You are on page 1of 11

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

Resistance of steel connections to low-cycle fatigue


A. Plumier & M.R.Agatino
Department of Civil Engineering, University of Liege, Belgium

A. Castellani, C.A. Castiglioni & C. Chesi


Department of Structural Engineering, Politecnico di Milano, Italy

Keywords: Earthquake, Steel, Welding, Frame, Connection, Ductility, Fatigue, Experimental Tests
ABSTRACT: following the northridge and Kobe earthquake, attention of researchers has been focused on the problem of the brittle collapse of beam to column connections in moment resisting
steel frames. large evidence has been given by the earthquake to this phenomenon which, although
not involving the frame global collapse, drastically changes the expectations about energy dissipation and ductile behaviour of connections. the paper presents an experimental testing program
which has been planned with the purpose of contributing to a deeper knowledge of the following
problems: the low cycle fatigue behaviour of welded joints and the implications of brittle failure in
welded connections on the global frame response.
1 INTRODUCTION
In the variety of cases of structural damage which were observed after the Northridge earthquake, a
most prominent one occurred in steel moment resisting frames (Bertero et al. 1994, Tremblay
1995); although not reaching collapse, this kind of frames presented brittle failures at moment connections in welded beam to column joints. The obvious relevance of this kind of behaviour has led
to a detailed mapping of such cases. Although similar failure situations had already been observed
in previous earthquakes and during laboratory experiments, the Northridge earthquake has given
special evidence to the problem of brittle failure of welded joints and to the need of clarifying the
conditions under which this kind of failure is likely to be induced by seismic actions. This implies
the study of the low cycle fatigue behaviour of beam to column connections. The same problem is
being addressed as it is of particular interest, at present, in the European Community Countries as
well, where common codes have recently been developed covering the various design fields,
among which the design of steel structures (CEN, Eurocode 3, 1994) and the design of buildings in
seismic areas (CEN, Eurocode 8, 1994). Codes are now subject to public debate and contributions
should be provided on critical design problems, like the case of welded joints subject to seismic
actions. In this context, funding has been accorded by the European Community for a wide project,
named Steelquake, which is under way at the moment and which is presented in this paper. The
Steelquake project concerns the analysis of the behaviour of civil engineering steel structures, of
the moment resisting category, under earthquake loads. The overall objective is to provide a better
insight into the actual behaviour of civil engineering steel structures, of the moment resisting category, under earthquake loads, implementing in such an analysis a quantitative reference to the failure of connections in terms of low cycle fatigue, considering the duration of the earthquake and the
corresponding number of cycles supported in dissipative (plastic) zones. The structural behaviour
factor characterising the structure (the so-called "q" factor) presently proposed in Eurocode 8, is not
related to the fatigue resistance of the potential failure zones, that is beam to column connections.
The aim is to allow the full design (weld size, thickness of components, dimensions) of steel frame
connections, taking into account the duration of the earthquake and the corresponding number of
cycles to be supported in dissipative (plastic) zones, that is to introduce q factors depending on the
considered earthquake time-history. A classification of connections, in terms of q values, is ex-

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

pected as a result of the project research. In order to achieve the above-said objective, full scale
testing of beam to column connections and testing on welds are carried out on different specimens
in Milan and Liege; shaking table tests at Athens and at ISMES Bergamo and pseudo-dynamic tests
at JRC Ispra are performed to provide complementary experimental data concerning the global behaviour of steel frames; parallel numerical analyses are performed with an aim to deeply interpret
the experimental behaviour both of the investigated single connections and steel frames.
The project has been conceived with the purpose of providing some contributions to two existing problems: 1) a clear understanding of the real behaviour of welded beam to column connections, a problem which still deserves accurate investigation despite of the work done in this area by
several researchers (Plumier 1994, Bernuzzi et al. 1997) and 2) the interpretation of the influence
of the connection failure on the global frame answer to seismic actions; in the Northridge earthquake no remarkable effect was noticed on the global frame behaviour as a consequence of joint
collapse. In this paper, only the research activities related to phase 1 is presented.
2
2.1

BEAM TO COLUMN CONNECTIONS AND WELDING PROCEDURES


Specimens selection and test set-up

The first part of the project, discussed here, is focused on the experimental response of welded
connections, for which the low cycle fatigue problem is investigated. The specimen selection aimed
at covering a significant variety of beam to column connections, frequently adopted in both European and American practice. Four typologies, as shown in Figure 1, were identified for this purpose, based on welded joints (type C) or partially welded joints (type A). All of these provide interesting cases; note, in particular, connection type A2, based on the concept of shifting the plastic
hinge zone away from the node, a joint design approach which has been recently proposed in the
U.S., and for which some results are already available, based on work of Ballio & Castiglioni
(1994).

