LTP20/0.5
Monotonic 1 1
Cyclic 63 2
II Corrugated Sheet LTP20/0.5 Gypsum Board
Monotonic 1 1
Cyclic 63 2
III
Monotonic 1 1
Cyclic 3 1
IV Door Corrugated Sheet LTP20/0.5
Monotonic 1 1
Cyclic 63 2
OSB I 10 mm OSB
Monotonic 1 1
Cyclic 3 1
OSB II Door 10 mm OSB
Monotonic 1 1
Cyclic 3 1
Total number of specimens 15
Loading velocity in cm/min for monotonically and min/cycles for cyclically tested specimens.
Fig. 2. Determination of conventional elastic limit
el
.
Cyclic testing methodology followed ECCS Recommenda-
tion [11]. Loading velocity for the cyclic experiments was
6 min/cycle for one specimen and 3 min/cycle for the second.
The main outputs of the experiments were shear-force versus
horizontal displacement at the top of the wall-specimen curves.
As in the case of the panels clad with corrugated sheet the
seams govern the failure, relative slip between two steel sheets
was also recorded. Load versus lateral displacement curves are
presented in Fig. 3 and, in order to illustrate the differences of
monotonic to cyclic results, stabilized envelope curves are also
presented for the cyclic experiments. As resulting characteristic
curves are very similar to cyclically tested specimens with
loading velocity 6 and 3 min/cycles, in Fig. 3 only one of the
two is reproduced.
Qualitatively observing comparative monotonic to cyclic
curves, a reduction of strength of about 10% can be identied in
the case of cyclic loading. Hence, if only monotonic response
is considered for an analysis (e.g. pushover analysis), the
performance of the panel will be overestimated.
2.2. Analysis of experimental results
As seen wall panels exhibited very complex, and highly
nonlinear behavior. In order to evaluate specic properties
like the elastic modulus, ultimate force or ductility, curves
have been interpreted according to the following established
procedures:
Method IInitial stiffness may be determined as secant
stiffness to the load level of 0.4F
max
. The evaluation
of the conventional yield limit was based on ECCS
Recommendation [18], at the intersection point of the elastic
line (K
o
) to a line of 0.1K
o
rigidity, tangent to the experimental
curve. Based on this conventional elastic limit (e
l
, F
int
)
the ultimate point (F
u
, D
u
) results at the intersection of
the horizontal yield line to the experimental curve in the
downloading branch (Fig. 4a).
Method II The second method has been adopted by Kawai
et al. [7]. Initial stiffness is deed as secant stiffness to the point
of drift angle corresponding to 1/400 (D
400
), while the yield
line is chosen in a way that the hatched parts in Fig. 4b have the
same area. The allowable strength is referred as the minimum
of the force at story drift angle 1/300 (F
300
) and 2/3F
max
.
The two methods, usually, yield similar results, with some
interesting peculiarities. Initial rigidity values are very similar
and it is important to realise that, ultimate load (F
u
) and
ductility are in direct relationship so if a method yields higher
ultimate load (F
u
) this automatically means lower ductility. For
dening design capacity, the minimum of 2/3F
max
and F
300
or
F
200
are relevant [20].
In the present paper only the results corresponding to
Method II (Table 2) are reported (in Ref. [14] the reader can
nd the results for both methods).
D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913 899
Fig. 3. Experimental curves [14].
Fig. 4. Methods for determining equivalent elasto-plastic model [14].
Differences between monotonic and cyclic values can be
observed as follows. Initial rigidity is not affected, values of
cyclic and monotonic tests range within a difference of less than
20%. The same can be noted for ductility, exception being in the
case of OSBspecimens where ductility is reduced by 10%25%
for cyclic results. One important observation concerns ultimate
load (F
u
), where cyclic results are lower than monotonic ones
by 5%10% even if we consider unstabilized envelope curve. If
we take into account stabilized envelope curves, the difference
can increase to 20%30%.
900 D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913
Table 2
Experimental results, Method II
Series-Curve K
o
(N/mm) F
400
(N) F
300
(N) D
curv
(mm) F
u
(N) D
uct
2/3F
max
F
200
(N)
I-1 Mon 4088.1 24 467.2 28 691.0 24.75 47 821.0 4.65 35 250.9 35 110.9
I-2
1(+)
3670.4 21 623.5 25 273.0
24.69 41 488.6 6.57
31 391.1 30 053.7
3(+) 12.65 32 127.5 4.59
1()
3554.5 21 324.3 24 772.3
20.93 38 685.0 4.46
29 374.0 30 367.5
3() 15.95 34 886.6 4.18
I-3
1(+)
3386.8 19 833.3 23 805.0
23.90 40 312.5 5.34
31 573.5 28 743.7
3(+) 16.92 34 589.1 4.36
1()
3174.9 19 363.7 22 495.9
20.55 38 300.2 4.15
29 384.8 28 536.8
3() 14.27 32 637.9 4.35
II-1 Mon 3311.5 20 088.5 24 349.9 40.96 53 801.3 5.03 39 810.3 30 508.6
II-2
1(+)
3766.2 22 654.0 26 872.8
30.66 49 185.6 5.22
38 530.6 33 635.1
3(+) 16.07 38 868.1 5.73
1 ()
4023.7 24 276.4 27 238.0
29.42 48 126.1 5.11
36 713.1 34 898.2
3() 14.75 38 751.3 5.68
II-3
1(+)
3936.3 23 389.8 27 339.5
30.15 50 985.9 5.20
40 022.8 35 284.9
3(+) 16.02 41 920.7 5.62
1()
3676.4 21 297.5 24 812.4
27.81 47 821.4 4.65
37 880.4 32 281.1
3() 15.48 39 734.3 5.14
III-1 Mon 4187.5 25 120.4 31 980.0 19.21 51 139.6 2.81 36 765.1 40 193.0
III-2
1(+)
3626.1 21 398.0 27 286.2
17.13 45 599.5 4.10
35 122.0 35 963.5
3(+) 13.52 40 602.5 3.35
1()
3627.6 21 120.3 26 626.8
25.94 50 427.8 6.41
36 243.3 36 230.7
3() 16.06 43 901.9 6.68
IV-1 Mon 1598.3 9349.6 13 723.8 41.67 35 532.7 3.79 26 813.5 18 048.4
IV-2
1(+)
1837.9 11 032.8 13 490.3
34.43 32 768.5 5.82
23 419.2 18 057.5
3(+) 22.99 26 756.3 6.38
1()
1808.9 10 953.4 13 318.1
34.39 31 849.7 5.68
23 320.7 17 689.1
3() 21.95 26 046.7 6.50
IV-3
1(+)
1616.7 9374.3 11 960.1
40.47 34 730.9 5.25
27 228.7 16 464.5
3(+) 24.89 26 908.7 5.55
1()
1801.9 10 305.0 12 710.8
39.36 33 719.6 5.73
26 684.4 16 897.3
3() 25.28 27 537.2 6.45
OSB I-1 Mon 3909.6 23 797.3 28 470.2 37.22 68 162.0 4.26 52 517.7 37 953.9
OSB I-2
1(+)
4406.2 25 807.2 30 159.0
21.32 55 142.0 4.07
49 810.4 38 457.9
3(+) 16.68 49 368.6 3.73
1()
3988.4 23 482.0 27 725.9
21.44 54 088.4 3.68
43 314.9 35 930.3
3() 18.30 48 520.5 3.61
OSB II-1 Mon 1814.9 10 702.5 13 779.6 36.45 37 014.8 3.19 29 586.5 18 732.5
OSB II-2
1(+)
Na. Na. Na.
