Professional Documents
Culture Documents
Connections in Steel Structures V. ESSC-AISC Workshop 2004 PDF
Connections in Steel Structures V. ESSC-AISC Workshop 2004 PDF
Steel Structures V
Behaviour, Strength & Design
held at
Edited by
F.S.K. Bijlaard
A.M. Gresnigt
G.J. van der Vegte
I
The information presented in this publication has been prepared in accordance with
recognized engineering principles and is for general information only. While it is believed to
be accurate, this information should not be used or relied upon for any specific application
without competent professional examination and verification of its accuracy, suitability and
applicability by a licensed professional engineer, designer or architect. The publication of the
material contained herein is not intended as a representation or warranty on the part of the
American Institute of Steel Construction, Inc. or the European Convention for Constructional
Steelwork, or of any other person named herein, that this information is suitable for any
general or particular use or is free from infringement of any patent or patents. Anyone
making use of this information assumes all liability arising from such use.
Copyright © 2005 by Bouwen met Staal, Boerhaavelaan 40, 2713 HX Zoetermeer, The
Netherlands (www.bouwenmetstaal.nl).
All rights reserved. No part of this publication may be reproduced without written permission.
ISBN-10: 90-9019809-1
ISBN-13: 978-90-9019809-5
II
FOREWORD
This book presents the Proceedings of the Fifth International Workshop on Connections in
Steel Structures: Behaviour, Strength and Design. The workshop was held at the Radisson
SAS Hotel in Amsterdam, The Netherlands, during the period 3-4 June 2004 under the
auspices of the Faculty of Civil Engineering and Geosciences of the Delft University of
Technology, The Netherlands.
The four prepreceding international workshops, alternately organised in Europe and the
USA, were held at (i) Ecole Normale Supérieure de Cachan, in Cachan, France, 25-27 May,
1987, (ii) Westin William Penn Hotel in Pittsburgh, Pennsylvania, 10-12 April, 1991, (iii) Villa
Madruzzo in Trento, Italy, 29-31 May, 1995 and (iv) Roanoke Convention and Conference
Center in Roanoke, Virginia, USA, 22-25 October 2000. Proceedings from the four preceding
workshops were published by Elsevier Applied Science Publishers (1988), the American
Institute of Steel Construction (1992), Pergamon (1996) and the American Institute of Steel
Construction (2002).
These five workshops on connections in steel structures express the importance of these
structural components. Connections have great influence on the structural behaviour and
safety of steel structures. Furthermore they determine almost 40% of the costs of steel
structures. Research in this field is therefore more than worthwhile. The success of these
workshops is also formed by the fact that the experts are selected and invited to present their
most recent research in the field of connections by the organising committee.
The American Institute of Steel Construction (AISC) and the European Convention for
Constructional Steel (ECCS) supported the idea of holding this fifth workshop. Financial
support for the workshop was generously provided by the sponsors as listed hereafter.
The Society Bouwen met Staal was the official host of the workshop. The effort put forth by
the staff at Bouwen met Staal, in particular Mrs Soraya van Beuzekom, made the workshop a
great success. This support is gratefully acknowledged. Further, the assistance by the
participants that served as session chairs helped the workshop to run smoothly. The
organising committee is indebted to Reidar Bjorhovde and Roberto Leon for their help in
selecting experts world wide. Special thanks are also due to Addie van der Vegte for his work
producing the proceedings in this book and on the CD.
The support and technical contributions of the 56 invited participants from 24 individual
countries all over the world resulted for the fifth time in a workshop of high quality and a
contribution to the profession. Without the efforts on the research, design and
implementation of steel connections worldwide, the workshop would not have been possible.
The continued efforts by the participants will no doubt result in another successful workshop
in the future.
III
THE FIFTH INTERNATIONAL WORKSHOP ON
CONNECTIONS IN STEEL STRUCTURES
was sponsored by
IV
TABLE OF CONTENTS
Evolution of shear lag and block shear provisions in the AISC specification 21
Louis F. Geschwindner, (USA)
Docking solution between a steel truss and a concrete tower at the ski jump in 45
Innsbruck
Aste, C., A. Glatzl, G. Huber, (Austria)
Modelling
V
Modelling, Deformation Capacity, Seismic
Modelling procedures for panel zone deformations in moment resisting frames 121
Finley A. Charney, William M. Downs, (USA)
Seismic
Effect of column stiffener detailing and weld fracture toughness on the 177
performance of welded moment connections
Dexter, R.J., J.F. Hajjar, (USA), D. Lee, (South Korea)
Analysis of bolted end-plate joints: cyclic test and standard approach 191
Dunai, L., N. Kovács, (Hungary), L. Calado, (Portugal)
VI
An alternative approach to design of eccentrically loaded bolt groups 273
Muir, L.S., W.A. Thornton, (USA)
Exploring the true geometry of the inelastic instantaneous center method for 281
eccentrically loaded bolt groups
Muir, L.S., W.A. Thornton, (USA)
Four-plate HEB-100 beam splice bolted connections: tests and comments 287
Zygomalas, M.D., C.C. Baniotopoulos, (Greece)
Bolted connections with hot dip galvanized steel members with punched holes 297
Valtinat, G., H. Huhn, (Germany)
High strength half round head and nut HV-bolts for ancient steel constructions 311
Valtinat, G. (Germany)
Tubular Connections
Strength and stiffness of RHS beam to RHS concrete filled column joints 403
Szlendak, J.K., (Poland)
VII
An effective external reinforcement scheme for circular hollow section joints 423
Choo, Y.S., J.X. Liang, (Singapore), G.J. van der Vegte, (The Netherlands)
The influence of boundary conditions on the chord load effect for CHS 433
gap K-joints
Van der Vegte, G.J., (Japan, The Netherlands), Y. Makino, (Japan), J. Wardenier,
(The Netherlands)
Conclusions 469
Stark, J.W.B., (The Netherlands)
VIII
CONNECTION DESIGN IN THE 2005 AISC SPECIFICATION
ABSTRACT
INTRODUCTION
The American Institute Steel Construction (AISC) introduced the first specification for the
design and construction of structural steel buildings in 1923, for the purpose of creating a
standard for the steel industry in the United States. This original document was a mere nine
pages approved by a committee of five, and it has grown to exceed 100 pages, undergoing
numerous revisions based on experience gained over the years and research; both analytical
and test-based. Today, the AISC Committee on Specifications consists of 40 members
currently working on the 2005 Specification for Structural Steel Buildings (1), hereafter
referred to as the 2005 Specification. This new document has a new format unlike any
previous versions, as it will combine both load and resistance factor design (LRFD) and
allowable stress design (ASD) methods into one. More specifically, many of the provisions
have been revised and updated in Chapter J, Design of Connections, since publication of the
most recent AISC specification, the 1999 Load and Resistance Factor Design Specification
for Structural Steel Buildings (2), hereafter referred to as the 1999 Specification. Although
the specification is still in draft form, with two remaining ballots, there are many issues that
can be discussed at this time. Some of the general connection design topics that will be
addressed are compression members with bearing joints, splices in heavy sections, beams
copes and weld access holes, combining bolts and welds, and limitations on bolted and
welded connections. The welding areas that will be revised are effective area and limitations
on effective throat area of groove welds, as well as, effective area, terminations, and strength
of fillet welds. Finally, some of the changes expected for design with bolts and threaded
parts occur in provisions for: the types of fasteners allowed, combined tension and shear
strength, design of slip-critical connections, block shear, and shear lag.
Before discussing the new revisions to the connection chapter, it is important to understand
the format of the 2005 Specification. The original 1923 document was based on the
allowable stress design format (ASD), which gives capacities in allowable stresses with the
safety factor incorporated. In 1986, AISC introduced their first load and resistance factor
Pn = FyAg
where the design tensile strength is φtPn and the allowable tensile strength is Pn / Ωt. The
safety factors were determined based on a live load-to-dead load ratio of 3, which results in
1.5 as the target effective load factor for the load combination of 1.2D+1.6L. Therefore, in
most cases, the safety factor is calculated as 1.5/φ and it is given to 3 significant digits. The
required strength or available strength are based on ASCE 7, Minimum Design Loads for
Buildings and Other Structures (4) factored load combinations for either LRFD or ASD,
depending on the method used. This arrangement will result in greater clarity, uniformity and
efficiency when applying AISC specifications. In the final analysis, the only difference
between the LRFD and ASD method of strength design is on the required strength side.
LRFD is based on factored load combinations given in ASCE 7 and ASD is based on service
load combinations in ASCE 7. Chapter J, Design of Connections, begins by stipulating the
design basis, similar to the above followed by more definitive design provisions as discussed
in the following.
GENERAL REQUIREMENTS
Chapter J of the 2005 Specification contains the majority of the connection design provisions
in that document. The first section entitled “General Provisions” contains revisions to such
topics as compression members with bearing joints, splices in heavy sections, beam copes
and weld access holes, bolts in combination with welds, and limitations on bolted and welded
connections.
The new provision permits that compression members, other than columns, be proportioned
for the less stringent of: 1. an axial tensile force of 50% of the required compressive strength
of the member or 2. the moment and shear resulting from a transverse load equal to 2
percent of the required compressive strength of the member. The application of this
transverse load should be at the splice location “exclusive of other loads that act on the
member. The member shall be taken as pinned for the determination of the shears and
moments at the splice.” This sub-section begins with a User Note that reminds the designer,
“All compression joints should also be proportioned to resist any tension developed by the
load combinations….” User Notes are a new feature of the 2005 Specification. They are
non-mandatory and are interspersed throughout the document to offer the designer concise
assistance with using the specification.
The special material toughness requirements for splices of heavy sections connected by
complete-joint-penetration groove welds have previously existed in the 1999 Specification.
The 2005 Specification will include clarification of these requirements. Shrinkage of large
welds between elements that are not free to move causes strains in the material adjacent to
the weld that can exceed the yield point strain. As the Commentary to the 2005 Specification
states, "In thick material the weld shrinkage is restrained in the thickness direction, as well as
in the width and length directions, causing triaxial stresses to develop…." and this can
prevent the steel from deforming in a ductile manner. Thus, special material toughness
requirements, and carefully prepared weld access holes and copes are required for heavy
tension members to prevent brittle fracture.
For both rolled and built-up shapes, special toughness requirements apply to shapes with
flanges or plates exceeding 2 in. (50 mm), when "used as members subject to primary tensile
forces due to tension or flexure and spliced using complete-joint-penetration groove welds
that fuse through the thickness of the member." The latter phrase was added to clarify the
extent of welding required for these provisions to be applicable. The verbiage in the 1999
Specification explaining how the impact test should be performed is replaced with a
reference to ASTM A6/A6M, Supplementary Requirement S30, Charpy V-Notch Impact Test
for Structural Shapes - Alternate Core Location (5). The impact test must meet a minimum
average value of 20 ft-lbs (27 J) absorbed energy at +70°F. The requirements do not apply if
the splices and connections are made by bolting, or if shapes with elements less than 2 in.
thick are welded to a heavy section, or to splices of elements of built-up shapes that are
welded prior to assembling the shape.” On the other hand, the provisions do "apply to built-
up cross section consisting of plates exceeding 2 in. that are welded with complete-joint-
penetration groove welds to the face of other sections."
When splicing hot rolled shapes with a flange thickness exceeding 2 in. (50 mm) and similar
built-up cross sections, special attention must also be paid to the formation of beam copes
and weld access holes. More detailed rules for the size of an access hole are given;
specifically, the height shall be 1 1/2 times the thickness of the material containing the
access hole, most likely the web thickness, but not less than 1 in. (25 mm) nor greater than 2
in. (50 mm). Room for weld backing must also be provided and no arc of the weld access
hole shall have a radius less than 3/8 in. For built-up shapes the access hole may
terminate perpendicular to the flange as long as the flange-to-web weld is held back at least
the weld size from the edge. The weld access hole details included in the 2005 Specification
are very similar to those in AWS D1.1, Structural Welding Code-Steel (6).
The design criteria for bolts in combination with welds in a joint are being completely revised
in 2005. Formerly, only bolts in slip-critical connections were permitted to share load with
welds. In the current draft, the provision reads as follows:
Bolts shall not be considered as sharing the load in combination with welds
except that connections with high-strength bolts installed in standard holes or
short slots transverse to the direction of the load are permitted to be
considered to share the load with longitudinally loaded fillet welds. In such
connections the strength of the bolts shall not be taken as greater than 50% of
the bearing strength of the bolts.
In other words, bolts in standard holes and short slots transverse to the direction of load can
share load with only longitudinally loaded fillet welds, but with a 50% reduction in the bearing
capacity of the bolts. This new provision is based on a recent research paper published in
the AISC Engineering Journal by Kulak and Grondin (7).
This section of Chapter J lists under what conditions pretensioned joints, slip-critical joints, or
welds are required. A similar section has existed in the AISC Specification for several
editions. For column splices, the height limitations and the language is being updated and
simplified, such that pretensioned joints, slip-critical joints, or welds are required in column
splices in all multi-story structures over 125 ft (38 m) in height. Formerly, the height limit was
based on the width of the building. The new provision is consistent with the height above
which connections of all beams and girders to columns are required to be pretensioned
joints, slip-critical joints, or welds. The remainder of the list remains unchanged, including
connections where live loads produce impact or reversal of stress and structures carrying
cranes over five-ton capacity.
A brief section on minimum strength of connections will be deleted. This section, also a
remnant of older versions of the AISC specification, stated a minimum factored load of 10
kips (44 kN) that all connections "providing design strength" should carry (2). The task
committee determined that these minimum loads have no technical basis and had the
potential of giving the designer the false idea that connections with this minimum design load
were adequate for fabrication and construction loads without further analysis.
WELDS
Weld provisions given in AWS D1.1 (6) apply under the 2005 Specification, with the
exception of those modified by the AISC sections listed in the preamble to Section J2. The
intention is for AISC to update their provisions to be consistent with the referenced version of
AWS D1.1. However, due to the different development cycles of the two standards, in some
cases, differences occur.
The most significant revisions to the weld provisions in the 2005 Specification occur in the
following areas: effective area and effective weld sizes for groove welds, and effective area,
terminations, and strength of fillet welds.
In line with AWS D1.1, as well as more recent research, the tables for effective throat of
partial-joint-penetration groove welds and effective weld sizes of flare groove welds are being
updated. Table J2.1 shown below has expanded to include more combinations of welding
processes and welding positions for partial-joint-penetration groove welds (new portions are
highlighted). The terminology to describe the effective throat thickness has been revised
from "depth of chamfer" to "depth of groove."
Effective weld sizes of flare groove welds are being increased based on a March 2003 report
by Packer and Frater (8) as shown in Table J2.2. This table applies when the flare groove
weld is filled flush to the surface of a round bar, a 90° bend in a formed section, or
rectangular tube. For flare groove welds filled less than flush, the values in Table J2.2 apply
minus the greatest perpendicular dimension measured from a line flush to the base metal
surface to the weld surface. Examples of Flare-V-groove and flare-bevel groove welds are
shown in Figure 1.
Effective throats larger than either Table J2.1 or J2.2 can be qualified by tests. For flare
groove welds the fabricator must establish by qualification the consistent production of such
larger effective throat thicknesses.
Fillet welds
The important revisions expected to the fillet weld provisions relate to effective throat, fillet
weld terminations, strength when fillet weld groups are oriented both longitudinally and
transversely to the direction of applied load. Regarding effective throat, historically, an
increase was permitted for submerged arc welding only. This increase has not been found to
be conservative for all process settings. Therefore, the new language allows an increase in
the effective throat using any welding process if consistent penetration beyond the root of the
diagrammatic weld is demonstrated by tests.
In the 1999 Specification, specific criteria for fillet weld terminations were incorporated into
the specification proper. This material has been "tweaked" slightly by revising the language
for terminations where cyclic forces exist. The previous text read, "For connections and
structural elements with cyclic forces, … fillet welds shall be returned around the corner for a
distance not less than the smaller of two times the weld size or the width of the part." This is
now only applicable to "double angle connections and structural elements subject to cyclic
forces." Additionally, the special termination for "fillet welds joining transverse stiffeners to
plate girder webs" now only applies to plate girder webs 3/4 in. (19 mm) thick or less - a less
stringent requirement.
A new provision not yet balloted by the committee at the time of this paper, relates to the
strength of fillet welds when fillet weld groups are concentrically loaded and consist of weld
elements that are oriented both longitudinally and transversely to the direction of applied
load. The combined nominal strength of the fillet weld group shall be determined as the
greater of:
where,
Rwl = the total nominal strength of longitudinally loaded fillet welds
Rwt = the total nominal strength of transversely loaded fillet welds
This new provision follows two existing sub-sections that allow higher design capacities
based on the angle of loading with respect to the weld longitudinal axis. Recent research by
Ng et al. (9, 10) has demonstrated that where fillet welds exist in the same weld group that
are oriented both transverse and longitudinal to the direction of applied load, the existing
provisions are unconservative (2).
Similar to design of welds in the AISC Specifications, use of high-strength bolts conforms to
another referenced standard, entitled Specification for Structural Joints Using ASTM A325 or
A490 Bolts, as approved by the Research Council on Structural Connections, hereafter
referred to as the RCSC Specification (11). For instances where the AISC Specification
differs from this referenced document, the AISC Specification controls. The new revisions
that will appear in the 2005 Specification include an expanded list of bolts or threaded rods
permitted, more liberal use of short-slotted holes, and revised procedures for combined
tension and shear in bearing-type connections and design for shear in slip-critical
connections.
High-strength bolts
Sometimes there is a need for larger diameter (greater than 1 ½ in. (38 mm)) or longer than
usual (greater than 12 diameters) high strength bolts, such as anchor rods for fastening
machine bases. Due to the diameter and length limitations of the more commonly accepted
bolts types, such as high-strength bolts, ASTM A325, A490, and F1852 (twist-off type), the
2005 Specification will permit the use of other specified ASTM bolts or threaded rods. Bolts
or threaded rods conforming to the following ASTM specifications are permitted in this case:
ASTM A354 Gr BC, A354 Gr BD, (Quenched and Tempered Alloy Steel Bolts, Studs, and
Other Externally Threaded Fasteners) or A449 (Quenched and Tempered Steel Bolts and
Studs). For slip-critical connections, it is important that the geometry of these special
fasteners, including the head and nut(s) is equal to or (if larger in diameter) proportional to
that provided by ASTM A325 or A490 bolts. Installation must comply with the RCSC
Specification with modifications as necessary to account for the increased diameter and/or
length to provide the design pretension.
In addition to permitting other ASTM bolt types, the new specification will relax the
requirements for hole types permitted. Previously, only standard holes were allowed in
member-to-member connections without the approval of the engineer of record (EOR). It is
proposed that additionally short-slotted holes oriented transverse to the direction of load may
also be used routinely without any special approval. This is in response to what is common
practice in the fabrication industry. Short-slotted holes oriented parallel to the load,
oversized holes, or long-slotted holes still require EOR approval.
Research has demonstrated that the strength of bearing fasteners subject to combined shear
and tension can be closely represented by an ellipse (12). Previous versions of the AISC
Specification have employed a straight-line representation of the ellipse as shown in Figure
2. In the 2005 Specification, the actual equation for the sloped portion of the approximation
is being given in the provisions, with the more exact elliptical equations given in the
commentary. The provisions read as follows:
⎛ ft ⎞ ⎛ fv ⎞
⎟+⎜ ⎟ = 1.3
A
⎜ (LRFD)
⎝ φFt ⎠ ⎝ φFv ⎠
⎛ Ωft ⎞ ⎛ Ωfv ⎞
⎜ ⎟+⎜ ⎟ = 1.3 (ASD)
⎝ Ft ⎠ ⎝ Fv ⎠ Pt. A = φFt or Ft / Ω
Pt. B = φFv or Fv / Ω
fv
B
One contentious issue in the current draft of the 2005 Specification is how to handle slip
resistance of slip-critical connections. The 1999 Specification gave two procedures for
calculating slip resistance: one method using factored loads and the other based on service
loads. For consistency with the format of the “unified” specification, only one procedure is
being proposed that is purported to give substantially the same results for ASD and LRFD.
This proposed procedure provides “resistance to slip at service loads or resistance to slip at
factored loads with a reliability appropriate for serviceability criteria.” The draft criteria can be
summarized as follows:
The design slip resistance φRn and the allowable slip resistance Rn/Ω shall be determined as:
φ= 1.00 Ω = 1.40
Rn = 1.13µ hscTbNs
where:
µ = mean slip coefficient for Class A (0.35) or B (0.50) surfaces, as applicable, or
as established by tests
hsc = hole factor based on the hole type (standard, oversize, etc.)
Tb = minimum fastener tension, kips (kN)
Rn = 0.6FyAgv + UbsFuAnt
where:
Ubs = 1 when tension stress is uniform; 0.5 when tension stress is nonuniform
Agv = gross area subject to shear, in.2 (mm2)
Ant = net area subject to tension, in.2 (mm2)
For a more detailed discussion of this limit state and its history, see Reference (13).
CONCLUSION
The 2005 Specification will not reach final approval until later in 2004, therefore the material
discussed in this paper is for information only and should not be applied until the final
document is announced. When that happens, the revised provisions in Chapter J for bolted
and welded connection design will be another step forward for the steel design and
fabrication industry in the United States. The AISC Committee on Specifications will
continue to work toward the goals of their mission statement:
Develop the practice-oriented specification for structural steel buildings that provides for
• life safety
• economical building systems
• predictable behavior and response
• efficient use
Based on new information from the areas of research and industry practice, the 2005
Specification for Structural Steel Buildings will allow for continued safe, as well as
economical and efficient steel building designs.
REFERENCES
(1) AISC, (2004). Specification for Structural Steel Buildings, Draft dated March 2004,
American Institute of Steel Construction, Chicago, IL.
(2) AISC, (1999). Load and Resistance Factor Design Specification for Structural Steel
Buildings, American Institute of Steel Construction, December 27, Chicago, IL.
(3) AISC, (1986). Load and Resistance Factor Design Specification for Structural Steel
Buildings, American Institute of Steel Construction, September 1, Chicago, IL.
(4) ASCE, (2002). Minimum Design Loads for Buildings and Other Structures, ASCE 7,
American Society of Civil Engineers, Reston, Virginia.
(5) ASTM, (2002), Standard Specification for General Requirements for Rolled
Structural Steel Bars, Plates, Shapes, and Sheet Piling, ASTM A6/A6M-02,
American Society of Testing Materials, West Conshohocken, Pennsylvania.
(6) AWS, (2002). Structural Welding Code -Steel, American Welding Society, AWS
D1.1/D1.1M:2002, Miami, Florida.
(7) Kulak, G.L. and Grondin, G.Y. (2003). “Strength of Joints that Combine Bolts and
Welds,” Engineering Journal, AISC, 4th Quarter.
(8) Packer, J.A and Frater, G.S., (2003). "The Effective Throat of Flare Bevel and Flare
V Groove Welds," Final Report to AISC and STI, March.
(9) Ng, A.K.F., Deng, K., Grondin, G.Y., and Driver, R.G., (2004). “Behavior of
Transverse Fillet Welds: Experimental Program,” Engineering Journal, AISC, 2nd
Quarter.
(10) Ng., A.K.F., Driver, R.G., Grondin, G.Y., (2004). “Behavior of Transverse Fillet
Welds: Parametric and Reliability Analyses,” Engineering Journal, AISC, 2nd
Quarter.
Reidar Bjorhovde
The Bjorhovde Group, Tucson, Arizona, U S A
ABSTRACT
A tension test is used to represent the properties of steel, but it has no meaning
for the response of the material in a structure. The uniaxial tension test was
developed as a consensus solution, to have a standard by which similar
materials could be compared to a common base. It does not represent the
actual behavior of the steel in a structure, and was never intended to do so. The
paper addresses the properties of a range of structural steels, how these are
incorporated into design standards and how the standards define deformation
characteristics and demands for bolted and welded connections.
INTRODUCTION
As a construction material, steel has significant advantages over many others: it offers high
strength and stiffness, has adequate deformation capacity and stress redistribution ability for
many applications, it does not crack or otherwise fracture under normal service conditions, and
is available in several strengths and geometric forms. Finally, for most practical purposes it may
also be regarded as isotropic, with resulting benefits.
On the other hand, many structures will experience "non-normal" conditions many times during
fabrication, construction or service. A dynamically loaded structure such as a bridge will
experience fatigue; seismic events impose major deformation demands on structural
components and details; fabrication methods such as welding place very high demands for local
deformation ability of the steel in certain regions of the structure. The state-of-the-art of
computation technology is such that it is possible to incorporate many of these effects explicitly
in the analysis phase, and the quality of fabrication and construction continues to improve as
staff training and equipment are enhanced. However, much of the advanced software is not
suitable for design purposes, and most of this work therefore continues to be strictly research-
oriented.
It is a major problem that the material itself is not adequately understood by the professionals
who specify its use for structural purposes. This includes the complexity of its chemical and
metallurgical makeup, as well as the fact that the models that are used by codes to represent its
mechanical response bear little resemblance to what the steel will experience under actual
fabrication and service conditions. For one, it is known that steel is anisotropic, as a result of
production operations as well as other plastic deformation effects. Although the anisotropy
normally is of no particular consequence, it will affect the response of the steel in extreme
loading and deformation demand situations. For another, the behavior of steel is a function of
deformation history, to the effect that an otherwise ductile steel may respond as a high strength,
low ductility material, given the prior occurrence of large displacements.
Many designers tend to consider the requirements of the materials standards as reflecting actual
performance ability. Two- and three-dimensional effects are not recognized, at least in part due
to the inability of design standards to correlate such effects with the elementary material
behavior models that are used. This is done in spite of the fact that multidimensional response
is the key to the behavior of some of the most important regions of the structure. In particular,
experience has shown that this is where problems tend to develop, much more than in any other
areas of the structure (1, 2, 3, 4).
Nevertheless, the focal point for designers continues to be the design codes. For rational
decisions and proper recognition of material abilities, it is essential to appreciate the
relationships between strength and deformation demands, and to assess which one that
governs the end result.
The uniaxial tension test uses engineering stress and strain to define the response of the steel,
for convenience in measurement and because the test results are ultimately intended only for
use in comparison with other steels. With P = applied axial load, A0 = original cross-sectional
area of tension specimen, A = cross-sectional area at load P, l0 = original gage length, l = length
(= instantaneous length) at load P, and ∆l = change in length = (l - l0), the stress and strain are
given by the elementary expressions of Eqs. (1a) and (1b):
σ = P/A0 (1a)
l = l0 (1 + ε) (1c)
Although convenient, and suitable as representations of the steel behavior up to and slightly
beyond yielding, these definitions do not recognize that the area changes as the load increases.
Based on the concept of an incompressible material, which is fundamentally correct for steel,
the following holds true
A0l0 = A l (2)
Similarly. the true strain, εtr, recognizes that the length changes continuously. The strain
increment, dεtr, for an instantaneous length l is given by
dεtr = dl/l
For all practical purposes, the engineering and the true strains are equal for small strains, and
the stress-strain curves coincide up to and slightly beyond the yield stress. After this point the
two curves will diverge, to the effect that the true stress will continue to increase until material
rupture.
The most common ductility measure for steel is the elongation at fracture, εu, which is defined in
terms of engineering strain. Somewhat inconveniently, this is a function of the gage length, and
mill test reports must therefore report the magnitude of l0, as either of the commonly used 50 or
200 mm lengths. True strain is clearly a better ductility measure, since it reflects the total
accumulated strain at the point of failure. For equal levels of engineering and true stress, the
true strain is significantly smaller than the engineering strain. Conversely, for equal levels of
engineering and true strain, the true stress is significantly higher than the engineering stress.
The preceding developments are based on one-dimensional material behavior, with no restraint
offered against deformations in the other orthogonal directions of the steel specimen. Once the
yield stress is reached, plastic deformation takes place, and necking occurs in the area of the
specimen where the failure ultimately will occur. For a more realistic assessment of the
response of the steel under multi-dimensional restraint conditions, it has been shown that the
true stress-true strain relationship is the appropriate representation, unless one proceeds to
utilize yield criteria for multi-dimensional states of strain. However, this is not practical,
especially in view of the need for fairly simple material definitions.
At the same time, the true stress and strain reflect the fact that the steel cannot supply the
amount of deformation indicated by the results of the simple tension test. Even under minor
restraint conditions, fracture takes place at strains that are significantly lower than those
specified by the materials standards.
Complicating the issues further is the fact that the yield ratio has been shown to play a major
role for the deformability of steel. Designating the yield ratio by Y, it is defined as
Yield Stress
Y = ------------------------ = Fy/Fu (4)
Tensile Strength
where Fy and Fu are the mechanical properties utilized by all common materials standards.
Examining a wide range of structural steels, Kato (5) showed that the deformability decreases
with an increasing value of Y, and the decrease is especially marked for Y-values in excess of
approximately 0.6. As an illustration, Table 1 gives the yield ratios and relevant mechanical
properties for a number of the current American structural steels. The standard numbers are
those of the American Society for Testing and Materials (ASTM); the relevant stress levels are
given in units of MPa (N/mm2). These steels have very similar counterparts within European
and Japanese steelmaking practices, to mention two major areas of steel production. Standards
published by the International Standards Organization (ISO) also reflect these types of
materials.
It is clear that adequate structural performance cannot be guaranteed by basing the material
choice only on the standards' basic properties. In fact, Kato (5) recommended that in order to
assure reasonable and reliable deformation capacity of steel members and connections, as a
Based on the above recommendation and substantial research work performed in the aftermath
of the 1994 Northridge earthquake, an enhanced 350 MPa steel grade is now produced by steel
mills in the United States, using a maximum Y-value of 0.85 (6). The specified minimum yield
stress is 350 MPa; the standard also requires a maximum value of Fy of 450 MPa, and the
minimum tensile strength is Fu = 450 MPa. Detailed chemistry requirements are provided, as is
an upper limit on the carbon equivalent, to ensure satisfactory weldability.
The preceding issues are further complicated by the way tension tests are performed and
reported by steel producers. Specifically, the upper yield point is commonly given as the value
of the representative Fy. Although the use of this property is understandable, from a production
viewpoint, it is not a dependable, realistic value, since it relies heavily on the specifics of the
method of testing. As noted by Lay (7) - "the upper yield stress is not of relevance in design, as
it is lost if small overloads or misalignments occur". Specifically, utilizing dislocation theory as
the basis for yielding or plastic deformation in steel, the upper yield point mobilizes dislocations,
and the lower yield point maintains the "movement" of the dislocations. In essence, yielding is
caused by crystal structure (lattice) defects, in the form of dislocations.
These issues have not been addressed by a number of studies, among them numerous seismic
projects, their reports noting that - "...... 50 percent of the material actually incorporated in a
project will have yield strengths that exceed these mean values. For the design of facilities with
stringent requirements for limiting post-earthquake damage, consideration of more conservative
estimates of the actual yield strength may be warranted" (4). The same reference notes that
"Design professionals should be aware of the variation in actual properties permitted by the
ASTM specifications. This is especially important for yield strength. Yield strengths for ASTM
A36 material have consistently increased over the last 15 years......" (4).
Several other factors also play important roles, such as the steel production methods (e.g.
conventional and thermomechanical control processes (8, 9)). The changes that have taken
place over the past 15 years in North America in going from iron ore and coke-based ingot steel
to scrap-based continuous cast steel have resulted in materials that are significantly different,
but clearly improved in metallurgical and mechanical sense. Steel chemistry and weldability,
and especially the carbon and alloy contents, further emphasize the complex problems facing
the designer and the fabricator in the steel selection process. Some of these issues have now
Current structural steels in the United States span the range from the mild, carbon-manganese
A36, to the high strength, quenched and tempered A514 and the quenched and self-tempered
A913. The traditional methods have largely given way to continuous casting; for structural
profiles in the US, current (2004) production is based entirely on this technology. As a result,
steel chemistry has changed perceptibly, such that steel now gains its strength less from carbon
and more from a variety of alloying elements. The current steels have significantly lower levels
of carbon than previous production runs. Values of C-content less than 0.10 percent are the
norm; this contrasts with a carbon content of 0.2 percent and higher for earlier steels.
The lower carbon and higher alloying elements contents result in steels with acceptable strength
and ductility characteristics, as defined by the material standards. Further, the lower carbon, in
particular, means that weldability is significantly improved. Fracture toughness is improved as
well, indicating that fatigue performance and resistance to brittle fracture should be enhanced
(9). Nevertheless, localized effects of cold straightening, for example, continue to affect the
performance of the steel, especially in connection regions. On the other hand, the issue of
through-thickness strength and ductility, which used to be regarded as critical for the
performance of seismic connections, for example, has been found to be unimportant (10).
The basic quality and variety of structural steels available to designers and fabricators have
therefore been improved significantly, yet problems persist. As demonstrated earlier in this
paper, this has at least partly been caused by misinterpretation of materials standards and what
they imply for the actual in-structure performance of the steel. Of equal importance are clearly
functions that are controlled directly by the designer and the fabricator: it is unrealistic to expect
the material to provide for all of the stiffness, strength and deformability that are needed by the
structure under all expected service conditions.
An evaluation of the ductility and deformability requirements of the current American structural
steel design specification (11) is provided in the following. It is emphasized that most of these,
where they exist, are implied rather than explicit. This is most likely the result of an engineering
tradition of focusing on stress and strength rather than strain and deformation.
General observations
Between structural strength, stiffness and deformability, the first two are supplied relatively
easily, although improvements continue to be made through higher strength and improved
production methods. Further, many structures are controlled by the need for stiffness, in the
form of deflection or drift limits or dynamic response characteristics. For these cases the use of
higher strength steel is not advantageous. Framing system, high redundancy, and less reliance
on a limited number of structural elements are keys to successful performance.
Possibly of the greatest significance are the problems and solutions for the variety of connection
types and details that are utilized in structures. These are the regions where the material will be
exposed to the highest degrees of restraint, during shop fabrication and field erection, as well as
during high-demand service conditions. The connections influence local ductility demands and
framing performance, as evidenced by numerous examples from the Northridge earthquake (4).
Many reports of fractured welds and base metal details have been publicized. The basic
The following examples examine some of the primary American design criteria for steel
members; it focuses on certain elements only.
Tension members
Chapter D of the AISC Specification (11) details the strength criteria for tension members,
possibly the simplest structural elements, and the ones whose performance is closest to the
uniaxial conditions of the basic tension test. The limit states of gross cross section yielding and
effective net section fracture are well defined, although the reliability of the fracture case is less
than that of the overall yield. The reason for this is the greater variability of the tensile strength
(Fu) of the steel, as well as the influence of the geometry of the net section and the shear lag
associated with the cross-sectional shape and the placement of the end connection.
Ductility is recognized through the reference to strain hardening, stress concentrations, and the
importance of large deformations accompanying the yielding of the gross cross section. These
observations are based on various full- and reduced-scale tension member tests, but no data
are presented on actual deformation demands. However, in view of the relatively simple (other
than within the end connection regions) condition of these members and their satisfactory
behavior over the long term, it is generally accepted that ductility and deformation needs have
been assessed correctly. Deformation data are judged to be roughly comparable to tension
specimen tests, although specific results in support of this finding are not presented. However,
it is understood that the deformations that will occur in full-size tension members will be larger
than those of the material tests, primarily due to residual stress, initial crookedness and
eccentric application of the axial load.
Columns
Chapter E of the AISC Specification (11) gives the design criteria for columns and other
compression members. Since column buckling is primarily a stability phenomenon that is not
related to local or overall deformation demands, the issues of material performance are not
central to the issues at hand.
Beams
Chapter F of the AISC Specification (11) addresses the design criteria for laterally supported
and unsupported beams, and sections of Chapter B details local buckling and other
compactness issues. Chapter I gives the criteria for composite members; these will not be
examined here.
The overall behavior of beams is based on ultimate limit states involving in-plane or out-of-plane
failure. For example, for a laterally supported, compact beam, the ultimate limit state is the
development of a plastic hinge at the location of the maximum moment. The strength in this
case is therefore governed by the fully plastic moment, Mp, of the cross section. As another
example, for a laterally unsupported beam with an unbraced length larger than Lr, the ultimate
limit state is governed by elastic lateral-torsional buckling.
However, in all of these cases there is no clear indication of a required deformation or rotation
capacity. This is implied only through the criteria used to define compactness or the capacity of
the cross section to rotate after reaching the fully plastic moment. Specifically, flange and web
width-to-thickness ratios are established to allow full yielding in the cross section. In addition,
the beam has to be capable of rotating a certain amount beyond what constitutes the theoretical
In Table 2 only the requirements for the flange of a W-shape are given, as an example. As an
aside, it is interesting to observe that the current seismic b/t-criterion was used for non-seismic
applications as recently as the 7th edition of the allowable stress design specification of AISC
(1970); the change of the constant from 0.30 to 0.38 was made in the 8th edition (1980).
In the table, the term ry is the y-axis radius of gyration; θu is the rotation developed before local
buckling or strain hardening occurs; and θp is the rotation developed as the fully plastic moment
is reached. The unbraced length criteria pertain to the maximum length that will allow the
development of the fully plastic moment for a laterally unsupported beam. Since plastic design
response characteristics are needed for seismic conditions, the requirements are much more
demanding.
The key data in Table 2 are given in the third line, as the rotation demand ratio, θu/θp. The
original data for the non-seismic ratio are based on numerous beam tests, as reported by Yura
et al. (12). The history of the seismic deformation demand is not as clear; the demand ratio
value of 7 to 9 is primarily based on studies by Popov and others, but specific references for this
work cannot be cited.
For beams in 250 and 350 MPa yield stress steel, the rotation demand ratios are governed by
the occurrence of local buckling or strain hardening in the compression flange. Limited studies
have been made of higher strength steel, but research work at the U. S. Steel Research
Laboratory in the 1960-s and early 70-s showed that for Fy = 700 MPa, tension flange fracture
governed the beam behavior (13). At the time, U. S. Steel was exploring the potential
development of hot-rolled shapes in A514 steel; this was unsuccessful as a result of the limited
rotation capacity.
Studies of beams with yield stress values from 380 to 550 MPa are very limited at this time, and
no definitive conclusions can be reached for such members. However, the practical utilization of
higher strength beams is questionable, especially in seismic areas. This is in part caused by the
"strong column, weak beam" concept, as well as the fact that beam size is frequently governed
by stiffness, rather than strength. Since the modulus of elasticity is independent of the level of
yield stress, using higher strength material beams is unnecessary.
Welds and bolts are addressed in Chapter J of the AISC Specification and connection details
are covered in Chapter K (11). These are the most complex sections of the specification and
the attention given to strength limit states as opposed to deformations is very substantial. This
is done in spite of the fact that the deformation response often controls the actual ultimate limit
state.
In the following only some of the requirements will be examined. However, in view of the severe
deformation demands that are placed on many types of connections, it would seem important to
assess all of these specification criteria in detail, to gain a clear understanding of what is
expected of the material when the connections are designed according to the Specification.
This is especially important for many types of beam-to-column moment connections and some
welded tension member splices, for which localized material deformation demands can be very
high (2, 3).
The criteria for welded splices in very heavy wide-flange shapes are qualitative, but clear
recognition is given to the fact that the combination of residual stress, localized high deformation
demand due to fabrication operations, high localized hardness, and low fracture toughness in
the core area of hot-rolled shapes have the potential for leading to cracks and propagation of
cracks (2, 3). The event that caused this change in the AISC Specification was the bottom
chord fracture in one of the trusses for the Orange County Civic Center in Orlando, Florida. The
core area problem is much less important now, since continuous cast shapes have smaller and
less pronounced cores. The shapes that cracked in the Florida structure were all ingot-based.
Single lap welded joints will rotate when subjected to axial forces in the longitudinal direction,
due to the eccentricity of one plate relative to the other. The Specification recognizes the need
for a certain length of overlap between the plates or members in the joint, equal to five times the
thickness of the thinnest part, but not less than 25 mm. If this is satisfied, ". the resulting
rotation will not be excessive....." (11). No specifics are given as regards rotation magnitudes.
Short bolted joints generally deform in such a fashion that localized yielding allows for a
redistribution of the bolt forces, to load each bolt equally. This is not the case with long bolted
joints, for which the non-uniform strain distribution leads to significant differences in the actual
bolt loads. In particular, the outermost bolts will have the higher loads, leading to the potential
for an "unzipping" type of failure. The AISC Specification recognizes this behavior by reducing
the tabulated bolt strength values by 20 percent for connections longer than 1270 mm.
However, the actual deformation demand is only accounted for qualitatively.
This is one of the few cases where deformation and strength limit states are explicitly
recognized. Research has shown that the hole deformation will increase beyond 6 mm when the
nominal factored bolt load exceeds 2.4dtFu (11). Under many circumstances this will be
unacceptable, due to the contribution of such deformations to overall connection deformations.
If the bolt load increases to 3dtFu, the limit state will be that of hole ovalization or "dishing".
The criteria for the design of the details of beam-to-column connections are given in Chapter K
of the AISC Specification (11). The section provides extensive ultimate limit state criteria for
local flange bending, local web yielding, web crippling, sidesway web buckling, compression
web buckling, panel zone web shear, unframed beam and girder ends, additional stiffeners
requirements for concentrated forces, and additional doubler plate requirements for
concentrated forces. It is only in the treatment of panel zone web shear that strength and
deformation are explicitly recognized. It is formulated qualitatively, to the effect that different
equations are used to determine the nominal strengths if panel zone deformations are
considered in the frame stability analysis.
The Commentary for the Specification gives qualitative observations involving deformation
needs, such as "... that flange must be sufficiently rigid to prevent deformation of the flange ......"
(11). A detailed evaluation is provided for the panel zone behavior and the importance of its
deformation as regards the story and overall drifts of the structure.
SUMMARY
The paper has presented a discussion of issues related to performance demands for steel in
structures, especially under high restraint and high dynamic load conditions. It is shown that the
use of elementary materials standards requirements, which reflect uniaxial tension test
response, is unacceptable as a means to assess the response of the steel in the structure
during actual operating conditions.
Designers and fabricators must fully understand the material behavior. However, it is also clear
that the steel cannot assure satisfactory behavior by itself. Only together, through (1) material
choice, (2) local and overall structural design, and (3) shop and field fabrication techniques and
operations, will overall performance demands be met. In all cases strict adherence to specified
procedures is essential, for what good does a specification do if it is not used? Future
developments may see improved material standards, particularly if upper and lower limits are
placed on the specified yield stress values, and/or yield ratios are defined and required. The
criteria of the A992 standard of ASTM reflect a novel and very significant improvement.
REFERENCES
ABSTRACT
INTRODUCTION
The planned 2005 AISC Specification for Structural Steel Buildings (1) will include provisions
for both allowable strength design (ASD) and load and resistance factor design (LRFD). The
overriding principle guiding the development of these new provisions is that the steel does
not know what method was used in its design. Thus, there should be a single approach for
determining member strength and the modification of that strength to be consistent with the
ASD and LRFD loading provisions of the governing building codes. The AISC Committee on
Specifications and its Task Committees were charged with evaluating the existing ASD and
LRFD provisions and incorporating the best of both standards into the new standard. In
addition, research results that have become available since publication of the previous
standards, ASD in 1989 (2) and LRFD in 1999 (3) should be incorporated. Thus, this new
specification should be a step forward for each design approach.
The original LRFD Specification (4) was calibrated to the then existing ASD Specification (5)
for a live load to dead load ratio, L/D = 3, in order that the new specification produce designs
that were comparable to the existing provisions. The ASD and LRFD design philosophies
are stated in the 2005 Specification (1) draft for ASD as
Rn
( D + L ) = Ra ≤ (1)
Ω
and for LRFD as
(1.2 D + 1.6 L ) = Ru ≤ φ Rn (2)
where Ra and Ru are the required strengths determined from the ASD and LRFD load
combinations, Rn is the nominal strength, φ is the resistance factor for LRFD, and Ω is the
safety factor for ASD . For the load combination of (1.2D + 1.6L) and L/D = 3, the effective
load factor becomes 1.5. Thus, the relationship between φ and Ω can be determined by
solving Eqs. 1 and 2 for Rn and setting the them equal, thus
1.5
Ω= (4)
φ
This relationship guides the development of factors of safety for the ASD provisions of the
unified specification, based on the LRFD resistance factors. It will be seen that the resulting
strength provisions for tension members in the 2005 Specification (1) will be the same as
they are in the current ASD (2) and LRFD (3) Specifications. The provisions for shear lag will
be modified slightly to account for some recent research results.
Block shear has seen several changes over the years since it was first introduced into the
specification. However, it will also be seen that the strength provisions for 2005 are
essentially the same as the ASD and LRFD provisions with a slight variation in the
controlling factor.
SHEAR LAG
Shear lag provisions were first introduced in the 1978 AISC ASD Specification (5). This was
to account for the research findings that the net section was not fully effective in providing
fracture strength when all elements of the tension member section were not attached to the
connecting elements The provisions of Section 1.14.2 simply stated that the effective net
area was to be taken as the net area times a reduction factor, thus
Ae = Ct An (5)
1. W, M, or S shapes with flange widths not less than 2/3 the depth, and structural tees
cut from these shapes, provided the connection is to the flanges and has no fewer
than 3 fasteners per line in the direction of stress. Ct = 0.90.
2. W, M, or S shapes not meeting the requirements of subparagraph 1, structural tees
cut from these shapes, and all other shapes, including built-up cross sections,
provided the connection has not less than 3 fasteners per line in the direction of
stress. Ct = 0.85.
3. All members whose connections have only 2 fasteners per line in the direction of
stress. Ct = 0.75.
The commentary to Section 1.14.2 indicates that these values are reasonable lower bounds
for profile shapes and connection means described in the research of Munse and Chesson
(6). In that research they proposed an equation
Ct ≈ 1 − x l (6)
that, although not a part of the actual specification provisions, was included in the
commentary.
The 2005 Specification (1) needed to consider how the combination of ASD and LRFD
provisions would impact tension member strength and how, if at all, the effective net area
provisions would need to change. Using the relation between φ and Ω presented in Eq. 4,
the design tensile strength, φPn, and the allowable tensile strength, Pn/Ω, for the limit state of
fracture can be taken from
Pn = Fu Ae (7)
This is the same provision as given in the 1999 LRFD Specification (3) and a comparison
with the 1989 ASD Specification (2) shows that it provides the same allowable strength.
Since effective net area is not a function of design approach, there will be no impact there.
There are two changes being considered for the 2005 shear lag provisions. The first is the
removal of the upper limit on U. There does not appear to be sufficient research results to
warrant retaining this limitation. The work of Munse and Chesson (6) did not include a
recommendation that the shear lag reduction factor have an upper limit, although their
testing program included only a few cases where this might have come into play.
The second change is the addition of a requirement that single angles, double angles, and
WT’s be proportioned so that U is equal to or greater than 0.6. Alternatively, a lesser value of
U is permitted if the tension member is designed for the effect of eccentricity through the use
of the interaction equations. Recent work reported by Epstein and Stamberg (8) suggested
that for WT sections, a lower bound of 0.65 be placed on U.
Since the 2005 Specification (1) will incorporate the previous separate specifications for
single angles and for HSS, the number of shear lag cases has increased. Thus, a table is
being provided to simplify the determination of appropriate U-values.
BLOCK SHEAR
As was the case for shear lag, block shear provisions first appeared in the 1978 ASD
Specification (5). These provisions were the result of the work of Birkemoe and Gilmor (9)
that was directed at the coped beam connection. The provisions, as stated in Section
1.5.1.2.2, indicate that the shear at beam end connections where the top flange is coped,
and in similar situations where failure might occur by shear along a plane through the
fasteners, or by a combination of shear along a plane through the fasteners plus tension
along a perpendicular plane on the area effective in resisting tearing failure, the stress was
to be limited to 0.3Fv This would have resulted in an allowable force equation of
The commentary provided an alternative where the tension and shear areas could be treated
separately as
Va = 0.3Fu Anv + 0.5 Fu Ant (9)
The 1989 ASD Specification (2) brought Eq. 9 from the commentary into the specification
and provided the direction that block shear should also be considered for welded
connections.
Block shear provisions are not as clearly articulated in the 1986 LRFD Specification (4). The
provision for shear fracture is given in the body of the specification but the discussion of
block shear is relegated to the commentary. In the commentary presentation, the possibility
of a combination of yielding on one plane and fracture on the other is introduced and the
following two equations are given
with φ = 0.75 and the largest value of φRn taken as the design strength. It is worth comparing
this first introduction of LRFD block shear provisions with the ASD provisions using the
relationship of φ and Ω presented in Eq.4. Since φ = 0.75, then Ω=2.0. Thus, dividing the
nominal fracture term from Eqs. 10 ands 11 by the safety factor of 2.0 and adding them,
yields Eq. 9.
The 1993 LRFD Specification (10) brought the two block shear equations forward into the
specification but altered the way that the controlling equation was selected. For these
provisions, the controlling factor was fracture. Shear fracture and tension fracture were to be
calculated and the larger fracture term was to be combined with the opposite yield term.
Additional research reported by Ricles and Yura (11) for coped beams and Hardash and
Bjorhovde (12) for gusset plates confirmed that the strength could be determined by the
summation of the shear and tension terms.
The 1999 LRFD Specification (3) continues to use the previously presented block shear
equations but sets an upper limit on the strength that is equal to the sum of the two fracture
terms and can be stated as
which is actually a return to the commentary equation of the 1978 ASD Specification (5),
since Rn from Eq. 12 divided by the factor of safety, 2.0, yields Eq. 9.
For the 2005 Specification (1), the currently proposed block shear provisions have
undergone another slight modification. For shear strength, either fracture or yield, the
relations remain unchanged. For tension strength, two revisions are recommended. The first
is the recognition of the influence of non-uniform tension, as would occur on the block shear
tension face for a coped beam with two rows of bolts, as identified by Ricles and Yura (11).
The second change is the use, for all conditions, of the tension fracture strength, rather than
either tension fracture or yield strength. The resulting provisions are
When the tension stress is uniform, Ubs = 1.0 and for cases where the tension stress is not
uniform, Ubs = 0.5. This is consistent with the recommendations of Kulak and Grondin (13).
Although the committee considered a more involved approach to the calculation of Ubs, it
was decided to simplify the term so as not to make its determination laborious.
CONCLUSIONS
An evolution of the shear lag and block shear provisions has taken place since their first
introduction in the AISC Specification in 1978 (5). Although the changes have been slight,
they reflect an improving understanding of the behavior that they are attempting to predict.
The intent has always been to provide specification provisions that are sufficiently accurate
so as to provide for safe and economical structures while at the same time providing design
methods that are simple and economical to apply.
The 2005 “unified” Specification (1) is scheduled to receive final approval by the end of
2004. This schedule has been set in order that it may be incorporated into the next revisions
of the NFPA and IBC building codes. Although the provisions presented here for the 2005
Specification are those under consideration at this time. The committee process allows for
changes to be made until the final Specification has received Committee on Specifications
approval. Thus, this description of what might be ahead for the designer should not be relied
upon for design.
NOTATION
1. AISC, (2004). Specification for Structural Steel Buildings, Draft dated March 2004,
American Institute of Steel Construction, Chicago, IL.
2. AISC, (1989). Specification for Structural Steel Buildings – Allowable Stress Design and
Plastic Design, American Institute of Steel Construction, June 1, Chicago, IL.
3. AISC, (1999). Load and Resistance Factor Design Specification for Structural Steel
Buildings, American Institute of Steel Construction, December 27, Chicago, IL.
4. AISC, (1986). Load and Resistance Factor Design Specification for Structural Steel
Buildings, American Institute of Steel Construction, September 1, Chicago, IL.
5. AISC, (1978). Specification for the Design, Fabrication and Erection of Structural Steel
for Buildings, American Institute of Steel Construction, November 1, Chicago, IL.
6. Munse, W. H., and Chesson, Jr., E. (1963), “Rivited and Bolted Joints: Net Section
Design,” Journal of the Structural Division, ASCE, Vol. 89, No. ST1, February, pp. 49-
106.
7. Easterling, W. S., and Giroux, L. G., (1993), “Shear Lag Effects in Steel Tension
Members,” Engineering Journal, AISC, Vol. 30, No. 3, 3rd Quarter, pp. 77-89.
8. Epstein, H., I., and Stamberg, H., (2002), “Block Shear and Net Section Capacities of
Structural Tees in Tension: Test Results and Code Implications,” Engineering Journal,
AISC, Vol. 39, No. 4, 4th Quarter, pp. 228-239.
9. Birkemoe, P. C., and Gilmor, M. I., (1978), “Behavior of Bearing-Critical Double –Angle
Beam Connections,” Engineering Journal, AISC, Vol. 15, No. 4, 4th Quarter, pp. 109-
115.
10. AISC, (1993). Load and Resistance Factor Design Specification for Structural Steel
Buildings, American Institute of Steel Construction, December 1, Chicago, IL.
11. Ricles, J. M., and Yura, J. A., (1983), “Strength of Double-Row Bolted Web
Connections,” Journal of the Structural Division, ASCE, Vol. 109, No. ST1, January, pp.
126-142.
12. Hardash, S. G., and Bjorhovde, R., (1985), “New Design Criteria for Gusset Plates n
Tension,” Engineering Journal, AISC. Vol. 22, No. 2, 2nd Quarter, pp. 77-94.
13. Kulak, G. L., and Grondin, G. Y., (2002), “Block Shear Failure in Steel Members – A
Review of Design Practice,” Connections in Steel Structures IV, Proceedings of the
Fourth International Workshop, AISC, pp., 329-339.
ABSTRACT
Nowadays engineers still tend to optimize a structure to a minimum of
kilograms, which is the only hard criterion that is available for an engineer
during the design of the structure. In the end this criterion leads to structures
that are expensive and have poor quality.
In this article it is illustrated that selection of a deeper top- and bottom chord
member and the use of 3D-analysis software leads to a reduction of costs of
the total project. Furthermore, it results in improvement of the quality of the
structure as a whole.
THE ISSUE
For most of the structures it applies that they are established in the same way for many
years. An architect makes an architectonic design, a design engineer designs the main
structure and a steel fabricator takes care that the construction will be detailed, fabricated
and erected.
These phases are mostly sequential and do not have much overlap. Consequence is that the
expertise and involvement of the several parties is restricted to his/her own specialty.
Because of this strict separation, but also because of ignorance and incapability, the
engineer often disregards the phases that come next. Most of the times the result is that the
detailing of the structure becomes very complex and that the structure will be difficult to
fabricate.
These things do not have a positive effect on the project financially, constructively,
aesthetical, as well as systematically. By keeping the detailing of the steel into account in an
early stage, a better and more cost effective structure can be designed.
From their own disciplines, ICCS bv and IDCS bv are involved with a lot of steel structures
and they observe that the abovementioned issue frequently occurs. Furthermore, it increases
the last couple of years because the quality of a main design decreases strongly.
STARTING POINTS
Inspired by many examples from the daily practice, the issue will be illustrated by discussion
of a standard truss.
Based on drawings of the architect the design engineer designs a steel structure. This
structure consists of a large amount of trusses that span a large space in two directions.
Therefore, both in longitudinal and transverse direction the main structure is modelled by
trusses that lay on a concrete substructure. For the analysis the centre lines of the profiles
are modelled. Members that come together in a node, thus theoretically intersect at one
point. Result is that there is a simple transfer of forces and no secondary forces or moments
Figure 1a. Centre lines intersect (actual). Figure 1b. Centre lines intersect (analysis model).
Figure 2a. Eccentricities e = 100 mm (actual). Figure 2b. Eccentricities (analysis model).
If the structure is made of concrete, the abovementioned approach can be followed. Because
of weight reductions, costs, construction time and erect onsite most of the time the trusses
are made of steel.
But the problem arises when the engineer has realised an “optimum” design on the basis of a
minimum of kilograms. For certain if, as in this outlined situation, high forces are involved. A
small eccentricity, by moving the centre lines, leads to a considerable extra bending moment.
NEN 6772, article 11.6.1 mentions that at determining the joint capacities the eccentricities
may be neglected if:
e e
− 0,55 ≤ ≤ 0,25 or − 0,55 ≤ ≤ 0,25
d0 h0
e=0 e=negative e=positive
Moreover, problems are not only caused by shifting the gridlines, but the area of force
application also deserves attention. When the diagonals and bottom- and top chord member
have similar profiling, it is very difficult to realise proper force lead-in without stiffeners.
The maximum allowable normal force in a profile is many times higher than the maximum
allowable shear force. Moreover, this is being reduced considerable when it concerns a web
that is sensitive for local buckling. Proper force lead-in without stiffeners is almost impossible
when the maximum allowable normal force in the diagonal is higher than the shear force
capacity of the bottom- and top chord member.
So not only does the bottom- and top chord member need to have an capacity left in order to
withstand the extra moments due to eccentricities, but, to prevent the need of stiffeners, they
also need to have a shear force capacity that is higher than the normal force capacity of the
diagonal.
The engineer often seems not to be aware of this. In the design he does not, or too little, take
this into account. If he chooses profiles with little reserve capacity in the design stage, it
results in fairly complicated, laborious and therefore expensive details.
The last couple of years the awareness that costs are caused largely by the connections has
sunk in. See former publications (1) regarding costs comparisons in the steel industry.
Additional stiffeners must be prevented because these components are labour-intensive.
Preference is given to simple connections with fin plates, angle cleats or endplates. The
recent price increase of the material is the reason that saving on material becomes an
important item. A good insight in total costs is and stays essential.
In order to come to a structure that is acceptable for all parties, there is a number of
solutions. Every party has its own preference for the ideal structure. An engineer does not
want (m)any eccentricities, a Steel fabricator wants the most simple fabrication possible
without a lot of labour and an architect wants an aesthetical justified structure, especially
when the steel structure remains visible. There is a number of methods to decrease the
amount of labour in the construction afterwards. Below you will find a few possible solutions.
Dealt with is the detail in the bottom chord member, but the same is applicable to
connections at the top chord member.
If the starting point remains that the centre lines have to go through one point, the profiles
can be rotated 90 degrees. If necessary, the thickness of profiles must be locally adjusted at
the joint because of the various profile dimensions.
In a truss it only concerns tension or compression members. By detailing the joint with two
plates on the flanges and not bothering the centre lines, no extra forces are introduced and
the connection can be realised with sawing and drilling as main labour. During the
processing stage the tolerance on the clearance between bolts and boltholes have to be
taken into account. This can be done by shortening the tension members or by use of
injection bolts. Furthermore the introduction of holes in the bars means a reduction of the
cross section area. If the tension members are used up till their full capacity, the weakening
due to holes lead to a reduction of the capacity, so that the profile will fail.
The thickening of the profiles, as mentioned above in connection with various profile
dimensions, gives as a side-advantage an increase of the tensile force capacity, so that the
profile complies again. Applying this solution mostly only make sense when the truss in the
nodes is directly loaded by purlins or floor beams.
If a floor or roof rests on the top or bottom chord member and this causes moments, it is
possible that the profiles do not meet the requirements anymore in connection with moments
about the weak axis of the profile.
When the diagonals are shifted away from the node simple endplate connections are made
possible and the connection no longer overlap. To enable this, the bottom chord member in
the truss must be deeper than is strictly necessary for transfer of the normal- and shear
forces that appear. This solution cuts both ways: because of the overcapacity of the bottom
chord member it is easy to take up eccentricities. The top flange moves upwards so that the
eccentricities of the diagonals become smaller.
ho e ≤ ho/4
Suppose: a truss with a span of 20 meter, completely built up from HE200A profiles both for
the bottom- and top chord member and the diagonals.
e = 100 mm
If the diagonals are connected to the bottom- and top chord member via endplates, the
eccentricity, that occurs at the connection of the diagonal on the bottom chord member will
have to be somewhat like100 mm. Suppose: a maximum normal force (Npl HE200A) in the
diagonal of 1200 kN. This gives a vertical force of 1000 kN and thus a moment of 100 kNm.
The maximum moment a HE200A can transfer is 101kNm, without keeping into account the
buckling lengths and moments that are already present as a result of the direct load. So the
“extra” moment cannot be withstood by the bottom chord member. In this realistic situation
the maximal normal force is 600 kN and thus the vertical force is 500 kN. As a result of the
eccentricity this leads to an extra moment of 50 kNm on the beam.
When the Head engineer has totally utilized the bottom- and top chord member, this moment
too cannot be withstood. Not any “extra” moment can be absorbed, so that the steel
fabricator is forced to connect the diagonals in the centre lines, which results in a
complicated and above all laborious connection design.
But what is the maximum normal force that can be absorbed without application of stiffeners
in the bottom and top chord member ? According to article 14 of the NEN6770 the bottom
chord member should be checked for:
⎧⎪ t f ⎛t ⎞⎛ c ⎞⎫⎪
Local buckling of the web Fu;2;d = 0.125 t w2 Ef y;d ⎨ + 3⎜ w ⎟ ⎜ ⎟⎬
⎪⎩ t w ⎜ t f ⎟ ⎜ h − 2t f ⎟⎪
⎝ ⎠⎝ ⎠⎭
⎧⎪ t f ⎛t ⎞⎛ c ⎞⎫⎪
Fu;2;d = 0.5 t w2 Ef y;d ⎨ + 3⎜ w ⎟ ⎜ ⎟⎬
⎪⎩ t w ⎜ t f ⎟ ⎜ h − 2t f ⎟⎪
⎝ ⎠⎝ ⎠⎭
From this follows a maximum normal force in the diagonal of 337 kN without the need of
stiffeners in the bottom chord member, whilst the maximum allowable normal force for this
diagonal is 860 kN (HE200A with a buckling length of 3.60m). By strengthening the bottom
chord member to HE240B the normal force in the diagonal becomes 700 kN without
application of the necessary stiffeners in the bottom chord member. A side-advantage is the
increase of the moment capacity of the profile. The moment capacity now becomes 247
kNm. The eccentricity that has become even smaller because of the deeper profile, now
does not give any capacity problems anymore. By only using a deeper profile for the bottom
chord member (and top chord member) the connections are much easier to fabricate.
If the first two solutions are not possible, web plates can also be chosen. The shear force
capacity of the profile will then increase. The truss built up from HE200A-profiles is taken as
an example again.
Increasing the web thickness of the HE200A into 20 mm (original 6.5 mm), the extra moment
as result of eccentricities can be absorbed without problems, as a result of the eccentricity.
With the available analysis tools it is simple to check if the selected profile complies.
The search for alternative detailing afterwards can be avoided totally, if the engineer in
his/her main design takes account of the fact that he tries to design a simple constructible
structure. In other words: a structure that not only is optimum according to his own criteria,
but is also perfect for the steel fabricator as well as the client. Nobody, not even the client, is
waiting for a design that has an optimum weight, but has complicated details that make the
steel structure unnecessarily expensive.
The conclusion is that all three outlined solutions show a considerable reduction of the total
costs in comparison to the original detail. Despite an increase of using material, due to
solutions 1 and 2, the connection has become a lot less complex in comparison with the
original connection design.
DESIGN STAGES
By going through the design stage of a construction once again, there will be indicated how
such problems can be prevented.
The client and the architect set up a list of requirements for the building that has to be
constructed. In the light of this list in combination with the building regulations and the
building code a certain material or a combination of materials to execute the design will be
selected. Every material has its own features. Timber and especially concrete are suitable to
withstand compressive forces, whilst steel is preferably suitable to withstand tensile forces.
The choice of material defines the shape of the main structure it is important that the right
material is used on the right spot.
Points for attention must be:
• Is there taken care of the stability of the structure by means of rigid plates, portals or
bracings
• Are high vertical forces transferred, by trusses with compression-, tension or
compression and tension members
• Is there any concentration of forces, as is the case with structures with large spans.
• How are forces lead in into the structure below.
In daily practice it often seems as if during the main design no account is taken at all of the
fact that the members must be joined.
When the engineer has made the first calculation and thus has an idea about the profiles, he
has to think about how the several profiles must be joined. By choosing for example a deeper
(=stronger) profile a lot of (detailing) sorrow can be prevented. An engineer also can insert
the eccentricities in his calculation and base profiling there-upon. But this is not the usual
procedure for calculations of steel constructions; moreover it gives a considerable increase
of the shear forces and moments in the top chord member and bottom chord member. Also
separating the nodes is a rather laborious for the constructor.
3 2
3 2 3 2
3 2
excentriciteit is 100 mm
3 2 3 2
These diagrams show clearly that the maximum moments and especially the shear forces
increase considerably by introducing eccentricities.
A design engineer, who thinks before designing about the way the beams and colums have
to be connected, enables the steel fabricator to construct an efficient and constructible
structure with the use of simple connections. This results in benefit for all parties concerned.
CONCLUSION
Nowadays engineers still tend to optimize a structure to a minimum of kilograms, the only
hard criterion that is available for an engineer during the design of the structure. Finally this
criterion leads to structures that are more expensive and have less quality. Working together
in a design and built-team in which the fabricator also has to assist in the design stage can
prevent these problems.
REFERENCES
(1) H.G.A. Evers, F. Maatje, ICCS bv, Cost based engineering and production of steel
constructions, Proceedings of the Fourth International Workshop, Connections in
steel structures IV, Behavior, strength and design, published 2002, Actas do II
Entrocontro Nacional de Construção Metálica e Mista 2, published 1999.
(2) F. Maatje, ECCS bv, Voorcalculeren met behulp van de computer (Cost estimation
with the computer), Bouwen met Staal 1989
ABSTRACT
INTRODUCTION
Education has always been seen as an essential part of the introduction and dissemination
of new methods for the design of steel connections. One of the first educational packages
specialised on connections was produced by Owens and Cheal (1) who prepared
educational material for structural connections. This material has been extended and is now
incorporated into a European educational package called the European Steel Design
Educational Programme, ESDEP (2). This programme is used today by educational
establishments throughout Europe. Other educational packages which build on the work of
ESDEP are available some of which include WIVISS (3) and (4), a set of lectures on CD,
SteelCall (5) and (6), a virtual steel designers office, and SSEDTA which consists of a set of
a) b)
For more than twenty year the European Convention for Constructional Steelwork’s
Technical committee for structural connections (ECCS TC10) has supported the
development and implementation of a common set of design rules for steel connections. It is
therefore not surprising to find that one of this committee’s priorities is to facilitate the
transition of EN1993-1-8 from a European pre-standard; see (10), (11) and (12), to a full
Euro-norm (13). Part of this activity is the development of the necessary educational material
to encourage designers throughout Europe to adopt EN1993-1-8. Consequently, a
programme called “Continuing Education in Structural Connections” (CESTRUCO) was
formed under the European Commission’s Leonardo initiative to collect commonly asked
questions on the background, implementation and use of prEN1993-1-8 and to publish
expert answers to these questions. The CESTRUCTO project was developed from an idea
by Mr. Marc Braham (Astron, Luxembourg), Prof. Jan Stark (TU Delft, The Netherlands) and
Mr. Jouko Kouhi (VTT, Finland) to provide designers with more detailed information on the
background and implementation of the design methods given in prEN1993-1-8.
Collection of questions
The collection was based on publications in national journals and on local seminars. The
questions were collected by project internet page, Figure 1, and in paper form as well. The
questions located during the conversion of ENV 1993-1-1 (including Annexes J (11), K (12),
L (10)) into EN 1993-1-8 were taken into account. Together 632 questions were collected
related to the topic.
Answering
All obtained questions were very good. Answers to 101 selected questions were chosen in
the second part of work based on its educative contribution. The agreement between all
partners was not reached for all answers. This is the reason that some nice questions are
missing in final material. The review of answers was prepared in tree steps: locally between
partners, by delegated partners and externally by members of ECCS TC10.
Dissemination
Under the third part of the project, dissemination, were prepared the educational materials
Textbooks and Internet/CD lectures. To facilitate easy of use questions/answers are split into
the following Chapters: Introduction, Bolts, Welding, Structural Modelling, Simple
Connections, Moment Resistance, Connections, Column Bases, Seismic Design, Fire
Design, Hollow Section Joints, Cold-Formed Member Joints, Aluminium Connections, and
Design Cases. Each chapter starts with a brief over-view of the method use in prEN1993-1-
8. This is followed by the commonly asked questions together with their answers. The
materials were localised for use in the partner’s national languages, Figure 2: Czech (300
copies of textbook was printed), Dutch (200 copies), English (1500), Greek (200),
Portuguese (1000), Rumanian (300) and Swedish (200). The translations into Polish and
Spanish are under preparation (written in February 2004). The pilot Seminars of the project
were organised in project partner’s countries to test the material in local environment and to
start to disseminate the material.
c) d)
Figure 3. Example from Textbook Chapter 13 Design cases - a), c) bad solutions of beam to
column joint, b), d) good cases.
2,29°
0,04 rad
a) b)
Figure 4. Example from Textbook Chapter 13 - Design cases a) bad solution of column
base, b) good design.
a) b)
INTERNET / CD LESSONS
The Internet/CD version, Figure 6, of project materials is based on Microsoft Windows help
format, which is a robust tool for education. The material prepared by RoboHelp tool (14)
allows the easy navigation in partner’s languages and in German and French. The
Internet/CD lessons are equipped with worked examples, Figures 3 and 4, PowerPoint
presentations, slides, videos, worked examples (15) and (16), animations of design cases,
educational software for connection modelling, and design tools available round the Europe.
Partners were so kind to equipped us with their demo version to show how easy may be nice
connection checked by EN1993-1-8. 3000 copies of CD first version were printed at the end
of the project in December 2003. The second version was disseminated in June 2004 based
on evaluation of first ones. Third upgraded version is intended to be published based of
interest of structural steel practice in June 2005.
c) d)
On the CD may be found the NASCon (Non-linear Analysis of Steel Connections) program
that offers a computer user-friendly tool for the component method. The tool allows modelling
the nonlinear behaviour of different components; see (17) Figure 7. The video film, see
Figure 8, demonstrates the correct design of T-stub connections and bolted splices to avoid
a fatigue failure of bolts. The related to the fire design is equipped with the PowerPoint
lessons including video/audio sequences, see Figure 9.
a) b)
Figure 8. The flow of the stresses in the connection in video film Statically Stressed Bolts
in Dynamically Loaded Connections.
a) b)
Figure 9. a) Front page of PowerPoint lesson ”Connection Design for Fire Safety”,
b) a slide from the lesson “Heating and Cooling of Structure”.
ACKNOWLEDGEMENT
The authors wish to dedicate this work to Mr. Martin Steenhuis, our good friend, who worked
with us in the field of structural connections for many years, launched this project, and who
tragically died in the summer of 2001.
REFERENCES
(1) Owens G. W., Cheal B. D.: Structural Steelwork Connections, Butterworths, London
1988, ISBN 0-408-01214-5.
(2) ESDEP, European Steel Educational Programme, SCI, London, 1994, www.esdep.org.
(3) Plank R.J.: Wider Vocational Initiative for Structural Steelwork, J. Construct. Steel
Research, 46, (1998), pp 278-279, ISBN 0-08-042997-1.