Figure 1. Beam to column connection typologies without (left) and with (right) node reinforcing plate.

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

Figure 2. Different welding procedures C1 and C3: with steel backing bar; C2: with polyester backing bar.

Figure 3. Beam to column connection


full scale specimen.

Figure 4. Full scale-small size specimens


for weld size testing.

The presence of stiffening plates in the panel zone of the column is also to be investigated; for
this purpose, each of the above typologies include one additional specimen, with a node stiffening
plate. The research, however, is mainly focused on the beam-column connection behaviour; for this
reason, an external T-shaped joint has been selected for the experimental program, in which collapse is expected to occur preferably at the beam-column interface. For connection type C different
welding procedures have been considered. These include the full penetration double bevel groove
weld and the full penetration single bevel groove weld with a steel or polyester backing bar, Figure
2. A V-shaped steel backing bar is also considered (C3). Some special care has been devoted to
welding execution, which was done by competent structural steel fabricators, using certified welders.
A typical view of the full-scale specimens is shown in Figure 3. During tests, the column is kept
horizontal, hinged at both ends, and the (horizontal) loading system applied at the top of the beam.
A total of 32 specimens have been fabricated, 16 specimens (type C1, C2 and C4) to be tested in
Milan and 16 specimens of types A, B and C3 to be tested in Liege. Testing has been completed. In
addition to the beam-column joints, full-scale small size specimens have also been fabricated with
the purpose of verifying whether the weld preparation and the specimen sizes are critical parameters. The specimens (Fig. 4) incorporate welded connections representing the welded beam to column flange connection. A total of 40 specimens have been constructed and subjected to cyclic
testing.
3

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

2.2

Testing procedure

In 1986, a recommended testing procedure was proposed by ECCS for cyclic testing of steel members and joints. The ECCS recommendations (ECN, 1986) proposed to perform cyclic tests applying to the specimen cycles of increasing amplitude with at least three cycles for the same amplitude, similarly to what is proposed by ATC 24 (1992).
Since then, a number of research programs were carried out in various European countries,
adopting such ECCS recommended procedures. Tests were carried out in Italy, Germany, Belgium,
Portugal and France, on members, connections and structural subassemblages. In particular a research project was carried out with the sponsorship of Arbed Research on beam to column connections, and tests were performed at different research centers in various European countries; see, for
instance, Ballio & Chen (1993). After nearly ten years, some criticism was raised with regard to the
1986 ECCS recommendations. In particular, with regard to the Testing Procedures, it has been recognised that:
1. Repeating three cycles for each cycle amplitude and then increasing the amplitude does not
provide direct information regarding the damage accumulation and strength degradation corresponding to one particular ductility demand; in fact, at increasing amplitudes corresponds an isotropic strain hardening effect resulting in an increment of the load carrying capacity of the member,
but this effect is opposite to the strength degradation due to local buckling and low cycle fatigue,
which are directly connected to the cycle amplitude and the number of imposed cycles;
2. Sometimes tests were performed under unusually large cycle amplitudes, i.e. simulating
situations and ductility demands unlikely to occur during earthquakes.
It was then proposed by various authors both in Europe (Ballio & Castiglioni 1994, Bernuzzi et
al. 1997) and in the U.S. (Krawinkler & Zohrei 1983) to perform tests with cycles of a constant
amplitude; such procedures present, with regard to the ECCS recommended testing procedure, the
advantage of allowing a clear understanding of the damage accumulation process as well as of the
parameters governing it.
The scope of the research is the development of a cumulative damage model for assessing the
performance of structural components under arbitrary loading histories and evaluating the effects of
inelastic cycles on a limit state of acceptable behaviour. A cumulative damage model is generally
based on a damage hypothesis and includes structural performance parameters to be determined
experimentally (Bernuzzi, et al 1997, Krawinkler & Zohrei 1983). Hence a multi-specimen testing
program is required, encompassing constant amplitude loading tests on identical specimens. For
each test, a new specimen must be used, since each specimen is to be tested to failure under deformation amplitudes covering the range of performance assessment.
Cyclic tests are performed following a constant amplitude loading history in the plastic range,
after a few cycles in the elastic range. In particular, it is proposed to adopt the loading history
schematically represented in Figure 5, vy being the normal yield displacement, to be computed
theoretically by means of usual structural mechanics.

Figure 5. Schematic of the loading history.