Na. Na. Na.
Na. Na.
3(+) Na. Na. Na.
1()
1610.5 9511.3 11 850.5
32.28 37 426.0 2.93
30 539.2 16 495.4
3() 27.76 33 908.3 3.11
1 unstabilized envelope curve; 3 stabilized envelope curve; (+) positive envelope; () negative envelope
Based on comparison of the medium values of monotonic
and cyclic results, the contribution of an opening, gypsum
board and other factors can be assessed with the following
conclusions:
Series ISeries II: Differences can be attributed to the
effect of the gypsum board. There is an increase of 16.2%
and 17.8% respectively in the ultimate load. As far as initial
values are concerned (K
o
, F
400
, F
300
, F
200
) there seem to be
no difference, but ductility is improved slightly.
Series ISeries IV: There is a signicant decrease of initial
rigidity (60.3%53.3%), for a lesser degree of ultimate load
(16.4%21.0%), but ductility values are essentially unaffected
Series ISeries III: Comparison is more qualitative because
of the different sheeting system. There are no differences as far
as initial rigidity is concerned; however an increase of ductility
had been expected. This was not possible as the failure mode for
the strap-braced specimens was not the most advantageous one,
the damage being concentrated entirely in the lower corners of
D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913 901
Fig. 5. Deformation pattern of corrugated sheet (a) Series I and (b) OSB I specimens [15].
the panel. Strap-braced wall panels have the advantage of stable
hysteretic loops, but also the disadvantage of higher pinching
than the sheeted ones.
Series ISeries OSB I: Comparison is more qualitative,
keeping in mind the different wall panel arrangements. Initial
rigidity is of similar magnitude, with increase of ultimate load.
Failure of OSB specimens under cyclic loading was more
sudden than in the case of corrugated sheet specimens where
degradation occurs gradually. This is also reected by the
reduced ductility for OSB specimens.
Series OSB ISeries OSB II: The effect of opening produced
similar results as in the cases of Series ISeries IV. Initial
rigidity decreased 64.6%59.1%, while ultimate load decreased
32.5%36.9%. There is also an important decrease of ductility,
probably highlighting the different failure modes of the two
wall panels.
3. Simplied models for 2D and 3D analysis of wall stud
structures
3.1. Main features of failure mechanisms
During experiments two distinct failure mechanisms were
identied for wall panels sheeted with corrugated sheeting and
OSB. The lateral deformation of a panel is dependent on: (1)
shear deformation of the sheeting material, (2) deformation
due to corner uplift and most signicantly on (3) nonlinear
deformation of the connections between shear panel and
skeleton.
In the case of wall panels with corrugated sheeting placed
horizontally (Fig. 5a) most of the nonlinear deformation was
due to the inelastic deformation of seam fasteners. Seam
connectors will be the ones to deform excessively, later load
being redistributed to the vertical screw lines connecting the
sheeting to the skeleton.
In the case of wall panels sheeted with OSB, as the skeleton
deforms into a parallelogram, the OSB panels have rigid body
rotation (Fig. 5b). As a consequence connections at the corners
of OSB panels will be the ones which have to accommodate the
largest slip and will be damaged.
3.2. Current design methodology
Based on the failure modes, calculation procedures based
on monotonic behavior models exist to predict initial rigidity
and elastic capacity of the panels. In the case of the panels
with corrugated sheeting the method from the European
Recommendations for the Application of Metal Sheeting Acting
as a Diaphragm [21] can conveniently be used keeping in
mind the following observations:
the method refers to behavior in the elastic range only and
hence the results one can obtain are initial rigidity and
elastic load bearing capacity;
for diaphragms with openings formulation is only approxi-
mate with limited accuracy.
The method is based on subdividing the panel into
components and summarizing the effect of each component
from the point of view of strength and stiffness. In the case of
steel sheeted panels the shear strength is characterised by the
following components: (1) sheet tearing along a line of seam
fasteners; (2) sheet tearing along a line of sheet/shear connector
fasteners; (3) sheet tearing in the sheet/purlin fasteners; (4)
end collapse of the sheeting prole; (5) shear buckling of
the sheeting; (6) failure of the edge member in tension or
compression. In the same way components that inuence panel
deformation in the elastic range are identied as: (1) prole
distortion; (2) shear strain in the sheeting; (3) slip in the
sheet/purlin fasteners; (4) slip in the seam fasteners; (5) slip
in the sheet/shear connector fasteners; (6) axial strain in the
longitudinal edge members.
The methodology has been applied to wall panels in
Series I and IV. Comparative results concerning design values,
experimental results and their interpretations using the two
equivalency methods described in Section 2.2 (Fig. 4a, b) are
presented in Fig. 6.
As observed the ECCS, methodology gives very good
estimates concerning initial rigidity and elastic load bearing
capacity of the panels. However substantial over-strength can
be identied, which has to be taken into account in any realistic
earthquake analysis attempt.
In the case of OSB panels a different procedure has been
adopted based on the observation that such panels behave like a
series of cells. Therefore, in a long wall sheeted with several
similar OSB panels the effect of this cells is cumulative and
load bearing capacity per unit length can be dened. The total
capacity of a wall is then the capacity per unit length multiplied
by the sheeted length of the wall [22]. In the case of walls
with openings this value is then reduced by a factor taking into
902 D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913
Fig. 6. Experimental curves versus ECCS prediction [15].