(4) Chladná M. - Wald F. - Burgess I. W. - Plank R. J.: Contribution of the Structural
Steelwork Educational Programme WIVISS, v Proceedings of the Conference
Eurosteel ´99, Studnička J., Wald F., Macháček J. ed., Vol. 2, Prague,
26 - 29 May 1999, ČVUT Praha, 1999, s. 137 - 140, ISBN 80-01-01963-2.
ABSTRACT
The crucial challenge in MBT (Mixed Building Technology) are the connection
elements between the different construction systems. A typical example of an
MBT realisation is the ski jump tower in Innsbruck where a steel cage of hollow
section trusses is cantilevering up to 12,5 m from the central concrete tower in
a height of about 35 m above ground. The transfer of the localised truss forces
into the concrete box section was solved by special pre-stressed docking
devices.
INTRODUCTION
The ski jump in Innsbruck known for the famous annual New Year ”Four-ski-jump-tour“ was
renewed. The original jumping tower (built for the 1976 Olympic winter games) was pulled
down and a new landmark similar to a lighthouse was erected. Zaha Hadid (London) won the
international architectural competition for this significant building. The constructional
realisation was ordered from aste konstruktion and was honoured with the “Austrian State
Award for Consulting 2002”. A speciality of this MBT are the high tension docking devices
between the cantilevering steel cage and the concrete tower.
PROJECT OVERVIEW
“Bergisel” - a glacier hill located to the south of Innsbruck (Figure 1) – has a significant history:
Offering place of the Celts, path marking for the roman-german emperors moving to Rome,
battle field of the Tyrolean war of liberation, centre of the Olympic Winter Games in 1964 and
1976 with the two flame basins and – already since more than 50 years - scene of the
international Four-ski-jump-tour at New Year. In the years 2001 and 2002 these facilities were
fully renewed and extended by new buildings (Figure 2, 3):
With 49 m above ground the ridge height of the concrete tower reaches 791 m above sea
level (Figure 4). The foundation was solved with a plate of 20 x 20 x 1,0 m at a level of –11 m
below ground with three basement storeys. The standard cross section of 7 x 7 m and a wall
thickness of 40 cm rises up for about 40 m above the foundation, stabilised with wall
diaphragms to the base plate limits. It contains the two elevators, the stairway and the supply
pit. From 29 m above ground the cross section tapers to 3,7 x 7 m making place for the
jumping access stairway.
Also at this level of cross section change the support girder for the ramp bridge cantilevers
4,5 m with a height of only 1,45 m. This slenderness was necessary to hide this girder in
between the steel truss flanges of the bridge.
The demand of fair-faced concrete in combination with the difficult access and supply
conditions resulted in the choice of a climbing formwork. Concreting started in June 2001.
The tower top is not ordinary - neither in architectural nor in constructional respect. A three-
level steel cap with a rescue level, a restaurant and an observation platform is docked to the
central concrete tower (Figure 5). Being 250 m above the city centre one has a fantastic view
on Innsbruck and the surrounding mountains.
The levels are cantilevering around the concrete core up to 12,5 m. Together with the steel
hollow section frames and the diagonal suspension tubes anchoring back to the concrete
core a steel cage is built (Figure 6). The horizontal stiffening to the core is realised by the
trapezoidal composite slabs. The transparency and elegance of the facade is supported by
the fact that diagonal bars within the front could be avoided and huge glass elements were
placed into the facade.
The crucial challenge in MBT construction are the connection elements between the different
construction systems. Thanks to the common, material-independent design and safety
concept of the constructional Eurocodes the interface problems at the level of design
methods and internal forces lost its deterrent effect and MBT solutions become more and
more usual in daily design practice. The effect is a more economical use of different
materials related to their constructional benefits (strength, stiffness, weight, prefabrication,
strengthening, dismanteling,...) and more innovative architectural solutions.
For the actual case of the Bergisel Ski Jump the considerable docking forces between the
steel cap and the concrete core had been a crucial challenge which was solved by special
pre-stressed steel brackets (Figure 7). These elements of at maximum 550 kg weight were
integrated into the formwork with a tolerance of less than 1 cm.
tension
compression
prestress
The load transfer into the concrete walls was handled with mutual pre-stressing cables
(interior tendons) from one docking point to the opposite one going through the tower.
Additional concentrated rebars in the local load introduction zones were provided to cover
the bursting forces and for crack distribution. The characteristic docking forces can be taken
from Figure 8.
The resulting necessary welding length led to the geometry of these docking brackets. By the
use of four longitudinal ribs which were welded on site to the slotted push-over hollow
sections the total bracket length could be minimised. The eventual negative influence of the
high welding temperatures on the end anchorage of the pre-stressing strands could be
dispelled by a test specimen. The maximum heat increase was measured to be only 50
degrees.
Attention has to be paid to the fact that the tendon head is no more accessible after
positioning of the steel cage. Therefore this application type is limited to quasi-static loading.
CONCLUSIONS
REFERENCES
(1) Aste, C., Glatzl, A., Huber, G. (2002). Ski jump „Bergisel“ – A new landmark of
Innsbruck. 3rd Eurosteel Conference, ISBN 972-98376-3-5, Lisbon.
(2) Aste, C., Glatzl, A., Huber, G. (2002). Schisprungschanze „Bergisel“ – Ein neues
Wahrzeichen von Innsbruck. Stahlbau Heft 3/02, S.171-177, ISSN 0038-9145,
Ernst&Sohn, Berlin.
(3) Aste, C., Glatzl, A., Huber, G. (2003). Steel-concrete mixed building technology at the
ski jump tower of Innsbruck, Austria. International Journal on Steel and Composite
Structures, ISSN 1229-9367, Technopress, Korea.
(4) Aste, C., Glatzl, A., Huber, G. (2003). Innsbruck ski jump: a triumph of mixed building
technology. Concrete Journal, The Concrete Society, Berkshire.
(5) Aste, C., Glatzl, A., Huber, G. (2003). Schisprungschanze „Bergisel“ – Ein neues
Wahrzeichen von Innsbruck. Zement + Beton, www.zement.at, Wien.
ABSTRACT
A systematic analytical procedure for characterization of the monotonic defor-
mation behaviour of single T-stub connections is presented and discussed. The
model is based on the prying mechanism implemented in Eurocode 3 and in-
cludes the deformations from tension bolt elongation and bending of the T-
element flange. The results of a series of tests (numerical and experimental) on
T-stubs that were conducted at the Delft University of Technology and the Uni-
versity of Coimbra are used to validate the procedures.
INTRODUCTION
The T-stub model is widely accepted as a simplified model for the idealization of the behav-
iour of the tension zone of bolted joints, which is often the most important source of deform-
ability of the whole joint (figure 1a) (1, 2, 3). The equivalent T-stub corresponds to two T-
shaped elements connected through the flanges by means of one or more bolt rows. The
system is loaded by applying a tensile force at the top of the webs that induces deformation
of the flanges (figure 1b). The load-carrying behaviour of this simple assembly is inherently
nonlinear and hence its characterization is not easily open to simple analytical formulations.
Tests (both experimental and numerical) are frequently carried out to obtain the actual re-
sponse, which is then modelled approximately by mathematical expressions that relate the
main properties: initial stiffness, ke.0, post-limit stiffness, kp-l.0, full plastic resistance, FRd.0, ul-
timate resistance, Fu.0 and deformation capacity, ∆u.0.
From a practical point of view, the characterization of the T-stub behaviour by means of ei-
ther of the above methods is not expedited. A simple methodology for prediction of the con-
nection response up to collapse is desired. The collapse is governed by fracture of the bolts
and/or cracking of the T-stub material. Because of the emphasis placed on connection ductil-
ity, the methodology must be able to predict the response of the T-stub well into its plastic
and strain hardening range with a reasonable degree of accuracy.
This paper presents simplified methods for determining the monotonic deformation response
of T-stub connections. First, existing models from the literature are discussed. Next, a two-
dimensional model is proposed and calibrated against test results from a database compiled
by the authors (4, 5). Some recommendations for modifications are also given. Finally, con-
clusions are drawn.
F
2
F ⎛∆⎞
2 ⎜⎝ 2 ⎟⎠
(C) (D)
(A)
B
Q
(a) T-stub identification in the particular case of an ex- (b) T-stub basic prying mechanism
tended end plate bolted connection (3). (one quarter-model).
PREVIOUS RESEARCH
The European code of practice for the design of structural steel joints in buildings, Eurocode
3 (6), provides design rules for the evaluation of ke.0 and FRd.0, based on elastic (3, 7, 8) and
pure plastic theories (1, 2), respectively. The post-limit stiffness, kp-l.0 is taken as zero, which
means that strain hardening and geometric nonlinear effects are neglected. Concerning the
component ductility, Eurocode 3 (6) presents some qualitative principles based on the main
contributions of the T-stub deformation: whenever the bending deformation of the flanges
governs the plastic mechanism, the ductility is infinite; should the bolt determine collapse, the
ductility is limited. These rules are, yet, insufficient. In fact, if simple plastic theory is applied,
the limit resistance and deformation are determined by yielding of the flange. Therefore, con-
sideration of strain hardening is crucial to carry out an ultimate analysis of the system up to a
fracture condition.
To fill in this gap, several theoretical approaches for the characterization of the overall behav-
iour of T-stubs have already been proposed in the literature (7, 9, 10). Essentially, they use
the same basic prying mechanism, which is also the model implemented in Eurocode 3 (6)
(figure 1b). The model is two-dimensional, i.e. the three-dimensional effects are not ac-
counted for. The system is statically indeterminate to the first degree. It is loaded by applying
a vertical force F/2 to the support (A), which is the critical section at the flange-to-web con-
nection. F is the applied force per bolt row. Only one quarter-model is taken into account due
to symmetry considerations. The contact points at the tip of the flange are modelled with a
pinned support and reproduce the effect of prying forces, Q. The T-stub flange behaves as a
rectangular cross-section of width beff and depth equal to tf. Such width, beff, represents the
flange plate width tributary to a bolt row that contributes to load transmission at pure plastic
conditions. This width accounts for all possible yield line mechanisms of the T-stub flange (1,
6) and cannot exceed the actual flange width, b. Despite these major simplifications, the
Jaspart approximates the nonlinear T-stub behaviour with a bilinear response (7, 8). The
characteristics of this bilinear behaviour are summarized as follows:
(i) The initial elastic region has a slope ke.o that is evaluated as in Eurocode 3 (6).
(ii) The swivel point in the bilinear relationship represents the full development of the yield
lines and the correponding force is FRd.0. This plastic resistance is also computed
according to the Eurocode 3 procedures (6).
(iii) In the plastic region, above the swivel point, the effects of material strain hardening are
dominant. The slope of this second linear region is then given by (7, 8):
k p − l .0 = Eh E ke.0 (1)
whereby E: Young modulus of the flange material and Eh: strain hardening modulus.
(iv) The point of maximum force, Fu.0, is determined by formally equivalent expressions to
FRd.0, by replacing the plastic conditions (index Rd) with ultimate conditions (index u).
This means that these expressions are based on the same geometric characteristics but
the plastic moment of the flange, Mf.Rd, is replaced with:
Mf .u = 0.25tf 2fu.f beff (2)
(fu.f: ultimate stress of the flange material) and the bolt strength, Bu, is:
Bu = fu.b As (3)
being fu.b: ultimate stress of the flange material and As: tensile stress area of a bolt. The
deformation capacity is readily determined by intersection of the plastic region, with
slope kp-l.0, with the maximum resistance, Fu.0, i.e.:
∆u.0 = FRd .0 ke.0 + ( Fu.0 − FRd .0 ) k p − l .0 (4)
b b b
Q Q Q Q
B B BRd BRd
BRd BRd
(=F1.Rd.0/2+Q)
n m m n n m m n n m m n
(a) Flange plastic mechanism (b) Combined bolt/flange (c) Bolt mechanism (type-3;
(type-1). mechanism (type-2; ξ ≤ 1). ξ ≤ 1).
Elastic, Kb
δb
δb.1 δb.2 δb.fract
Swanson developed a prying model that includes: (i) nonlinear material properties, (ii) a vari-
able bolt stiffness that captures the changing behaviour of the bolts as a function of the loads
they are subjected to, (iii) partially plastic hinges in the flange and (iv) second order mem-
brane behaviour of thin flanges (10). The bolt behaviour is incorporated by means of an ex-
tensional spring located at the inside edge of the bolt shank. This spring is characterized by
a piecewise linear force-deformation, B-δb, response (figure 3). The bolt elastic stiffness, Kb,
is given by:
Eb
Kb = (5)
Ls Ab + Ltg As
being Eb: Young modulus of the bolt material, Ls: bolt shank length, Ab: nominal area of the
bolt shank and Ltg: bolt threaded length included in the grip. The bolt fracture deformation,
δb.fract, is computed as follows:
0.90Bu Ls ⎛ 2 ⎞
δ b.fract = + ε u.b ⎜ Ltg + ⎟ (6)
Ab Eb ⎝ nth ⎠
whereby εu.b: ultimate bolt strain and nth: number of threads per unit length of the bolt. The
flange mechanistic model assumes an elastic-yielding-plastic (with strain hardening) constitu-
tive relationship for the steel. Plastic hinges will develop at the flange-to-web connection and
at the bolt axis. Swanson derived the stiffness coefficients and corresponding prying gradi-
ents, qij,k = ∂Q/∂∆, by using the direct stiffness method. Both parameters are used in an in-
cremental solution technique. First, the initial stiffness and prying gradient are determined.
Next, several checks are made to determine which limit is reached first (bolt force or flange
PROPOSED MODEL
The models described above afford some basis for the development of a model for the
evaluation of the deformation capacity and the load-carrying capacity of single T-stub con-
nections. The mechanical model is similar to that depicted in figure 1b and is represented in
figure 4. A model that uses geometrical and mechanical properties consistent with the Euro-
code 3 (6) prying model is desired so as to make implementation easier.
Fracture conditions
The two possible ultimate (fracture) conditions are: (i) fracture of the bolt and (ii) cracking of
material of the flange near the web, i.e. at section (A) (figure 4). Normally, M(C) ≤ M(A) (section
(C) is the flange section at the bolt line, figure 4). In the context of a two-dimensional model,
whereby the flange is modelled as a rectangular cross-section, the latter condition may be
too severe.
0.8r
r
F
2
(C)
(A*)
(A) B-δb
m n
⎧⎪r → HR T-stubs
m = d − 0.8 ⎨
⎪⎩ 2aw → WP T-stubs
n = min (1.25m; e )
The bolt elongation response is based on the Swanson’s proposals (figure 3). For the com-
putation of δb.fract, Eq. (6), however, the parameter nth is disregarded.
With reference to this equation, it was developed for short-threaded bolts (10). If a full-
threaded bolt is considered instead, this expression seems to overestimate the bolt fracture
deformation. Therefore, in this case, the evaluation of δb.fract by application of the above ex-
pression should be cautious.
The flange material constitutive law is modelled by means of a piecewise stress-strain rela-
tionship that accounts for the strain hardening effects. This law is a true stress-logarithmic
strain relationship, i.e. it is defined in natural coordinates in order to capture the actual mate-
rial behaviour. Piluso et al., in fact, adopt the same approach in (9) since the prediction of the
plastic deformation capacity of compact sections is more accurate if natural stress-strain co-
ordinates are used. From a design point of view, the constitutive law can be idealized by
means of multilinear models that reproduce well the material strain hardening and fracture
ranges. Gioncu and Mazzolani (11) present alternative models for the stress-strain curve of
structural steels.
The above model is not proposed for a hand calculation. Instead, a numerical finite element
(FE) method is used to determine the structural response. Consequently, the piecewise con-
stitutive law may contain several branches. Figure 5 shows the stress-strain relationship im-
plemented for three different steel grades that will be referred to later in the text (4, 5). The
FE code Lusas (12) implements a beam element that belongs to the Kirchoff beams group
(with quadrilateral cross-section). Shear deformations are excluded in this element formula-
tion.
900
S690
True stress (MPa)
750
S355 (1)
600
S355 (2)
450
300
150
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Logarithmic strain
Figure 5. True stress-logarithmic strain characteristics for different steel grades.
First, the proposed model is analysed and validated in the elastic domain by using the speci-
mens from the authors’ database (4, 5, 13) (table 1). The initial stiffness of a T-stub connec-
tion is evaluated and compared with the actual predictions, which correspond to experimental
or (three-dimensional) numerical values. It was found out that the model yielded a very stiff
response in the elastic domain if the actual Young modulus of the flange was employed (13).
It was also verified that in this domain, disregarding the shear deformability of the flange
would introduce an error of 15% in the predictions of the initial stiffness. Therefore, a reduc-
tion of the Young modulus, Ered, was proposed in order to make the necessary corrections
(13). Such value was defined as follows:
2 2
2⎛m⎞ ⎡ 3 ⎤ E ⎛m⎞ ⎡ 3 ⎤
Ered = 0.5E × ⎜ ⎟ ⎢ 1 + − 1⎥ = ⎜ ⎟ ⎢ 1 + − 1⎥ (7)
( m tf ) ⎥⎦ 3 ⎝ tf ⎠ ⎢⎣ ( m tf ) ⎥⎦
2 2
3 ⎝ tf ⎠ ⎢
⎣
so that the shear deformability was taken into account (13, 14). This reduction does not have
much influence at ultimate conditions. In fact, the effect of shear on the moment resistance of
the flange is apparently beneficial (15).
Table 1. Geometrical and mechanical characteristics of the specimens (Fl: flange; Bt: bolt;
Bts: Bolt stripping; Flange steel grade, fy.f: S355 (1), regular values; S355 (2), underlined val-
ues; S690: values in bold).
Test ID Fracture ele- Geometrical characteristics Mechanical characteristics
ment
Ac- Model beff tf m n φ fy.f εp.u.f fu.b δb.fract Kb
tual pred. (mm) (mm) (mm) (mm) (mm) (MPa) (MPa) (mm) (N/mm)
T1 Bt Bt 40.0 10.7 29.45 30.00 12 431.0 0.284 974.0 0.97 6.92×105
P1 Bt Bt 40.0 10.7 34.45 25.00 12 431.0 0.284 974.0 0.97 6.92×105
P2 Bt Bt 40.0 10.7 24.45 30.56 12 431.0 0.284 974.0 0.97 6.92×105
P3 Bt Fl (A*) 35.0 10.7 29.45 30.00 12 431.0 0.284 974.0 0.97 6.92×105
P4 Bt Bt 52.5 10.7 29.45 30.00 12 431.0 0.284 974.0 0.97 6.92×105
P5 Bt Bt 60.0 10.7 29.45 30.00 12 431.0 0.284 974.0 0.97 6.92×105
P6 Bt Bt 35.0 10.7 29.45 30.00 12 431.0 0.284 974.0 0.97 6.92×105
P7 Bt Bt 45.0 10.7 29.45 30.00 12 431.0 0.284 974.0 0.97 6.92×105
P8 Bt Bt 40.0 11.0 27.10 33.88 12 431.0 0.284 974.0 0.98 6.78×105
P9 Bt Bt 40.0 14.0 28.75 35.94 12 431.0 0.284 974.0 1.36 5.45×105
P10 Fl Fl (A*) 40.0 7.0 30.75 30.00 12 431.0 0.284 974.0 0.75 9.37×105
P11 Bt Bt 40.0 14.0 28.75 30.00 12 431.0 0.284 974.0 1.20 5.57×105
P12 Fl Fl (A*) 40.0 10.7 29.45 30.00 16 431.0 0.284 974.0 1.01 1.19×106
P13 Bt Fl (A*) 40.0 10.7 29.45 30.00 12 355.0 0.284 974.0 0.97 6.92×105
P14 Fl Fl (A*) 40.0 10.7 29.45 30.00 12 275.0 0.284 974.0 0.97 6.92×105
P15 Fl Fl (A*) 40.0 10.7 24.45 30.56 16 431.0 0.284 974.0 1.01 1.19×106
P16 Bt Bt 70.0 10.7 29.45 30.00 12 431.0 0.284 974.0 0.97 6.52×105
P17 Bt Bt; fl (A*) 70.0 10.7 29.45 30.00 16 431.0 0.284 974.0 1.09 1.11×106
P18 Fl Fl (A*) 70.0 10.7 29.45 30.00 20 431.0 0.284 974.0 1.09 1.73×106
P19 Bt Bt; fl (A*) 70.0 10.7 29.45 30.00 16 431.0 0.284 974.0 1.09 1.11×106
P20 Bt Bt 70.0 14.0 28.75 30.00 16 431.0 0.284 974.0 1.37 9.54×105
P21 Bt Fl (2) 92.5 10.7 29.45 30.00 20 431.0 0.284 974.0 1.37 1.49×106
P22 Bt Bt 70.0 15.0 32.34 30.00 20 431.0 0.284 974.0 1.52 1.40×106
P23 Bt Bt 70.0 18.9 31.59 30.00 20 431.0 0.284 974.0 1.98 1.14×106
Weld_T1(i) Bt Fl (A*) 40.0 10.7 37.43 30.00 12 431.0 0.284 974.0 0.97 6.92×105
Weld_T1(ii) Bt Fl (A*) 40.0 10.7 33.42 30.00 12 431.0 0.284 974.0 0.97 6.92×105
Weld_T1(iii) Bt Bt 40.0 10.7 30.14 30.00 12 431.0 0.284 974.0 0.97 6.92×105
From the comparison of the model predictions for the above specimens and the actual initial
stiffness values, it was observed that the stiffness prediction was sufficiently accurate leading
to an average overestimation equal to 18% and to a coefficient of variation equal to 0.16 in
the case of HR T-stubs. For WP T-stubs, the results are further improved: estimation error of
1% and coefficient of variation of 0.18 (13).
Analysis of the model in the elastoplastic domain and comparison with the other
models
The proposed model for the F-∆ curve is compared with the actual curve in figure 6 for some
representative specimens. The comparison with experimental and numerical (three-
dimensional model) evidence (13) shows a good agreement of results in terms of stiffness
and resistance (approximate errors: underestimation of Fu.0 in 15%, for those specimens
whose cracking of the flange determines collapse and overestimation of circa 10% if collapse
is governed by bolt fracture). In terms of ductility, the model predicts the deformation capacity
accurately if the cracking of the flange is critical (see figures 6c,6d,6e). (For specimen WT1
(figure 6d), however, the bolt also fractured in the tests.) The exceptions in this case are the
specimens made up of steel grade S690. Apparently, if the cracking condition is imposed as
the attainment of εu at section (A*), ∆u.0 is clearly underestimated (figure 6h). For those
specimens whose plastic collapse is of type-2 and eventually the bolt fractures (specimen
P16, for instance – figure 6b), the prediction of ∆u.0 is good (average overestimation of 9%
with a coefficient of variation of 11%). At last, for those specimens exhibiting plastic mecha-
nism type-1 and fracture of the bolt at ultimate conditions, the agreement between actual and
predicted values is not that satisfactory (predicted values nearly twice the actual values) (fig-
ure 6a).
Figure 6g illustrates the above mentioned setback with the bolt model applied to full-threaded
bolts. In specimen WT57_M12 the flange plates are fastened by means of two full-threaded
bolts. At collapse, the bolt model estimates a fracture elongation of 4 mm (table 1), which
seems rather high. Since bolt governs fracture of this specimen, the post-limit behaviour pro-
ceeds until this deformation of 4 mm is attained, leading to an overall deformation 8.1 mm
and ultimate resistance of 174.5 kN, corresponding to 1.43 times the maximum resistance
from the tests.
Load, F (kN)
100 100
80 80
60 Actual response 60 Actual response
Simplified response (Beam model) Simplified response (Beam model)
40 Jaspart approximation (actual Eh) 40 Jaspart approximation (actual Eh)
Jaspart approximation (nominal Eh) Jaspart approximation (nominal Eh)
20 Piluso et al. approximation
20 Piluso et al. approximation
0 0
0 2 4 6 8 10 12 14 16 18 20 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Deformation, ∆ (mm) Deformation, ∆ (mm)
(a) Specimen T1. (b) Specimen P16.
280 105
240 90
Load, F (kN)
Load, F (kN)
200 75
160 60
120 45 Actual response
Actual response Simplified response (Beam model)
80 Simplified response (Beam model) 30
Jaspart approximation (actual Eh) Jaspart approximation (actual Eh)
40 Jaspart approximation (nominal Eh) 15
Piluso et al. approximation Jaspart approximation (nominal Eh)
0 0
0 3 6 9 12 15 18 21 24 27 30 0 2 4 6 8 10 12 14 16 18 20
Deformation, ∆ (mm) Deformation, ∆ (mm)
(c) Specimen P18. (d) Specimen WT1.
160 180
140
150
120
Load, F (kN)
Load, F (kN)
120
100
80 90
Actual response Actual response
60 Simplified response (Beam model)
Simplified response (Beam model) 60
40 Jaspart approximation (actual Eh)
Jaspart approximation (actual Eh) Jaspart approximation (nominal Eh)
30
20 Jaspart approximation (nominal Eh) Piluso et al. approximation
0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Deformation, ∆ (mm) Deformation, ∆ (mm)
(e) Specimen WT7_M16. (f) Specimen WT7_M20.
180 280
150 240
Load, F (kN)
Load, F (kN)
200
120
160
90
Actual response 120 Actual response
60
Simplified response (Beam model) 80 Simplified response (Beam model)
30 40
Jaspart approximation (actual Eh) Jaspart approximation (actual Eh)
0 0
0 1 2 3 4 5 6 7 8 9 10 0 2 4 6 8 10 12 14 16 18 20 22 24
Deformation, ∆ (mm) Deformation, ∆ (mm)
(g) Specimen WT57_M12. (h) Specimen WT57_M20.
Jaspart has shown that a significant increase in the resistance of single T-stubs that fail ac-
cording to a plastic mechanism type-1 (figure 2a) can be expected due to the influence of the
bolt action on a finite contact area (7). This effect is also taken into account in the proposed
methodology, as a modification of the original model. The bolt is then modelled as a spring
assembly in parallel, as indicated in figure 7. The length of this assembly is the bolt diameter.
The behaviour of this spring assembly is the same as the original single spring, i.e. the spring
stiffness and force values are divided by the number of springs in the assembly.
The application of this modified model provides significant enhancement of results in terms of
resistance, particularly for the evaluation of Fu.0 and ∆u.0, as well. Figure 8 compares the two
modelling approaches in terms of overall behaviour for some of the former specimens.
0.8r
r
φ
F
2
(C)
(A*)
(A) B-δb
m n
Figure 7. T-stub model accounting for the bolt action on a finite contact area.
CONCLUSIONS
The proposed beam model yields an accurate prediction of the F-∆ response of bolted T-stub
connections, despite the simplifications inherent to a two-dimensional modelling of the be-
haviour. In the model, the dominant effects are the strain hardening of the flange and the bolt
elongation behaviour. The behaviour predicted by this model is rather good in terms of resis-
tance. With respect to ductility, it reflects an overestimation of test results that is within an
acceptable error (13).
Load, F (kN)
Load, F (kN)
100 75
80 60
Actual response Actual response
60 45
Simplified response (simple Simplified response (simple
40 30 beam model)
beam model)
Simplified response accounting 15 Simplified response accounting
20 for the bolt action for the bolt action
0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Deformation, ∆ (mm) Deformation, ∆ (mm)
Load, F (kN)
Load, F (kN)
200
100
Actual response 160
80 Actual response
120
60 Simplified response (simple
Simplified response (simple
beam model) 80
40 Simplified response accounting beam model)
Simplified response accounting
20 for the bolt action 40
for the bolt action
0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 22 24
Deformation, ∆ (mm) Deformation, ∆ (mm)
(c) Specimen WT7_M16. (d) Specimen WT57_M20.
Figure 8. Comparison of the force-deformation response for the simple proposed model and
with the inclusion of the modelling of the bolt.
The applicability of the model was well demonstrated within the range of examples shown in
this paper. The modification for inclusion of the bolt action provides an enhancement of re-
sults but introduces an additional complexity. From a design point of view, the methodology
should be further simplified so that it can be used in an expedite way, as the Jaspart’s simple
proposal. This can be achieved by modelling plasticity phenomena in the flange by means of
rotational springs at the critical sections that capture the overall behaviour. Current work on
this particular subject is being carried out by the authors.
ACKNOWLEDGEMENTS
Financial support from the Portuguese Ministry of Science and Higher Education (Ministério
da Ciência e Ensino Superior) under contract grants from PRODEP and FCT (Grant
SFRH/BD/5125/2001) for Ana M. Girão Coelho is gratefully acknowledged.
REFERENCES
(1) Zoetemeijer P, (1974). A design method for the tension side of statically loaded bolted
beam-to-column connections. Heron; 20(1):1-59.
(2) Packer JA, Morris LJ, (1977). A limit state design method for the tension region of
bolted beam-to-column connections. The Structural Engineer; 55(10):446-458.
(3) Yee YL, Melchers RE, (1986). Moment-rotation curves for bolted connections. Journal
ABSTRACT
In the current paper the experimental results related to welded aluminium alloy
T-stub components under quasi-static monotonic and cyclic loading are
provided. The experimental study refers to 26 different welded specimens. In
particular, 2 boundary conditions, 4 T-stub geometry (varying in-plane
dimensions, plate thickness, number and location of bolts), 3 aluminium alloys
and 3 types of bolt have been considered. The obtained results allow the
influence of the above parameters on the observed failure plastic mechanism
as well as on the deformation and dissipative capacity of the examined joint
type to be pointed out.
INTRODUCTION
The correct prediction of joint structural response is an important issue in metal structures (1).
For aluminium structures, current standards provide rules to estimate the influence of joints on
the global structural behaviour through the joint classification (2), but the application of such
methods is related to the correct evaluation of the main joint mechanical properties. On the other
hand, it is recognized that the behaviour of connections is very complex, because it includes
many local effects and the system is highly indeterminate, giving rise to complicate stress
redistribution among the basic components. Therefore, for a reliable understanding of
connection behaviour and setting up of relevant calculation methods the direct experimentation
is required. While several laboratory experiments have been carried out for steel, allowing some
consistent calculation procedures to be set up (EC3 – Part 1.8), only few are concerned with
aluminium alloys (3). On the basis of the above considerations, a large research project dealing
with aluminium joints has been carried out at the University of Naples Federico II, sponsored by
the Italian Ministry of Education, University and Research (MIUR). It was mainly concerned with
the behaviour of bolted T-stub joint components, which has been deeply examined under
analytical, numerical and experimental points of view (4).
T-stub is one of the more important joint components. It may be regarded both as stand-
alone connection (5) and as a part of several, more complex bolted joint configurations (6). In
previous papers, the authors have already proposed an analytical method to determine both
the collapse mechanism type and the corresponding load bearing capacity of aluminium T-
stubs based on the existing formulations adopted into EC3-Part 1.8 (7). It is important to
remind that for steel T-stubs, three possible failure modes are recognized: type-1
mechanism, with complete yielding of the flange through the onset of four plastic hinges, two
of which are located at bolt axis location; type-2 mechanism, when the bolt failure takes
place together with the yielding of the flange at the sections corresponding to flange-to-web
connection; type-3 mechanism, due to bolt failure with the overall up-lift of the flange. Due to
The above method has been firstly calibrated on the basis of numerical FEM analyses only
(8). Then, monotonic loading tests have been carried out for several aluminium T-stub
specimens (9), allowing the above analytical model to be checked and re-calibrated (10).
Finally, in the present paper, experimental results of the same specimens previously tested
under monotonic loading are analysed in comparison with results obtained under cyclic
loading (11), so to evaluate the dissipative capacity of the considered specimens as well as
the relevant failure mechanisms under repeated deformations.
The experimental study, which has been carried out at the Laboratory of the Department of
Structural Analysis and Design at the University of Naples Federico II, concerns the
behaviour of isolated aluminium T-stubs having different geometrical configurations and
material properties of the basic components. Both monotonic and cyclic loading tests have
been carried out. The test specimens were fabricated starting from T-shaped profiles, which
were previously pre-assembled by means of welding made at flange-to-web connection. The
basic plate elements constituting the isolated T-stub specimens are made of three different
wrought aluminium alloys (namely, AW 6061, AW 6082 and AW 7020), having different
thickness (namely, 10 mm, 12 mm and 12 mm, respectively), which were artificially aged
through the application of the T6 heat treatment. Nominal values of the main mechanical
properties of the above alloys are summarised in Table 1. All welds between plate elements
were previously made using standard automatic MIG welding process and were carried out
by the commercial fabricator ALCOA-Italia of Novara.
By combining available flange and web plate elements, 12 different T-stub configurations
have been obtained (Fig. 1), corresponding to 4 different geometrical configurations (varying
flange in-plane dimensions and thickness) and, for each shape, to several flange-to-bolt
material combinations. In particular, in Figure 1, letters A, B and C refer to the three different
materials used for the T-stub flanges. Besides, it has to be noted that labels 'Type 2' and
'Type 3' have been applied to specimens having the same geometry for allowing the
execution of tests with two different coupling conditions. T-joints were made by a bottom Y
partial penetration weld, connecting the bottom part of the two flange elements to each other
and to the web plate element, and by two complementary top fillet welds (5 mm throat
section), connecting the upper part of the flanges to the web. This welding procedure has
been selected in order to reduce the out-of-straightness of flanges. Only for specimen Type-
It is worth noting that normal engineering procedures were adopted during fabrication and
assembly of specimens. In particular, in all cases, according to current practice, bolts were
preloaded to a torque corresponding to an initial tensile stress in the bolt equal to the 80% of
the bolt nominal yield resistance, the latter being evaluated according to EC9. Furthermore,
no special effort was taken to get perfect fit-up of mill cutting during specimen preparation
and to centralize the bolts in the bolt holes during specimen assemblage.
135
135
135
22 36 22
35 30 35 30 35 30
80
12
12
Ø 11 mm 142 142 142
135
135
135
22 36 22
80
30 20 30 20 30 20
AW 6061-T6 AW 6082-T6 AW 7020-T6
10
12
12
TYPE 4 TYPE 5
10 10
122
128
28 28 10 56 AW 6061-T6 AW 6061-T6
24 35 10 59
130
124
35
60
35 24
70
AW 6061-T6 28 28 AW 6061-T6
120
35
10
10
60
Ø 11 mm
122 128
Ø 11 mm
The performed experimental tests are related to two different types of T-stub assemblage
(Fig. 2). In particular, in some cases, two isolated aluminium T-stub specimens are
assembled (coupled T-stub arrangement), while, in some other cases, the isolated T-stub
specimen is joined to a purposely de-signed steel support having negligible deformability
(single T-stub arrangement). The former configuration is typical of end-plate beam splice
joints, where the connected elements have the same flexural stiff-ness. Conversely, the latter
configuration is representative of joints where the connected plate elements are different
from each other, such as in the case of beam-to-column joints. Furthermore, in order to
investigate the influence of bolts on the stiffness as well as the failure mechanism of
aluminium T-stubs, different locations, number and types of bolts have been considered,
taking into account the possibility to use both aluminium bolts (type AA 7075) and steel bolts
(grade 4.8 and 10.9). In all the cases, the bolt diameter is 10 mm while the hole diameter is
T-stub flange
T-stub web
LVDT 4 LVDT 2
LVDT 6
a) b) c)
Tests on selected specimens were performed by a Material Test System (MTS 810) 500 kN
universal testing machine, under displacement control, using displacement rate equal to
0.005 mm/s in the first elastic loading and 0.05 mm/s in the remaining part of the loading
process. For cyclic tests, the Complete Testing Procedure provided by ECCS
Recommendations (1986) has been used. For this purpose, the conventional limiting elastic
displacement used for the definition of the loading history of each sample has been
established on the basis of the results obtained from monotonic tests (9).
Tie forces were applied to the webs of the connected T-stub elements, which were tightened
by the jaws of the testing machine for a length of 50 mm, having the axial force applied
collinear with the midpoint of the connection. During the test, the applied load was
automatically measured by the testing machine, while displacements were measured by
using displacement transducers having accuracy of 0.01 mm. In particular, according to
Figure 2a, on each side of the joint, a 500 LVDT was used to measure the flange-to-flange
relative displacement at the centreline of the connection (LVDTs 5 and 6) while two 200
LVDTs were used to measure the flange-to-flange relative displacements at the bolt locations
(LVDTs 1, 2, 3 and 4). Both plate material and bolt tensile tests have been carried out to
characterize the mechanical behaviour (4, 9). For each alloy, average values of material
properties are reported in Table 2, while obtained results for bolts are given in Table 3. In
such a table, Steel 10.9 bolts are not considered, because they are never involved in
collapse mechanism of the specimen.
TEST RESULTS
General
In the following, for both monotonic and cyclic loading tests, the obtained results are provided
in terms of applied force (F) versus relative displacement (∆) curves. The relative
displacement (∆) between T-stub elements has been evaluated at the connection centreline,
as average of the two measuring points. In particular, in Figures 3-4, results are provided for
the single T-stub arrangement, while in Figure 5, the results related to the coupled T-stub
arrangement are presented. In addition, in Table 4 for each test specimen, the main
measured and determined characteristics of F-∆ curves are summarized, together with the
revealed types of failure mechanism. It is worth mentioning that for each T-stub type, the
sequential numbering for samples refers to the specimen identification stated in De Matteis
et al. (4), where, the actual geometrical properties together with the measured main
imperfection parameters (flange out-of-straightness and uplift) for each specimen are
supplied. In the above table, the peak strength (Fmax), the corresponding displacement level
(∆max) and the ultimate displacement (∆ult), the latter evaluated either as the value
corresponding to a sudden drop off of the bearing capacity of the specimen or as the value at
the end of the test in case the relevant failure was gradual and not clearly identified, are
given. Also, for monotonic tests, the conventional elastic strength (Fel), evaluated according
to ECCS Recommendations (1986), and the elastic displacement (∆el) associated to Fel are
specified.
The examination of Table 4 reveals that tested specimens gave rise to all the possible failure
mechanisms, showing important differences in terms of strength and deformation capacity
due to T-stub geometry, bolt type and flange material. In particular, as it could be expected,
the T-stub geometry has an important influence on the failure mechanism: T-stub Type 1 and
Type 2-3 present weak-flange failure mechanism, while T-stub Type 4 and T-stub Type 5 are
characterised by a weak-bolt failure mechanism. Furthermore, as a rule, the specimens
endowed with steel grade 10.9 bolts provide an ultimate strength remarkably higher than the
other cases. On other hand, specimens with bolt type AA 7075 and steel grade 4.8 present
similar strength, even though the former are generally characterized by a lower value of the
ultimate deformation capacity. On the other hand, the influence of the flange material on the
overall response of the examined structural component has to be related to the relevant
collapse mode, it appearing more important for specimens exhibiting a weak-flange failure
mechanism (failure mechanism type 1 and type 2a).
A deeper examination of results evidences that there is a strict dependence between the
failure mode and the type of applied bolt. In fact, all specimen type 1 and 2-3 equipped with
steel bolts grade 10.9 exhibited a type 1 failure mechanism, with limited deformation
1C1 1C4 AW-7020 Steel 4.8 2a 88.4 65 0.38 2.11 7.34 2a 81.9 3.4 4.0
1C2 1C5 AW-7020 AA 7075 2c 91.6 42 0.15 0.96 0.96 2c 93.5 1.2 1.2
1C3 1C6 AW-7020 Steel 10.9 1 145.2 64 0.13 2.46 2.77 1 132.2 1.6 2.6
2A2 3A2 AW-6061 AA 7075 2c 87.8 43 0.15 1.17 1.17 2c 72.0 1.1 1.1
2A3 3A3 AW-6061 Steel 10.9 1 118.7 67 0.10 1.94 2.38 1 94.1 1.1 1.9
2B2 3B2 AW-6082 AA 7075 2c 105.2 52 0.08 1.08 1.08 2c 90.6 1.2 1.2
2B3 3B3 AW-6082 Steel 10.9 1 154.2 78 0.09 1.43 1.66 1 117.4 1.1 3.1
2C2 3C2 AW-7020 AA 7075 2c 102.9 49 0.05 0.85 0.86 2c 82.0 1.5 1.5
2C3 3C3 AW-7020 Steel 10.9 1 153.2 78 0.14 2.08 2.24 1 128.2 1.2 2.5
4-1 4-4 AW-6061 Steel 4.8 3 44.9 33 0.49 1.47 2.93 3 47.1 1.6 4.0
4-2 4-5 AW-6061 AA 7075 3 50.5 23 0.09 0.63 0.63 3 56.4 1.4 1.4
4-3 4-6 AW-6061 Steel 10.9 2b 110.7 79 0.69 3.90 4.54 2b 103.9 3.5 5.4
5-1 5-3 AW-6061 Steel 4.8 2c 52.9 38 0.04 1.97 7.01 2c 46.8 2.4 5.0
5-2 5-4 AW-6061 AA 7075 2c 52.3 29 0.03 0.90 0.91 2c 66.3 1.8 1.8
2A4 + 3A4 + AW-6061 AA 7075 2c 88.2 72 0.21 1.56 1.56 2c 79.4 1.5 1.5
2A6 3A6
Coupled T-stub assemblage
2A1 + 3A1 + AW-6061 Steel 10.9 1 119.4 54 0.15 3.32 3.67 1 98.8 2.1 3.2
2A5 3A5
2B4 + 3B1 + AW-6082 AA 7075 2c 92 50 0.24 2.22 2.24 2c 85.7 2.2 2.2
2B5 3B5
2B1 + 3B4 + AW-6082 Steel 10.9 1 147.8 88 0.24 2.44 2.57 1 102.4 1.6 2.9
2B6 3B6
2C4 + 3C4 + AW-7020 AA 7075 2c 89.9 44 0.12 1.56 1.56 2c 95.0 1.8 1.8
2C5 3C6
2C1 + 3C1 + AW-7020 Steel 10.9 1 137 66 0.18 2.72 2.97 1 114.9 1.5 2.8
2C6 3C5
Obviously, the above circumstances has an effect also on the dissipative capacity of the
joint; in particular, specimens equipped with steel bolts 4.8 show a larger residual plastic
deformation and amount of energy dissipation per cycle with respect to the same specimens
equipped with aluminium bolts 7075. It is also important to observe that specimens endowed
The difference between aluminium AA 7075 and steel bolt class 4.8 is also evident by
analyzing curves related to T-stub type 4 and type 5, the collapse mechanism being related
essentially to the bolt failure (Fig. 6d). On the other hand, it is evident that the response of
specimens with steel bolts is better than the one of specimens with aluminium bolts, the latter
exhibiting significant strain hardening but a limited ductility and dissipative capacity. Finally, it
is important to observe that the combination of T-stub type 4 with steel bolt class 10.9 gave
rise to an unexpected failure mode, which was characterized by significant yielding of the
flanges with yield lines developed in a similar way as in the non-circular pattern provided by
EC3-Part 1.8 but with final rupture of bolts under tension.
Figure 3. F-∆ curves for T-stub Type 1 and 2 (single T-stub assemblage).
In case of T-stub type 2-3, it is also possible to investigate the effect of different coupling
conditions (Fig. 5). Obtained results show that the qualitative behaviour is very similar in
case of single and coupled T-stub assemblage. Really, in some cases, a small reduction of
strength was noticed for the latter configuration, which could be ascribed to the following two
reasons: (1) due to the increment of global deformation, the flexural effects on the bolts are
more important, especially in case of type 2 collapse mechanism; (2) due to high
imperfections characterising the specimens, a statistical scatter among results concerning
equal specimen has to be expected, with a major probability that lower strength is of concern
for coupled T-stub assemblage.
In order to highlight the influence of flange material on the overall response of the joint, in
Figure 7, with reference to T-stub type 1 and 2, the relevant monotonic F-∆ curves are
grouped for each type of bolt. Obviously, such an influence is related to the type of failure
mechanism, and it appears more clearly for failure mechanism type-1, when the flange itself
180 F [kN] Type 1 (bolt S 4.8) 180 F [kN] Type 1 (bolt AA7075) 180 F [kN] Type 1 (bolt S 10.9)
180 F [kN] Type 2-3 (bolt AA7075) 180 F [kN] Type 2-3 (bolt S 10.9) 180 F [kN] Type 2-3 (bolt S 10.9)
coupled
150 A (AW 6061) 150 150
C
B (AW 6082)
120 C (AW 7020) 120 120
B A B C A
90 90 90
A (AW 6061) A (AW 6061)
60 60 B (AW 6082) 60 B (AW 6082)
C (AW 7020) C (AW 7020)
30 A 30 30
C B
∆ [mm] ∆ [mm] ∆ [mm]
0 0 0
0 0.5 1 1.5 2 0 1 2 3 4 0 1 2 3 4
Figures 3, 4 and 5 show also the comparison between monotonic and cyclic test results.
Preliminarily, it should be observed that for every tested specimen, the failure mode detected
for monotonic loading tests was observed also for cyclic loading tests (see Table 4). This
means that the loading type has not a significant influence on the collapse mode of the
specimen.
From the examination of the above figures, it can be noted that cyclic tests generally gave
rise to a structural performance lower than the corresponding one gained by monotonic
loading tests. This can be ascribed to the mechanical and geometrical damage of the
specimen accumulated during the sequential plastic excursions to which the specimen itself
is subjected. As a result, the ductility of specimens under cyclic loading is reduced with
respect to monotonic tests. This is particularly evident for specimens failing with a type 1 or
type 2a failure mechanism, which involves the plastic collapse of the flange at the weld
location. In such a case, it has been already observed that the collapse of the specimen was
essentially due to crack initiation in the weld. Obviously, the application of cyclic loads
favourites the crack propagation, anticipating the failure of the specimen. On the contrary,
the damaging effect due to low cycle fatigue is limited for specimens collapsing due to bolt
failure (type 2c or type 3 failure mechanism), which are mainly subjected to tensile forces,
exhibiting a not-ductile failure mechanism even for monotonic loads.
Finally, it should be observed that among analysed specimens it seems that there are some
anomalous cases. In fact specimen type 4 (Bolt S 4.8), specimen type 4 (Bolt AA 7075) and
specimen type 5 (Bolt AA 7075) present a performance (strength and ductility) for cyclic
loading higher than for monotonic loading. On the other hand, all these cases refer to
specimens failing due to bolt fracture, where the influence of loading history is not important.
In particular, as it could be expected, it is apparent that the T-stub geometry has an important
influence on the failure mechanism exhibited by the structural component. In fact, among the
analysed specimens, T-stub Type 1 and Type 2-3 presented a weak-flange failure
mechanism, while T-stub Type 4 and T-stub Type 5 were characterised by a weak-bolt
failure mechanism.
Furthermore, it has been pointed out that the specimens endowed with steel grade 10.9 bolts
provide an ultimate strength remarkably higher than the other cases. In such a case, a type 1
failure mechanism, with limited deformation occurring in the bolts and a notable plastic
engagement of flange material, was evidenced. On other hand, specimens with bolt type AA
7075 and steel grade 4.8 present a similar strength, even though the former are generally
On the other hand, the influence of the flange material on the overall response of the
examined structural component has to be related to the relevant collapse mode, it appearing
more important for specimens exhibiting a weak-flange failure mechanism (failure
mechanism type 1 and type 2a).
Eventually, the comparison between monotonic and cyclic test results allows us to conclude
that the loading type has not a significant influence on the collapse mode of the specimen.
Nonetheless, cyclic tests generally gave rise to a structural performance (strength and
ductility) lower than the corresponding one gained by monotonic loading tests, due to low-
cycle fatigue. This is particularly evident for specimens failing with a type 1 or type 2a failure
mechanism, which involves the plastic collapse of the flange at the weld location. On the
contrary, the damaging effect due to low cycle fatigue is limited for specimens collapsing due
to bolt failure (type 2c or type 3 failure mechanism), which exhibited a not-ductile failure
mechanism even for monotonic loading tests.
ACKNOWLEDGEMENT
The current experimental research has been carried out in the framework of the Research
Project “Damage of Connections in Metal and Composite Constructions”, funded by the
Italian Ministry of Education, University and Research (MIUR).
NOTATION
f0.2 conventional elastic strength for aluminium flange and bolt material
fu ultimate strength for aluminium flange and bolt material
fy yielding strength for steel bolt material
εu ultimate (uniform) deformation for aluminium flange material
F applied T-stub tensile force
Fel T-stub conventional elastic strength derived from monotonic tests
Fmax T-stub peak strength measured from tests
∆ relative displacement between T-stub elements at the connection centerline
∆el T-stub elastic displacement associated to Fel
∆max T-stub displacement level corresponding to Fmax
∆ult T-stub ultimate displacement
REFERENCES
(1) Mazzolani, F.M. 1995. Aluminium Alloy Structures, E & FN SPON, London.
(2) Mazzolani, F.M., De Matteis, G., Mandara, A., 1996. Classification system for
aluminium alloy connections, in Proc. of the IABSE Int. Col. on Semi-Rigid Structural
Connections, Istanbul, Vol. 75, 83-94.
(3) Matusiak, M. 1999. Strength and Ductility of Welded Structures in Aluminium Alloys,
ABSTRACT
The successful modeling of partially-restrained beam-to-column connections
requires less accurate and less complicated constitutive laws for the separate
elements when using mechanical modeling approaches, when compared to
traditional finite element approaches. Nevertheless, the adopted constitutive
law needs to conform to certain accuracy limits, lest the simulated behavior be
unsatisfactory. The paper presents some results from ongoing experimental
tests aimed at obtaining simplified constitutive laws for T-Stub connections
suitable for use within a previously published mechanical model. Also, the
paper investigates the behavioral differences between hot-rolled and built-up
T-Stubs. The experimental data is presented and discussed, and some
numerical comparisons are also discussed.
INTRODUCTION
The mechanical modeling approach for bare-steel and steel-concrete composite beam-to-
column connections allows for the utilization of simple constitutive laws for the single
components of the joint, by concentrating the deformation and stiffness response into a finite
set of non-linear springs, which need to simulate all significant contributions to the behavior
of the connection (1, 2). If for a bare-steel connection it would be theoretically possible to use
one single properly defined non-linear rotational spring to account for the monotonic and
cyclic behavior of a beam-to-column joint, sacrificing the completeness of the output results
for the ease of modeling, in the case of a steel-concrete composite connection such
modeling is not anymore a viable approach. In fact, it has been shown (3, 4) that for steel-
concrete composite connections the presence of the reinforced slab forces an interaction
between the two sides of a beam-to-column joint, which are not behaving independently
anymore. For this reason, the one-spring approach is not sufficient anymore for the
simulation of such complex behavior and modeling formulations capable of accounting for
that interaction are needed. One viable approach is the modeling schematization presented
in (2), which takes into account most of the deformability contributions, and which is suitable
for both static and dynamic analyses (Figure 1).
Such modeling approach offers the advantage of a limited number of aimed output results
(as opposed to a classic finite-element modeling, which is sometimes hindered by the sheer
amount of results to manage). Also, the considered modeling approach, provided that it takes
into account most of the deformability and stiffness contributions, is capable of simulating
quite closely the behavior of a beam-to-column connection, without the need for very
complicated constitutive laws for the single components.
5 5
1 2 2 1
1 2 2 1
100 100
50 50
0 0
-0.09 -0.06 -0.03 0.00 0.03 0.06 0.09
-0.09 -0.04 0.00 0.04 0.09
-50 -50
Despite the lower threshold of acceptability for component constitutive laws, typical of
mechanical models, some issues have emerged concerning the simulation of end-plate and
welded connections. It is worthwhile to recall at this point that the approach followed in the
Annexes J to Eurocodes 3 and 4 (5, 6) for the evaluation of force and stiffness for a single
connection component is based on the evaluation of equivalent T-stubs, corresponding from
case to case to the portion of column subjected to the connection forces, or to the steel
connection, etc. It is therefore important within this framework to use an analytical expression
suitable for the simulation of T-stubs, whatever their fabrication process may be.
The rationale for the ongoing experimental campaign herein presented lays in the need to
MODELING ISSUES
In order to be able to take correctly into account the mentioned behavioral differences among
different types of connections, various constitutive laws have been investigated for the
simulation of PR beam-to-column connections.
F F
δ δ
Figures 4a, b. Tri-linear constitutive law without and with stiffness degradation.
The simplest cyclic constitutive law considered was an elastic-plastic curve with constant
hardening, which provides the benefits of a simple definition at the price of a somewhat
rough reproduction of experimental data. The logical refinement of such constitutive law is
represented by a cyclic tri-linear curve, which is capable of more closely reproducing the
experimental behavior of components (Figure 4a). Such constitutive law is used in the
mechanical model described in (2) with satisfactory results. The advantages of the tri-linear
constitutive law are represented by a reasonable ease of definition associated with a
reasonably accurate simulation of the real behavior of the component. The simulation
accuracy tends to deteriorate during cyclic events, well within plastic range, when the
material starts showing damage, like stiffness and strength degradation. In this circumstance,
in fact, the tri-linear curve overestimates the strength values in ultimate conditions, eventually
leading to the prediction of values of stiffness and strength of the connections larger than the
actual ones. On the other hand, the tri-linear curve proved to be accurate for the simulation
of monotonic events, and to be able to provide good results in post-elastic range for
connections characterized by a stable behavior.
Considering that the scope of the simulations encompasses limit analysis of a structure (for
instance, in the framework of a performance-based design approach), it is also important to
accurately assess the behavior of structural components well into plastic range, in order to
be able to predict their failure parameters. To this end, a refined version of the tri-linear law
has been considered, which includes some consideration for stiffness degradation in post-
yielding range (Figure 4b). This constitutive law is based on the definition of the “classic” tri-
In order to test the effectiveness of the proposed constitutive law, and to investigate the
behavioral differences among differently fabricated specimens, a series of component tests
on rolled and built-up T-stubs, as well as on cleated connections has been conducted, and
the results have been compared to the simulations outcome.
The test program involves four different nominal T-stub sizes (original depth of the hot-rolled
HEA section equal to 120, 200, 240, and 300 mm, respectively) obtained in three different
ways: cut from hot-rolled shapes, built-up by welding plates together, and built-up by bolting
angles to a plate (cleated connection). The geometric dimensions of the three differently
fabricated specimens have been chosen in order to obtain easily comparable data. An
example of the typical specimen is shown in Figure 5. In this paper, only some results of the
T-200 and T-240 series will be reported, as the experimental campaign is ongoing.
HEA200
200
150
100
50
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44
The loading frame, shown in Figure 6 with a specimen installed, was made of two steel
plates (top 50 mm thick and bottom 30 mm thick with stiffeners), connected by means of four
high-strength steel 50-mm-diameter threaded rods. The specimens were directly bolted to
the top plate, and the stem was bolted to the loading device, consisting in a two-way
hydraulic jack (435kN in tension, 933 kN in compression, manually operated).
The force and the deflection of the specimens were measured by means of a 400-kN load-
cell and two linear motion position sensors, installed at each side of the T-stub.
The initial tests were conducted monotonically, with unloading cycles to verify the effects of
strain-hardening and eventual losses of strength or stiffness. For each set of specimens,
cyclic tests were also conducted. The loading history follows the spirit of the ECCS 1992 (7)
document; the load cycles are function of a theoretical yielding value, and are shown in
Figure 7. Figures 8 to 10 show specimen T240L before and after testing.
RESULTS
The obtained results can be classified under three different aspects: a) behavioral
differences among same-sized T-stubs fabricated with different methods; b) behavioral
differences among different-sized T-stubs with the same fabrication method; and c)
differences between the cyclic and the monotonic behavior of the specimens.
For what concerns the first aspect of the problem, Figure 11 shows the comparison of the
monotonic skeleton curve for the differently fabricated T-stubs. The elastic stiffness appears
to be the same in all specimens, as expected, whereas the ultimate behavior is remarkably
different: the welded Tee results weaker than the rolled one, due mainly to the smaller
amount of material in the k-zone of the profile. On the other hand, the bolted Tee is stronger
than the hot-rolled one, as a result of the thicker compound web, and of the larger k-zone.
Moreover, the offset of the two angles (8 mm) reduces the distance between the bolt-holes
and the theoretical location of the plastic hinge, therefore resulting in smaller stresses.
Figure 12 shows, on the other hand, a comparison of the monotonic skeleton curves for hot-
rolled T-stubs T-200 and T-240. It is evident that the ultimate strength of the specimens
increases proportionally to the size, as expected; it is nevertheless possible to notice how
there is little or no difference as far as elastic and post-elastic stiffness and yielding points
are concerned. This is mostly due to geometric differences between the two specimens, in
particular to the bolt-hole positioning.
Figures 16 to 18 show the comparison of the experimental results with some of the analytical
simulations outlined above, using either the basic or the modified tri-linear curve.
The tri-linear curve without deterioration seems to be more than suitable for the simulation of
monotonic behavior, and also of cyclic behavior, provided that the response of the
component remains stable. It is evident how in the later stages of the simulation, when
degradation of strength and stiffness are not negligible, the simulated curve deviates quite
noticeably from the measured curve.
On the other hand, the adoption of the modified constitutive law (Figure 18) seems to provide
a closer simulation of the plastic behavior of the investigated components. Both loading and
unloading curves follow the experimental data quite closely, and a visual inspection of the
cyclic curves shows how the total dissipated energy in the process seems to be predicted
quite well by the model. The diagrams deviate once the strength degradation starts evidently
affecting the experimental behavior.
350
F [kN]
300
250
200
150 T200L-M
T200B-M
100 T200S-M
50
δ [mm]
0
0 5 10 15 20 25 30
300
250
200
T240L-M
150
T200L-M
100
50
δ [mm]
0
0 5 10 15 20 25 30
400
F [kN]
300
200
100
-100
-200 T200L-M
-300 T200L-C
-400
δ [mm]
-500
0 2 4 6 8 10 12 14 16 18 20
400
F [kN]
300
200
100
-100
T200B-M
-200
T200B-C
-300
-400
δ [mm]
-500
0 2 4 6 8 10 12 14 16 18 20 22
200
100
-100
T200S-M
-200
T200S-C
δ [mm]
-300
0 2 4 6 8 10 12 14 16 18 20 22 24 26
350
F [kN]
300
250
200
T240L-M EXPERIMENTAL
150
T240L-M NUMERIC
100
50
δ [mm]
0
0 5 10 15 20 25 30
300
F [kN]
200
100
-100
-200
T200L-C TRI-LINEAR WITHOUT DEGRADATION
T200L-C EXPERIMENTAL
-300
δ [mm]
-400
0 2 4 6 8 10 12 14 16 18 20
Figure 17. T-200L hot-rolled Tee cyclic numerical vs. experimental comparison.
200
100
-100
-200
T200L-C EXPERIMENTAL
T200L-C TRI-LINEAR WITH DEGRADATION
-300
δ [mm]
-400
0 2 4 6 8 10 12 14 16 18 20
CONCLUSIONS
Some results from a series of ongoing experimental tests on T-stub components have been
presented and discussed. The interest in the T-stub component resides in the conventional
use that Annexes J to Eurocodes 3 and 4 (5, 6) make of equivalent T-stubs to take into
account deformation contributions to the overall behavior of beam-to-column connections.
Furthermore, it has been noted that the behavior of connections realized by means of built-
up T-stubs results in some cases different from comparable connections that used hot-rolled
T-stubs.
A refined constitutive law has been proposed for the simulation of T-stub connections, which
takes into account degradation of stiffness in ultimate conditions, and should enable accurate
predictions of the ultimate behavior of PR beam-to-column connections under cyclic and
dynamic loading conditions. As soon as data for the remaining tests will be available, it will
be possible to verify the suitability of the new constitutive law for the simulation of the ideal T-
stub component in a steel or steel-concrete beam-to-column connection. Also, yet another
constitutive law will be investigated, namely a tri-linear curve with both strength and stiffness
degradation, in order to better simulate the ultimate behavior of the components.
ACKNOWLEDGEMENTS
This ongoing research project is part of the doctoral work of the first author. Funding has
been obtained through a grant from the Italian Ministry of University and Scientific Research.
REFERENCES
(1) Huber G., Tschemmernegg F., (1998), Modelling of beam-to-column joints, Journal of
Constructional Steel Research, Vol. 45, No. 2, pp. 199-216
(2) Rassati G.A., Leon R.T., Noè S. (2004). Component modeling of PR composite joints
under cyclic and dynamic loading, Journal of Structural Engineering, Vol. 130 (2):343-
351
(3) Amadio C., Benussi F., Noe’ S., (1994), Behaviour of unbraced semi-rigid composite
ABSTRACT
Among finite element (FE) programs, two different categories can be
distinguished depending on the methodology to compute the nodal
displacements : implicit and explicit solvers. This study first briefly describes
and compares both the implicit and explicit approaches. Various issues such
as the solution strategy, computational time as function of model size, and
convergence aspects of non-linear contact analyses are addressed. In
addition, explicit FE simulations are made of two experiments on large-scale,
partially bolted beam-to-column connections. A comparison between the
experimental and numerical hysteresis loops is presented. In addition, the
advantages of the use of the explicit method relative to the implicit approach
are highlighted.
INTRODUCTION
Because of the high costs of experiments, the number of tests carried out is usually limited.
On the other hand, in the last decade, the FE method has developed as a cost-effective and
reliable tool to conduct research beyond the scope of experimental programmes. Numerical
simulations of bolted connections include a range of aspects such as material and geometric
non-linearities and “rigid body movements”. Implicit FE packages as ABAQUS/Standard (2)
are not able to analyse these problems effectively. In contrast, explicit solvers as
ABAQUS/Explicit (3) are well-suited to simulate multiple contact interactions between
independent bodies.
After having carried out numerical simulations on bolted beam splices, numerical research of
bolted connections by van der Vegte et al (4) then focused on the simulation of actual bolted
beam-to-column connections under cyclic loading. In 2002, Miura et al (5) reported the
results of three large-scale tests on partially bolted beam-to-column connections without
Because of the detailed description of the experiments conducted by Miura et al, these tests
were selected to serve as a reference to assess the reliability of FE models analysed with the
explicit solution technique. The present study describes the test specimens, the numerical
models and the comparison between the numerical and experimental results. However, as
an introduction to explicit approach, this study first briefly addresses the most important
features of the implicit and explicit solution techniques.
Bolted end-plate connections have been widely applied in steel structures due to the
simplicity and economy of their fabrication and assembly. Figure 1a shows an example of a
bolted end-plate connection. Because moment-rotation curves of partially restrained
connections are of great importance for designers, in the past, many research programmes
were conducted to generate such data. In general, most of the “older” studies consist of
experiments in combination with an analytical approach. Along with the introduction of
computers, in more recent publications, another tool has emerged to obtain the moment-
rotation behaviour of bolted end-plate connections : the FE method.
Krishnamurthy who can be considered as one of the pioneers in the field of FE simulations of
bolted connections, reported his findings in several publications (6, 7). Because 3D FE
models were computationally expensive, Krishnamurthy and Graddy (6) analysed selected
benchmark connections by 2-D and 3-D models. The correlation between these results was
then used for the prediction of other 3-D models based on the results of corresponding 2-D
models.
Sherbourne and Bahaari (8, 9) also presented a number of publications evaluating the
moment rotation behaviour of bolted end-plate connections generated with the FE
methodology. Although the models were 3D, some simplifications were made. At first,
Sherbourne and Bahaari assumed a continuous connection between the nodes of the bolt
head and nut and the nodes of the contacting plates. As a result, relative motions between
bolt, column flange and end-plate were restrained. A further limitation of Sherbourne and
Bahaari‘s FE models was caused by the use of truss elements to model the bolt shank.
Because the interface between the bolt shank and the hole boundary was neglected, the
model would not be suitable if the bolts would go into bearing action.
As computers became more powerful during the last decade, a steady increase in the size of
the FE models became possible. This enabled the analyses of less simplified and thus, more
realistic 3D models.
As an example may serve the work of Choi and Chung (10), whose FE models, including the
bolts, were completely based on solid elements. Because bolted connections consist of
various components with complex interaction between the various entities, it is understood
Not only enhanced the capacity of the hardware, but also the capabilities of the software
improved. New solution techniques were derived, making numerical analyses much more
efficient in terms of computational time. The introduction of sophisticated contact algorithms
further widened the application of numerical methods. “Contact” in its simplest form can be
described by the use of so-called “gap elements” which impose displacement compatibility
between user-defined pairs of nodes. However, such elements can only be used when
friction can be ignored. In addition, modelling of such elements is a tedious and time-
consuming task. To overcome these problems, commercial FE packages developed more
user-friendly options, such as contact between surfaces and interface elements instead of
the node-to-node contact definition required by gap elements. The improved contact
algorithms also enabled the modelling of shear and friction between the contacting surfaces.
In 1997, Bursi and Jaspart (11, 12) presented their numerical results on isolated bolted tee-
stub connections. Unlike most of the previous studies, the bolt head and the flange plate
were now modelled as individual components and no longer connected through common
nodes, enabling the relative movement between these components.
The same approach was followed by Wheeler et al (13) who carried out numerical
simulations on bolted end-plate connections subjected to pure bending. Figure 1a illustrates
one of the configurations analysed by Wheeler. Prior to the FE simulations of the assembled
model, Wheeler first conducted a series of numerical analyses to determine the most efficient
type of element and corresponding mesh density for each of the components (i.e. end-plate,
bolts, weld beam section). Wheeler recommended the use of eight noded, hybrid elements
with four elements modelled through the thickness of the end plate. The same element type
was employed to model the bolts. To avoid problems associated with rigid body movements,
loading of the connection was carried out in five steps, such as pre-tensioning of the bolts,
applying moment to the tip of the beam section, etc.
(a) Example of bolted end-plate connection (b) FE mesh analysed by Willibald et al (14)
Most of the references so far considered end-plate connections, in which the load transfer
primarily takes place through axial loading of the bolts. FE simulations become more
complicated and expensive in terms of computational time when bolted connections are also
subjected to shear load, simply because of the specification of additional contact surfaces.
An example of the use of shear bolts, including the difficulties encountered in the FE
analyses is described by Swanson et al. After having conducted a series of experiments (15),
Swanson et al (16) then carried on with FE simulations on bolted T-stub connections. The
configuration, illustrated in figure 2, consists of a T-stub section which was attached to a
column flange through tension bolts. Shear bolts were used to connect the beam flange to
the T-stub. The FE mesh, almost exclusively modelled by quadratic, solid elements,
contained approximately 7300 elements, resulting in 53,000 degrees of freedom. Although
the loading of the connection was merely monotonic, the authors reported that a full run on a
Pentium II - 450 Mhz took approximately 36 hours. In addition, convergence problems
caused by rigid body motions, had to be overcome by assuming artificial boundary conditions
on the T-stub section.
Column flange
T-stub
Beam
As described in the previous section, in the last decade, the FE method has become a
popular tool to solve a wide range of structural mechanics problems. Two different types of
solution strategies can be distinguished : the explicit and the implicit solution procedures.
This section briefly describes and compares the following aspects of both approaches : the
solution strategy, computational time as function of model size, and convergence in a non-
linear contact analysis. Further information is presented by van der Vegte et al (4).
Solution strategy
IMPLICIT APPROACH : The implicit method is based on static equilibrium and is characterized
by the simultaneous (implicit) solution of a set of linear equations. For a model to be in static
equilibrium, the net force acting on each node must be zero. After a full set of linear
equations has been assembled, this set of equilibrium equations is then solved
simultaneously and the unknown nodal displacements are obtained. The assembly of a
As might be clear from the explicit solution algorithm, explicit simulations may yield results
significantly affected by dynamic effects. A possible tool to identify whether or not this occurs,
involves monitoring the various components of the energy balance throughout the loading
process.
IMPLICIT APPROACH : Although for the implicit method, the computational cost as a function of
model size is rather difficult to predict, experience shows that for many problems, the CPU
time is approximately proportional to the square of the number of degrees of freedom.
EXPLICIT APPROACH : Using the explicit method, the computational cost is proportional to the
number of elements and roughly inversely proportional to the smallest element dimension.
Implicit
CPU time
Explicit
0
0
A qualitative comparison between CPU time and model size for both approaches is
illustrated in figure 3. Although for small models, the implicit method may be favourable, for
larger models the explicit approach becomes more attractive. This becomes even more valid
when disk space and memory requirements are taken into account. As result, for very large
models, the maximum size of an implicit model may be controlled by the available memory
and disk space instead of the required computational time.
IMPLICIT APPROACH : Figure 4 shows a flow chart for one increment in a non-linear contact
analysis. From this diagram, it becomes clear that within each increment, two iterative loops
can be distinguished. Only after both criteria regarding contact constraints and equilibrium
are satisfied, the analysis proceeds with the next increment. In case of severe discontinuity, it
EXPLICIT APPROACH : In a non-linear explicit analysis including contact, it is first assumed that
no contact occurs between the various entities. If, at the end of the increment, overclosure
(i.e. penetration due to contact) is found, corrections for the kinematics of the contacting
bodies are calculated and adopted that would have been required to prevent penetration
from taking place. Unlike the implicit approach, no convergence criteria are considered and,
as a result, contact simulations proceed smoothly.
Start of increment
Perform iteration
Check changes in
contact
No changes
Check equilibrium
Convergence
End of increment
In 2002, Miura et al (5) reported the results of three large-scale tests on bolted beam-to-
column connections without weld access holes. Because of the detailed description of the
experiments, these tests were selected to serve as a reference to evaluate the reliability of
the explicit FE predictions.
The connection details of two of the three T-shaped configurations are shown in figure 5. In
each of these specimens, a wide flange beam was partially welded and bolted to a cold-
formed square hollow section column. Specimen A contains an internal diaphragm plate
welded inside the column located at the bottom flange of the beam. The web and bottom
flange of the beam are bolted to a shop-welded shear tab and flange plate respectively. The
top flange of the beam is field-welded to the external diaphragm plate. In Specimen C, the
beam web is bolted to a shop-welded shear tab, whereas both flanges are field-welded to the
external diaphragm plates.
The tip of the beam is subjected to reversed loading with increasing amplitude, while both
column ends are fixed against displacements.
Numerical model
FE MESH : Because the configurations tested by Miura et al lack symmetry, a full scale FE
model had to be developed for each specimen. The FE meshes, primarily modelled with 8-
noded solid elements with reduced integration (ABAQUS element type C3D8R), follow the
mesh layout recommended by van der Vegte et al (4). Based on the numerical results of a
simplified connection between three plates and a single bolt, it was found that, in order to
obtain reliable simulation results (i) at least 24 elements should be modelled around the
perimeter of a bolt and bolt hole, and (ii) at least 4 layers of solid elements should be
considered through the thickness of each plate. In addition, the aspect ratio of the linear solid
elements should not exceed 10. Figure 6 illustrates the mesh adopted for the bolt and
washers. For clarity, only one of the washers is displayed. Figure 7 displays the FE mesh
generated for Specimen A. The FE mesh of Specimen C is similar. The FE mesh of
Specimen A (with 17 bolts) contains approximately 135,000 nodes.