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

Two cycles will be performed for each cycle with a displacement range v < 2 vy, while cycles
in the plastic range will be continued until complete failure of the specimen. Various values of the
maximum displacement range are considered: v = 4 vy , 6 vy , 8 vy. Experience gathered in previous testing programs, when it was assumed a cumulative damage model based on an S-N curve
approach, indicated that such procedure allows for a correct identification of low cycle fatigue
curves for each typology. To run the test, a definition of the failure of the specimen must be given
in cyclic testing, in particular when cycles of constant displacement amplitude are applied, the
failure may be very long process if a zero resistance is chosen as the definition of failure. Refering to practical applications in structures, it is clear that the structure is out of use before a zero resistance is reached. Conventionally in our testing, failure is determined by a drop in resistance of
50 % of the initial value obtained at the same displacement.
3 PRESENTATION OF LOW-CYCLE FATIGUE TEST RESULTS
The parameter characterizing high cycle fatigue is the number N of cycles at failure and the stress
tange , defined as follow for overall elastic deformations:

M P d P P d
=
=

W
W
P
W

(1)

P v
=
P
v
=

(2)

v P d

v
W

(3)

A working hypothesis is developed for low cycle fatigue. is the basic parameter characterizing
fatigue; in the dissipative zone, an equivalent stress * can be computed as:
* = E =

v Py d

vy
W

(4)

As test on each specimen are run at constant v and as Py and vy are known, it is easy to compute
* and to present the results in typical diagrams used for the presentation of high cycle fatigue
with * in ordinate and the number N of cycles to failure as abciss.
Expressed in the way, the results allow the definition of low cycles fatigue (Whler) curves
which can be compared to curves concerning the resistance to high cycle fatigue (classical). In the
presentation of the test results, reference is made to the design curves of Eurocode 3 (design of
steel structures).

Figure 6.

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

4 TEST RESULTS
All the tests in Lige and in Milan were constructed in early 1998. Some first significant results are
presented hereafter, but the analysis of the results is still going on, involving numerical evaluation
and considering experimental work still going on at the other testing sites.
4.1 Specimens type A1
Specimens A1 correspond to a rigid full strength bolted connection. The end plate of the beam is
welded with K preparation. The failure takes place in the beam: flange buckling or cracks in the
beam close to the weld. The results are presented in terms of low cycle fatigue (*, N) (Fig. 7).
Results A15 (95 mm) is situated on a higher curve than A12 (38 mm) and than A13 (57 mm). This
corresponds to two different failure modes: without apparent deformation in the flange for A12 and
A13, with buckles in the flange for A15. The result of specimen A1B2 (33 mm), which had no
double plate in the web of the column, lies higher than A12 (38 mm); this may be explained by the
existence of yielding of the panel zone that brings an increase in resistance to low cycle fatigue.
Globally, it can be concluded that connection A1, which is typical European design involving
welding at the shop and bolting on the site, has a very good behaviour. It corresponds to yielding in
the flange of the beam or to yielding in the panel zone of the column (specimen A1B).
4.2 Specimens type A2
The specimens A2 were subjected to several problems, which are not related to failure in welds:
Failure of bolts in the original design with 4 bolts in the connection (specimens A2 I , A2 II)
Lateral buckling, caused by a combination of web and flange buckling in the section where the
hunch starts (specimen A23); consequently the lateral support were increased in the testing rig
Even with increased lateral support (specimen A25, A26) the combination of local buckling of
web and flange in the section where the haunch starts remains the factor governing failure; this
can be explained by particular state of stress at this place; indeed the haunch introduces a thrust
onto the web; this was known and complete web stiffener would be needed to really prevent
the web buckling.
As a general comment on haunched beams the following can be expressed:
Putting a haunch increases the rigidity and the resistance of the beam
However, the haunch beam cost is very high because it requires much preparation of stiffeners,
a very high and thick plate and very long welds; the cost of connection A2 is about 80% higher
than the cost of A1.
The results are presented in terms of low-cycle fatigue in Figure 8.
4.3 Specimen B
Connection B is a semi rigid partial strength connection in which bolts are perpendicular to the
bended beam. The bolts are prestressed to a 9 kNm torque.
It is intended to develop energy dissipation through the ovalization of boltholes and through
friction between the web of the UPN 300 beam and the flange of the HE 300 M column. Both
mechanisms work:
- friction gives a constant resistance throughout the displacement,
- bearing resistance provides increases in resistance at both left and right side of the M - diagrams (Fig. 9).
These mechanisms are not subjected to degradation of strength up to number of cycles which can
be considered very high (over 40) in the earthquake context.