Fig. 7. Possibilities of wall panel modeling and effect [15].
account the ratio of openings from the total wall area [23]. The
disadvantage of the procedure is that load bearing capacities
per unit length have to be derived on the basis of full-scale
experimental results which are available for a limited number
of sheeting congurations.
A more general approach is to relate sheeting-to-frame
connector slip to the lateral displacement of the wall panel.
In this way a panel can be analysed under increasing lateral
displacement and based on the individual connector properties,
the load bearing capacity can be calculated. As the number
of connectors is usually large, it is convenient to perform the
analysis by computer [24].
As experimental curves are nonlinear from the beginning,
the elastic design capacity of the panels can be dened only
in a conventional way. Obviously, any assumption of elastic
design limit, like in the case of the ECCS Recommendation, is
to be related to a tolerable deformation of the relevant group
of connectors subjected to the highest forces. Therefore the
design capacity of the panel is mainly based on serviceability
than strength criteria. Consequently, there will be an important
strength reserve beyond any design limit considered, due to
load bearing capacity of the remaining connectors when the few
most damaged ones have excessive deformations or fail.
3.3. Hysteretic models [15]
In order to model the complex hysteretic behavior of the
panel different numerical techniques can be employed and they
range, depending on the desired accuracy, from simple bilinear
to highly complex nonlinear models (Fig. 7a, b, c).
The bilinear and trilinear models presented hereafter are
not capable of taking into account strength deterioration due
to repeated loading. In order to cover this shortcoming they
have been dened based on the stabilized envelope of the cyclic
curve. The nonlinear model was built starting from the proposal
of Della Corte et al. [25], based on a RichardAbbott type
curve, and has very good capability in characterising all aspects
of the panel behavior see also [11,12]. However, this model
depends on a large number of parameters, and needs to be
calibrated based on relevant experimental results.
To compare and assess the effects of hysteretic model on the
seismic response a dynamic nonlinear analysis, considering the
D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913 903
(a) Trilinear hysteretic model. (b) Modeling with equivalent bracing.
Fig. 8. Wall-panel idealization [15]
Table 3
Calibrated values for the nite element model [15]
Series I II IV OSB I OSB II
Wall panel scheme
Sheeting Corrugated sheet Corrugated Sheet + Gypsum Corrugated Sheet OSB OSB
Initial rigidity (N/mm) 3446.6 3850.6 1766.3 4197.3 1610.5
Elastic limit (F
el
/D
el
) (N & mm) 24 086/6.99 26 566/6.90 12 8670/7.28 28 942/6.89 11 850/7.36
Yield limit (F
yield
/D
yield
) (N & mm) 33 560/14.95 39 819/15.58 26 812/23.78 48 944/17.49 33 908/27.76
Ultimate limit (D
ult
/D
uct
) (mm) 42.61/4.37 57.29/5.54 94.35/6.22 42.85/3.67 65.57/3.11
Kobe-JMA record has been performed using all three models
(Fig. 7d). The differences seem to be negligible, but one can
clearly identify the benecial effect of taking into account the
realistic behavior in the case of the nonlinear model considering
higher capacity, but also strength deterioration due to repeated
loading.
Whichever model is used, it is important to take into
account the three main characteristics of the panel behavior;
(1) pinching, (2) over-strength as difference between allowable
elastic design limit and actual capacity and (3) plastic
deformation capacity.
In further studies the trilinear model, based on DRAIN-
3DX [26] computer code was used, together with the full
nonlinear model employed for comparison. A single degree
of freedom system (SDOF) with a ber-hinge accommodating
the desired hysteretic behavior has been constructed and
calibrated using the experimental results (Fig. 8a). According
to this approach the wall shear panel can be replaced with
an equivalent bar model (e.g. pinned rectangular frame with
dissipative diagonals, see Fig. 8b).
As all column ends are hinged the frame itself is a
mechanism and it does not contribute to load bearing
capacity. Braces are modeled as TYPE 8 ber-hinge (FH)
beamcolumn elements with FH to accommodate the hysteretic
behavior. In order to calibrate the nite element model
experimental results from the full-scale testing program
described in Section 2 was used (Table 3).
Fig. 9. Elastic spectra of records (damping factor = 5%).
3.4. Incremental dynamic analysis
For the purpose of earthquake analysis ve (Fig. 9)
earthquake records have been selected. Normalised elastic
spectra with a damping ratio of 5% from the critical have been
compared to Eurocode 8 elastic spectra for A, B and C soil
conditions [27].
As one observes the three records (EL, NE, KO) resemble
reasonably the proposed Eurocode 8 elastic spectra, while
the other two (SH and VR) reect extremely stiff and soft
soil conditions respectively. Therefore, results from these two
records representing very lowand very high corner periods have
to be treated with some caution.
Using the SDOF system with the simplied trilinear
hysteretic model, time history analyses have been carried out
904 D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913
Fig. 10. Denition of limit states on the envelope curve.
Fig. 11. IDA curve example.
using all tested wall panels with acting masses of 2000, 2500,
3000, 3500, and 4000 kg, and records being scaled from 0.05 g
to 2 g.
Due to lack of reliable values, and in order to take into
account the hysteretic characteristics only, damping has not
been considered even if values as big as 6% have been
suggested at the level of an entire structure [7]. To account for
second-order effects a vertical force equal to 30% of the mass
value has been also used in the model.
The procedure utilised is known as incremental dynamic
analysis (IDA) or dynamic pushover (DPO) and it is a common
analysis procedure which is both structure and accelerogram
sensitive. Outputs of the analysis consist in IDA curves relating
a Performance Parameter (PP) of the structure to an Intensity
Measure (IM) of the record [28].
Based on the displacement values corresponding earthquake
IM levels have been identied for the different panel
congurations and earthquake records. The three limit states
(Figs. 10 and 11) correspond to the following states for the
wall panel under consideration: D
el
elastic design limit of the
panel up to which behavior can be considered elastic and it is
the conventional capacity to be used in design; D
yield
yield
limit of the wall panel, where the panel lost its load bearing
capacity, but it is still capable of deforming under constant load,
D
ult
ultimate state, the panel is not capable of sustaining a
constant load level, and its capacity is decreasing.
The last two limit states can be identied fairly accurately,
alternative methods of determination yielding similar values.
The elastic design limit is more a conventional value accepted
in practice and depends on what is to be considered tolerable
from the point of view of local connection damage.
If elastic design is assumed, the limit of D
el
is the basis
of engineering calculations, even if the panels have important
post-elastic capacity. In European seismic design practice this
post-elastic behavior is to be accounted for by the behavior
factor q, used for design purposes to reduce the forces obtained
from linear analysis, in order to account for the nonlinear
response of the structure.