CONTACT, FRICTION AND PRE-LOADING : Similar to the explicit analyses on the single-bolted
configuration by van der Vegte et al (4), each set of a bolt and two washers considers 6
“contact pairs”. In addition, the following interactions are modelled : (i) contact between the
beam web and the web plate and (ii) contact between the beam flange and the flange plate.
As a result, the FE model of Specimen A with 17 bolts, includes 104 contact definitions.
Friction is included, whereas the friction coefficient is taken as either 0.3 (for contact
involving the bolts) or 0.6 (for contact between rusted surfaces). Following the
recommendations of AIJ (17), the bolts are pre-loaded by a tensile stress of 520 N/mm2.
KINEMATIC HARDENING : The kinematic hardening model adopted in the current study
considers linear hardening only. For the hardening modulus, a value of E/100 is assumed.
LOADING PROCEDURE : In the experiments, the amplitude of the beam rotation was increased
as 2θp, 4θp, 6θp ... up to failure, whereas θp is a theoretical, calculated value of the beam
rotation at full plastic beam moment (1). For each displacement increment, the beam was
subjected to two cycles. However, in the numerical analyses, the beam is subjected to only
one cycle for each displacement increment, due to constraints in computational time.
Numerical results
Because of the large demand in computational time, the numerical simulations of Specimens
A and C were terminated when the beam tip rotation was between 6θp and -6θp, providing
sufficient data to assess the reliability of the FE model.
Figure 8 displays both the numerical and experimental results, summarized in hysteresis
loops. The moment in the beam (at column face) Mm, made non-dimensional by the full
plastic beam moment Mp, is plotted against the non-dimensional beam-end rotation θm/θp.
Figure 8 reveals that for both specimens, the outer points of the experimental and numerical
loops (i.e. points of load reversal) are close. The slopes of both sets of curves near the points
where the load reverses, also match well. However, the numerical loops overestimate the
stiffness of the specimens shortly after the load reverses. The divergence between the
experimental and numerical loops is likely caused by the relatively simple, linear kinematic
hardening model adopted in this study. These observations are well in line with the
conclusions made by Goto et al (18) who conducted numerical research into the hysteretic
behaviour of thin-walled columns.
Even more important, unlike the frequent occurrence of convergence failures observed in
implicit contact simulations, the explicit analyses of highly non-linear FE models A and C
proceeded smoothly without any computational difficulty.
1.5 1.5
Num erical Numerical
1.0 Experim ent 1.0 Experim ent
0.5 0.5
M m /M p
M m /M p
0.0 0.0
-0.5 -0.5
-1.0 -1.0
-1.5 -1.5
-6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6
θm / θp θm / θp
CONCLUSIONS
After having presented a brief overview of the implicit and explicit FE solution techniques,
explicit FE simulations have been conducted on bolted beam-to-column connections. Based
on the results, the following conclusions can be made :
a. The explicit solution technique is able to provide reliable predictions of the hysteretic
response of actual bolted beam-to-column connections, even when the simple, linear
kinematic hardening model is used. However, for further improvement of the shape of
the numerical hysteretic loops, a more sophisticated material model should be
considered.
b. The explicit method is a suitable tool to simulate the behaviour of bolted connections
effectively. Unlike implicit analyses of bolted connections, the explicit solution technique
does not require simplifications to be made in order to generate a reliable solution.
c. The explicit solution algorithm enables the numerical simulation of FE models much
larger than the model size attainable by the implicit strategy.
d. The explicit method requires a more intense evaluation of the results than the implicit
methodology. The explicit solution, if not correctly applied, may yield results significantly
affected by dynamic effects. For the explicit analyses, various components of the
energy balance should be monitored throughout the loading process. Hence, the use of
the explicit method is only recommended for advanced FE users.
ACKNOWLEDGEMENTS
The first author would like to thank the Japan Society for the Promotion of Science for the
opportunity and financial support to carry out the research reported.
(1) Kurobane, Y., (1998). Improvement of I-Beam-to-RHS Column Moment Connections for
Avoidance of Brittle Fracture. Proc. of the 8th International Symposium on Tubular
Structures, Singapore, Keynote Paper, pp. 3-17.
(2) ABAQUS/Standard, (2000). Version 6.1, Hibbitt, Karlsson & Sorensen, U.S.A.
(3) ABAQUS/Explicit, (2000). Version 6.1, Hibbitt, Karlsson & Sorensen, U.S.A.
(4) Vegte, G.J. van der, Makino, Y. and Sakimoto, T., (2002). Numerical Research on
Single-Bolted Connections Using Implicit and Explicit Solution Techniques. Memoirs of
the Faculty of Engineering, Kumamoto University, Vol. 47, No. 1.
(5) Miura, K., Makino, Y., Obukuro, Y., Kurobane, Y., Vegte, G.J. van der, Tanaka, M. and
Tokudome, K., (2002). Testing of Beam-to-RHS Column Connections Without Weld-
Access Holes. International Journal of Offshore and Polar Engineering, Vol. 12, No. 3,
pp. 229-235.
(6) Krishnamurthy, N. and Graddy, D.E., (1976). Correlation between 2- and 3-Dimensional
Finite Element Analysis of Steel Bolted End-Plate Connections. Computers and
Structures, Vol. 6, pp. 381-389.
(7) Krishnamurthy, N., Huang, H.T., Jeffrey, P.K. and Avery, L.K., (1979). Analytical M-θ
Curves for End-Plate Connections. Journal of Structural Division, ASCE, Vol. 105, No.
1, pp. 133-145.
(8) Sherbourne, A.N. and Bahaari, M.R., (1994). 3D Simulation of End-Plate Bolted
Connections. Journal of Structural Engineering, Vol. 120, No. 11, pp. 3122-3136.
(9) Sherbourne, A.N. and Bahaari, M.R., (1997). Finite Element Prediction of End-Plate
Bolted Connection Behaviour. I : Parametric Study. Journal of Structural Engineering,
Vol. 123, No. 2, pp. 157-164.
(10) Choi, C.K. and Chung, G.T., (1996). Refined Three-Dimensional Finite Element Model
for End-Plate Connection. Journal of Structural Engineering, Vol. 122, No. 11, pp. 1307-
1316.
(11) Bursi, O.S. and Jaspart, J.P., (1997). Benchmarks for Finite Element Modelling of
Bolted Steel Connections. Journal of Constructional Steel Research, Vol. 43, Nos. 1-3,
pp. 17-42.
(12) Bursi, O.S. and Jaspart, J.P., (1997). Calibration of a Finite Element Model for Isolated
Bolted End-Plate Steel Connections. Journal of Constructional Steel Research, Vol. 44,
No. 3, pp. 225-262.
(13) Wheeler, A.T., Clarke M.J. and Hancock, G.J., (2000). FE Modelling of Four Bolt,
Tubular Moment End-Plate Connections. Journal of Structural Engineering, Vol. 126,
No. 7, pp. 816-822.
(14) Willibald, S., Packer, J.A. and Puthli, R.S., (2001). Bolted Connections for RHS Tension
Members. CIDECT Report 8D/8E-10/01, University of Toronto, Canada.
(15) Swanson, J.A. and Leon, R.T., (2000). Bolted Steel Connections : Tests on T-Stub
Components. Journal of Structural Engineering, Vol. 126, No. 1, pp. 50-56.
(16) Swanson, J.A., Kokan, D.S. and Leon, R.T., (2000). Advanced Finite Element Modelling
of Bolted T-Stub Connection Components. Journal of Constructional Steel Research,
Vol. 58, Nos. 5-8, pp. 1015-1031.
(17) AIJ, (2001). Recommendation for Design of Connections in Steel Structures.
Architectural Institute of Japan (in Japanese).
(18) Goto, Y., Wang, Q. and Obata, M., (1998). FEM Analysis for Hysteretic Behavior of
Thin-Walled Columns. Journal of Structural Engineering, Vol. 124, No. 11, pp. 1290-
1301.
ABSTRACT
Four-parameter power model for predicting moment relative-rotation relation of
semi-rigid connections has been established which is composed of initial
stiffness, strain-hardening stiffness, reference connection moment, and shape
parameter. These parameters for total 168 experimental test data of the end-
plate type connections stored in the connection database were determined. An
applicability of the model using the parameters determined here to frame
analysis was investigated comparing with the numerical analysis results
obtained directly using experimental test data.
INTRODUCTION
To rationally perform structural analysis and design calculation for flexibly jointed frames, it is
very important to establish a simple model for estimating nonlinear moment relative rotation (M-
θr) relation for semi-rigid connections. The authors proposed the use of three-parameter power
model for predicting nonlinear M-θr curves of angle type connections. However, this model may
not be much applicable for estimating M-θr curves of the end-plate type connections because
clear-cut strain-hardening connection stiffness is indicated in these experimental test data. On
the other hand, it is well known that four-parameter power model can present M-θr curves of
semi-rigid connections with strain-hardening connection stiffness. This is one expanded model of
three-parameter power model.
From this point of view, in this paper, in order to rationally estimate nonlinear M-θr curves of the
end-plate type connections stored in the connection database, four parameters for each 168
experimental test data were decided, and an applicability of the model was investigated
comparing with the numerical analysis results for a flexibly jointed frame obtained using
experimental test data for the end-plate connections.
END-PLATE CONNECTIONS
The end-plate connections, usually shop welded and field bolted, are widely used in practice due
to its simplicity of fabrication and erection. Depending upon the relative length of the end-plate to
beam height, there are three types of the end-plate connections in practice: extended end-plate
connections (Figure 1a), flush end-plate connections (Figure 1b), and header plate connections
(Figure 1c).
Observing a significant number of experimental M-θr curves of extended and flush end-plate
connections in the database, it is clearly understood that these connections possess
pronounced strain-hardening connection stiffness. It means that four-parameter power model
including strain-hardening connection stiffness rather than three-parameter power model should
be applied to estimate these M-θr relations.
Now, a total of 168 experimental test data for these types of connections are installed in the
database (5, 6). The references and corresponding number of tests are listed in Table 1.
A four-parameter power model, which is proposed by Richard and Abbott (7) for modelling
elastic-plastic stress-strain relation, is applied. Using this power model, the connection moment
M and relative rotation θr, and tangent connection stiffness Rkt are represented as follows:
dM (Rki − Rkp )
Rkt = = + Rkp (3)
(n +1)
dθr
(1 + ( θr
θ0
n
) ) n
where Rki = initial connection stiffness; Rkp = strain-hardening connection stiffness; n = shape
parameter; θ0 = reference relative rotation [= M0/(Rki-Rkp)]; and M0 = reference connection
moment.
Figure 2 shows the M-θr curves of a connection when a magnitude of shape parameter n is
varied from small value to infinity. From this figure, it is seen that if shape parameter n is taken to
be infinity, the model is reduced to a bilinear one with the initial connection stiffness Rki and
strain-hardening connection stiffness Rkp. The model has the following merits:
1. The model is composed of only four parameters and its formulation is simple and
straightforward;
2. The model requires less computing time in a nonlinear structural analysis, since not only
connection moment M but also relative rotation θr, and tangent connection stiffness Rkt can
be represented using simple closed form equations as indicated in Eqs (1) to (3);
3. If three parameters of Rki, Rkp and M0 were determined using simple structural mechanics,
an empirical equation for shape parameter n might be determined by using the database
and M-θr relation for the end-plate connection with arbitrary parameters might be evaluated.
The initial connection stiffness Rki is defined as the linear relationship between moment M and
relative rotation of connection θr in the elastic region. In general, the experimental moment-
rotation curve is expected to be continuously smooth over the initial range of loading. In such
ideal case, initial connection stiffness Rki can be estimated using tangent connection stiffness at
the initial loading point. However, there may be some experimental M-θr data with discontinuous
even in the elastic region as shown in Figure 3. To rationally estimate initial connection stiffness
Rki for all cases including the above mentioned experimental test data, here, the maximum
secant connection stiffness among all data from initial to final loading is taken as the stiffness:
in which Mi and θi are the i-th experimental test data from initial loading.
The strain-hardening connection stiffness Rkp is evaluated taking tangent connection stiffness
obtains drawing straight line backward from the point of the maximum connection moment (Mk,
θk). The cases softened in the post peak region are excluded from the consideration because
these cases need more complicate procedure in frame analysis. Strain-hardening connection
stiffness Rkp for each experimental test data can be obtained using following equation:
Mk − M j
Rkp = (5)
θk − θ j
min
in which suffixes k and j are the point at the maximum connection moment and arbitrary point
less than point k.
The reference connection moment M0 is the value where the straight-line representing strain-
hardening connection stiffness Rkp intersects the moment axis in the moment and relative-
rotation diagram as shown in Figures 2 and 3. It is represented as:
M0 = Mk − Rkp θk (6)
Shape parameter n
The shape parameter n is introduced for fitting the curve obtained using the power model with
experimental test data under three parameters being determined as shown in Figure 2. Here, it
is determined numerically applying least-mean-square technique for the connection moment
obtained using Eq. (1) and experimental test data
The four parameters determined based on the above procedures for each experimental test data
are listed in Table 2. Figure 4 shows the comparison among experimental test data and two
curves obtained using four-parameter power model proposed here and modified exponential
model proposed by Kishi and Chen (5). From these figures, it can be seen that M-θr curves
predicted using four-parameter power model agrees rather well with experimental test data, and
those from the modified exponential model represent precisely experimental test data.
Figure 4. Comparisons of the prediction models with test data in connection database.
program (8) is used for this investigation which is capable of rationally analyzing flexibly jointed
frames with nonlinear connection stiffness. Three kinds of frame responses are used for this
investigation: 1) the nodal displacement under service load, 2) the fixed end moment of column
under factored load, and 3) the beam end moment under factored load. These response values
are normalized with reference to the corresponding values obtained using modified exponential
model as shown in Eqs (7) and (8).
Normalized displacement:
These normalized displacement d* and normalized moment m* are then plotted taking
logarithmic initial connection stiffness Rki .
Figure 6 shows an example of normalized displacement d* versus log10 Rki obtained for node
3. In this figure, µ and σ indicate a mean value and standard deviation among all response
values respectively. In this case, since µ = 1.0 and σ = 0.02, it is seen that nodal
displacement under service load can be better predicted using four-parameter power model.
Table 3 summarizes the numerical results for normalized nodal displacement and shows that
all mean values µ’s are almost unity and the results for 94 % of connection data distribute in
the region less than 5% error.
Figure 8 shows the distributions of non-dimensional beam end moments for elem. no. 6. From
this figure, it is seen that the moment in the region of weak connection stiffness are scattered in
the area more than 10 % error. However, in the region of stiffer connection stiffness, the moment
has a tendency to concentrate in unity. The numerical results of beam end moment are listed in
Table 5. Standard deviations σ’s for corresponding beam end moments of elem. no. 6 are
distributed in the range of 0.01 to 0.05. More than 92% connection test data are distributed in
region less than 5 % error.
Standard deviation σ of non-dimensional beam end moment for node no. 3 of elem. no. 5 is very
large comparing with those of the other end moments. This may be due to the reasons that 1)
the moment is at windward beam end and becomes smaller than that of leeward one; and then
2) its sensitivity becomes higher than those in cases of the other columns. However, from the
engineering point of view, the windward beam end moment may not be much important. Indeed,
it has been confirmed that the windward beam end moments for all connection data are almost
1/80 – 1/2 those at the leeward node.
(a) Elem. no. 6, node no. 5 (b) Elem. no. 6, node no. 6
Then, from Figure 8 and Table 5, it can be verified that four-parameter power model for
estimating M-θr relation of end-plate connections can be applicable in semi-rigid frame analysis
in lieu of real experimental test data and/or modified exponential model.
A four-parameter moment-rotation model has been proposed for estimating moment M and
relative rotation θr relation of the end-plate beam-to-column connections. The first three
parameters, initial connection stiffness Rki, strain-hardening connection stiffness Rkp, and initial
connection moment M0 were determined based on the experimental test data. The fourth
parameter, shape parameter n, is numerically determined applying the least-mean-square
technique for connection moments between the proposed model and experimental test data. It is
confirmed that the proposed model with four parameters decided here can better simulate M-θr
relation for experimental connection test data installed in the connection database.
NOTATION
REFERENCES
(1) Kishi, N. and Chen, W.F. (1990). “Moment-rotation relations of semi-rigid connections
with angles.” Journal of Structural Engineering, ASCE, 116(7), 1813-1834.
(2) Kishi, N., Chen, W.F., Goto, Y. and Hasan, R. (1997). “Study of Eurocode 3 steel
connection classification”, Engineering Structures, 19(9), 772-779.
(3) Hasan, R., Kishi, N., Chen, W.F., and Komuro, M. (1997). “Evaluation of rigidity of
extended end-plate connections”, Journal of Structural Engineering, ASCE, 127(12),
1595-1602.
(4) Kishi, N., Hasan, R., Chen, W.F. and Goto, Y. (1994). “Power model for semi-rigid
connections.” Steel Structures, Journal of Singapore Struct. Steel Society, 5(1), 37-48.
(5) Kishi, N and Chen, W.F. (1986). “Data base of steel beam-to-column connections.”
Struct. Engrg. Report No. CE-STR-86-26, Purdue Univ., West Lafayette, Ind.
(6) Hasan, R. (1996). “Evaluation of connection stiffness and modeling of semi-rigid
connections.” Thesis presented to Muroran Institute of Technology, in Partial
Fulfillment of the Requirements for the Degree of Doctor of Philosophy.
(7) Richard, R.M. and Abbott, B.J. (1975). “Versatile elastic-plastic stress-strain formula.”
Journal of the Engineering Mechanics, ASCE, 101(4), 511-515.
(8) Goto, Y. and Chen, W.F. (1987). “On the computer-based design analysis for flexibly
jointed frames.” Journal of Constructional Steel Research, Special Issue on Joint
Flexibility in Steel Frames (W.F. Chen Ed.), Vol. 8, 203-231.
ABSTRACT
To establish a numerical analysis method for appropriately evaluating moment
relative-rotation relations of top- and seat-angle connections under monotonic
loading, three-dimensional elasto-plastic finite element analysis was performed
under following considerations: 1) all the connection components are modeled
using solid element; 2) a contact surface algorithm is applied between every
adjacent two components; and 3) bolt pretension is introduced. An applicability
of this analysis method for this type connections was discussed comparing
with the experimental results. From this study, it can be concluded that
applying the proposed analysis method, moment relative-rotation relations of
top- and seat-angle connections can be accurately estimated up to the ultimate
state.
INTRODUCTION
To perform the semi-rigid frame analysis and design, it is important to evaluate the moment
relative-rotation behavior of the beam-to-column connections. Many experimental studies
have addressed prediction of the moment relative-rotation characteristics of top- and seat-
angle type connections all over the world (1, 2, 3, 4). On the other hand, over the past few
years, several numerical simulations of angle type connections have been conducted by
means of elasto-plastic finite element (FE) analysis method. Kishi et al (5) and Citipitioglu et
al (6) investigated the effects of pretension of bolts and friction coefficient between
connection components on moment relative-rotation curves of top- and seat-angle type
connections in detail. In these studies, moment relative-rotation curves of the connections in
the initial loading area can be approximately predicted by using the FE analysis. However, in
the large deformation area including plastic deformation, connection moment relative-rotation
relation may not be appropriately estimated.
In this study, in order to establish a numerical analysis method for appropriately evaluating
moment relative-rotation relations of top- and seat-angle with/without double web-angle
connections in the finite deformation area under monotonic loading, three-dimensional
elasto-plastic FE analysis was performed under following considerations: 1) all the
connection components (beam, column, angle, etc.) are modeled using solid element; 2) a
contact surface algorithm is applied between every adjacent two components; and 3) bolt
pretension is introduced. Here, ABAQUS code (7) was used for numerical analysis of the
three kinds of connection specimens, in which dimensions of the web angles were taken as
variables. An applicability of this analysis method for this type of connections was discussed
comparing with the experimental results conducted by the author’s laboratory (4). Moment
relative-rotation behavior of the connection, strain relative-rotation relations of top and/or web
EXPERIMENTAL OVERVIEW
Specimens
Test setup
Dimensions and geometry of the specimens are shown in Figure 3. The specimens were
attached to the column (H408×408×21×21), which was set on the steel bed and was
strengthened welding vertical stiffeners to keep from local buckling. The lateral load was
applied to the top end of cantilever beam through a swivel using a 500 kN screw jack. The
loading point was at a height of 1,500 mm from the column face. Two guide rails were used
to control the loading direction.
The lateral load and displacement were measured by using a load cell (capacity: ± 300 kN)
and laser type Linear Variable Displacement Transducers (LVDTs) (maximum stroke: ± 250
mm), respectively. The vertical displacements at the four side-edges of beam flanges bolted
to angles were measured by using laser type LVDTs to evaluate the relative-rotation of
connection (Figure 3). Strains at some locations of angles were measured to investigate
strain and deformation behavior of angles. All of these measured data were continuously
recorded using digital data-recorders.
Lateral load was surcharged following a displacement control method. No axial load was
applied to the beam. To ensure the safety of experiment, monotonic load was increased until
displacement at the loading point reaches about 145 mm except the case of W00 specimen
which reached ultimate state due to bolts failure. Therefore, ultimate moment capacity of
connection for W18 and W29 specimens may not be precisely evaluated in this experiment.
Evaluation of relative-rotation
M = PH ⋅ h (2)
where PH = laterally surcharged load; h = a distance from the center of swivel to column face
(h = 1,500 mm).
Relative connection rotation θr and connection moment M discussed later was estimeted
using Eqs (1) and (2), respectively.
FE MODEL
In this study, ABAQUS code was used to estimate the connection moment relative-rotation
behavior of angle type connections. Figure 5 shows an example of FE model for W29
specimen. A half of specimen referring to the symmetrical axis was modeled. All connection
components (angle, beam, column, bolt) were modeled using eight-node solid elements.
Total number of nodal points and elements for W29 specimen are 55,810 and 33,729,
respectively. The bottom flange of column is assumed to be perfectly fixed according to the
experimental boundary conditions.
To accurately simulate the connection behavior, small sliding occurred between every
adjacent two components is explicitly considered applying a contact surface algorithm
prepared in ABAQUS code. These are 1) between bolt shank and bolt hole, 2) between bolt
head/nut and connecting components, and 3) between angles and beam/column flanges. All
components including bolts are completely independent from each other as assemblages in
real connection. Bolt holes which are 2 mm larger than the bolt size (D = 20 mm) were drilled.
178 kN pretension force was introduced into each bolt according to the experimental
conditions. Figure 6 shows the contact surface of top angle and Mises stress contour after
pretension being introduced into bolts. From this figure, it is confirmed that the pre-stress
was transfer from bolt to angle.
Figure 10. Comparison of stress relative-rotation relation in top angle between numerical
and experimental results.
Figure 11 shows the deformation configuration and equivalent plastic strain distribution of top
angle of W00 specimen. From this figure, it is observed that the toe of the fillet of angle in
beam flange and the area around bolt hole of angle in column flange have been yielded from
the initial loading stage, and these areas are gradually extended with an increase of relative-
rotation θr. Since a similar behavior was also observed from experimental results, the
proposed numerical analysis method can appropriately estimate the strain distribution of the
top angle.
Figure 12. Comparison of stress relative-rotation relations in web angle of W29 specimen
between numerical and experimental results.
Figure 13 shows the deformation configuration and equivalent plastic strain distribution for
web angle of W29 specimen. From this figure, it is seen that the toe of the fillet of angle in
beam web is yielded first at θr = 11 mrad, and then yielded area is developed from tension
side (c1 section) to compression side (c5 section) with an increase of θr.
Figure 14 shows the deformation configuration and Mises stress distribution of top angle
including bolt at four loading stages. From these figures, it is observed that the heel of top
angle is apart from the surface of column flange at θr = 5 mrad, and then top angle is
gradually deformed with an increase of θr. At θr = 39 mrad, the area from edge of bolt hole
through angle’s heel is completely apart from the surface of column flange due to the prying
action of angle. Moreover, the yielded area is developed to the middle of bolt shank, in which
Mises stress reached up to more than 1,000 MPa. At θr = 85 mrad, yielded area is developed
to the whole of the bolt shank.
Figure 15 shows the deformation configuration and Mises stress distribution of the top angle
including bolt at θr = 114 mrad of relative-rotation at which bolt failure was occurred in
experiment, and a look of failure bolt after the experiment. From this figure, it is observed that
the bolt shank was significantly deformed in numerical analysis, and the deformation
configuration of bolt is good correspond to a look of fractured bolt in experiment. Thus, it is
indicated that the collapse mode of the top- and seat-angle type connections may be
appropriately estimated using the deformations and Mises stress distribution of connection
components.
CONCLUSIONS
In this study, in order to establish a numerical analysis method for appropriately estimating
moment relative-rotation relation of top- and seat-angle with/without double web-angle
connections from initial stage to finite displacement stage under monotonic loading, three-
dimensional elasto-plastic finite element analyses were performed. The results obtained from
this study are as follows:
1) applying the proposed FE analysis method, moment relative-rotation relations of top- and
seat-angle type connections can be better estimated regardless of longitudinal size of
web angle;
2) strain distribution of top- and/or web-angle can be estimated by using the proposed
analysis method qualitatively, and
3) failure mode of connection may be estimated using the deformation and Mises stress
distribution of connection components.
ACKNOWLEDGEMENT
This research was supported in part by a grant from ‘The Japan Iron and Steel Federation’.
The support is gratefully acknowledged.
dl horizontal distance between two laser type LVDTs set at tension and compression
side flange of beam
Es Young's modulus of steel
fu tensile strength of steel
fy yield stress of steel
h distance from the center of swivel to column face
M connection moment
PH lateral surcharged load
δ1 vertical displacement at tension side edge of beam flange bolted to angle
δ2 vertical displacement at compression side edge of beam flange bolted to angle
θr relative connection rotation
νs Poisson's ratio
REFERENCES
(1) Kukreti, A.R. and Abolmaali, A.S. (1999). Moment-rotation hysteresis behavior of top
and seat angle steel frame connections, Journal of Structural Engineering, 125(8), 810-
820.
(2) Calado, L., Matteis, G.D., and Landolfo, R. (2000). Experimental response of top and
seat angle semi-rigid steel frame connections, Materials and Structures, 33, 499-510.
(3) Azizinamini, A. (1985). Cyclic Characteristics of bolted semi-rigid steel beam to column
connections, PhD thesis, University of South Carolina, Columbia.
(4) Komuro, M., Kishi, N., and R.Hasan (2003). Quasi-static loading tests on moment-
rotation behavior of top- and seat-angle connections, Proceedings of the Conference
on Behaviour of Steel Structures in Seismic Areas, Naples, Italy, June 9-12, 329-334.
(5) Kishi, N., Ahmed,A., Yabuki, N., and Chen, W.F. (2001). Nonlinear finite element
analysis of top- and seat-angle with double web-angle connections., International
Journal of Structural Engineering and Mechanics, 12(2), 201-214.
(6) Citipitioglu, A.M., Haj-Ali, R.M., and White, D.W. (2002). Refined 3D finite element
modeling of partially restrained connections including slip, Journal of Constructional
Steel Research, 58, 995-1013.
(7) ABAQUS/Standard user’s manual (1998). Ver. 5.8, Hibbitt, Karlsson & Sorensen.
ABSTRACT
Elastic and inelastic deformations in the panel zone of the beam-column joint
region of moment resisting frames are responsible for a very significant
portion of the lateral flexibility of such systems. This paper provides a brief
theoretical basis for computing panel zone deformations, and compares
results obtained from two simple mechanical models to each-other and to
those obtained using detailed finite element analysis. It is shown that the
simplest mechanical model, referred to as the Scissors model, produces
results that are comparable to those obtained from the more complex
mechanical model, and also correlates well with the results computed from
the detailed finite element model.
INTRODUCTION
The influence of panel zone deformations on the flexibility of steel moment resisting frames
is very significant. This is true for elastic response, and particularly for inelastic response
when yielding occurs in the panel zone. Structural analysis should always include such
deformations.
While the state of stress in the panel zone is extremely complex, the sources of deformation
can be divided into three parts: axial, flexural, and shear. For low to medium rise frames,
axial deformations are negligible, flexural deformations are minor but significant, and shear
deformations are dominant. This paper concentrates on the shear component of panel
zone deformation. See Downs (1) for a detailed discussion on modeling approaches for
axial and flexural deformations within the panel zone.
Mathematical modeling procedures for panel zone deformation are typically based on
simple mechanical analogs which consist of an assemblage of rigid links and rotational
springs. The principal challenge in the derivation of such models is the development of the
transformations from shear in the panel zone to rotation in the springs of the analog. Two
mechanical models were studied in the research reported herein. These are the
“Krawinkler Model” (2) and the “Scissors Model”. When properly used, the results obtained
from these models are essentially identical, even though the kinematics of the Krawinkler
model is significantly different than that of the Scissors model. Unfortunately the Scissors
model is often misused in practice because analysts tend to compute spring properties that
were derived for the Krawinkler model and use them in the Scissors model. A complete
description of the mechanical models is presented in the next section of this paper.
Results obtained from structures implementing the simple mechanical models were
compared with those computed from a detailed finite element model of a full beam-column
subassemblage. The detailed model was created using ABAQUS (3). Both elastic and
inelastic analysis was performed. It was found that good correlation was obtained between
Terms α and β represent the ratios of the effective depth of the column to the span length,
and the effective depth of the girder to the column height, respectively. The effective depth
of a section is defined as the distance between the centers of the flanges. Use of these
terms in lieu of the actual physical dimensions greatly simplifies the derivation of the
properties of the models.
VC
VC H
VC / H
L
H βH
VVCC/HH
L
VC
αL
L
The total drift in the subassemblage, ∆, is defined as the lateral displacement of the top of
the column with respect to the bottom of the column under the load VC. Following the
procedure described by Charney (6), this drift may be divided into three components, one
for the column, one for the girder, and one for the panel zone.
∆ = ∆C + ∆G + ∆ P (1)
The column and girder displacement components are due to axial, flexural and shear
deformations occurring in the clear span region of the respective sections. The panel
contribution to displacement may also be divided into axial, flexural, and shear components:
∆ P = ∆ PA + ∆ PF + ∆ PV (2)
If it is assumed that the moment in the girder at the face of the column is resisted entirely by
the flanges of the girder, it can be shown by simple statics that the horizontal shear force in
the panel zone is
VC (1 − α − β )
VP = (3)
β
VC H (1 − α − β )
τP = (4)
∇P
This shear stress is uniform throughout the volume of the panel zone. The term ∇ P , which
represents the volume of the panel zone, appears repeatedly in the following derivations.
To determine the panel zone contribution to subassemblage drift, equal and opposite unit
virtual forces are applied in lieu of the actual column shears VC. The shear stress in the
panel due to the unit virtual shear force is
H (1 − α − β )
τ1 = (5)
∇P
The contribution of panel zone shear strain to subassemblage drift is obtained by integrating
the product of the real strains and the virtual stresses over the volume of the panel. The
uniformity of stress and strain over the volume of the panel simplifies the integration.
τ Pτ 1 VC H 2 (1 − α − β ) 2
∆ PV = ∫ dV = (6)
V
G G∇ P
The Krawinkler model and the Scissors model must have the same panel zone shear
contribution to displacement as given by equation 6.
Research performed by Krawinkler (2) has shown that the strength of the panel zone
consists of two components; shear in the panel itself, and flexure in the column flanges.
The larger of these components is the panel zone shear, which is resisted by the web of the
column acting in unison with the doubler plate, if present. If it is assumed that the yield
stress in shear is 1 / 3 ≅ 0.6 times the uniaxial yield stress and that the column and doubler
plate are made from the same material, the yield strength of the panel in shear is
0.6 Fy ∇ P
VYP = 0.6 FyαLt P = (7)
βH
Force-deformation response
The assumed force-deformation behavior of the beam column joint is illustrated in figure 2.
In the figure the deformation is the racking displacement over the height of the panel. The
moment-rotation aspect of figure 2 is used later.
Shear, V
Moment, M
VYP
Total
βH
VYP MYP
Panel
MYP
δY
VYF βH
Flange θY
MYF
The total response is equal to the sum of the response of the panel and the column flanges.
Following Krawinkler (2) it is assumed that the flange component yields at four times the
yield deformation of the panel component. It should be noted that figure 2 shows that the
flange component of the resistance is effective immediately upon loading. Krawinkler
assumes that this component of resistance does not occur until the panel yields in shear.
We have used the relationship shown in the figure as it simplifies the implementation of the
mechanical models without compromising accuracy.
The Krawinkler and Scissors models must be proportioned such that yielding is consistent
with figure 2.
The Krawinkler model is shown in figure 3. The model consists of four rigid links connected
at the corners by rotational springs. The springs at the lower left and upper right corners
have no stiffness, and thereby act as true hinges. The spring at the upper left is used to
represent panel zone shear resistance, and the spring at the lower right is used to represent
column flange bending resistance. A total of twelve nodes are required for the model (there
are two nodes at each corner). The number of degrees of freedom in the model depends
Rotational Spring
for Panel Shear
Rigid Link
The properties of the springs in the Krawinkler model are easily computed in terms of the
physical properties. Looking at only the panel spring, for example, the moment in the spring
is equal to the panel shear times the height of the panel. (See the diagram at the right of
figure 2.) The rotation in the spring is equal to the shear displacement in the panel divided
by the panel height. Hence,
M P , K = VP βH (9)
VP βH 1 V βH
θ P ,K = = P (10)
GαLt P βH G∇ P
Note that the “K” subscript in the above expressions refers to the Krawinkler model.
The stiffness of the rotational spring representing the panel in the Krawinkler model is the
moment divided by the rotation;
M P ,K
S P ,K = = G∇ P (11)
θ P ,K
The yield moment in the spring is simply the panel shear strength times the height of the
panel. Using equation 7,
As seen in figure 2, the stiffness of the flange bending component of the Krawinkler model
is equal to the yield moment in the flange bending component divided by 4.0 times the yield
rotation of the panel component. The yield rotation of the spring representing the panel
component is
M YP, K FY
θ YP, K = = 0.6 (13)
K YP, K G
The yield moment is equal to the yield strength times the panel height;
M YF , K
S F,K = = 0.75(bCf )(t Cf ) 2 G (15)
4θ YP, K
In summary, Expressions 11 and 12 and 14 and 15 are all that are needed to model the
panel spring and the flange spring, respectively, in the Krawinkler model. If desired, a strain
hardening component may be added.
The Scissors model is shown in figure 4. This model derives its name from the fact that the
model acts as a scissors, with a single hinge in the center. Only two nodes are required to
model the joint if rigid end zones are used for the column and girder regions inside the
panel zone. The model has four degrees of freedom. As with the Krawinkler model, one
rotational spring is used to represent the panel component and the other is used to
represent the flange component of behavior.
The properties of the Scissors model are determined in terms of those derived previously
for the Krawinkler model. First, consider the displacement participation factor for panel
shear as derived in equation 6. Noting that the denominator of this equation is the same as
the panel spring stiffness for the Krawinkler model, equation 6 may be rewritten as
VC H 2 (1 − α − β ) 2
∆P = (16)
S P ,K
For the Scissors model, the moment in the spring under the column shear VC is simply VCH.
If the Scissors spring has a stiffness SP,S, the rotation in the spring is VCH/SP,S. The drift over
the height of the column is the rotation times the height, thus for the Scissors model,
VC H 2
∆ P ,Scissors = (17)
S P ,S
As this displacement must be identical to that given in equation 16, it is evident that the
relationship between the Krawinkler spring and the Scissors spring is as follows:
S P ,K
S P ,S = (18)
(1 − α − β ) 2
Similarly, when the moment in the Krawinkler spring is VPβH, the moment in the Scissors
spring is VCH. Using equations 3 and 9
M P , K = VP βH = VC H (1 − α − β ) (19)
M P,K
M P,S = (20)
(1 − α − β )
M F ,K
M F ,S = (22)
(1 − α − β )
Rotational Spring
For Panel Shear Rigid
Link
As an example, consider the case where α and β are 0.1 and 0.2, respectively, the Scissor
spring must be approximately twice as stiff and 1.43 times stronger than the Krawinkler
spring. Many analysts erroneously use the springs derived for the Krawinkler model in the
Scissors model. This will produce models that are more flexible than the true structure, and
that prematurely yield in the panel zone regions.
One should note from equations 18 and 20 that while the properties of the Scissors models
are dependent on the quantities α and β, those of the Krawinkler model are not. Since it
was explicitly assumed that the columns and girders on both sides of the joint are of equal
height and span, and these terms are reflected in α and β, the Scissors model may not be
used when this condition is violated. There is no such restriction on the use of the
Krawinkler model.
The deformed shape of the Krawinkler and Scissors models are shown in figure 5. In this
figure all of the deformation is assumed to be in the panel, with the girder and column rigid.
The most striking difference in the behavior between the two models is the offset in the
centrelines of the columns and girders in the Krawinkler model, which are not present in the
Scissors model.
A series of analyses were carried out using DRAIN-2DX (7) to determine the effect of the
kinematic differences on the pushover response of a series of assemblages and planar
frames which had yielding in the panel zone and at the ends of the girders. A variety of
girder spans were used, but the column height remained constant. Analysis was performed
with and without gravity load, and with and without P-Delta effects. For simple
subassemblages analyzed using the Krawinkler and the Scissors models, the pushover
responses were identical. For structures created by assembling subassemblages into a
rectilinear frame, but with real hinges at the midspan of the girders and midheight of the
columns, the pushover responses were again identical. Minor differences in the pushover
responses were obtained when the midspan/midheight hinges were removed. It was
Offsets
To evaluate the effectiveness of the elastic modeling techniques developed above, a series
of comparisons was performed using test results provided by Ricles (5) from Phase II of the
SAC Steel Project. The properties used for one of the test specimens are provided in Table
1. Additionally, the SAC subassemblage was modeled using ABAQUS. The ABAQUS
model of the subassemblage is shown in figure 6.
CONCLUSIONS
Simplified mechanical models such as the Krawinkler model and the Scissors model are
extremely effective in representing both elastic and inelastic panel zone deformations in
steel frame structures. While the Krawinkler model is considerably more complex and has
significant kinematic differences with the Scissors model, the results obtained using the two
models are essentially identical. The results from the simple mechanical models also
correlate well with more advanced finite element analysis, and with experimental results.
The Scissors model, however, is limited to use in frames with equal bays widths and equal
story heights. The Krawinkler model has no such restriction. Very significant errors will
occur in the analysis if the properties derived for the Krawinkler model are used in a
Scissors model. Such inconsistent use of the models was found in several references
reviewed by the authors.
NOTATION
REFERENCES
(1) Downs, William M, (2002). Modeling and Behavior of the Beam/Column Joint Region
of Steel Moment Resisting Frames, M.S.Thesis, Department of Civil and Environmental
Engineering, Virginia Tech, Blacksburg, Virginia.
(2) Krawinkler, H., (1978), “Shear in Beam-Column Joints in Seismic Design of Frames”,
Engineering Journal, v15, n3, American Institute of Steel Construction, Chicago,
Illinois.
(3) Hibbit, Karlson, and Sorensen, (2001). ABAQUS User’s Manual, Verson 6.2.
(4) FEMA (2000). Recommended Seismic Design Criteria for New Steel Moment Frame
Buildings, FEMA-350. Federal Emergency Management Agency, Washington D.C.
(5) Ricles, J. M., (2002). “Inelastic Cyclic Testing of Welded Unreinforced Moment
Connections” Journal of Structural Engineering, ASCE, v128, n4.
(6) Charney, Finley A., (1993). "Economy of Steel Frame Buildings Through Identification
of Structural Behavior", Proceedings of the Spring 1993 AISC Steel Construction
Conference, Orlando, Florida.
(7) Prakesh, V. and Powell, G. H., (1993). DRAIN 2D-X Users Guide, University of
California, Berkeley, California.
ABSTRACT
The paper describes the reasons that cause rotation capacity of moment
connections strongly dependent on material properties of components of such
connections. Then a statistical evaluation of the rotation capacity of typical
moment connections with respect to the variation of the material properties of
individual components was performed. The Monte Carlo method was used and
10,000 calculations were performed in each analysis. Log-normal distribution
was assumed for yield strength fy and ultimate strength fu with the range of
typical mean values and coefficients of variation from literature. The results
show that considerable variation in rotation capacity is obtained in some cases.
INTRODUCTION
Besides strength and stiffness, rotation capacity is also very important characteristic of
moment connections of steel structures. Until recently there have been no general methods
for the determination of rotation capacity available, except tests and sophisticated FE
analysis. Also EN 1993-1-8 (1), a very detailed standard on connections in steel structures,
only briefly addresses this topic by giving some requirements to achieve sufficient ductility.
For these reasons it is not surprising that several research groups are working on this
problem (2, 3, 4, 5) and the authors of this paper together with I. Vayas have recently
developed an analytical method that enables the assessment of rotation capacity of moment
connections based on a component approach from EN 1993-1-8 (6, 7). From test results and
extensive numerical simulations for each relevant component, deformation capacities and
simplified bilinear force-displacement relations were established. By suitable mechanical
model the components were assembled to model connection behaviour and to determine the
rotation capacity.
Studying the problem and applying the developed method, it became evident that rotation
capacity is very sensitive to the changes of basic material properties, yield stress fy and
tensile strength fu, of individual components. It was observed that nominal values of fy and fu
that are usually known at the design stage do not always give satisfactory results, as the real
values are usually higher and most likely also some differences between individual
components take place.
The rotation capacity of the joint is governed by its weakest component. When the weakest
component reaches its ultimate resistance, also the resistance of the joint is exhausted. This
is exactly true for a component method, but also in reality (tests, FE simulations) there is very
limited possibility for shedding of forces between different parts of the connection, and the
resistance of the connection usually drops after the resistance of the critical part is reached.
One of the most important issues when determining the rotation capacity is the strength of
components and especially relative difference in strength between components. This is
demonstrated in Figure 2. For the sake of simplicity the connection is composed of only two
components, represented by bilinear inelastic force-displacement diagrams, one in tension
and one in compression (Figure 2a).
Such scenario can easily happen in reality. At the design stage only characteristic values of fy
and fu are known and the rotation capacity is determined based on these values (case in
Figure 2b). Real values of fy and fu are higher than the characteristic values, as characteristic
values are more or less lower guaranteed values. For grade S235 the increase of strength
goes up to 30% and for S355 the increase is somewhat smaller. The case in Figure 2c can
be regarded as a situation, where the resistance of component C1 was increased due to
differences between characteristic and real material properties, while the resistance of
component C2 remained almost unchanged. The result is much smaller rotation capacity. In
this case the actual behaviour is on the unsafe side compared to the design values.
Certainly, it is possible to find the opposite situation, where the real behaviour is on the safe
side. The main conclusion is that the rotation capacity is very sensitive to the variability of
material strength parameter. This is not a specific problem of a component method, but is
immanent to the behaviour of joints when ductility is considered.
Statistical analysis of typical moment connections was performed to assess how the variation
of basic material properties fy and fu of individual components influences the rotation capacity
of a connection. For basic random variables yield strength of column web, column flange,
end-plate and tensile strength of bolts was selected. Log-normal distribution was applied to
all random variables with the following four combinations of mean my and coefficient variation
vy:
• A – my = 1.18 fy, vy = 0.08
• B – my = 1.18 fy, vy = 0.10
• C – my = 1.12 fy, vy = 0.05
• D – my = 1.12 fy, vy = 0.08
where fy is a nominal value from EN 1993-1-1.
Different authors (8, 9, 10) also propose log-normal distribution and case A as the most
suitable for structural steel. Case B gives larger scatter and cases C and D smaller scatter of
random variables than case A. For bolts of grade 8.8 in all calculations case C was taken into
account to allow for more favourable statistical distribution of tensile strength in this case. In
addition to the analysis with all four random variables, for case A also the Monte-Carlo
simulation with only one active random variable at a time was performed. Other parameters
were kept constant at their nominal (characteristic values).
DATA ON CONNECTIONS
Five different end-plate moment beam-to-column connections were analysed. They were
designed for the purpose of statistical analysis, each with different decisive component with
the smallest resistance. At the fifth connection all components have approximately equal
resistance. The relevant data on connections are given in Table 1 and Figure 3.
For all five connections rotation capacity ϕu was determined at nominal values of material
parameters with two methods: FE simulation (ABAQUS (11), solid finite elements) and own
analytical method (6), where the resistance of components was determined in two ways –
according to EN1993-1-8 and by replacing fy with fu when relevant (6) to get more accurate
resistances. Comparison to numerical simulations (case c in Table 2) shows very good
results when more realistic resistance of components is used – case b in Table 2 (very good
agreement in three cases and conservative results in two cases), but some unsafe results
when resistances from EN1993-1-8 were used (case a in Table 2).
Criteria for reaching ultimate rotation ϕu were stated by maximum equivalent plastic strains
reached in different parts of connection: 0.1 for column web in tension, compression on
shear and 0.2 for column flange and end-plate in bending.
Due to the limited number of pages of this paper it is not possible to present all the results
obtained from the analysis. For connection 1 more results are given and for other
connections only some results for case A of statistical parameters are presented (only for
case a in Table 2).
250 SWC
200 CWC
150 TWC
EP
100
CF
50
d ,f
2.5 5 7.5 10 12.5 15
Connection 1
From moment-displacement diagram (Figure 4) it is clear that column flange evidently has
the lowest strength and should influence the rotation capacity to a large extent. In Figures 5
to 11 the calculated values of rotation capacity are plotted in diagrams M - φ (left) and the
obtained distribution function of the rotation capacity is plotted at the right side. From these
results the following observations can be made:
- The results for the ultimate capacity are in the range of 0.05 – 0.138 rad with the most
probable result around 0.057 rad (0.055 rad at nominal values).
- Comparison of results for cases A, B and C shows that the results are very similar, but
clearly reflect the statistical input parameters. Case B (Figure 6) gives the worst results
and case C (Figure 7) gives the most favourable results.
- From Figures 9 – 11 it is evident that only the change of material properties of column
flange strongly influences the rotation capacity, while other components have no
influence. This is expected because their resistance is much larger than the resistance of
column flange, and therefore they do not contribute significantly to the rotation capacity of
the connection.
2500
100
80 2000
60 1500
40 1000
Monte Carlo
20 EC3 - 1- 8 500
Abaqus
f Cd f Cd
0.025 0.05 0.075 0.1 0.125 0.15 0.025 0.05 0.075 0.1 0.125 0.15
MRd n
80 3000
2500
60
2000
40 1500
1000
20 Monte Carlo
EC3 - 1- 8 500
f Cd f Cd
0.025 0.05 0.075 0.1 0.125 0.15 0.025 0.05 0.075 0.1 0.125 0.15
MRd n
70 4000
60
3000
50
40
2000
30
20 Monte Carlo 1000
10 EC3 - 1- 8
f Cd f Cd
0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.02 0.04 0.06 0.08 0.1 0.12 0.14
MRd n
7000
50 6000
40 5000
30 4000
3000
20
Monte Carlo 2000
10 EC3 - 1- 8
1000
f Cd f Cd
0.02 0.04 0.06 0.08 0.02 0.04 0.06 0.08
MRd n
50
8000
40
6000
30
4000
20
Monte Carlo
10 EC3 - 1- 8 2000
f Cd f Cd
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.01 0.02 0.03 0.04 0.05 0.06 0.07
MRd n
50
8000
40
6000
30
4000
20
Monte Carlo
10 EC3 - 1- 8 2000
f Cd f Cd
0.01 0.02 0.03 0.04 0.05 0.06 0.01 0.02 0.03 0.04 0.05 0.06
MRd n
200
1500
150 1250
1000
100
750
Monte Carlo
50 EC3 - 1- 8 500
Abaqus
250
f Cd f Cd
0.025 0.05 0.075 0.1 0.125 0.15 0.025 0.05 0.075 0.1 0.125 0.15 0.175
125 1000
100 800
75 600
Monte Carlo
50 400
EC3 - 1- 8
25 Abaqus 200
f Cd f Cd
0.05 0.1 0.15 0.2 0.05 0.1 0.15 0.2
MRd n
2000
100
80 1500
60
1000
40 Monte Carlo
EC3 - 1- 8 500
20 Abaqus
f Cd f Cd
0.05 0.1 0.15 0.2 0.05 0.1 0.15 0.2
MRd n
2500
120
100 2000
80 1500
60 EC3 - 1- 8
SS tlak 1000
40 PS upogib
ÈP upogib 500
20 Abaqus
f Cd f Cd
0.05 0.1 0.15 0.2 0.05 0.1 0.15 0.2
Connections 2 to 5
At connection 3 column web in shear is decisive and the results are similar as at connection
1.
Also at connection 4 one component, end-plate, has evidently lower resistance and the
influence of this component is large, while other components have little influence.
CONCLUSIONS
Obviously, statistical analysis can give more information about realistic behaviour in such
conditions. For our own analytical method for the calculation of the rotation capacity we
performed statistical evaluation, where material properties of individual components were
random variables.
These results show that the deterministic approach to the calculation of the rotation capacity
based on nominal values of material properties may give unsatisfactory results. A possible
solution could be statistical approach. Based on this approach reduction factors could be
determined to be used in the simple deterministic calculation of the rotation capacity.
REFERENCES
(1) Eurocode 3: Design of steel structures, Part 1.8: Design of joints, prEN 1993-1-8: 2000.
(2) Coelho A., Bijlaard F., da Silva L. On the deformation capacity of beam-to-column
bolted connection. Report AG-XXIII, ECCS-T10, April 2002, Ljubljana. 2002.
(3) Huber G., Tschemmernegg F. Modeling of beam-to-column joints. Journal of
Constructional Steel Research, Vol. 45, 199-21. 1998.
(4) da Silva L. S., Coelho A. M. G. A ductility model for steel connections. Journal of
Constructional Steel Research, Vol. 57, 45-70. 2001.
(5) Kuhlmann U., Sedlacek G., Kuhnemund F., Stangenberg H. Mitteilungen: Vorhandene
ABSTRACT
Actual design codes and recommendations for ductile Moment Resisting Steel
Frames in seismic zones require to beam-to-column joints to sustain a total
rotation of 0.035 rad (in Europe) to 0.040 rad (in United States). The problem is
that in codes, there are no analytical methods to predict the rotation capacity of
MR connections, not even for monotonic loading, which is simpler then
compared with seismic action of cyclic type. Even if last years appeared some
proposals for analytical or numerical based methods to evaluate rotation
procedure for their qualification in terms of ductility, still remain the pre-
qualification tests, both for static and seismic actions.
In the present paper, based on extended experimental studies, comparatively
conducted under monotonic and cyclic loading on different types of
connections, both for steel and composite MR frames, a reduction factor to be
applied to monotonic rotation capacity is proposed in the purpose of evaluation
of cyclic rotation capacity.
INTRODUCTION
Modern design codes characterise the behaviour of a connection by three basic parameters
i.e. strength, stiffness and ductility. In MR connection the ductility is measured in terms of
rotation capacity. From this point of view, connections may be classified, similar to the
classification of sections (EN 1993-1-1 [1]) in terms of their ductility (EN 1993-1-8 [2]) as
follows:
Class 1 joints: Ductile joints.
A ductile joint is able to develop its plastic moment resistance and
to exhibit a sufficiently large rotation capacity.
Class 2 joints: Joints of intermediate ductility.
A joint of intermediate ductility is able to develop its plastic moment
resistance but exhibits only a limited rotation capacity once this
resistance is reached.
Class 3 joints: Non ductile joints.
Premature failure (due to instability or to brittle failure of one of the
joint components) occurs within the joint before the moment
resistance based on a full plastic redistribution of the internal forces
is reached.
The parameters of this classification are shown in Figure 1 (EN 1993-1-8 [2]).
* at present Assoc. Prof. to INSA Rennes France
For welded beam-to-column connection, in which the web is stiffened in compression, but
unstiffened in tension, under monotonic load only, EN 1993-1-8 [2] provide the following
formula to estimate the design rotation capacity, ΦCd:
Φ Cd = 0.025 h c h b (1)
where hb is the depth of the beam and hc is the depth of the column.
The deformation capacity of components has been studied by several researchers. Faella at
al. [19] carried out tests on T/stubs and derived analytical expression for the deformation
capacity of this component. Kuhlmann and Kuhnemund [22] performed tests on the column
web subjected to transverse compression at different levels of compression axial force in the
column. Some authors have tried to extract the information of the behaviour of single
components from the tests on a whole joint. Bose et al. [14,15] determined only the strength
of the most important components, while da Silva et al. [23] tried to determine all three
important parameters, stiffness, strength and deformation capacity, at different levels of axial
force in a beam.
Recently, Beg et al. [12] proposed a simple analytical method for calculation of rotation
capacity of MR connection which is compatible with the “component method” used in EN
1993-1-8 to provide the stiffness and strength.
Also Gioncu [20] suggested to use the “local mechanism method” to evaluate the rotation
capacity of top-and-seat cleat bolted MR connections. This method could be used for welded
connections, but probably not for extended-end-plate bolted connections, or other types.
However, even, as shown, there are tentatives to find analytical methods for evaluation of
monotonic rotation capacity, some of these referred in this paper being promising, the basic
procedure provided by design codes on this purpose still remain the prequalification tests.
In case of seismic resistant connection the problem is much more complex, and if the
relevant design codes have specific provisions to request precisely the lower bound values
for the necessary rotation capacity, they do not provide a specific evaluation methodology,
except, again testing.
EN 1998-1 [3] has introduced three levels of structural ductility class in connection with
design concepts and range of reference values of the behaviour factors as presented in
Table 1.
Dissipative semi-rigid and/or partial strength connections are permitted, provided that all of
the following conditions are satisfied: a) the connections have a rotation capacity consistent
with the deformations; b) members framing into the connections are demonstrated to be
stable at ultimate limit state (ULS); c) the effect of connections deformation on global drift is
taken into account using non linear static (pushover) global analysis or non linear time history
analysis.
The overstrength condition for connections need not apply if the connections are designed in
a manner enabling them to contribute significantly to the energy dissipation necessary to
achieve the chosen q-factor.
The moment frame connections design should be such that the plastic rotation capacity θp in
the plastic hinge is not less than 35 mrad for structures of ductility class DCH and 25 mrad
for structures of ductility class DCM with q>2.
Even with the limits mentioned above for the rotation capacity of joints, EN 1998-1 does not
specify any formula for this evaluation except testing and design experience.
According to American codes UBC-97 [5] or AISC-2002 [6], the purpose of the earthquake
provisions is primarily to safeguard buildings against major structural failures and loss of life,
not to limit damage or maintain function.
Structures and portions thereof shall, as a minimum, be designed and constructed to resist
the effects of seismic ground motions.
Moment Frame (MF) defined as a building frame system in which seismic shear forces are
resisted by shear and flexure in members and connections of the frame are divided as
follows (AISC-2002): Special Moment Frame (SMF), Intermediate Moment Frame (IMF) and
Ordinary Moment Frame (OMF).
Even if the rotation capacity of the beam-to-column joints is connected with the classification
of frames, the AISC code is not providing any formula for the evaluation of this very important
characteristic.
Anyhow, American FEMA 350 document [8], in Chapter 3: Connection Qualification provides
pre-qualification data and design procedures for alternative types of welded, fully restrained,
The Japanese code AIJLSD-90 [7] provides specifications concerning only the frame
classification, according to the classification of members and member cross-sections, storey-
drift and overstrength of joints in comparison with the connected members. Any specification
concerning the rotation capacity of beam-to-column joints is not made.
In this chapter four different experimental studies aiming to evaluate performances of both
steel and composite beam-to-column joints under cyclic loading will be summarised.
Aribert & Grecea [11] have developed an experimental research program at INSA Rennes.
This research program dealt with 8 beam-to-column welded joints under monotonic and
repeated cyclic loading. The specimens were major axis joints with a symmetrical cruciform
arrangement comprising an H or I column connected to two cantilever beams. No transverse
stiffener was welded in the compression zone of the column web.
The tests were performed according to the Recommended Testing Procedure of ECCS
(1985). Each type of joint was subject first of all to a monotonic loading (CPP11, CPP13,
CPP15 and CPP17, then to cyclic reversal loading (CPP12, CPP14, CPP16 and CPP18).
The cyclic moment-rotation curves for the four tested joints are presented in Figure 2.
Experimental values of ultimate resistance moment and rotation capacity obtained in both
monotonic and cyclic reversal loading can be compared in Table 2.
From the moment-rotation curves, it is observed that the ultimate moment and the initial
stiffness of the joints are not strongly influenced by the repeated cyclic loading, so that in
seismic design the corresponding formulae given in EN 1993-1-8 for the case of static
loading can be used, as reasonable approximations. On the other hand it appears clearly
that the rotation capacity of the joints is systematically reduced by a factor about 2.
200 150
100
100
50
Φ [rad] Φ [rad]
0 0
-0,03 -0,02 -0,01 0 0,01 0,02 0,03 0,04 -0,04 -0,03 -0,02 -0,01 0 0,01 0,02 0,03 0,04
-50
-100
-100
-200
-150
-300 -200
100 200
Φ [rad] Φ [rad]
0 0
-0,02 -0,01 0 0,01 0,02 0,03
-100 -0,05 -0,04 -0,03 -0,02 -0,01 0 0,01 0,02
-200
-200
-400
-300
-400 -600
Table 2. Comparison between joint characteristics under cyclic and monotonic loadings.
CPP11 CPP12 CPP13 CPP14 CPP15 CPP16 CPP17 CPP18
Mu [kNm] 230.0 253.0 166.9 180.3 349.2 368.5 467.5 486.2
Φu [rad] 0.064 0.031 0.045 0.023 0.045 0.020 0.052 0.030
Three typologies of beam-to-column joints have been tested from a total of 12 specimens. All
the joints are double sided. For all the specimens the design steel grade was S235 (fy=235
N/mm2, fu=360 N/mm2), beams being IPE 360 and columns HEB 300.
Two types of loading were applied: symmetrical and anti-symmetrical and three connection
typologies were tested (Figure 3).
10M20 gr 10.9 Equal strength weld 3M20 gr6.6 Supplementary web plate
200
MOMENT AT THE
150
MOMENT AT THE
100
50
0
-0.06 -0.04 -0.02 -50 0 0.02 0.04 0.06
-0.06 -0.04 -0.02 0 0.02 0.04 0.06
-100
-150
-200
-250
-300
-350
-400 TO TAL JOINT RO TATIO N [rad]
TOTAL JOINT ROTATION [Rad]
150
MOMENT AT THE
MOMENT AT THE
2 00
1 00 50
0
-0 .0 6 -0 .0 4 -0 .0 2 -5 0 0 0 .0 2 0 .0 4 0 .0 6
-0.0 15 -0 .0 1 -0.0 05 0 0.0 05 0 .0 1 0.0 15 0 .0 2
-1 00
-1 5 0
-2 00
-3 00 -2 5 0
-4 00
-3 5 0
-5 00
T O T A L J O IN T R O T A T IO N [ra d ]
TO T A L J O IN T R O T A TIO N [ra d .]
X S -C W P 1 X U -C W P 1 250
400
COLUMN FACE [kNm]
COLUMN FACE [kNm]
150
MOMENT AT THE
MOMENT AT THE
200
50
0
-0 .0 4 -0 .0 2 0 0 .0 2 0 .0 4 0 .0 6 -0 .0 7 -0 .0 5 -0 .0 3 -510
-0 .0 0 .0 1 0 .0 3 0 .0 5 0 .0
-2 0 0 -1 5 0
-4 0 0 -2 5 0
-3 5 0
-6 0 0
T O T A L J O IT R O T A T IO N [ra d .] T O T A L J O IN T R O T A T IO N [ra d ]
The joints (three different joint configurations and two load types) have been designed
according to EN 1993-1-8. The loading history was made according to the ECCS
Recommendations simplified procedure. The cyclic testing curves are presented in Figure 4.
Comparing the experimental and computed values of joint moment capacity, it can be
observed that generally, close values are obtained for the XS series. In the case of XU
series, all experimental values are lower than the ones computed by EN 1993-1-8.
In what concerns the initial stiffness of the joints, numerical and experimental results agree
fairly well for the XU series, while significant differences are noticed for XS series. Anyway,
stiffness is much lower for the anti-symmetrical joints both from experimental and computed
stiffness values. This fact is again given by the deformability of the panel zone.
Joint rotations are considerably higher for XU series, generally in the ratio of 1:2. Improved
ductility in the case of XU joints is given by good rotation capacity and stable hysteresis
loops of the web panel in shear. Anti-symmetrical joints have generally increased energy
dissipation capacity with respect to the symmetrical ones.
Rotation capacity for bolted steel and composite joints (UP Timisoara)
Column
Column 8 top & bot. fl.
r.c. slab
8M20 gr.10.9 Beam
120
R20 R20
290
290
Beam
12
12
5 .
5 . 6M20 gr.10.9
8 top & bot. fl.
170
2Ø12 PC52
14
600
3Ø10 PC52
2Ø12 PC52
20 360 20
(a) (b)
Figure 5. Connection configurations for (a) BX-S series and (b) BX-C series of joints.
The testing program comprised six specimens: three joints under symmetrical loading (BX-
SS and BX-CS), and three joints under anti-symmetrical loading (BX-SU and BX-CU). Tests
were performed in accordance with the ECCS Recommendations complete procedure
(ECCS 1985). The first specimen from each series was tested monotonically, The load was
applied quasi-statically, under displacement control.
Figures 6, and 7 synthetically present the main experimental results expressed in terms of
maximum plastic rotation ϕ max and maximum moment attained M max , for steel and
composite joints. Tables 4 and 5 present the joint characteristics obtained from the tests and
computed analytically in accordance to EN 1993-1-8 and EN1994-1-1 section 8, respectively.
MOMENT AT THE
100 100
0 0
-0.1 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1 -0.1 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1
-100 -100
-200 -200
-300 -300
-400 -400
TOTAL JOINT ROTATION [rad] TOTAL JOINT ROTATION [rad]
(a) (b)
Figure 6. Moment - rotation relationships for cyclic specimens of the BX-S series.
350 350
BX-CS-C1 250
BX-CU-C1 250
COLUMN FACE [kNm]
150 150
MOMENT AT THE
MOMENT AT THE
50 50
-150 -150
-250 -250
-350 -350
TOTAL JOINT ROTATION [rad] TOTAL JOINT ROTATION [rad]
(a) (b)
Figure 7. Moment - rotation relationships for cyclic specimens of the BX-C series.
FULL SHEAR
SUPPLEMENTARY
AREA shear area WEB PLATE
shear area
(a) (b)
Figure 8. (a) Full shear approach and (b) EN 1993-1-8 approach.
Table 5. Comparison of test to the analytical results (EN 1994-1-1) for composite joints.
Total Energy φmax+ φmax- Mmax Mmin Sj,ini+ Sj,ini- φy+ φy- My+ My-
3
Specimen kNm rad mrad kNm x10 KNm/rad mrad KNm
Symmetrically loaded joints
EC4 Mom - --- --- --- 57.11 2.34 133.80
EC3* Mom + --- --- --- 99.68 2.30 230
BX-CS-M1 16.60 86.7 244.25 98.47 1.27 155.51
BX-CS-M2 8.4 32.7 316.09 105.27 2.52 231.32
BX-CS-C1 32.0 44.8 14.1 305.5 196.9 102.5 75.05 1.60 1.73 195.2 149.9
Anti-symmetrically loaded joints
EC4-comb.** --- --- --- 41.78 3.84 160.55
BX-CU-M 15.2 90.1 198.24 47.19 2.34 126.23
BX-CU-C1 104.5 60.5 53.8 187.1 193.3 36.87 37.92 3.49 3.38 142.67 137.3
BX-CU-C2 32.2 32.7 30.9 191.3 210.5 -- -- -- -- -- --
EC3* - computed according to EN 1993-1-8, by translation of the centre of compression
EC4-comb.** - mean value between the positive and negative values
The EN 1994-1-1 section 8 does not give the possibility of computing composite joints
subjected to positive moments. For this case, the values of stiffness and moment resistance
presented in Table 5 are computed according to EN 1993-1-8 by a translation of the centre of
compression from the upper beam flange to the middle of the concrete slab (considered
without corrugated sheet). These assumptions lead to comparable values to the tests in
terms of resistance and stiffness. For the case of cyclic loading, there can be observed a
decrease in both resisting moment and stiffness due to rapid slab degradation.
Analytical values of moment resistance and stiffness for anti-symmetrical loading have been
obtained by the mean values for the two connections subjected to anti-symmetrical loading,
taking into account the full shear area approach. This prediction remains only an attempt of
computing the composite joints under anti-symmetrical loading.
Ciutina [16] has tested some composite connections at INSA Rennes (Figure 9) under
monotonic and cyclic loads. G14 and G15 are composite joints under monotonic and cyclic
loading, and G17 and G18 are composite joints with haunch under monotonic and cyclic
loading.
Column Column
Beam Beam
M M
Dissipative zone
Dissipative zone
(connection)
Haunch (beam)
Tests were performed in accordance with the ECCS Recommendations complete procedure
(ECCS 1985).
Obtained results concerning the joint characteristics are presented in Figure 10 and Table 6
for test specimens G14-G15 and Figure 11 and Table 7 for G17-G18 respectively.
In this case of tests, it is quite clear again that the rotation capacity of joints subjected to
cyclic loading could be estimated as approximately 0.5 of the rotation capacity of joints
subjected to monotonic loads.
400
Rotation Globale
Global rotation
300 Moment [kNm]
200
100
0
-0.1 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04
-100
Rotation [rad.]
-200
-300 Essai
Test G15
Essai
Test G14
-400
Figure 10. Comparison between tests G14 and G15.
Moment [kNm]
Rotation Globale
Global rotation 600
400
Essai
Test G18 200
Essai
Test G17
0
-180 -160 -140 -120 -100 -80 -60 -40 -20 0 20 40 60 80
-200
-400
-600
Rotation [mrad.]
-800
Figure 11. Comparison between tests G17 and G18.
A small difference should be remarked for G14 and G15, but this is due to the test conditions
and the fact that the test could not be developed symmetrically. Anyhow, it can be seen that
the sum of the negative and positive rotation is practically identical with the rotation capacity
of the joint subjected to monotonic loads (24+68=92≈91.8 mrad).
In the second case, for G17 and G18, where negative and positive rotations are similar, it
can be seen that 40+32=72≈88 mrad.
CONCLUSIONS
It is clear for all specialists of the field that the rotation capacity of beam-to column joints is
one of the main characteristics which are influencing the seismic behaviour of the steel MR
frames. That is the reason why in the last period, this characteristic has been introduced with
some arbitrary values of reference for different types of frames. Unfortunately, formulae for
evaluating this characteristic are very few. It is evident that the researches have to be
continued in both directions experimental test research and numerical modelling, to establish
new definitions for the evaluation of the rotation capacity of beam-to-column joints.
Rotation capacity of joints is very difficult to establish except of test procedures. From
analytical point of view, the evaluation in static range should be possible, using the method of
components. Contrary, in cyclic range, Zandonini and Bursi [26] have shown that the use of
this method is impossible and have recommended the use of some macro-components,
incorporating some of the interaction effects among elemental components. This method
should be more adequate, because some macrocomponents are ductile ensuring the ductile
behaviour and the rotation capacity, and some others are fragile, needing to ensure them a
certain level of overstrength, in order to ensure the development of the rotation capacity.
Anyhow the ductile components of the joint are the endplate and the column web panel. After
An important observation in our opinion should be that taking into account the complex
behaviour of the moment resisting joints, and the conclusions of Stojadinovici [24] and
Zandonini and Bursi [26], concerning the pre-qualification and the introduction of
macrocomponents, under seismic loads, it is more adequate to refer to rotation capacity of
the joint using the concept of “macrocomponent method”, instead of rotation capacity of the
connection using the concept of “component method” utilised for monotonic loads.
So, what could we do in order to establish the rotation capacity of the joints, without
experimental cyclic tests? Analysing several tested joints under monotonic and cyclic loads,
the authors of this paper have observed that the rotations under cyclic and monotonic loads
are usually in the ratio of 1:2. Using this conclusion, the answer would be very simple.
For the evaluation of the rotation capacity under cyclic loads, we could use an analytical
method for evaluation under monotonic loads, or the well-known computer code DUCTROT
of Gioncu and then affect the results with the correction coefficient for cyclic loads equal to
0.5.
For the same evaluation, the correction coefficient of 0.5 could be used also to affect the
values of rotation capacity of joints determined under monotonic loads. This proposal is
made in order to use the large number of results, existing in literature, thinking also to some
data bank like SERICON.
From the same analytical point of view, we are reminding the importance of the
macrocomponents method to be developed and used after some more numerical
simulations.
It is important also to remind the prequalification connections what FEMA 350 have made, in
order to make a similar prequalification in Europe, with the usual connections utilised by
designers in Europe.
REFERENCES
1. EN 1993-1-1. 2003. Design of steel structures. Part 1-1: General rules and rules for
buildings. European standard.
2. EN 1993-1-8. 2003. Design of steel structures. Part 1-8: Design of joints. European
standard.
3. EN 1998-1. 2003. Design of structures for earthquake resistance. Part 1: General rules,
seismic actions and rules for buildings. European standard
4. EN 1994-1-1. 2003. EUROCODE 4: Part 1.1. General rules and rules for buildings;
Section 8: Composite joints in frames for buildings. European standard
5. Uniform Building Code, Volume 2, Structural Engineering Design Provisions. International
Conference of Building Officials, Whittier, California, USA, 1997.
6. Seismic Provisions for Structural Steel Buildings. American Institute of Steel
Construction, Inc. Chicago, Illinois, USA, 2002.
7. Standard for Limit State Design of Steel Structures. Architectural Institute of Japan, 1990.
8. FEMA 350. 2000. Recommended Seismic Design Criteria for New Steel Moment-Frame
Buildings, SAC Joint Venture.
ABSTRACT
In recent years, the pursuit of increased competitiveness of steel construction
has led to the quest for simpler and more economical joint details. The under-
lying concept of semi-continuity requires that adequate rotation capacity is
available, that is very much dependent on a good balance between ductile and
brittle components. In this paper, a probabilistic approach based on the Monte
Carlo method is implemented to investigate the sensitivity of end-plate beam-
to-column joints to (i) the differences between nominal and actual material
properties of steel and (ii) the difference between actual moment-rotation be-
haviour and the elastic-perfectly plastic response usually assumed in terms of
codes of practice.
INTRODUCTION
The rotation capacity of steel joints constitutes an essential property for the safe response of
steel structures. In non-seismic regions where gravity loading usually controls the behaviour,
the shift to semi-continuous design means that available rotation capacity must safely ex-
ceed the rotation demand. In seismic regions, this becomes an even higher requirement,
coupled with sufficient energy dissipation.
Current design procedures for steel joints do not include explicit evaluation of rotation capac-
ity. In particular, the Eurocode 3 provisions for rotation capacity (1) do not establish any spe-
cific procedures for the evaluation of the rotation capacity of bolted or welded joints. How-
ever, they state the need to ensure adequate rotation capacity either by testing in accor-
dance with EN 1990 (2) or alternatively, using appropriate calculation models based on the
results of tests. Additionally, it provides sufficient conditions that ensure adequate rotation
capacity (3, 4).
