Journal of Constructional Steel Research 128 (2017) 498–511

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Welded fuses for dissipative beam-to-column connections of composite

steel frames: Numerical analyses
Marco Valente ⁎, Carlo A. Castiglioni, Alper Kanyilmaz
Department of Architecture, Built Environment & Construction Engineering ABC, Politecnico di Milano, Piazza Leonardo da Vinci 32, 20133 Milan, Italy

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

Article history: An innovative repairable fuse device for moment-resisting composite steel frames is numerically investigated in
Received 29 March 2016 this study developed within the European research project “Fuseis”. The fuse device consists of steel plates
Received in revised form 30 August 2016 welded to the web and bottom flange of the beam with a specifically detailed gap in the concrete slab. In case
Accepted 14 September 2016
of damage after a severe seismic event, the repair work is limited only to the replacement of the fuse. The results
Available online xxxx
of preliminary numerical analyses carried out on detailed finite element models of beam-to-column sub-assem-
Keywords:
blages show that the dissipative fuses are able to concentrate plastic deformations in the steel plates protecting
Welded fuse devices the other structural elements. Parametric investigations are performed to examine the influence of some relevant
Energy dissipation geometric characteristics of the flange plate on the fuse behavior. Then, a simplified numerical model of the de-
Numerical analyses vice is developed and calibrated using the results of experimental tests carried out on beam-to-column sub-as-
Seismic performance semblages. The effects of the application of the device on the seismic performance of multi-storey composite
Composite steel frames steel frames are investigated through non-linear dynamic and static analyses. The results of the numerical anal-
yses show the effectiveness of the fuse devices to dissipate large amounts of plastic energy and to preserve the
other structural elements from damage. The influence of both the main mechanical characteristics and different
locations of the fuses is also discussed in terms of relative seismic performance of the composite steel frames
under study.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction The work described in this paper is developed within a European re-
search program named “Fuseis” [2–5]. The aim of the “Fuseis” project is
Under strong earthquakes building structures may experience se- to develop innovative fuse devices for moment-resisting composite
vere damage in the main structural elements that are not easily repair- steel frames, with the following functional objectives: 1) inelastic defor-
able. A simple and cost-effective repair of the damaged parts of the mations are concentrated in the devices that act as dissipative fuses; 2)
structures is a critical issue for seismic strengthening of structures and the fuses can be easily fabricated, installed and removed, limiting repair
still represents an interesting research subject in earthquake-resistant costs and time required to make the structure operational after severe
design. According to the provisions of modern seismic codes, in mo- earthquakes. A simplified scheme of the fuse device placed in a mo-
ment-resisting frames beams are the dissipative zones and damage ment-resisting composite steel frame is sketched in Fig. 1. The innova-
may occur at the beam ends near the beam-to-column connections. tive system is a moment-resisting frame in which the beams are cut
Failures of beam-to-column connections of steel buildings in past earth- near their ends and continuity is provided by means of additional
quakes generated concern about the reliability and high repair costs of steel plates (fuse devices) that are welded to the web and the lower
conventional steel moment connections. Different alternative methods flange of the beams. Potential plastic hinges are supposed to take
have been suggested for improving the seismic behavior of connections, place in the middle of the fuse device away from the connection. In con-
moving the likely plastic hinges in beams away from the column face ventional moment-resisting frames beams generally have large and
[1]. A ductile fuse can be created to accommodate the plastic deforma- heavy sections, carry gravity loads and are difficult to be replaced. In in-
tion that is required for seismic energy dissipation while protecting novative seismic-resistant steel frames with dissipative fuses, repair
the beam-to-column joints. Structural systems that may be easily and work, if needed, is limited only to the replacement of the additional
economically repairable are considered as effective and attractive in steel plates.
high seismicity regions due to the reduction of possible repair costs. Within the “Fuseis” research project, experimental investigations on
the welded fuse behavior under cyclic loading were performed at
⁎ Corresponding author. Instituto Superior Tècnico of Lisbon and detailed results are extensively
E-mail address: marco.valente@polimi.it (M. Valente). reported in [2]. The work described in this paper presents the numerical

http://dx.doi.org/10.1016/j.jcsr.2016.09.003
0143-974X/© 2016 Elsevier Ltd. All rights reserved.
 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511 499

study provided useful information about the dimensions of both the

fuse system and the experimental specimens.