As shown by tests, wall panels behavior is characterised by
important strength reserve over the accepted allowable design
strength and it can be expected that this over-strength plays
an important role in the post-elastic performance. Based on
previously dened limit states (D
el
, D
yield
and D
ult
), partial
behavior factors can conveniently be dened as the ratios of the
corresponding IM-s. Following this idea; q
1
has been dened
as the ratio of Sa
yield
and Sa
el
, and it is primarily a measure
of performance due to panel over-strength, while q
2
, the ratio
of Sa
ult
and Sa
yield
is to be understood as a performance
parameter due to ductility (Table 4). The behavior factor q can
be calculated as q = q
3
= q
1
q
2
.
It is important to mention that q
1
is highly dependent on
the elastic design limit dened (D
el
), limit that is on the other
hand conventional. As the design strength denition was based
on a 1/300 story drift angle [19], for specimens with low
initial rigidity (e.g. specimens with door openings, Series IV
and Series OSB II) the criteria is very severe and a very low
Sa
el
is identied. This particularity yields high and unrealistic
values of q
1
and consequently q
3
. The value of q
2
instead is
less dependent on conventional values. Some dependency of q
1
and q
2
with earthquake records was observed, but dispersion of
the results is not unacceptable keeping in mind the uncertainties
related to the phenomena.
For this reason, in the case of panels with door openings
we suggest to take q = q
2
, and to account in fact for the
dissipative capacity due to ductility only. It is, perhaps an
empirical assumption, but for sure, it is on the safe side.
4. Experiments on connections
4.1. Description of experimental program [16]
After observing the paramount importance of the connection
details on the behavior of the wall panels, an additional
experimental program dealing with the component materials
and the connections has been designed [16].
The rst series of tests had the aim of determining the
properties of the base materials (i.e. steel, OSB panels). These
experiments revealed good correlation between nominal and
measured mechanical properties of the base steel, and a big
scatter in the measured properties of the OSB material.
The second series of experiments was carried out to
determine the mechanical properties of the connections that
were identied to have a crucial effect on the behavior of
the panels. From this point of view, the self-drilling screws
connecting the sheeting material to the steel skeleton and seam
connections are important and were of the following typologies:
(1) connection between the corrugated steel sheeting and steel
proles; (2) sheeting-to-sheeting seam connections and (3)
connections between the OSB panels and steel proles.
Steel-to-steel connections using self-drilling screws are
largely used in everyday practice, and both testing and
D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913 905
Table 4
Sa and performance parameters q
1
, q
2
and q
3
Series I Series II Series IV Series OSB I Series OSB II
2t 2.5t 3t 3.5t 4t 2t 2.5t 3t 3.5t 4t 2t 2.5t 3t 3.5t 4t 2t 2.5t 3t 3.5t 4t 2t 2.5t 3t 3.5t 4t
Sa
el
(g) 0.62 0.47 0.41 0.39 0.36 0.97 0.60 0.51 0.42 0.39 0.36 0.29 0.21 0.22 0.19 0.97 0.61 0.47 0.45 0.43 0.34 0.29 0.19 0.20 0.14
Sa
yield
(g) 1.63 1.15 0.97 0.85 0.79 1.75 1.39 1.08 0.97 0.81 1.09 0.87 0.78 0.69 0.60 2.01 1.74 1.26 1.18 1.04 1.11 0.94 0.67 0.72 0.58
Sa
ult
(g) 2.06 1.65 1.53 1.31 1.14 2.64 1.99 1.83 1.67 1.49 2.04 1.84 1.75 1.61 1.58 2.84 2.09 1.77 1.69 1.43 1.87 1.75 1.36 1.18 1.19
q
1
2.84 2.57 2.42 2.25 2.39 1.89 2.35 2.17 2.38 2.26 3.04 3.12 3.80 3.09 3.37 2.14 2.97 3.04 2.70 2.47 3.46 3.43 3.80 3.95 4.29
q
2
1.25 1.49 1.59 1.53 1.45 1.51 1.44 1.74 1.72 1.83 1.92 2.12 2.40 2.54 2.82 1.42 1.21 1.44 1.43 1.37 1.73 1.85 2.02 1.65 2.15
q
3
= q
1
xq
2
3.58 3.64 3.92 3.46 3.48 2.90 3.36 3.60 4.10 4.09 5.72 6.46 8.71 7.44 9.90 3.11 3.62 4.23 3.99 3.39 5.86 6.42 7.60 6.24 8.66
Average
q
1
q
2
q
3
q
1
q
2
q
3
q
1
q
2
q
3
q
1
q
2
q
3
q
1
q
2
q
3
2.50 1.46 3.62 2.21 1.65 3.61 3.28 2.36 7.65 2.66 1.38 3.67 3.78 1.88 6.96
Table 5
Summary of tested connections
Base sheet thickness (mm) Connected sheet thickness (mm) Screw diameter (mm) Loading velocity
Slow (V1 = 1 mm/min) Fast
(V2 = 420 mm/min)
Nr. Code Nr. Code
1.42 0.417 4.8 4 +2 I-TP-M-V1-1. . . 4, 6, 7 4 I-TP-M-V2-1. . . 4
0.417 0.417 4.2 4 I-TS-M-V1-1. . . 4 3 I-TS-M-V2-1. . . 3
Fig. 12. Dimensions of the steel-to-steel connections.
calculation methods are well established. Specimens were
chosen in such a way that they represent (1) and (2) connection
typologies used in the tested wall panels. The dimensions of
the specimens were chosen according to the ECCS (1983)
Recommendations [29], the dimensions of the connecting
plates being 50 220 mm (Fig. 12). Tests were conducted on
two types of specimens (Table 5): (1) connecting corrugated
sheet to skeleton (0.4171.42 mm sheet) using SD3-T15-4.8
22 (4.8 mm) screws and (2) corrugated sheet to corrugated
sheet (0.417 mm to 0.417 mm sheet) using SL2-T-A14-4.8
20 (4.8 mm) screws. The edge distance of the sheets in the
direction of the loading was 30 mm, in order to facilitate
bearing failure of the thinner sheet, failure mode that was
observed during the panel test.
Test connections were executed with the same materials and
in similar conditions as the ones in the panels. In the case of the
thin-to-thick sheet specimens, the thinner sheet was equipped
with a supplementary plate to ensure centric transmission of
the loading of the screw.
The minimum number of tests of the same typology was 3.