For end-plate beam-to-column steel joints, these conditions basically impose that either the
column web panel in shear controls the behaviour of the joint or, alternatively, the end-plate
or the column flange in bending and are reproduced below:
d < 69ε (1)
tw
t < 0.36d fub f y (2)
where d denotes the bolt diameter in the unthreaded part of the shank, tw is the thickness of
the column web, t is the thickness of the end-plate or the column flange, fub is the ultimate
tensile strength of the bolt, fy represents the yield stress of the plate and
ε= 235
fy (3)
Given that small variations of the post-limit stiffness (in particular for the critical component)
result in large variations of the maximum rotation of the joint (12), it is the aim of this paper to
assess the influence of the various component properties (post-limit stiffness, yield force and
failure deformation) on the available ductility of the joint.
General description
The Monte Carlo simulation technique (13), applied to steel joints (12) consists of the systematic
solution of the relevant component model, carried out in a deterministic way, to yield a locus of
statistically valid solutions that provide a probabilistic description of the joint response. To
achieve this objective, one of the main tasks in a simulation procedure is the generation of ran-
dom numbers from prescribed probability distributions. This can be achieved using a random
generator available in any computer according to the following procedure: (i) generation of a uni-
formly distributed random number between 0 and 1, and (ii) application of appropriate transfor-
mations in order to obtain the corresponding random number with the specified probability distri-
bution. The Box and Muller Method was adopted in this paper (13), complemented with adequate
checking of both the independence and the randomness of the generated numbers. Finally, it is
noted that many simulations are usually necessary to obtain results with acceptable precision, al-
though resource to sample reduction techniques such as the latin hypercube may be used to limit
the number of simulations.
Computational implementation
(iii) For each simulation, the mr values for each random variable are combined, resulting
in a sample containing m-dimensional values, where
r
m = ∏mj (4)
j=1
(iv) Using the appropriate mechanical model for the chosen joint configuration and the
program NasCON, m moment-rotation curves are obtained for each simulation (Step
4).
(v) Steps (ii) to (iv) are repeated to generate s independent simulations.
(vi) Finally, direct probabilistic assessment is carried out on the s available simulations
for the desired indicators (failure rotation or sequence of yielding of the various
components, for example) – Steps 5 and 6.
Although steel is among the most uniform construction materials, and is produced with strin-
gent quality control procedures, the production requirements as specified by the relevant
standards of steel production only demand that minimum values for yield and ultimate stress
are guaranteed. Thus, in practice, all structural steels always present higher yield stress than
the nominal values, usually with a significant scatter between the nominal value and the ac-
tual mean or characteristic yield stress (15). This over-strength has always been regarded as
beneficial because most emphasis in the design of joints was on strength. As ductility be-
comes an issue, excessive over-strength may result in unforeseen behaviours and must be
assessed carefully.
The force-displacement relation for the various components is frequently simplified as a bi-
linear curve, characterized by four independent properties: initial stiffness (Ke), resistance
(FY), post-limit stiffness (Kp) and collapse displacement (∆f). All these properties depend on
the mechanical properties of steel, the major dependencies being highlighted in the following:
Ke= Ke(E), FY= FY(E, fy), Kp= Kp(E, Ep, fy), ∆f= ∆f (E, Ep, fy, fu) (5)
where E, Ep, fy, fu are, respectively, the elastic stiffness, the plastic stiffness, the yield stress
and the ultimate stress of steel. Consequently, being a major source of randomness in the
characterization of the various components, the variability of the yield stress of steel is as-
sessed by considering either a normal distribution or a mixed distribution, obtained by a lin-
ear combination of two normal distributions (16).
Because there are no accepted estimates for the post-limit stiffness and collapse displace-
ment of most components, it was necessary to establish representative statistical estimates
for these properties. To achieve this objective, a representative set of experimentally tested
sample joints covering flush and extended end-plate connections taken from the database
SERICON (17) were selected. Subsequently, using the deterministic values of initial stiffness
and resistance obtained according to part 1-8 of EC3 and a genetic algorithm coupled with
NASCon (18), average values of Kp and ∆f were established for each component. The
adopted probability distributions for the component properties are summarized below:
(i) Initial elastic stiffness (Ke): deterministic values from EC3 (1);
(ii) Resistance (FY): given the direct dependence of component resistance on the
yield stress of steel (eq. (5)), the adopted probabilistic distribution is taken as the
product of the deterministic value (obtained according to EC 3 – part 1.8) by the
histogram of the corresponding steel grade divided by the nominal yield strength
value. The two yield stress distributions described above are used: normal or
mixed distribution.
(iii) Post-limit stiffness (Kp): a Gaussian distribution is proposed to simulate its varia-
tion. For components with limited ductility, a coefficient of variation equal to 100%
was selected (so that the generated Kp values could become negative). For com-
ponents with high ductility, a coefficient of variation equal to 50% was chosen (so
that the generated Kp values could approach zero stiffness). Finally, in case of
brittle components the post-limit stiffness, resulting from the linearization of the
behaviour law, is not considered.
(iv) Collapse displacement (∆f): given the scarce experimental evidence, a normal dis-
tribution corresponding to the range suggested in (6) is used.
Reference configuration
In order to explore a typical design condition, a flush end-plate beam-to-column joint tested
by Lima at the University of Coimbra (19) was selected. It presents a critical tension zone,
with compression zone overstrength of 143 % in nominal terms and 164 % in real terms.
Having a critical tension zone, it provides an excellent example to explore the possibility of
brittle components such as the bolts in tension having a significant probability of being
critical. Table 1 describes the flush end-plate connection. It is further noticed that, according
to EC 3, this connection does not meet the requirements for sufficient rotation capacity.
Table 1. Test details for Flush Endplate joint tested in Coimbra - FE1 (19).
Test arrangement Geometry
T-configuration, column with 2 supports, cantilever beam
Component characterization
60 60
40 40
Figure 2. Comparison between experimental and numerical results for the flush end-plate.
joint for two types of component behaviour: bilinear and smoothed multi-linear.
Fy Fy
Nominal [kN] Measured [kN]
Ratio C/T 143% 164%
Ratio S/C 108% 116%
Ratio S/T 155% 189%
Simulation schedule
Three cases were tested. Case FA is intended to explore the influence of the post-limit stiff-
ness variability on the joint overall behaviour. Case FA was sub-divided into 3 possibilities: (i)
FA.1 (simulation of the post-limit stiffness of the critical components – column web in tension,
column flange in bending and end-plate in bending - using the measured real values for steel
properties), (ii) FA.2 (same as FA.1 but adding the variability of the post-limit stiffness of the
column web panel in shear) and (iii) FA.3 (same as FA.1 but using the nominal properties of
steel). Case FB adds the variability of the resistance of the components, closely linked to the
distribution of the yield stress of steel. The chosen components now have their resistance FY
simulated following either using a Normal law (FB.1) or a Mixed distribution (FB.2). Finally,
Case FC explores the influence of the collapse displacement variation on the global behav-
iour.
Starting with an analysis of the variability of the post-limit stiffness (Case FA), it is noted that a
high percentage of the samples reach 100 mrad, failures ranging from 18.2% to 43.5%. Using
measured values of yield stress of steel instead of nominal values leads to an increase of fail-
ures of the bolts in tension (6.7% to 30.7%). The histograms of rotation at failure show the same
Reference configuration
Table 3. Test details for extended end-plate joint tested by Humer (17).
Mechanical properties
fy fu
Type of joint: extended endplate [ N/mm2 ] [ N/mm2 ]
with 2 bolts per row Column
Flange 275,9 400,50
Column orientation: strong axis Web 306,60 445,00
Profiles sizes: IPE 450, HE240 B Beam
Number, size and grade of bolts: Flange 284,60 413,00
6 bolts M24 of grade 10.9 including Web 315,60 298,00
washers. Endplate 323,00 n/a
Bolts 900,00 1000,00
ep = 65 bp = 239 ep = 65
w = 109
tbp = 12
ex = 35 lp = 84
185 p = 115
tp = 41
hp = 553
320
HEB 240R
IPE 450R
tb = 14.0
100
= 240 wb = 10.4
bb = 192
wc = 10.4
hc = 242 hb = 454
(mm)
The extended end-plate joint, described in detail in Table 3, was tested by Humer at the Uni-
versity of Innsbruck (17). This joint presents a critical shear zone, with compression zone re-
Component characterization
Table 4 reproduces the component properties both for nominal and measured values of yield
stress. Figure 3 compares the experimental results with a calculated result based on a bilin-
ear characterization of the various components and calibrated values of Kp.
5.2 - End-Plate in
462.29 635.40 2.45 E+07 2.45E+07 200
Bending
8.2 - Beam Web in
677.52 909.90 ∞ - 200
Tension
10.2 - Bolts in Tension 635.00 635.00 1.31 E+06 5.91E+06 3
Critical 567.27 688.87
1 - Column Web Panel in
Shear -416.83 -543.83 6.11 E+05 1.83E+05 200
Shear
Zone
Critical -416.83 -543.83
2 - Column Web in
-461.65 -602.31 2.46 E+06 1.23E+06 12
Transverse Compression
Compres- 7 - Beam Flange in
sion Zone -993.52 -1203.22 ∞ - 10
Compression
Critical -461.65 -602.31
Fy Fy
Nominal [kN] Measured [kN]
Ratio T/C 128% 128%
Ratio T/S 142% 142%
Ratio C/S 111% 111%
Simulation schedule
As for the flush end-plate joint, Case EA is intended to explore the influence of the post-limit
stiffness variability and its implications in the joint overall behaviour. Case EA was subdivided
into 3 possibilities: (i) EA.1 (simulation of the post-limit stiffness of the critical failure compo-
nent – column web in compression - and characterization of components using the nominal
400
M (kNm)
300
200
100 NASCon
Experimental
0
0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00
Rotation (m rad)
Figure 3. Comparison between experimental results for the extended end-plate joint and
numerical results obtained using NASCon and bilinear component behaviour.
Starting with cases EA.1 and EA.2, that compare the influence of actual vs. nominal steel
properties, 94% of failures were recorded (less than 100 mrad of rotation). Failure of the col-
umn web in compression occurs in 60.9% and 57.5% of the cases (nominal and real, respec-
tively), as expected in a joint where the tension zone is 128% stronger than the compression
zone. However, because of the over strength effect of steel, a shift in failure modes is ob-
served whenever the tension zone controls the behaviour. For nominal steel properties, 33%
of the simulated connections fail in the column web in tension while for actual steel proper-
ties, 36% of failures result from the component bolts in tension. Case EA.3 explores the in-
fluence of the post limit stiffness of the column web panel in shear, the component with the
lowest resistance. A decrease in failures (47%) of the column web in compression together
with an increase of failures of the bolts in tension (44%) is observed, together with some
marginal failures of the column web in tension (4%). Comparison of cases EB.1 and EB.2
shows that the use of the mixed distribution versus a normal distribution for the resistance of
the compression and shear components does not significantly affect the results. A slight in-
crease of column web in compression failures (72% to 79%) is noted, together with a small
decrease of ductility (39 mrad to 29 mrad). The simulation of the component column web in
tension for FY and Kp brought no major changes. The consideration of the variability in the
failure deformation of component column web in compression yields a big influence on the
joint ductility, reducing the average rotation capacity to 25 mrad. Globally, taking 30 mrad as
a ductility limit state for sufficient rotation capacity, the nine cases that were analysed yield
probabilities of not reaching the required rotation ranging from 26% to 62%.
CONCLUSIONS
The two examples of joint configuration that were explored in this paper have revealed some
interesting features. Firstly, the flush end-plate beam-to-column joint designed to be critical in
tension (mode 1: end-plate in bending), exhibits a significant probability of bolt failure in ten-
REFERENCES
(1) Eurocode 3, prEN 1993-1-8: 2003, Part 1.8: Design of Joints, Eurocode 3: Design of
Steel Structures, Stage 49 draft., 5 May 2003. CEN, European Committee for Stan-
dardisation, Brussels, 2003.
(2) ENV 1090: 2001. Execution of Steel Structures. CEN, European Committee for Stan-
dardisation, Brussels, 2001.
(3) Zoetemeijer P. Summary of the research on bolted beam-to-column connections. Re-
port 25-6-90-2. Delft University of Technology, Faculty of Civil Engineering, Stevin
Laboratory – Steel Structures, 1990.
(4) Jaspart, JP. Recent Advances in the Field of Steel Joints – Column Bases and Further
Configurations for Beam-to-Column Joints and Beam Splices. Thèse présentée en vue
de l’obtention du grade d’Agrégé de l’Enseignement Supérieur, Année Académique
1996-1997.
(5) Weynand K, Jaspart JP and Steenhuis M. The Stiffness Model of Revised Annex J of
Eurocode 3. In: Connections in Steel Structures III, Proceedings of the 3rd International
Workshop on Connections (eds.: Bjorhovde R, Colson A, Zandonini R), Trento, Italy,
May 8-31, pp 441-452, 1995.
(6) Simões da Silva L, Santiago A and Vila Real P. Post-limit Stiffness and Ductility of End-
Plate Beam-to-Column Steel Joints. Computers and Structures, 80, pp 515-531, 2002.
(7) Swanson JA. Characterization of the strength, stiffness and ductility behavior of T-stub
connections. PhD dissertation, Georgia Institute of Technology, Atlanta, USA, 1999.
(8) Faella, C., Piluso, V., Rizzano, G. Structural Steel Semirigid Connections Theory De-
sign and Software. CRC Press, 2000.
(9) Girão, A., Bijlaard, F., Gresnigt, N. and Simões da Silva, L. Experimental assessment
of the behaviour of bolted T-stub connections made up of welded plates. J. of Con-
structional Steel Research, 60, 269-311, 2004.
(10) Kuhlmann U, Kuhnemund F. Ductility of semi-rigid steel joints. In: Stability and ductility
of steel structures (SDSS 2002); Budapest, Hungary, 2002.
(11) Beg D, Zupančič, Vayas I. On the rotation capacity of moment connections. J. of Con-
structional Steel Research 60(4), 601-620, 2004.
(12) Gervásio, H., Simões da Silva, L., and Borges, L. Reliability assessment of the post-
limit stiffness and ductility of steel joints. J. of Constructional Steel Research 60(4),
635-648, 2004.
(13) Ang, A and Tang, W. Probability concepts in Engineering planning and design. Vols. 1-
2, Wiley, 1975-1984.
(14) Borges, L. Probabilistic Evaluation of the Rotation Capacity of Steel Joints. MSc The-
sis, University of Coimbra, 2003.
(15) Fajkus, M., Melcher, J., Holicky, M., Rozlivka, L., Kala, Z. Design Characteristics of
ABSTRACT
Performance-based seismic design (PBSD) produces structures that meet
multiple performance objectives. Economical application of PBSD produces
structures that:
• have adequate strength and stiffness to remain serviceable during
small, frequent earthquakes, and
• develop cyclic nonlinear deformations while assuring life safety and
collapse prevention during large infrequent earthquakes.
Special concentrically braced frames (SCBFs) and buckling restrained
concentrically braced frames (BRCBFs) can meet these diverse objectives if
the gusset plate connection provides adequate performance. Current
connection design provisions attempt to ensure adequate connection
resistance to avoid premature failure, but the resulting connections may be
massive and uneconomical or provide unacceptable performance. An
analytical and experimental research study to develop improved design
methods for these gusset plate connections is described. A rational, PBSD
procedure is proposed. Future directions of the research study are noted.
INTRODUCTION
Large, infrequent earthquakes induce huge elastic forces in building structures. Therefore,
seismic design of buildings employs relatively small earthquake design forces to assure that
the structure remains serviceable during frequent seismic events, and cyclic, inelastic
ductility is used to prevent loss of life and structural collapse during large seismic events.
This concept is simple and results in economical design, but it is difficult to reliably and
accurately apply in engineering practice. Performance based seismic design (PBSD) is a
recently developed design concept that formalizes this procedure for meeting multiple
design objectives. Concentrically braced frames (CBFs) are stiff, strong steel structures,
which are economical systems for seismic design. The inelastic lateral response of CBFs is
dominated by axial yielding and post buckling deformation of the braces, and special
concentrically braced frames (SCBFs) are designed under guidelines which are intended to
assure good inelastic performance from the brace. These SCBF requirements control the
local and global slenderness of the brace to prevent concentration of local damage during
post buckling deformations. They require that the strength of the connection be greater than
the yield capacity of the brace, and they establish geometric clearances intended to develop
the connection rotations needed to develop brace buckling. Innovative bracing systems,
such as unbonded or buckling restrained braced frames (BRCBFs), have also been
proposed, since they hold promise for improved seismic performance. The seismic
performance of these CBF systems depends on the brace, the connection, and the framing
members. To achieve a superior level of seismic performance, the design of the connection
SCBFs (AISC, 2002) provide economical strength and stiffness and are commonly used for
seismic design. The brace provides lateral stiffness to the frame, and so the brace attracts
large axial forces during earthquake loading. As a result, the cyclic inelastic deformation of
an SCBF is achieved through post buckling deformation and tensile yield of the brace as
illustrated in Zones 0-A, A-B, B-C, C-D and D-E in Figs. 1a and 1b. Plastic hinges form
within the brace after buckling, because of the P-δ moments. These hinges cause
permanent plastic deformations and ultimately deterioration of brace resistance. When the
brace is subjected to a tensile force during load reversals (Zones B-C, C-D and D-E)
significant axial deformation is needed to recover the full tensile stiffness and resistance.
This leads to the one-sided axial force-deflection behavior of the brace seen in Fig. 1a, and
SCBFs use braces in opposing pairs to achieve the system inelastic hysteretic behavior
illustrated in Fig. 1c.
Braces are normally joined to the beams and columns of the frame through gusset plate
connections as illustrated in Fig. 2. SCBF brace post-buckling behavior places significant
cyclic load and deformation demands on these connections as illustrated by the brace end
rotation shown in Zone A-B of Fig. 1b. These connection rotation demands vary depending
upon whether the brace has in-plane or out-of-plane buckling. Figure 3 is a photo of a
buckled braced from a steel frame damaged during a past earthquake, and it shows the
large end rotations that can occur during inelastic seismic deformations.
Brace buckling may result in deterioration of stiffness and resistance and pinched hysteretic
behavior shown in Fig. 1a and 1c. BRCBFs have been developed to increase ductility and
reduce deterioration in brace resistance. BRCBFs are patented systems where an axially
loaded bar is encased into a stiff tube but is not bonded to the tube. As a result, the bar
yields in tension and compression without brace buckling, and the resulting cyclic inelastic
performance of the brace is symmetric without deterioration as illustrated in Fig. 4. Many
engineers currently prefer BRCBFs for seismic design to SCBFs, because of this inelastic
performance. A significant body of research (Clark et al. 2000, Ando et al 1993, Connor et
al. 1997, and Inoue et al. 2001) has been completed on BRCBF performance. However, the
past research and professional opinions on BRCBF braces are based on the assumption
that braced frames behave as a truss, and the brace has pure axial deformation with no
bending moment. This idealization is frequently not achieved in practice, because of the
gusset plate performance.
Figure 4. Axial force-deformation hysteresis curve for an BRCBF (Clark et al. 2000).
Past research (Kahn and Hanson 1976, Foutch et al. 1987, Astaneh-Asl et al. 1982, Lee and
Goel 1987, and Aslani and Goel 1989) shows that SCBFs can provide good seismic
performance if premature fracture or tearing of the brace and the connection is avoided.
The AISC Seismic Design Requirements for the SCBF system (AISC 2002) provide
guidelines toward meeting these goals. Slenderness limits for the brace and different brace
geometry and configuration requirements are provided to avoid concentration of inelastic
strain that leads to early tearing or fracture. The AISC seismic provisions also require that
the connection be designed to be stronger than the brace, and with out-of-plane buckling,
geometric limits are established to permit the expected end rotation on the connection.
These design rules lead to the common conception that a stronger gusset plate connection
is better, and some very uneconomical and impractical connections such as illustrated in
Fig. 5 have been built. Further, it is difficult to satisfy the out-of-plane buckling geometry
requirements for most practical gusset plate connections. The past performance of braced
frames during earthquakes has been mixed. In some cases, economical and serviceable
performance during earthquake loading has resulted, but in others, the apparent resistance
and ductility were significantly smaller than expected when brace or connection failures such
as illustrated in Fig. 6 occurred.
Prior discussion demonstrates the need to understand and improve the BRCBF and SCBF
gusset plate connection design to efficiently and economically achieve multiple performance
objectives for seismic design. The stiffness and strength of CBF systems clearly aid in
satisfying the Operational and Immediate Occupancy performance levels, and they provide
economical building design. However, the SCBF and BRCBF connections must
accommodate significant inelastic deformation and strength demands for the Life Safety and
Collapse Prevention performance levels. These later design limit states are less easily
satisfied, because they place both force and deformation demands on the gusset plate
A rational hierarchy of yield mechanisms within the frame and the connection are required to
meet these diverse force and deformation demands, because a simple connection
resistance check as presently used in SCBF seismic design leads to variable seismic
performance. The simple resistance checks in the AISC requirements do not assure the
most desirable connection behavior nor do they prevent the least desirable performance.
The proposed design procedure must address this complex braced frame behavior while
retaining a simple design method.
The proposed procedure employs concepts developed for PBSD of steel moment-resisting
frames (Roeder 2001 and 2002) during the SAC Steel Project. This design procedure
developed balance conditions to assure desirable yield mechanisms, restrict undesirable
failure modes, and achieve the desired seismic performance. Occurrence of a yield
mechanism changes the stiffness and provides inelastic deformation of the structure without
significant loss in resistance. The occurrence of a failure mode can lead to fracture, loss in
resistance, and reduced inelastic deformation capacity. A single failure mode will produce a
significant reduction in resistance or deformation capacity, but multiple failure modes are
usually required to produce complete connection failure. Similar design procedures are
being developed to achieve the performance objectives of SCBF and BRCBF structures.
The yield mechanisms and failure modes for the CBF systems as illustrated in Fig 9 are
considered in this evaluation. The controlling yield mechanisms for SCBFs are expected to
be inelastic shortening due to post-buckling deformation and tensile yielding of the brace.
The axial load capacity of the BRCBF brace is similar in tension and compression, and the
brace should not buckle. As a result, the controlling yield mechanism for BRCBF systems
will be this axial yield deformation. Secondary yield mechanisms are also possible with
braced frame gusset plate connections. These include local yielding of the gusset plate,
local yielding of the beam and column adjacent to the gusset plate and elongation of
boltholes in brace and gusset plate. These secondary yield mechanisms can contribute
some plastic deformation during large earthquakes, but their deformation capacity is limited
so that they are not capability of providing primary sources of inelastic deformation. CBFs
have many different possible failure modes including tearing or fracture of the brace, net
section fracture of the brace or gusset plate, weld fracture, shear fracture of the bolts, block
shear, excessive bolt bearing deformation, and buckling of the gusset plate.
A rational connection design procedure for seismic design can be based upon the expected
or mean yield resistance of the controlling yield mechanism, and its comparison to the
Where Rfailure is the failure mode resistance for an individual failure mode, and Ag is the
gross cross sectional area of the brace. Ry is a factor for adjusting the nominal yield stress,
Fy and Fcr, to the expected or average yield and critical buckling stress, respectively. Two
balance equations may be required for SCBFs because both brace buckling and tensile yield
serve as controlling yield mechanisms for some failure modes. BRCBFs have identical yield
resistance in both tension and compression, and so only Eq. 1a is required.
The β factor used in these balance equations is similar to the φ factor in LRFD design (AISC
2001) in that both are less than 1.0 and both are based upon the performance and variability
of the structural elements. However, φ is based solely upon strength, safety, and statistically
extreme considerations, while β depends upon balancing the expected inelastic seismic
behavior to meet the performance requirements and inelastic deformation capacity.
Different β values will be required for different design limit states and for different failure
mode-yield mechanism combinations, because brace connection design must consider the
full range of seismic behavior at all performance levels. Connection strength and stiffness
must be adequate to insure serviceable seismic performance, but once this adequate
Some failure modes may be controlled by prescriptive design measures. For example,
fracture or tearing of the SCBF brace is theoretically controlled by global and local
slenderness limits. In addition, geometric constraints are employed with SCBF gusset plate
connections to assure that connection can tolerate brace end rotations due to post-buckling
brace deformation. BRCBFs require great control over the connection deformation to avoid
premature brace damage, and so additional constraints may be needed for out-of-plane
stiffness of these BRCBF connections. This research work focuses on improving the
seismic design of these gusset plate connections through:
CONTINUING WORK
This research work is in progress and will continue for the next two years. At present, past
research is being compared and evaluated and this database will be used to develop and
verify accurate but simple models for predicting connection behaviors. The inelastic
performance achieved with different yield mechanism and failure mode conditions for SCBF
and BRCBF systems in these past research studies will be assessed. Balance conditions (β
values) necessary to assure the proper combination of behaviors are achieved in practice
will then be developed to establish a proposed gusset plate connection design procedure.
The proposed connection design procedure will then be evaluated experimentally. The
experimental results will be used to modify the design procedure to assure that it provides
the greatest design economy combined with good seismic performance at all performance
levels. Follow-up experiments will be conducted to evaluate the design modifications.
The experiments will be designed to evaluate bolted and welded gusset plate connections.
The test matrix will be developed to consider variation in the type of brace, type of
The test specimens will be constructed and attached to the laboratory strong floor as
depicted in Figure 10. Each specimen will consist of end gusset plates attached to a brace
and the surrounding beam and column framing. This configuration should insure realistic
boundaries for the test specimens. The imposed displacement history will include cyclic
deformation with multiple cycles of increasing story drift such as employed with the ATC-24
testing protocol. Initial cycles will be at deformations below the initial yield and buckling
loads of the brace to examine Operational and Immediate Occupancy performance limit
states. Subsequently, multiple cycles will be completed at and slightly above the buckling
load and tensile yield load of the brace. Finally multiple cycles will then be completed with
increasing inelastic story drift until ultimate failure of the brace or the connection occurs.
These later cycles will be documented with emphasis upon the Life Safety and Structural
Collapse Prevention performance limit states. Work on these tests will begin in Spring 2004.
ACKNOWLEDGEMENT
This research work is funded by the National Science Foundation through Grant CMS-
0301792, Performance-Based Seismic Design of Concentrically Braced Frames. Dr. Steven
L. McCabe is the Program Manage for this research. This financial support is gratefully
acknowledged.
REFERENCES
(1) Ando, N. Takahasi, S. and Yoshida, K., (1993) "Behavior of Unbonded Braces
Restrained by Reinforced Concrete and FRP," ASCE, Composite Construction II, New
York, pgs 869-882.
(2) AISC (2001) "Manual of Steel Construction, Load and Resistance Factor Design," 3rd
Edition, American Institute of Steel Construction, Chicago, IL.
(3) AISC (2002). "Seismic Provisions for Structural Steel Buildings," American Institute of
Steel Construction, Chicago, IL.
(4) Astaneh-Asl, A., Goel, S.C., and Hanson, R.D., (1982) "Cyclic Behavior of Double
Angle Bracing Members with End Gusset Plates," Research Report UMEE 82R7,
Department of Civil Engineering, University of Michigan, Ann Arbor, MI.
(5) Clark, P.W., Kasai, K., Aiken, I.D., and Kimura, I., (2000) "Evaluation of Design
Methodologies for Structures Incorporating Steel Unbonded Braces for Energy
Dissipation," Proceedings 12th WCEE, Auckland, New Zealand.
(6) Connor, J.J., Wada, A., Iwata, M., and Huang, Y.H., (1997) " Damage-Controlled
Structuresl I: Preliminary Design Methodology for Seismically Active Regions," ASCE,
Journal of Structural Engineering, Vol. 123, No. 4, pgs 423-31.
(7) Foutch, D.A., Goel, S.C. and Roeder, C.W., (1987) Seismic testing of a full scale steel
building - Part I, Journal of Structural Division, ASCE, No. ST11, Vol. 113, New York,
pgs 2111-29.
(8) Goel, S.C. (1992). "Earthquake Resistant Design of Ductile Braced Steel Structures,"
Stability and Ductility of Steel Structures Under Cyclic Loading, edited by Y. Fukumoto
and G.C. Lee, CRC Press, Boca Raton, Florida.
(9) Inoue, K., Sawaizumi, S., and Higashibata, Y., "Stiffening Requirements for Unbonded
Braces Encased in Concrete Panels, ASCE, Journal of Structural Engineering, Vol
127, No.6, pgs 712-19.
(10) Kahn, L.F., and Hanson, R.D., (1976). "Inelastic Cycles of Axially Loaded Steel
Members," Journal of Structural Division, ASCE, No. ST5, Vol. 102, pgs 947-59.
(11) Lee, S., and Goel, S.C., (1987). "Seismic Behavior of Hollow and Concrete Filled
Square Tubular Bracing Members," Research Report UMCE 87-11, Department of
Civil Engineering, University of Michigan, Ann Arbor, MI.
(12) Popov, E.P. Takanashi, K., and Roeder, C.W. (1976) Structural Steel Bracing
Systems, EERC Reprot 76-17, University of California, Berkeley, 1976
(13) Roeder, C.W., (2001) “State of Art Report – Connection Performance”, FEMA 355D,
Federal Emergency Management Agency, Washington, D.C.
(14) Roeder, C.W., (2002) "Connection Performance for Seismic Design of Steel Moment
Frames," approved for publication, ASCE, Journal of Structural Engineering.
ABSTRACT
Full-scale cyclically loaded cruciform experiments with weak panel zones and
welded unreinforced flange-welded web (WUF-W) prequalified moment
connections performed well provided that the weld metal has minimum Charpy
V-Notch (CVN) toughness. Welds with low CVN and brittle fracture was
obtained in one specimen despite using weld metal certified to meet minimum
CVN requirements. Unstiffened columns perform well and alternative stiffening
details all performed well, indicating that variations in column stiffening details
did not affect the potential for fracture or low-cycle fatigue. Low-cycle fatigue
performance is compared to strain range-cycles curves extrapolated from high-
cycle S-N curves.
INTRODUCTION
Following the Northridge earthquake of January 17, 1994, extensive damage to steel
moment connections was reported (1-3). This damage most often consisted of brittle
fractures of the bottom girder flange-to-column flange Complete Joint Penetration (CJP)
groove welds. The fractures were caused by the use of low toughness welds combined with
a number of other connection detailing and construction practices that were typical prior to
the earthquake (3-5). Additionally, column stiffening practices have been cited as a possible
contributor to the fractures, largely as a result of observations that many of the connections
fractured during the Northridge earthquake lacked continuity plates and that some had weak
panel zones (6). Finite element analyses (7-10) also have shown an increase in stress and
strain concentrations in the girder flange-to-column flange CJP welds associated with
excessively weak panel zones or insufficient continuity plates. It is presumed that these
stress and strain concentrations increase the potential for fracture. As a result of these
observations, there has subsequently been a tendency to be more conservative than
necessary in designing and detailing of the continuity plates and doubler plates in steel
moment connections.
Recommendations for the seismic design of new steel moment-frame buildings (3) provide
equations for determining whether continuity plates are required, and indicate that any
required continuity plates must be of equal thickness to the girder flange for interior
connections (thinner continuity plates are permitted for exterior connections), unless
connection qualification testing demonstrates that the continuity plates are not required.
Furthermore, the connection of the continuity plates to the column flanges must be made
with CJP welds, and reinforcing fillet welds should be placed under the backing bars.
Design criteria for the limit states related to column stiffening are presented in the AISC
The tendency towards being more conservative than necessary in column stiffener design
has raised concerns about economy as well as the potential for cracking of the k-area in the
column web near the web-flange junction during fabrication due to high residual stresses
caused by highly restrained CJP welds on the continuity plates or doubler plates (14,15).
Therefore, a combined experimental and computational research study was conducted at the
University of Minnesota to reassess the recent column stiffener design and detailing
provisions and recommendations, and to provide economical alternative stiffener details that
minimize welding along the column k-line while retaining superior performance for non-
seismic and seismic design (16). This paper examines the effect of variations in column
stiffening, including no stiffening as well as more economnical alternative stiffening details,
on the fracture and low-cycle fatigue performance of Welded Unreinforced Flange-Welded
Web (WUF-W) moment connections (3).
The design and results of these tests are reported elsewhere, including an assessment of
LFB and LWY limit states (17) and the cyclic panel zone behavior and design (18). Related
research included nine pull-plate experiments (19-22) that investigated the limit states of LFB
and LWY, primarily for non-seismic design, and further tested the alternative doubler plate
and continuity plate stiffener details. Finite element analyses of all experimental specimens
were also conducted as part of this research as well as parametric studies to extend the
results to member sizes and details not tested (23).
Six full-scale, girder-to-column cruciform specimens were tested (Table 1). The SAC (24)
loading history was applied including six cycles at each interstory drift level of 0.375%, 0.5%,
and 0.75%, four cycles at 1.0% interstory drift level, and two cycles at each interstory drift
level of 1.5%, 2.0%, 3.0%, and 4.0%.
ASTM A992 wide-flange sections and A572 Grade 50 plate was used. The specimens used
the pre-qualified (3) WUF-W connection detail (Figure 1). The column stiffening was varied
in these specimens including three alternative doubler plate details (i.e., back-beveled fillet-
welded doubler plate, square-cut fillet-welded doubler plate, and groove-welded box doubler
plate; see Figure 2) and a fillet-welded ½ in. thick fillet-welded continuity plate detail.
Table 2 presents the design strength-to-demand ratios using minimum specified material
properties for the LFB and LWY limit states. Note that the design strength for LFB and LWY
is the design strength of the column shape alone and does not include the column
reinforcement, if any. The demand is calculated with various methods:
Table 2 shows that 1992 AISC Seismic Provisions (12) and AISC Design Guide No. 13 (25)
provide similar demand values. For these cases, the demand exceeds the capacity for LFB
for all specimens except CR1, whereas only specimen CR3 has continuity plates (although
the box detail functions as a continuity plate as well as a doubler). The remaining specimens
CR2 and CR5 are underdesigned for LFB for seismic demand.
Figure 1. Typical welding details used for cruciform specimens (Specimen CR1).
However, the girder flange demand predicted by the latter two methods is very conservative
and can be put in perspective by comparing to the maximum possible uniaxial tensile
strength of A992 steel. The stress in the flange is 1.8 times 50 ksi or 90 ksi, well above the
likely tensile strength of A992 steel. For example, a survey of more than 20,000 mill reports
from (26,27) showed that A992 steel has a mean tensile strength of 73 ksi. The 97.5
percentile tensile strength was 80 ksi, and the maximum value reported was 88 ksi.
Also shown in Table 2 are the panel zone capacity-to-demand ratios (including the strength
of the doubler plate in the capacity) and column-girder moment ratios calculated from the
Figure 2. Doubler details: (a) back-beveled fillet-welded doubler (Detail I), (b) square-cut
fillet-welded doubler (Detail II), (c) box (offset) doubler (Detail III).
Table 2. Nominal capacity/demand ratios of PZ yielding, LFB, and LWY limit states.
LFB φRn/Ru LWY φRn/Ru
∑ M *pc PZ
∑ M *pb φvRv/Ru 1.8Yield a 1.8Yield a
a a a
(AISC) Yield 1992 DG13 Yield 1992 DG13 a
(AISC)
Seismic Seismic
3.04 1.69 1.64
CR1 1.50 0.72 2.38 1.32 1.29
(3.38)b (1.88) (1.82)
1.47 0.82 0.80
CR2 0.99 0.66 2.20 1.22 1.19
(1.63) (0.91) (0.89)
1.22 0.68 0.66
CR3 0.89 0.74 2.51 1.39 1.36
(1.36) (0.76) (0.73)
CR4 1.22 0.68 0.66
0.89 0.93 3.19 1.77 1.73
CR4R (1.36) (0.76) (0.73)
0.84 0.47 0.46
CR5 0.73 0.74 2.34 1.30 1.27
(0.93) (0.52) (0.51)
a
Equation used to calculate demand, Ru
b
Values in parentheses reflect use of φ = 1.0
Weld details
E70T-1 (Lincoln Outershield 70) wire with 100% CO2 shielding gas was used for all shop
welding. The girder flange-to-column flange CJP groove welds were made in the flat position
with E70T-6 (Lincoln Innershield NR-305) wire. Welds made with E70T-6 wire are required
by AWS A5.20 (29) and AISC 2002 Seismic Specifications (28) to have notch toughness of
20 ft-lbs at -20°F. FEMA 350 (3) has recommended minimum notch toughness requirements
at two temperatures, 20 ft-lbs at 0°F and 40 ft-lbs at 70°F. According to the Lincoln Electric
Company product family literature, the typical values for NR-305 are 21 to 35 ft-lbs at -20°F
and 21 to 54 ft-lbs at 0°F. As shown in Table 3, Specimens CR1 and CR4 were fabricated
with a 5/64 in. diameter NR-305 wire and the remaining were fabricated with 3/32 in.
diameter NR-305 wire. All CJP welds were ultrasonically tested by a certified inspector in
conformance with Table 6.3 of AWS D1.1-2000 (30) for cyclically loaded joints.
The out-of-position field welds, including the CJP welds connecting the girder web to the
column flange and all reinforcing fillet welds were made with 0.068 in. diameter E71T-8
(Lincoln Innershield NR-203MP) wire for Specimens CR1 and CR4, and 5/64 in. diameter
The shear tab was designed to extend approximately 0.25 in. into the top and bottom access
holes and acted as the backing bar for the CJP welds of the girder web to the column flange.
This extension acted as a short runoff tab, allowing the weld to extend the full depth of the
girder web. Ricles et al. (31) recommended that these runoff tabs of the vertical web weld be
ground smooth, which is labor intensive. Since it was felt that this might not be necessary,
these runoff tabs were not ground smooth in the specimens tested in this work.
BRITTLE FRACTURE
Specimen CR4 exhibited premature brittle failure in three of four girder flange-to-column
flange CJP welds in the early stage of the SAC (1997) loading history and was stopped after
one-half cycle at 2.0% interstory drift. It was found that this specimen was unintentionally
prepared with low-toughness weld metal, as shown in Table 3. Note that the AWS
Certificate of Conformance for this wire indicated that the weld metal meets the minimum
toughness requirement of AWS A5.20 (28) of 20 ft-lbs at -20°F (32).
Specimen CR4R was essentially a replicate test of Specimen CR4, except that the weld
metal used for Specimen CR4R met the minimum requirements of FEMA 350 (3). Specimen
CR4R not only performed acceptably according to the SAC (24) requirements, it performed
as well as any of the other specimens successfully tested in this experimental study. The
fact that the box (offset) doubler plate detail performed well in Specimen CR4R indicates that
the detail itself was probably not a factor in the fracture that occurred in Specimen CR4.
Following the premature brittle failure in Specimen CR4, it was found that the previously
tested Specimen CR1 also had relatively low weld-metal notch toughness, with an average
of 2.6 ft-lbs at 0°F and 19.3 ft-lbs at 70°F as presented in Table 3. Specimen CR1, which
was welded using the same wire and the similar welding procedure as Specimen CR4, but
had marginally better notch toughness, performed very well, experiencing 14 cycles of 4.0%
interstory drift before the significant strength degradation. It is important to note that this
Specimen CR1 had no doubler plates or continuity plates, even though doubler plates would
be required as shown in Table 2. Thus this test shows that column stiffeners are not
absolutely required to avoid brittle fractire or low-cycle fatigue, even with this very poor notch
toughness. These two tests have closely bracketed the minimum notch toughness required
for adequate performance of CJP welds.
It is believed that the FEMA 350 requirements (3) for minimum notch toughness are
LOW-CYCLE FATIGUE
The remaining connections exhibited no brittle fractures, but low-cycle fatigue failures occurred
after significant cyclic loading. Figures 3 and 4 show low-cycle fatigue cracks forming at the
beam flange weld. These beam flange weld cracks were the only type of low-cycle fatigue
cracks that actually propagated to cause failure, which is defined here as significant strength
reduction. Low-cycle fatigue cracks did originate at the weld of the beam web to the column
flange and at the weld access hole, as shown in Figure 5. However after propagating for a short
distance they arrested and did not propagate further or lead to failure, so therefore they are not
structurally significant.
(a) (b)
Figure 3. Low cycle fatigue crack developing at the toe of the beam flange weld in a moment-
frame connection after (a) 11 cycles and (b) 17 cycles of 4% drift.
Table 4 shows the cycles at 3% or 4% drift when the first crack was first visible in the CJP
welds, the final cycles at 4% drift when the strength was reduced, and the measured strain
ranges. The pairs of measured strain ranges for each specimen are from the west top flange
and the east bottom flange, respectively (except for CR1 where only the west top flange data
were avaialble). The average of five gages across the width was used to eliminate some of
the scatter and the effect of strain gradients. It is believed that the variation in the measured
strain ranges is random, and that the strain range at 4% drift was relatively consistent among
the specimens, averaging approximately 4.1%.
The performance requirement is that the connections must complete 2 cycles at 4% drift without
a significant reduction in strength in order to be prequalified connections (3). One could
conclude that all these connections (except the original CR4, which experienced brittle fracture)
met this performance requirement and that therefore the performances of the specimens are
equally good. However, there may be some significance to the final number of cycles before
LCF Crack
Column Column
Web Flange
Slag Inclusion,
LOF
Figure 4. Cross section of beam flange weld showing low cycle fatigue crack
developing at the weld toe.
Figure 5. Low cycle fatigue cracks forming at the end of the beam web to column flange
weld and at the weld access hole.
Most past research on low-cycle fatigue has involved pressure vessels and some other types
of mechanical engineering structures. Since low-cycle fatigue is an inelastic phenomenon,
the strain range is the key parameter rather than the stress range as in high-cycle fatigue.
The Coffin-Manson rule (33) has been used to relate the strain range in smooth tensile
specimens to the fatigue life. Manson suggested a conservative lower-bound simplification,
called Manson’s universal slopes equation (34):
σu
∆ε = 3.5 N −0.12 + ε 0f .6 N −0.6 (1)
E
where: ∆ε is the total strain range, σu is the tensile strength, and εf is the elongation at fracture.
Note that the first term in Equation 1 is the elastic part of the total strain range (which is relatively
insignificant when there are fewer than 100 cycles) and the second term is the plastic part of the
total strain range. Figure 6 shows a plot of Manson’s universal slopes equation where σu is 450
MPa and εf is 25%, typical minimum properties for Grade 50 structural steel. Many studies have
shown that Manson’s universal slopes equation is conservative compared to experimental data
from smooth specimens (34,35). However, because of buckling at greater strain ranges, most of
the experimental data are for strain ranges less than 1%, i.e., for cycles greater than 1000.
Limited data exist at higher strain ranges – some are shown in Figure 6 for A36 steel smooth
specimens machined from the flanges of wide-flange sections (35).
At this time, very little is understood about low-cycle fatigue in welded or bolted structural details.
For example, it is a very difficult task to predict accurately the local strain range at a location of
cyclic local flange buckling. However, Krawinkler and Zohrei (36) and Ballio and Castiglioni
(37,38) showed that the number of cycles to failure by low-cycle fatigue of welded connections
could be predicted by the local strain range in a power law that is analogous to an S-N curve.
Ballio and Castiglioni (37,38) showed that the power law would have and exponent of 3, just like
the elastic S-N curves. Krawinkler and Zohrei (36) also showed that Miner’s rule (39) could be
used to predict the number of variable-amplitude cycles to failure based on constant amplitude
test data.
Therefore, it may be possible to predict and design against low-cycle fatigue using strain-range
vs. number-of-cycles curves that are extrapolated from the high-cycle fatigue design S-N curves.
Figure 6 shows the AISC S-N curves (11) for Categories A and C, converted from stress range
to strain range by dividing the stress ranges by the elastic modulus, and extrapolated up to one
cycle.
There are only limited data to support this approach. Figure 6 shows the strain range-number of
cycles data from these WUF-W beam-to-column connection tests. The number of cycles plotted
in Figure 6 is the number of cycles at 4% drift. If the effect of all the previous cycles is included
using Miner’s rule (39), they add up to an equivalent of about one additional cycle at 4%. Since
flange strains right at the weld toe were used rather than nominal values, this is analogous to a
hot-spot approach for high-cycle fatigue. For high-cycle fatigue, the Category C S-N curve is a
suitable baseline S-N curve for the hot-spot approach (40,41). It appears that the Category C S-
N curve is also a good lower bound to these low-cycle fatigue data. The scatter in the data is
substantial, as is also true in high-cycle fatigue.
Category A
10
Category C
0.1
1 10 100 1000 10000
Cycles
Figure 6. Comparison of standard S-N curves presented in terms of strain range and Manson’s
universal slopes equation for Grade 50 (350 MPa yield strength) steel to low-cycle fatigue test
data and the connection test data.
1. When properly detailed and welded with notch-tough filler metal, the WUF-W steel
moment connections can perform adequately even though relatively weak panel zones and
low local flange bending strengths were chosen.
2. The failure mode of the specimens other than the original CR4 was low-cycle fatigue
(LCF) crack growth and eventual rupture of one or more girder flange-to-column flange
complete joint penetration (CJP) groove welds. Low cycle fatigue may be conservatively
predicted using strain-range vs. cycles curves derived from the stress based S-N curves for
high-cycle fatigue.
3. Specimens CR1 and CR4 were unintentionally prepared with weld metal that had CVN
values that were much lower than the minimum requirements. The premature brittle failure of
specimen CR4 reconfirmed that achieving the required minimum CVN toughness in the
girder flange-to-column flange CJP welds is critical. These low toughness welds occurred
despite the certification of the filler metal; the certification is only required annually, unlike the
way that each heat of steel is tested. A study should be conducted to fully characterize the
typical variability in the CVN and other properties of the weld. An evaluation of the need for
lot testing should be performed. Consideration should also be given to use of filler metals
with a distribution of CVN such that there is a sufficiently small probability of not meeting the
minimum required values, and therefore lot testing may not be required.
4. Application of the alternative column stiffener details (i.e., back-beveled fillet-welded
doubler plate detail; square-cut fillet-welded doubler plate detail; groove-welded box doubler
plate detail; fillet-welded 1/2 in. thick continuity plates) in the WUF-W steel moment
connections was successfully verified. No cracks or distortions were observed in the welds
connecting these stiffeners before the rupturing of the girder flange-to-column flange CJP
ACKNOWLEDGEMENT
This research was sponsored by the American Institute of Steel Construction, Inc. and by the
University of Minnesota. In-kind funding and materials were provided by LeJeune Steel
Company, Minneapolis, Minnesota; Danny’s Construction Company, Minneapolis,
Minnesota; Braun Intertec, Minneapolis, Minnesota; Nucor-Yamato Steel Company,
Blytheville, Arkansas; Lincoln Electric Company, Cleveland, Ohio; and Edison Welding
Institute, Cleveland, Ohio. Supercomputing resources were provided by the Minnesota
Supercomputing Institute. The authors wish to thank T. V. Galambos and P. M. Bergson,
University of Minnesota, L. A. Kloiber, LeJeune Steel Company, and the members of the
technical advisory group on this project for their valuable assistance.
REFERENCES
ABSTRACT
Experimental program is performed on bolted end-plate type joints of 19 steel
and steel-concrete composite specimens under cyclic loading. The test
specimens are designed to study and characterise the typical cyclic behaviour
modes of this type of joints. The paper summarises the details of the
experimental program, the cyclic joint behaviour and the cyclic characteristics.
The joint design characteristics (moment resistance and initial stiffness) are
determined by the Eurocode standard and compared to the experimental joint
parameters. After the verification of the standard procedure a parametric study
is completed to study the influence of the structural details on the design
parameters.
INTRODUCTION
The paper presents an experimental and Eurocode standard based analyses of end-plate
type joints. The experimental program is performed in a co-operation project between the
Budapest University of Technology and Economics and the Instituto Superior Tecnico in
Lisbon. The subject of this experimental program is to test bolted end-plate type steel and
composite joints (19 specimens) under cyclic reversal loading. The experimental program is
published in (1), in this paper only the short summary of the specimen characteristics, the
test set-up and procedure and the behaviour modes are presented.
The aim of the research is to observe the typical cyclic behaviour modes of the studied
connection type, characterize it quantitatively and find structural solution for the improved
cyclic behaviour. On the basis of the experimental results a calculation method is under
development for the determination of the monotonic and cyclic design parameters of the
joints. In the first phase of it Eurocode 3 and 4 based numerical study is performed on the
monotonic joint characteristics (the moment resistance and the initial stiffness). The paper
presents the standard procedure with the comparison of the design values and the
experimental results. Since cyclic tests are performed the envelope curve of the hysteretic
moment–rotation diagram is applied to characterize the monotonic behaviour. The
experimentally verified Eurocode procedure is to be extended to describe cyclic behaviour.
The developed, Eurocode based, calculation method is also presented in the paper with the
comparison of the experimental results.
After the verification of the proposed method a parametric study is performed to study the
influence of structural details on the design moment resistance and initial stiffness. In the
further step of the research the cyclic characteristics (degradation characteristics, absorbed
energy, ductility, etc.) of the joint will be evaluated to gain the improved structural solution for
cyclic behaviour of this type of joints.
The test programs are performed on end-plate type bolted joints with different parameters of
each specimen. In case of the first test program the aim is to have information about the
cyclic behaviour modes of the joint (2). During the second experimental program the main
parameters of the specimens are identical with those of the first case; the only difference is
the composite member section (3). In case of the third program the focus is on the cyclic
local buckling of the slender composite section (4).
Altogether three test sets are performed on end-plate type joints. In case of the first test set
the specimens are designed with H-shaped steel element (hot-rolled or welded) as it is
illustrated in figure 1 a). In the second and third test set composite member – steel section
with concrete filling and reinforcement between the flanges – is applied as figure 1 b) shows.
Table 1 contains the main characteristics of each specimen.
The test setup is developed at the Instituto Superior Tecnico to test beam-to-column joints of
steel frames in cantilever type arrangement (1), as it can be seen in figure 2. The loading
history is defined in accordance with the ECCS Recommendations (5). It is decided to use
the same loading history in all test cases to be able to compare the results of similar
specimens. Inductive displacement transducers are used to measure the displacements
during the testing procedure.
On the basis of the measured data the cyclic moment-rotation curves are determined. The
calculation of the moment-rotation diagrams are performed in the so-called ‘joint reference
section’. In this way the disturbing effect of the local buckling of the flanges is eliminated. The
moment-rotation diagrams illustrate the various hysteretic behaviours and characterise
qualitatively the joints by reflecting the tendencies of the rotational stiffness, the moment
carrying capacity, the rotational capacity and the absorbed energy. The behaviour modes are
detailed in (1) and summarised as the follows.
‘Pure’ bolt-failure:
The ‘pure’ bolt-failure occurred when the end-plate is thick enough to resist the failure without
significant deformation (CCB1 and CCB3). The governing phenomenon is the plastic
deformation and bolt fracture after pinching of the bolt shanks. The most representative
phenomenon is the rigid-body type rotation of the specimen, which is caused by the plastic
elongation of the bolt shanks and appears as large deformations on near-zero moment level
as it is shown in figure 3 a). This type of joint has small deformation and moment carrying
capacity limited by the ductility and grade of the bolts. While the bolts govern the behaviour
of the specimen the concrete filling has no significant effect on the cyclic behaviour since the
load carrying and deformation capacities are limited by the characteristics of the bolts. The
specimen after failure is presented in figure 3 b).
60 rigid-body
moment [kNm]
40 type rotation
20
0
-20
-40
-60
-50 -40 -30 -20 -10 0 10 20 30 40 50
rotation [mrad]
‘Pure’ plate-failure:
The ‘pure’ plate-failure type behaviour mode occurs when the end-plate is thin comparing to
the other elements (CCB2). The governing phenomenon is the elastic/plastic deformation
and plastic hinge of the end-plate. In the followings subsequent cycles cracks occurred and
propagated in the plate near to the welding of the flanges, which caused resistance
degradation. Plastic deformation of the bolts is not significant. Figure 4 a) shows the
representative moment-rotation diagram of the ‘pure’ plate-failure. The deformation capacity
of this joint is reached higher value than in case of the bolt-failure. The absorbed energy is
also higher. The crack of the end-plate is shown in figure 4 b).
60
40
moment [kNm]
20
0
-20
-40
-60
-60 -45 -30 -15 0 15 30 45 60
rotation[mrad]
a) M-Θ diagram b) Specimen after failure
Figure 4. Plate-failure.
40
20
0
-20
-40
-60
-60 -45 -30 -15 0 15 30 45 60
rotation [mrad]
a) M-Θ diagram b) Specimen after failure
Figure 5. Combined plate+bolt-failure.
Plate buckling-failure:
In case of slender welded composite member the plate buckling-failure of the flanges is
observed (the sections belong to the Class 4 in Eurocode 3). The phenomenon of the local
plate buckling is the following: the elastic buckling of the compressed flanges is appeared in
early cycles. After this the phenomenon became plastic plate buckling and later plastic lines
are developed and the phenomenon turned into a yield mechanism. The final collapse
caused by the cracking and fracture of the tension flanges due to low cycle fatigue. The
buckling behaviour is the symmetric buckling of the flanges since the typical asymmetric
buckling pattern of the steel sections cannot be developed due to the supporting effect of the
concrete. The final failure is the cracking and fracture of the tension flange and also the
cracking of the web. The representative moment-rotation diagrams of these behaviour
modes are plotted in figure 6 a). The moment carrying and rotational capacity of these joints
are favourable and besides this small degradation of the absorbed energy is observed. The
significant resistance decreasing can be seen after the cracking of the flange and falling out
of the crushed concrete (figure 6 b).
75
50
moment [kNm]
25
0
-25
-50
-75
-100 -75 -50 -25 0 25 50 75 100
rotation [mrad]
50
S jini
25
moment [kNm]
M jRd
0
-25
-50
-60 -40 -20 0 20 40 60
xd cd rotation[mrad]
The EC3 gives the following formula to calculate the moment resistance:
M j,Rd = Σh r Ftr,Rd (1)
In our case we have only one connecting member, which connected to a rigid support, so the
half of the joint is studied. For this reason the tension resistance of the bolt-row (Ftr,Rd) is
defined as the resistance of end-plate in bending (Ft,ep,Rd), which formulae depends on the
failure mode. The failure modes are clearly defined from experimental tests as previous
sections discuss (Plate-failure – Mode1; Combined plate+bolt-failure – Mode2; Bolt-failure –
Mode3). The Eurocode 4, 1.1 (7) gives the moment resistance of the composite section, as it
is illustrated in figure 8.
tg fck fs fy
tf
Fc
xc
dt
Fs Fy
M pl,Rd
h1
h
F'y
zc
zy
zs
F's
z'y
z's
bp
The rotational stiffness of the joint is calculated from the following formula:
2
hr
Sj = E ⋅ (2)
1
µ⋅Σ
ki
Two stiffness coefficients of EC3 are taken into account: end-plate in bending (k5) and bolts
in tension (k10).
The moment resistance and initial stiffness are calculated for each specimen for both positive
(Mj,Rd,exp1) and negative (Mj,Rd,exp2) hemicycles. The following notifications are drawn for the
calculation method:
- The yielding stress (fy) of the steel material is from material tests and it is increased by
10% to consider the hardening of the material.
- The experimentally determined rotational stiffness (Sj,ini,exp1 and Sj,ini,exp2) is evaluated from
the unloading part of the hysteretic curve, since initial part of the loading path is unsteady
caused by the uncertainties of the specimen and the test equipment.
The experimental and numerical moment resistances are shown in figure 9. In case of the
local buckling type failure the standard values reach 82-105% of the experimental moment
resistances, which are in good coincidence (see figure 9 a). When the failure is occurred in
the connecting elements (Mode1-Mode3) significant differences between the experimental
and the standard values (56-87%) are found, as figure 9 b) shows.
100 100
80 x 80
moment [kNm]
moment [kNm]
60 60
40 40
20 20
0 1 2 3 4 5 6 7 8 9 10 10 12 14 16 18 20 22 24 26 28 30
flange/web thickness [mm] end-plate thickness [mm]
exp.results EC results exp.results EC results mod EC results
On the bases of the experimental results the modification of the design method is proposed
to apply the standard EC3 design procedure for cyclic behaviour: During cyclic loading the
bolts have residual plastic deformation/elongation. Due to this effect the lever arms are
modified as follows (see figure 10).
F F F
hr1
hr2
hr3
Mode1: 'pure' plate-failure. No significant bolt elongation, but the bolt heads have
deformation; hr1 is the distance between two bolt rows (see figure 10 a).
Mode2: combined failure. Plastic bolt elongation and end-plate deformation occurred; hr2
lever arm is measured between the tensioned bolt row and the half distance of the
compressed bolt row and edge of the end-plate (see figure 10 b).
Mode3: 'pure' bolt-failure. The bolts have plastic elongation without significant deformation of
the end-plate; hr3 is the distance between the tensioned bolt row and the lower edge
of the end-plate.
70 70
60 60
S jini [kNm/mrad]
S jini [kNm/mrad]
50 50
40 40
30 30
20 20
10 10
10 20 30 40 10 20 30 40
end-plate thickness [mm] end-plate thickness [mm]
exp. results EC results mod. EC results exp. results EC results mod. EC results
In case of the presented experimental program, the initial stiffness Sjini depends on two
stiffness parameters (k5,k10). The modification of these parameters is as the follows:
t 4p
k 5 = 0,9 ⋅ Σl eff,1 ⋅ (3)
m 4x
As
k 10 = 1,1 ⋅ (4)
Lb
The above modification of the stiffness coefficients gives better agreement of design and
experimental values in case of specimen CCB. But in case of specimens CCF with the
formulae (3) and (4) the stiffness is also overestimated. In this case the application of larger
M24 bolts increases extremely the initial stiffness also in case of the thin end-plates, where
the k5 coefficient should have influence. For this reason the reduction of the multiplication
factor of the k10 coefficient in accordance with the bolt diameter gives better coincidence with
the experimental and design values.
In figure 12 the envelope moment-rotation diagrams are shown with the results of the EC3
calculation and the developed modified design method. In case of CCF the initial stiffness is
overestimated as shown in figure 12 a). Good agreement is found in case of CCB1 as figure
12 b) shows.
100 100
75 75
50 50
moment [kNm]
EC3 results
moment [kNm]
25 25 EC3 results
mod. EC3 results
0 0 mod. EC3 results
a) CCF1 b) CCB1
Figure 12. Experimental and design values.
After the EC3 procedure is studied and modified, the evaluation of the monotonic joint
parameters (moment resistance and initial stiffness) is extended to apply them to
characterize cyclic joint behaviour. The original and modified standard procedures are used
to prepare parametric study with the following program:
In figure 13 the moment resistance is presented via the end-plate thickness and shows the
results of the standard EC3 procedure with the proposed modification. Figure 13 indicates
the region of the certain failure modes (Mode1, Mode2 and Mode3). Due to the modification
of the EC3 procedure the limits of the failure modes have a small alteration. The failure
modes of the modified EC3 procedure are always in accordance of the tests results.
100
Mode1 Mode2 Mode3
80
moment [kNm]
60
40
20
0
10 12 14 16 18 20 22 24 26 28 30
end-plate thickness [mm]
EC3 results mod. EC3results
The effect of the bolt grade and the steel material is shown in figure 14 a) and b),
respectively. Figure 14 a) shows the application of the higher bolt grade modifies the limits of
the failure modes by moving the curve right and up. This means that the failure occurred in
thicker end-plate on higher moment level. The application of altering bolt grades does not
change the shape and the tangent of the diagram. If we increase the grade of the steel
material the diagram is moved to left, the failure occurred in thinner plate, but approximately
on the same moment level as figure 14 b) shows.
100 60
80 50
moment [kNm]
moment [kNm]
40
60
30
40
20
20 10
0 0
10 12 14 16 18 20 22 24 26 28 30 10 12 14 16 18 20 22 24 26 28 30
end-plate thickness [mm] end-plate thickness [mm]
M16 8.8 M16 10.9 M16 12.9 S235 S275 S355
In case of the composite section the studied structural details are the flange and the web
thicknesses as figure 15 shows. The application of higher grade of steel material causes
the increase of the tangent of the curve, which means that it has significant effect mainly in
case of larger flange thickness, as figure 15 a) shows. In case of the web the increasing of
the steel grade move the curve up without significant changing of the tangent, so it has
remarkable effect also in case of smaller web thickness (see figure 15 b).
moment [kNm]
200 100
150 75
100 50
50 25
0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
flange thickness [mm] web thickness [mm]
S235 S275 S355 S235 S275 S355
The parametric study includes the investigation of the initial stiffness. Figure 16 a) shows the
effect of the modified k5 and k10 parameters. Coefficient k5 has the influence on the initial
stiffness in case of relatively thin end-plate and k10 in case of thicker plate region. If formula
(3) is used to calculate k5 it decreases the initial stiffness in case of relatively thin plates, but
it has no significant effect on thicker end-plates. Formula (4) of k10 coefficient decreases the
curve initial tangent and move down the original curve. For these reasons the modification of
both coefficients is required to have better agreement between the experimental and design
values.
50 100 M30
x
Sjini [kNm/mrad]
40 M27
75
Sjini [kNm/mrad]
30 M24
M22
20 50
M20
10 M16
25
0 M12
M10
10 15 20 25 30 35 40 45 50 M8
0
end-plate thickness [mm] 10 15 20 25 30 35 40 45 50
EC3 results
mod. k10
mod. k5
mod. EC3 results
end-plate thickness [mm]
The effect of the bolt diameter is presented in figure 16 b). In case of the studied joint the
larger bolt diameters (over M16) have significant effect on the initial stiffness, since it
remarkable increases the stiffness. When the bolt diameter is relatively small the behaviour
in governed by the end-plate.
SUMMARY
The paper presents the comparison of the design and the tested characteristics of bolted
end-plate type joint. The performed test program and the behaviour modes are summarized.
The design methods of Eurocode 3 and 4 are followed to evaluate the moment resistance
and initial stiffness of the end-plate type joint and the composite section. The EC3 gives
method to evaluate monotonic joint behaviour. Since the performed tests are cyclic tests the
envelope curves of the hysteretic diagrams are assumed to cover the monotonic behaviour.
From the comparison it is found that in case of the studied joint the EC3 procedure does not
give good agreement between the design and experimental results. The moment resistance
ACKNOWLEDGEMENT
The research work was conducted under the financial support of the Portuguese – Hungarian
Intergovernmental Science and Technology Cooperation Program ICCTI/OM TÉT P-4/99,
TÉT P-11/01 and the National Scientific Research Foundation OTKA –F 037869.
NOTATION
REFERENCES
(1) Kovács N, Calado L. and Dunai L., (2004). Behaviour of bolted composite joints;
Experimental study, Journal of Constructional Steel Research, Volume 60, Issues 3- 5
March-May 2004. (Special issue: Eurosteel 2002 Third European Conference on Steel
Structures), pp 725-738.
(2) Ádány S., (2001). Numerical and experimental analysis of bolted end-plate joint under
monotonic and cyclic loading, Ph.D. Dissertation, Budapest University of Technology
and Economics, Department of Structural Engineering.
(3) Kovács N., Ádány S., Calado L. and Dunai L., (2001). Experimental Program on Bolted
End-plate Joints of Composite Members, Report ICIST, DTC No. 15/01, Lisbon,
Portugal.
(4) Kovács N, Ádány S, Calado L and Dunai L., (2002). Experimental program on the
bolted end-plate type joints of slender section columns, Report ICIST No. 10/02,
Lisbon, Portugal.
(5) ECCS (1986). Recommended Testing Procedure for Assessing the Behaviour of
Structural Steel Elements under Cyclic Loads, TWG 1.3, No. 45.
(6) prEN 1993-1-8:2003, Eurocode 3: Design of steel structures, Part 1.8: Design of joints
(7) prEN 1994-1-1:1992, Eurocode 4: Design of composite steel and concrete structures,
Part 1.1: General rules and rules for buildings
(8) Ciutina A. L. and Dubina D., (2003). Influence of column web stiffeners on the seismic
behaviour of beam-to-column joints, Proc. of Conference on Behaviour of Steel
Structures in Seismic Areas, STESSA, pp 269-275. Naples, Italy.
ABSTRACT
This paper explores strength, stiffness and rotational capacity rotational in thick
top-and-seat (cleated) angle connections subjected to monotonic and cyclic
loads. The results of test on two full-scale connections are described first, and
are then compared to published curve-fitting models for these types of
connections. The data indicate that the curve-fitting constants of some existing
mathematical models cannot be extrapolated to thick angles. The
experimental results also show that these connections are capable of providing
very ductile behavior and constitute an ideal back-up structural system in steel
frames.
INTRODUCTION
Two major earthquakes in 1994 and 1995, the Northridge and Kobe earthquakes, highlighted
the vulnerability of welded connections in steel moment frames and drew attention to the role
that partial strength, partial restraint connections played in the post-fracture behavior of those
structures. As a result, several research projects were carried out to investigate the
behavior of weak to moderately stiff bolted connections as part of the SAC Program (1). As
part of SAC (2), the behavior of bolted T-stub and top-and-seat angle was studied both at the
component and full-scale connection level. This paper discusses the results of the two full-
scale tests (3,4); the companion component tests have been described elsewhere (5). The
results of these tests, along with other modern tests by Azizinamini et al. (6), are then
compared with the predictions from several curve-fitting models available in the literature
(7,8). The latter include the Fry and Morris (1975) model based on odd-power polynomial
formulas (9), the Richard and Abbott (1975) four-parameter model (10), the Chen and Lui
(1985) exponential model (11), and the Kishi and Chen (1986) linear model (12,13). None of
these models limits the rotational capacity, as they are based on tests with flexible angles in
which a full plastic mechanism develops in the angles. This paper investigates the
extension of these models to situations where thick angles may lead to tension bolt failures.
EXPERIMENTAL PROGRAM
Two full-scale top-and-seat angle connections were tested as part of the SAC project (4).
As there was extensive data available on the performance of connections with thin angles (t
< 16 mm) and shallow beams (d < 360 mm) subjected to cyclic loads (3), the emphasis was
on testing connections with thicker angles (t = 25 mm) and deeper beams (d = 460 mm). The
specimens consisted of FE 350 (Grade 50) W460x60 (W18x40) beams bolted to W360x216
4000
400
2000
200
Moment (kN-m)
Moment (kip-in)
0 0
-200
-2000
-400
-4000
-0.04 -0.02 0.00 0.02 0.04
4000
400
3000
2000
200
Moment (kN-m)
Moment (kip-in)
1000
0 0
-1000
-200
-2000
-3000
-400
-4000
-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06
Specimen FS-02 (Figure 2) produced a tension bolt failure as in FS-01. However, higher
displacements were reached due to the movement of the tension bolts away from the k-line
of the angles. Much more visible prying action was observed and less damage to the beam
was incurred. While plastic hinges did form above the k-line of the angle leg to the beam,
very little yielding occurred near the tension bolt line in the column. The angles in these tests
were extremely thick relative to the size of the beam and the tension bolts were clearly not
strong enough to allow complete formation of the plastic hinges in either the beam or angles.
CURVE-FITTING MODELS
In the USA, when top-and-seat connection capacities are required, it is common to resort to
published equations for moment-rotation curves. The latter are derived from curve-fitting to
test data. A good curve-fitting model requires a simple formulation that represents, to some
degree, the mechanics of the problem, but uses a minimum number of parameters to achieve
good results. The models proposed by Ang and Morris (1974), Frye and Morris (1975),
Richard and Abbott (1975), Chen and Lui (1985), and Chen and Kishi(1986) cover a wide
range of connection types (9-13). The data used, however, are typically from tests on small
specimens and some data relate to very old tests in which the degree of pretension and
actual material properties is uncertain. Figure 3 shows the standardized parameters
typically used for the mathematical models described in Table 1.
f (Fastener Dia.) = w lt = l
tt = t Flange Angle
Top angle
ga
kt gt
Web Angle
la d
ks p gs
Seat angle
Table 2 gives details of eight modern specimens tested under monotonic and cyclic loads. Six
specimens come from Azizinamini (6), with companion specimens at two thickness (10 mm
and 13 mm) tested under both static and cyclic loads (14S1/C1 and 12S2/C2); the other two
specimens represent different web detail connections (14S3 and 14S4). The remaining two
Standardized Ramberg-Osgood
General Form of the
φc = C1(KM) +C2(KM) +C3(KM)
1 3 5
function
model
[φc/φo] = [KM/(KM)o]×[1+(KM/(KM)o)n-1]
Calibration Tests Rathbun(1936), Mains(1944) Hetchman and Johnston(1947)
Power Model
Model
Three Parameter Four Parameter
Developer Chen and Kishi (1987) Richard and Abbott (1975)
Rki, Mu,
Parameters n=0.827((φo<1.9x10-3rad) kp, k, Mo, n=2
n =1.398log10φo+4.93(φo>1.9x10-3rad)
lt, tt, kt, gt, w, d, ls, ts, ks, gs, da, ta, ka, lt, tt, kt, gt, w, d, ls, ts, ks, gs, da, ta,
Geometric Parameter
ga, la (Including web angle) ka, ga, la (Including web angle)
General Form of the M=[(k-kp) φc] / [(1+|(k-kp) φc / Mo|n )n-1]
φc = (M/Rki)⋅(1/(1-(M/Mu)1/n))
model + kpφc
Calibration Tests Azizinamini et al.(1985)
Specimen Type of Beam Top and Bottom Flange Angle Web Angle
Moment(K-in)
.
500
Moment(K-in)
700
400 600
500
300 14S1 14S2
400 F- M M o.
F- M M o.
200 300 A- M M o.
A- M M o.
200 C- K M o.
C- K M o.
100 100 R- A M o.
R- A M o.
0 0
0 0.01 0.02 0.03 0.04 0 0.01 0.02 0.03 0.04
Rot at i on( r ad) Rot at i on( r ad)
800 1000
700 900
800
600
700
.
.
Moment(K-in)
Moment(K-in)
500 600
400 500
14S3 400 14S4
300 F- M M o.
F- M M o.
300
200 A- M M o. A- M M o.
200 C- K M o.
C- K M o.
100 100 R- A M o.
R- A M o.
0 0
0 0.01 0.02 0.03 0.04 0 0.01 0.02 0.03 0.04
Rot at i on( r ad) Rot at i on( r ad)
The four specimens subjected to static load shown in Table 2 will be first compared using four
models (Figure 4). For the Ang and Morris model, regardless of the dimension of each
specimen, an identical reference moment and rotation angle are used. This model is
capable of providing a good, if somewhat larger, estimate of initial stiffness in spite of
neglecting the effect of the double web angles. On the other hand, the Fry and Morris model
consistently underestimates both strength and stiffness of the connections. The power
models have parameters based on the specimen’s dimensions and show good agreement
with the experimental moment-rotation curves even in the nonlinear range. When n, the
shape parameter, is increased, the behavior of the connection becomes closer to an
idealized elasto-plastic model. The connection has no further moment resistance when the
moment approaches ultimate moment (Mu). The Richard and Abbott model needs four
parameters: the initial stiffness (k), the strain hardening stiffness (kp), a reference moment
(Mo) and a parameter n defining the shape of the curve. These values are shown in Table 3
based on the results of the Azizinamini tests.