2.1. Finite element models

Detailed three-dimensional finite element models were developed

through the computer code Abaqus [9] in order to reproduce the behav-
ior of the fuse device under monotonic and cyclic loading and to per-
form parametric analyses. First, an introductory three-dimensional
finite element model of a steel beam with composite slab was created
to identify the mechanical and geometric parameters influencing the
cyclic performance of the fuse device, Fig. 2. Then, more refined three-
Fig. 1. Schematic representation of a fuse device placed in a moment-resisting composite
dimensional finite element models of a composite beam connected to
steel frame.
a column were developed to anticipate the behavior of the specimens
tested at Instituto Superior Tècnico of Lisbon, Fig. 2. The different flange
investigations carried out in support of the experimental campaign in plates of the fuse device used in the experimental tests were modeled
order to study the cyclic behavior of the welded fuse devices. Then, nu- and represented in Fig. 3. Finally, a three-dimensional finite element
merical analyses are performed to assess the effects of the application of model of a portion of a storey of a composite steel frame was used to cal-
welded fuse devices on the seismic performance of multi-storey com- ibrate and simulate the experimental tests performed at Politecnico di
posite steel frames. Milano, Fig. 4.
The fuse device behavior is studied by means of two different nu- The elements used in the different models are all chosen from those
merical approaches. First, the refined three-dimensional finite element available in ABAQUS/Standard library. All steel parts and concrete slab
modeling technique, for which the computational effort is very expen- were modeled using the solid (continuum) elements named C3D8R.
sive, is employed. Detailed finite element models using solid elements These elements are linear displacement interpolation solid elements
are developed to have a better understanding of the behavior of the with reduced integration. Reduced integration elements are chosen in
fuse system and to demonstrate that the whole plasticization occurs order to reduce computational time, which would be excessive in the
only in the device. Then, a lumped plasticity modeling approach is case of higher order elements. The mesh is finer in the parts of the
employed and global numerical models are created through a computer model where high stress concentrations are expected. The steel plates
code commonly used by structural engineers in order to analyze the of the fuse device are tied to the web and bottom flange of the beam
seismic response of different multi-storey composite steel frames through the weld parts. Shear connection between slab and beam and
equipped with fuse devices. Special non-linear link elements are used all other constraints are modeled by using the “tie” option. Using such
to simulate the non-linear behavior of the fuse derived from the results a constraint type, the degrees of freedom (rotational and translational)
of experimental tests. After the calibration of the models according to between the nodes of the two surfaces that are “tied” to each other are
the test results, non-linear dynamic and static pushover analyses are constrained to maintain the same values.
carried out to assess and compare the effectiveness of the fuse devices Structural steel and reinforcing rebars in the slab are assumed to be-
with different geometric and mechanical characteristics. have like an elastic-plastic material with hardening both in compression
and in tension and the Von Mises yield criterion is adopted. The steel
2. Preliminary numerical analyses of detailed finite element models grades S275 and A500 are used, respectively, for structural elements
and reinforcing rebars of the finite element model representing the
An extensive preliminary numerical study was carried out before the test specimens.
execution of the experimental tests of the “Fuseis” research project. The Numerical analyses and experimental tests showed that the behav-
aims of these introductory numerical analyses were: 1) to define the ior of the fuse was mainly dependent on the yielding and buckling of
main geometric and mechanical properties of both the devices and the the steel plates and no major cracking of the slab was observed. Conse-
specimens to be experimentally tested; 2) to identify the optimal con- quently, in preliminary numerical models concrete was modeled with
figurations for the welded fuse solution studying the influence of the an elastic behavior, thus reducing the high computational demand.
most relevant variables; 3) to anticipate the overall behavior of the Buckling of the flange plate was properly simulated at hogging rotation.
test specimens with fuse devices. A three-dimensional finite element In this study the finite element models were analyzed in terms of de-
idealization allows for the representation of the local details of the formed shape and equivalent plastic strain (PEEQ) contour plot. The
members geometry and its use is essential to detect stress concentra- qffiffiffiffiffiffiffiffiffiffiffiffi
tions and local damage patterns, [6–8]. This preliminary numerical equivalent plastic strain PEEQ is defined as PEEQ ¼ 23 εpij εpij and it

(a) (b)

Fig. 2. (a) Preliminary finite element model of a composite steel beam with fuse device; (b) Final proposal of the configuration of a beam-to-column sub-assemblage with fuse device.
500 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511

Fuse A Fuse B

Fuse C Fuse D

Fig. 3. Finite element models of the specimens used for experimental tests: dimensions of the web (yellow) and flange (blue) plates (dimensions in mm). (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)

represents the equivalent plastic strain accumulated during the dis- compared under sagging and hogging bending for different joint rota-
placement-controlled loading history. tions. Parametric analyses were performed to identify the influence of
some geometric and mechanical variables and to define effective config-
2.2. Preliminary parametric investigations of the fuse device urations for the fuse device. In particular, the cyclic behavior of the fuse
was analyzed changing selected geometric and mechanical parameters,
The initial simplified finite element model consisted of a composite such as the geometric slenderness and the resistance capacity ratio.
steel beam with an IPE240 section profile supporting a 150 mm thick re- The geometric slenderness λ of the flange plate is defined as:
inforced concrete slab. A welded fuse device was inserted in the beam
and a specifically detailed gap was created in the concrete slab in L0
order to allow for large plastic rotations in the fuse without causing λ¼ ð1Þ
t fuse
crushing in the concrete slab. Steel rebars were continuous over the
gap and were designed in order to force the plastic neutral axis to lie
within the slab thickness. A displacement-controlled cycling loading where L0 is the free buckling length of the fuse and tfuse is the thickness
history was applied at the beam edge and the cyclic response of the of the flange plate. The free buckling length L0 is given by the horizontal
fuse device was numerically investigated. The results obtained from distance between the edges of the lap welds on each side of the gap on
the simplified numerical models were qualitatively analyzed and the fuse section.

Fig. 4. Finite element model of a portion of a storey of a composite steel frame with different fuse devices.
 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511 501