If during tests technical problems appeared this number was
supplemented to 5 or 7. Two loading velocities were applied,
V
1
= 1 mm/min for quasi-static loading conditions and
V
2
= 420 mm/min for high-velocity tests. The extensometer
Fig. 13. Experimental curves for I-TP-M-V1.
captured the elongation of an 80 mm portion from the middle
part of the specimens.
The testing methodology was adapted to simulate the
behavior of connections into the panel [16].
Figs. 1316 show comparatively the experimental curves for
specimen series I-TP-M-V1 and I-TM-M-V2, and respectively,
I-TS-M-V1 and I-TS-M-V2.
In the case of sheeting-to-framing specimens, the failure
mode was, generally, as expected through tearing of thinner
material, while, in the case of thin-to-thin sheet connections,
which were modeling seam fasteners, the failure mode was due
to tilting and pull-out of the screw, the same failure modes were
observed during wall panel tests.
The load bearing capacity (F
nom
) and the rigidity (K
serv
)
of the connections can analytically be calculated using several
methods, the subject of connections using self-drilling screws
being well documented in the technical literature and in design
906 D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913
Table 6
Comparative results of experiments and analytical methods [16]
v (mm/min) EC 3 Fan Experiment ECCS
0.4171.42 mm
1
F
nom
(N) 1961.9 2618.2 2206.4 1775.11
K
serv
(kN/mm) 0.760 8.624 (4.312
a
) 2.86
b
420
F
nom
(N) 2037.3 2718.8 2200.3 1843.30
K
serv
(kN/mm) 0.760 8.615 (4.313
a
) 2.86
b
0.4170.417 mm
1
F
nom
(N) 881.2 1074.3 1091.3 798.58
K
serv
(kN/mm) 0.428 13.249 (6.615
a
) 4
420
F
nom
(N) 915.0 1115.6 1247.7 829.25
K
serv
(kN/mm) 0.428 9.526 (4.763
a
) 4
F
nom
is the reference bearing capacity; the design strength results by dividing F
nom
with the relevant safety factor. K
serv
is the design rigidity considered in the
serviceability limit state check.
a
In the literature it is suggested that the rigidity of a connection with a single screw is twice as big as the rigidity (reduced to a screw) of a connection with
multiple screws.
b
The closest value to be assimilated is for d = 5.5 mm screw with neoprene.
Fig. 14. Experimental curves for I-TP-M-V2.
Fig. 15. Experimental curves for I-TS-M-V1.
standards. However, calculation methods usually refer to the
calculation of the capacity and in some cases rigidity, the
ductility of such connections being considered of less interest.
In this case three alternative calculation methods were applied
for the connection typologies under investigation. Using the
design method of Eurocode 3-1.3 [30] only the load bearing
Fig. 16. Experimental curves for I-TS-M-V2.
capacity of the connections can be calculated. Using the
ECCS method [31] the load bearing capacity can be calculated
and the rigidity of the connection is included for typical
connection typologies as empirical values. The proposal of Fan
et al. [32] is an improvement of the Eurocode 3-1.3 method and
allows for the calculation of more varied connection typologies
and contains specic recommendations for the evaluation of
rigidity. The three methods have been applied for the tested
connections, comparative results being presented in Table 6.
A third connection typology used in the wall panel test was
the one connecting OSB to the steel skeleton. In order to test
this typology of connections, specimens as presented in Fig. 17
were prepared.
The testing of these specimens yielded very inhomogeneous
results (Fig. 18) depending on the direction and density of bers
in the vicinity of the screw and between the screw and the
margin of the OSB panel. No generalizing conclusion can be
drawn from these experiments; besides that OSB connections
possess less ductility, the most likely reason for the low ductile
failure mode of the wall panels sheeted with OSB.
D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913 907
Fig. 17. OSB-to-steel skeleton connection specimens.
Fig. 18. Experimental curves of the OSB-to-steel connections.
4.2. Performance criteria
An important aspect of performance-based seismic design
philosophy is to dene acceptable damage levels and
relate them to the performance objectives. Performance
objective proposals are based on three or four goals [20]:
(1) Serviceability under ordinary occupancy conditions; (2)
Immediate occupancy following moderate earthquakes; (3) Life
safety under design-basis events; (4) Collapse prevention under
maximum considered event. Such vague goals can be translated
into practice by relating performance objective to deformations,
using for horizontal loads the inter-story drift (), as measure.
Fig. 20. Performance criteria for OSB-to-steel connection [16].
In the case of the wall panels with corrugated sheeting the
main damage was concentrated in the seam fasteners. It is
important to establish an acceptable level of deformation at
connection level and, for different wall typologies, relate this
to the overall deformation of the wall panel. To establish global
performance criteria the following acceptable deformations in
the seam fasteners are suggested:
If slip of the seams does not exceed the elastic limit (D
e
,
Fig. 19), corresponding to 0.6F
max
of the seam connection,
damage is limited and can be considered negligible. In
this case the integrity of the cladding is fully preserved,
no repairs are required; it corresponds to serviceability
conditions.
If slip is limited to the diameter of the screw (D
r
=
4.8 mm, Fig. 19) the cladding requires some repair. There is
damage, but not excessive and by minor interventions, like
replacing screws with larger diameter ones, the structure can
be repaired. This could correspond to immediate occupancy.
In the case of life safety criteria any kind of damage is
acceptable, without endangering the safety of occupants.
This criterion corresponds to the attainment of the ultimate
force (F
ult
) and the starting of the downwards slope.
As it can be observed from the experiments the behavior of
the connections is nonlinear. Both initial rigidity and especially
Fig. 19. Performance criteria at steel-to-steel connection level [16].
908 D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913
design capacity can only be assigned as conventional value in
relationship with the curves, especially in the case of thin-to-
thick steel sheet connections. Any such assumed value of the
design capacity will inevitably lead to some supplementary
strength of the connection not considered. Furthermore, the
load distribution in the connections of the panel is uneven.
While some connections are loaded to failure others are loaded
with much lower loads, and when some connections fail the
load will be redistributed to other groups of connections.
Based on these assumptions at connection level the
following performance criteria are suggested for wall panels
clad with corrugated sheet: (1) fully operational ( < 0.003);
(2) partially operational ( < 0.015); (3) safe but extensive
repairs required ( < 0.025).
The rst performance level does not provide ductility,
because shear panel work is elastic. This could be the design
criteria for frequent, but low intensity earthquakes. In the case
of rare but severe earthquakes, the last two design criteria can
be used and some ductility will be available.
In the case of OSB-to-steel connections, which are
characterised by a fragile behavior, the design has to be
controlled by the elastic limit only (D
e
Fig. 20). In such
a case multiple performance levels cannot be applied.