14C1 14C2
600 800
600
400
400
.
.
Moment(K-in)
Moment(K-in)
200
200
0 0
-0.015 -0.01 -0.005 0 0.005 0.01 0.015 -0.015 -0.01 -0.005 0 0.005 0.01 0.015
-200
-200
F- M M o. F- M M o.
-400
-400 A- M M o. A- M M o.
C- K M o. -600
C- K M o.
-600 -800
Rot at i on( r ad) Rot at i on( r ad)
3000 3000
2000 2000
.
.
Moment(K-in)
Moment(K-in)
1000 1000
0 0
-0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 -0.04 -0.02 0 0.02 0.04 0.06
-1000 -1000
-2000 F- M M o. F- M M o.
-2000
A- M M o. A- M M o.
-3000 -3000
C- K M o. C- K M o.
-4000 -4000
Rot at i on( r ad) Rot at i on( r ad)
The use of models developed for monotonic loading to predict cyclic behavior is based on the
observation that the envelopes of cyclic tests match closely the envelope for static tests; i.e.,
the behavior of specimens 14S1 and 14S2 match closely the envelope from 14C1 and 14C2
respectively. For the thick clip angle specimens (FA-01 and FA-02), the Chen and Kishi
Tables 4 through 7 show a comparison between the experimental results and the
mathematical models in terms of both the stiffness and moment. Four parameters are used
for comparisons in these tables. First, *K-secan is the elastic stiffness of a secant line to 4.0 x
10-3 rad resulting from a polynomial model or a power model. Similarly, **M-secant is moment
at 4.0 x 10-3 rad. After reaching the yield moment, the slope of the moment rotation curve
remains constant. A slope parameter (***Kp) was measured at 24x10-3 rad and used in the
comparisons of the monotonic tests. To compare with experimental curves after yield, the
static moment at 24x10-3 rad was used. For the cyclic tests with thick angles, the moment at
20x10-3 rad was used as this is still in the hardening range of the envelope. The Richard-
Abbott model requires parameters obtained from experimental results, and thus its results
can also be included in the comparisons.
14S1 746 x107 39.3% 298 x105 39.3% 111 x107 -68.9% 657 x105 13%
7 5 7 5
14S2 862 x10 49.7% 345 x10 49.7% 121 x10 15% 759 x10 28.8%
7 5 7 5
14S3 746 x10 25.6% 298 x10 25.6% 111 x10 -36.1% 657 x10 10.9%
7 5 7 5
14S4 746 x10 46.7% 298 x10 46.7% 111 x10 15.3% 657 x10 29.3%
7 5
14C1 746 x10 49.2% 298 x10 49.2% - - - -
7 5
14C2 862 x10 38.5% 345 x10 38.5% - - - -
8 5 6
FS01CA01 207 x10 61.5% 827 x10 61.5% - - 170 x10 58.3%
8 5 6
FS01CA02 207 x10 54.8% 827 x10 54.7% - - 170 x10 50%
Ave(%) 45.65% 45.65% -18.7% 31.72%
Var(%) 10.42% 10.42% 35.75% 17.49%
14S1 162 x108 -31.7% 647 x105 -31.7% 120 x107 -82% 115 x106 -53%
8 5 7 6
14S2 190 x10 -10.4% 757 x10 -10.4% 119 x10 15.8% 135 x10 -26.2%
8 5 7 6
14S3 162 x10 -61.4% 647 x10 -61.4% 120 x10 -47.2% 115 x10 -56.7%
8 5 7 6
14S4 162 x10 -15.5% 647 x10 -15.5% 120 x10 27.7% 115 x10 -24.3%
8 5
14C1 162 x10 -10.2% 647 x10 -10.2%
8 5
14C2 190 x10 -9.8% 757 x10 -9.8%
8 6
CA01 432x10 24.3% 173 x10 24.3% 307 x106 24.4%
8 6 6
CA01 432x10 6% 173 x10 6% 307 x10 9.3%
Ave(%) -13.6% -13.6% 21.42% -21.08%
Var(%) 23.8% 23.79% 60.71% 29.76%
14S1 111 x108 9.2% 443 x105 9.2% 231 x106 64.7% 759 x105 -0.005%
8 5 6 6
14S2 194 x10 -13.3% 775 x10 -13.3% 446 x10 68.6% 118 x10 -10%
8 5 6 5
14S3 162 x10 -62.5% 399 x10 -62.5% 257 x10 68.3% 655 x10 11%
7 5 6 5
14S4 997 x10 28.8% 574 x10 28.8% 378 x10 59.7% 958 x10 -3.2%
8 5
14C1 111 x10 24.6% 443 x10 24.6%
8 5
14C2 194 x10 11.3% 775 x10 11.3%
9 6
CA01 124 x10 -132% 497 x10 -132% 629 x106 -54.5%
9 5 6
CA01 127 x10 -176% 506 x10 -176% 643 x10 -89.6%
Ave(%) 38.73% 38.73% 65.5% -24.3%
Var(%) 72.58% 72.58% 3.6% 33.2%
CONCLUSIONS
Except for the Richard – Abbott model, which has to show good agreement by definition
since its constants were based on the data used for the comparisons, large discrepancies
between the experimental results and the models are obvious. In general, the results for
thick clip angles evidence a larger percent difference than those for thin angles. As the table
and figure indicate, the Chen-Kishi model based on Azizinamini’s experimental data shows
good agreement for the thin to medium clip angle cases. However, it does not provide
satisfactory results for the behavior of the thick clip angle connections. Most important, none
of the curve-fitting models is capable of predicting the rotational capacity when the tension
capacity of the bolts governs the behavior. Ongoing work will compare these conclusions with
results from both Eurocode and another polynomial models developed by Citipitioglu et al.
(14).
REFERENCES
1 Liu, J., and Astaneh-Asl, A. (2000). Cyclic Tests of Simple Connections Including the
Effects of Slabs, J. of Structural Engineering, ASCE, v. 126, n. 1, pp. 32-39.
2 Swanson, J.A., and Leon, R.T. (2002). Bolted Steel Connections: Tests on T-stub
Subcomponents, J. of Structural Engineering, ASCE, v. 126, n. 1, pp. 50-56.
3 Leon, R.T., Swanson, J.A., Schrauben, C., and Smallidge, J. (2000). Experimental Test
ABSTRACT
An experimental study was conducted to investigate the seismic behavior of
reduced beam section (RBS) moment connections to a deep wide flange
column. The test matrix for the experimental program consisted of six full-scale
interior RBS connections, where the column for the specimens ranged in depth
from a W24 to a W36 wide flange section. All but one of the specimens had a
composite floor slab. The results from the study show that a composite floor
slab provides restraint to the top flange of the beams; reducing the magnitude
of beam top and bottom flange lateral movement in the RBS, column twist, and
strength degradation due to beam instability in the RBS. The performance of
each of the test specimens was found to meet the seismic connection
qualification criteria in Appendix S of the AISC Seismic Provisions, and thereby
have sufficient ductility for seismic resistant design. The results of the
experimental study, along with a nonlinear finite element study were used to
develop seismic design recommendations for RBS connections to deep
columns.
INTRODUCTION
RBS beam-to-column moment connections are often utilized in the design of special steel
moment resistant frames (SMRFs). The details of a typical RBS connection are shown in
Figure 1(a), where the flanges of the beam are reduced in width, away from the column face.
Complete joint penetration (CJP) groove welds attach the beam flanges to the column. The
beam web often is welded to the column flange with a CJP groove weld. By design, the RBS
connection develops inelastic deformations primarily in the region where the beam flange
width has been reduced (referred to herein as the RBS), limiting the inelastic strain developed
in the beam flange-to-column CJP groove welds. With the reduction of the beam flange width,
an RBS connection is more prone to inelastic local buckling of the beam web and flanges in
the RBS. For economical reasons, design engineers in the U.S. prefer to use deep columns
in SMRFs (as large as 914 mm in depth corresponding to a W36 wide flange section) in order
to control seismic drift. Previous tests on RBS connections have been performed primarily on
columns with depths corresponding to a W12 and W14 wide flange section (Roeder (1)),
where the depth was about 305 mm to 356 mm. Some tests using W27 wide flange column
sections (686 mm depth) were conducted by Chi and Uang (2), where the connection was an
exterior connection (i.e., only one beam was connected to the column). It was observed in
these tests, that as a result of inelastic beam web and flange local buckling in the RBS, a
lateral displacement of the beam compression flanges occurs. Shown in Figure 1(b) is the
movement of the compression flanges (the top and bottom flanges of the right and left-hand
beams, respectively), where F1 and F2 represent the beam flange compression forces of the
TEST MATRIX
The test matrix for the experimental program is given below in Table 1, where some of the
details of the six full-scale RBS beam-to-deep column connection specimens are
summarized. All specimens represented an interior RBS connection in a perimeter SMRF
with a composite floor slab, with the exception of SPEC-6 which did not have a composite
floor slab. The parameters investigated in the experimental program included: (1) column
size; (2) beam size; (3) the floor slab; and (4) supplemental lateral brace at the end of the
RBS.
The beam and column section sizes for each specimen were selected on the basis of
introducing different degrees of torsional effects, predicted by the recommended design
procedure of Chi and Uang (2), while also satisfying the weak beam-strong column criteria in
the ASIC Seismic Provisions (4). The design procedure by Chi and Uang considers the total
normal stress in the column at 4% story drift due to axial load, flexure load, and torsion. The
predicted total normal stress in the column flange is shown plotted in Figure 2 for various
column sections, including those of the test specimens. Figure 2 indicates that SPEC-2,
SPEC-4, and SPEC-5 are predicted to develop column flange yielding. The columns for all
specimens and the beams for SPEC-3 through SPEC-6 were fabricated from A992 steel. The
beams for SPEC-1 and SPEC-2 were fabricated from A572 Gr. 50 steel. Both A992 and A572
Gr. 50 have a nominal yield strength of 345 MPa.
6"
No Run- Beam
300 6" 6" off Tabs
6"
SPEC-6 W24x131 E70T-1
(supplemental bracing) E71T-8
200
Figure 2. Column total stress per Chi and Uang Figure 3. Specimen typical connection
(2) versus column section weight. details (Note: 1 inch = 25.4 mm).
The elevation of a typical connection detail is shown in Figure 3. Each specimen was
designed in accordance with the criteria recommended by Engelhardt (5) for RBS
connections, where the design moment in the beam at the column face is limited to Mpn,
where Mpn is the nominal plastic capacity of the beam. For the six specimens in the test
matrix, the average value of the beam design moment at the column face was equal to
0.973Mpn. The reduction in flange width at the center of the RBS for each specimen was 50%
of the original flange width, which complied with the design criteria of Engelhardt. The RBS
was flame cut, with the burned surface ground to a surface roughness of 500 micro-inches,
as recommended by FEMA 353 (6). Each specimen had continuity plates the same thickness
as the beam flanges and designed for a balanced panel zone condition. Complete details are
given in Ricles et al. (3). The weld procedure specifications used in the fabrication of the
connections were prequalified in accordance with AWS D1.1/D1.1M:2002 (7). All welds were
done using the flux core arc welding procedure, and conformed to the AWS 5.20-95
Specification (8). The beam flange-to-column flange CJP groove field welds and beam web-
to-column CJP groove field welds utilized E70T-6 and E71T-8 electrodes, respectively. All
shop welds (e.g., shear tab to the column, doubler and continuity plates) were performed
using E70T-1. The run off tabs for the beam flanges were removed following the placement of
The specimen composite floor slab had a total thickness of 133 mm, and consisted of 27.6
MPa nominal compressive strength concrete cast on a 20-gage zinc coated metal deck. A
W4x4 welded wire mesh with wire 152 mm on center was placed in the floor slab prior to
pouring the concrete. The width of the floor slab was 1220 mm to one side, with a 305 mm
overhang on the other side to simulate the conditions for a perimeter SMRF. The ribs of the
decking ran parallel to the main beam (i.e., the beams with the RBS connections) of each test
specimen. To develop the composite action, 19 mm diameter shear studs were placed
outside the RBS region at 305 mm spacing along the beams to attach the deck to the main
beams as well as transverse W14x22 floor beams. These transverse beams were placed at a
spacing of 3048 mm to provide lateral bracing to the main beams and column, where the
distance of 3048 mm satisfied the AISC Seismic Provisions (4).
SPEC-6, which had no composite floor slab, had a supplemental lateral brace at the end of
the RBS in addition to the other lateral bracing noted above for the beams. The lateral
bracing was attached to a W36x150 section that was placed parallel to the beams of the test
specimen to simulate a parallel beam in the prototype building. This parallel beam in the test
setup was allowed to move horizontally with the test specimen, but restrained from out-of-
plane movement. The corresponding stiffness of the lateral bracing setup satisfied the AISC
LRFD Specification (9). SPEC-3 also had supplemental lateral braces, but these were
anchored in the floor slab.
The test setup is shown in Figure 4 (a), with the lateral bracing detail given in Figure 4(b) for
the main beams. The ends of the members in the test setup had pin-connected boundary
conditions, using cylindrical bearings to simulate inflection points at the beam midspan and
column midheight in the prototype frame. The ends of each beam away from the column were
supported by instrumented rigid links, which simulated a roller boundary condition and
enabled horizontal movement of the end of each beam. The lateral bracing detail shown in
Figure 4(b) was used to prevent out-of-plane movement of the beams and column (the
diagonal double angles were not used at the column), and designed for strength and stiffness
in accordance with the AISC LRFD Specification (9). The top of the column was braced
against torsion, while at the base of the column a clevis was used to create the pin boundary
condition. The beams were also braced at the rigid links in order to stabilize the test setup.
The torsional bracing provided at both ends of the column in the test setup was evaluated
using a nonlinear finite element model (3) to examine whether the stiffness would be
representative of the torsional restraint at the column inflection points in the prototype
structure. It was found to be satisfactory and not influence the test results by over-restraining
the ends of the column from twisting.
The specimens were tested by imposing a cyclic story drift history based on the loading
sequence defined in Appendix S of the AISC Seismic Provisions (4). The loading protocol
consisted of initial elastic cycles of story drift, followed by cycles of increased amplitude to
cause inelastic response. A test was terminated when either a fracture occurred, resulting in
21"
Load Cell
Actuator
Column Shear Stud
Floor Beam 12" 48" Floor Slab
(North Side Only) Floor Beam(North Side Only)
6'-6"
51 4"
2
A325
3
4"
diam.
Beam (East) Double W14x22
Beam (West) (TYP)
6'-6"
14'-9" 14'-9"
29'-6"
Setup Lateral Bracing
Figure 4. (a) Test setup and (b) beam lateral bracing detail for specimens with a
composite floor slab (Note: 1 inch = 25.4 mm).
TEST RESULTS
A summary of test results for each specimen is given in Table 2, where Rv/Vpz, θmax, Mf/Mpn,
Kφ,col, φ, δflg, and bf are equal to the ratio of panel zone shear capacity-to-panel zone shear
force corresponding to the plastic flexural moment developing in the RBS, specimen drift from
the last cycle prior to any fracture or strength deterioration to below 80% of the specimen
nominal capacity, ratio of maximum measured beam moment developed at the column face-
to-nominal beam flexural capacity, column elastic torsional stiffness, specimen column twist
at 4% story drift, lateral displacement of the beam bottom flange at the RBS at 4% story drift,
and beam flange width, respectively. Typical observed behavior during the testing of a
specimen consisted of yielding in the RBS and the panel zone, followed by cyclic local web
and flange buckling in the RBS. Following the development of local bucking in the RBS,
lateral movement of the bottom beam flange began to occur in the RBS of specimens with a
composite floor slab at 2% to 3% story drift. The combined effect of cyclic local buckling and
lateral flange displacement resulted in a gradual deterioration in specimen capacity to occur
during subsequent cycles where the story drift amplitude was increased. This is evident in the
lateral load-story drift hysteretic response of SPEC-4 shown in Figure 5. The lateral
displacement of the bottom beam flange occurred when it was in compression, and caused
some column twist to develop. Figure 7(a) and (b) shows photographs of SPEC-4 at 4% and
6% story drift, where the yielding in the members and panel zone in the connection region
and lateral beam flange movement in the RBS are visible. The maximum column twist among
the specimens with a floor slab at 4% story drift was 0.037 rads. (SPEC-4). 4% story drift is
the drift at which connections are judged for qualification for seismic use by the AISC Seismic
Provisions (4). SPEC-4, like the other specimens, developed a flange fracture in the RBS
where extensive local flange buckling had occurred (see Figure 7(c)). This occurred at a story
drift of 6%, and was caused by local buckling in the beam flange that led to large cyclic
strains, resulting in a low cycle fatigue failure. SPEC-6, which had a supplemental brace and
lateral bracing attached to the beam that is parallel to the test beam, had minimal
deterioration in capacity as well as column twist (0.004 rads. at 4% story drift), see Figure 6.
1500 1500
1000 1000
500 500
0 0
-500 -500
-1000 -1000
-1500 -1500
-2000 -2000
-6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6
Story Drift (% rad.) Story Drift (% rad.)
fracture
(b) Beam bottom flange lateral movement (c) Beam bottom flange fracture
at RBS, 4% story drift at RBS, 6% story drift
Figure 7. Photographs of SPEC-4 during testing.
The reduced amount of deterioration in the capacity of SPEC-6 was due to the specimen
having a weaker panel zone than the other specimens. As noted in Table 2, for SPEC-6 the
ratio of Rv/Vpz is equal to 1.03. Rv is based on the ASIC Seismic Provisions (4). All other
specimens have a value of 1.14 or greater for the ratio of Rv/Vpz. Consequently, these other
specimens developed a larger amount of yielding and local buckling in the RBS than SPEC-6,
leading to local buckling and deterioration in specimen capacity.
In Table 2, all specimens are shown to have a value for θmax that exceeds 0.04 rads., which is
the current criteria in Appendix S of the AISC Seismic Provisions (4) for qualifying a
connection for seismic use. A summary of the ratio of Mf/Mpn in Table 2 indicates that the
maximum beam moment developed at the column face in the specimens exceeded the
design value of Mpn for which the specimens were designed, with SPEC-5 having the
maximum value of 1.2. The increase in the moment Mf is attributed to the composite floor
slab increasing the moment capacity in the RBS. SPEC-6 had a valve of Mf equal to 1.0, and
had no composite floor slab.
6000
SPEC-4
SPEC-6
4000
Strain (microstrain)
2000
εy=1765 µε
0
-2000
-4000
-6000
-150 -100 -50 0 50 100 150
Distance across column flange (mm)
An examination of the results for column twist φ in Table 2 reveals that column twist tends to
increase when the elastic torsional stiffness of the column Kφ,col is reduced. However, for a
smaller beam section size φ is reduced, although the column torsional stiffness is smaller
(e.g., SPEC-5). This phenomenon is associated with a smaller demand on the column when
a smaller beam is used. The column twist is reduced significantly in specimens with a
supplemental brace (SPEC-3 and SPEC-6). The reduction in column twisting in SPEC-6 is
also attributed to a weaker panel zone, which reduced the amount of yielding and local
buckling in the RBS, and subsequently less lateral movement in the RBS. An examination of
the measured specimen beam flange lateral displacement δflg in Table 2 shows these results
to be less than the value of 0.2bf, which is the value recommended by Chi and Uang (2) for
determining the design torque T applied to the column. Consequently, the use of the value of
0.2bf for determining the design torsional loading on the column from the RBS will result in a
larger column design torque. This is evident by comparing the column total normal stress at
the connection based on Chi and Uang’s recommendation with the measured specimen
response (see Table 3). The criterion by Chi and Uang anticipates column flange yielding
occurring in SPEC-2, 4, and 5; see Figure 2, where the nominal yield stress is 345 MPa. The
measured column flange longitudinal strains in these specimens indicated no yielding in
SPEC-2 and 5, with some minor yielding occurring in SPEC-4 (a maximum strain of 2 to 4
times the yield strain developed). The measured longitudinal strains across the column flange
just below the connection are shown in Figure 8 for SPEC-4 and SPEC-6. These results are
DESIGN RECOMMENDATIONS
The strains in the beam bottom flange near the column face were examined to evaluate the
stress distribution across the beam flange that leads to a torque T applied to the column.
Shown below in Figure 9(a) is the distribution of longitudinal stress across the beam bottom
flange at 4% story drift. These stresses are based on measured longitudinal strains in the
specimens. These results correspond to a negative beam moment at the column face (i.e.,
when the bottom flange of the beam is in “compression”). Similar results for longitudinal
stress across the beam compression flange were obtained from finite element studies (see
Figure 9(b)). The results in Figure 9 show a trend where the stress distribution across the
beam flange has a reduction in stress, which is due to a moment in the plane of the beam
flange caused by the lateral movement of the beam flange at the RBS. This moment is
equivalent to the torque T that is applied by the beam flange to the column. Shown in Figure
10(a) is an idealized uniform longitudinal stress distribution prior to lateral movement of the
beam flange in the RBS (at 2% story drift). The idealized longitudinal stress distribution at 4%
story drift based on the measured and finite element analysis results is given in Figure 10(b).
At 4% drift local buckling and lateral beam flange movement has occurred in the RBS.
Elastic-perfectly stress-strain behavior is assumed in Figure 10, where Fye is the yield stress.
400 100
SPEC-1
300
SPEC-2
SPEC-3 0
(b)
SPEC-4
Longitudinal stress (MPa)
200 SPEC-5
-100
Stress (MPa)
100
0 -200
-100
-300
-200
(a)
-400
-300
-400 -500
-150 -100 -50 0 50 100 150 -150 -100 -50 0 50 100 150
Distance across beam flange (mm) Distance across beam flange (mm)
Figure 9. Longitudinal stress distribution across beam flange for (a) all test
specimens, and (b) finite element analysis of SPEC-2.
Fye Fye
(a) (b)
Figure 10. Idealized longitudinal stress distribution across beam bottom flange at
(a) 2% story drift and (b) 4% story drift.
where Fye, bf, and tf are equal to the expected beam flange yield stress (1.1Fy), the beam
flange width, and beam flange thickness, respectively.
A design procedure was thus developed in order to determine the total design longitudinal
stress ftotal in the column flange that is attached to an RBS connection. The procedure
involves determining the elastic warping normal stresses fw that develop in the column flange
due to the torque T (10) and superimposing them with the column flange normal stresses
due to bending (fb) and axial loading (fa) to obtain the total normal stress ftotal, where
In Equation (2) E, WnO, and θ” are equal to the Young’s modulus, normalized warping
function at the column flange tip (10), and the second derivative of the angle of twist in the
column (10), respectively, where θ” is a function of the torque T.
The total stress ftotal is compared to the criteria in the AISC LRFD Specification (9), Equation
(H2-1), where
in which φ and Fy are the resistance factor (0.9) and nominal yield stress of the column
flange, respectively. The above design procedure is similar to that developed by Chi and
Uang (2), except for the method in which the torque is determined.
Table 3. Comparison of column normal flange compression stresses with design procedure.
Experimental
Warping
Axial Total normal results, total
stress fw
load Bending stress ftotal (MPa) stress & strain,
(MPa)
SPEC Column Beam stress stress 4% story drift
fa fb (MPa) Chi Chi
Pro- Pro- Strain Stress
(MPa) and and
posed posed (µε) (MPa)
Uang Uang
1 W36x230 0 190 128 66 318 256 1277 255
2 W27x194 0 299 182 101 481 400 2151 372(1)
W36x150
3 W27x194 0 332 0 0 332 332 1797 356(1)
4 W36x150 0 337 321 163 658 500 3296 365(1)
5 W27x146 0 252 180 95 432 347 1598 319
W30x108
6 W24x131 0 347 0 0 347 347 2525 334(1)
Note: (1) Yield stress of the column flange.
The total normal column flange stress based on the above procedure is compared in Table 3
to the measured stress of the test specimens, as well as the stress predicted using the
procedure by Chi and Uang (2). The comparisons in Table 3 indicate that a more accurate
prediction of the total normal stress in the column flange is made using the above procedure
compared to the procedure developed by Chi and Uang (2). The difference between the two
Based on the experimental study, the following main conclusions are noted:
1. A composite floor slab can significantly reduce the lateral displacement of the beam
bottom flange in the RBS and the amount of twist developed in the column. The slab
appears to be effective in reducing the twist in deeper columns attached to an RBS
connection, and enables the cyclic strength of the beam with an RBS connection to
be better sustained.
2. All of the specimens were able to satisfy the criteria in the AISC Seismic Provisions
(4) for qualifying the connection for seismic use.
3. A weaker panel zone in a deep column RBS connection will not develop as much
column twist and strength degradation as a connection with a stronger panel zone.
However, a weaker panel zone can significantly increase the potential for ductile
fracture of the connection (3). It is recommended that connections be designed with a
balanced panel zone strength condition.
4. A supplemental brace at the end of the RBS significantly reduced the transverse
movement of the beam flanges in the RBS and column twist that leads to cyclic
degradation in specimen capacity.
5. Basing the column torque on a transverse movement of the beam flange in the RBS
of 0.2bf for calculating column flange warping stresses appears to be conservative. A
new procedure for estimating the torsional load applied to the column due to the local
and lateral buckling in the RBS shows improvement in predicting the correct column
flange normal stress.
ACKNOWLEDGEMENTS
The research reported herein was supported by a grant from the American Institute of Steel
Construction (Mr. Tom Schlafly program manager) and from the Pennsylvania Department of
Community and Economic Development through the Pennsylvania Infrastructure Technology
Alliance (PITA) program. The following companies donated materials for the experimental
testing conducted in this research project: Arcelor International America of New York, NY
(steel sections); Nucor Vulcraft Group of Chemung, NY (metal decking); and the Lincoln
Electric Company of Cleveland, OH (welding wire). The support provided by the funding
agencies and companies is greatly appreciated.
(1) Roeder, C. W. (2000). “Connection Performance State of Art Report,” Report No.
FEMA-355D, FEMA, Washington, D.C.
(2) Chi, B. and Uang, C.-M. (2002). “Cyclic Response and Design Recommendations of
Reduced Beam Section Moment Connections with Deep Columns,” Journal of
Structural Engineering, ASCE, 128(4): 464-473.
(3) Ricles, J., Zhang, X., Lu, L.W., and J. Fisher, (2004). “Development of Seismic
Guidelines for Deep-Column Steel Moment Connections,” ATLSS Report No. 04-13,
ATLSS Engineering Research Center, Lehigh University, Bethlehem, PA.
(4) “Seismic Provisions for Structural Steel Buildings,” (2002). American Institute of Steel
Construction, Chicago, Illinois.
(5) Engelhardt, M. D. (1999). “The 1999 T. R. Higgins Lecture: Design of Reduced Beam
Section Moment Connections,” Proceedings: 1999 North American Steel Construction
Conference, American Institute of Steel Construction, Toronto, Canada, pp. 1-1 to 1-29.
(6) “Recommended Specifications and Quality Assurance Guidelines for Steel Moment-
Frame Construction for Seismic Applications,” (2000). Report No. FEMA 353, Federal
Emergency Management Agency, Washington D. C.
(7) “Structural Welding Code – Steel,” (2002). AWS D1.1/D1.1M:2002, American Welding
Society, Miami, Florida.
(8) “Specification for Carbon Steel Electrodes for Flux Cored Arc Welding,” (1995).
ANSI/AWS A5.20-95, American Welding Society, Miami, Florida.
(9) “Manual of Steel Construction-Load and Resistance Factor Design,” (2001). Third Ed.,
AISC, Chicago, Illinois.
(10) Seaburg, P., and C. Carter, (1997). “Torsional Analysis of Structural Steel Members,”
American Institute of Steel Construction Steel Design Guide Series, ASIC, Chicago,
Illinois.
ABSTRACT
Eccentrically braced steel frames represent a suitable solution for multi-storey
buildings located in seismic areas. A bolted connection between the link and
the beam is suggested to facilitate replacement of damaged dissipative zones
(links) after a moderate to strong earthquake, which reduces repair costs. A full
scale testing program was carried out in order to demonstrate the feasibility of
this concept and to evaluate the performance of bolted links. The paper
summarises the results of the testing program.
INTRODUCTION
ed
e
On the other hand, capacity-based design, applied in most of the current seismic design
codes allows design of structures that promote plastic deformation in predefined areas only,
called dissipative zones. In the case of a bolted connection between dissipative zones and
the rest of the structure it is possible to replace the dissipative elements damaged as a result
of a moderate to strong earthquake, reducing the repair costs. Application of this philosophy
to eccentrically braced frames, where link elements serve as dissipative zones, is presented
in figure 1. The connection of the link to the beams is realized by a flush end-plate and high-
strength bolts. Bolted connection allows he link element to be fabricated from a lower-yield
steel grade, assuring an elastic response of the elements outside removable link element.
This system may be applied to both homogeneous structures (eccentrically braced frames
alone) and dual ones (eccentrically braced spans combined with moment-resisting spans).
The latter system has the advantage of more uniform transient and smaller permanent lateral
displacements, which is beneficial for replacing damaged links, as well as for the building
function, Stratan and Dubina, 2002 (2). Extended end-plate bolted connections for
EXPERIMENTAL PROGRAM
An experimental program was carried out to determine cyclic performance of bolted links and
to check the feasibility of the suggested solution. The removable link was fabricated from
IPE240 profile of S235 grade steel, while the rest of the structure – from S355 grade steel.
Four link lengths were considered (e=400, 500, 600 and 700 mm, see figure 1), to study the
influence of moment to shear force ratio. All links are classified as short ones according to
AISC, 1997 (4). Another parameter considered was the spacing of web stiffeners, provided to
prevent web buckling and to improve rotation capacity of the link. Two limit values of stiffener
spacing were considered to AISC, 1997 (4): "close" spacing - 30tw-h/5, for a rotation capacity
0.08 rad, and "rare" spacing - 52tw-h/5, for 0.02 rad rotation capacity.
support
actuator
link
For combination of link length and stiffener spacing, three specimens were tested: one
monotonically and two cyclically. A total of 24 specimens were thus obtained, each being
denoted as L[x][n]-[t], where: [x] – L for "rare" spacing of stiffeners, H for "close" spacing of
stiffeners; [n] – 7, 6, 5, 4 for link length; [t] – m for monotonic, c1 for cyclic 1 and c2 for cyclic
2 specimens. Thus, LL7-m specimen is one with rare spacing of stiffeners (L) of length 700
mm (7), monotonically loaded (m). The complete ECCS 1985 (5) loading procedure was
applied, consisting of one monotonic and two cyclic tests. The monotonic test was used to
determine the yield force Fy and displacement Dy, at the intersection of the initial stiffness
and the tangent to the F-D curve having 10% of the initial stiffness. Yield displacement was
determined for each monotonic test with rare stiffeners, and used to apply cyclic loading to
the specimens of the same length. The cyclic tests consisted of four cycles in the elastic
range (±0.25Dy, ±0.5Dy, ±0.75Dy and ±1.0Dy), followed by groups of three cycles at
amplitudes multiple of 2Dy (3x±2Dy, 3x±4Dy, 3x±6Dy, etc.) The loading was applied quasi-
statically, in displacement control. Bolts were preloaded to 100% of the full preload value for
friction-grip bolts in the case of the monotonically loaded (m) and the first of the cyclically
loaded (c1) specimens, and 50% for the second cyclically loaded specimen (c2).
Previous experimental research on beam-column joints with end plates, Dubina et al., 2000
(6) showed a series of problems that undermined their cyclic performance: (1) fillet welds are
inappropriate in the case of cyclic loading; (2) full-penetration 1/2V weld with the root at the
exterior part of the beam cross-section promotes fragile ruptures, due to cracks initiated at
weld root; (3) weld-access hole acts as a stress concentrator, causing brittle ruptures of the
Table 1. Characteristics of the materials used for fabrication of removable link specimens.
component fy (Reh), N/mm2 fu, N/mm2 fu/fy A, %
IPE240 flange 268.0 401.9 1.50 29.2
IPE240 web 337.8 426.7 1.26 30.8
t=25 250.8 413.1 1.65 36.3
Standard tensile tests were performed on coupons extracted the materials used to fabricate
the link specimens. Results presented in table 1 revealed a higher yield strength of the web
in comparison with flanges of the link.
Design of connections
Bolted connection between the link element and the beam is located in a zone of maximum
stresses. There are two possible strategies for connection design. The first one is to provide
a sufficient overstrength of the connection over the link shear resistance. The second one is
to assure a ductile behaviour of the bolted connection itself. The former strategy was
followed in this case, as it facilitates replacement of damaged link elements.
Capacity design of the connection involves two steps: determination of the yield strength of
the dissipative element (link plastic shear resistance), and of the overstrength to allow for
strain hardening. Two design provisions available at the date of the experimental program
set-up were considered: Eurocode 8, 1994 (7) and AISC, 1997 (4). Though plastic shear
resistance is determined using similar formulations in the two codes, the European seismic
design provisions, referring to Eurocode 3 (8), consider the contribution of the fillet radius to
the shear area, resulting in a capacity 40% higher than the one of the American code, which
considers only the web area. The overstrength required for elements outside links also differ
substantially. Previous experimental research, Kasai and Popov, 1986 (9), indicated link
ultimate shear resistance about 1.5 times the plastic shear resistance. Eurocode 8 requires
an overstrength of only 1.2, while AISC 1997 results in overstrength factors between 1.38
and 1.88. Reduced overstrength factor in European codes is counterbalanced by higher
plastic shear resistance, the maximum shear force estimated to the two codes having similar
values. A relatively conservative estimation of maximum shear force was adopted in this
study (1.75 factor, applied to the web area, corresponding to a 1.25 factor applied to the
Eurocode 3 shear area):
Vmax = 1.75 ⋅ Vy = ( h − 2 ⋅ tf ) ⋅ tw ⋅ fy / 3 (1)
Design of connections to the forces determined to equations (1) and (2) was based on
Eurocode 3, 1997 and its Annex J (8). M20 gr.10.9 high-strength bolts were used. End plate
thickness (25 mm) was chosen so as to provide a mode 3 (bolts in tension) failure mode of
the equivalent T-stub, preventing excessive deformation of the end plate. A linear distribution
of bolt forces was then assumed, and the bolts checked for tension, shear, combined tension
and shear resistance, assuming a partial safety factor γMb=1.25. Demand to capacity ratio for
combined tension and shear ranged from 0.7 for the LH4 and LL4 specimens to 0.98 for the
LH7 and LL7 specimens. Additionally, bolt slip resistance was checked.
The instrumentation consisted of the actuator load cell, and a series of displacement
transducers used to measure both absolute and relative displacements. The basic force-
displacement relationship used to characterize the monotonic and cyclic response of bolted
links was actuator force (equal to link shear force) and total displacement DT, which includes
slip in connection and endplate deformations. Response of link elements is characterised by
the shear distortion angle γ - shear force F relation. For classical links, the distortion γ is
determined as the difference of end displacements divided to the link length, Engelhardt and
Popov, 1992 (10). With the notations from figure 3, γ is expressed as:
γ = DT / b (3)
Assuming that the edges of the panel bounding the link remain straight after deformation, the
same angle γ may be determined from the deformations of the diagonals (DD1 and DD2):
a 2 + b 2 ⋅ ( DD2 − DD1)
γ = (4)
2⋅a⋅b
d+
d
DD
π D1
d+D
2
γ
+γ
d a 2
-γ1
a
π
DT b
b
θ
DALS
γL
γΤ
θ γ
DALJ
γ
θ L
γ
DT
(a) (b)
Figure 4. Deformation of a bolted link (a) and its idealisation (b).
Values of angle γ determined according to equations (3) and (4) have close values in the
case of classical links. However, in the case of removable bolted links, the behaviour of the
link is more complex, and angle γ determined from equations (3) and (4) will be different.
Total link deformation is given by the sum of: (1) shear distortion of the link panel - γ, (2)
rotation in the two connections θM=θS+θj, and (3) slip in the connections, characterised by the
equivalent rotation γAL=(DALS+DALJ)/ed, and can be expressed as:
γ T = γ + θM + γ AL (5)
BEHAVIOUR OF SPECIMENS
Strength characteristics obtained from nominal and measured geometry and strength are
presented in Table 2. Account was taken of the different flange and web yield strength in
determining the link plastic moment: M y = Wpl ,w ⋅ fy ,w + Wpl * ⋅fy ,f .
Table 2. Yield and maximum forces evaluated from nominal and measured characteristics.
Wpl, Wplw, Wpl*, My, 1.6My/Vy, Vmax, Mmax,
specimen Vy, kN
cm3 cm3 cm3 kNm mm kN kNm
LH7, LL7 366.6 75.29 291.31 185.4 86.2 743 278.1 83.4
LH6, LL6 366.6 75.29 291.31 185.4 86.2 743 278.1 69.5
nominal
LH5, LL5 366.6 75.29 291.31 185.4 86.2 743 278.1 55.6
LH4, LL4 366.6 75.29 291.31 185.4 86.2 743 278.1 41.7
LH7, LL7 366.6 75.43 291.2 266.7 103.5 621 400.1 120.0
LH6, LL6 366.6 75.43 291.2 266.7 103.5 621 400.1 100.0
measured
LH5, LL5 366.6 75.43 291.2 266.7 103.5 621 400.1 80.0
LH4, LL4 366.6 75.43 291.2 266.7 103.5 621 400.1 60.0
Note: Mmax determined per equation (2)
400 400
LL7−c1 LH7−c1
200 200
F, kN
F, kN
0 0
−200 −200
−400 −400
−0.2 −0.1 0 0.1 0.2 −0.2 −0.1 0 0.1 0.2
γT, rad γT, rad
Figure 5. Force-total deformation relationship F-γT for specimens LL7-c1 and LH7-c1.
400 400
LL4−c1 LH4−c1
200 200
F, kN
F, kN
0 0
web
−200 breathing −200
−400 −400
−0.2 −0.1 0 0.1 0.2 −0.2 −0.1 0 0.1 0.2
γT, rad γT, rad
Figure 6. Force-total deformation relationship F-γT for specimens LL4-c1 and LH4-c1.
Smaller length of Lx6 reduced the damage to connections and the pinching behaviour.
Failure was attained by complete damage to bolts (see figure 7a), but also by web cracking
after repeated plastic web buckling in the case of LL6-c2 specimen, with rare stiffeners.
Starting with Lx5 specimens, connections were characterised by a more stable response,
plastic web buckling being more important and preceding the one of the flanges. Failure of
LL5-c1 and LL5-c2 specimens, with rare stiffeners, was attained by tearing of the web on
three edges, at the cracks initiated in the base metal at the web-stiffener and web-end plate
welds. Closer stiffener spacing in the case of LH5-c1 and LH5-c2 specimens reduced web
Response of specimens from the Lx4 series was dominated by web shear. Connection had a
quasi-elastic response. Flange buckling was observed only after important web buckling.
Hysteretic response was characterised by "full" cycles with high energy dissipation capacity
(see figure 6). Due to higher web slenderness of the LL4-c1 and LL4-c2 specimens, web
buckling was pronounced, and plastic web "breathing" was observed, as web buckling wave
was changing direction at reversals of load direction (figure 6). Repeated buckling lead to
web tearing along the diagonals (see figure 7b). Close spacing of stiffeners at the LH4-c1
and LH4-c2 specimens prevented this phenomenon, failure initiating through web tearing
along the stiffener weld, which extended on three edges of the web.
High stresses are present at the beam to brace welded connection, next to the beam to link
bolted connection. Higher grade steel of the elements outside removable link did not provide
sufficient overstrength in this zone. Due to repeated cyclic loading, the lower beam to brace
welded connection fractured during the test of LL7-c1 specimen. Removing the weld and
applying a new weld did not help, and the lower beam-brace assembly was completely
replaced to due to extensive damage in the zone between the lower connection and the
brace. To mitigate this problem after similar failure of the new subassembly during testing of
the LL6-c1 specimen, stiffeners were added in the affected zone to increase the shear area
and provide a smooth transfer of stresses from the bolted link element to the brace and the
beam (see figure 7c). The performance of the subassembly modified in this way was
satisfactory for the rest of tests.
Elastic response of links was characterised by the total initial stiffness KγT, determined from
V-γT relationship, as well as shear stiffness of the web Kγ, stiffness of connections KθJ and
KθS, determined from M-θJ, and M-θS relationships. Initial shear stiffness of the link (Kγ) was in
good correlation with the theoretical one (Kγth=G⋅As), and not influenced much by the different
considered test parameters. There was an important scatter in experimental values of
connection rotational stiffness. Full preloading increased the stiffness of connection by
approximately 50%. Upper connection resulted more flexible in comparison with the lower
connection. Unsymmetrical distribution of moments and lack of fit at the upper connection
may be attributed to this behaviour. Reduction of total initial stiffness of the bolted link in
comparison with the classical solution is important, as a result of both the semi-rigid end-
plate, and slip in the connection. Therefore, either explicit modelling of the semi-rigid
connection behaviour, or consideration of an equivalent link stiffness is necessary for global
analysis of frames with bolted links.
Yield force determined from V-DT relationship was not influenced by the test parameters and
was controlled by shear response of the web. Lower experimental values (see table 3) are
partially explained by the procedure used to determine yield force, which underestimates it
for high initial stiffness. On the other hand, experimental maximum force presents an
increase from the longer to the shorter links (effect of connection strength) and is higher for
closer stiffeners (prevention of web plastic buckling).
The maximum moment determined from equation (2) was lower than the theoretical one
used to design the connections. Poor performance of connections could be explained by the
fact that vertical displacement in the experimental set-up was constrained, which generated
supplementary tension in the connections at large displacements. Further research is
needed to validate this assumption and to check its application to real structures. Following
the experimental observations in this study, in order to reduce damage in bolted connections,
it is recommended to limit the length of bolted links to ed ≤ 0.8 ⋅ My Vy , which corresponds to
links LL4 and LH4.
Ultimate link displacement DTu, representing the stable hysteretic response is presented in
table 4. Cyclic loading reduced by 40% to 70% rotation capacity, with the maximum reduction
for short links. A slight reduction of ultimate displacements was observed for short links. In
terms of deformations (γTu), rotation capacity increases slightly for shorter links, with the
exception of LL4 and LH4 specimens. With the exception of longer links with rare stiffeners
(LL7), specimens showed a stable deformation capacity of at least 0.1 rad. Ductilities larger
than 10 were observed, with a number of 16 to 22 cycles in the plastic range. Bolt preloading
did not affect rotation capacity, as oppose to closer spacing of stiffeners, which improved link
deformation capacity.
Behaviour of long specimens was much influenced by the response of the bolted connection,
characterised by a gradual reduction of strength due to bolt thread stripping, and a pinching
cyclic response. The latter effect reduced the energy dissipated in the group of cycles of
constant amplitude. Full bolt preloading reduced partially this effect. Response of short
specimens was controlled by the shear of the link web, characterised by important hardening
and energy dissipation capacity, but a more rapid degradation of strength after web tearing.
Stiffener spacing had maximum importance for short links. Their effect was to limit plastic
local buckling of the web, increasing the maximum force and deformation capacity, and
providing a more stable cyclic response. However, after the attainment of ultimate
deformation, failure of LH4 specimens was more rapid in comparison with LL4 specimens.
Distribution of ductility demands between end pate and link web resulted in improved overall
deformation capacity in comparison with "pure" failure modes, determined by concentration
of plastic deformations in connection or web alone. This effect is characteristic of
CONCLUSIONS
Longer links and closer stiffener spacing imposed higher demands on the connection. Cyclic
response elements for which connection represented the weaker element were characterised
by: (1) a reduction of maximum force in comparison with elements dominate by web shear;
(2) a pinching behaviour with stiffness and strength degradation in cycles of constant
amplitude; (3) failure by gradual strength degradation due to bolt thread stripping. Post
elastic connection response was ductile, due to thread stripping. This failure mode is not
generally characteristic for bolts. Bolt failure by shank rupture would have caused a more
brittle response of long links.
Response of short links was governed by web shear, stiffener spacing being important for
their performance. In the case of rare spacing of stiffeners, inelastic response of short links
was determined by plastic web buckling, which lead to strength degradation by alternative
buckling in the direction of the two diagonals. Closer stiffener spacing limited plastic web
buckling, leading to: (1) attainment of the maximum possible shear strength; (2) a stable
hysteretic response; (3) a larger rotation capacity, but also (4) a more rapid failure by web
tearing on the panel edges.
With the exception of very short links, connections were partial-strength. On the basis of
present experimental program, in order to prevent excessive connection damage, it is
recommended to limit link length ed to 0.8⋅My/Vy. Design strength of short removable links
limited to this length may be computed as for classical short links. Full bolt preloading
resulted in higher initial stiffness, a more stable hysteretic response and a larger deformation
capacity, and therefore is recommended for removable short links. Semi-rigid connections
with flush end plate reduce substantially initial stiffness of removable short links in
comparison with classical solution. Global analysis of eccentrically braced frames with
removable links requires either explicit modelling semi-rigid connections, or consideration of
an equivalent shear stiffness of the removable link.
ACKNOWLEDGEMENT
Support of the Romanian National Education Ministry (MEC-CNCSIS) and World Bank
through the C16 Grant “Reliability of Buildings Located in Strong Seismic Areas in Romania"
and MEC-CNCSIS grant AT10/218 "Seismic response of dual eccentrically braced frames
with removable links" is gratefully acknowledged.
e, ed – clear length of the link between braces, and length of the bolted link
tf, tw, h – flange thickness, web thickness, and cross-section height
Dy, Fy – yield displacement, yield force
fy (Reh), fu, A – (upper) yield stress, tensile strength, elongation at rupture
Vy , My – plastic shear resistance, plastic moment
Vmax, Mmax – maximum shear force, maximum moment
γMb – partial safety factor for bolt resistance
DT – total link displacement
a, b – link panel dimensions
γ – link shear distortion angle
γAL – equivalent link rotation angle due to connection slip
θM, θS, θj – average, bottom, and top connection rotation
γT – total link distortion angle
DD1, DD2 – measurements of link diagonal displacement transducers
DALJ, DALS – measurements of link slip displacement transducers
fy,w, fy,f – web and flange yield stress
Wpl,w, W*pl – plastic modulus of the web and flanges (W*pl = Wpl - Wpl,w, Wpl)
KγT, Kγ, KθJ and KθS – total initial, web shear, and connection stiffness
DTu, γTu – ultimate displacement, ultimate deformation
REFERENCES
ABSTRACT
Rack Structures are constructed easy to be assembled, that is
the connections must be temporary elements in order to allow
the owner to change the lay out according to the needs, which
makes that bolted and welded connections do not qualify for
these purposes. This paper presents experimental findings about
the beam-column connection under static and cyclic loads. The
connection is extremely flexible, so the rack structures should be
classified as Partially Restrained Construction, according to
AISC design code. The similarity of the static and cyclic failure
modes indicate that the failure is controlled entirely by the
connecting elements.
INTRODUCTION
The design of rack supporting structures has become increasingly complex in recent years;
the properties of the connections determine the behaviour of the structure and the
performance of the structural system in the event of a destructive earthquake occurrence.
There is not much information available about these systems and there are some evidence
showing a poor performance of the racks supporting systems during the past earthquakes.
Rack structures in seismic zones are requested to comply with the local building codes and
they must be engineered to meet the code requirements of the building structures. Even
though rack structures are quite different to the buildings they use to be placed inside a
building, so it is necessary to control the lateral deflections in order to avoid the hammering
of the structures, the collapse of the racks and eventually the collapse of the surrounding
structure. The connections designed to be easily unlocked as well as the user needs makes
difficult to keep control on the amount of beams to be removed, which can produce at last
low redundant structures. The rack supporting structure studied in this paper has been used
successfully in Chile for several years and they have survived the March 3, 1985 Chilean
earthquake. The beam is connected to the column by using hooks that are fabricated with
the beam; these hooks are inserted into columns slots, so they can be easily disconnected
from the column. Details and connecting elements are shown in Fig. 1, the thickness of the
elements is usually 2 mm, however, in some columns this thickness is increased up to 3 mm.
LABORATORY TESTS
In order to study the behavior of this type of rack structures, there were conducted several
static and cyclic tests on these connections. Moment-rotation curves were determined as a
result of the tests.
HOOK
OPENING FOR
BEAM
SAFETY CLIP
301
BEAM BEAM 48 64
80
The mechanical properties of the connecting elements were obtained from the testing of
several coupons. The set up is shown on Fig. 2, the axial displacement measurements were
obtained from high resolution photography. The yielding and rupture values are shown on
Table 1.
The properties of the steel shapes are presented on Table 2, according to the notation
indicated in Fig. 3.
Test A B C D E F G H I J K L M N O
1 80 22 46 28.5 74.5 15.5 50 2 53 121 2 48 64 301 2
2 80 22 46 28.5 74.5 15.5 50 2 53 121 2 48 64 301 2
3 80 22 46 28.5 74.5 15.5 50 2 53 121 2 48 64 301 2
4 80 23 48.7 27.6 76.3 17.3 50 3 53 121 2 48 64 301 2
5 80 23 48.7 27.6 76.3 17.3 50 3 53 121 2 48 64 301 2
Cíclico 80 22 46 28.5 74.5 15.5 50 2 53 121 2 48 64 301 2
O
B
HOOKS
C
E BEAM OPENING FOR
SAFETY CLIP
H D
K J
F G F N
L M
Fig. 4 shows the set up of the tests, the testing frame is placed on the floor, supported on
wheels in a self reacting fashion, that is, there is no need of fixed points to take the reactions.
The testing frame allows the testing of two beams at the same time, one of them in a normal
position and the other one either in the normal position or upside down. The inverted position
was included in some of the tests to take into account the possibility of a sign change of the
seismic loads, able to produce the unlocking of the connection.
LVDT transducers were installed on the beams and strain gauges to measure the axial
deformations on the columns; they were discarded afterwards because the column axial
deformation measurements turn out to be negligible. The load was measured by a load cell
placed between the jack and the beam. There were conducted five static tests and three
cyclic ones. Fig. 5 shows a sketch of the geometry of the specimen, the table on the right
indicates the load and the beam position.
Normal Beam T
Right Side Connection x P
TEST x(m) R.S.B. L.S.B.
Static Loads
Stat. 1.4 Normal Inverted
Tests 1 and 4: x= 1.40
Test 2: x= 0.70 Stat. 0.7 Normal Inverted
Tests 3 and 5: x= 0.20
Stat. 0.2 Normal Inverted
Cyclic Loads
Stat. 1.4 Normal Inverted
x=0.70
Stat. 0.2 Normal Normal
Cyc. 0.7 Normal Normal
Left Side Connection
P Tests 1, 2, 3 and 4: Inverted Beam
x
Test 5 y cyclic test: Normal Beam
Three positions of the load were considered to take into account the influence of the shear
force in the moment-rotation curve. In a first arrangement, both beams were connected in
the same position they work. In a second arrangement, one of the beams was connected in
an inverted position. Fig 6 shows a detail of the beam to column connection, the transducers
are fixed to the beam and there is a reference point, fixed to the column, in front of every
transducer. Under the left transducer it is placed a safety clip fitted in a special hole. In some
cases the builder does not include the clip or this is replaced by a bolt.
There were conducted five tests to determine the behavior and the collapse mechanism of the
connection. The tests were conducted as far as the collapse of any of both connections or up to
the maximum displacement of the loading device. An Electrack 2000 jack was used the
maximum displacement capacity of this device is 300 mm.
In the first test a bolt was placed instead of the safety clip at the inverted beam to avoid
unlocking. In order to detect potential yielding of the column, deformations in three points of the
column sections were measured. The failure occurred at the beam in the normal position; the
The second test was basically the same than the first one but the eccentricity of the load was
smaller (0.70 m). The failure was again after the outermost hooks of the connection of the
beam placed in the normal way, were broken, the failure pattern is shown in figure 7d. It can
be observed that the first hook (on the top) is missing and the second and fourth (on the
bottom) failed. The column deformations seems to be unimportant, except for the local dents
that can be observed at the hooks holes after the failure.
Test N° 4 intended to confirm the results of the first one, the column selected was thicker
(3mm instead 2mm) and the specimen included the safety clip. The failure occurred at the
inverted beam, all the hooks yielded. Figure 7d shows that the load was applied until three
hooks became torn, and the fourth one, close to the safety clip, finished absolutely deformed.
The clip did not suffer any damage, just like in the case of the normal beam connection.
Specimen five was arranged with both beams on the normal position. The failure mode was
similar and the hooks became broken at the end of the test.
All of the tests exhibit similar curves, a comparison of the moment-rotation curves obtained
from connections with beams in a normal position are shown on figure 8. A more detailed
comparison for connections with the beam in an inverted position is presented by Irisarri (1).
2500
2000
Moment [N-m]
1500 Test 1
Test 2
1000 Test 3
Test 4
500 Test 5A
Test 5B
0
0 0.01 0.02 0.03 0.04 0.05
Rotation [rad]
The set up arrangement was modified to perform the cyclic tests; a double action Sheffer
jack, with a 500 mm displacement stroke was used. The transducers were also changed to
allow a larger deformation range. Additional transducers were included to measure the gap
of the connection and the displacement of the application point of the load. The loading
history included series of 3 cycles of equal maximum displacement (± 30, ±60 and ± 90 mm),
the maximum displacement of the series was increased until the connection failed. The
behavior of the samples was essentially the same of the behavior observed in the static
tests. A first failure occurred at the outermost hooks, and the connection kept a residual
strength given by the central hook, the safety clip, and the end plate. Figure 9 shows a
typical moment-rotation curve.
The similarity of the static and cyclic failure modes indicate that the failure is controlled
entirely by the hooks. Such a result could be expected, because the bearing strength of the
plate is normally larger than the shear strength of the hook (when both components are
made with the same steel). The following zones can be identified:
1. A first horizontal line (zero moment) represents a gap; it reaches about 0.02 rad
after several cycles. The gap increases as long as the loading cycles increases;
this effect is a result of the yielding of the hooks and the residual deformations.
2. There is a second zone with a large stiffness, it occurs after the gap when all the
hooks are in the elastic range.
3. The third zone occurs after the yielding of the outermost hooks. It can be
appreciated a reduction of the stiffness, this smaller stiffness is provided by the
hooks who still remain in the elastic range.
4. The stiffness of the unloading path is the same than the elastic one. The yielding
process increases the gap of the connection.
5. When all the hooks have yielded, the moment keeps constant.
The slopes of the curves are easy to be identified, thus, it is possible to describe the behavior
of the Moment-Rotation Curve in terms of several straight lines. The hysteretic model is
shown in red on Fig. 10. The aforementioned behavior is a result of the progressive failure of
the hooks, starting from the outermost of them.
The gap is represented by the horizontal zero line, which is about 0.002 radians in the first
loading step. R1 is the elastic stiffness before the occurrence of the failure of the hooks. A
reduced stiffness follows the elastic loading process as it is indicated in the third loading
path; this new reduced stiffness was called R2. The model assumed that R2 represents the
stiffness from the failure of the outermost hooks up to the maximum strength.
All the tested connections show a similar strength range, from 2000 N-m up to 2500 N-m, the
only exception was the Specimen N° 3, whose strength was 1484 N-m. It was found no
difference on the behavior when the connection is normal compared to the inverted
connection. The hysteretic curve is stable, as long as the number of cycles increases the
initial 0.002 radians gap increases. Even though the connection hooks yield, the deformed
shape was the same during the rest of the loading process. Table 3, shows the maximum
value of the moment, rotation and load, for the static loading case.
2000
1500
R2
1000
Moment [N-m]
R1
500 R 1
0
-0,03 -0,02 -0,01 R1 0 0,01 0,02 0,03 0,04 0,05 0,06 0,07
R1 -500
-1000
R2
-1500
-2000
Rotation [radians]
The connection of the inverted beam with bolt (IBB) has more resistance than the normal
beam without either bolt or clip (NBW) that was observed during the first two tests where the
failure occurred at the normal beam. The rest of the tests were performed with the safety clip
in both connections, in such a case, the inverted beam had a smaller resistance than the
normal beam so the failure occurred at the inverted beam, except in the last test where both
beams were connected in a normal position.
CONNECTION CLASIFICATION
The connection can be classified as semi-rigid, but it exhibits a different behavior than other
type of connections. Fig. 11 is a non-dimensional M-θ curve comparison with some other
typical cases, the vertical axis is the moment of the connection divided by the plastic moment
of the beam, and the horizontal axis is the rotation of the connection divided by the plastic
rotation of the beam.
Bjorhovde, Colson, and Brozzetti (2), based on a series of 55 set of connection data tests
collected by Kishi and Chen (3), proposed a criteria to determine whether a connection is
rigid, semi-rigid or flexible. They defined three zones assuming that the connection is a beam
segment whose length defines the connection type, for semi-rigid connections the segment
should be between two and seven times the beam height (2d~7d). Fig. 12 shows the non-
dimensional M-θ curve for the connection. It is based on the aforementioned definition, the
axis have the same meaning than the axis of figure 11.
The boundary between the semi-rigid and the flexible zone in terms of moment ratio is 0.2,
the rigid zone falls out of the limits of the graphic.
According to the classification, the equivalent length for this connection is 170d, that is, ten
times larger than the length of a double angle connection, which is extremely flexible. The
curve shows a continuing hardening, so it starts flexible and it moves to the semi-rigid zone
at larger deformation levels.
0.6
0.5
0.4 le=10d
M/Mp
0.3 Clipped
Bolted
0.2
0.1
0
0 20 40 60 80
θ/θp
Figure 12. Moment – rotation curve classification.
1. Since the rack connections presented are essentially flexible, the rigid approach for the
analysis is not correct; the proper way to analyze the structure is taking into account the
non linear properties of the connection.
2. Even though several types of connections have been tested, and there are several
predictive equations, there is not too much information about these kinds of connections.
It is necessary to perform more tests in order to have a deeper knowledge and a
calibrated model of these rack structures.
3. The connection requires a minimum locking which can be obtained by using the clip, a
bolt or another safety device. The clip seems to be a good solution, because is easier for
the purpose of locking-unlocking the rack structure. However, the easy unlocking of
beams can produce less redundant structures, which makes the structure more
vulnerable to lateral loads.
4. Once a hook has yielded that portion of the connection stop working, the connection
become less redundant and a redistribution of the bending moments towards the center
of the beam span takes place. As a consequence an earlier failure of the beam is
probably to occur.
5. The failure mode is controlled entirely by the hooks, so the failure takes place at the
beam. The beam is a component easy to be replaced and the hooks acting as fuses
prevent the columns failure.
6. The hysteretic loops are stables with a gap near the zero moment of the curve. The
connection unloading path follows a line parallel to the initial loading part of the curve,
and during the next loading steps, it follows the same initial line. The gap increases with
the loops and the rotation increases too.
7. The shear forces seem to have no effect on the moment-rotation curve.
8. The ability of the connection to support vertical loads, the loss of redundancy, and the
type of the failure in a brittle fashion makes them useful for vertical loads but they should
not be used for seismic loads. The connection can reach more than 50% of the beam
capacity before the failure, just like a semi-rigid connection. However the rotation level is
just like a typical flexible connection which is about 0.02 radians when the strength is
about 10% of the full beam capacity.
REFERENCES
ABSTRACT
The paper is related to a frame-type cold-formed structural system that is
developed as a load bearing structure of smaller industrial-type and residential
buildings. In the structure C-section structural members are connected to each
other by simple site connections using self-drilling screws. The design is based
on test, where the local (screw) and global (joint) behaviour modes are
investigated. In the paper the experimental programs are emphasized. The test
results of the two-level test programs are shown and discussed as an
interacting phenomena: the joint behaviour is derived and explained by the
local phenomenon. The behaviour modes are classified and the effects of
different joint characteristics are determined.
INTRODUCTION
In the framework of a joint project of Lindab Hungary Ltd. and Budapest University of
Technology and Economics a new cold-formed structural system is under development. The
aim of the research and development activities to give alternative, cold-formed steel
structural solutions for different functional purposes, such as smaller industrial-type buildings,
residential buildings, floor sub-systems, flat roof covering sub-systems. The current phase of
the development is based on the available Lindab C/Z and sheeting profiles and self-drilling
screw type connections. In the applied structural solutions there is a priority on the traditional
frame-type of arrangements due to the traditional Lindab fabrication and erection technology
of the region. In the second step of the R&D project product development is planned to
increase the efficiency of the structural systems.
The load bearing structure of the building system is cold-formed frame, built-up from C-
profiles using self-drilling screws. The configurations of the frames are as follows (figure 1):
• two-hinged portal frames,
• two-hinged portal frames with tie bar,
• two-storey frames with floor beam.
The beams are single and the columns are single or double C-profiles. The double C-profiles
are used in box arrangement. The joints are solved by back-to-back self-drilling screw
connections between the webs of the connecting C-profiles, resulting semi-rigid and partial
strength characteristics. A piece of C-profile as an additional element is used in the eve joint.
The design of the cold-formed frame is based on the pertinent parts of Eurocode 3 Part 1.1
(1) and 1.3 (2) standards as presented by Fóti and Dunai (3). The design is supported by
experimental tests as follows:
• beam-to-column joint tests,
• self-drilling screw shear tests,
• compression member tests,
• composite floor beam tests.
The paper deals with the first two – interacting – test programs. Bellow the experimental
studies are summarized. Then the evaluation strategy of the test results is discussed. The
joint experienced behaviour modes are shown with a focus on the interaction of the failure
phenomena of the local connecting element and the global joint. Finally the conclusions and
the further steps of the ongoing research and development activities are drawn up.
Test program
Full-scale tests are performed on beam-to-column frame corner joints as a part of the
development of the first prototype of cold-formed Lindab frames; see Dunai et al (4). The
results of the studies can be extended for the behaviour of the eve – beam-to-beam – joint,
too. Altogether 38 experiments are performed and the investigated parameters are as
follows:
• size of the fasteners (diameter: 5.5 and 6.3 mm),
• arrangement – number and placing – of fasteners (5-22 pieces of self-drilling
connectors),
• thickness of connected C-profiles (1.0 – 3.0 mm),
• height of the connected C-profiles (150 – 300 mm).
Test arrangement
Figure 2 shows the two types of test setups which are used in the experiment, including
typical joints. In test setup A the lengths of the column and beam elements are 1.0 and 1.5 –
2.0 meters, respectively. The concentrated load is applied on 1.0 – 1.5 meters arms for
different profile heights. Lateral bracings are applied on the top flange by hat profiles (purlins)
and a C-profile (bracing) on the compression side of the joint.
The test setup B is designed for residential frame system, using a 1.0 m long boxed column
element and a 2.5 m long beam section. The C-profiles have different upper and lower flange
widths; the boxed column is build up from two C-profiles slipped into one another and joined
partly on the flanges by self-drillers. Lateral bracing is solved by two pairs of hot rolled
columns keeping the beam in the plane of the frame by fork supports.
F
19
20
25
00
58
0
1050
1000
165
200
Measurement system
The main purpose of the experiments is to obtain the moment-rotation relationships of the
joints. The rotations are measured in three points of the joint region by rotation measurement
devices: on the column under the connection and on the beam both sides of the connection
(see figure 2, numbered boxes). The displacements are measured in four places: vertically
under the force and in the middle of the beam; horizontally, transverse direction in the joint;
and horizontally, lateral direction in the joint. Strains are also measured; strain gauges are
placed on the flanges of the beam near to the joint.
Isolated self-drilling screws of the frame corner joints are tested under shear, in single
overlapped condition in the “local” experiments; see Dunai et al (5). The screw types and
sizes are the same as applied in the “global” tests. The plates of the specimens are partly cut
out from the global specimens and partly obtained from base material of the C-profiles. Cold-
formed steel strips nominal basic yield strength of 350 MPa and ultimate tensile strength of
420 MPa, with thickness from 1.0 to 3.0 mm are used. In the program one, two and three
screwed specimens are tested, as it is shown in figure 3. In the tests the force-displacement
relationship of the specimens are measured.
The test results of the “global” joint and “local” screw tests are evaluated and presented by
Fóti and Dunai (6), and Fóti (7), respectively. In the papers the typical behaviour modes are
identified and illustrated. In the current paper the results of the two types of tests are
combined and presented with the purpose to obtain correlation between the component and
the joint behaviour. This background is required to develop a component based design
method for the joint as it is presented by Fóti and Dunai (8).
In the traditional approach a bolted connection under shear is designed by separating it into
connector and base material, and checked on the bases of the “independent” failure by the
application rules of the given standard. The results are connector resistances, from which the
number of connectors can be determined. In the investigated details the joint is assumed as
a multi-component structure. The interaction in the element and connector behaviour
requires another level of design methodology. The connector and its region are considered
as a unit. The difference between the separations is similar to the difference between the
meanings of connection and joint. This approach follows the philosophy that is used in the
moment resisting end-plate joint design of Eurocode 3.
If the screw is separated with its region, the interactive failure modes that are affected by
both screw and material properties can be followed. These possible interactive failure modes
are obtained from the tests, as follows:
• tilting and pull-out,
• tilting and bearing,
• tilting and shear and
• tilting and tension failure of fastener.
It means, that only two failure modes: pure bearing failure (not experienced) and pure screw
shear failure are non-interactive – independent modes. In all other cases the properties of
the screws and the base material affect the behaviour of each other.
In the following sections the combined (local and global) test results are presented and
discussed for typical joint failure modes.
This failure mode is occurred when relatively small number of screws is applied in relatively
thick profiles. In this case the tilting and pull-out of fasteners cannot take place.
The first mode of shearing failure is shown on the moment-rotation relationship of figure 4.a.
The assumed resistance of the profile is nearly reached in the first mode when the dominant
screw is sheared suddenly. The downward steps over the top equals to the shearing
resistance of the screws, one by one. Certain load bearing capacity is experienced on a
lower load level since the remaining screws are still active.
15
8
„a” „b”
Moment [kNm]
6 10
Moment [kNm]
4
5
2
0 0
0 0,03 0,06 0,09 0,12 0,15 0,00 0,01 0,02 0,03 0,04 0,05
Rotation [rad] Rotation [rad]
At the maximum level of the moment the shearing failure is occurred in more screws at the
same time in the second mode and it resulted in a sudden collapse of the whole joint, as
shown in figure 4.b.
Force [kN] 10
0
0 10 20 30
D is p la c e m e n t [m m ]
This failure mode is experienced when relatively small number of screws is applied in a
relatively thin profile. The typical moment–rotation curve can be seen in figure 7.
2
1 ,5
Moment [kNm]
0 ,5
0
0 ,0 0 0 ,0 4 0 ,0 8 0 ,1 2 0 ,1 6 0 ,2 0
R o ta tio n [ra d ]
Figure 8 shows the force-displacement relationship of the reference local screw tests using
the same thickness of plates, failed by tilting and pull-out mode. The failure mode is
illustrated by the photo. As it can be seen the jumping phenomenon of the screw behaviour is
inherited in the global joint behaviour.
9
6
Force [kN]
0
0 15 30 45
Displacement [mm]
General
This phenomenon may appear in the column or in the beam, or in both of them. In all of
these cases, the connectors are strong enough to avoid the failure but due to tilting and/or
bearing the behaviour is combined. The place where local buckling appears is determined
basically by the arrangement of the connectors. Using spread screws the profile’s failure is
experienced in three places: next to the joint in the beam (figure 9), just bellow the fasteners
in the column (figure 10), and near to the joint in the column.
This failure mode is obtained when strong enough connection is applied in thinner plates and
the dominant behaviour is appeared in the element by plate buckling. Typically obtained in
case of four groups of self-drilling screws. Figure 11 shows that the initial part of the
moment–rotation curve is nearly linear then the stiffness is gradually decreasing due to
plastic deformation of the components and tilting of screws. In the ultimate behaviour under
the screw group in the compression side of the joint local buckling is experienced, with a
stronger interaction of the screw deformation and local buckling under the group of screws
on the compression side.
6,0
Moment [kNm]
4,0
2,0
0,0
0 20 40 60 80 100 120
Displacement [0,001 rad]
The reference screw local behaviour is illustrated in figure 12. The force-displacement
diagram starts with a clear linear part, followed by plastic-like behaviour, caused by the
interaction of tilting and bearing. A jump on the diagram shows a pull-out effect, directly
followed by bearing failure.
12
8
Force [kN]
0
0 5 10 15
Displacement [mm]
This failure mode is experienced when the tilting/pull-out resistance of the connection and
the buckling resistance of the element are close to each other. It is typical when spread of
self-drilling screws is used and in the behaviour the most effected screw (farthest from the
centroid of the screw arrangement) is dominant. Due to the large deformation capacity of the
screws by tilting, a long plateau of the moment-rotation curve at the ultimate level is
obtained, as it is shown in figure 13. After significant plastic deformation, local buckling
appeared in the beam due to the concentrated bolt force effect on the compression side of
the profile. Due to the increased rotations of the joint in the dominant screws pull-out
phenomenon is experienced.
Moment [kNm]
2
0
0,00 0,05 0,10 0,15 0,20
Rotation [rad]
Figure 13. Interaction of tilting/pull-out and buckling.
Distortion typically appeared in specimens of test setup B when the connectors are strong
enough to avoid failure, as it is shown in figure 14. This failure mode is in certain correlation
with the buckling resistance, but also greatly effected by the eccentricity of the setup. The
phenomenon is the following: first the web of the beam moves out from the initial plane due
to the eccentricity, and the column web acts as a bended plate. It causes decrease in the
stiffness of the whole joint, as it is shown in figure 15. In combination with it tilting of the
screws appears, and the jumps on the curve refer to pull-out of screws. The reference screw
local behaviour is illustrated in figure 8.
20
16
Force [kN]
12
0
0 100 200 300 400
Displacement of the beam end [mm]
In the paper the results of an ongoing research and development work on light-gauge cold-
formed frame joints are presented. On the bases of the results of the R&D activities the
following conclusions can be drawn.
The research and development work is continued in two directions: further joint tests are in
progress on glued and bolted specimens with the purpose to extend the system to bigger
spans, and test-based design model is under development for more general application.
AKNOWLEDGEMENT
The research work is conducted under the financial support of the OTKA T035147 project
and OM ALK00074/2000 R&D project with the cooperation of BUTE and Lindab Ltd.
REFERENCES
(1) Eurocode 3, (1992). Design of Steel Structures, Part 1.1 General Rules and Rules for
Buildings.
(2) Eurocode 3, (1996). Design of Steel Structures, Part 1-3. General rules –
Supplementary rules for cold-formed thin gauge members and sheeting.
(3) Fóti, P. and Dunai, L. (2001). Design aspects of cold-formed portal frames, 3rd
International Conference on Thin-Walled Structures, Eds. Zaras, J., Kowal-Michalska,
K., Rhodes, J., Elsevier Publisher, pp. 203-208, Krakow, Poland.