Table 1 of the distance between the edges of the welds decreases the possibility
Main dimensions and geometric slenderness λ of the welded flange plates of the fuses. of the flange plate buckling.
Flange plate A B C D The third series of analysis was conducted on numerical models with
Thickness [mm] 10 10 12 8
different plate cross-sections obtained by means of an increase of the
Width [mm] 80 130 110 100 plate thickness. An enlargement of the cross-section of the flange
Cross-section area [mm2] 800 1300 1320 800 plate causes an increase of both the moment capacity and the energy
Geometric slenderness 17 17 14.1 21.25 dissipation of the fuse. However, large increases of the flange plate
cross-section are not suggested because they may alter the strength hi-
erarchy that prevents concentration of plastic deformation in the irre-
The resistance capacity ratio α of the fuse is defined as: placeable parts of the beam: in this case, the aim of the project is not
achieved because the whole plastic deformation is not concentrated
M max;fuse only in the fuse.
α¼ ð2Þ
Mpl;beam The preliminary analyses of the simplified numerical models proved
the effectiveness of the fuse devices, highlighted the influence of some
where Mmax,fuse is the maximum moment developed by the fuse device geometric and mechanical parameters, and concluded with the defini-
and Mpl,beam is the plastic resisting moment of the unreinforced area of tion of a suitable final configuration for the fuse device, see Fig. 2. A com-
the composite beam cross-section. prehensive description of the results of the preliminary numerical
The finite element model was subjected to imposed cyclic displace- parametric investigations is reported in [10].
ments applied at the beam edge and the overall behavior of the joints
was summarized by means of moment-rotation diagrams in order to 2.3. Numerical simulations of the beam-to-column sub-assemblage model
represent the hysteretic response. The cyclic behavior of the joints
was overall stable and ductile; the strength loss under hogging bending The results of the preliminary analyses of the simplified finite ele-
was caused by the buckling of the lower flange plate. The effects of the ment models of the composite steel beam were used to define different
variation of the plate thickness, the free buckling length and the plate suitable configurations of the fuse devices for the specimens of the lab-
section were investigated by performing parametric analyses. For the oratory tests. This section presents the results of the numerical simula-
sake of conciseness, in this paper only a qualitative description of the tions of the detailed finite element models representing the specimens
main results of these preliminary parametric analyses is reported. of the experimental tests. The different parts of the models were care-
The first series of analysis was performed on numerical models with fully created to match the same dimensions of the test specimens. The
the same free buckling length (L0 = 150 mm) of the welded plates and finite element model was a typical beam-to-column sub-assemblage,
the flange plate thickness was assumed as the variable parameter. The comprising a HEB 240 steel column and a composite beam with an
increase of the thickness was obtained maintaining the same cross-sec- IPE300 section profile supporting a 150 mm thick and 1450 mm wide
tion area of the plate by changing the plate width. The increase of the reinforced concrete slab. The fuse device consisted of steel plates
moment of inertia of the plates causes an enlargement of the cycle welded to the web and bottom flange of the beam with a specifically de-
and an enhancement of the plastic energy dissipated by the model. tailed gap in the concrete slab. The longitudinal steel rebars of the con-
The second series of analysis was carried out on numerical models crete slab were continuous over the gap in order to ensure transmission
with the same thickness of the plates and the free buckling length of of stresses. To assess the performance of the devices, monotonic and cy-
the welded plates was assumed as the variable parameter. An increase clic analyses were carried out on the models equipped with fuses with
of the free buckling length causes both the loss of stiffness and the de- different geometric properties. The same web plate was used for the
crease of the plastic energy dissipated by the model. Clearly, a reduction fuse device in all the models, while the geometric dimensions

Plate A Plate B

Plate C Plate D

Fig. 5. PEEQ contour plot for the models with different flange plates under sagging rotation (imposed top displacement = 80 mm).
502 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511

Plate A Plate B

Plate C Plate D

Fig. 6. PEEQ contour plot for the models with different flange plates under hogging rotation (imposed top displacement = 80 mm).

(thickness and width) of the flange plates were different. The same free the plates is achieved successfully. It is apparent that strength in the
buckling length, equal to 170 mm, was initially assumed for all the con- hogging configuration is controlled by buckling, which occurs as a func-
figurations of the fuses. Table 1 summarizes the dimensions (thickness tion of both the geometric properties of the flange plates and the free
and width) and the geometric slenderness of the welded flange plates buckling length of the fuse. Numerical results show that the hogging re-
of the fuses investigated in the numerical analyses. sistance of the fuse is more sensitive to a geometry variation of the
The results obtained from the numerical simulations carried out flange plates than the sagging resistance. Buckling of the flange plate
through a displacement-controlled increasing (monotonic) loading his- of the fuse is clearly observed for the models with plates A and D,
tory applied at the top edge of the beam are reported in this study. while in the models with plates C and B buckling of the flange plate
Fig. 5 shows the deformed shape and the PEEQ contour plot for the under hogging rotations is less evident.
models with different plates under sagging rotations (imposed top dis- When sagging and hogging behavior is compared, a more significant
placement = 80 mm). In all the models plastic deformations concen- loss of stiffness occurs under hogging rotation due to buckling of the
trate in the web and flange plates, ensuring that the plastic hinge fuse plate. The severity of yielding and buckling of the plates has a fun-
occurs at the fuse section and protecting the remaining parts of the damental influence on the performance of the fuse device.
joint. High values of PEEQ and stress concentrations are registered The effects of different free buckling lengths were also investigated
near the welds of the web plates and in the middle of the flange plates. in the beam-to-column sub-assemblage models. Fig. 7 shows the PEEQ
Typical failure modes observed in the specimens tested in laboratory contour plot at sagging (left) and hogging (right) rotations (imposed
consisted of the development of cracks at the mid-section of the flange top displacement = 80 mm) for the model with flange plate A and
plate under tension. The models with the weakest flange plates exhibit free buckling length equal to 140 mm. Comparing the results with the
the highest values of PEEQ in the fuse devices. model with the same plate and free buckling length equal to 170 mm,
Fig. 6 presents the deformed shape and the PEEQ contour plot of the it can be noted that buckling of the flange plate is less evident, plastic
fuse device under hogging rotation (imposed top displacement = deformations are concentrated in a smaller region of the plates and
80 mm) for the different models. The results of the analyses indicate higher values of PEEQ are registered in the fuse device.
that both the column and the composite beam remain in the elastic Fig. 8 presents the PEEQ contour plots in the models with plates B
range for all the different plates. The pronounced curved shapes of the and D under sagging and hogging rotations at the onset of plastic defor-
plates indicate that the attempt of modeling the buckling behavior of mations. In all the models the plastic deformations involve the flange

Fig. 7. PEEQ contour plot for the model with plate A and free buckling length equal to 140 mm under both sagging (left) and hogging (right) rotations (imposed top displacement = 80
mm).
 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511 503

Plate B Plate B

Plate D Plate D

Fig. 8. PEEQ contour plot for the models with plates B and D under both sagging (left) and hogging (right) rotations at the onset of plastic deformations.