4.3. Numerical modeling of panels accounting for the behavior
of connections
Based on the component characteristics determined experi-
mentally, an attempt of Finite Element (FE) modeling (Fig. 21)
was made to reproduce the behavior of the entire wall panel. As
the rst step the wall panels sheeted with corrugated sheeting
were considered because component behavior was more homo-
geneous in this case.
The bars of the skeleton were modeled as elastic beam
elements taking into account that these elements were not
heavily deformed in the post-elastic range. The corrugated sheet
was modeled as an equivalent orthotropic plate in order to take
into account the fundamentally different mechanical properties
of the corrugated sheet in the two principal directions and
the distortion of the corrugated sheet when loaded in shear.
The equivalent elastic modules (E
l
longitudinally and E
t
transversally) have been determined through preliminary
modeling of a single corrugation loaded in tension in the
two principal directions. The equivalent shear modulus (G
eff
)
has been calculated taking into account the end distortion
of the corrugated sheeting. Connections, both between the
skeleton and the sheeting and seam connections were modeled
using nonlinear spring elements taking average deformation
properties of the tested connection loaded at 1 mm/min
(Figs. 13 and 15). Besides these components it is important to
take into account the uplift deformation characteristics of the
wall panel corners, without which any attempt of modeling is
unrealistic.
The FE model was subjected to increasing horizontal
loading at the upper part of the panel similarly to wall panels
during the full-scale test. The deformation pattern and the
nonlinear behavior curve obtained with the FEM model was
compared with the monotonic curves obtained from the wall
panel experiments (Fig. 22).
Fig. 22 and 23 show a remarkable similarity between FE
models and experiments both in terms of deformation pattern
and nonlinear behavior up to large displacements. Using the
described FE technique the performance of wall panels with
different congurations, but using the same basic materials and
connection typologies can be evaluated. The correspondence
between the experimental and numerical drift values associated
with the three performance levels dened in the previous
section is quite perfect. So, the possibility of using such a FE
modeling reduces considerably the amount of full tests, even if
a few experiments are still necessary to calibrate the FE model.
5. Conrmation of design methodology and global perfor-
mance by in situ measurements
5.1. Designed structure
The structure presented in Fig. 24 belongs to a one story
single family house.
The dimensions of the building are 9m (axes AD) by
10.5 m (axes 14), of which 1.5 m is between axes 3 and 4.
Each level is about 2.75 m high and the slope of the roof is 30

.
The structural skeleton was made of cold-formed C-shaped
proles with a core thickness of 1.5 mm (C150/1.5) placed at
600 mm intervals. The elements of the skeleton were connected
Fig. 21. Elements to be taken into account for the FE modeling.
D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913 909
Fig. 22a. Comparison of deformed shape (experiment versus FEM Series I).
Fig. 22b. Comparison of deformed shape (experiment versus FEM Series IV).
Fig. 23. Comparison of deformed shape (experiment versus FEM Series I, Series IV).
using 4.8 mm diameter self-drilling screws. The height of the
C prole was 150 mm value governing the thickness of the
walls. The load bearing beams in the slab are C200/1.5 proles
placed at 600 mm. The size of these proles resulted from the
condition to control the vibrations of the oor not from strength
consideration. Roof purlins are Z150/1.5 proles placed at
1200 mm intervals.
The skeleton was stiffened using 10 mm thick OSB plates
xed to both sides of structural walls (Fig. 25). The oor
diaphragms were originally designed also to be based on OSB
covering. This solution has later been changed into sheathing
with trapezoidal steel sheaths, both in the slab and at the level
of the roof. No concrete was used in the slab.
The self-weight of the elements was evaluated to:
0.45 kN/m
2
for the roof; 0.70 kN/m
2
for the slab; 0.60 kN/m
2
for exterior and 0.20 kN/m
2
for internal walls. Other loadings
for the design were determined according to the relevant
Romanian standards. Live load on the slab was 1.50 kN/m
2
,
snow load on the roof was 1.20 kN/m
2
and wind load on the
surface exposed to maximum lateral pressure was 0.40 kN/m
2
.
The earthquake design was done for a Peak Ground
Acceleration (PGA) of 0.25g and allowing no reduction
of the seismic forces (q = 1). Such a severe condition
had to be fullled due to local regulations which do not
allow for reduction of seismic forces for cold-formed steel
structures different from the structural scheme. One important
particularity of the seismicity in the region is the unusually high
value of the corner period T
C
= 1.5 s. For the design spectral
analysis has been used.
One of the problems during the design of such a structure is
the evaluation of the load bearing capacity and of the rigidity of
the sheathing system of walls and slabs. For this case, the slabs
can be covered with general provisions (e.g. some detailing
conditions) due to the low level of stresses, but the realistic
910 D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913
Fig. 24. Steel skeleton of the structure.
Fig. 25. The skeleton with the structural OSB sheeting.
analysis for the walls is crucial. In this case extrapolation from
existing experimental results on shear walls was the basis of the
evaluation of the load behavior properties. Walls with similar
skeleton sheathed on one side using identical OSB panels
(t
OSB
= 10 mm), with similar xing plan (e.g. 105 mm screw
intervals) and the same diameter screws (d
screw
= 4.2 mm)
were previously tested monotonically and cyclically [14]. The
tested panels were 2.55 mhigh, 3.6 mlong and the ultimate load
obtained from the test was F
max
= 69.8 kN (or 19.4 kN/1m).
The elastic rigidity of the panel was K
ini
= 4197 kN/mm.
The wall panels used in this case were sheathed on both sides
with OSB and the corner details were strengthened compared
to the tested case. The design capacity of the walls used for the
construction has been evaluated to F
des.
= 17.6 kN/1m.
5.2. Numerical simulation
The rigidity generated by the sheathing has been modeled
by equivalent cross bracings and considering the frame to be
pinned. The cross section area of the brace has been calculated
to produce the same rigidity as the sheathing it replaced. Once
the brace area was calibrated for the dimensions of the tested
wall panels (2.55 3.6 m), the wall panels in the structure
have been divided into sub-panels of approximately this size
and the sub-panels were braced with the brace (A = 64 mm
2
).
Differences of up to 1015% of the height and length of the
Table 7
Modal parameters obtained by FE modeling [17]
Stage T
1
(s) Type
a
T
2
(s) Type
a
T
3
(s) Type
a
1 0.44 Tr 0.39 To 0.35 Lo
2 0.19 Lo 0.18 Tr 0.13 To
3
b
0.33 Lo 0.31 Tr 0.23 To
a
Note
1
: Tr transversal, Lo longitudinal, To Torsional.
b
Note
2
: Only the masses changed from Stage 2 to Stage 3.
wall panel had to be accepted due to the conguration of the
structure.