(4) Dunai, L., Fóti, P., Kaltenbach, L. and Kálló, M. (2000). Experimental study on frame
corner joints built-up from cold-formed C-profiles, Department Reports (1-3) (in
Hungarian), Technical University of Budapest, Department of Steel Structures,
Hungary
(5) Dunai, L., Fóti, P. and Kálló, M. (2001). Experimental study on screwed connections
under shear, Department Reports (1-2) (in Hungarian), Budapest University of
Technology and Economics, Department of Structural Engineering, Hungary
(6) Fóti, P. and Dunai, L. (2000). Interaction phenomena in the cold-formed frame corner
behaviour, 3rd International Conference on Coupled Instabilities in Metal Structures,
CIMS 2000, Eds. Camotim, D., Dubina, D., Rondal, J., Imperial College Press, pp. 459-
466, Lisbon, Portugal,.
(7) Fóti, P. (2002). Behaviour modes of screwed connections under shear – experimental
research, (in Hungarian), Scientific Publications of the BUTE Department of Structural
Engineering, Műegyetem Kiadó, pp. 29-36, Budapest, Hungary.
(8) Fóti, P. and Dunai, L. (2002). Test based design method of moment resisting joints in
cold-formed structures, International Colloquium on Stability and Ductility of Steel
Structures, Prof. O. Halász Memorial Session, Ed. Iványi, M., Akadémia Kiadó, pp.
211-218, Budapest, Hungary.
ABSTRACT
Steel storage pallet racks are three-dimensional framed structures, similar to
multi-storey building structures. For practical reason (possibility of easy access
to stored products) pallet racks are not braced in down-aisle direction, so the
only source of the stiffness required for down-aisle stability is the stiffness of the
connections between columns and beams, and the stiffness of the column
bases.
The paper presents selected test results of beam-to-column joints of one of the
commercially available pallet racks system in Poland. The component method
was used to assess main joint properties. Results obtained using proposed
model were compared with test results.
INTRODUCTION
Steel pallet racks are regular, 3D multi-storey, multi-bay structures, used in industry and
warehouses for the storage of palletised goods. Design analysis of such structures is carried
out on 2D sub-assemblies, separately in the cross- and down-aisle directions. Stability in the
cross-aisle direction is provided by bracings. Because of lack of bracings in the down-aisle
direction, structure analysis is carried out by adopting a semi-continuous sway frame model,
i.e. unbraced frame with semi-rigid joints.
Most of the recent design codes and papers, e.g. Baldassino and Zandonini (1), Baldassino
and Bernuzzi (2), Markazi et al. (3) recommend an experimental-theoretical approach and
require experimental tests of beam-to-column connections to obtain semi-rigid joint
characteristics that can be applied in the global analysis. Such experimental tests were
conducted for one of the commercially available steel racks system in Poland. Observation of
joint behaviour during tests suggests possibility of using the component method to evaluate
the main join properties: moment resistance and initial stiffness. The aim of the paper is to
present preliminary model of the component method for investigated rack joints.
A sketch of tested beam-to-column connection with its dimensions is shown in figure 1. The
joint consists of a beam-end connector with tabs, made of a 4-mm thick cold-formed angle,
welded to each end of the beam. The leg of the angle with tabs is in contact with column web
after assembly, while there is a 2-mm gap between the second leg, perpendicular to the
beam axis and the column web (figure 1).
Tests were carried out on five joint groups. Every tested joint group had the same connector and
250
19
50
37
90
Testing procedures, instrumentation and detailed test results of analysed joints can be found
in Kozłowski and Ślęczka (4); below the main test results are summarized.
M pl,exp
1
M u,exp
3.0
2.0
experimental curve
1.0
0.0 φ [rad]
0 0.02 0.04 0.06 0.08 0.1
The plastic flexural resistance M pl ,exp was obtained as a point of intersection of a straight
line representing initial stiffness S j ,ini , with the line of slope 0.1 S j , ini , which is tangent to the
non-linear part of the curve obtained in test (figure 2).
The component method is now one of the most effective methods to analyse and predict the
rotational behaviour of different types and different configurations of connections. It is mainly
used in case of steel joints made from hot rolled sections, prEN 1993-1-8 (5), and composite
joints, prEN 1994-1-1 (6), but recently has found also other applications e.g. Fink et al (7).
The application of the component methods is usually performed in three stages. The first
stage is the identification of components in the analysed joint, where the complex joint is
subdivided into parts. The second stage is predicting for each component its individual initial
stiffness, strength and deformation capacity. The behaviour of each component is described
by a bilinear relationship between displacement and force. Components that do not affect the
stiffness of the joint are modelled as rigid-plastic, while other components are modelled as
elasto-plastic elements. The third stage is the evaluation of flexural strength and rotational
stiffness of the whole joint. In this stage the lever arms hi should also be predicted for every
group of components.
In the case of analysed storage rack joint, the following components can be identified.
A part of the column web situated between the slot and the flange of the column is subjected
to a distributed load transmitted by the tabs in the connector. Fixed-ended beam of a span
equal to the height of the slot, loaded by contact stresses can be used to model the
behaviour of this component (figure 3). The resistance of the component depends mainly on
shear stresses. This component is active only in a tension part of the joint.
V
0.06 F
A 19 A
5
F 0.94 F
A-A AV 2.0
12
Figure 3. Model to determine the resistance and stiffness of column web in tearing.
This component is active both in tension and compression zones of the joint (figure 4).
Tab
19
4
In tension
In compression zone zone
Figure 4. Local stresses due to the pressure acting on the hole walls.
The resistance of the column web in bearing can be evaluated using analogy to bolted
connections, according to EC3:
Fcw, b = 2.5α f u , cw d t cw (3)
where bolt diameter should be replaced by the thickness of the tab d = t tab = 4.0 mm .
In the case of joint with tabs α = f u , co / f u , cw = 0.935 . Finally, the resistance of the column
web in bearing can be predicted as Fcw, b = 8.07 kN .
The stiffness of the component can be also estimated, as suggested by EC3, for snug
tightened bolts.
k cw, b = 24 k b k t d f u , cw (4)
where d = t tab = 4.0 mm , k b = 1.25 and k t = 1.5 t cw / d M 16 ≤ 2.5 . Using the above
equations it can be found that k cw, b = 90.7 kN / cm .
A part of the column web is subjected to tension or compression. The buckling resistance of
the column web in compressions can be computed according to Chen and Newlin (8) as:
3
t cw
Fcw, c = 10766.1 f y, cw = 42.2 kN (5)
d wc
The resistance of the tensioned part of the column web can be determined according to EC3
as:
Assuming the tangent of the spreading angle 1:2.5 (figure 5), beff = 25 mm and
Fcw, t = 19.2 kN .
1:2.5
tcw=2.0 tcw=2.0
1:1
5 5
be‘ ff
beff
dwc=172 mm dwc=40
Figure 5. Model to determine the resistance and stiffness of column web in tension
and compression.
The axial stiffness of the column web in tension and compression can be evaluated by the
relationship
'
k cw = E beff t cw / d wc (7)
'
Effective width of column web for stiffness calculation beff in compression zone is equal the
length of zone where distributed load from bearing is acting ( 5.0 mm ). The effective width in
tension zone is predicted considering an 45D angle of the load spreading (figure 5). So, the
axial stiffness in the compression zone is k cw, c = 525 kN / cm , and in the tension zone is
k cw, t = 488 kN / cm .
Tabs in shear
Another component affecting the rotational behaviour of the rack joint is the tab of the
connector, subjected to bending and shear from local bearing stresses. To predict the
resistance and stiffness of this component, the model of a cantilever beam, subjected to a
concentrated load can be used (figure 6). The distance between the load and fixed end of
beam, measured by the developed length of tabs is equal l = 7.0 mm . The resistance of the
component can be evaluated as shear resistance:
Ft , s = Av ( f u , co / 3 ) (8)
Area of bearing
l
F
13 ttab
Behaviour of the connector can be modelled taking into account a cantilevered part of the
connector, protruding over the flange surface of the beam, as shown in figure 7.
δ
X
F 4
Ext.
A-A 70
Int.
A A
6 13 x
37
The initial stiffness k co of this component is calculated as the ratio of the force to the
deflection of the component under this force.
Because of the openings (slots) in connector, the deflection of the component was evaluated
as for shear-wall structures under lateral loads acc. to Benjamin (9), including the effect of
bending and shear. Every tested group of joints have different height of the beam, so the
initial stiffness of the component is different for each group. In every group two values of
initial stiffness has been calculated: for external slot (force is acting in tab more distant to the
beam), and for internal slot (force is acting in tab closer to the beam). Predicted values of the
initial stiffness for each group are listed below.
Regarding the tension and compression zone of the connector web, the resistance can be
estimated according to EC3 as:
where the effective width beff of the connector web is found accounting for the spreading of
the stresses transmitted by the beam flange (figure 8). Assuming the tangent of the
spreading angle 1 : 2.5 , it can be adopted that beff = 27 mm and Fcow, t = Fcow, c = 37.4 kN .
dwc=33
beff
tco=4.0
The axial stiffness of the connector web in tension and compression can be evaluated from
the relationship:
'
k cow = E beff t co / d wc (13)
Assuming an angle of 45D for the load spreading, the value of the initial stiffness
k cow, t = k cow, c = 2800 kN / cm can be calculated.
beff
The last component, beam flange in tension and compression, has to be considered only
in the evaluation of the joint resistance. Its resistance can be computed as:
hb-tfb
hb-tfb
hb-tfb
50 50 hc
S1 S2 Sj,ini
50 kcow,c 50
kcw,b kcw,c
Figure 10. Mechanical model and procedure for evaluating the rotational stiffness.
The mechanical model adopted to predict the initial rotational stiffness is shown in figure 10a,
and the procedure to evaluate the rotational stiffness is depicted in figures 10b-10e. The first
step is the computation of effective stiffness of each row (figure 10b), next predicting the
equivalent overall stiffness in compression and in tension zone and lever arms (figure 10c).
Third step is change of two springs with axial stiffness into one spring with rotational stiffness
(figure 10d) and finally predicting the initial stiffness of the whole model, (figure 10e).
Adequate relationships to make conversion of mechanical model from figure 10a to model
depicted in figure 10e, can be found in mechanics handbooks or in Faella et al (10). In the
calculation it is assumed, that contribution of springs situated in the centre part of the joint is
small, so influence of this row was neglected. Comparison of predicted values of initial
rotational stiffness with values from tests is shown in table 3.
Table 3. Prediction of the initial stiffness and comparison with experimental data.
Group C Group B Group E
Test results [kNm / rad ] 116.7 139.4 123.1
Component method [kNm / rad ] 133.0 148.4 156.4
Difference: (component–test)/test [%] 13.9 6.4 27.0
Behaviour of such joints should be considered in two stages: before direct contact of
connector with column flange (elastic stage) and after this contact occurred (plastic stage).
Mechanical model adopted to predict the elastic flexural resistance and lever arms for each
group of components are presented in figure 11a. Because the weakest component governs
the resistance of each row, model depicted in figure 11a can be simplified into model
presented in figure 11b. The resistance of components marked “1” in figure 11b is equal
F = Fcw, tear = 6.36 kN when component is in tension (column web in tearing) or
F = Fcw, b = 8.07 kN when component is in compression (column web in bearing). The
resistance of the connector in bending (modelled as rotational spring) is equal to
M co,b = 42.22 kNcm , and the resistance of components marked “2” (beam flange) is equal
F = Fcof , b = 29.9 kN . Model presented in figure 11b, represents only the initial behaviour of
hb-tfb
hb-tfb
hb-tfb
50 50 hi
50 Fbf,c 50 “2”
“2”
Fcow,c
Fcw,b Fcw,c
Figure 11. Mechanical models for the elastic flexural resistance evaluation
Group E Group B Group C
Fcw,tear Fcw,tear ψ Fcw,tear
Fcw,tear
170
150
160
110
120 100
120
80
100
60
60
70
Mco,b
28
Mco,b Mco,b
65
25
85
75
22
Plastic flexural resistance of joints in group E and B is governed by resistance of column web
in tearing Fcw, tear in tension zone and connector in bending and shear M co, b in
compression zone (figure 12). Bearing stresses between connector and column (when the
gap disappeared) are non-uniformly distributed – it is assumed triangle distribution. Plastic
flexural resistance can be predicted using equation:
3
M pl = Fcw, tear ∑ hi + M co, b (15)
i =1
where hi is the distance of i − th component from the centre of compression. Predicted
values are presented in the table 4. In case of group C moment resistance can also be
calculated using formula (15), but now the resistance of the first row is reduced to ψ Fcw, tear
due to essential influence of the component M co, b also in the tension zone.
Table 4. Prediction of the flexural resistance and comparison with experimental data.
Group C Group B Group E
Test results [kNm] 2.36 3.19 3.20
Component method [kNm] 2.48 3.0 3.12
Difference: (component–test)/test [%] 5.1 -5.9 -2.5
Comparison of predicted by means of the component method results with those obtained
from tests shows the high level of accuracy, both for the stiffness and for the flexural
resistance, in case of analysed joints.
One of the main advantages of component method is the possibility to predict mechanical
characteristic of joints without expensive experimental work. Traditionally, up till now, behaviour
of such complex joints as steel pallet rack joints is predicted by testing. Presented in this paper
component approach can be used as complementary method.
The second important advantage of the component method is the possibility of clear and precise
specification of the influence of each component on the resistance and stiffness of the joint. The
weakest component can be easy identified and improved. It makes easier to conduct the
optimisation of joints, especially in such elements, which are produced in long series, as steel
storage pallet racks.
NOTATION
AV shear area
beff effective width
d wc depth of the column web
f y , cw , f u , cw yield and ultimate stress of column section
f y, co , f u , co yield and ultimate stress of connector
t co , t cw , t bf , t tab thickness of the connectors, column web, beam flange and tab
REFERENCES
(1) N. Baldassino, R. Zandonini. (2002). Design by testing of steel storage pallet racks.
Proceedings of the 3rd European Conference on Steel Structures, Coimbra 2002,
Volume I, pp. 689-698.
(2) N. Baldassino, C. Bernuzzi. (2000). Analysis and behaviour of steel storage pallet
racks. Thin-Walled Structures 37 (2000) 277–304.
(3) F. D. Markazi, R. G. Beale, M.H.R. Godley. (1997). Experimental Analysis of Semi-
Rigid Boltless Connectors. Thin-Walled Structures, Vol.28, No.1, pp.57-87.
(4) A. Kozłowski, L. Ślęczka. (2002). Experimental analysis of beam-to-column joints in
steel storage pallet racks. Proceedings of the 3rd European Conference on Steel
Structures, Coimbra 2002, Volume II, pp. 897-906.
(5) prEN 1993-1-8 (2003). Eurocode 3: Design of Steel structures. Part 1.8: Design of
joints. (Stage 49 draft). CEN, Brussels.
(6) prEN 1994-1-1 (2002). Eurocode 4. Design of composite steel and concrete structures.
Part 1.1: General rules and rules for buildings. CEN, Brussels.
(7) J. Fink, D. Rubin, K. Hollmann. (2003) Anwendung des Innsbrucker
Komponentenmodells bei der Optimierung eines modernen Deckenschaltischs.
Stahlbau 72, Heft 1.
(8) W.F. Chen, D.E. Newlin. (1973) Column Web Strength of Beam-to-Column
Connections. Journal of the Structural Division. ASCE. Vol.99, No ST9.
(9) J. R. Benjamin. (1959). Statically Indeterminate Structures. McGRAW-HILL.
(10) C. Faella, V. Piluso, G. Rizzano. (2000). Structural steel semirigid connections. Theory,
Design and Software. CRC Press.
ABSTRACT
The paper summarises the results of an experimental program carried out at
the Politehnica University of Timisoara in order to evaluate the performance of
eaves (knee) and ridge (apex) joints of pitched roof cold formed steel portal
frames under monotonic and cyclic loading. Three different configurations of
ridge and knee joints have been tested. The behaviour and failure mechanisms
of joints have been observed in order to evaluate their rigidity, strength and
ductility.
INTRODUCTION
Previous studies by Lim and Nethercot, 2004 (1) and Chung and Lau, 1999 (2) showed that
bolted joints in cold formed steel portal frames have a semi-rigid behaviour. Also, this type of
joints are partially resistant (Lim and Nethercot 2003 (3), Wong and Chung 2002, (4)). An
important contribution to the global flexibility of the joints, besides the bearing effect (e.g. bolt
hole elongation), is due to the deformation induced by the local buckling or distortion of the
thin walled profiles. In an unwisely configured joint premature local buckling can cause the
failure of the joint itself well below the expected load bearing capacity.
Previous studies focused on monotonic tests only. Present paper summarises the testing
programme developed at the "Politehnica" University of Timisoara, on a series of knee and
apex (ridge) joints used for pitched roof cold-formed steel frames. The joints were tested
under both monotonic and cyclic loading aiming to observe the effect of loading type on the
response parameters. In this paper authors extend the results already reported, Dubina et.
al. (5), with the moment-rotation curves and ductility factors characterising the monotonic and
cyclic behaviour of the joints.
TESTING PROGRAM
Specimens
In order to be able to define realistic specimen configurations a simple pitched roof portal
frame was first designed with the following configuration: span 12 m; bay 5m; height 5m and
roof angle 10°. This frame was subjected to loads common in the Romanian design practice
as follows: self weight 0.35 kN/m2 (γULS=1.1), technological loads 0.15 (γULS=1.1) and
symmetric snow load 0.72 (γULS=2.0). These loads were totalling an approximately 10 kN/m
uniformly distributed load on the frame. The frame was analysed and designed according to
the current EN 1993-1-3 (6) rules. The size of the knee and the ridge specimens, and the
From the design the elements of the portal frame resulted back-to-back built up sections
made by Lindab C350/3.5 (SUB350 - fy=350N/mm2). In accordance to these cross sectional
dimensions three alternative joint configurations, using welded connecting gusset elements
(S235 - fy=235N/mm2), were proposed (see figure 1 and figure 2).
1025 65
220
65
83 -RSG- 1025 65
220
45
M20 gr 6.6 (FB-only) 2 M20 gr.6.6 M20 gr 6.6 (FB-only)
1458 1420
2-2
220
45
1025 65
1420
1535
2354
2354
2354
M20 gr.6.6
M20 gr.6.6 M20 gr.6.6
1 2 3
65
270 65
270 65
65
65
419
419
The connecting bolts are subjected to shear and their design was carried out assuming the
rotation of the joint around the center of the bolt group and a linear distribution of the arising
forces in each bolt, depending on the distance from the rotational center. In the design of the
One group of specimens (KSG and RSG) used spaced gussets, (figure 3.c). In this case,
bolts were provided only on the web of the C350 profile. In the other cases, where two
different details were used for the connecting bracket – e.g. welded I sections only (KIS and
RIS), and welded I section with plate bisector (KIP and RIP), respectively - bolts were
provided on the web only (figure 3.a), or both on the web and the flanges (figure 3.b). The
case where bolts were also on the flanges had in their name the distinctive FB (see table 1).
Test setup
Monotonic and cyclic experiments were made for each specimen typology, all specimens
being tested statically. Figure 4 shows the test setup and specimen instrumentation. For
monotonically loaded specimens the loading velocity was approximately 3.33mm/min, and
the ‘yield’ displacement was determined according to the ECCS procedure (figure 5).
αy vu
vy vmax v
For the cyclic tests several alternative loading procedures were used: (1) the standard ECCS
cyclic procedure, ECCS, 1985 (7), (2) a modified cyclic procedure, suggested by the authors,
which is based on the ECCS proposal and (3) a cyclic procedure for low cycle fatigue. Some
inadequacies were found as far as the ECCS procedure is concerned, and will be discussed
later. As seen in figure 4 several parameters of the response were monitored, but in this
paper only the basic force versus displacement are discussed. All graphs and numeric
values refer to Fact as force and the average of Dglft and Dglsp as displacement.
TEST RESULTS
Monotonic tests
The monotonic tests identified failure modes of the different joint configurations. All
specimens had a failure due to local buckling of the cold formed profiles; however two
distinctive modes were identified for specimens with flange bolts and those without (figure 6
and figure 7).
(a) (b)
Figure 6. Failure of ridge specimens RIS-M (a) and RIP-FB-M (b).
If no bolts are provided on the flange of profile, initially minor bearing elongation of the bolt
holes were observed, the failure being due to stress concentration in the vicinity of first bolt
row. The resulting concentration of compressive stress in the C profile causes the local
buckling of the web (figure 8.a) followed by web-induced flange buckling. This phenomenon
occurred in a similar way in the case of RSG and KSG specimens. No important differences
were observed between specimens where no bolts were provided on the flanges. In the case
of the specimens with flange bolts, the stresses concentrated in the vicinity of the firs bolt row
on the flange. In this case no initial elongation of the bolt holes were observed; the buckling
was firstly initiated in the flange, and only later was extended into the web (Figure 8.b).
(a) (b)
Figure 8. Failure mode of specimens with and without bolts on the flanges (FB).
Comparative experimental curves for ridge and knee connections are presented in figure 10.
There are no significant differences among the specimens without flange bolts (RSG-M, RIP-
M, and KSG-M, KIS-M). The explanation for this is because the connecting bolts had higher
rigidity and capacity compared to the other components of the joint. On the other hand, there
is important gain in load bearing capacity when bolts are installed also on the flanges,
although this solution is more difficult to fabricate (RIS-FB-M and KIS-FB-M).
200 80
RSG-M KSG-M
70
RIP-M KIS-M
160
RIS-FB-M 60 KIS-FB-M
50
120
P (kN)
P (kN)
40
80 30
20
40
10
0 0
0 20 40 60 80 0 50 100 150
v (mm) v (mm)
(a) (b)
Figure 10. Comparative results from monotonic tests for ridge (a) and knee (b) joints.
Based on notations in figure 5 the yield displacement (vy), and corresponding force (Py), were
determined for all monotonically tested specimens. These values are summarised together
with the values of maximum force (Pmax), and the corresponding displacement (vmax) in table
2. The load bearing capacity of the specimens with bolts on the flanges was higher, and only
in this case the capacity of the joint reached the capacity of the connected C profile. It can
also be observed that all specimens have limited ductility. In table 2, PRdth represents the
calculated capacity of the member, while PRd,bth represents the bearing capacity of the bolted
connection. All strength values were calculated using measured characteristics of the cold-
formed steel (fy=452N/mm2 , fu=520N/mm2).
Obviously, the specimens with unbolted flanges that failed prematurely by web buckling due
to stress concentration around the first row of bolts, would be the weakest part of such a
frame. Consequently, this joint typology is not recommended to be used in practice. Table 3,
associated with figure 11 show the characteristic values of moment-rotation curves for
flange-bolted specimens (RIP-FB and KIP-FB).
200 200
150 150
M, kNm
M, kNm
100 100
50 RIP-FB-M 50 KIP-FB-M
0 0
0.000 0.020 0.040 0.060 0.080 0.100 0.120 0.000 0.020 0.040 0.060 0.080 0.100 0.120
φ TOT, rad φ TOT, rad
Comparing figure 10 and figure 11 one observes that the shapes of the F-d and M-φ curves
are similar, as expected. The ductility µ = φu φ y is 1.19 for knee joint KIP-FB-M, and 1.16 for
apex joint – RIP-FB-M. The reduction of the maximum moment (Mmax) and of ultimate rotation
(φu) in the case of RIP-FB-M specimen is due to the effect of axial compression, which is
significant in this case.
Cyclic tests
In case of the cyclic loading, the degradation of the specimens initiated with elongations in
the bolt holes caused by bearing. Compared to monotonic loading, in this case the
phenomenon was, amplified due to the repeated and reversal loading. However, the failure
occurred also by local buckling, as in case of monotonic tests, but at the repeated reversals,
the buckling occurred alternatively on one and the other side of the profile. This repeated
loading caused the initiation of a crack at the corner of the C profile, in 2-3 cycles following
the buckling, closed to the point where the first buckling wave was observed in the flange
(figure 12). The crack gradually opened in the flange and web causing an important decrease
of the load bearing capacity in each consecutive cycle.
An important observation during the cyclic testing was that the recommended ECCS
procedure for cyclic testing proved to be unsuitable for the limited ductility subassemblies.
The increase of displacement from 1vy directly to 2vy is too sudden making it impossible to
asses the characteristics of the cyclic behaviour (figure 13.a - RIP-C1). As it can be
observed, before the attainment of the displacement limit vy there was no significant damage
in the specimen, while at the displacement level 2vy it already failed. Therefore the failure
mode was not different from the one observed during the monotonic tests. For the other
cyclically tested specimens a modified ECCS procedure was used. In this procedure half of
the difference between vmax and vy was used as increment after the reaching of the
200 90
150
60
100
30
50
P (kN)
P (kN)
0 RIP-C1 0
-60 -30 0 30 60 -105 -35 35 105
-50
-30
-100
-60
-150 RIP-C1 KIS-FB-C
RIS-FB-C1 KIP-FB-C
-200 -90
v (mm) v (mm)
(a) (b)
Figure 13. Comparative results from cyclic tests.
The hysteretic curves show a stable behaviour up to the yield limit (vy) with a sudden
decrease of the load bearing capacity afterwards. Therefore the low ductility of the
specimens must be underlined again. Further, the cycles show the effect of slippage in the
joint (i.e. pinching) and strength degradation in repeated cycles. The strength degradation is
stronger in the first repetition, while in the consequent cycles the behaviour is more stable.
Based on the unstabilised envelope of the cyclic curves the strength, capacity and ductility
characteristics of the joints have been determined, and are reported in Table 4. Again, joints
without flange bolts were weaker.
For this reason, as previously, the moment-rotation curves parameters are shown in table 5
and figure 14 for RIP-FB and KIP-FB joints only. Apparently, the cyclic ductility values are
greater than the monotonic ones, at least for positive cycles in case of knee joints. In fact,
this is not true, because the ultimate rotation capacity for KIP-FB-C (positive) is similar with
that of KIP-FB-M, but the yield rotation is reached earlier in cyclic loading. Practically, cyclic
loading makes weaker the knee joint, and the maximum moments clearly show that.
M, kNm
50 50
M, kNm
0 0
-0.100 -0.050 -500.000 0.050 0.100 -0.100 -0.050 -500.000 0.050 0.100
-100 RIFB-C1-env+ KIP-FB-C1-env+
-100
RIFB-C1-env- KIP-FB-C1-env-
-150 -150
RIFB-C1 KIP-FB-C
-200 -200
φ TOT, rad φ TOT, rad
CONCLUDING REMARKS
The experiments described in this paper have been carried out recently and the
interpretation of the results is underway, but already some conclusions can be drawn as
follows.
The calculation model for the connection, based on the linear distribution of the force on
each bolt is not correct. The force distribution is unequal due to the flexibility of the
connected member. In fact, the force is an order of magnitude bigger in the outer bolt rows
compared to most inner one. There are two main components, namely the bearing of the
bolts and the local buckling of the connected profile, which interact and determine both the
rigidity and the load bearing capacity of the joint. A correct model for the behaviour must
include both these components.
A connection with bolts only on the web of the profiles is always partial strength. If the load
bearing capacity of the connected beam is to be matched by the connection strength, bolts
on the flanges become necessary.
The ductility of the connection is limited both under monotonic and cyclic loads and the
design, including the design for earthquake loads, should take into account only the
conventional elastic capacity corrected with safety factors. Because there is no significant
post-elastic strength, there are no significant differences in ductility and capacity of cyclically
tested specimens compared with the monotonic ones. However, if the joints are loaded
under the limit of their maximum capacity, even cyclically, their strength is not too much
affected. Consequently, if the joint detailing and connection components sizing may provide
at least 20% overstrength, the cold-formed steel pitched-roof frames could be classified as
class L of ductility (low) according to EN 1998-1 (8).
NOTATION
REFERENCES
(1) Lim, J.B.P. and Nethercot, D.A. (2004). "Stiffness prediction for bolted moment-
connections between cold-formed steel members", Journal of Constructional Steel
Research, Vol.60, No.1: 85-107
(2) Chung, K.F. and Lau, L. (1999). "Experimental investigation on bolted moment
connections among cold formed steel members", Engineering Structures, Vol.21,
No.10: 898-911
(3) Lim, J.B.P. and Nethercot, D.A. (2003). "Ultimate strength of bolted moment-
connections between cold-formed members", Thin-Walled Structures, Vol.41, No.11:
1019-1039
(4) Wong, M.F. and Chung, K.F. (2002). "Structural behaviour of bolted moment
connections in cold-formed steel beam-column sub-frames", Journal of Constructional
Steel Research, Vol.58, No.2: 253-274
(5) Dubina, D., Stratan, A, Ciutina, A., Fulop, L., Zsolt, N. (2004). "Monotonic and cyclic
performance of joints of cold formed steel portal frames". 4th International Conference
on Thin-walled Structures, ICTWS'2004, Loughborough, UK, 23-24 June 2004 (in
print).
(6) EN 1993-1-3 (2001). "Eurocode 3: Design of steel structures. Part 1-3: General
Rules. Supplementary rules for cold-formed thin gauge members and sheeting". CEN
- European Committee for Standardization.
(7) ECCS (1985). "Recommended Testing Procedure for Assessing the Behaviour of
Structural Steel Elements under Cyclic Loads", European Convention for
Constructional Steelwork, TWG 13 Seismic Design, Report No. 45, 1985
(8) EN 1998-1 (2003). "Eurocode 8: Design of structures for earthquake resistance. Part
1: General rules, seismic actions and rules for buildings". CEN - European Committee
for Standardization.
ABSTRACT
INTRODUCTION
The usual approach to eccentrically loaded bolt groups assumes that the bolts have less
strength than the plates or members they connect. This is called the “weak bolt/strong plate”
model.
When thin plates or members are used, and bolts are close to the edges of plates or
members, the bearing strength of the plate or member at these bolts can be less than the
shear strength of the bolts.
In this case, the “weak bolt/strong plate” model is not correct, because the force distribution
based on bolt shear strength, assumed with this model cannot be achieved.
This paper presents a “weak plate/strong bolt” model which can be used to supplement the
“weak bolt/strong plate” model. The capacity of the connection will be the greater of the
values that result from the two models, but will not exceed the capacity obtained using the
“weak bolt/strong plate” considering only bolt shear.
This is the model that assumes that an eccentrically-loaded bolt group rotates about an
“instantaneous center of rotation” (AISC, 2001). This instantaneous center (ic) determines
the forces and direction of these forces on each bolt. The method can use an elastic or an
inelastic constitutive equation, and the compatibility equation usually assumes that bolt
deformation is linearly proportional to distance from the ic. This method is well established,
However, this model does not allow for the situation that exists when bolts close to an edge
cannot develop the force dictated by the ic location. This is a strong bolt/weak plate situation
where the bolt force is dictated by the bearing strength of the plate, rather than the shear
strength of the bolt.
If the force induced in a bolt close to an edge exceeds the edge distance bearing strength, all
bolts in the group must have their forces reduced in the ratio of the edge distance strength to
the bolt shear strength. This must be done to guarantee that the calculated location of the ic
does not change. If the above ratio is 50%, the capacity of the connection is reduced to 50%
of what the weak bolt/strong plate model would otherwise predict. It can be seen that the
strength of one bolt in a group can degrade the strength of the entire group. The weak
plate/strong bolt model is introduced to mitigate this strength degradation.
Figure 1. Typical bolt group and plate – weak plate/strong bolt model.
Consider Fig. 1. This shows a typical plate and bolt group. In the following, bold face
symbols are used to represent vectors. Point O is any arbitrary point used as an origin. The
centroid of the bolt group will usually be used. Fi is the force on the i-th bolt, ri is the position
of the i-th bolt with respect to the origin, and ei is the edge distance for the i-th bolt,
measured from the edge of the bolt hole to the edge of the plate, along as the line of action
of the force Fi. P is the applied load, ex is the eccentricity with respect to the origin, and i and
j are unit vectors in the x and y directions, respectively.
Note that in Fig. 1, the bolt forces Fi are not necessarily perpendicular to the location radii ri.
As noted above, the point O, used as an origin for the ri vectors, is completely arbitrary, but
the bolt group centroid is a convenient point to use.
There is no ic involved in this method. Instead, the bolt forces are each allowed to achieve
any magnitude and direction that will maximize the design strength of the connection, subject
In the notation of Fig. 1, the problem of the weak plate/strong bolt can be formulated as:
n
Find Fi such that P = ∑ Fi • j → max subject to the constraints:
i =1
n
∑ Fi • i = 0
i =1
n
∑ Vi xFi − e x ixPj = 0
i =1
Fi • Fi ≤ (φ1.2Fu t )2 e 2i i = 1, 2, ---, n
Fi • Fi ≤ φR bs 2 i = 1, 2, ---, n
In the above, t is the plate thickness, Fu is the plate strength, d is the bolt diameter,
e i 2 = e i • e i , and n is the number of bolts. The bolt shear constraint is added to insure that
bolt shear limit state Rbs is not exceeded.
Because of the Lower Bound Theorem of Limit Analysis, any solution to the above problem
will be less than the collapse solution. The problem is formulated as a non-linear
programming problem. The solution, which is the greatest lower bound, approximates the
actual collapse solution and will be less than or at most equal to the collapse solution.
Let the solution to the above problem, the weak plate/strong bolt problem, be denoted by Pp.
If the conventional weak bolt/strong plate problem solution is denoted by Pb, then the
{
proposed capacity is P = max Pp , Pb . }
The simplest of these cases, the two-bolt group subjected to a vertical eccentric load, Case
(a) in Fig. 2, provides a good example of the procedures which can be used to optimise the
connection. First the equations limiting the connection capacity are derived which are
independent of the direction of the forces. These limit states are bolt shear and bearing, and
are defined by the following equations, where η represents the percentage of the vertical
force resisted by the first bolt, s is the bolt spacing, and ex is the eccentricity:
Since the equation for the bolt shear limit state represents the strong-plate/weak-bolt model,
and the bolt capacity is optimised when the bolts vertical load is distributed evenly between
the bolts (η = 0.5) its validity can be proven by re-writing it as a solution for the C-value given
in the Table 7-17 (AISC, 2001) as follows:
1
C=
2
⎛e⎞
0.25 + ⎜ ⎟
⎝s⎠
It should be noted that the values obtained using this equation are slightly higher than those
shown in Table 7-17 because of the empirical nature of the load deformation equation used
by AISC, which results in a maximum normalized force of 0.982 instead of 1.
Next equations for the connection capacity based on bolt tear-out are derived. Under
downward vertical loading it is clear that the top bolt may tear-out through the top of the
plate, while the bottom bolt may tear-out through the side of the plate. This results in the
following capacities:
⎡ ⎤
⎢ ⎥
⎢L φh ⎥
Capacity to − top = φ(1.5)(Fu )(t )⎢ cv − ⎥
⎢ η 2 ⎥
⎛e⎞
⎢ 2 ⎜ ⎟ + η2 ⎥
⎢⎣ ⎝s⎠ ⎥⎦
⎡ ⎤
⎢ ⎥
⎢L φh ⎥
Capacity to −side = φ(1.5)(Fu )(t )⎢ ch − ⎥
⎢ 1− η 2 ⎥
⎛e⎞
⎢ 2 ⎜ ⎟ + (1 − η)2 ⎥
⎢⎣ ⎝s⎠ ⎥⎦
In the above φh is the hole diameter and Lcv, Lch are the vertical and horizontal clear edge
distances.
Having derived the equations for each the of the connection capacities based on each of the
The maximum capacity is found when η = 0. The individual limits are as follows:
26.1
Capacitybs = = 26.1kips
2
⎛ 3⎞
02 + ⎜ ⎟
⎝ 3⎠
26.1
Capacitybs = = 18.5kips
2
(1 − 0)2 + ⎛⎜ 3 ⎞⎟
⎝ 3⎠
⎡ ⎤
⎢ ⎥
⎢ 1.0625 ⎥
Capacity to − top = 0.75(1.5)(58)(0.25)⎢1.5 − ⎥=∞
⎢ 0 2 ⎥
⎛3⎞
⎢ 2 ⎜ ⎟ + 02 ⎥
⎢⎣ ⎝3⎠ ⎥⎦
⎡ ⎤
⎢ ⎥
⎢ 1.0625 ⎥
Capacity to −side = 0.75(1.5)(58)(0.25)⎢ 1.5 − ⎥ = 18.3
⎢1−0 2 ⎥
⎛3⎞
⎢ 2 ⎜ ⎟ + (1 − 0 )2 ⎥
⎢⎣ ⎝3⎠ ⎥⎦
Therefore the capacity of the connection is the minimum value of 18.3 kips. AISC in Table
10-9 reports the capacity of this connection as 23.0 kips. Using η = 0.5 which maximizes the
bolts results in a capacity of 16.7 kips.
It should be noted that when checking bearing and tearout a hybrid of equations J3-2a and
J3-2b in the form of R n = 1.5Lc tFu ≤ 2.4dtFu is employed. This approach has been used for
several reasons, First, the weak plate/ strong bolt model is predicated on the Lower Bound
Theorem of Limit Analysis which guarantees that the applied external forces in equilibrium
with the internal force field are less than or, at most, equal to the applied external force that
would cause failure, provided that all the limit states are satisfied. This theorem is, strictly
speaking, only valid as long as the connection is sufficiently ductile to allow redistribution of
the forces. The use of the 1.5 factor is consistent with this approach since considerable
deformations will occur as the forces in the bolts are distributed. Also since this is a weak
plate/strong bolt model, the use of the 1.5 factor helps to minimize the thickness of the plate,
thereby reducing the likelihood of bolt fracture and increasing the ductility of the connection.
Further, the instantaneous center method itself is based on an assumption that rather large
It appears from Table 1 that the greatest benefit is achieved for cases with fewer bolts, this is
because the horizontal component is larger and the bolt tearout becomes more critical. Of
course as the eccentricity increases, the weak plate/strong bolt model becomes more
advantageous to larger number of bolts as well. The model is also suited to analyze
connections subjected to simultaneous axial and shear loads. A 1/4” plate was chosen for
this example because it can be easily compared to the existing values presented in AISC
Table 10-9. However it is not just the connection plate which can be the critical element, and
some beam webs are thinner than 1/4”. In such cases the weak plate/strong bolt model can
be used to optimize the connection to even greater advantage.
One concern that arises is that the weak plate/strong bolt model does not take into account
the load-deformation behavior of the bolts. It is assumed that the load can redistribute freely
to optimize the connection. This is not a major concern for the simple cases shown here,
because when the bolt shear and not plate bearing or tearout governs the capacity, the
capacity is consistent with the instantaneous center method, which incorporates the load-
deformation behavior of the bolts. However for more complicated geometries the capacity
should be limited to that obtained using the instantaneous center method.
As has been shown, the weak plate/strong bolt model can be used to mitigate the effects of
the new AISC bearing and bolt tear-out requirements. However because of the nature of the
problem, the number of equations to be solved increases exponentially as the number of
bolts involved increases. The authors have developed ad hoc solutions to the three cases
shown in Figure 2, as well as two further cases involving both shear and axial loads.
However a non-linear programming solution still needs to be developed to increase the
usefullness of the model for increased number of bolts with larger eccentricities.
REFERENCES
1. AISC, 2001, Load and Resistance Factor Design, 3rd ed., American Institute of Steel
Construction, Chicago, Illinois, pages 7-7,8 & 7-38 to 7-85
ABSTRACT
INTRODUCTION
Figure 1. Figure 2.
The AISC LRFD, 3rd Edition Manual (1) presents Figure 7-2, reproduced here as Figure 1, to
demonstrate the geometric relationship assumed when using the Instantaneous Center of
Rotation Method to calculate the ultimate capacity of eccentrically loaded bolt groups. This
figure indicates that the instantaneous center is located along a line perpendicular to the
applied load and passing through the center of gravity of the bolt group. Crawford and Kulak
(2) also indicate that this is the intended geometry in their paper. It will be shown however
that it is impossible to satisfy equilibrium and the requisite nonlinear load-deformation
relationship of the bolts with this geometry, except for the case where the load is parallel to
one of the symmetry axes of a doubly symmetric bolt group, or a linear elastic constitutive
INVESTIGATION
In order to prove that equilibrium cannot be satisfied while maintaining both the geometric
constraints of Figure 1 and the load-deformation constraints represented by Figure 2, we will
look at the simple case of a two-bolt group, illustrated in Figure 3 (see section Notation).
From Figure 2 the required load-deformation relationship, based on empirical data, is:
0.55
R i = R ult ⎛⎜1 − e −10∆ ⎞⎟
⎝ ⎠
Since the total deformation at each bolt is assumed to vary linearly with its distance from the
instantaneous center, and the bolt furthest from the instantaneous center is assumed to
reach its ultimate stress when ∆=0.34 inches, the force on each bolt can be calculated as
follows:
0.55
⎛ ⎛ r ⎞⎞
⎜ −10(0.34)⎜⎜ 1 ⎟⎟ ⎟
⎜ ⎝ r2 ⎠ ⎟
R1 = R ult ⎜1 − e ⎟
⎜ ⎟
⎜ ⎟
⎝ ⎠
0.55
R 2 = R ult ⎛⎜1 − e −10(0.34) ⎞⎟ = 0.9815R ult
⎝ ⎠
For simplicity assume that R 2 = 1 . Note that theoretically R 2 should be unity but R 2 =0.982
because of the empirical nature of the load-deformation equation.
0.55
⎛ ⎛ r ⎞⎞
⎜ −10(0.34)⎜⎜ 1 ⎟⎟ ⎟
⎜ ⎝ r2 ⎠ ⎟
Also for simplicity introduce η = ⎜1 − e ⎟ .
⎜ ⎟
⎜ ⎟
⎝ ⎠
Rx
The tangent of the applied load from Figure 3 is tan θ =
Ry
From the geometry the component forces on the bolts can be calculated as:
r2 y r
R 2x = R ult R 2 y = R ult 2 x
r2 r2
r1y r
R 1x = R ult η R 1y = R ult η 1x
r1 r1
r1y r2 y
η +
tan θ R =
∑ Rx
=
r1 r2
∑Ry ⎡η 1 ⎤
r0 x ⎢ + ⎥
⎣ r1 r2 ⎦
Since tan θ R must equal tan θ for equilibrium to be satisfied it can be shown that:
r1y r2 y ⎡η 1 ⎤
η + = r0 y ⎢ + ⎥ (1)
r1 r2 ⎣ r1 r2 ⎦
It can also be shown that this equation can only be satisfied when r1y = r2 y = r0 y which only
occurs when θ = 0 with a doubly symmetric bolt group or when η = r1 / r2 which is the linearly
elastic case.
Having proven that the prescribed geometry cannot be satisfied for the two bolt case, it can
also be demonstrated for other bolt groups by moving the instantaneous center along the line
perpendicular to the applied load and passing through the centroid of the bolt group and then
plotting the resultant theta angle, θ R , as shown for a three bolt group in Figure 4. As can be
seen from the graph θ R is asymptotic to the value of θ , but will never equal θ for a finite r0.
Therefore, equilibrium cannot be achieved.
Figure 4.
Finding errors in the presentation of the theory underlying the calculation of the
instantaneous center calls into question the values in Tables 7-17 through 7-24 presented in
the AISC Manual (1). Fortunately these values were produced by a program based on
Brandt’s work (Brandt (3, 4)). Brandt’s procedure never restricts its search for the
Figure 3.
1. AISC, 2001, Load and Resistance Factor Design, 3rd ed., American Institute of Steel
Construction, Chicago, Illinois, pages 7-7,8 & 7-38 to 7-85
2. Crawford, S. F. and G. L. Kulak Eccentrically Loaded Bolted Connections Journal of
the Structural Division, ASCE, Vol. 97, ST3, March 1971, pages 765-783.
3. Brandt, G. Donald: Rapid Determination of Ultimate Strength of Eccentrically Loaded
Bolt Groups. Engineering Journal, American Institute of Steel Construction, Vol. 19, No.
2, 2nd Quarter 1982.
4. Brandt, G. Donald: A General Solution for Eccentric Loads on Weld Groups.
Engineering Journal, American Institute of Steel Construction, Vol. 19, No. 3, 3rd
Quarter 1982.
ABSTRACT
The present research work concerns the laboratory testing of a full strength
bolted splice connection for a HEB100 profile. Taking as reference results
those obtained from a continuous intact beam subjected to a concentrated
force at the middle of its span, eight simply supported beams with full strength
bolted splice connections have been tested. The obtained test results
significantly diverge from the reference ones obtained from the intact beam.
INTRODUCTION
The use of non-pretensioned bolts in full strength splice connections is a common practice in
several countries (1)(2). However, in such splices first due to the difference between the
diameter of bolts and the respective holes and second due to a possible preliminary
imperfection (rotation of the connected parts of the beam), an additional significant deflection
is caused as soon as the loading is applied on the beam and before its value reaches the
calculated value.
Scope of the present research work was to experimentally define the relationship between
the moment at the middle of the span and the deflection at the same point. As a matter of
fact, the latter is tightly connected to the Serviceability Limit State of the beam, even before
the total loading at the middle of the span has been applied (3).
For the laboratory tests, HEB100-beams have been selected because the chosen profile is
very often used as a structural member in a plethora of steel structures (cf. e.g. the purlins of
steel roofs). The material used was Fe 360 and the non-pretensioned bolts were M12-8.8
and M14-8.8. Four groups of different connections were used, once with bolts M12 and once
with bolts M14. Two steel plates having cross section 6x80 mm were used to connect the
webs, whereas another two with cross section 12x100 mm to connect the flanges.
The full strength bolted splices have been designed and constructed so that they exhibit at
least the same load-bearing capacity as the intact beam (4).
The maximum moment and shear force that the selected profile HEB 100 can carry under
the plastic limits of strength are equal to:
For the connection of the two separated parts of the beam, two horizontal steel plates with
cross section 10*1,2 cm were used for the flanges and two vertical ones with a cross section
0,6*5,5 cm for the web (cf. e.g. Figure 1).
Figure 1. The splice connection (left) and the vertical additional plate (right).
The additional horizontal plates on the flanges were used to transfer the moment in the
connection area, whereas the additional vertical ones were used to carry the shear forces.
Because of the symmetry of beam and loading, only half of a vertical plate was drawn and
calculated. In Figure 1 a half of one of the two additional vertical plates is shown where the
geometrical position of the center of the holes is the point of the load application. Note that in
such splices the shear forces are not the principal problem. The same position and the same
kind of bolts were used for all specimens. The experimental results showed that these
vertical additional parts of the connection did not take or carry any force during the
experiment.
In the following equations, calculations and symbols from EC3 were used (4). A 6 bolt
connection was used for M14-8.8 bolts for the half of the connection of a flange. The position
of the bolts is shown in Figure 1. The horizontal distance between the two connected pieces
of the beam was equal to 1,5 cm for all the specimens.
Vx=0 Vy=12,828+34,635=47,463 kN
Two groups of four different positioning of the bolts were used for the specimens. For the first
group, bolts type M12-8.8 were used calculated for the connections on the flanges. For the
second group, bolts type M14-8.8 were used in an effort to optimize the deformation of the
connected areas. Calculations were not repeated because the change in the diameter of the
bolts was in the safe side.
SPECIMENS B1 AND B2
The suggested position of the bolts is shown in Figure 2 (diameters for the holes do =13 and
do =15 mm) corresponding to specimens B1 and B2.
Figure 2. The position of the holes in the additional horizontal splice plates and flanges
(specimens B1 and B2).
SPECIMENS B3 AND B4
In an effort to add more stiffness to the area of the connection in order to obtain better results
during deformation, it was decided that the distance between the holes to be greater in the
specimens B3 and B4 than the distance in the specimens B1 and B2. The positions of the
bolts in specimens B3 and B4 is shown in Figure 3.
Similarly to the calculation for specimens B1, B2, the present one leads to safe results.
SPECIMENS B5 AND B6
Figure 4. The position of the holes in the additional splice plates and the flanges
(specimens B5 and B6).
Figure 5. The position of the holes in the additional splice plates and the flanges for
specimens B7, B8.
SPECIMEN B9 (REFERENCE)
This specimen was used as an intact beam without any kind of splice connection. The
respective test results were used as reference for the rest experimental results of the
specimens B1 - B8.
All the experiments were performed at the Laboratory of the Institute of Steel Structures,
Department of Civil Engineering, Aristotle University of Thessaloniki, Greece. All the
specimens used were simply supported beams with a span of 1 meter and a concentrated
vertical force in the middle of their span.
A composite hydraulic machine was used with an upper limit of force equal to 500 kN. A
computer was used to record and store 1000 pairs of values corresponding to the force and
the vertical displacement for each one of the specimens. In the sequel, the diagrams
corresponding to each one of the specimens from B1 to B8 in parallel with the diagram of the
reference beam B9 in Figures 6 and 7 are depicted. Figure 8 shows the diagrams for all the 9
specimens including the reference one in order to give a general comparison of the splices
test results with those of the intact beam.
250 250
200 200
B9
kN
kN
150 150
100 100
B9
50
B1 50
B2
0 0
0 10 20 30 40 50 0 10 20 30 40 50 60 70
mm mm
300 300
250 250
200 200
B9
kN
kN
150 150
100
B9 100
50
B3 50
B4
0 0
0 10 20 30 40 50 0 10 20 30 40 50 60 70
mm mm
Figure 6. Figures of the specimens and the diagrams for B1, B2, B3, B4 and B9.
250 250
200 200
B9
kN
kN
150 150
100 100 B9
50
B5 B6
50
0 0
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
mm mm
300 300
250 250
200 200
kN
kN
150 150
100 B9 100
B9
50
B7 50
B8
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60 70
mm mm
Figure 7. Figures of the specimens and the diagrams for B5, B6, B7, B8 and B9.
250
200
kN
150
100
50
0
0 10 20 30 40 50 60 70
mm
B1 B2 B3 B4 B5 B6 B7 B8 B9 (REF.)
Figure 8. Force-deflection diagrams for the connected beams from B1 to B8 and the
reference beam B9.
The diagrams of the test results (Figures 6, 7, 8) for the connected beams from B1 to B8
gave a significant deflection that corresponds to a much lower force than this that
corresponds to the reference curve of beam B9. In particular, only the 1/10 of the load used
for reference beam B9 was used to give the same deflection to the connected beams.
After the end of the testing program, it was observed that all the vertical additional pieces of
the web remained undeformed (intact) due to the fact that they didn’t carry any loads. The
latter certifies the initial assumption taken into account in the calculation of the connection.
Having in mind that it is always critical to check the serviceability limit state of such beams,
the failure of the specimens due to excessive deflection around the area of the splice
connection is obvious. The maximum acceptable percentage according to EC3 § 4.3.2 is
equal to 1/200=0.05% for total loading (4). The experimental results gave a deflection of
approximately 15/1000=1.5% below the 1/10 of the total load.
Two were the reasons for the observed significant deflections around the area of the splice
connections: The first is the difference of 1 mm between the diameters of the holes and
those of the bolts. The existence of the clearance gives a kind of freedom of sliding as soon
as a small (1/10) load is applied on the beam. At this first step, the force loaded only the two
additional horizontal plates on the flanges which reacted as autonomous simply supported
beams instead of the full cross-section of the beam. This is the reason for the existence of
linear part at the beginning of the diagrams for the connected beams from 0 to near 15 mm.
The second reason of the observed significant deflection is due to the insertion of the thread
of the bolts into the mass of the steel in the vicinity of the contact area of the holes, as shown
in Figure 9.
Figure 9. The holes (with the traces of the insertion of the thread into the steel mass) after
the experiment in an additional horizontal splice plate and in the beam flange.
ACKNOWLEDGEMENTS
The work reported here has been partially supported by the European Union Research and
Training Network (RTN) “Smart Systems. New Materials, Adaptive Systems and their
Nonlinearities. Modeling, Control and Numerical Simulation”, with contract number HPRN-
CT-2002-00284.
NOTATION
d diameter
fu ultimate stress
fy yield stress
t thickness
α distance
REFERENCES
(1) Ivanyi, M. & Baniotopoulos, C.C. (2000) (eds), Semi-rigid Joints in Structural Steelwork,
Springer Wien, New York, p. 350.
(2) Baniotopoulos, C. C. & Wald, F. (2000) (eds), The Paramount Role of Joints into the
Reliable Response of Structures. From the Classic Pinned and Rigid Joints to the
Notion of Semi-rigidity, Kluwer, Dordrecht, p. 480.
(3) Kontoleon, M. J., Kaziolas, D. N., Zygomalas M.D. & Baniotopoulos, C. C. (2003),
Analysis of Steel Bolted Connections by Means of a Nonsmooth Optimization
Procedure, COMPUTERS & STRUCTURES 81, 2455-2465.
(4) ENV1993-1-1 (1993). Eurocode 3, Design of Steel Structures. CEN, Brussels.
ABSTRACT
Bolted connections of hot dip galvanized steel members with punched holes
are usually taken for mast and tower constructions. These structures are often
loaded by wind, so it makes good sense to examine the fatigue behaviour of
the connections. One of the findings described in this paper is that the
punching of the holes and the galvanizing process have a negative influence
on fatigue behaviour. This poses the question: how can the fatigue resistance
of bolted steel connections of galvanized steel members with punched holes
be improved? The main idea is to use preloaded bolts. A preload perpendicular
to the surface of the members protects the area around the hole by reducing
the notch stresses. This means the structures can be strengthened and their
lifetime will be prolonged.
INTRODUCTION
Another important influence on these properties is the production of the bolt holes
themselves. For the profiles used, such as angle profiles, it is much cheaper to punch the
holes than to drill them. Punching produces a very heavy impact on the steel until a severe
material hardening occurs around the holes. The ductility is reduced tremendously, the
ultimate tensile strength is usually affected and cracks may be initiated. The area affected
extends about 2 to 3 mm around the holes. This is the reason why reaming of the holes is
often required with thicker plate material. Hot dip galvanizing (with its high temperature) that
takes place after punching may also promote the aging of the material. All these influences
have a negative effect the fatigue behaviour of the connections.
STATIC TESTS
Before we investigated this special problem it was necessary to gain knowledge on the
influence of punching and hot dip galvanizing on the load carrying behaviour of bolted
Figures 2 and 3 show the difference of the rupture modes of two connections which were
designed to break by net section rupture. In figure 2 we see the net section of a member with
a drilled hole; only one crack runs through the section with a high reduction in plastic
thickness reduction due to Poisson's ratio. By contrast, figure 3 shows the net section of a
member with a punched hole. Many initial cracks start from the hole wall, and one of them
finally runs through the net section but without any reduction in plastic thickness. This
confirms the previous assumption that the ductility is very much reduced.
Figure 2. Net section rupture of a member Figure 3. Net section rupture of a member
with a drilled hole in a static with a punched hole in a static
short term test. short term test.
The structure of the material at the edge of a punched hole was examined (3). Microscopic
photos were taken at different levels of the thickness of the plate. They show how much the
material along the hole wall changes its grain structure when the punch is driven through it.
Just after the entrance of the punch a special zone at the upper part of the plate is formed,
figure 4a. In this first zone the material flows in the cutting direction of the punch from the
edge into the hole. The plastic flow passes over to the second zone in the middle of the plate
thickness, figure 4b. Then shear cutting takes place, and is followed by the third zone where
the punch leaves the plate, shown in figure 4c. Here a rougher surface is built which is
widened conically.
In addition, the distribution of the hardness in different sections parallel to the surface at
intervals of 0.5 mm has been worked out (4). In an adequate distance from the hole edge the
average value of the Vickers hardness is 150 HV 0.2 for the uninfluenced area. From 2.5 mm
distance up to the hole edge we found an increase of the hardness up to a value of about
330 HV 0.2. Figure 5 shows the effect of cold-work hardening due to punching.
Many slip tests under static load have been carried out over the years. In the following some
actual friction coefficients µ of double lap joints with different surface treatments are
represented
Specimen with pure zinc on its surfaces were dipped into the zinc bath without removing the
pure zinc layer afterwards, they had a glossy appearance. To separate and remove the pure
zinc layer the members were put in a centrifuge after galvanizing. The appearance of these
specimen was matte and lacklustre.
Friction coefficients for different surface treatments which result from former slip tests can be
find in the literature.
As a result it can be seen that the friction paint produces a remarkable increase of the friction
capacity. But it is necessary to check under which percentage of that value µ the connection
does not slip-creep and maintains its rigidity under static long term loads. Therefore, we
examined the creep behaviour of such connections (5).
After these tests the question came up as to whether the friction capacity µ may be
influenced by a long term loading, such as included by the creep tests. Therefore, the
behaviour of the slip connection with hot dip galvanized contact surfaces having alkali-zinc-
silicate friction paint has been tested after creep tests, while in the creep tests the actual
friction has been exploited only up to 60% till 70%. The subsequent slip tests were performed
until slip occurred and they should give the remaining friction capacity. Figure 8 shows in the
left diagram the results of the initial short term slip tests with HVM16 and HVM20 bolts. The
series II and III in the middle of the diagram show the ultimate slip capacities after long term
tests (no reduction!). Series IV at right hand in the diagram show the slip capacities after
fatigue tests with two million load cycles (no reduction!).
Figure 8. Behaviour of the slip coefficient of hot dip galvanized surfaces with alkali-
silicate-zinc-paint during long term testing and fatigue loading.
A last question: Is there constancy or an increase of the displacement during fatigue loading
up to more than one million (M) cycles? The solution is illustrated by figure 9. Under high
exploitation of about 70% of the static slip capacity and after 1.1M to 2M cycles the
displacement ended in a maximum 0.12 mm. We see a kind of hysteresis which converges
on a value much below the limit of 0.300 mm. Nine more tests show the same, or even
better, results.
Further investigation have been performed on hot dip galvanized steel members with
punched slotted holes, friction paint and high strength preloaded bolts. Results can be taken
from the literature (6).
A frequently asked question is: Can we rely on a certain constancy of the bolt preload FV or
do we need to take into account a considerable reduction? This takes place since in a steel
package there are at least four galvanized zinc layers on the members, another four
galvanized zinc layers on the washers and two layers on the bolt head and nut when
clamped together. The total amount of zinc layer may add up to 0.5 mm, which surely creeps
under the high preload of the bolt. Therefore, a second tightening procedure should go over
the connections after two hours or an overtightening of the bolts should be envisaged.
But what is the amount of the preload reduction? Since in our tests the preload is
permanently measured, we can give relevant information on this topic. Apart from that there
is the possibility of overtightening the bolts by a well defined percentage to cover the creep of
the bolt force. Rotation-preload diagrams of six tightening tests with HV bolts M 20x100 -
10.9 show that an additional rotation of 20° makes an increase of about 10% of the required
preload according to DIN 18800-7, EC 3 and EN 1090, which usually covers the creep
influence.
Masts and towers are, in addition to their dead load, mainly loaded by wind. The wind may
come from different directions, leading to oscillating movements and stresses in the structure
and in the members as well as in the joints. The resulting number of cycles can be very high
and is in the range of the fatigue strength. The use of bearing-type connections in hot dip
galvanized members with punched holes leads to questions about the load carrying
behaviour of these joints under cyclic loading. The knowledge of the fatigue behaviour of the
joints is necessary to predict the lifetime of the structures.
In earlier research projects at the TUHH, low-cycle-fatigue and high-cycle fatigue tests were
carried out. Members with holes and bearing-type connections with both punched and drilled
holes, but without any preload of the fasteners were examined (7, 8). The test specimens
consisted of S 235 JR G2 (formerly: RSt 37-2) and the loading was of simple sinus wave
form, while the ratio κ between the lower and upper tension in the net section was +0.1. The
test specimen for the shear connection (type 1) and the member with a hole (type 2) are
shown in figure 10. The bolts were hand tightened after they had contact with the wall of the
hole in order to exclude the influence of friction due to preloading of the bolts. The
dimensions of the test specimens were selected primarily to produce a net section fatigue
failure of the middle member, and not a failure in the bolts or cover plates.
After the fatigue failure of the test specimens we examined the crack surfaces to get
information about the crack initiation. Figure 11 shows two representative net section failures
of members with a hole. It was remarkable that the starting point of the crack front for
punched holes lies in the first zone on the upper corner of the hole edge (figure 11a) and
never in the third zone where the punch leaves the plate. From here the fatigue crack front
runs through the material until the net section is so much weaken that finally a static crack
causes the rupture. In members with drilled holes a surface crack at the wall of hole is
predominant (figure 11b).
The analysis of the fatigue tests happens grafically as S-N curve (stress S over load
cycles N) in log-log diagrams. The higher the curve in the diagram the higher is the fatigue
resistance of the test specimen.
The experimental S-N curves of different hot dip galvanized test specimens are shown in
figure 12. Members with hole and bearing-type connections are compared. As expected, the
members with a hole were able to withstand a higher stress range ∆σ at the same number of
cycles N up to failure than the joints. A comparison between the test specimen with punched
holes and the test specimen with drilled holes shows the negative influence of punching. The
S-N curve for both different structural members with punched holes lies under the
corresponding S-N curve for drilled holes.
Figure 12. S-N-curves of hot dip galvanized members with hole and shear-
bearing connections with drilled and punched holes, FV = 0
(L = member with a hole, V = bearing-type connection s = punched hole,
b = drilled hole, f = hot dip galvanized, κ = stress relation).
Fatigue tests of hot dip galvanized and non-galvanized bolted connections with punched and
drilled holes were carried out to determine the influence of galvanization performed after
punching or drilling. The results are compared in figure 13. It can be seen clearly that the
fatigue life decreases owing to galvanizing and punching. This is the case for members with
The relation of these results to the corresponding S-N curves of the EC 3 is of great interest
with regard to the position of the points of failure and the slope of the S-N curve relative to
the S-N curves of EC 3. The slope of the fatigue strength curves of bolted joints lies between
m = 5.9 and m = 6.7. The results show that the slope of the S-N curves with m = 3.0 is too
high for these experimental curves.
In literature (9) we found other fatigue tests of connections in comparison with EC 3. The
result of the experimental tests is the same as before. Most of the fatigue life data lies below
the S-N curve of EC 3. This leads to the question: Is the EC 3 is suitable for use? Because
the detail category 112 is specifically permitted to non-preloaded high strength bolts.
The idea of covering the area around the hole by a high strength preloaded bolt can be
transmitted to the problem of fatigue behaviour of hot dip galvanized steel members with
punched holes. In a 2003 finished research project on this topic many fatigue tests were
carried out at the TUHH. The aim was the strengthening of the fatigue resistance of the
constructions. For this the negative influence of punching and galvanizing on material data
near to the hole must be reduced. To ensure the comparison between the fatigue tests of the
earlier research project of the members without any preload, we use the same geometry of
the test specimen for this project (see figure 10).
In the diagram of figure 14 the results of various fatigue tests can be seen. Hot dip
galvanized bearing-type connections with punched holes were compared with friction-type
connections having 50% and 100% preloaded high strength bolts and with an unnotched
specimen. The open circles and the lowest S-N-curve show the load cycles of connections
with non preloaded bolts. The full grey circles represent the results of equivalent connections
with preloaded bolts tightened up to 50% of the required preload for a high strength bold
M16. Even a preload of only 50% increases the fatigue behaviour enormously. The increase
can be better expressed by the stress range ∆σnet than by the number of cycles. We found a
step from 72 N/mm2 to about 180 N/mm2 at 1 million cycles. There was a further increase of
fatigue strength up to 233 N/mm2 at a preload of 100% of the required preload (full black
circles). This value is near to the fatigue failure of unnotched bars. They reach a stress range
∆σnet of 280 N/mm2 also regarded at 1 million cycles.
These regarded S-N curves for the friction-type connections have a friction coefficient µ of
0.38 between the middle member and the cover plates. This value is valid for hot dip
galvanized surfaces without any pre-treatment. As a consequence of the influence of friction
identified previously, we painted the surface of the members with alkali-silicate coating to
increase the friction coefficient. With this treatment we achieve a value of µ = 0.6. The results
of the fatigue tests are also shown in figure 14. The open grey circles represent the results of
tests with 50% preloaded bolts and the open black circles the results with 100% preload. It
can be seen that it is possible to achieve a further increase of the fatigue resistance in
comparison with the corresponding S-N curves for friction-type connections without any
treatment of the hot dip galvanized surface.
In the next diagram (figure 15) we compared the S-N curves again with the corresponding S-
N curves for non-galvanized connections with drilled holes and with the S-N curve of
EUROCODE 3 (detail category 112). As seen before, the preload increases the fatigue
behaviour of the connections. One interesting scientific finding is that the negative effect of
punching in combination with galvanizing is neutralized because the experimental S-N
curves for 50% and 100% are nearly identical. And these S-N-curves now lie above the S-N
curve of the detail category 112 of EC 3 for high number of load cycles.
The problem which is still exists is the evaluation of the slope with m = 3.0. The real
behaviour of the tests shows that the slope is smaller than m = 3.0. In particular, the preload
causes a smaller slope, which can be explained by the crack growth. The crack can only be
observed in the protected area under the washers. If the crack leaves this area the rupture of
Besides the fatigue resistance the crack growth is of great interest. The literature gives many
information about the stress distribution of members with cracked and uncracked net
sections but we found nothing about the influence of the pressure under the washers of a
preloaded high strength bolt relating to this problem. Based on experimental tests and finite
element parameter studies we did further investigations to clear the effects of the preload on
the notch stresses near to the hole, the crack initiation and the crack growth. Finally a
simplified calculation method for the lifetime prediction of the steel members was developed
(11, 12).
With that a large test programme with static, low-cycle- and high-cycle-fatigue tests is
finished which had the aim to give answer about the bearing capacity and the fatigue
behaviour of hot dip galvanized steel members with punched holes.
CONCLUSION
The idea of protecting the net section area around a punched hole of a hot dip galvanized
member by the use of preloaded high strength bolts with two washers has shown that a
remarkable influence on the fatigue life can be achieved. The advantage of this method is the
ease of handling with maximum of efficiency. Even the negative influence of punching in
combination with hot dip galvanizing can be neutralized.
ACKNOWLEDGEMENTS
REFERENCES
ABSTRACT
More and more ancient and historic steel structures need repair and restora-
tion. Sometimes the authorities are on the way to decide to take the complete
building down and replace it by a new one. In many cases the “Authority for the
Preservation of Ancient Monuments” (Denkmalschutz-Behörde) insists on
equal looking and identical constructions such as trusses, arches and so on.
Especially the connections have to look similar to riveted connections. To seek
workers on the market who are skilled for the job of hot driven riveting is hope-
less. Therefore the rivets have to be replaced by specially designed high
strength bolts with round heads and similar looking nuts. One German com-
pany specialised in this field and created the so called “Rundkopf-HV-
Schrauben AF” (Round head HV-bolt assembly AF) which looks like a rivet
when installed, because the hexagon for the tightening process has been re-
duced to a very small height. From a distance of 1 to 2 m the hexagon nearly
cannot be identified. These bolts have been accepted in a recent greater re-
construction work at the Frankfurt central station. In tightening tests these bolts
behave different from normal HV-bolts as you will see in the document.
INTRODUCTION
More and more ancient steel constructions come under control because of their further ser-
viceability for the origin purpose. I name some of them here such as the big four bay hall of
the Frankfurt main station (Figure 1), the equivalent buildings in Cologne and in Hamburg-
Dammtor as well as in Liverpool many years ago. Very often the experts come to the conclu-
sion that the building, especially the steel construction, needs a complete restoration since
corrosion has done its work over many years and nobody took any action to start earlier with
the restoration. Not seldom the discussion ends with the result, that at least a replacing of
more or less parts of the construction is necessary. Before such actions start usually the au-
thority for the preservation of historic monuments (in Germany the Denkmalschutzbehörde)
play an important role about the aspect of such buildings after the repair. And they put their
recommendations and requirements on the table, which generally say:
After the reconstruction the steel construction must look the same way as before.
This means that the design engineers cannot work as in the case of a new building but they
must deal with historic methods of calculation, fabrication, erection and material. Often old
books from former times help, but sometimes special skills are gone and nobody has any
idea how to recover them.
Such an ancient skill p. e. is the hot riveting of steel constructions on site. It needs special
tools and a clever group of craftsmen on one hand and it is much to expensive today as
every handwork is. If the main construction parts of these halls such as the main girders are
riveted spatial lattice truss girders, arches or multi bay arches the construction firm cannot
but produce similar looking members either in the workshop for transport or on the site. The
individual members have to be connected on site with simple and cheap rivet similar looking
fasteners. The authority for the preservation of historic monuments (Denkmalschutzbehörde)
will not at all give any permission to change something and therefore the bolt manufacturers
have to have new ideas to make a bolt looking like a rivet, make it high strength, make every
wanted diameter and not only the bolt but also the nut, even if one of them, bolt head or nut,
is positioned inside a spatial truss girder which cannot easily be seen.
The well known German bolt manufacturer August Friedberg in Gelsenkirchen developed
such a type of bolt, which should be installed in the steel construction building in Frank-
furt/Main central railway station. The bolts have a half round head with a very small but high
enough hexagon part on it next to the washer. The nut was formed similarly. The length had
to be adjusted so that when tightened the bolt end fits into the half round of the nut. All bolts,
nuts and washers should be hot dip galvanized.
Since the dimensions of those bolts and nuts differ in a certain way from the normal dimen-
sions of high strength bolts the building authority required an independent expertise about
their installation and performance. This expertise was not possible without a test series on
M16 and M20 bolts. We have been asked to perform such test series and evaluate all nec-
essary tightening and working data.
The many many discussions with other expert colleagues in the international group ECCS
TC 10 Connections and in the CEN/TC 185/WG 6 - most of these colleagues I see in the
auditorium - and a lot of similar tests in cooperation with friends of the „bolt experts family“
and literature (1, 2, 3, 4) gave us the capability to do the tests also in a very short time.
The Figures 2, 3 and 4 show drawings of the bolts, nuts and special distance washers.
We performed the following 4 test programmes tightening either the nuts or the bolt head
The continuous tightening was performed with motor and gearbox, so that the speed was
about 6 rounds per minute. While tightening the following measurement were taken stepwise
all 15° of rotation:
The distance between the two measurements was always 15° rotation. They have been reg-
istered by the computer and immediately after the individual test presented on the screen.
Figure 5 shows such a type of test diagram for the rotation(ϕ)-preload(FV)-behaviour.
We have collected all three important test data as rotation(ϕ), torque(MA) and preload(FV).
They can be drawn one against the other in diagrams. We produced three diagrams of each
test data set such as
150
100
preload [kN]
50
0
0 60 120 180 240 300 360 420 480 540 600
rotation [°]
From these diagrams all required proof data, which are specified in the prEN 14399-1, and
prEN 14399-2 (Fitness for purpose documents) can be read.