plates and the bottom part of the web plate. In the case of sagging rota- sub-assemblage and the PEEQ contour plot near the beam-to-column
tions, uniform distributions of plastic deformation are observed in the connections. The results obtained with the frame model confirm the
flange plate. In the case of hogging rotations, plastic deformations con- outcomes observed in the beam-to-column sub-assemblage models.
centrate near the end of the weld and in the central part of the flange The objectives of the “Fuseis” research project were achieved because
plate. all the potential damage was concentrated only in the fuses. In fact, plas-
tic deformations were registered only in the device regions, while both
2.4. Numerical simulations of the frame sub-assemblage model the columns and the composite beams remained in the elastic range.
Buckling of the flange plate can be observed in the internal beam-to-col-
The finite element model of the fuse was then inserted in a portion of umn connection under hogging rotation.
a storey of a composite steel frame in order to anticipate the behavior of
the specimen tested at Politecnico di Milano. The frame consisted of two 3. Experimental tests and calibration of simple models of the fuse
HEB240 steel columns, two IPE300 steel beams and a 150 mm thick re- device
inforced concrete slab, Fig. 4. Welded fuse devices were included in this
preliminary numerical model of the frame sub-assemblage, though the 3.1. Experimental tests
experimental tests were carried out with bolted plates. In the finite ele-
ment model the columns bottom edges were restrained with hinges After the preliminary numerical analyses carried out with refined fi-
and displacements were imposed at the columns top edges. Fig. 9 nite element models, within the “Fuseis” research project, laboratory
shows both the deformed shape of the numerical model of the frame tests were performed on beam-to-column sub-assemblages at Instituto

Fig. 9. Deformed shape and PEEQ contour plot in the frame sub-assemblage model (imposed top displacement = 80 mm).
504 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511

Fig. 10. Experimental test set-up and fuse device configuration.

Superior Tècnico of Lisbon in order to characterize the experimental be- unwelding the damaged plates and welding the new ones. A compre-
havior of the welded fuse devices. Fig. 10 shows the experimental test hensive description of the results of the extensive experimental cam-
set-up, the specimen and the fuse device configuration. The dimensions paign can be found in [2].
and the geometric slenderness of the welded flange plates of some fuses
investigated in the experimental tests are reported in Table 1. The 3.2. Development and calibration of simple numerical models
beam-to-column sub-assemblages equipped with welded flange plates
E and F were the last specimens to be tested during the experimental Simplified numerical models of the beam-to-column sub-assem-
campaign and they have not been considered in this study for the cali- blages tested at Instituto Superior Tècnico of Lisbon were developed
bration of the numerical models because the experimental results ob- by using the computer code SAP2000 [11]. The numerical models
tained from these plates were significantly affected by the effects of were later used to further investigate the effects of the application of
accumulated damage induced by previous tests. the fuse devices on the seismic response of different multi-storey com-
The laboratory tests were conducted by applying cyclic variable am- posite steel frames. The models consisted of a composite steel beam,
plitude displacements at the top of the beam of the specimens. The ex- equipped with fuse devices, connected to a column with the same ge-
perimental tests confirmed the main results obtained through the ometry used in the experimental test set-up. The potential non-linear
preliminary numerical analyses on detailed finite element models. The behavior of the beam was simulated with a lumped plasticity approach
proposed devices were able to concentrate plastic deformations in the by defining a non-linear plastic hinge at the beam end connected to the
steel plates and to dissipate large amounts of plastic energy through sta- column. The fuses were modeled as non-linear link elements with a
ble cyclic behavior. The failure modes of the cyclically tested specimens length equal to the free buckling length of the device. The link element
consisted of the development of cracks at the mid-section of the flange is a non-linear spring with six independent internal deformations for
plate under tension, whereas both the column and the composite beam which a non-linear generalized force-deformation relationship can be
remained in the elastic range. The devices could be replaced by defined. The multi-linear plastic pivot model was used as hysteresis

Fig. 11. Comparison of experimental and numerical moment-rotation diagrams of the welded fuses tested at Instituto Superior Tècnico of Lisbon and investigated in this study.
 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511 505

Fig. 12. Resistance capacity ratio (left) and elastic stiffness (right) for sagging and hogging rotations for the welded fuses with different flange plates.

rule including asymmetrical cross-section behavior with pinching and The values of the resistance capacity ratio and elastic stiffness of the
strength degradation. The link behavior was defined by a moment-rota- welded fuses utilized in this study are reported in Fig. 12, for sagging
tion curve characterized by positive and negative moment capacities and hogging rotations. The highest resisting moments were assumed
and initial stiffness of the fuse device. The non-linear behavior was for plates B and C because of the large resistant sections of the plates.
assigned only to the rotational degree of freedom of the link with re- Some discrepancies between numerical and experimental values of
spect to the major axis of inertia. The constitutive law adopted for the elastic stiffness were observed for some fuses. In this study it was decid-
non-linear link was able to represent the dissipated energy, the stiffness ed to adopt the values of initial stiffness that more accurately approxi-
and the maximum moment of the fuse during the cyclic loading history. mate the experimental results for the following analyses on multi-
The initial input parameters of the monotonic moment-rotation dia- storey composite steel frames.
gram of the fuses were obtained from simplified analytical models de- After the calibration of the numerical models of the welded fuses
scribed in [2]. The computations of the resistance and stiffness values tested at Instituto Superior Tècnico of Lisbon, another moment-rotation
for all fuses were based on the material properties measured during curve, hereafter denoted as plate M, was adopted for the welded fuse
the experimental campaign. Then, the models were calibrated and re- devices in this study. The input parameters of the monotonic mo-
fined by using the results of the experimental tests carried out on the ment-rotation curve for plate M were obtained from the 3D FE models.
specimen sub-assemblages. Consequently, the initial mechanical characteristics of the fuse were not
The comparison of the experimental and numerical moment-rota- affected by the deterioration of the specimen sub-assemblage observed
tion diagrams of the welded fuses is shown in Fig. 11. The models during the experimental tests. The numerical moment-rotation diagram
were able to accurately capture the non-linear behavior of the different assumed for plate M and the monotonic curve derived from the 3D FE
devices in terms of stiffness, resistance, ductility and dissipated energy. model are shown in Fig. 13. The main characteristics (resistance capac-
The moment-rotation diagrams show that the hysteretic behavior of the ity ratio and elastic stiffness) of the numerical model of plate M, along
fuses was overall stable and dissipative. It was also characterized by with all the other welded fuses, are reported in Fig. 12. Moreover, Fig.
pinching effects on the hysteresis loops of cycling loading due to buck- 13 compares the numerical moment-rotation curve of plate M with
ling of the fuse plates under hogging rotation. The buckling of the fuse those of the other plates. It can be noted that plate M presents the
plate and the presence of the concrete slab explain the asymmetry of highest resistance capacity ratio and elastic stiffness for sagging
the diagram in terms of moments. All fuses were able to achieve rota- rotation.
tions at least equal to 35 mrad, which is the minimum value recom-
mended by EN 1998 for structures of high ductility class [12]. The 4. Steel frames under study and numerical modeling
experimental and numerical results clearly show the different behavior
between plates with extreme values of the resistance capacity ratio Three different multi-storey steel frames with composite beams are
(plates A–D and plates B–C, respectively), mainly due to the pinching analyzed in this study in order to understand the effects of the applica-
phenomenon and to the differences in strength, which directly affect tion of the fuse devices on different types of structures. The seismic re-
the energy dissipation capacity. A more pronounced stiffness degrada- sponse of the frames with and without fuse devices is evaluated and a
tion with cycling was registered for devices with lower values of the re- comparison of the seismic performance of the innovative seismic-resis-
sistance capacity ratio, especially under sagging rotation. Moreover, tant steel frames with different fuse plates is carried out too.
flange plates A and D proved to be more sensitive to buckling due to The frames under study were extracted from composite steel build-
their geometry under hogging rotation, showing more marked pinching ings with different number of storeys. Dead loads consist of the weights
effects. of structural components and partitions, and live loads are assumed to