The masses were calculated from the earthquake combina-
tions (1 self-weight +0.4 live load +0.3 snow load) accord-
ing to Romanian standards. The rst three vibration modes of
the structure have been calculated using the FE program AXIS-
VM (*** 2000) and the periods of vibration were obtained to
be: T
1
= 0.38 s, T
2
= 0.35 s and T
3
= 0.26 s.
The structure is subjected to torsion during earthquake
because the walls in axes 1 and D are fully sheathed while the
ones in axes 34 and A had to accommodate openings. In these
conditions the most critical wall panel is the one at the ground
oor in axis A. The maximumshear force in this wall was 43 kN
(12.8 kN/1m).
The dynamic properties of the structure for the design have
been presented earlier. However, in the proposed measurement
stages the structure was much lighter than in the earthquake
design situation:
In Stage 1 only the self-weight of the steel proles has to be
taken into account.
In Stage 2 part of the self-weight is present, but the self-
weight due to nishing, live load and snow load has to be
eliminated.
In Stage 3 all the self-weight is acting on the structure (but
still there is no live or snow load).
The self-weight of the steel skeleton (i.e. excluding the
trapezoidal sheathing) has been evaluated to 4600 kg. The
masses of the sheeting and nishing elements are M
1
= 700 kg
(Stage 1), M
2
= 4650 kg (Stage 2) and M
3
= 25 200 kg (Stage
3). In the earthquake design situation the mass of the structure
is M
design
= 32 700 kg. Taking into account the contribution
of these masses the vibration periods presented in Table 7 were
predicted using FE analysis. Note that Stage 3 (i.e. the structure
with nishing) differs from Stage 2 only from the point of
view of the supplementary mass generated by the nishing; the
stiffness contribution of the secondary and nishing elements
has not been included in the modeling.
All mode shapes of the structure contain some degree
of torsion because the center of rigidity of the structure is
shifted towards the more heavily sheathed walls. For a visual
evaluation the mode shapes for Stage 2 are presented in Fig. 26.
5.3. In situ measurements
The dynamic properties of the building have been studied for
small amplitude vibrations by measurements in the 3 stages of
construction described earlier.
D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913 911
Fig. 26. Mode shapes in stage 2 (FE modeling): (a) rst mode T
1
, (b) second mode T
2
, (c) third mode T
3
.
During the measurements the working activities at the
building site have been stopped, only the ambient vibrations
being transmitted to the structure. For each of the three stages
of construction, 10 different measurement schemes with 10
velocity sensors were used and for each scheme two vibration
recordings of 2 minutes duration were done with a sampling
frequency of 100 Hz [33].
For the vibration measurement a GEODAS mobile
acquisition station (Buttan Service-Tokyo & Tokyo Soil) has
been used coupled with velocity sensors type CR2-5 H and
CR2-5 V.
Because of measuring only the response to ambient
vibrations (as excitation of the structure) not even the friction
forces in the connections have been exceeded.
Based on the analysis of recorded vibrations, the buildings
modal frequencies (periods) have been estimated using two
identication techniques: (i) analysis of the Fourier spectra
and (ii) spectral and correlation analysis. For the estimation
of the modal frequencies and corresponding damping ratios,
the Eigen-system Realisation Algorithm (ERA) has been used,
which also validated the modal periods obtained by the above
mentioned techniques [34].
The rst set of measurements has been done when the
construction was in the stage when the skeleton of the building
was almost entirely completed and included 10 schemes with
10 sensors at different locations in the building. The corrugated
sheet of the slab was in place together with the OSB panels over
it (i.e. the ones which support the nishing layers of the slab).
However, the roof of the structure was not nished, not even
the purlins for the roof being in place. Practically, a exible
diaphragm was provided at the level of the rst slab but not at
the level of the roof.
The second set of measurements was done when the building
skeleton was nished and most of the structural OSB panels
were xed to it. Some OSB panels, only, were still missing
in the attic area. The structural part of the roof has also been
nished with the outer corrugated sheath being in place. No
work on the nishing (i.e. external thermal insulations, inner or
outer wall nishes, slab nishing) has been undertaken at this
stage.
The third set of measurements was done, before the moving
in of the occupants. Therefore, all nishing has been completed
but furniture has not been moved in. This can be considered the
nal stage of construction without live load.
Based on the analysis of the recorded vibrations the modal
frequencies (periods), damping ratios and mode shapes of the
structure have been identied (Table 8).
It has to be underlined that the presence of the stiffening
OSB panels not only increased the rigidity of the structure
considerably, but the weakest rigidity direction has been also
changed. In Stage 1 the rst mode shape was longitudinal
(Table 8) with the period T
1,St 1
= 0.546 s, while in Stage
2 the rst mode became transversal (T
1,St 2
= 0.103 s). It is
interesting to observe that from Stage 1 to Stage 2 the damping
ratio has also increased considerably. In Stage 3 no important
change of the vibration properties can be observed. This means
that the supplementary mass introduced with the nishing is
counterbalanced by the stiffness increase generated by these
nishing elements themselves.
When live load will be added (i.e. furniture etc.) the period of
vibration will slightly increase compared to the value T
3,St 2
=
0.101 s. Most probably the period T
1
= 0.150.2 s will be
reached when the house is fully furnished.
912 D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913
Table 8
Modal parameters based on the analysis of ambient vibrations [17]
S Mode 1 Mode 2 Mode 3
T
1
(s)
1
(%) Type T
2
(s)
2
(%) Type T
3
(s)
3
(%) Type
1 .546 1.2 Lo .437 1.1 Tr .456 1.3 To
2 .103 3.4 Tr .096 3.7 Lo
3 .101 4.1 Tr .096 3.8 Lo .072 4.1 To
Note: Vibration shapes: Tr transversal, Lo longitudinal, To Torsional
Another observation is that the damping ratio = 5% is a
reasonable estimate for a light-gauge steel house. Even if the
values of 3.8%4.1% obtained here are slightly unconservative,
it is suggested by Kawai et al. [19] that with the increase of
the excitation level both the damping ratio and the period of
vibration will also increase.
6. Concluding remarks
Cold-formed steel framing houses represents, perhaps, the
best structural solution for such a type of building located in
seismic areas.