250
HV-Schrauben
200 nach DIN 18800 T7:
FV = 160 kN
150
preload [kN]
100 10,9-Halbrundkopf-HV-
Schrauben
red.FV = 120 kN
50
0
0 60 120 180 240 300 360 420 480 540 600
rotation [°]
250
MA = 450 Nm = required torque HV M 20
HV-Schrauben
200 acc. DIN 18800 T7:
FV = 160 kN
preload [kN]
150
50
0
0 100 200 300 400 500 600 700 800 900 1000
torque [Nm]
900
800
700
10,9-Half-round-head HV bolt
600
torque [Nm]
M A = 500 Nm
500
400
HV-bolts
acc. DIN 18800 T7:
300
M A = 450 Nm
200
100
0
0 60 120 180 240 300 360 420 480 540 600
rotation [°]
Figure 8. Rotation (ϕ) versus torque (MA)-diagram for 15 bolts M20, nut tightening.
HV-bolts
200 acc. DIN 18800 T7:
FV = 160 kN
150
preload [kN]
100
10.9-Half round head HV bolts
red.FV = 100 kN
50
0
0 60 120 180 240 300 360 420 480 540 600
rotation [°]
250
M A = 450 Nm = required torque for HV M 20
HV-bolts
200 acc. DIN 18800 T7:
F V = 160 kN
preload [kN]
150
100
10.9-Half round head HV Scatter ∆M A = 71 Nm
bolts, redF V = 100 kN for reduced preload red.F V = 100 kN
50
0
0 200 400 600 800 1000 1200
torque [Nm]
Figure 10. Torque (MA)-preload (FV)-diagram for 6 bolts M20, head tightening.
1300
1200
1100
1000
900 10.9-Half round head HV bolt
800 MA = 700 Nm
torque [Nm]
700
600
500
400
HV-bolts
300 acc. DIN 18800 T7:
200 MA = 450 Nm
100
0
0 60 120 180 240 300 360 420 480 540 600
rotation [°]
Figure 11. Rotation (ϕ) – torque (MA)-diagram for 6 bolts M20, head tightening.
150
HV-bolts
acc. DIN 18800 T 7:
FV = 100 kN
100
preload [kN]
0
0 60 120 180 240 300 360 420 480 540 600
rotation [°]
150
HV-bolts
acc. DIN 18800 T 7:
FV = 100 kN
AF-Half round head HV
100 bolts M16x100:
red.FV = 75 kN
preload [kN]
Scatter ∆MA = 75 Nm
for red. preload red.FV = 75 kN
50
0
0 50 100 150 200 250 300 350 400 450 500 550 600
torque [Nm]
600
550
500
450
400 AF-Half round head HV
bolts, M16x100
torque [Nm]
350
MA = 275 Nm
300
250
200 HV-bolts,
acc. DIN 18800 T7:
150
MA = 250 Nm
100
50
0
0 60 120 180 240 300 360 420 480 540 600
rotation [°]
HV-bolts
acc. DIN 18800 T 7:
FV = 100 kN
100
preload [kN]
0
0 60 120 180 240 300 360 420 480 540 600
rotation [°]
150 MA = 250 Nm
HV-bolts
acc. DIN 18800 T 7:
FV = 100 kN
100
preload [kN]
0
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700
torque [Nm]
700
600
500
torque [Nm]
400
200
0
0 60 120 180 240 300 360 420 480 540 600
rotation [°]
In the diagrams Figures 6 to 17 horizontal and vertical full lines mark the required data for the
preloads and torques as they are mentioned in the German standard, while the dashed lines
show the results for half round head HV-bolts and nuts.
The bolts were always put from the outside into the hole, the nut was inside the spatial truss
girder. In those cases, where the full preload was necessary tightening from the inside by the
nut has been proposed with the result: 120 kN for M20 bolts and 75 kN for M16 bolts. In
those cases where only a partial preload was necessary such as in bearing type connections
the tightening could be managed by bolt head rotating from the outside. Since in those cases
no further lubrication has been used the scatter in preload was much greater than usual, the
recommended torque had to be higher and the reliable preload was smaller: 90 kN for M20
bolts and 60 kN for M16 bolts.
The reason why the full preload of normal high strength bolts could not be reaches, was the
fact that the size of the nut is much different, the nut has less material to withstand the tan-
gential tension stress with very small deformation (see Figure 3).
CONCLUSION
Half round head HV-bolts and nuts can be used instead of rivets for the restoration of old his-
toric steel constructions to preserve them as an ancient building monument. The carrying ca-
pacity is somewhat smaller than known from usual high strength bolts. I would like to rec-
ommend to perform more tests to get a good knowledge about the special behaviour of such
bolts in erection and in service.
(1) Joint Program: Fitness for Purpose Tests of High Strength Bolt Assemblies with High
Strength Nuts according to prEN 783. Report des Arbeitsbereichs Stahlbau und Holz-
bau der Technischen Universität Hamburg-Harburg, Hamburg 1994.
(2) Report no. 1: International Tightening Tests with High Strength Preloaded Bolts of the
Quality 10.9 according to the System HR. Common Report of the Research Centre of
the Belgian Metalworking Industries, Liège, and the Department of Stahlbau und Holz-
bau der Technischen Universität Hamburg-Harburg, Hamburg 1997.
(3) Report no. 2: International Tightening Tests with High Strength Preloaded Bolts of the
Quality 10.9 according to DIN 6914 – 6916. Common Report of the Research Centre of
the Belgian Metalworking Industries, Liège, and the Department of Stahlbau und Holz-
bau der Technischen Universität Hamburg-Harburg, Hamburg 1996.
(4) Valtinat, G., E. Piraprez and E. Greff: International Tightening Tests with High Strength
Bolts M20x100 of the Systems HR and HV. Heft 4 der Schriftenreihe Stahlbau und
Holzbau der Technischen Universität Hamburg-Harburg, Hamburg 1998.
ABSTRACT
INTRODUCTION
Block shear failure is characterized by a combination of rupture on the tension plane and
yielding on the shear plane(s) of a block of material. Although this mode of failure can occur
in either welded or bolted connections, it is more common in the latter because of the
reduced area that results from the bolt holes. This paper deals only with bolted connections.
Design rules in various codes and standards base block shear capacity on a combination of
yield and rupture strength of the net or gross areas in shear and tension on the failure
planes. North American design provisions are examined in order to assess the level of safety
provided by each design standard for gusset plates, angles, tees, and coped beams. The
Eurocode 3 design equations are also examined as one alternative to current North
American practice. Finally, a unified approach that provides a uniform level of safety for all of
these cases is proposed.
DESIGN PROVISIONS
CSA–S16–01
In the current Canadian design standard, CSA–S16–01 (1), the block shear capacity of
gusset plates and angles is taken as the lesser of the following two equations:
Pu = A nt Fu + 0.6 A gv Fy (1)
Pu = A nt Fu + 0.6 A nv Fu (2)
where Fy and Fu are the yield and the tensile strength of the material, respectively, A nt and
A nv are the net tension and shear areas, respectively, and A gt and A gv are the gross
tension and shear areas, respectively. Equations (1) and (3) are based on the observation
that rupture on the tension plane occurs before rupture on the shear planes. This is
supported by several test programs, including the University of Alberta test observations
presented below. Equations (2) and (4) are provided in order to limit the capacity of the shear
planes to the rupture strength of the net shear area.
AISC 1999
The block shear provisions in the AISC LRFD 1999 Specification (2) make use of two
equations that depend on the relative strength of the net tension and shear areas of the
connection:
when Fu A nt ≥ 0.6 Fu A nv , Pu = Fu A nt + 0.6 Fy A gv ≤ Fu A nt + 0.6 Fu A nv (5)
where all the variables are as defined above. The block shear capacity combines either
rupture on the net tension area, A nt , with yielding on the gross shear area, A gv , or yielding
on the gross tension area, A gt , with rupture on the net shear area, A nv . Both equations
have an upper bound corresponding to the combination of rupture on both the net tension
and net shear areas. Although equation (5) represents a plausible failure mode, the validity
of equation (6) is questionable (3). Although not yet adopted, the current proposal for the
2005 AISC Specification is to use the S16-01 equations (equations (1) through (4)), except
that the 0.5 coefficient for the tension plane is used for angles as well as for coped beams.
Eurocode
Eurocode 3 (4) combines shear yield acting over the gross shear area with rupture on the net
tension plane. Thus, the block shear capacity of a gusset plate or tension member can be
obtained as follows:
Pu = Fu A nt + Fy ( )
3 A gv (7)
(Eurocode uses the von Mises shear yield stress equal to Fy 3 , whereas the North
American standards simply take this as 0.60 Fy .)
Although equation (7) is not given specifically in Eurocode, it can be deduced from the
description of the mode of failure. A more elaborate procedure is presented for block shear
failure in beam webs, where a series of equations, when combined, gives the following
result:
1 1
Pu = w (L gt − k dh ) Fu + A gv F (8)
3 3 y
where w is the web thickness, L gt is the gross tension length, k is a tension area coefficient
(0.5 for one line of bolts or 2.5 for two), and dh is the bolt hole diameter.
There are a large number of gusset plate tests reported in the literature for which block shear
is the failure mode. Table 1 presents a summary of the results for 133 gusset plate tests from
eight different sources.
All gusset plate tests show that the ultimate load is reached when the tensile ductility of the
gusset plate material at the first (i.e., inner) transverse row of bolts is exhausted. This is true
even in cases where oversize holes were used and in cases where the connection was short
(i.e., little shear area available). Recent tests conducted at the University of Alberta (5) show
clearly that fracture on the net tension plane takes place before shear fracture on the shear
planes, as illustrated in Figure 1 on the left. The displacement of a block of material is seen
only when the test is continued until the parts separate (Figure 1, right), and this occurs well
past the ultimate load capacity of the connection. Figure 1 also illustrates that shear rupture
does not take place at the net area, as assumed in some design equations.
Tension fracture
Although many tests have been conducted on gusset plates, the range of geometries tested
in the past is fairly limited. Most tests have been conducted on gusset plates with two lines of
bolts with from two to four bolts in each line. Huns et al. (5) have extended the number of
bolts in a line to eight using finite element analysis. The 10 tests reported in Table 1 from
reference (5) include five results from a finite element model that has been validated against
existing test results.
The test-to-predicted ratios for all 133 gusset plates presented in Table 1 average 1.18 for
CSA–S16–01, 1.19 for AISC LRFD, and 1.10 for Eurocode 3. Thus, all design standards
provide a conservative estimate of the block shear capacity of gusset plates.
Experience and test results show that block shear is potentially a failure mode for angles and
tees, particularly when the connection is short. Tests of 41 such members (including nine
structural tees from Orbison et al. (16) connected at the stem that are considered to give the
same failure mode as angles) are shown in Table 1. Although Epstein (14) reported the
results of 144 tests, only the three specimens that failed in block shear with the tension face
perpendicular to the load direction are included in Table 1. Other test programs have also
Gusset Plates
1.20 1.22 1.19 1.05
Hardash and Bjorhovde (6) 28
(0.07) (0.08) (0.06) (0.06)
1.22 1.22 1.05 0.96
Rabinovitch and Cheng (7) 5
(0.06) (0.06) (0.05) (0.05)
1.18 1.18 1.06 0.96
Udagawa and Yamada (8) 73
(0.05) (0.05) (0.09) (0.06)
1.35 1.35 1.07 1.00
Nast et al. (9) 3
(0.02) (0.02) (0.01) (0.01)
1.21 1.21 1.05 0.95
Aalberg and Larsen (10) 8
(0.04) (0.04) (0.18) (0.10)
Swanson and Leon (11) 1 1.18 1.18 0.99 0.88
1.14 1.13 1.10 0.98
Huns et al. (5) 10
(0.17) (0.19) (0.14) (0.13)
1.13 1.15 1.16 1.02
Mullin (12) 5
(0.04) (0.05) (0.03) (0.05)
Angles and Tees
1.00 1.01 0.89 0.87
Barthel et al. (13) 13
(0.04) (0.05) (0.04) (0.04)
Epstein (14) [only mean reported] 3 0.98 0.98 1.00 1.02
0.97 0.96 0.93 0.92
Gross et al. (15) 13
(0.05) (0.05) (0.07) (0.06)
1.09 1.12 1.09 1.08
Orbison et al. (16) [includes 9 tees] 12
(0.05) (0.07) (0.05) (0.05)
Coped Beams
Birkemoe and Gilmor (17) 1 1.17 0.95 1.18 0.90
1.22 1.03 1.25 0.96
Yura et al. (18) 3
(0.16) (0.13) (0.16) (0.13)
1.02 0.70 1.13 1.00
Ricles and Yura (19) 7
(0.10) (0.08) (0.06) (0.12)
1.33 1.13 1.21 1.01
Aalberg and Larsen (10) 8
(0.12) (0.08) (0.23) (0.11)
1.20 1.07 1.21 1.02
Franchuk et al. (20, 21) 17
(0.10) (0.13) (0.09) (0.12)
Note : The number in parentheses is the standard deviation.
In contrast to the number of gusset plate test results available, there are relatively few tests
of coped beams. Of the 36 tests shown in Table 1, ten had connections with two lines of
bolts and 26 had a single line. A recent test program at the University of Alberta (20, 21)
provided nearly one-half of the available test results and expanded the number of test
parameters by including several variables related to connection geometry and loading
conditions. Seventeen full scale tests were conducted on coped beams and included the
effect of beam end rotation, gross shear area, bolt end and edge distances, number of bolt
lines and bolt rows, bolt diameter, section depth, connection depth, and double copes vs.
single cope. Analysis of the load vs. deformation results showed that none of these
parameters affected connection capacity significantly, apart from the associated changes in
tension and shear areas. It was concluded that only the tension and shear areas need be
considered in a design equation to reasonably represent the failure mode. The test results
indicated that rupture on the tension plane occurs before shear rupture, as shown in Figure 2
(left). It was also evident from the tests that shear rupture does not take place on the net
shear area. As shown in Figure 2 (right), shear failure takes place on a plane that intercepts
the holes very near to their edges.
Tension
fracture
The ratio of test ultimate load to the AISC predicted ultimate load is significantly non-
conservative for the tests by Ricles and Yura (19), which were all tests on coped beams with
two lines of bolts. The average AISC test-to-predicted ratio for the three test specimens of
Franchuk et al. (20, 21) with two lines of bolts is 0.90, compared to 1.10 for the test
specimens with one line. Both the Canadian and European standards seem to predict the
test results conservatively, although a large standard deviation is observed in some cases.
RELIABILITY ANALYSIS
Although the test-to-predicted ratios presented in Table 1 provide an indication of the ability
of each equation to predict block shear capacity, they do not by themselves quantify the level
of safety being provided. Therefore, the suitability of the various design equations was
assessed using reliability methods. Since the equations presented above do not contain a
performance factor, the following analysis derives the performance factors required to
achieve pre-selected levels of safety. Because the desired level of safety is a matter of
where β is the safety index, which is directly related to the probability of failure during the life
of the structure. It is desirable to have a higher safety index (lower probability of failure) for
connections than for structural members, such as beams, that tend to be more ductile. As
such, members are usually assigned a safety index of about 3.0, while connections have
typically been assigned a value of approximately 4.5 (22). The bias coefficient for resistance,
ρR , is given as:
ρR = ρM ρG ρP (10)
where ρM is the ratio of the mean measured to nominal material strength, ρG is the ratio of
mean measured to nominal connection geometric properties, and ρP is the professional
factor, or mean ratio of measured to predicted resistance, which reflects the ability of the
equation to predict the capacity. The coefficient of variation for the resistance, VR , is given
as:
There is an interdependence between φ and the load factors, shown by Fisher et al. (23).
This means that the use of a safety index other than 3.0 in equation (9) requires that a
modification factor be applied. As shown in (20, 24), equation (9) must be modified to:
Material Factor
The material factor reflects the difference between the nominal and measured material
strengths (yield or ultimate). The bias coefficient and coefficient of variation for the material
properties were obtained using data from Schmidt and Bartlett (25) for plates and rolled wide
flange sections. The material factors selected for gusset plates and angles and tees are
presented in Table 2. The values selected for coped beams are presented in Table 3.
Geometry Factor
The geometry factor accounts for the difference between the nominal and measured plate
thickness and hole layout dimensions. Since insufficient data exist to evaluate the geometry
factors for gusset plates, angles and tees, the bias coefficient and coefficient of variation for
the geometry factor proposed by Hardash and Bjorhovde (6) was used. The values selected
for coped beams were obtained from measurements on the 17 specimens tested by
Franchuk et al. (20, 21, 24) and data reported by Kennedy and Gad Aly (26). The values of
ρG and VG adopted are presented in Tables 2 and 3.
The professional factor (the ratio of the measured capacity, obtained either by laboratory
testing or from a validated finite element analysis, to the capacity predicted by the equation
using measured material properties and geometry) represents the ability of a model to
predict the block shear capacity. The mean and coefficient of variation of the test-to-
predicted ratios for the individual test programs presented in Table 1 have been consolidated
for gusset plates, angles and tees, and coped beams and are summarized in Tables 2 and 3.
Although the North American standards require the inclusion of a 2 mm allowance when
punched holes are used, it was not included in the professional factors presented in this
work. This is because in many cases it is not known how the holes were made. The effect of
this approximation was assessed for coped beams by Franchuk et al. (24) and found to be
relatively small. The performance factor, φ , was calculated for three levels of safety index,
namely, 3.5, 4.0, and 4.5.
Table 2. Reliability parameters for gusset plates and angles and tees.
Using the performance factor adopted by each standard (0.9 or 0.75 for the Canadian and
American standards, respectively) the current safety indices for each type of connection can
be determined. Use of the Canadian standard gives a safety index of 4.5 for gusset plates,
3.4 for angles and tees, and 4.1 and 3.0 for one- and two-line coped beam connections. Use
In order to address the shortcomings and inconsistencies of existing design equations, a new
equation for the design of gusset plates, angles, tees, and coped beams is proposed. It uses
laboratory observations of the failure mode as its basis. A single equation is considered
appropriate because the failure mode observed over a large variety of tests consists of
rupture on the tension plane after yielding has occurred on the shear plane, but prior to shear
rupture. Although most of the existing design equations predict the test capacity for angles
reasonably well, the capacity of gusset plates is generally under-predicted by a considerable
margin, and none of the equations adequately predict the block shear capacity of all coped
beams. It is also desirable to have a single equation rather than multiple equations, as is
currently used in the design standards. Although the stress on the shear plane is lower than
the rupture stress when rupture on the tension plane occurs, it is likely significantly greater
than the yield stress. Simply taking the mean of the yield and ultimate shear stresses
provides excellent correlation with test results. Another issue that must be taken into account
is that shear failure has been observed to occur on a plane that intercepts the holes very
near to their edges (see Figures 1 and 2). Therefore, the gross area is deemed to be more
appropriate for determining the strength of the shear components of the block.
The proposed unified equation combines effective stresses on both the net tension area and
the gross shear area:
⎛ Fy + Fu ⎞
Pu = R t Ant Fu + R v A gv ⎜ ⎟ (13)
⎜ 2 3 ⎟
⎝ ⎠
where R t and R v are the tension area and shear area stress correction factors, respectively,
as given in Table 4. These factors account for the non-uniform stress distributions that occur
in some cases on the tension and shear planes of the block shear failure surface. These
factors are semi-empirical in that they have been selected to optimize the consistency of the
safety indices, although they have support from finite element studies (20).
The proposed model is simplified over those of the existing North American and European
standards in that only one equation need be checked for a particular connection type.
Nevertheless, it not only provides an accurate prediction of connection capacity in an
average sense, with mean professional factors varying from 0.96 to 1.07 for all types of
connections considered, it also results in coefficients of variation that are among the lowest
of all procedures investigated. These values are tabulated in Tables 2 and 3. The excellent
correlation is attributed to the consistency of the structure of the proposed unified equation
with experimental observations.
Tables 2 and 3 give the performance factors, φ, required to achieve safety indices of 3.5, 4.0,
A comprehensive research program was carried out to evaluate the level of safety currently
being provided by design standards for determining the block shear capacity of gusset
plates, angles, tees, and coped beams. A reliability analysis indicates that the current North
American standards do not provide an adequate level of safety for all coped beam
connections, as compared to the traditional target safety index for connections of 4.5. Both
standards provide a sufficient level of safety for gusset plates and coped beams with one line
of bolts. For angles and tees and for coped beam connections with two lines of bolts, the
level of safety in the Canadian standard is unacceptable. The safety index provided by the
1999 AISC Specification for coped beams with two lines of bolts is far below the target value
of 4.5 for connections and even lies well below the target for ductile member failure of 3.0.
On the other hand, the equations currently proposed for the 2005 AISC Specification result in
safety indices very much higher than 4.5 for most common connections, while the value for
coped beams with two lines of bolts is considerably lower. The Eurocode equation provides
an inconsistent level of safety when considering various connection types.
The proposed unified design model (Equation 13) is recommended for the prediction of block
shear capacity. It is simpler than existing procedures and reflects the failure mode observed
consistently in a wide variety of tests. Moreover, with a single performance factor the model
provides a consistent and adequate level of safety for all types of connections investigated in
this study.
ACKNOWLEDGEMENT
The research conducted at the University of Alberta was funded by the the Steel Structures
Education Foundation, the Natural Sciences and Engineering Research Council of Canada,
and the C.W. Carry Chair in Steel Structures Research.
REFERENCES
ABSTRACT
In most of the design codes, very few recommendations exist to ensure the
ductile behaviour of connections with non preloaded bolts under shear loading.
The engineers are therefore used to consider steel as a “good” material with
sufficient ductility, so allowing generally to reach at ultimate state a uniform
distribution of the internal forces consecutive to a plastic redistribution amongst
the bolt rows. But in reality this requires a sufficient deformation capacity from
the constitutive plates in bearing and connectors in shear. In the present paper,
all these aspects are discussed and ductility requirements allowing to reach
uniform distributions of internal forces in shear connections are proposed.
These investigations have to be seen as a preliminary step on the way to the
proposal of more general ductility criteria covering other important aspects as
high strength steels, tolerances of fabrication, oversized and slotted holes, …
When a bolted connection is submitted to shear loading (Figure 1), a distribution of the
external applied force takes place amongst the connectors. Locally, forces are transferred
from one plate to another (to the others) by plate-to-bolt contact, so leading to bearing and
tension forces in the plates as well as shear forces in the connectors.
F
F
F F
F/2
F
F/2
Component approach
For the modelling and calculation of the connection shown on Figure 4.a, the component
method (1), prescribed by Eurocode 3 - Part 1.8 (2), can be efficiently applied. The
components to be considered for shear connections are presented in Table 1 (bolts in shear,
plates in bearing and plates in tension). These ones are subjected to an internal force R and
elongates accordingly by δ; their characteristic stiffness (Sc) and resistance (Rc) properties
(Figure 3.a), given in Part 1.8 of Eurocode 3, are reported in Table 1. On the contrary, no
information is provided as far as the available deformation capacity (δc) is concerned.
Plate in
Spl = EA/pb Rpl=min(A fy; 0,9 Anet fu)
tension
Bolt in
Sb = 8 d² fub/dM16 Rb = αvfub Ab
shear
Plate in
Sp = 12 kb kt d fu Rp = k1αb fu d t
bearing
Equivalent Seq =
Req = min(Rb, Rp1, Rp2)
component (Sb-1 + Sp1-1 + Sp2-1)-1
R R
Rc (p, pl or b)
Req
Sc (p, pl or b)
δ Seq
δc (p, pl or b) δ
δav
a – Individual component b- “Equivalent bolt zone” component
In Figure 4.b, a mechanical model is built where all the constitutive components are
represented by extensional springs. When the forces pass through the bolts from the lower
plate to the upper plate, the bolt holes in the lower plate deform by bearing, the bolts are
sheared and finally bearing ovalise the holes in the upper plate. This transfer of forces is
modelled by three springs in series in Figure 4.b. These three springs may be merged
Failure modes
The following possible failure modes may be observed: (i) plate in tension (in the net area or
gross area) or (ii) failure of an equivalent bolt zone in shear. The second one is linked either
to the bearing resistance of one of the connected plates, or to the resistance of a bolt in
shear. In the present study, the properties of the connected plates are considered to be such
that the failure by “plate in tension” is not at all relevant.
To illustrate how internal forces distribute amongst the components when the connection is
progressively loaded, a simple lap connection with 5 bolts is considered (Table 2). Again an
elasto-plastic behaviour of the components is assumed (Figure 3.b).
Step 1: For a low level of load, all the components are in their elastic range of behaviour and
the two exterior components are more loaded than the interior ones.
Step 2: For a higher applied force, the exterior components reach their strength limit Req.
They can not support any further load. The other components are still elastic.
Step 3: When the applied load increases, plastic redistributions of forces take place as the
exterior components cannot transfer further load; forces are reported on the other
components C1 and C2 until the C2 components reach also Req. This requires plastic
deformation capacity from the C1 components.
Step 4: Plastic redistribution may progress as long as the deformation capacity is available in
the C1 and C2 components. Finally, the maximum resistance of C3 is reached. A complete
plastic redistribution is then obtained and the connection resistance equals 5Req. To reach
this ultimate load, the exterior components have to exhibit a sufficient deformation δreq.
C1 C2 C3 C2 C1
∆ δ δ δ
δav δav δav
R R R
F F=Fe
Fpl Req Req Req Req Req
Fe
C1 C2 C3 C2 C1
δ δ δ
∆ δav δav δav
F Fe<F<Fpl R R R
Fpl
Req Req Req Req Req
Fe
C1 C2 C3 C2 C1
∆ δ δ δ
δav δav δav
R R R
F F=Fpl
Fpl Req Req Req Req
Fe
C1 C2 C3 C2 C1
δ δ δ
∆ δreq δav δav δav
Table 2. Response of a “5-bolts” shear connection and of its constitutive components under increasing applied loading.
The here-above presented model has been run numerically, through the use of the Liège
home-made non-linear FEM software FINELG, for many different situations. Some
experimental tests on shear connections performed by Valtinat in Hamburg (3) have also
been successfully simulated.
If the components exhibit an infinite deformation capacity, a total plastic redistribution of the
internal forces can take place and each equivalent component can reach its maximum strength
Req. In such a case, the shear resistance of the connection is defined as equal to Fpl = n Rc.
If, on the contrary, the “equivalent bolt” components do not exhibit good deformation
capacity, the resistance of the connection will have to be limited to a lower value, which is the
one transferred by the connection just when the maximum available deformation capacity δav
is reached in the most heavily deformed component.
In other words, a shear connection with n bolts (or bolt-rows) only reaches its full plastic
capacity nReq if the “equivalent bolt” components possess a deformation capacity δav higher
than the required one, δreq:
The elastic distribution of internal forces amongst the components depends on the relative
stiffness of the steel plate Spl and that of the equivalent bolt components Seq (see Figure 5).
δ0 δ1 δ0
δ0 = F1/Seq (2.a)
δ1 = (F-2F1)/Seq (2.b)
By expressing the compatibility of displacements between the plates and the bolt zones, the
distribution of shear forces amongst the components in the elastic range may be derived
(figure 5):
⎛ 1 p ⎞ 1
⎜ + b⎟
⎜S EA ⎟ S
F1 = F ⎝ ⎠ (3.a and 3.b)
eq eq
F − 2 F1 = F
⎛ 1 p ⎞ ⎛ ⎞
⎜3 +2 b ⎟ ⎜ 3 1 + 2 pb ⎟
⎜ S EA ⎠⎟ ⎜ S EA ⎟⎠
⎝ eq ⎝ eq
If the plates in tension are stiff (EA → ∞), the distribution of forces tends to be uniform; for a
fully flexible plate (EA → 0) , only the external components transfer the shear force.
When the external components reach their maximum strength (F1 = Req), the maximum
elastic resistance of the connection is obtained. The corresponding applied load is equal to:
⎛ 1 p ⎞
⎜3 +2 b ⎟
⎜ S EA ⎟⎠
Fel = Req ⎝
eq
(4)
⎛ 1 p ⎞
⎜ + b⎟
⎜S ⎟
⎝ eq EA ⎠
For this level of shear load, the elongations in the external and internal equivalent bolt
components are respectively equal to:
The maximum strength of the internal bolt is not yet exhausted and the additional force that it
can still transfer is:
1
S eq
Req (1 − (6)
⎛ 1 p ⎞
⎜ + b⎟
⎜S ⎟
⎝ eq EA ⎠
In order to reach a full plastic distribution of internal forces, to which would be associated the
following connection resistance:
VRd = 3 Req (7)
the internal bolt component has therefore to undergo an additional elongation:
And finally, through the equations of displacement compatibility, the total elongation of the
external bolt components required to enable a full plastic redistribution to take place may be
expressed as follows:
Req pb
δ 0,req = + Req (9)
S eq EA
The procedure followed in the previous section may be generalised to connections with n
bolts or bolt-rows. Table 3 gives values of the required deformation capacity up to n=8.
Table 3. Maximum required deformation capacity from the equivalent bolt components.
Req
2 S eq
⎛1 p ⎞
Req ⎜ + b⎟
3 ⎜S ⎟
⎝ eq EA ⎠
⎛ 1 p ⎞
4 Req ⎜ +2 b ⎟
⎜S EA ⎟⎠
⎝ eq
⎛ 1 p ⎞
5 Req ⎜ +4 b ⎟
⎜S EA ⎟⎠
⎝ eq
⎛ 1 p ⎞
6 Req ⎜ +6 b ⎟
⎜S EA ⎟⎠
⎝ eq
⎛ 1 p ⎞
7 Req ⎜ +9 b ⎟
⎜S EA ⎟⎠
⎝ eq
⎛ 1 p ⎞
8 Req ⎜ + 12 b ⎟
⎜S EA ⎟⎠
⎝ eq
In a more general way, the required deformation capacity may be evaluated as follows:
⎛ 1 p ⎞
δ req = Req ⎜⎜ +ρ b ⎟ (10.a)
⎝ S eq EA ⎟⎠
n/2
with: ρ = ∑ (n − 2i) for an even value of n (10.b)
i =1
( n −1) / 2
ρ= ∑ (n − 2i)
i =1
for an odd value of n (10.c)
From these values, the available deformation capacity of the “equivalent bolt zone
component” may be derived as follows:
If the shear resistance of the bolt is significantly higher than that of the plate(s) in bearing,
the whole plastic deformation capacity results from that of the plate(s) in bearing and:
If the shear resistance of the bolt is significantly lower than that of the plate(s) in bearing,
the whole plastic deformation capacity results from that of the bolt in shear and :
In intermediate cases, the plastic deformation capacity results from both the bolt and
plate(s) deformations; as a rough approximation, the following formulae are suggested:
The following worked example (Figure 6) illustrates the content of the paper and allows to
draw some first conclusions about the capacity of shear connections to undergo plastic
redistributions of internal forces. It is also the occasion to identify further research needs.
The numbers of bolt rows and of bolt lines are varied. All the other geometrical and
mechanical parameters are fixed. The variation of the required and available deformation
capacities with the number of bolt rows and files is reported in Figure 7.
p2 = 40mm
b = 216mm n2=1 → 5
eb = 35mm pb = 80mm
n1 = 6
practical criterion
The values reported in Figure 7 are calculated for “ideal” connections, i.e. with no
geometrical lack-of-fit (exact position of the holes, same hole diameters for all the bolts,
equal hole and bolt diameters,...). To reach the full plastic redistribution and consequently the
full resistance of the connection, the curve δav has to be above the curve δreq.
It may be seen that δreq increases with the number of bolt rows (see also Formula 10). On the
contrary δav is not affected by this parameter as it only depends on the local properties of the
equivalent component. The required deformation capacity increases also with the number of
bolt lines n2. Indeed, this last parameter influences the values of Seq and Req in Formula 10.
CONCLUSIONS
In the present paper, a methodology is introduced to check whether sufficient ductility for full
plastic redistribution of internal forces is available in shear connections with non preloaded
bolts. The development and preliminary validation of a behavioural model for shear bolted
connections is presented. Design requirements are expressed and formulae to evaluate the
available deformation capacity in the vicinity of the bolts are proposed for connections with
normal bolt holes. Finally, the justification of the apparently safe character of the EC3 ductility
criteria (and others) has been demonstrated. However it has to be mentioned that the
influence of fabrication tolerances as well as the effect of the gap between the bolt shanks
and the hole edges may significantly influence the resistance of the “ideal” studied
connections. In further studies, the effects of fabrication tolerances will be implemented and
the calculation model presented in the paper will be extended to slotted and oversize holes.
Finally, it is expected that these works will lead to new improved ductility requirements to be
included in a later stage in Eurocode 3.
REFERENCES
(1) Jaspart J.P., “Design of structural joints in building frames”, Progress in Structural
Engineering and Materials, Vol. 4 N° 1, January-February 2002, pp. 18-34.
(2) Eurocode 3 “Design of steel structures” Part 1.8 “Design of joints”, prEN 1993-1-8,
CEN, Brussels, April 2002.
(3) Valtinat G. and Wilhem M., “Schrauben-verbindungen im Stahlbau“, Last-
Verschiebungs-Diagramme, Hamburg University, Germany, 1988 – doc. ECCS-TC10-
89-301.
(4) Karmalin V.V. and Pavlov A.B., “Load capacity and deformability of bearing and friction-
bearing connections”, Proceedings of the International Colloquium on Bolted and
Special Structural Connections, USSR, Moscow, May 15-20, 1989, pp. 52-60.
(5) Jaspart J.P., “Etude de la semi-rigidité des noeuds pouter-colonne et son influence sur
la résistance et la stabilité des ossatures en acier”, Ph.D. Thesis, M&S Department,
Liège University, Belgium, 1991.
ABSTRACT
Current practice for computing block shear capacity of steel angles with bolted
connections is to assume the stress distribution along the “block” gross shear
plane to be uniform, which may not always be true. Based on previously
published experimental data and nonlinear finite element analysis, it is
suggested here that area along “block” gross shear plane should be suitably
modified taking into account the stagger of bolts (keeping the “block” net
tension area unaltered). This improved approach leads to adequate prediction
of block shear capacity of single as well as double angles, with staggered as
well as non-staggered bolts.
INTRODUCTION
Steel tension members are probably the most common and efficient members in structural
applications. It is relatively easy to fabricate and erect structures, or a part thereof, comprising
of angles because of the basic simplicity of its cross-section. As such, steel angles are
frequently used as tension members in a majority of applications, in spite of their complex
behaviour.
Often, it is not practicable to connect both the legs of the angle with the gusset plate. Angles
are generally used as single or as a pair, symmetrically placed about a gusset plate that
passes between them. The connection between the angle and gusset may be made by
welding or by bolting. Bolting requires drilling or punching of holes, which results in reduced
area at the critical section. The bolts may be placed along one or more gauge lines, and may
or may not be staggered along the length of the connection. When staggered, well known
s2/4g rule is applied to compute the least net area.
Angles with bolted connections normally observe net section failure for relatively longer
connections. However, for relatively shorter connections, the mode of failure may be block
shear, wherein a ‘block’ of the connected element may separate from the remainder of the
element (Figure 1). The failure path in this case lies entirely within the connected leg;
outstanding leg is stressed very little. On account of this, the load carrying capacity is
significantly reduced when the mode of failure is block shear. The angle section must be
checked for block shear failure also, even if the end, edge and pitch distances are within the
specified limits.
Block shear mode of failure has drawn the attention of researchers from most parts of the
world. As a result of the ongoing research, the provisions for block shear have been modified
in almost every revision of various specifications. However, even the current provisions in
various specifications are not the perfect reflection of experimental results, as envisaged in
the recent research in this area. Thus, there is a scope for improvement in the way the block
shear strength of angles under tension is predicted.
In the present study, block shear model based on the failure mechanism observed during the
experiments is evaluated on the basis of relevant test results available in the literature.
These test results include single as well as double angles, having bolt holes in one or more
rows, and with staggered and non-staggered holes. An improvement in this approach of
computing block shear capacity of steel angles is then proposed.
The failure mechanism observed during the experiments consists of rupturing of the block
net tension plane, accompanied by inelastic deformation of the block gross shear plane
(Figure 2).
The yield strength in shear is taken as 1 / √3 times the yield strength in tension, according to
Von Mises yield criteria, so that, fys = fy / √3. Kulak and Grondin (4), (5) are of the view that
this equation be used for angles, on the rationale that it is based on the behaviour observed
during the experiments on angles. In their evaluation, they have included only the angles in
which bolts are not staggered and have shown that rules in AISC specification are in better
agreement with the angle tests than Equation (1). In their discussion, Gupta and Gupta (6)
included few angles that had staggered holes.
In the following, it is shown that this simple equation adequately predicts the block shear
capacity of angles with non-staggered bolts and with staggered bolts as well, after
improvement in the way the block gross shear area is computed in certain cases.
One of the simplest and straight forward ways to evaluate the prediction of block shear
strength by Equation (1) would be to work out the ratio of experimental failure load to the
predicted ultimate load, termed as professional factor, for previously published test results.
Professional factor less than 1.0 indicates non-conservative load prediction, while factor
exceeding 1.0 is produced by conservative prediction.
In Table 1a, the specimen numbers used are same as reported in the original research. In
this table, only those specimens from previously published literature have been included that
showed block shear failure. All these specimens follow all provisions regarding minimum
pitch, edge and end distances. The specimens which violate one or more of these provisions
are not included here. Angles composed of high strength steel are also not included here.
Specimens of Epstein (7) had double angles with two rows of bolts. In a majority of these
specimens, the bolts are staggered. Specimens of Gross et. al. (8), Orbison et. al. (9) and
Gupta and Gupta (10) are single angles with one row of bolts.
The professional factors are worked out in Table 1a, using the actual material properties of
each of these specimens. Average and standard deviation of professional factors have also
been worked out. The average professional factor of 1.04 for all tests indicates adequate
representation of block shear strength by Equation (1). The average professional factor as
per AISC specification (3) is 1.02 while it is as high as 1.26 for Eurocode3 (11).
In Table 1a, there are certain specimens in which the bolts are not placed along the standard
gauges. It is observed that the professional factors for few such specimens, as per Equation
(1) and AISC specification (3) are on the non-conservative side. There is an increase in the
in-plane eccentricity as the bolt line shifts towards the toe of the angle section, and it is
reconfirmed here that the block shear capacity decreases with an increase in the in-plane
eccentricity. More non-conservative predictions may be expected, if there is further increase
in the in-plane eccentricity. As such, angles in which the bolts are not placed along standard
gauges must be given due consideration in design and the capacity should be judiciously
reduced, depending upon the eccentricity.
This evaluation points that the block shear equation (Equation 1) predicts the capacity of
angles more conservatively than the AISC specification (3), when the mode of failure is block
shear. AISC approach involves comparatively more computations and in many cases, the
governing mechanism does not agree with the experimentally observed mechanism.
Eurocode3 (11) provisions predict this capacity quite conservatively.
There are other specimens tested by Esptein (7), in which the mode of failure was partially
block shear. A majority of such specimens had bolt holes staggered in two rows such that the
end distance along the row of bolts towards the heel was more than the end distance along
the row of bolts towards the toe of the angle (Figure 3a).
e p p p e p p p
Agv Agv
Agv = ( p + p + p + e ) * t Agv = ( p + p + e ) * t
(a) (b)
It is observed from results reported in Epstein (7) that, in many cases, the experimental
failure load of specimens with bolt hole configurations as in Figure 3a and Figure 3b, is about
the same. Whereas, the predicted block shear capacity, as per Equation (1), is much more in
specimens of Figure 3a than that of specimens of Figure 3b, because of increased block
gross shear plane area. The bolt hole configuration of specimens considered in Table 1b is
Hardash and Bjorhovde (12), in their study on gusset plates, observed that the effect of
varying the connection length is important, and proposed that it must be incorporated in a
rational and complete gusset plate model. They have suggested a linear interpolation
function that is applied on shear stress acting on gross connection length area. A deep
insight into the bolt hole configurations of specimens of Table 1b reveals that the stress
distribution along the block gross shear plane may not be uniform. The stress is likely to be
more towards the lead bolt hole and the stress near the end of block gross shear plane may
be quite less. This non-uniform stress distribution is illustrated in Figure 5.
A nonlinear finite element model was prepared to verify this concept and the resulting shear
stress distribution along block gross shear plane is shown in Figure 6. This finite element
analysis is largely based on the methodology adopted by Kulak and Wu (13) for modelling of
angles under tension.
Outstanding leg
Minimum
Gusset
Plate Zero
Maximum
Connected leg
Figure 6. Non uniform stress distribution along block gross shear plane.
e p p p e p p p
Sample Calculations:
Agv = ( p + p + p + e ) * t
Agv A*gv
A*gv= ( p + p + p ) * t
The professional factors with the AISC approach, Eurocode3 approach and improved
approach suggested herein (Equation 2) are also presented in Table 1b. The average
professional factor of these specimens as per AISC approach is only 0.86 and indicates
unconservative prediction of block shear capacity. The Eurocode3 provisions for such
specimens are not as conservative as for specimens of Table 1a. It is observed that the
professional factors as per Equation (2) represent the block shear capacity adequately in
most of the cases with a mean value of 1.06. Thus, there is a considerable improvement in
values of professional factors when the concept of effective block gross shear area is used in
the computations.
It is to be noted that the data available on block shear failure of angles with staggered holes
is from a single source only. More such tests may be conducted to further validate the
CONCLUSIONS
This study examined the block shear capacity of steel single as well as double angles, for
bolt holes in one or more rows, and with staggered and non-staggered holes. Only those
specimens that follow all provisions regarding minimum pitch, edge and end distances were
included in this study. Angles composed of high strength steel are not included in this study.
An improved approach, to compute the the block gross shear area of specimens that had
bolt holes staggered such that the end distance along the row of bolts towards the heel is
relatively more, is suggested. This area, A*gv, as per this improved approach is termed here
as effective block gross shear area and is somewhat less than the block gross shear area,
Agv, as per current practice. There is a considerable improvement in values of professional
factors when the concept of effective block gross shear area is used in the computations.
The following simple equation was found to give adequate results for single as well as double
angles, for bolt holes in one or more rows, and with staggered and non-staggered holes.
The data available on block shear failure of angles with staggered holes is from a single
source only. More such tests may be conducted to further validate the findings of this study.
The proposed approach is specific to angles, and may likely be incorporated in the AISC
specification (2) on angles.
NOTATIONS
REFERENCES
(1) Gupta Mohan, “Study of Effective Areas of Steel Angles under Tension”, Ph. D.
Thesis, Visvesvaraya National Institute of Technology, Nagpur, India, May 2003.
(2) AISC LRFD:2000, “Load & Resistance Factor Design Specification for Single Angle
Members”, Second Edition, American Institute of Steel Construction, Chicago, Illinois.
(3) AISC LRFD:1999, “Load & Resistance Factor Design Specification for Structural
Steel Buildings”, Third Edition, American Institute of Steel Construction, Chicago,
Illinois.
ABSTRACT
Flange tip connections have to transfer primarily shear forces from a
secondary beam to a main beam through the main beam flanges since
stiffeners are omitted for economic reasons. In the flanges of the main beam,
longitudinal bending stresses occur due to beam behaviour. But
simultaneously also transverse bending stresses occur in these flanges at the
connection resulting in a complex stress state in the main beam flanges. The
work reported in the paper includes experimental, analytical and numerical
research on flange tip connections and is aiming at obtaining design rules for
these connections.
INTRODUCTION
Flange tip connections are a special type of beam-to-beam connections, for example in a
floor or roof structure, where a secondary beam is connected to a main beam. Usually these
connections have to transfer primarily shear forces from the secondary beam to the main
beam (figure 1). There are several alternatives in case the top of the upper flanges are to be
at identical levels. One of the economically attractive ones is by adding a connection plate to
the side of the flanges of the main beam and a (short) header plate to the end of the
secondary beam thus connecting both beams by bolts through the connection plate and
header plate. From point of view of force transfer it would be logical to add a stiffener
between flanges, web and connection plate of the main beam. However, this stiffener is
preferably left out for economic reasons and so the secondary beam is connected to the
flange tips of the main beam only, hence the name ‘flange tip connection’.
In the flanges of the main beam, longitudinal bending stresses occur due to beam behaviour.
But simultaneously also transverse bending stresses occur in these flanges at the connection
due to the presence of the secondary beam. Also shear stresses are present resulting in a
complex stress state in the connection and complex connection behaviour. In general there
is interaction between connection and beam behaviour. Several research projects have been
carried out at Eindhoven University of Technology (1,2,3) to analyse flange tip connections.
This is done by simplifying the problem to first focussing on the flange tip connection itself
and then on the interaction between connection and beam behaviour. The work reported in
the paper includes analytical, numerical and experimental research on flange tip connections
and is aiming at design rules for these connections. The work is still in progress so only
preliminary results are presented here.
Secondary beam
M line
APPROACH
A structural floor system or even a sub-system consisting of one main and two secondary
beams is relatively complex to investigate experimentally. Therefore, the secondary beams
are omitted and the load is applied to the connection plates connected to the flange tips of
the main beam (figure 2). Also, the problem is further simplified by testing as a first step the
connection only. This is done by supporting the main beam along its length under the web at
the connection. This situation is called a supported connection and only connection
behaviour is investigated. As a second step, the main beam is not supported under the web
at the connection. The main beam can be seen now as a simply supported beam in three-
point bending with the load applied at the connection plates. This situation is called an
unsupported connection and now the interaction between beam and connection behaviour is
investigated.
Supported connections of three types with an IPE330 as main beam are investigated as
shown in figure 2 (2). In type 1 only the top flange is investigated. In type 2 a thick connection
plate is used forcing yielding into the flanges and distributing the load equally over top and
t = 12 mm
t = 40 t=6
Typical yield line failure mechanisms are shown in figure 5 (2,3). Using first order plastic
theory, the following failure loads per flange tip can be derived for the mechanisms 1, 2 and 3
of the figures 5a, 5b and 5c respectively:
Mp,fl = ¼ t2 fy (4)
The formulas (1-3) with β = 1 hold for type 1 connections. The three failure loads of these
mechanisms are very close, but formula (1) is the governing one. In case of type 2
connections another yield line forms along the connection plate and then formulas (1-3) with
β = 2 are valid. For type 3 an additional yield line develops in the connection plate resulting in
a modification of the formulas (1-3). A value of (k/d Mp,hp) has to be added to the failure loads
of formulas (1-3) with β = 1 ( fy = 308 N/mm2 for the connection plate material). The failure
loads calculated this way are listed in table 1. For type 3 the calculated theoretical failure
load overestimates the experimental failure load substantially. This is due to the fact that the
bottom flange does not contribute to the failure load and the connection plate fails instead of
the bottom flange. Therefore, the failure load was estimated on the basis of connection plate
failure (figure 5d) with mechanism 1 in the top flange resulting in the failure load indicated
between brackets in table 1.
a) mechanism 1 b) mechanism 2
c) mechanism 3 d)
Tests 1.1 and 2.1 have numerically been simulated (3) using the finite element method
(FEM) programme Ansys. The finite element model of test 1.1 is shown in figure 6. Figure 6a
shows the tested specimen, figure 6b shows the shell elements used in the section and
figure 6c gives an impression of the mesh used. Here also the theoretical failure mechanisms
1 and 2 are shown. The material model used is represented in figure 7. Hardening is
neglected. The calculations carried out include material and geometrical non-linearity. It
turned out that the effect of geometrical non-linearity is substantial.
t=40
The finite element model of test 2.1 is shown in figure 10 and is comparable with that of test
1.1 (figure 6). Again the material model of figure 7 is used. The calculations carried out
include material and geometrical non-linearity. However, in this case it turned out that the
effects of geometrical non-linearity can be neglected. The load-displacement diagram (3) is
not shown in this paper. The failure load calculated with FEM is Fnum = 93 kN and is also
presented in table 1. Equivalent stresses calculated at failure are shown in figure 11. Large
yield areas can be observed in top and bottom flange. Figure 12 shows qualitatively the
numerically determined displacements at failure. The numerically found failure mechanism
corresponds to theoretical mechanism 3 with lines of equal displacement not ending
perpendicular to the flange tip but under an angle of approximately 70o.
Figure 11. Equivalent stresses at failure (Fnum = 93 kN) for test 2.1.
In figure 13 measured strains are compared with numerically determined strains and they
correspond reasonably well, with measured strains extending beyond calculated strains.
The numerically determined failure load underestimates the measured failure load
considerably and therefore the numerical model was refined according to figure 14. In figure
14a the original model is shown. The value d is used in the theoretical models. In the finite
element model, the weld thickness is neglected and the connection plate is shifted compared
to the test. The refined model is shown in figure 14b. Now the welds are represented by a
thicker shell element and the connection plate is in the right position. The numerically
determined failure load is now Fnum = 109 kN, listed between brackets in table 1.
Load F (kN)
FEM
Test
Strain in %
Figure 13. Comparison of equivalent strain at rosette R5 for test 2.1.
Theory
Experiment
Experiment
FEM refined
FEM original
a) b)
In case the beam is very short, the behaviour will be similar to a supported connection and
connection failure will govern behaviour. For this case the theoretical failure load per flange
tip Fthe,1 can be calculated using formula (1) with β = 2.
In case the beam is very long, the behaviour will be governed completely by beam
behaviour. Using first order plastic theory assuming a plastic hinge at mid span, the failure
load of the beam can be calculated as follows:
M = Mp,b = ¼ Pp L (6)
Pp = 4 Mp,b / L (7)
Pp = 4 Fthe.b (8)
Combining formulas (7) and (8) results in the theoretical beam failure load per flange tip:
Both Fthe,1 and Fthe.b are upper bounds of the theoretical failure load that includes beam
connection interaction. These values are also listed in table 2.
Numerical research on unsupported connections has not yet been carried out but is planned.
DISCUSSION
The failure loads for supported flange tip connections are summarised in table 1. In all cases
the lowest theoretical failure load is the one for mechanism 1 as could be expected
considering formulas (1-3). Therefore, the theoretical failure load according to mechanism 1
(formula (1)) will be used for comparisons.
Table 1. Failure loads per flange tip for supported flange tip connections.
Type Test fy [N/mm2] Fexp [kN] Fthe,1 [kN] Fthe,2 [kN] Fthe,3 [kN] Fnum [kN]
1 1.1(2) 340 65 72 82 76 70
2.1(2) 340 118 94 103 96 93 (109)
2.2(1) 340 93 97 108 101 n.a.
2
2.3 289 121 82 90 85 n.a.
2.4 289 101*) 82 90 85 n.a.
3 3.1(2) 340 61**) 78 (45) 88 82 n.a.
*) weld failure before flange failure **) connection plate failure
For test 1.1 the theoretical failure load is 11% unsafe with respect to the experimental failure
load. The numerical failure load is close to the theoretical failure load but also on the unsafe
side (8%) when compared to the experimental failure load.
For test 2.2 also the theoretical failure load is on the unsafe side but by just 4%.
For test 2.1 the theoretical failure load is on the safe side (20%). The numerical failure load
obtained with the original finite element model is however very close to the theoretical failure
load. Refining the finite element model taking the weld geometry into account (figure 14),
improves the numerical failure load, however, to be still on the safe side (8%).
For the tests 2.3 and 2.4 the results seem to be comparable with those of test 2.1, however,
the theoretical failure loads are even more conservative. Note that premature weld failure
governed the failure load of test 2.4.
For test 3.1 connection plate failure determined the failure load and this explains why the
theoretical failure load obtained with formula (1) is too high. Adopting a different failure
mechanism accounting for connection plate failure gives a far too low theoretical failure load.
It can be concluded that further research is necessary to improve the model (failure
mechanism) to obtain the theoretical connection failure load, making use of more tests and
finite element calculations and taking the weld geometry into account.
The failure loads for unsupported flange tip connections are summarised in table 2.
Test 4.1 is a relatively long specimen and it can be seen that beam behaviour governs the
failure mode. The theoretical beam failure load is very close to the experimental failure load, 5%
on the unsafe side. The theoretical connection failure load is relatively high so that the
connection is very strong and does not influence the behaviour.
Tests 4.2 and 4.3 are similar. Here, the theoretical connection failure load and the theoretical
beam failure load are relatively close so that both were expected to influence the behaviour in
this test, leading to an even lower experimental failure load than the theoretical upper bounds.
This is not confirmed by the tests since the experimental failure load exceeds the theoretical
upper bounds (22% and 8%). However, bearing in mind the fact that the theoretical connection
Table 2. Failure loads per flange tip for unsupported flange tip connections.
Type Test fy [N/mm2] L [m] Fexp [kN] Fthe,1 [kN] Fthe,b [kN]
4.1(1) 340 6.65 39 97 41
4 4.2 289 2.40 106 83 97
4.3 289 2.40 105 83 97
When two proper upper bounds are obtained after modifying the models for connection
failure and beam failure, the next step would be to develop a model predicting the combined
connection and beam interaction failure load, taking the complex stress state in the main
beam flanges at the connection into account.
Flange tip connections between main and secondary beams have been investigated
experimentally, theoretically and numerically to try to obtain design rules. A complex stress
state exists in the flanges of the main beam due to overall beam bending and local
connection behaviour. Simplifications have been made omitting the secondary beam and
investigating the connection only by supporting the main beam over its full length. Several
theoretical failure mechanisms have been considered and compared with experimental and
finite element results. Further research is necessary to improve the connection failure model,
making use of more tests and finite element calculations and taking the weld geometry into
account. Also, the interaction between beam and connection behaviour was investigated for
beams in three-point bending. So far, this interaction has not been found in the tests carried
out. It seems that this interaction only is relevant for relatively short beams. Further research
on short beams is therefore planned. A simple beam failure model was developed which
requires improvement. Further research is necessary to find a model predicting the combined
connection and beam interaction failure load.
NOTATION
a length, a=0.378d
b length, b=1.512d
bf width of the flange
d distance between root radius and flange tip
E Young's modulus
fy yield stress
F load on flange tip
Fexp experimental failure load per flange tip
Fnum numerical failure load per flange tip
Fthe,1 theoretical failure load per flange tip for mechanism 1
Fthe,2 theoretical failure load per flange tip for mechanism 2
Fthe,3 theoretical failure load per flange tip for mechanism 3
Fthe,b theoretical beam failure load per flange tip
k width of connection plate
REFERENCES
ABSTRACT
INTRODUCTION
The 1994 Northridge and 1955 Kobe Earthquakes revealed that beam-to-column
connections in steel moment resisting frames were susceptible to brittle fracture. In
Northridge brittle fractures most frequently occurred at the beam bottom flange groove
welds with cracks initiating at roots of the welds. Unfused regions between steel
backup bars and column flanges created sharp notches at the weld roots. In Kobe
cracks frequently initiated at toes of the beam copes. The occurrences of these cracks
were found to be reduced by improving profiles of the beam copes. Cracks also
frequently emerged from notch roots formed by the steel weld tabs and beam flanges at
the terminations of CJP groove welded joints. The improvements of the weld tab regions
were found to be made by replacing the steel tabs by flux tabs or removing the steel tabs
after welding and then grinding smooth the ends of the welded joints. However, the
improvement of the weld tabs only is insufficient to avoid a brittle fracture initiating from
the terminations of the CJP welds at the beam ends. Further improvements in welded
joints are required by some means.
The shop-welded connections proposed here utilize PJP groove welds with reinforcing
fillet welds for the welded joints at the beam flange ends. The beam webs are
fillet-welded to the column flanges. Backing bars and beam copes are unnecessary for
PJP groove welded joints, however, inevitably contain discontinuities that may act as
sharp notches. Thus, the design strength of PJP groove welds recommended in the
Eurocode 3 (3) and AISC LRFD Specification (4) are generally equivalent to the design
strength of fillet welds. Eurocode 3 allows that a tee-butt joints, consisting of a pair of
PJP groove welds reinforced by superimposed fillet welds, is calculated as a CJP groove
weld, if the total nominal throat thickness is not less than the thickness t of the part
forming the stem of the tee joint, although the stringent limitation that the unwelded gap is
not more than (t/5) or 3 mm is imposed. The AIJ LSD Recommendations (5), in contrast,
specify the design strength of PJP welds equivalent to that of CJP welds.
The design strength according to AIJ is based on extensive test results for plate-to-plate
welded joints. However, details of the joints and loading conditions are different between
the beam-to-column joints and plate-to-plate joints. Especially, the former joints are
subjected to cyclic loading. Further experimental investigations are required to establish
a reliable design methodology for PJP welded beam-to-column joints. This paper
presents test results for two connections, previously reported, and for two additional
connections. Based on these test results, the behavior and design of PJP welded
beam-to-column connections will be discussed hereafter.
Full-size beam-to-column connections with PJP groove welded joints were tested to
failure under cyclic loads. Four specimens, two with wide-flange section columns,
designated as H1 and H2, and two with RHS columns, designated as R3 and R4, were
tested. All the specimens were of one-sided connections, reproducing beam-to-exterior
column connection assemblies (See Fig.1).
Beam flange
Beam flange
7 8
8
8
7 8
H1 R3
Stiffener Diaphragm
(PL-16) 12 (PL-25) 10
12 4 12
12 4 10
Beam flange
Beam flange
10
12
WF Beam WF Beam
(500x200x10x16) 12 (500×200×10×16) 10
H2 R4
Column flange
Beam top flange
flange
Beam web
The beams are wide-flange sections with the nominal dimensions of 500x200x10x16 mm
and with the steel grade JIS (Japanese Industrial Standard) SN490B (specified minimum
yield strength=325 N/mm2) for all the specimens. The beam top and bottom flanges
were PJP groove-welded to the column flanges using the single-bevel grooves with the
groove angle of 45 degrees. The depth of the grooves are 8 mm (one half of the nominal
flange thickness) for Specimens H1 and R3 and 4 mm (one quarter of the nominal flange
thickness) for Specimens H2 and R4. Reinforcing fillet welds were added all around the
perimeters of the beam sections contacting the column faces. The nominal dimensions
of the welds are shown in Fig.1.
Incomplete penetration of PJP groove welds was found in Specimens H1 and H2 after
testing. Sliced sections of the welded joints are shown in Fig. 2. Portions of the weld
metal near the root of the weld are not fused with the base metal in Specimen H1. The
lack of penetration from the root of the joint was of about 5 mm for Specimen H1 and less
than 1 mm for Specimen H2. These specimens were welded in a flat position.
Preliminary trials showed that better penetration of the weld metal was achieved by
welding in a horizontal position than in a flat position. For Specimens R3 and R4 the
beam ends were welded in a horizontal position after placing the columns horizontally.
The penetration of the weld metal was satisfactory for these specimens. The GMAW
(Gas Metal Arc Welding) electrodes used for fabricating Specimens H1 and H2 were of
the grade JIS YGW11 with the specified minimum yield strength of 390 N/mm2, while
those for fabricating Specimens R3 and R4 were of the grade JIS YGW18 with the
specified minimum yield strength of 430 N/mm2.
u1
L=2400
td u2
Hd
v1 v2
Specimens H1 and H2 have continuity plates welded to the column flanges and webs at
the positions of beam flanges. Specimens R3 and R4 have internal diaphragms CJP
welded to the columns at the positions of beam flanges.
Material properties determined by tensile coupon tests and measured plate thicknesses
are summarized in Table 1. All-weld-metal tension testing was conducted for the
electrode YGW18, the results of which are also included in Table 1. The notch
toughnesses of materials used were measured by Charpy impact testing and summarized
in Table 2. Charpy specimens were taken from plates welded under the conditions close
to those used for fabricating connection specimens. Notch roots were located at the
base metal, weld metal and HAZ.
The both ends of the column were fixed to a strong floor, while the cyclic shear load was
applied at the end of the beam by a double action hydraulic ram. The loading
arrangements are shown in Fig. 3. Lateral bracing systems were provided at two
positions of the beam.
Displacement measurements taken were not only the horizontal displacement u1 at the
loading point but also the vertical displacements v1 and v2 at the column face where the
continuity plates or diaphragms exist and the horizontal displacement u2 at the column
end. The rotation of the beam θf was calculated by the following equation.
U1 − U2 V1 − V2
θf = − (1)
L Hd − t d
where L and (Hd-td) denote the distance from the loading point to the column face and the
The load was applied in the following sequences: firstly a few cycles of loads were applied
in the elastic region and then the amplitude of the beam rotation was increased as 2θp,
4θp, 6θp,,…with 2 cycles of load application at each amplitude, up to failure. θp is herein
defined as the elastic component of the beam rotation when the beam moment at the
column face reaches the fully plastic moment Mp. The elastic rotations measured were
slightly different from those calculated using a simple beam theory. The elastic beam
rotations were obtained from experimental M versus θ relationships. The values of Mp
and θp, calculated using the measured material properties, dimensions and rotations are
shown below.
The moment versus rotation hysteresis loops measured during the test are shown in Fig.
4. The moment is the moment at the column face Mf and is non-dimensionalized by the
full-plastic moment Mp of the beam. The moment takes a positive value when the bottom
flange is in tension. The rotation θf, which was determined by Eq. 1, is also
non-dimensionalized by θp of the beam. Several failure events observed during the test
are described below. The numbers shown in the hysteresis loops indicate the occasions
when these events occurred.
Specimen H1
The beam top flange ruptured along the weld toes (1). The load was reversed. A
ductile crack was found at the toes of the welds at the edge of the beam bottom flange (2).
Local buckles were found in the beam top flange (3). The cracks in the beam bottom
flange grew rapidly (4). Complete rupture of the beam bottom flange (5).
Specimen H2
The beam top flange buckled locally (1). Buckles of the beam top flange and web
became large (2). The beam bottom flange buckled extensively (3). Small cracks were
found along the weld toes in the bottom flange.
Specimen R3
The beam top flange buckled locally (1). A crack was initiated along the weld toes in the
top flange (2). The bottom flange buckled locally. The crack in the top flange extended
over 1 mm. (3). Local buckles of the top flange grew (4). Local buckles of the bottom
flange grew and cracks in the top flange extended significantly (5). The top flange
buckled accompanying lateral buckling (6). The bottom flange buckled accompanying
lateral buckling (7).
Specimen R4
The beam bottom flange buckled locally (1). Two cracks were found at the weld toes in
the beam bottom flange (2). A crack initiated at the weld toe in the bottom flange (3).
Local buckles of the top flange and web progressed. The cracks in the top flange
extended (4). The top flange buckled locally and laterally. The cracks in the bottom
flange grew extensively (5). Extensive lateral buckling of the beam (6).
Mf/Mp
0
Mf/Mp
0
-0.5
-0.5
-1
1
-1 3
-1.5
-1.5 -2
-4 -2 0 2 4 6 8 10 -12 -8 -4 0 4 8 12
θ f /θ p θ f /θ p
0.5 0.5
Mf/Mp
Mf/Mp
0 0
-0.5 -0.5
6
-1 -1
7 2 1
3 4 2
5 -1.5
-1.5
-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10
θ f /θ p θ f /θ p
ROTATION CAPACITY
Although several criteria have been proposed for the assessment of plastic deformation
capacities of beam-column assemblies, cumulative plastic deformation factors for the
connections discussed herein are evaluated and shown in Table 3. The values of
cumulative plastic deformation factors evaluated from the hysteresis loops are
comparable to those for connections with improved details (6) and meet ductility
requirements for the seismic design. Specimen H1 failed prematurely and is the
exception to the above statement. The definition of the cumulative plastic deformation
factor can be found in other literature (for example, reference (7)).
Two failure modes were identified in this series of tests: 1. combined local and lateral
buckling of beams; and 2. tensile failure of welded joints including ductile crack growth.
The ultimate strengths are evaluated focusing only on the latter failure mode.
The ultimate strengths of the welded joints at the beam flange ends are calculated mainly
using the recommendations by AIJ. The resistance factor is excluded. The AIJ
recommendations show no formula for calculating the strength of a welded joint with
combined PJP and superimposed fillet welds. Although Eurocode 3 shows a design rule
that this type of joint can be calculated as a fillet weld with deep penetration, this rule
gives rather conservative predictions as compared with test results.
a2
a1
a3
Beam web
The ultimate strength of the welded joint is calculated as the sum of the ultimate strengths
of one PJP weld and two fillet welds, each being calculated as an independent weld (See
Fig. 7). Namely, the ultimate tensile strength of the welded joint Pu is calculated by
1.4 1.4
Pu = a1l1 Fw,u + a 2l2 Fw,u + a3l3Fw,u (2)
3 3
where the symbols a and l denote the effective throat thickness and length of the weld,
respectively. The subscripts 1, 2 and 3 distinguish among the PJP weld, superimposed
fillet weld and fillet weld on the opposite side of the groove. The factor 1.4 in Eq. 2 is to
increase the design strength of the fillet weld when the axis of the weld is perpendicular to
the direction of loading. Fw,u represents the ultimate tensile strength of the weld metal.
The ultimate strength of the welded joint is governed also by the tensile strength of the
beam flange given by
Pu = BfltflFfl,u (3)
where Bfl, tfl and Ffl,u denote the width, thickness and tensile strength of the beam flange,
respectively. Pu takes the smaller value of those given by Eqs. 2 and 3.
in which Wweb,net signifies the plastic section modulus of the net area of the beam web
considering reduction of the cross section due to the cope holes (no cope hole exists in
the present examples) and Fweb,y represents yield strength of the beam web, while the
symbol m represents the dimensionless moment capacity of the welded web joints. The
second term of the right-hand side of Eq. 4 represents the ultimate flexural load carried by
the welded web joint considering flexibility of the column flange. Further details of Eq. 4
are referred to elsewhere (7,8).
The ultimate flexural capacities calculated by Eq. 4 are summarized in Table 3. When
evaluating ultimate capacities of connections, the measured dimensions, including the
sizes of the welds, and material properties were used. All-weld-metal tension testing of
the electrode YGW11 was not conducted. However, extensive investigations on material
properties of weld metal have already been conducted (9). Especially, many data exist
for welded joints with the base metal of SN490B and the filler metal of YGW11. From
these data, the data for welded joints having the same heat-input as that for the
specimens H1 and H2 were selected. The material properties for the electrode YGW11
that were inferred from the existing data and used for the calculation are: Fw,y=450 N/mm2
and Fw,u=550 N/mm2. The ultimate capacity was governed by the strength of the weld
metal only in Specimen H1. The ultimate capacities of the Specimens H2, R3 and R4
were determined by the strength of the base metal (Eq. 3).
In Table 3, the rows “top flange” and “bottom flange” distinguish the cases when the top is
in tension and when the bottom flange is in tension. As for Specimen H1 the top flange
ruptured at the same load as the load predicted. When the load was reversed, the
connection supported a much higher load than the predicted load. The reason for this
increased resistance is not known. The neutral axis may have moved after failure of the
top flange. Specimen H2 showed about the same ultimate loads as the predicted
capacities. This specimen sustained extensive buckling but cracks found were small.
On the other hand, Specimens R3 and R4 sustained not only extensive buckling but also
significant crack growth. The flanges of these two specimens seemed to have nearly
reached their tensile capacities. The experimental ultimate loads are a little higher than
the predicted ultimate loads. This underprediction is understandable because Eq. 4 is
ignoring the effects of biaxial stress state at the beam flange ends and cyclic hardening of
materials on the flexural strength of connections.
However, the previous investigation used the nominal dimensions of the welds for
modeling the specimens. A significant lack of penetration was observed in the PJP
welds of Specimen H1 as was mentioned earlier. A reanalysis was conducted using
measured dimensions of the welds of Specimen H1. The software used is the ABAQUS
(2003) general purpose FE package.
Figure 9. Contour plot of equivalent plastic strain in cross section on plane of symmetry.
in which σh denotes the hydrostatic stress and σeq denotes the Mises’s equivalent stress.
The triaxiality represents the level of plastic constraint at crack tips. As seen in this
figure, the stress triaxialiy first reaches a maximum value of about 2.0 under small scale
yielding conditions and then decreases as yielding progresses, showing that the
constraint is at a very low level under fully yielded conditions. The J integral reached
about 250 N/mm at the stage when the connection ruptured during the test.
2.5
1.5
Ts
0.5
0
0 50 100 150 200 250
J (N/mm)
The numerical analysis demonstrated that the roots of the PJP welds are nearly under a
plane stress state. The loss of constraint leads to enhanced resistance to the cleavage
fracture. Although the notch toughness of the base metal of Specimen H1 is not
measured, the J integral of 250 N/mm is not large enough to induce cleavage fracture
(See reference (11)). Therefore, it is unlikely to have a brittle fracture starting from the
roots of the PJP welds. If a ductile tensile failure occurs after extensive yielding of the
joint, the ultimate strength of the joint can be predicted based on a simple plastic analysis,
like the analysis presented in the previous section.
CONCLUSIONS
Three connections of these four showed sufficient strength and plastic deformation
capacity. Plastic hinge formed at the beam ends accompanying local and lateral
buckling of the beams. Ductile cracks initiated at the weld toes at the edges of the beam
flanges and grew stably. The previous investigation indicates that it is unlikely that brittle
fractures start from the roots of the PJP welds or cracks at the edges of the beam flanges.
One connection sustained a ductile tensile failure prematurely. This premature failure
was caused by the lack of penetration in the PJP welded joints. It was found important to
select an appropriate joint detail, welding position and other welding conditions for
achieving sufficient penetration in PJP welded joints.
The evaluation of the ultimate strength of the connections based on a simple plastic
analysis was found to be accurate and applicable to the connection design.
A non-linear FE analysis showed that tips of unfused gaps existing in PJP welded tee-butt
joints are not subjected to high plastic constraint. The stresses were not high enough to
induce a brittle fracture starting from the tips.
ACKNOWLEDGEMENTS
The proposals made here are based on joint experimental programs between Sojo and
Kumamoto Universities. The authors wish to thank H. Shinde, M.Sc. student and the
other 4th year students for their hard work in laboratories. This work was partly supported
by the Japanese Society for the Promotion of Science Grant-in-Aid for Scientific Research
under the number 13650645.
REFERENCES
(1) Azuma, K., Kurobane, Y. and Makino, Y., (2000). Cyclic testing of beam-to-column
connections with weld defects and assessment of safety of numerically modeled
connections from brittle fracture. Engineering Structures, Vol. 22, No. 12,
pp.1596-1608.
(2) Azuma, K., Kurobane, Y., Dale, K. and Makino, Y., (2003). Full-scale testing of
beam-to-column connections with partial joint penetration groove welded joints.
Tubular Structures X, M.A. Jurrieta, A. Alonso and J.A. Chica eds., Balkema, Lisse,
The Netherlands, pp. 419-427.
(3) CEN, (1992). Eurocode 3: Design of steel structures, ENV 1993-1-1: Part 1.1
General rules and rules for buildings. Comité Européen de Normalisation, Brussels,
Belgium.
(4) AISC, (2000). Load and resistance factor design specification for structural steel
buildings. American Institute of Steel Construction, Chicago, Ill., USA.
(5) AIJ, (1998). Recommendation for limit state design of steel structures. Architectural
Institute of Japan, Tokyo, Japan. (in Japanese)
(6) Kurobane, Y., Azuma, K. and Makino, Y., (2003). Fully restrained beam-to-RHS
column connections with improved details. Tubular Structures X, M.A. Jurrieta, A.