Fig. 13. Numerical moment-rotation diagram for plate M: comparison with the 3D finite element model and with other plates.
506 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511

be equal to 2 kN/m2 for all storeys. Storey masses include dead loads and
a percentage of live loads (30% according to Eurocode 8 for common res-
idential and office buildings). The steel grade used for all the structural
members is S275. The three frames consist of three bays and different
number of storeys ranging from three to nine. The storey height and
the bay width are equal to 3.5 m and 5 m, respectively. The column sec-
tion dimensions change as a function of the number of storeys of the
frames and range from HE320B for the three-storey frame to HE400B
for the nine-storey frame. The composite beams present an IPE300 sec-
tion profile supporting a 150 mm thick reinforced concrete slab in ac-
cordance with the specimen of the experimental test set-up. The
geometric dimensions and section profiles of the different frames
under study are illustrated in Fig. 14.
Numerical models of both the conventional steel frames without
fuse devices and innovative seismic-resistant steel frames with fuse de-
vices were developed using the computer code SAP2000. Hereafter,
conventional frames are denoted as “F1 − n” and innovative seismic-re- Fig. 15. Response spectra of the artificial accelerograms.

sistant frames are denoted as “FPi − n”, where n and i indicate the num-
ber of the storeys of the frame and the plate type used in the model, assumed in the model in accordance with the geometry of the experi-
respectively. mental specimens.
A lumped plasticity modeling approach was employed for the non-
linear models of the frames. Beam and column elements were modeled 5. Numerical analyses and results
as frame elements and non-linearity was concentrated in plastic hinges
at their ends. To characterize the non-linear behavior of a plastic hinge, The seismic performance of the frames under study was assessed
the force-displacement properties suggested in FEMA 356 [13] were im- through non-linear dynamic analyses with response spectrum-compat-
plemented. Flexural moment hinges were adopted for the beams, while ible artificial accelerograms. The set of artificial ground motions
plastic hinges accounting for the interaction between axial force and consisted of three different records that were generated so as to
bending moment were defined for the columns. Beam-to-column joints match the Eurocode 8 response spectrum (Type 1, soil type A, 5% vis-
were considered as rigid in accordance with the connection detailing of cous damping) using the computer code SIMQKE [14]. The response
the experimental tests. spectra of the artificial accelerograms used in this study are shown in
The models of the different fuse devices were implemented in the nu- Fig. 15. The non-linear dynamic analyses were performed for three dif-
merical models of the frames. The length of the beams was subdivided ferent peak ground accelerations (PGA = 0.3g, PGA = 0.5g and PGA =
into different elements in order to take into account both the presence 0.7g) of the artificial earthquake records. In addition, non-linear static
of the fuse devices and the part of the beam reinforced with additional (pushover) analyses were performed in order to assess the distribution
welded plates. The fuses were modeled as non-linear link elements of plastic hinges and to check the possible collapse mechanisms.
inserted in the beams with a length equal to the free buckling length of
the welded plates. The multi-linear plastic pivot model was used as hys- 5.1. Top displacement and base shear
teresis rule for the fuses. The values of the parameters used for the hys-
teretic model were obtained from the calibration of the models of the The results of the non-linear dynamic analyses in terms of maximum
specimens tested at Instituto Superior Tècnico of Lisbon. roof displacement and base shear are presented using the average
The part of the beam reinforced with additional plates, aimed at values obtained through the three artificial records.
avoiding spreading of plasticity into the connection, was reproduced The maximum top displacements of the different frames are shown
in the numerical models by using different cross-sections and plastic in Fig. 16 for different peak ground accelerations. As can be noted, the
hinges properties around the device. The length of these regions was maximum top displacements of the innovative frames with fuses are

Fig. 14. Steel frames under study.