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

Figure 7. Whler curve for specimen type A1.

Figure 8. Whler curve for specimen type A2.

Figure 9. M - diagram for specimen B2.

The imposed displacements in test of specimens B have been chosen on the basis of a maximum
5 % drift of the building in which they would be present. In the range of displacements considered,
the friction force is quite constant and independent of the maximum displacement. It is around 50
kNm / (4 x 0,113 x 2) = 55 kN / bolt in both directions.
This last value is coherent for a M27 10.9 bolt with partial prestress and a friction coefficient
around 0,25. Once the bearing resistance is involved, the total resistance becomes higher, by a factor of 4 at the beginning of the cycles; this correspond to a total resistance of about 4 x 55 = 220 kN
/ bolt. This is very similar to the design resistance computed in the design of the specimens (217
kN / bolt). In reality, the phenomenon explaining the resistance of the connection is more complicated than a simple addition of "friction" and "bearing resistance", because at the 1st cycle, the
yielding realized by the ovalization of holes generates a local increase in the thickness of the web
of the UPN 300, which brings two additional factors: an increased prestress because the bolt is
forced to elongate and an increased friction because of the uneven surface produced by this "forging" phenomenon. The complete understanding of the resistance mechanism in such a connection
requires an extensive study and numerous small tests.

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

4.4 Specimen C1
Specimen type C1 were realized with a full penetration single bevel groove weld with a steel
backing bar. Figure 10 shows the hysteresis loops for test C1-50, while Figure 11 shows the hysteresis loops for C1-100 one. These specimens were tested under displacement ranges v = 50
mm and v = 100 mm. The yield displacement vy being approximately 23-25 mm, these displacement ranges resulted in ductility ranges v/ vy = 2 and 4 respectively. These specimens
are similar to those adopted for the moment resisting frame structures in which failures were reported during the Northridge Earthquake. Specimen C1-50 collapsed by failure of the beam flange
at the weld toe; this failure can be considered a brittle failure, because it took place suddenly, without noticeable deterioration in the load carrying capacity of the specimen. Only limited local buckling was evident at the end of the test in the beam flanges. Completely different was the behaviour
of specimen C1-100, which collapsed by failure of the beam flange due to large plastic deformations in the buckles, at the plastic hinge location. In the case of this specimen, the deterioration is
obvious and represents a clear warning sign of an incipient collapse. The different behaviour of the
two specimens is also evident at Figure 14, which presents the trend of the absorbed energy at each
cycle (E) normalized on the energy at the first cycle in the plastic range (E0). It can be noticed that
in the case of specimen C1-100, the ratio E/E0 shows an evident reduction during all the test duration, and before the last cycles leading to failure, it is already reduced below a value of 0.50. On the
other hand, the same ratio remains higher than 0.8 for specimen C1-50, dropping then suddenly, in
the last two cycles, below 0.2.

Figure 10. Hysteresis loops for specimen C1-50.

Figure 11. Hysteresis loops for specimen C1-100.

In addition, it should also be noticed that the conventional failure for specimen C1-50 can be
considered attained after 17 cycles in the plastic range, while that of specimen C1-100 was attained at cycle n. 12. This difference in the life of the specimens, that is very small, can be explained only by the different failure modes. From these results, it seems that cycles with a small
amplitude can lead to brittle fracture and may be more dangerous because they provide much less
visual warning about the degradation of the connection than cycles of large amplitude. These ones,
on the contrary, induce the formation of a plastic hinge with local buckling giving a very ductile
behaviour and a large energy dissipation.
4.5 Specimens C2
Specimens C2 are identical to specimen C1 but for the material of the backing bars, which is polyester. All the observation made for C1 are relevant. The distinction between two behaviours, one
brittle (C2B50, C2B75) and one ductile ( C2B100, C2B125) is even clearer (Figs. 13, 15).
8

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

Figure 12. Whler curve for specimen type C1.

Figure 13. Whler curve for specimen C2.

Figure 14. Energy absorption in Hysteresis loops


for specimens type C1-B.

Figure 15. Energy absorption in Hysteresis loops


for specimens type C2.

Figure 16. Whler curve for specimens C4.