In terms of performance, for fully operational and
immediate occupancy levels, they could be designed as Low
Dissipative Structures corresponding according to Eurocode
8-1 [35] taking q factor of 1.5 to 2.0, while, for life safety
and collapse prevention, these structures could be framed as
Moderate Dissipative with a corresponding factor of 23. Both
test results and numerical simulations sustain this assumption.
Moreover, the in situ measurements provide evidence for
a signicant damping effect due to the nishing materials,
which also contribute to resist earthquakes in the case of these
constructions.
References
[1] Pekoz T. Building design using cold formed sections. Publication of the
Swedish Institute of Steel Construction no. 154; 1995.
[2] AISI, Shear Wall Design Guide, Publication RG-9804. The American Iron
and Steel Institute; 1998.
[3] Serette RL, Ogunfunmi K. Shear resistance of gypsum-sheeted light-
gauge steel stud walls. Journal of Structural Engineering, ASCE 1996;
122(4).
[4] Serrette RL, Hall G, Nygen J. Shear wall values for light weight steel
framing. AISI; 1996.
[5] Salenikovich AJ, Dolan JD, Easterling WS. Racking performance of
long steel-frame shear walls, In: Proceedings of the 15th international
speciality conference on cold-formed steel structures, 2000. p. 471480.
[6] Serrette RL. Seismic design of light gauge steel structures: A discussion,
In: Fourteenth int. specialty conference on cold-formed steel structures.
1998.
[7] Kawai Y, Kanno R, Uno N, Sakumoto Y. Seismic resistance and design
of steel framed houses. Nippon Steel Technical report, No. 79; 1999.
[8] Gad EF, Chandler AM, Dufeld CF, Hutchinson G. Earthquake ductility
and overstrength in residential structures. Structural Engineering and
Mechanics 1999;8(4):36182.
[9] Gad EF, Dufeld CF. Lateral behaviour of light framed walls in residential
structures. In: 12th world conference on earthquake engineering. 2000.
[10] De Matteis G. The effect of cladding in steel buildings under seismic
actions. Ph.D. thesis. Universita degli Studi di Napoli Federico II; 1998.
[11] Landolfo R. et al. Theoretical and experimental study on the seismic
performance of lightweight cold-formed steel low-rise residential
building. Research report PRIN 2001 Innovative steel structures for
seismic protection of buildings. University Federico II Naples; 2004.
[12] Landolfo R, et al. Seismic behaviour of sheathed cold-formed structures:
physical tests. Journal of Structural Engineering, ASCE 2006;132(4):
57081.
[13] Serrate R, et al. Cold-formed steel frame shear walls utilising structural
adhesives. Journal of Structural Engineering, ASCE 2006;132(4):5919.
[14] F ul op LA, Dubina D. Performance of wall-stud cold-formed shear panels
under monotonic and cyclic loading Part I: Experimental research. Thin-
Walled Structures 2004;42(2).
[15] F ul op LA, Dubina D. Performance of wall-stud cold-formed shear panels
under monotonic and cyclic loading Part II: Numerical modeling and
performance analysis. Thin-Walled Structures 2004;42(2).
[16] Fulop LA, Dubina D. Design criteria for seam and sheeting-to-framing
connections of cold-formed steel shear panels. Journal of Structural
Engineering, ASCE 2006;132(4):58290.
[17] Dubina D., et al. In: Mazzolani MF, Wada A, editors. Proc. of seismic
performance of cold-formed steel framed houses 5-th international
conference on steel structures on seismic areas. Taylor & Francis, London,
2006. 429436.
[18] ECCS, Recommended testing procedure for assessing the Behaviour of
structural steel elements under cyclic loads, 1985.
[19] Kawai Y, Kanno R, Hanya K. Cyclic Shear Resistance of Light-Gauge
steel framed walls. Poland: ASCE Structures Congress; 1997.
[20] FEMA-273. NEHRP guidelines for the seismic rehabilitation of buildings;
Prepared by the Building Seismic Safety Council for the Federal
Emergency Management Agency, Washington, DC; 1997.
[21] ECCS P88, European recommendations for the Application of Metal
sheeting acting as a Diaphragm, Publication 88. European Convention for
Constructional Steelwork. Brussels; 1995.
[22] AISI, Shear Wall Design Guide, Publication RG-9804. The American Iron
and Steel Institute; 1998.
[23] Philip L. Perforated shear wall design. Wood Design Focus 2002;12(2).
[24] Ekwueme G, Hart GC. Non-linear analysis of light-framed wood
buildings. In: Proceedings of the 12th world conference in earthquake
engineering. 2000.
[25] Della Corte G, De Matteis G, Landolfo R. Inuence of connection
modeling on the seismic response of moment resisting steel frames, In:
Mazzolani FM, editor. Moment resistant connections of steel frames in
seismic areas. E & FN Spoon; 2000.
[26] Prakash V, Powell GH. Drain-3DX Base program description and user
guide, Version 1.10. Department of Civil Engineering, University of
California at Berkley, 1994.
[27] ENV-Eurocode 8. Design provisions for earthquake resistance of
structures; 1998.
[28] Vamvatsikos D, Cornell CA. Incremental dynamic analysis. Earthquake
Engineering and structural Dynamics 2002;31(3):491514.
[29] ECCS (1983), P21 - european recommendation for steel construction: The
design and testing of connections in steel sheeting and sections. ECCS;
1983.
[30] Eurocode 3 Part 1.3. (2001), Supplementary rules for cold formed thin
gauge members and sheeting. 20 July; 2001.
[31] ECCS (1995), P88 European recommendations for the application
of metal sheeting acting as a diaphragm. European Convention for
Constructional Steelwork. Brussels; 1995.
D. Dubina / Journal of Constructional Steel Research 64 (2008) 896913 913
[32] Fan L, Rondal J, Cescotto S. Finite element modeling of single lap screw
connections in steel sheeting under static shear. Thin Walled Structures
1997;27(2):16585.
[33] Aldea A, Demetriu S, Vacareanu R, Oprea R. Ambient vibration
measurements and analysis for a cold-formed steel framed house,
Contract report. Technical University of Civil Engineering, Bucharest,
22p + Annexes 154p. 2005.
[34] Demetriu S. Modal parameter identication of instrumented buildings
from earthquake records. In: Lungu, D, Wenzel, F, Mouroux, P, Tojo, I,
editors. Earthquake loss estimation and risk reduction (Proceedings of the
international conference, Vol. 2); 2004. p. 113130.
[35] EN1998-1 Eurocode 8: Design of structures for earthquake resistance Part
1: General rules, seismic actions and rules for buildings, CEN - Brussels,
Belgium; 2004.