 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511 507

Fig. 16. Maximum top displacement of the frames under study for different peak ground accelerations.

larger than those of the conventional frames because the insertion of the are reported as average values of the ones obtained through the non-
fuse devices has the consequence of reducing the lateral stiffness of the linear dynamic analyses using different accelerograms. The maximum
frames. values of the base shear of the innovative frames with fuses are always
Comparing the seismic performance of the different innovative smaller than those of the conventional frames. The decrease of the base
frames, it can be argued that the initial stiffness and resisting moments shear of the innovative frames depends mainly on the resistance capac-
adopted for the constitutive laws of the fuse devices affect the results. In ity ratio of the fuse devices. The highest values of the base shear are ob-
the case of seismic intensity levels equal to PGA = 0.3g, frames FPM and served for the frames equipped with fuses with the highest resisting
FPD generally exhibit smaller top displacements because the moment- moments (frames FPM, FPB and FPC). The innovative frames FPA and
rotation curves adopted for the fuse plates M and D present higher elas- FPD present large decreases of base shear.
tic stiffness. In the case of seismic intensity levels equal to PGA = 0.7g,
smaller top displacements are registered for frames FPM, FPB and FPC 5.2. Plastic hinge pattern and energy dissipation
because the moment-rotation curves adopted for the fuse plates M, B
and C present higher resisting moment. In the case of severe ground The distribution of plastic hinges is evaluated for conventional and
motions (PGA = 0.7g), the conventional frames and the innovative innovative frames at the end of the non-linear dynamic analyses. Fig.
frames FPM show similar values of top displacements for all the differ- 18 compares the plastic hinge patterns in conventional frames F1 and
ent frames. Numerical results highlight that a significant reduction of in frames with fuses FPM under Acc-1 with PGA = 0.7g. As can be
the mechanical characteristics of the fuse (initial stiffness and resisting noted, the number and the distribution of plastic hinges are similar for
moments) can be detrimental to the seismic performance of the innova- the two types of structures.
tive frames in terms of top displacements. In particular the innovative In conventional frames plastic deformations occur at the beam ends
frame FPA presents large increases of top displacements for different and at the base of the ground level columns. This result reflects the de-
seismic intensity levels. sign of the frames according to the provisions of Eurocode 8. Neverthe-
Fig. 17 shows the maximum values of the base shear of the frames less, for high values of peak ground acceleration some plastic hinges are
for different peak ground accelerations. As already explained, the results sometimes observed at columns ends too. For instance, the three-storey

Fig. 17. Maximum base shear of the frames under study for different peak ground accelerations.

F1-3 FPM-3 F1-6 FPM-6 F1-9 FPM-9

Fig. 18. Plastic hinge patterns in conventional frames (F1) and innovative frames (FPM) under Acc-1 (PGA = 0.7g).
508 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511

Fig. 19. Plastic energy dissipated in conventional frames (F1) and in innovative frames with fuses (FPA, FPB, FPC, FPD, FPM) under PGA = 0.7g.

Fig. 20. Plastic energy dissipated in conventional frames (F1) and in innovative frames (FPM) under different peak ground accelerations.

frame F1–3 presents plastic hinges both at the base of the ground level dissipated by beams in conventional frames. The ground level columns
columns and at the second level columns. of innovative frames dissipate smaller amounts of hysteretic energy
Comparing the results with conventional frames, it can be argued than those of conventional frames for severe seismic excitations
that in all cases the application of the fuse devices prevents the forma- (PGA = 0.5g and PGA = 0.7g). The total amount of dissipated energy
tion of plastic hinges in columns, except at the columns base, and the is always larger in innovative frames with fuses than in conventional
possible activation of a weak-storey collapse mechanism. In innovative frames.
frames plastic deformations concentrate only in the fuses and the beam As well known, the global ductility of the frames significantly de-
ends are protected remaining in the elastic range. The fuses act as dissi- pends on the available local ductility. High inelastic deformations and
pative devices, preventing the spreading of damage into the other struc- large amounts of plastic energy dissipation require high values of rota-
tural elements and the pattern of yielding is in perfect agreement with tion capacity of the plastic hinge regions. Fig. 21 indicates the maximum
the global mechanism. Finally, numerical results show that no kinemat- values of rotations computed in beams and fuses of the first storeys for
ic mechanism is observed for all the frames under study subjected to the different frames under study subjected to severe seismic excitation
accelerograms with PGA = 0.7g. (PGA = 0.7g). As can be noted, the maximum rotations registered in
Fig. 19 compares the amount of plastic energy dissipated in conven- fuses of innovative seismic-resistant frames are larger than those ob-
tional and innovative seismic-resistant frames with different plates served in beams of conventional frames. For all the different frames
under severe seismic excitation (PGA = 0.7g). The plastic energy dissi- the maximum value of rotation is registered for the fuse A, which ex-
pated in columns, beams and fuses is reported; it is computed as the hibits the smallest rotational stiffness; the fuse M presents the smallest
area enclosed inside all the cycles during the earthquake excitation. As maximum rotation among the different fuses. The deformation capacity
can be noted, the total amount of hysteretic energy dissipated in inno- of the fuses was demonstrated by the results of the experimental tests
vative frames with fuses is higher than in conventional frames. This re- carried out at Instituto Superior Tècnico of Lisbon, where all specimens
sult can be explained by the higher values of hysteretic energy showed stable hysteresis loops and were able to achieve rotations equal
dissipated by fuses in innovative frames than by beams in conventional to 35 mrad. The maximum rotations attained by the fuses of the innova-
frames. The plastic deformation demand and the plastic energy dissipat- tive frames in non-linear dynamic analyses were about 30 mrad, a value
ed in columns slightly decrease when fuse devices are present. More- for which the fuses of the experimental tests showed good energy dissi-
over, the resistance capacity ratio of the fuses affects the plastic pation capacity.
energy dissipated in fuses and columns of innovative frames.
The largest amount of plastic energy is dissipated by frames FPM and
FPC for all the different structures. This result confirms the dissipative 40
behavior of these plates shown by the experimental tests. The energy F1 FPA FPB FPC FPD FPM
dissipated by frames FPD is smaller than the other innovative frames be-
Rotation [mrad]

30
cause of the low energy dissipation capacity of this type of fuse due to
buckling of the flange plate. The reduction of the plate thickness may
lead to pinching effects on the hysteretic loops characterizing the cyclic 20
behavior of the fuse. This entails a loss of dissipated energy that could be
prevented by reducing the slenderness of the flange plate.
10
Fig. 20 compares the plastic energy dissipated in conventional
frames and in innovative frames with fuse FPM for different peak
ground accelerations. In the case of seismic intensity levels equal to 0
PGA = 0.3g, the innovative frames dissipate small amounts of plastic F-3 F-6 F-9
energy in fuses, whereas the conventional frames practically remain in
the elastic range. For each level of peak ground acceleration, the plastic Fig. 21. Maximum rotations in beams (for frames F1) and fuses (for innovative frames FPA,
energy dissipated by fuses in innovative frames is larger than that FPB, FPC, FPD, FPM) under PGA = 0.7g.
 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511 509

Fig. 22. Base shear and top displacement corresponding to the activation of the plastic mechanism for the different frames from pushover analyses.