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

4.6 Specimens C3
Specimens C3 are similar to C1, but for the shape of the backing bars which are triangular. The
backing bar tested in specimens C3 did not allow a complete butt weld to be made. Yielding was
restricted to the weld material and it did not last long: 2 or 3 cycles, to be compared for instance to
30 and 77 in specimens A1.
The high yield strength of the IPE 450A material (405 N / mm instead of the expected 300
N/mm) may also be participating to this early failure, since the weld material had not been chosen
specifically for a yield strength higher than expected.
4.7 Specimens C4
Specimens C4 correspond to a K preparation and welding from both sides of the flanges, which are
conditions for high quality welds.
Here again two different behaviours can be observed, to which two levels of Whler curves correspond, Figure 16:
- Brittle for displacement up to 50 mm
- Ductile above 50 mm
4.8 Comparison between various design
In the above table, the index of EC3 curves that would represent the behaviour of tested specimens
is given. For each specimen type two values are presented, which correspond to the two different
mechanical behaviour observed:
Low cycle fatigue excursions, no buckling, rather brittle failure
High yield excursion, buckling, progressive decrease of resistance.
Table 1. Comparison for of EC3 fatigue line (MPa) between various designs.
SPECIMEN TYPE
A1
A2
C1
C2
C3
C4

Small yield
displacement v = 2 vy

High yield
displacement v = 4 vy

42
/
21
32
10
29

50
40
56
63
18
63

5 CONCLUSIONS
The tests realized with the objective to define low cycle fatigue curves corresponding to various
design detail of connection used in moment frames allow several practical conclusions.
The results obtained set forward the validity of a quantification of cyclic tests in term of low
cycle fatigue
On that basis, the various connection design tested can be classified in terms of average of resistance, but also of scatter; in this formulation, the word design includes the many factors influencing the behaviour, like base material, weld material, groove type, welding sequence, type
of backing bar, flexibility of eventual end plates, yielding resistance of the panel zone of the
column
Two basic different behaviours have been observed, independently of the type of connection
design. One takes place when the plastic strains realized during the cyclic test are small,
10

11th European Conference on Earthquake Engineering 1998 Balkema, Rotterdam, ISBN 90 5410 982 3

around two times the yield stains, and no buckling of the beam is observed; then, the strength
degradation is small until a relatively sudden failure takes place. The other type of behaviour
happens when large plastic cyclic excursions are realized; then the resistance progressively
drops, as buckling develops.
ACKNOWLEDGEMENTS

Financial support to the project has been provided by the European Community Commission for
Environment and Climate Programme, Directorate XII, Science, Research and Development;
Contract N. ENV4-CT96-0278; Proposal N. PL950671.
REFERENCES
ATC-24 1992. Guidelines for Cyclic Seismic Testing of Components of Steel Structures, Applied Technology Council.
Ballio, G., Chen, Y. 1993. An Experimental Research on Beam Column Joints: Exterior Connections,
Giornate C.T.A., Viareggio.
Ballio, G., Castiglioni, C.A. 1994. Seismic Behaviour of Steel Sections, Journal of Constructional Steel
Research, Vol. 29, pp. 21-54.
Bernuzzi, C., Calado, L., Castiglioni, C.A. 1997. Behaviour of Steel Beam-to-Column Joints under Cyclic
Reversal Loading: an Experimental Study, SSDS, Proceedings, Nagoya, Japan.
Bernuzzi, C., Calado, L., Castiglioni, C.A. 1997. Low Cycle Fatigue of Structural Steel Components: Methods for Re-Analysis of Test Data and a Design Approach Based on Ductility, submitted for publication
on the Journal of Earthquake Engineering.
Bertero, V., Anderson, J.C., Krawinkler, H. 1994. Performance of Steel Building Structures during the
Northridge Earthquake, EERC, University of California, Berkeley, Rep. n. UCB/EERC-94/09.
European Committee for Standardization (CEN) 1994. ENV 1993-1-1 Eurocode 3: Design of Steel Structures.
European Committee for Standardization (CEN) 1994. ENV 1998 Eurocode 8: Design Provisions for Earthquake Resistance of Structures.
European Convention for Constructional Steelworks 1996. Recommended Testing Procedures for Assessing
the Behaviour of Structural Elements under Cyclic Loads, Technical Committee 1, TWG 1.3 - Seismic
Design, Publ. N. 45.
Krawinkler, H., Zohrei, M. 1983 Recommendations for Experimental Studies on the Seismic Behaviour of
Steel Components and Materials, The John Blume Earthquake Engineering Research Center, Stanford
University, Rep. n. 61.
Plumier, A. 1994. Behaviour of Connections, Journal of Constructional Steel Research, Vol. 29, N. 2, pp.
95-119.
Tremblay, R., Timler, P., Bruneau, M., Filiatrault, A. 1995. Performance of Steel Structures during the
Northridge Earthquake, Canadian Journal of Civil Engineering, Vol. 22, N. 2, April.

11

You might also like