5.3. Collapse mechanism assessment location of dissipative zones and the type of post-elastic response
in these zones. Global collapse mechanisms with stable and dissipa-
Non-linear static (pushover) analyses were performed in order to tive plastic hinges within all the fuses and at the column bases are
study the plastic collapse mechanisms of the different frames. The push- preferred due to the high energy dissipation capacity of the frame.
over analyses were conducted under monotonically increasing horizon- It should be observed that an excessive increase of the resistance ca-
tal loads taking into account both geometric and mechanical non- pacity ratio of fuses close to unity may generate some plastic hinges
linearities and confirmed the reduction of the lateral stiffness of the in- at the column ends also in innovative frames.
novative frames with fuses when compared with the conventional
frames.
Fig. 22 shows the values of base shear and top displacement corre- 5.4. Effects of different locations of the fuse devices
sponding to the activation of the plastic mechanisms for the different
frames under study. The values of base shear obtained from pushover The efficiency of the fuse devices depends on proper design rules,
analyses are in a good agreement with those obtained from non-linear which include the location of the fuse along with the resistance capacity
dynamic analyses for all the structures analyzed. As expected, the inno- ratio. The effects of three different locations of the fuse devices on the
vative seismic-resistant frames with dissipative fuses present lower seismic response of the innovative frames were investigated for the
values of base shear than conventional frames. The decrease of the six-storey frames with plates A, C and M. Numerical models of the
base shear in the innovative seismic-resistant frames depends on the frames with different distances of the fuse (L1 = 25 cm, L2 = 35 cm,
mechanical characteristics of the fuse devices. A reduction of the bend- L3 = 50 cm) from the column face were created and analyzed through
ing moment capacity of the fuses causes a decrease of the base shear of non-linear dynamic analyses. The results of the numerical analyses con-
the innovative frames. firm that the location of the fuse device is an important design parame-
Moreover, it can be argued that innovative frames with fuses charac- ter influencing the seismic response of the frames. When the fuse
terized by higher values of the resistance capacity ratio generally pro- distance measured from the column face increases, the top displace-
vide higher performance levels in terms of ultimate top displacements ments generally decrease and the base shear enhances, as shown in
than conventional frames. Only for the innovative frames FPA-3 and Fig. 23 and Fig. 24. Moreover, the formation of the first plastic hinges
FPA-6 the plastic mechanism is achieved earlier than the conventional in the devices is deferred and occurs early at the column bases. It can
frames. In all other cases the top displacements corresponding to the be noted that the variation of the fuse location has smaller effects on
development of the plastic mechanism are much larger for the innova- top displacements for innovative frames FPM than for other innovative
tive frames with fuses than for conventional frames. The innovative frames.
seismic-resistant frames with fuses characterized by the highest resis- However, it is necessary to avoid too large distances of the fuses from
tance capacity ratios (FPM, FPC and FPB) present the largest values of ul- the column face, since damage would be more easily concentrated in
timate top displacements. the irreplaceable parts of the frame, such as beams and columns, than
The collapse mechanisms obtained from pushover analyses are in the dissipative devices. Fig. 25 shows a decrease of plastic energy dis-
similar for conventional frames and innovative frames with dissipa- sipated in fuses and the appearance of damage in beams for the case of
tive fuses. A global collapse mechanism involving the beams and the fuses with distance equal to L3 = 50 cm.
bases of the ground level columns is registered for both the types of It is apparent that the location of the fuse should be chosen as a
frames. Nevertheless, in conventional frames some plastic hinges are function of the resistance capacity ratio of the device. The results of
observed in the columns of different storeys. On the contrary, in in- the analyses show that a satisfactory seismic performance can be
novative frames with dissipative fuses plastic hinges are located achieved selecting a distance of the fuse equal to about the beam
only in the fuses and at the base of the ground level columns. As depth and values of the resistance capacity ratio within the range
well known, it is highly desirable in seismic design to control the of 0.6–0.7.

Fig. 23. Maximum top displacement of the six-storey frames (F1, FPA, FPC, FPM) with different locations of fuses for different peak ground accelerations.
510 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511

Fig. 24. Maximum base shear of the six-storey frames (F1, FPA, FPC, FPM) with different locations of fuses for different peak ground accelerations.

6. Conclusions – The energy dissipation capacity of the innovative frames is larger

than that of the conventional frames. This result can be explained
In this paper an innovative fuse device for dissipative beam-to-col- by the higher values of hysteretic energy dissipated by the fuses in
umn connections applicable to multi-storey composite steel buildings innovative frames than by beams in conventional frames. The plastic
has been investigated through numerical simulations. The fuse device deformation capacity of the fuses is demonstrated by the results of
consists of steel plates welded to the web and bottom flange of the the experimental tests in which all specimens show stable hysteresis
beam with a specifically detailed slab gap, providing the advantage of loops and are able to achieve the minimum rotations prescribed by
replacement in case of damage after severe earthquakes. the code. The use of fuse devices in composite steel frames reduces
the plastic deformation demand in columns.
– Preliminary numerical investigations carried out by means of de- – The collapse mechanisms of the innovative frames with fuses are
tailed finite element models of beam-to-column sub-assemblages global, ductile and very dissipative, involving all the fuses and the
show the effectiveness of the fuses. Energy dissipation and plastic bases of the ground level columns. Plastic deformations and damage
deformations concentrate only in the devices, without any signifi- concentrate in the fuse devices that are easily replaceable after se-
cant damage in the other structural elements. The effects of some vere ground motions. The top displacements corresponding to the
design parameters on the cyclic response of the fuse devices are collapse mechanisms are larger in innovative frames with fuses
evaluated to support the following experimental tests. These prelim- than in conventional frames.
inary results validate the initial hypothesis for the fuse and confirm – The efficiency of the fuse devices depends on the implementation of
that both the steel beam and column remain in the elastic range. proper design rules, which include the choice of the resistance ca-
The most relevant aspects of the fuse behavior observed in the fol- pacity ratio and the location of the fuse. The comparison of the seis-
lowing experimental tests are conveniently reproduced by the finite mic performance of the innovative frames with different fuse
element model. The hysteretic behavior of the fuse is mainly affected devices highlights the importance of the resistance capacity ratio.
by buckling of the flange plates under hogging rotation. Fuse devices with values of the resistance capacity ratio within a
– Simple numerical models of the dissipative beam-to-column con- range of 0.6–0.7 generally provide high seismic performance levels
nections are developed in order to assess the effects of the applica- in terms of top displacements and ductility, preventing plastic
tion of the fuse devices on the seismic response of multi-storey hinge formation in beams and columns. Higher values of the resis-
composite steel frames. The properties of the non-linear link ele- tance capacity ratio of the fuses entail higher resistant capacities of
ments for the different types of fuse devices are first derived from the frame, but don't guarantee a complete safeguard of the structur-
simple analytical models and then calibrated in order to fit the re- al elements from damage. Lower values of the resistance capacity
sults of the experimental tests carried out on beam-to-column sub- ratio of the fuses are more efficient to prevent the spreading of dam-
assemblages. age in the irreplaceable parts of the frame, but can undermine the
– The insertion of fuse devices reduces the lateral stiffness of multi- seismic performance of the frame mainly in terms of top displace-
storey composite steel frames. A decrease of the maximum base ments increase. The distance of the fuse from the column face should
shear and an increase of the top lateral displacements are observed be chosen as a function of the resistance capacity ratio of the device.
for the innovative frames with dissipative fuses when compared Numerical results show that satisfactory global seismic performance
with the conventional frames in non-linear dynamic analyses. A can be achieved selecting a fuse distance equal to about the beam
proper choice of the mechanical characteristics of the fuses may depth and values of the resistance capacity ratio within the range
limit the enhancement of the maximum lateral drifts. of 0.6–0.7.

Fig. 25. Plastic energy dissipated in six-storey frames (FPA, FPC, FPM) with different locations of fuses under PGA = 0.7g.
 M. Valente et al. / Journal of Constructional Steel Research 128 (2017) 498–511 511

Acknowledgments [6] G. Vasdravellis, M. Valente, C.A. Castiglioni, Behavior of exterior partial-strength

composite beam-to-column connections: experimental study and numerical simu-
lations, J. Constr. Steel Res. 65 (1) (2009) 23–35.
The studies reported in this paper were conducted within the scope [7] G. Vasdravellis, M. Valente, C.A. Castiglioni, Dynamic response of composite frames
of the FUSEIS (Dissipative Devices for Seismic Resistant Steel Frames, with different shear connection degree, J. Constr. Steel Res. 65 (10–11) (2009)
2050–2061.
reference RSFR-CT-2008-00032) research project, financed by the Re- [8] D. Giannuzzi, R. Ballarini, A. Hucklebridge, M. Pollino, M. Valente, Braced ductile
search Fund for Coal and Steel, of the European Commission. shear panel: new seismic-resistant framing system, Journal of Structural Engineer-
ing ASCE 140 (2) (2014) 1–11.
[9] Simulia, ABAQUS Theory Manual, USA, 2014.
References
[10] I. Mazza, F. Pedrazzoli, Numerical Modeling for Innovative Type of Seismic Moment
Resistant Frames with Dissipative, Easy Replaceable, Joints(MSc Thesis) Politecnico
[1] A. Plumier, The dogbone: back to the future, Engineering Journal (New York) (1997) di Milano, 2008.
61–67. [11] SAP2000 Static and dynamic finite element analysis of structures. Computers and
[2] L. Calado, J.M. Proenca, M. Espinha, C.A. Castiglioni, Hysteretic behavior of dissipative Structures Inc., Berkeley, California.
welded fuses for earthquake resistant composite steel and concrete frames, Steel [12] EN 1998, Eurocode 8: Design of Structures for Earthquake Resistance, 2004.
Compos. Struct. 14 (6) (2013) 547–569. [13] FEMA 356, Prestandard and Commentary for the Seismic Rehabilitation of Buildings,
[3] C.A. Castiglioni, A. Kanyilmaz, L. Calado, Experimental analysis of seismic resistant Federal Emergency Management Agency, Washington DC, USA, 2000.
composite steel frames with dissipative devices, J. Constr. Steel Res. 76 (2012) 1–12. [14] E.H. Vanmarcke, C.A. Cornell, D.A. Gasparini, S. Hou, SIMQKE: A Program for Artificial
[4] L. Calado, J.M. Proenca, M. Espinha, C.A. Castiglioni, Hysteretic behavior of dissipative Motion Generation, Civil Engineering Department, Massachusetts Institute of Tech-
bolted fuses for earthquake resistant steel frames, J. Constr. Steel Res. 85 (2013) nology, Massachusetts, 1976.
151–162. [15] M. Dicleli, A. Mehta, Efficient energy dissipating steel-braced frame to resist seismic
[5] D. Danai, G. Dougka, I. Vayas, Seismic behavior of frames with innovative energy dis- loads, Journal of Structural Engineering ASCE 133 (7) (2007) 969–981.
sipation systems (FUSEIS1-2), Eng. Struct. 90 (2015) 83–95.