Alonso and J.A. Chica eds., Balkema, Lisse, The Netherlands, pp. 439-446.
(7) Kurobane, Y., Packer, J.A., Wardenier, J. and Yeomans, N.F., (2004). Design guide
for structural hollow section column connections. CIDECT/Verlag TÜV Rheinland,
ABSTRACT
INTRODUCTION
The design philosophy for tension members is largely an attempt at replication of the basic
tensile behavior found in a tensile coupon and early tension members were made as pinned
ended where the reduced member length yielded and eventually ruptured under increasing
axial load. While this behavior may not always be achieved in modern designs of bolted or
welded members, the treatment of the limit state of gross member yielding as a goal and a
secondary limit of ultimate rupture or fracture of the member itself as an additional limit state
is common practice. A connected member may have its behavior controlled by rupture as
influenced by connection parameters and never achieve the gross yield limit state albeit at a
reduced rated capacity or resistance for design.
In addition to the necessary or purposeful reduction of the net cross sectional area of a
member, the additional concerns that are common to bolted and welded connections are
those associated with “shear lag,” “block shear” and “eccentricity.” When the long
connections and the eccentricities of force at the connection are minor, the problems
associated with reduced section capacity are minimized or absent, but shorter connection
lengths tend to be controlled by these reduced capacity failure modes.
The current study is examines the behaviour of simple single plate, fillet welded, connections
to single and paired channels used as tension members. The results of full scale tests on
channels are studied here and compared to tests on members of similar geometry but
consisting of slotted hollow structural sections. An immediate goal is to evaluate these results
to characterize the behavior for design, but a longer term aspect is the measurement of
strain fields during the fracture at ultimate and the anticipated development of reasonable
analytical limits for advanced analysis applied to more complex connection configurations.
In the examination of past work on this and closely related problems the examination of
parameters that were not addressed or varied intentionally, led to a series of tests that
demonstrated the effect of those parameters and helped explain some of the variations in
historical results. This paper will emphasize the parametric effects that were found to
potentially have the largest impact on design and will introduce the ongoing development of
large strain measurement in steel near rupture.
EXPERIMENTAL APPROACH
In all cases, the design of the specimens followed the design or proportioning of practical
tension members. This entails the design of tension members that are of similar length,
strength and cross section as those commonly used in current steel construction. However,
the constraints of the testing equipment imposed a maximum specimen length of about 2
metres. This length was thought to be representative of tension members commonly used in
small trusses. While many tension members would be longer, it was considered reasonable
to assume that the additional length would have an insignificant effect on the results of the
investigation. The length of the tension member was one of the parameters that had been
overlooked in many earlier works while the member and connection proportions were
realistic; economy and equipment limitations were likely influences.
A channel section, (C150 x 12) was selected for two reasons: its strength would be sufficient
that it could be reasonably used in various real applications, and its geometry had a
theoretically significant shear lag effect.
The connection length was chosen as a balance between two competing factors: realism and
efficiency. Excessively short connections did not seem realistic as they did not seem to be on
the order of what might be reasonably designed in practice. To ensure, however, that shear
lag governed the design, the connection length had to be short. A compromise was reached
with a connection length equal approximately to the depth (d) of the member. This produced
reasonable design and a calculated efficiency that would cause the net section strength limit
state strength to govern
The specific designs of specimens were done so that differences in the connection efficiency
could be assumed to result from the specific changes to specimen orientation, symmetry or
geometry. Figure 1 illustrates the variation in specimen design and the philosophy behind
Figure 1. Cross section variation for channel-like tension members to single plates.
All specimens were similar on either connection end and thus possessed a symmetry about
the mid-length. All cross sections were symmetric about at least one axis; the eccentricity of
a member centroid was varied by orientation with the connecting plate as seen in the figure.
The test matrix was designed so that for any of the variables being examined, specimens
were tested that varied the parameter of interest at a rational extreme (for example, there
were specimens with high weld metal strength and low weld metal strength). All other details
of the specimens were identical. In this way, a variation in strength between otherwise
comparable specimens was reasonably attributed to the chosen parameter variation.
Two test series were planned; the second was intended to be guided by the results of the
first. A simple naming convention for each series is outlined as follows:
Series 1
The naming convention used the following template: prefix-a-suffix-A or B where
prefix: MH1 for all series 1 specimens
a: 6 or 8, the design weld size
suffix: 480 or 550, the design weld metal strength
A or B: used to identify replicate specimens
Series 2
A modified naming convention used the following template: prefix-abc-suffix-A/B where
prefix: MH2 for all series 2 specimens
a: C or H, section type, C for channels, H for HSS
b: T, B or X, orientation, T for toe to toe channel sections, B for back to back channel
sections and X for HSS sections
The baseline connection was a pair of longitudinal welds of length equal to the nominal
channel section depth (150mm). For two specimens (MH1-6-480A and MH2-CBS-19B), an
additional transverse weld was added across the web/gusset at the end of the sections.
HSS sections were slotted and connected similarly to pairs of channels with a pair of
longitudinal welds on each side of the gusset plate. A close up of the baseline connection is
shown as built in Figure 2.
All specimens were prepared for a photometric strain measurement technique which was
being evaluated in this study. A visible grid was applied to the surfaces of the specimen in
the region of the connection using contrasting inked dots on a painted surface. Before
loading high resolution pictures of the grid were then taken as a zero reference. Resistance
type strain gauges were also applied to the specimens at their centres at this time. These
strain gauges were used to verify gross section yielding during the testing.
Connection specimens were mounted in an MTS 3000 kN universal test frame using a
special gusset plate assembly to maximize total specimen length. Linearly variable
differential transformers (LVDT) were mounted over the connection length. Load was
applied under displacement control until the initiation of failure that is typically defined by the
A view of one of the connections (boxed channels) showing the initiation of the net section
rupture (top left), the “strain grid,” and indications of yielding.
Figure 3. Close up of boxed channel at failure showing fracture initiation (upper left).
TEST RESULTS
Discussion of results
On the basis of these results, several comparisons were undertaken to quantify and observe
the effects of specific variations of connection parameters.
Current design standards do not consider weld size (or connected area) in their estimation of
connection efficiency. This reflects their basis in experimental results from bolted and riveted
specimens. In this research programme specimens were tested whose design efficiencies were
identical, but whose weld sizes were significantly varied. Specimens MH1-6-480B and MH1-6-
480C were connected with 6 mm welds where specimens MH1-8-480A and MH1-8- 480B were
identical except that they were connected using 8 mm welds. Their global behavior is very
consistent as seen in the Load-Displacement plots of Fig. 4.
There is clearly a small increase in the efficiency of the connection with increasing weld size.
In fact, the increase shown above (from an average efficiency of 90% to 95%) is the
minimum increase as neither of the MH1-8-480 series specimens failed at the weaker end as
indicated by design but practically reached the ultimate strength of the member section.
Current design standards that account for shear lag reduction in welded tension members do
not account for member symmetry with respect to connection loading. For example, the
design strength of specimens comprised of back-to-back channels using CAN/CSA S16-01
(1) is the same as that for specimens comprised of a single similarly connected channel.
These design codes assume that the effect of the eccentricity at the connection will be small
Single member specimens without axial symmetry experience realignment (through bending)
of their sections and connections so that the eccentricity of the section and connection is
minimized. Since they have no net eccentricity, the symmetric members do not experience
this realignment to the same degree. Realignment of the section is proposed to have two
competing contributions to the shear lag effect: a decrease in the degree of stress
redistribution required to shed the load into the connection and an increase in extreme fiber
stresses and strains caused by the bending. For small eccentricities, unsymmetric
specimens can have higher net section efficiencies as a result of this realignment. If the
eccentricity of the section is high however, the extreme fiber stress/strain caused by bending
can have a substantial reduction on net section efficiency.
1800
1600
1400
1200
1000
Load (kN)
800 MH1-8-480B
MH1-8-480A
600
MH1-6-480B
MH1-6-480C
400
200
0
0 20 40 60 80 100 120 140 160
Global Displacement (mm)
Figure 4. Load vs. global displacement for specimens with varying weld sizes.
As illustrated by the following results, Fig. 5, overall member symmetry with respect to the
connecting plate can have significant effects on the efficiency in welded tension connections.
The eccentricity of the force remains a significant value even after some realignment and relief.
The amount of reduction in strength is a function of the stiffness of the member and the restraint
that is provided by the connecting ends. The connections here have little rotational restraint
since they are only loaded and attached at the edge of the connecting plate.
The use of slotted HSS sections is popular in modern steel design. It was decided that a
study of the comparative efficiencies of slotted HSS members and similar boxed channels
would provide more information on very similar configurations. Specimens of the series
MH2-HXL-64 and MH2-CTS-12 have this global configuration similarity. It was postulated
that the less restrained boxed channel sections may deform (move towards each other) and
achieve a better efficiency. The results would demonstrate any effect of the apparent
restraint given by the continuity of the HSS that is absent between the two channels as they
enter the connection.
600
500
Load (kN)
400 MH2-CTU-12A
MH2-CTU-12B
300 MH2-CTS-12A
MH2-CTS-12B
200
100
0
0 10 20 30 40 50 60 70 80 90
Global Displacement (mm)
Figure 5. Load per channel vs. global displacement for (double channel) symmetric and
(single channel) unsymmetric connections using identical members.
These slotted HSS members did not deform plastically as did the double boxed channel
sections.. This is clearly evident in their efficiency (load) vs. global displacement graphs
shown in Fig. 6. Their stiffness is nearly linear until failure. However, the boxed channel
sections clearly show a high degree of plastic deformation before failure. Their freedom to
deform inward just out side of the connection permits redistribution and this inward
movement was visibly apparent during testing. The efficiency here is the percentage of the
full member yield that was achieved. Further comparisons of efficiencies related to design
are found in Table 1.
1.4
1.2
1
Efficiency (Load/[(Ag)(Fy)])
0.8
MH2-HXL-64B
0.6 MH2-HXL-64A
MH2-CTS-12B
0.4
MH2-CTS-12A
0.2
0
0 10 20 30 40 50 60 70 80 90 100
Figure 6. Comparison of the efficiency for slotted HSS and boxed channel sections.
It is clear from the results of this series of tests that the reduction in ductility caused by the
weld across the web end of the channel had a similar effect as the increase in connected
area caused by the additional weld. In all cases, the sections with the additional weld
exhibited very similar efficiencies to those without it. Clearly the ability for the section to
distribute stress/strain across the section through plastic deformation is as important as the
weld length. The important aspect of this comparison is that the addition of a weld across the
web of the section did not have a significant effect on the efficiency of the connection.
CONCLUSIONS
This experimental study has demonstrated the influence of a few of the parameters in welded
connection design that have an influence on the strength and ductility of tension members.
This data and the detailed displacement/strain distribution data that was not reported here
provide insight and support for new analytical modeling capabilities of the nonlinear response
of complex geometries up to the development of fracture and ultimate strength.
An increase in weld size causes an increase in efficiency in shear lag effected welded
tension connections. This finding seems intuitive since connections with larger welds have
Member symmetry (with respect to the load or connection plate) can have a profound effect
on efficiency in shear lag effected welded tension connections. Member sections oriented so
that their individual eccentricity was maximized (boxed channels) with respect to the gusset,
and exhibited large reductions in efficiency when tested singly as a non-symmetric member.
But when tested as part of a symmetric member (double boxed channels) the efficiency was
increased. Many current design standards do not prescribe variations in shear lag reductions
to account for this effect of global symmetry on efficiency. The current standards penalize the
symmetric members in that their design efficiencies are representative of the lower
experimental values for single unsymmetrically loaded member.
Slotted HSS sections although geometrically similar to boxed channel sections in shear lag
affected welded tension connections show distinctively different behavior. The continuity of
the sides of the HSS section beyond the slot restrains shape of the section; the more ductile
behaviour of the channel sections occurred in part as a result of the freedom of the individual
channel halves to realign a small amount. The limited results, shown here, were for HSS that
were not welded at the end of the slot.
In a simple comparison of the results of this series with the simplified empirical shear lag
reduction factor (1-x/L), the results of this study (6) of welded connections are plotted in Fig.
8 with a selection of past research (2, 3, 4, 5) results for various details of a similar nature.
This compilation is not complete but does indicate a trend as well as a significant scatter.
Improvement by reduction of scatter has been accomplished by reinterpreting and changing
the rules of application of the reduction factor. Further analytical and experimental work is
underway to find a non-empirical solution to this problem that will lead to the improved
advanced analysis of strength and deformation for this and for more complex connection
details.
ACKNOWLEDGEMENT
The substantial portion of the support for this study is from a grant from the SSEF (Steel
Structures Education Foundation) and from the NSERC (Natural Sciences and Engineering
Research Council of Canada). All of the tests were performed in the large Structural Test
Facilities at the University of Toronto, Department of Civil Engineering. Support in the form
of material, fabricated specimens and practical technical interaction is greatly appreciated
and crucial to this effort; CANRON Eastern Division Ltd., M&G Steel Ltd, Leroux Steel Inc.,
CoSteel LASCO Ltd., and Hobart Inc. are especially cited for their generous contributions.
REFERENCES
110
Gonzalez Giroux and Easterling - 1989
Sarvinis - 1989
100
Davis and Boomslitter - 1934
80
70
60
50
0,5 0,6 0,7 0,8 0,9 1
1-xbar/L
Figure 8. Experimental efficiency correlation with empirical shear lag reduction factor.
ABSTRACT
Current design models for fire resistance of structures are based on isolated
member tests subjected to standard fire conditions. Such tests provide raw
data for numerical models and to develop design methods, however, they do
not reflect the behaviour of a complete building. Many aspects of behaviour
occur due to the interaction between members and cannot be predicted or ob-
served in isolated tests.
In order to study this global structural behaviour a research project was con-
ducted on the 8-storey steel-concrete frame building at the Cardington labora-
tory. This paper summarises the experimental programme and presents the
thermal and mechanical results of the studied joints.
INTRODUCTION
Significant developments have been made in the analysis of the behaviour of steel framed
structures under fire conditions over the last ten years. Due to the high cost of full-scale fire
tests and size limitations of existing furnaces, these studies are based on the observation of
real fires and on tests performed on isolated elements subjected to standard fire regimes,
which serve as reference heating, but do not model the natural fire. However, the failure of
the World Trade Centre on 11th September 2001 and, in particular, of building WTC 7,
alerted the engineering profession to the possibility of connection failure under fire condi-
tions. Many aspects of behaviour occur due to the interaction between members and cannot
be predicted or observed from isolated tests, such as global or local failure of the structure,
stresses and deformations due to the restraint to thermal expansion by the adjacent struc-
ture, redistribution of internal forces, etc. Otherwise, it is known that even nominally ‘simple’
connections can resist significant moments at large rotation. At the severe deformation of a
structure in fire, moments are transferred to the connection and to the adjacent members,
and hence they may have a beneficial effect on the survival time of structure.
Unlike the standard fire curve, a natural fire is characterized by three phases: a growth
phase, a fully developed phase and a decay phase. It is necessary to evaluate not only the
effect on the structural resistance during the heating phase, but also the high cooling strains
in the joints induced by distortional deformation of the heated elements during the decay
phase.
In order to study this global structural behaviour a collaborative research project (Tensile
membrane action and robustness of structural steel joints under natural fire, European
Community FP5 project HPRI - CV 5535) was conducted on the 8-storey steel-concrete
The experimental results on the response of steel connections under fire conditions are relatively
recent and limited, partly because of the high cost of the fire tests and the limitations on the size
of furnace used. Only a few connections tests have been performed and they have concentrated
on obtaining the moment-rotation relationships of isolated connections. It is doubtful whether
these results will be useful when dealing with the behaviour of frame connection. The develop-
ment of the Cardington Laboratory of the Building Research Establishment (BRE) has provided
the opportunity to carry out several research projects that included full-scale fire tests. A brief
state-of-art description of the steel joints fire tests is presented in Table 1.
The BRE’s Cardington Laboratory is a unique worldwide facility for the advancement of the
understanding of whole-building performance. This facility is located at Cardington, Bedford-
shire, UK and consists of a former airship hangar with dimensions 48 m x 65 m x 250 m. The
Cardington Laboratory comprises three experimental buildings: a six storey timber structure,
a seven storey concrete structure and an eight storey steel structure.
A B C D E F
9000 9000 9000 9000 9000
4
6000
3
Fire compartment
9000
2
6 950
6000
10 900
1
Figure 1. The Cardington fire test (9) with identification of the fire compartment.
Fire compartment
The fire test was carried out in a centrally located compartment of the building, enclosing a
plan area of 11 m by 7 m on the 4th floor.
The mechanical load was simulated using sandbags, each weighing 1100 kg. In addition to
the self-weight of the structure, sand bags represented the following mechanical loadings:
remaining permanent actions, 100% of variable permanent actions and 56% of live actions.
The fire load was provided by 40 kg/m2 of wooden cribs (moisture contents < 14 %) covering
the compartment floor area. The fire compartment was bounded with three layers of plaster-
board (15 mm + 12.5 mm + 15 mm) with a thermal conductivity around 0.19-0.24 W/mK. In
the external wall (gridline 1) the plasterboard was fixed to a 1.5 m high brick wall. An opening
1.27 m high and 9 m long simulated an open window to ventilate the compartment and to al-
low the observation of the behaviour of the various elements. The columns, external joints
and connected beam (1.0 m from the joints only) were fire protected to prevent global struc-
tural instability. The material protection used was 15 mm of Cafco300 vermiculite-cement
spray, with a thermal conductivity of 0.078 W/mK.
Structural arrangement
The steel structure exposed to fire consists of two beam sections (356x171x51 UB for the
edge beams and the 6 m primary beams and 305x165x40 UB for the internal secondary
beams) and two columns (305x305x198 UC and 305x305x137 UC) as shown in figure 2a
D E wp c
End plate connection
eob
eo
P8-260x140 Secondary beam
p1 e1
305x165x40UB
eop
2
End plate connection
P8-260x150 Primary beam
aw aw
hp
356x171x51UB
Primary beam
356x171x51UB Secondary beam
305x165x40UB
Fin plate connection g
P10-260x100 Fin plate connection
P10-260x100
e2 e2
Secondary beam N
356x171x51UB
1 End-plate Fin-plate
Figure 2. (a) Arrangement of members in the fire compartment; (b) connections geometry.
Table 3. Material properties of steel and steel connectors at ambient temperature (12).
Material Ultimate Stress (MPa) Yield Stress (MPa)
nominal measured nominal measured
S275 430 469 275 303
Steel
S355 510 544 355 396
Plate Grade 43 430 ----- 275 -----
Bolts 8.8 800 869 640 -----
The main requirements of the instrumentation were to measure the temperature, the distribu-
tion of internal forces, the deflected shape of the floor and main structural elements. The in-
strumentation included thermocouples, strain gauges and displacement transducers (figure
3). Additionally, ten video cameras and two thermo imaging cameras recorded the fire and
smoke development, the structural deformations and the temperature distribution with time.
Thermocouples (TC)
D2 major axis TC + HSG in Bolt D2 minor axis
HT Strain Gauges (HSG)
EXPERIMENTAL OBSERVATIONS
Temperature variation
Figure 4 compares the temperatures recorded in the compartment with the parametric curve
presented in Eurocode 1, Annex A (11). The quantity of thermal load and the dimensions of
the opening on the facade wall were designed to achieve a representative fire in the office
building. The maximum recorded compartment temperature near the wall (250 mm from D2)
was 1107,8 ºC after 54 minutes, while the maximum temperature predicted by the parametric
curve was 1078 °C after 53 min., see figure 4.
Temperature, °C
Prediction prEN 1991-1-2, Annex A
1200 1108
Back in fire compartment c d
300
1000 1078 c
500 500
800
Fire compartment (DE, 1-2)
600
Average temperature
400
In front of fire compartment
53 54 d
200
Time, min
0
0 15 30 45 60 75 90 105 120 135 150
Measurements of the temperature in the connections were taken on the beam adjacent to
the connection, in the plate and in the bolts. A summary of the temperatures recorded in the
connections is presented in figures 5, 6 and 7.
In the heating phase, the joint temperature is significantly lower than the remote bottom flan-
ge, which is usually the critical element that defines the limiting temperature of the beam; in
contrast, the cooling down in the joints was slower. At the maximum temperature, the joint
temperature was around 200 ºC lower than the limiting temperature of the beam. The first
bolt row from the top was significantly cooler than the lower bolts, because of shielding by
the adjacent slab and column. The end-plate was hotter than the bolts at the same level due
to the ratio of the bolt diameter to the end-plate thickness (20 mm).
Figure 5a. Temperature variation within the beam-to-beam fin plate connection D1/2-E1/2.
g
c 30 min.
plate 63
bolts 106 min. 45 min. 90 min. 75 min. 60 63 min. 60 min.
h beam
e
end
f
i
defgh i
800
c
600
N
bott. flange g
400 plate 4th row f
D1 E1
4th bolt f
200 upp. flange h
1st bolt c
plate 1st row c Time, min
0
0 15 30 45 60 75 90 105 120 135 150
c
60 min. 90 min. 75 min.
d
e plate
bolts 45 min. 106 min.
f
beam 60 min.
106 min.
end
g 45 min. 75 min.
Temperature, °C
400 °C 500 °C 600 °C 700 °C 800 °C
600 N
D1 E1
400
200
Time, min
0
0 15 30 45 60 75 90 105 120 135 150
h
c
plate
d bolts
e beam
end
45 min. 106 60 min. 90 min. 75 min.
f
Local buckling of the beam lower flange was one of the main mechanisms. It is observed in
the lower beam flange and web adjacent to the joints, see figure 8, the concrete slab having
restrained the upper flange. This local buckling occurs during the heating phase after about
23 min. of fire (observed by thermo imaging camera), due to the restraint to thermal elonga-
tion provided by the adjacent cooler structure and the structural continuity of the test frame.
D2 E2
D1 E1
D2 E2 D2 E2
N N
(a) D1 E1 (b) D1 E1
Figure 9. (a) Beam web in shear; (b) buckling of column flange in compression.
The local buckling of the column flange in compression was observed in the major axis
beam-to-column joints, see figure 9b. This behaviour results from the small column flange
thickness (t = 21.4 mm) and the small distance between the bolts, the bolted end-plate be-
having as a welded joint. This behaviour was observed in both columns flanges of the two
beam-to-column joints (D2; E2).
Fracture of the end-plate along the welds was observed, caused by the horizontal tensile
forces during cooling of the connected beam under large rotations associated with flexible
end-plate joints, see figure 10. The fracture occurred along one side of the connection only,
while the other side remained intact. After one side has fractured, the increased flexibility al-
lowed larger deformations without further fracture. This behaviour was observed in both the
major axis beam-to-column joints (D2-D1; E2-E1) and the minor axis beam-to-column joints
(D2-C2).
The elongation of the holes (in top bolt row +12 mm and in bottom bolt row -10 mm in con-
nection D1/2-E1/2) in the beam web in the tension/compression part of the fin plate connec-
tion is due to the associated large rotations, see figure 11. The elongation of the holes oc-
curred on the web of connected beam, while the fin-plate remained intact: the beam web
thickness (6 mm) is smaller than the fin-plate (10 mm); again, the elongation of the holes of
the beam web leads to increased joint flexibility, allowing larger deformations without further
fracture.
D2 E2
D1 E1
Figure 11. Elongation of holes in the beam web in fin plate connection.
CONCLUSIONS
Globally, we did not succeed to reach the collapse of structure or its parts was not reached
2
for the fire load of 40 kg/m , which represents the fire load in a typical office building, to-
gether with a mechanical load greater than standard approved cases. The structure showed
good structural integrity. The test results supported the concept of unprotected beams and
protected columns as a viable system for composite floors.
More specifically, with respect to the performance of the joints, local buckling of the lower
flanges of beams was observed after 23 minutes of fire. Fracture of the end plates occurred
under cooling in the heat affected zones of welds without losing the shear capacity of the
connections. The fin plate connections behaved in a ductile fashion due to elongation of
holes in bearing.
The detailed behaviour of the beams and connections in the real boundary conditions is cur-
rently being investigated to refine the analytical and numerical prediction models, preliminary
published results being available in (14, 15). Further experimental tests on subassemblies
using the boundary conditions measured on the Cardington frame test are currently being
prepared.
ACKNOWLEDGEMENTS
The project has been supported by the grant of European Community FP5 HPRI - CV 5535
REFERENCES
(1) Kruppa, J. “Resistance au feu des assemblages par boulons haute resistance”, St.
Remy-les-Chevreuse, Centre Technique Industriel de la Construction Metalique,
France, 1979.
(2) “The performance of beam/column/beam connections in the BS476: art 8 fire test”, Brit-
ish Steel (Swinden Labs), Reports T/RS/1380/33/82D and T/RS/1380/34/82D.
(3) Lawson, R.M., “Behaviour of steel beam-to-column connections in fire”, The Structural
Engineer, 68(14), London, pp. 263-271, 1990.
(4) Leston-Jones, L.C., Burgess, I.W., Lennon, T. and Plank, R.J.: “Elevated temperature
moment-rotation tests on steelwork connections”. Proceedings of Institution of Civil
Engineers. Structures & Buildings; 122(4): pp. 410-419, 1997.
(5) Al-Jabri, K.S., Burgess, I.W. and Plank, R.J.: “Behaviour of steel and composite beam-
to-column connections in fire - volume 1”. Research Report DCSE/97/F/7, Department
of Civil and Structural Engineering, University of Sheffield, 1997.
(6) Al-Jabri, K.S., Lennon, T., Burgess, I.W. and Plank, R.J., “Behaviour of steel and com-
posite beam-column connections in fire, Journal of Constructional Steel Research, 46,
pp. 1-3, 1998.
(7) Spyrou, S. and Davidson, J.B., “Displacement measurement in studies of steel T-stub
connections”, Journal of Constructional Steel Research, 57, pp. 647-659, 2001.
(8) Beneš, M.: Equivalent T stub in tension under fire, Ph.D, thesis, CTU in Prague, 2004,
p. 124.
(9) Wald, F., Simões da Silva, L., Moore, D., Lennon, T., Chladná, M., Santiago, A., Beneš
M. and Borges, L.: Experimental behaviour of steel structure under natural fire, The
Structural Engineer (submitted for publication, 2004).
(10) Bailey, C.G., Lennon, T., Moore, D.B.: The behaviour of full-scale steel-framed building
subject to compartment fires, The Structural Engineer, Vol.77/No.8, 1999, p. 15-21.
(11) Draft prEN - 1991-1-2: 200x, Part 1.2: General actions – Actions on structures exposed
to fire, Eurocode 1: Actions on structures, Final Draft, 2002, CEN, European Commit-
tee for Standardization, Brussels, 2002.
(12) Bravery, P.N.R.: Cardington Large Building Test Facility, Construction details for the
first building, Building Research Establishment, Internal paper, Watford 1993, p. 158.
(13) Wald F., Santiago, A., Chladná, M., Lennon, T., Burgess, I.W. and Beneš, M.: Tensile
membrane action and robustness of structural steel joints under natural fire, Internal
report, Part 1 - Project of Measurements; Part 2 - Prediction; Part 3 – Measured data;
Part 4 – Behaviour, BRE, Watford, 2002-2003.
(14) Santiago, A., Simões da Silva, L., Vila Real, P., Franssen, J.M.: Effect of cooling on the
behaviour of a steel beam under fire loading including the end joint response, in Pro-
ceedings of the 9th International Conference on Civil and Structural Engineering Com-
puting, ed. Topping, B.H.V., Civil-Comp Press, Stirling, United Kingdom, paper 65,
2003.
(15) Sokol, Z., Wald, F., Pultar, M. and Beneš, M.: Numerical simulation of Cardington fire
test on structural integrity, in Mathematical and computer modelling in science and en-
gineering, ed. Kočandrlová M., Kelar V., CTU, 27-30.1.03, p. 339-343.
ABSTRACT
Composite connection made with RHS chord or column filled by concrete and
branches with RHS steel profile are studied herein. The aim of this paper is
deriving a simple theoretical formula for calculating the strength and stiffness
of such joints. Test results of twelve connections in natural scale are
described. Geometry and material properties of the tested joints are given.
Theoretical solution of the joint strength and stiffness are proposed and the
comparisons between theoretical and experimental results are presented.
INTRODUCTION
European Code EC 4 (1) makes possibility to design much more effective structures which
combined advantages of steel structural sections and concrete structures. However many of
structural problems are not included in this regulation. If the steel Vierendeel girder should be
loaded by the significant load the interesting solution is such design where RHS or box
chords section are concrete filling. From the structural point of view the box chords section
ought to have the possibly large dimensions and their wall thickness ought to as small as
possible. However in such situation two problems arise. Local instability of section walls
leads to degradation of chord resistance and very thin walls decrease the strength and
stiffness of joints. It leads to decreasing the overall carrying capacity of such structure. The
strengthening of joints is possible by the steel plate welded to the face of chord. This
however, does not strength the slender webs of box section. The other possibility is concrete
filling of hollow section. Such filling leads also to increasing the thermal capacity of structure
and its fire resistance. The comparison of these two ways of strengthening is given in (2).
Strength and stiffness of T concrete filled joints made with RHS are the aim of this paper.
Test rig is shown in figure 1. Twelve joints in natural scale were tested here up to failure.Ten
of specimens, made with RHS, have the concrete filling chords. Two additional specimens
are not concrete filling and are used for comparison how the concrete filling is effective
compare with the pure steel RHS joints. The compression load equal to 420 kN, simulating
the load in real structure, was applied to chord before the branch was loaded. Therefore in
several steps the branch was loaded up to the reach the failure load. After each loading step,
the joint was unloaded to measure the permanent deformations of the tested specimen.
Typical type of joint failure was the inelastic deformation of the flange in the tension zone and
finally cracking of welds, see figure 2.
In Table 1 the geometry of the specimens, mechanical properties, and failure moment are
given. The mechanical properties are the medium value from three tension coupons tests.
The concrete mechanical properties were obtained from tests of five concrete standard
cubes 100x100x100mm. Results obtained shown that the filling concrete has characteristic
stress 42 MPa. Thickness of welds was equal to a = 1,2 tn.
Strength prediction
For prediction the theoretical strength of filled joints, from the observations which were done
during experimental tests, the following assumptions are adopted:
xb o
M ip,1,Ed f 2
f f to
bo bn 3 1
d Legend: Φ
φ 1, φ 2 , φ 3
- virtual
0,65h n f rotations in
2
plastic hinges
δ- virtual
ho displacement
hn concrete 0,65hn - range
of the tension
zone
1. Yield line mechanism, which is created in the tension zone of joint, is deceived. Erasing
inelastic deformations leads finally to situation that steel loaded flange looses the contact
with filled concrete.
2. In compression zone the connection is almost absolutely stiff. So, for the simplicity could
be assumed that this part of joint is compact.
3. In tension zone range of yield line mechanism is larger then in compression one. From
the tests the assumption is adopted that in tension zone range of yield line mechanism is
equal to 0,65hn
For the prediction of theoretical strength the yield line mechanism is proposed, similar to that
as for unstrengthen steel joints (3). Proposed theoretical model is shown in figure 3.
From the equation that the virtual work dissipated in the hinges by inner forces on the virtual
rotations and deformations is equal to outer forces work on the virtual displacements the
formula to predict strength of joints is given
M ip ,1, Rd 2 8
= + ( x + 0 ,65 ) + 3,08 (1)
η bo m pl x 1− β
1− β
x= (2)
2
M ip ,1, Rd 8 0 ,65
= (1 + ) + 3,08 ( 3)
η bo m pl 1− β 1− β
Initial stiffness
Initial stiffness Sj,ini is a coefficient in the linear function between the bending moment applied
to the joint and its local rotation (M = Sj,ini Φ).
For pure steel joints the power function is assumed to predict the initial stiffness of the joints
when β > 0,4, see (4). Analysis of the influence of particular parameters leads to the
following formula:
However, for the concrete filled joints, after the numerical simulation, the increasing
coefficient 1,3 is suggested and the design value of the joints initial stiffness could be
calculated as below:
Secant stiffness
According the recommendations which are given in EC-3, see part 5.1.2 (5), as a
simplification, the rotational, secant stiffness may be taken as Sj,ini /η in the analysis for all
values of the design moment. Therefore, the secant stiffness of concrete filled joints is
suggested to be calculated using coefficient η = 2, see Table 5.2 (5), as below:
In figure 4 to 15 the moment-rotation curves (M - Θ) for each tested joints are presented.
They are shown not only loading but also unloading curves registered by LVDT and dial
gauges. Unloading curves gives possibility to obtain the end of its elastic behaviour and
show the arising of the joint permanent deformations. In Table 2 the comparison between the
theoretical prediction and the test results is presented.
Moment kNm
30
8
Moment kNm
25
6 LVDT Loading
20
LVDT Unloading
LVDT Loading
Dial gauge Loading
LVDT Unloading 15
4
Dial gauge Loading Dial gauge Unloading
Figure 4. Joint BS1 β = 0.57, λo=26.9. Figure 5. Joint BS3 β = 0.86, λo=19.9.
BS5
BS4
16 27
Moment kNm
Moment kNm
24
21
12
18 LVDT Loading
15 LVDT Unloading
Figure 6. Joint BS4 β = 0.71, λo=27.5. Figure 7. Joint BS5 β = 0.71, λo=19.9.
BS6 BS7
15 15
Moment kNm
Moment kNm
12 12
LVDT Loading
LVDT Loading
9 9 LVDT Unloading
LVDT Unloading
Dial gauge Loading
Dial gauge Loading
Dial gauge Unloading
6 Dial gauge Unloading 6
Welds failure
Welds failure
Initial stiffness (6)
Initial stiffness (6)
3 3 Secant stiffness (7)
Secant stiffness (7)
Design load (3)
Design load (3)
0 0
0 Rotation
1,5x 10-2 rad3 4,5 6 0 Rotation
1,5 x 10-23rad 4,5 6 7,5
Figure 8. Joint BS6 β = 0.57, λo=19.7. Figure 9. Joint BS7 β = 0.71, λo=26,2.
Moment kNm
Moment kNm
18 30
15 25 LVDT Loading
LVDT Loading
LVDT Unloading
LVDT Unloading
12 20
Dial gauge Loading
Dial gauge Loading
9 Dial gauge
Dial gauge Unloading 15
Unloading
Welds failure
Welds failure
6 10 Initial stiffness (6)
Initial stiffness (6)
Secant stiffness (7)
3 Secant stiffness (7) 5
Design load (3)
Design load (3)
0 0
0 Rotation
1,5 x 10-23 rad 4,5 6 7,5 0 Rotation
0,5 x 10-21rad 1,5 2 2,5
Figure 10. Joint BS8 β = 0.71, λo=20. Figure 11. Joint BS9 β = 0.86, λo=20.
BS10 BS11
27 15
Moment kNm
Moment kNm
24
12
21
LVDT Loading LVDT Loading
18
LVDT Unloading LVDT Unloading
9
15
Dial gauge Loading Dial gauge
Loading
12 Dial gauge Dial gauge
Unloading 6
Unloading
9 Welds failure Welds failure
Initial stiffness (6) Serie6
6 3
Secant stiffness (7) Serie7
3
Design load (3) Serie8
0 0
0 1,6x 10-2 rad
Rotation 3,2 4,8 6,4 0 1,5
Rotation x3 10-24,5
rad 6 7,5 9 10,5
Figure 12. Joint BS10 β = 0.86, λo=26,9. Figure 13. Joint BS11 β = 0.57, λo=20.
S2 S12
22 21
Moment kNm
Moment kNm
18
16,5
15 LVDT Loading
LVDT Loading
LVDT Unloading
LVDT Unloading
12
Dial gauge
11 Dial gauge Loading
Loading Dial gauge
Dial gauge 9 Unloading
Unloading Welds failure
Welds failure
6 Initial stiffness (5)
5,5 Initial stiffness (5)
Secant stiffness
Secant stiffness 3
Design load
Design load
0 0
0 1,5x 10-2 rad
Rotation 3 4,5 6 0 1,5 x 10-2
Rotation 3 rad 4,5 6 7,5
Figure 14. Joint S2 β = 0.71, λo=19,7. Figure 15. Joint S12 β = 0.71, λo=19,6.
Theoretical strength of pure steel joints S2 and S12 is estimated from the formula presented
in (3) but initial stiffness is calculated from formula (5). Secant stiffness of these joints are
equal to Sj,ini /η , where η = 2.
CONCLUSIONS
a. Table 2 shown that formula (3) good predicts the strength of T RHS joints which chords
are filled by concrete.
b. Filled joints could be classified as the joints with full strength (see joints BS 3 and BS 9 in
Table 2), where the parameter β < 0,85 and if the wall slenderness of the chord section
λo is not slender then 20. With regard to the unfilled, pure steel joints, the joints could be
classified as the full strength if they are more compact i.e. parameter β < 1 and λo < 16.
c. As it could be expected, it was noticed that the rotation capacity of filled joints is much
smaller than the adequate steel joint. However, all the tested joints have the rotation
capacity over 1,5 x 10-2 rad, what guarantee to reach the serviceability limit of beam. For
very flexible joints, for example BS 1, the rotation capacity is only equal to 6 x 10-2 rad,
see figure 4, when for the adequate unfilled steel joint it is much larger and equal at least
to 20 x 10-2 rad.
d. After the opening the RHS it was observed very good condition of the concrete. It was not
worse then the similar concrete curing in more wet environment.
e. As it is shown, see formula (6), that the stiffness of filled joints is about 30 % larger then
adequate steel joint. The test results show that such estimation could be accepted.
However, more tests are needed for confirmation of this data.
ACKNOWLEDGEMENT
NOTATION
REFERENCES
(1) prEN 1994-1-1 "Design of composite steel and concrete structures". Part 1.1 "General
rules and rules for buildings", CEN, 18 March 2002.
(2) Szlendak J. Improve the joints strength in steel frames with RHS Columns by concrete
filling. Proc. Int. Conf. on “Steel Structures of the 2000’s”, Istanbul, 2000, pp.345-352.
(3) Szlendak J., Bródka J. "Investigation into the static strength of welded T moment
unreinforced joints in rectangular hollow sections“, International Institute of Welding,
IIW-Doc. XV 538-83, March 1982.
(4) Szlendak J.K.: Design models of welded joints in steel structures with rectangular
hollow sections. DSc thesis, Bialystok Technical University Press, 2004 (in Polish)
(5) prEN 1993-1-8 Eurocode 3: Design of steel structures: Part 1.8: Design of joints, CEN,
31 January 2003 (Stage 34 draft).
ABSTRACT
Nominally pinned joints where longitudinal plates are welded to the walls of
RHS column are studied herein. The earlier results of such experiments made
by Jarrett, where tension loading is applied to the joints, are reminded. A
design formula for prediction the strength of joints, more optimistic, than one
given by Jarrett is proposed. Results of experimental tests, where connections
are loaded by the shear forces and secondary bending moments, are further
discussed. Three types of joint failure were observed in tests. Models for these
joint failures are given. For inelastic failure of chord face, proposed formulas
for prediction the strength and stiffness of the joint have not reference to EC-3.
INTRODUCTION
Braced frameworks with nominally pinned joints could be the proper solution for the saving of
labour, manufacturing and erection costs. If in such structures the columns are made with
box section (RHS), and the beam is the I - type profile, one of the easiest joint between them
is the longitudinal plate welded to column, figure 1. This plate usually is connected with web
of beam by using a bolted connection. Such joints are given in Table 7.13 (1). It ought to be
remember that they are very flexible (2, 3) and sensitive on the cracks in the tension area.
So, they should be rather used for the structures with predominantly static loading. Such
joints are usually loaded by shear forces and secondary bending moments. Furthermore,
they should be also able to carry the incidental tension load (2). If a carrying capacity for
such loading is too small the failure of joint, as in figure 2, is available.
When the semi-continuous framing is considered, than the design formulas to predict the
strength, stiffness and rotation capacity of joints ought to be developed. From the reason of
complex behaviour of described here joints such formulas are often semi-empirical.
Test results
Tests of joints, where the tension load Nt is applied from the I beam web by the steel
longitudinal gusset plate to the wall of RHS, are rather rare. One of it as in figure 2, for 13
joints, has been undertaken by Jarrett (2). In figure 3 results of these experiments for two
joints sign 6T and 12T are shown. For joint 6T n = 0 but for 12T n = 0,52. It is easy to noticed
that the prestressing of chord n = No/Aofyo decreased the ultimate load of joint. It could be
also seen that such joints have a significant “overstrength”, which increases for large
deformations of the loaded flanges of joint, see figures 2 and 3.
Prediction of the joint strength is given from the yield line bending-squash mechanism, which
occurs on the face walls of the RHS. In this mechanism the initial energy is dissipated not
only by the bending moments but also by the membrane forces in the yield lines, see
Groeneveld (4), Szlendak and Brodka (5) and Szlendak (6).
250
Joint 6T ( n=0)
200
Joint 12T ( n = 0,52)
150
jt,ult
t
N jt from (10)
e
N
for joint 6T
100
50
t
N jt from (10) for
joint 12T
0
0 5 10 15 20 25 30 35 40 45
Deformation of loaded wall "w" [mm]
Figure 3. Experimental ultimate load Nejt, ult as a function of (w) 6T and 12T, (2) and
theoretical strength Ntjt estimated from (10).
(
N t jt = 4m pl η C n + 2 Dn 2(1 − n 2 ) ) (1)
where coefficients Cn and Dn are calculated as below:
Dn = 3 (3)
G 2 (1 − β )
2.5
G2 = 1 + ( 4)
(1 − β ) 2 λo
2
but when
2λo − 1.5 17λo − 10
≤β<
2λo 17λo
1 + 2 B(1 − β )λo
Cn = (5)
(1 + B )(1 − β )
1
1 + B(1 − β )λo +
3B(1 − β )λo
Dn = ( 6)
(1 + B )(1 − β )
1 + 1 + (1 − β )λo [17(1 − β )λo − 10]
B= ( 7)
(1 − β )λo [17(1 − β )λo − 10]
Furthermore, when
17λo − 10
< β ≤1
17λo
Cn = 2λo , Dn = λo (8)
When the pure bending model occurs, what is the typical assumption (1), the particular
solution is obtained for which Cn is calculated as below:
2
Cn = ( 9)
1− β
If a deformation of loaded wall “w” arises then the strength of the joint increases due to the
membrane effect. As a good approximation could be assumed that for deformations such
that w < to/2, the coefficient increasing the strength of the joint due to the membrane effect is
equal to (1+4(w/to)2), (4). If one assumes that the maximum deformations of the loaded walls
is equal to 1% bo, (3), this coefficient could be written as (1 + 4 (0,01bo/to)2) = (1 + 0,0004
λo2). Then, more optimistic estimation of the joint strength is suggested. Proposed below
formula is the extension of formula (1) and includes the increasing coefficient from the
membrane effect:
[ ](
N t jt = 4m pl η C n + 2 Dn 2(1 − n 2 ) 1 + 0,0004 λ o
2
) (10)
In figure 4 the comparison between this theoretical estimation and the experimental strength
of joints tested by Jarrett (2), is shown. The experimental strength Nejt,pl, for the moment of
creating the yield line mechanism and ultimate strength Nejt,ult, for large deformations and full
membrane action is given for the joints 6T and 12T. As the illustration the results of ultimate
strength Nejt,ult for the other from 13 joints are also included.
120
tj,pl/mpl
100
e
Experimental results N
80
Jarrett (2) - ultimate load (13 joints)
0
0 20 40 60 80 100 120 140
t
Theoretical estimation N tj/mpl
Figure 4. Comparison: theoretical estimation (10) and the experimental strength from (2).
Test results
Test rig is shown in figure 5. Six joints were tested here in natural scale.
Eight LVDT gauges were used to measure the displacements and the rotations of each
gusset plate. Registrations of the results were made permanently (one registration per one
second) during the full loading and unloading process up to failure. After each loading step
the joint was unloaded to measure the permanent deformations of the tested specimen.
Three types of joint failure, as described below, were observed in tests.
40 Ø17
35
25
Flange yield failure occurred for specimens’ 3D/3 and 2D/3. This failure occurs when walls of
section are slender, here λo = 30. Inelastic mechanism arises on the loaded flanges and
sufficient membrane action was noticed, see figure 8. Large permanent deformations were
observed during the unloading process. In figure 10 the examples of moment - rotation
curves are shown. They are obtained for three of the gusset plates of joint 3D/3.
13
12
11
10
9
Moment kNm
0
0 25 50 75 100 125 150 175 200 225
Local rotation x mrad
Figure 9. Moment-rotation curves (bearing of gusset plate material and welds cracking).
6
Moment kNm
5
Results for plate nr 1
4 Results for plate nr 3
Results for plate nr 4
3
Initial stiffness (19)
2 Secant stiffness (20)
Design load (16)
1
0
0 50 100 150 200 250 300
Local rotation x mrad
Strength prediction
Formulas for design resistance for individual fasteners subjected to shear are shown in Table
1, (1). For Category A (bearing type) connections the below relations occur:
Fv,Ed ≤ Fv,Rd
Fv,Ed ≤ Fb,Rd
where,
α v f ub A
Fv,Rd = (11)
γ M2
For strength grades 4.6, 5.6 and 8.8: αv = 0,6
and,
k1 a b f u d t
Fb,Rd = (12)
γ M2
f ub
where αb is the smaller of αd; or 1,0;
fu
Model 2: Welds failure
According the formula (4.1) part 4.5 (1), the resistance of the fillet weld will be sufficient if the
following are both satisfied:
[σ┴2 + 3 (τ┴2 + τ║2)] 0,5 ≤ fu / (βw γM2 ) and σ┴ ≤ fu / γM2 (13)
Estimation of strength the RHS joints with longitudinal gusset plates is included, loaded by
normal forces, is given in EC-3 (see Table 7.13 (1). However, up to now, the design formula
for such joints, loaded by shear forces and bending moments, is not proposed there.
Theoretical investigations into the static strength of RHS to RHS beam-column welded joints
were undertaken in earlier works see e.g. (3, 6). The new design formula, presented in (6),
could be used in case of joints studied here, when β < 0,2. Proposed formula has easy form
for the direct calculation by the designers as below:
f yo
M t jy , pl = k M f M ( β ,η , λo ) M n pl , y (14)
f yn
From different possible functions fM the power function is chosen. Then, the strength of joint
Mtjy,pl could be calculated from formula:
f yo
M t jy , pl = k M β y1η y 2 λo
y3
M n pl , y (15)
f yn
After the numerical simulations the following exponents were obtained: y1=1/6, y2=1/2, y3= -
4/3. For eliminating the false results the Chauvenet rule was used, see (6).
For assumption that the level of confidence will be 0,95 the coefficient kM = 27,5 was
obtained. Furthermore, when the coefficient γM5 = 1,1 ( in EC-3 (1) it is assumed γM5 = 1)
then the design value of the joints strength could be calculated as below:
βη 3 f yo n
M t
jy , pl = 25 6 M pl , y (16)
λo 8 f yn
Comparison of this estimation with 186 test results of the static strength of RHS to RHS
beam-column welded connections, collected in the data-bank (7), is given in (6). Results of
27 tests are ignored as false if level of confidence will be 0,95. This simplified estimation
good predicts the experimental results (6).
Initial stiffness
Initial stiffness Sj,ini is a coefficient in the linear function between the bending moment applied
to the joint and its local rotation (M = Sj,ini Φ).
The power function is assumed to predict the initial stiffness of the joints when β > 0,4, see
(6). Analysis of the influence of particular parameters leads to the following formula:
For assumption that the level of confidence will be 0,95 and when coefficient γM5 = 1,1 the
coefficient ks = 6 was obtained. Then, the design value of the joints initial stiffness could be
calculated as below:
However, for the longitudinal gusset plate joints parameter β should follow the condition β <
0,2; see Table 7.13 (1). Even if the thickness of gusset plates changed the value of β is still
very small and influence of that parameter on the initial stiffness of joint is negligible. So, in
the formula (18) it is assumed that β2 = 0,04. Therefore for 0,03 < β < 0,2 the initial stiffness
of longitudinal gusset plate joints could be calculated as below:
Secant stiffness
According the recommendations which are given in EC-3, see part 5.1.2 (1), as a
simplification, the rotational, secant stiffness may be taken as Sj,ini /η in the analysis for all
values of the design moment. From Table 5.2 (1) the stiffness modification coefficient is
equal to η = 2. Therefore for 0,03 < β < 0,2 the secant stiffness of longitudinal gusset plate
joints could be calculated as below:
3,0
2,5
2,0
Moment kNm
1,5
Results for plate nr 1
Results foir plate nr 3
1,0
Results for plate nr 4
Initial stiffness (19)
0,5 Secant stiffness (20)
Design load (16)
0,0
0 5 10 15 20 25 30 35 40 45 50
Local rotation x mrad
a. New design formula (10) for joints loaded by tension load Nt, is suggested. Strength of
the joint increases due to the membrane effect. Proposed estimation is more optimistic,
than one given by Jarrett (2) and in Table 7.13 EC 3 (1). For compact RHS, when λo = 10
this strengthening is negligible and equal to 4%. However for slender wall RHS, when λo
= 30 it is significant and equal to 36%.
b. Joints loaded by shear load and by additional bending moment failed from: bearing of
plate material, shear of bolts and welds cracking. However, the yield failure of loaded
flanges was specially research in this paper. From figure 9 and 10 could be easy noticed
that the rotation of joints is large enough to reach the serviceability limit state. It is
reminded that for typical load the rotation of beam supports about 15 mrad leads to
exceeding this limit. Deflection of beam could be calculated with including the initial
stiffness of beam supports, which is given by (19).
c. For ultimate limit state the minimum strength of joints calculated from formula (11-13)
and (16) ought to be checked. Moreover, when the semi-continuous framing is
considered the secant stiffness of joints given by the formula (20) could be included.
Design model of such joint is given in figure 11.
NOTATION
ACKNOWLEDGEMENT
REFERENCES
ABSTRACT
INTRODUCTION
This paper presents an external reinforcement scheme, termed a collar, for strengthening
circular hollow section (CHS) joints. Choo et al. (1) first investigated plate reinforcement
schemes which may be used for field installation of auxiliary structures for offshore
structures. The collar plate may be suitable to provide reinforcement to a pre-fabricated joint
that is found to be under-designed. This concept may also find potential applications for
reinforcing joints in older offshore platforms and large span structures.
Fig. 1 illustrates the schematic arrangement for the collar reinforcement for a X-joint which
may be found to be under-strength. In the figure, the collar plate reinforcement, assumed to
be square in this case, is shown to be placed outside the foot-print of the brace-chord
intersection, with thickness tc and length lc. The usual notations for the outside diameter and
wall thickness of the brace (d1 and t1) and chord (d0 and t0), and associated geometric ratios
are also indicated. The details 1 and 2, with additional weld shown hatched in Fig. 1, are
indicative welding arrangement to connect the collar plate to the brace and chord. The
edges of collar plate can be profiled to accommodate the existing full penetration weld at the
brace-chord intersection.
For the externally placed collar reinforcement plate which may be bent to be compatible with
the chord curvature, Fig. 2 shows three possible schemes: 4 parts, 2 parts (parallel) and 2
parts (perpendicular). For the 4 part scheme, for example, the solid lines indicated in Fig. 2
denote the lines of weld connecting the collar plate to the brace and chord. Indicative weld
details shown in Fig. 1 (details 1 and 2) can be sized appropriately for the design
requirements. For a joint loaded predominantly by in-plane bending, the 2 parts (parallel)
arrangement may be an option if the welding requirement needs to be minimized. For a joint
loaded predominantly by brace axial load or out-of-plane bending, the 2 parts
(perpendicular) arrangement may be considered. For collar plates with large lc/d1 ratio,
additional slot welds may be placed within the boundaries to provide supplementary ties
between the collar plate and the chord.
t1 β = d1/d0 τc = tc/t0 t1
2γ = d0/t0 lc/d1
lc 1
tc
t0
2
lc
d0
d0 collar chord
plate brace
Detail 1 Detail 2
Figure 1. Collar plate reinforced CHS X-joint.
A B
This paper presents results of numerical studies on the behaviour of CHS T- and X-joints
with collar plate reinforcement. The accuracy of the numerical results is verified against the
T-joint tests reported by Choo et al. (1, 2). The results show that significant strength
enhancement for collar reinforced joints can be achieved through proper proportioning of the
reinforcement plate. Selected plots are presented to demonstrate the strength enhancement
of X-joints under brace axial compression, in-plane and out-of-plane moments.
Choo et al. (1) presented results from an experimental programme investigating the strength
enhancement to a simple T-joint by provision of reinforcement around the intersection
region, in the form of a doubler plate or a collar plate. The experimental programme
consisted of eight tests with brace axial load with four pairs of tests, each pair with brace
compression and tension. The chord length was chosen such that joint failure occurred prior
to chord member failure, with particular reference to recommendations by Zettlemoyer (3).
In this paper, the experimental result for the collar reinforced T-joint specimen EX-03 and the
calibration of the nonlinear finite element model are provided for illustration. Details can be
referenced in our papers (2, 4).
For a particular joint subjected to given loading, an analyst can consider the appropriate
symmetry in geometry, loading and boundary conditions to determine the finite element (FE)
model for analysis. For a X-joint subjected to brace axial load, only one-eighth of the joint
modelled (as shown in Fig. 3) with appropriate symmetry conditions and load specification is
required. For each FE model, more refined mesh is generated where stress gradient is more
critical. The automatic mesh generator for reinforced joints in this study is an extension of
that presented by Qian et al. (5).
Y
Figure 3. FE model for one eighth of a collar plate reinforced CHS X-joint
2γ = 50.8, β = 0.64, lc/d1 = 1.50 and τc = 1.0.
For the present FE models, two layers of 20-noded solid elements, type C3D20R with
reduced integration in ABAQUS (6), are specified through the thickness of all members to
provide good description of possible non-linearity in the thickness direction. Depending on
the actual joint geometry, 500 to 1000 elements are created to represent one-eighth of a
whole joint. Such mesh density has been proven to be able to produce results with good
accuracy (7).
Weld geometries
As three-dimensional solid elements are used in the FE models, it is possible to simulate the
weld geometries with high accuracy. The actual geometric definition of the welds is included
in all FE models. The geometry of the penetration weld between the brace and the chord is
modelled following the American Welding Society (8) recommendations. The depth of the
fillet welds between the reinforcing plate and the chord surface is taken the same as the
thickness of the plate, with two layers of finite elements specified. The welds connecting the
collar plate and chord (along the chord circumferential or longitudinal directions) are not
explicitly modeled. These are reflected in the FE model by specifying the appropriate
spatially common nodes to be tied.
The geometrical non-linearity is included to predict possible buckling in the chord wall
through the NLGEOM parameter in the *STEP option in ABAQUS input file. The material
nonlinearity is specified using the “true stress” and associated logarithmic strain to define the
plasticity with isotropic hardening (6).
Contact interaction
When a collar plate reinforced joint is loaded, contact may occur between the bottom of the
collar parts and the chord outer surface. The contact interaction plays an important role in
the load transferring mechanism of plate reinforced joints and thus non-linear contact
analysis is required. Since both of the reinforcing plate and the chord wall are deformable
bodies, a deformable-deformable contact interaction was defined using a “master-slave”
algorithm in the numerical analysis (6).
Fig. 4a shows the cut-section of the collar-reinforced Specimen EX-03 after completion of
the test. It can be observed that the collar reinforcement has relocated the chord plastic
hinges away from the brace-chord intersection, and that the brace has deformed extensively
adjacent to the intersection. Fig. 4b shows the deformed shape predicted by the nonlinear
FE analysis, and very good agreement with the experimental result is observed.
a b
Figure 4. Comparison between test and FE results, (a) Cut-section of EX-03 after test, and
(b) FE prediction.
The load-ovalisation curves (in which ovalisation at particular load level is based on the
change in diameter of the chord section) for Specimen EX-03 are shown in Fig. 5. The
numerical prediction is found to correspond very closely with the experimental curve, and
this serves to verify the accuracy of the numerical method.
Programme set-up
Parametric studies to investigate the static strength of collar plate reinforced X-joints have
been conducted by the authors. The chord diameter of all joints was taken as do=508 mm,
with β varying from 0.25 to 0.80 (β= 0.25, 0.43, 0.64 and 0.80), α = 12, and 2γ= 31.8 and
50.8. The brace-to-chord thickness ratio τ =1.0, and the brace length was kept at 4d1. The
thickness of the reinforcing plate was assumed equal or larger than the chord wall thickness
t0. For each combination of 2γ and β ratios, three values of plate thickness parameter
(τc=1.00, 1.25 and 1.60) and five values of plate length parameter (lc /d1=1.25, 1.5, 2.0, 2.5
and 3.0) have been considered. The corresponding un-reinforced joints were also included
to provide the appropriate reference strength. A total of 8 un-reinforced joints and 120 collar
The un-bent collar plate was assumed to be square in shape, except for large β cases,
where the plate width exceeded half the perimeter of the chord section, and for this case, the
plate width was limited to half the chord perimeter with welds along its edges.
In the following sections, selected results shown for the various loading conditions are
focussed on joints with 2γ = 50.8 and β = 0.25 and 0.64.
400
300
Load [kN]
200
100
0
0 20 40 60 80 100 120
Ovalisation [mm]
Figure 5. Experimental and numerical load-ovalisation curves for EX-03.
Fig. 6a to 6b show the deformed shapes of two collar reinforced X-joints subjected to axial
brace compression. Due to the weld at the brace-chord intersection, and the collar-chord
segments along the longitudinal (crown) and circumferential (saddle) segments, the collar
plate is effective in stiffening the chord and enhancing the load transfer from the brace.
β = 0.25 β = 0.64
τ = 1.00 τ = 1.00
lc = 2.00d1 lc = 2.00d1
τc = 1.00 τc = 1.00
Figure 6. Deformed shapes of collar reinforced X-joints with lc=2.0d1 and 2γ=50.8 with
(a) β = 0.25, (b) β = 0.64.
30
10
F/fy0*t02
F/fy0*t02
20
5
10
0 0
0.00 0.02 0.04 0.06 0.08 0.00 0.02 0.04 0.06 0.08
δ/d0 δ/d0
Figure 7. Normalised load-indentation curves for collar reinforced X-joints with different plate
width to brace diameter ratios (a) β = 0.25 (b) β = 0.64.
The deformation limit proposed by Yura et al. (9), which is defined as 60fyd1/E, is adopted to
determine the ultimate strength of a joint without a pronounced peak value in the load-
displacement curve. It is noted that the collar plate reinforcement can provide substantial
strength enhancement to the joint.
Fig. 8a and 8b present the strength enhancement due to provision of collar plate for joints
with 2γ=50.8 and β=0.25 and 0.64, with the corresponding un-reinforced joint strength as
reference strength. Each of the strength ratios is plotted against the plate parameters τc and
lc/d1 in a three-dimensional diagram for each β. As noted in Fig. 8b, the reinforced joint
strength, obtained by the provision of an appropriately dimensioned collar plate, can be up to
3 times of the strength of an un-reinforced joint. The strength of a collar plate reinforced joint
may be improved either by increasing the collar plate length or by using a thicker plate. For
joints with small values of lc/d1, the effect of the plate thickness is insignificant. The effect of
plate thickness becomes more important as the collar plate length increases.
In this section, the failure mechanisms for un-reinforced and collar reinforced X-joints under
in-plane bending are presented to highlight the differences. The geometric parameters of the
joints considered are 2γ = 50.8 and β = 0.25 and 0.64. More details are reported by Choo et
al. (10).
2.2 3.5
2.0 3.0
1.8
2.5
Fu,c /Fu,u
Fu,c /Fu,u
1.6
2.0
1.4 3.0 3.0
1.2 2.5 1.5 2.5
1
1
ld /d
lc /d
1.0 1.0
1.2 1.5 1.2 1.5
1.4 1.25 1.4 1.25
τc 1.6 τc 1.6
Figure 8. The effects of τd and ld/d1 on the strength of axially loaded collar plate reinforced
X-joints with 2γ = 50.8 (a) β = 0.25 (b) β = 0.64.
Failure mechanisms
Fig. 9a and 9d show the deformed shapes of collar plate reinforced joints with different
combination of β and lc/d1. The collar plate reinforced joint is observed to fail with relatively
large plastic zones formed near the brace-chord intersection. Because of the welds between
the collar plate parts and the chord surface parallel to the chord axis, the collar plate acts
closely with the chord wall on both compressive and tensile sides.
For joints with short collar plates (Fig. 9a and 9c), plastic hinges are observed near the
welds between the collar plate and the chord. The strength enhancement due to the short
collar plate may be regarded as an equivalent increase in β. No obvious plastic hinge is
found for a joint with long collar plates (Fig. 9b and 9d).
a β = 0.25 c β = 0.64
lc = 1.25d1 lc = 1.25d1
b β = 0.25 d β = 0.64
lc = 2.00d1 lc = 2.00d1
Figure 9. Deformed shapes of collar plate reinforced X-joints under in-plane bending.
Fig. 10a and 10b present the strength enhancement due to provision of collar plate for joints
with 2γ=50.8 and β=0.25 and 0.64. As noted in Fig. 10b, the reinforced joint strength,
obtained by the provision of an appropriately dimensioned collar plate can be up to 2.8 times
of the strength of an un-reinforced joint. For joints with small values of lc/d1, the effect of the
plate thickness is insignificant. The effect of plate thickness becomes more important as the
collar plate length increases and more deformation of the collar plate takes place.
3.0 3.0
2.5 2.5
Mi,u,c /Mi,u,u
Mi,u,c /Mi,u,u
2.0 2.0
3.0 3.0
1.5 2.5 1.5 2.5
1
1
ld /d
lc/d
1.0 1.0
1.2 1.5 1.2 1.5
1.4 1.25 1.4 1.25
τc 1.6 τc 1.6
Figure 10. The effects of τc and lc /d1 on the strength of collar plate reinforced X-joints under
IPB with 2γ = 50.8 (a) β = 0.25 (b) β = 0.64.
Failure mechanisms
Fig. 11a and 11b show the deformed shapes of collar plate reinforced joints loaded by out-
of- plane bending. It can be observed that the weld connecting the collar plate to the chord,
from the saddle positions along the chord circumferential direction is effective in transferring
the brace moment.
Figure 11. Deformed shapes of collar plate reinforced X-joints under out-of-plane bending
with 2γ = 50.8 (a) β = 0.25 (b) β = 0.64.
Fig. 12a and 12b show the potential strength enhancement for collar plate reinforced X-
joints. It can be seen that the strength ratio of the reinforced joint to the corresponding un-
reinforced joint varies from 1.6 to 3.6. The plate thickness parameter τc and length parameter
lc /d1 have significant effects on the strength of the reinforced joints for cases with large lc/d1
ratios. Equivalent strength enhancement can be obtained by either increasing the plate
length or by using a thicker collar plate.
4.0 4.0
3.0 3.0
Mo,u,c /Mo,u,u
Mo,u,c /Mo,u,u
2.0 3.0 2.0 3.0
2.5 2.5
1
1
ld /d
lc/d
1.0 1.0
1.2 1.5 1.2 1.5
1.4 1.25 1.4 1.25
τc 1.6 τc 1.6
Figure 12. The effect of τd and ld/d1 on the strength of collar plate reinforced X-joints under
OPB with 2γ = 50.8 (a) β = 0.25 (b) β = 0.64.
Extensive numerical studies have been conducted to evaluate the behaviour of circular
hollow section (CHS) X-joint reinforced with a collar plate, subjected to axial brace
compression, in-plane bending or out-of-plane bending respectively. From the presented
results of un-reinforced and collar plate reinforced CHS T- and X-joints, the following may be
concluded:
1. The collar plate is an effective reinforcement scheme, and can improve the static strength
of CHS T- and X-joints considerably.
2. Each of the parameters: the brace-to-chord diameter ratio β, the plate-to-chord wall
thickness ratio τc, and the plate length-to-brace diameter ratio lc/d1 have significant
influence on the strength of collar plate reinforced joints.
3. For a reinforced joint with fixed brace and chord dimensions, equivalent strength
enhancement can be obtained by either appropriately increasing the plate length or using
a thicker reinforcement plate.
Based on the present studies, the following are possible recommendations for future
research on the collar plate reinforced joints:
1. Since only part of the geometric parameters and loading conditions have been covered in
the current study, more extensive parametric studies will provide a comprehensive
understanding of collar plate reinforced joints.
ACKNOWLEDGEMENTS
REFERENCES
1. Choo, Y.S., B.H. Li, G.J. van der Vegte, N. Zettlemoyer & J.Y.R. Liew (1998). Static
strength of T-joints reinforced with doubler plate or collar plate. Tubular Structures VIII:
Proceedings Eighth International Symposium on Tubular Structures, Singapore, pp. 139-
145.
2. Choo, Y.S., G.J. van der Vegte, B.H. Li, N. Zettlemoyer & J.Y.R. Liew (2005). Static
strength of T-joints reinforced with doubler or collar plates - Part I: Experimental
investigations. Journal of Structural Engineering, ASCE, Vol. 131, No. 1, pp. 119-128.
3. Zettlemoyer, N. (1988). Developments in ultimate strength technology for simple tubular
joints. Proc. Offshore Tubular Joints Conference (OTJ’88), Surrey, UK.
4. van der Vegte, G.J., Y.S. Choo, J.X. Liang, N. Zettlemoyer and J.Y.R.Liew (2004). Static
strength of T-joints reinforced with doubler or collar plates - Part II: Numerical
simulations. Journal of Structural Engineering, ASCE (accepted for publication).
5. Qian X.D., Romeijn A., Wardenier J. and Choo Y.S. (2002). An automatic FE mesh
generator for CHS tubular joints. Proc. 12th International Offshore and Polar Engineering
Conference. Kita-Kyushu, Japan.
6. Abaqus/Standard User’s Manual Version 6.2 (2001). Hibbitt, Karlsson and Sorensen
Inc., Rhode Island, USA.
7. van der Vegte, G.J. (1995). The static strength of uniplanar and multiplanar tubular T-
and X-joints. PhD thesis. Delft University Press.
8. A.W.S. (1996). Structural Welding Code, AWS D1.1-96. American Welding Society Inc.,
Miami, USA.
9. Yura, J.A., N. Zettlemoyer & I.F. Edwards (1980). Ultimate capacity equations for tubular
joints. Proc. Offshore Technology Conference, Paper OTC 3690, Houston, U.S.A.
10. Choo Y.S., Liang J.X., van der Vegte G.J., Liew J.Y.R. (2004). Static strength of collar
plate reinforced CHS X-joints loaded by in-plane bending. Journal of Constructional
Steel Research, Vol. 60, No. 12, pp. 1745-1760.
ABSTRACT
In the framework of a larger programme to establish new chord load functions
for circular hollow section joints, this study evaluates the effects of various sets
of boundary conditions and chord pre-load on the static strength of axially
loaded gap K-joints. The influence of boundary conditions on the chord stress
contours is made clear for four different combinations of the geometric
parameters β and 2γ. It is concluded that a better understanding of the effects
of chord pre-stress on the strength of K-joints is obtained by considering the
maximum chord stress as the governing variable, instead of the chord stress
due to externally applied pre-loads.
INTRODUCTION
In current design rules, insufficient emphasis is put on the consistency of various design
equations. For circular hollow section (CHS) joints, the external chord “pre-load” (i.e. the
additional load in the chord which is not necessary to resist the horizontal components of the
brace forces) is used to account for the effects of chord loading. However, for rectangular
hollow section (RHS) joints, the chord stress formulation is based on the maximum chord
stress i.e. the stresses as a result of axial forces and (where applicable) bending moments.
As reported by van der Vegte and Makino (1), in the past, research into the effects of chord
pre-load on the strength of tubular joints was limited. Three experimental studies are
available in the literature regarding the effects of pre-load on the ultimate strength of tubular
X-joints. Although for CHS K-joints, the number of experiments outnumbers the data
available for uniplanar X-joints, in most of the K-joint tests, chord stress was simply a result
of horizontal equilibrium loads and was not meant as a prime variable. Only a few
researchers e.g. Kurobane and Makino (2) and de Koning and Wardenier (3) explicitly
applied a chord pre-load to the joints. More recent investigations into the effects of chord pre-
load on CHS joints were conducted by Dier and Lalani (4) and Pecknold et al (5, 6).
Since numerical tools offer the flexibility to vary various parameters and at the same time
Figure 1. Boundary conditions for K-joints used by van der Vegte et al (7).
Various researchers assessed the effects of boundary conditions on the ultimate strength of
uniplanar K-joints and came to the conclusion that the influence of restraints could be
significant. A brief overview of some of the investigations is presented in the section
hereafter. Additional FE analyses were conducted to evaluate the effects of various sets of
boundary conditions on the static strength of uniplanar K-joints subjected to chord pre-load.
The current study addresses the research programme, the FE strategy and the failure
criteria. Finally, the numerical results are presented as a function of either the chord pre-
stress due to external loads or the maximum chord stress.
In 1989, Connelly and Zettlemoyer (9) performed numerical research into the static
behaviour of various overlap and gap K-joint configurations. Each joint was analysed twice :
at first the K-joints were restrained and loaded in a manner consistent with laboratory tests
on such joints. In the second analysis, the K-joint models were mounted into a braced frame
with the load being applied directly to the frame instead of the K-joint. Connelly and
Zettlemoyer found that for this specific configuration with β = 1.0, the frame-mounted K-joints
showed axial capacities which were between 11 to 26 % higher compared to the isolated
joints. The authors suggested that, if possible, future tests on isolated joints should consider
a more accurate replication of the boundary conditions found in actual frames.
In 1992, Bolt et al (10) conducted numerical analyses on a single gap K-joint geometry using
different boundary conditions for chord and braces. Variations in capacities of up to 10 %
were observed while the post-peak load-deformation responses also varied significantly
among the cases considered. However, Bolt reported that confidential research on K-joint
configurations with different sets of geometric parameters suggested an even much greater
dependency of boundary conditions.
As part of a larger study into the behaviour of overlapped K-joints, Healy (11) performed
various numerical simulations to assess the effect of chord and brace end restraints on the
axial capacity. Healy evaluated two sets of boundary conditions often used in experiments.
Figure 2a illustrates “single” boundary conditions : one chord end is left free, while one brace
end reacts the load applied to the other. Figure 2b depicts the “double” set of boundary
conditions where both chord ends are restrained. Healy concluded that, when lateral
movements of the braces are restricted, the differences between both sets of boundary
conditions are negligible for the joints considered. On the other hand, Healy mentioned that
experiments carried out by Bjornoy (12) revealed a strong dependency of boundary
conditions on the ultimate strength of K-joints, especially for eccentric overlapped joints.
In 1998, Liu et al (14, 15) investigated the effect of boundary conditions and chord load on
the capacity of selected uniplanar and multiplanar RHS gap K-joints. Similar to the findings of
the previous researchers, the authors concluded that boundary conditions have a significant
influence on the ultimate strength of RHS K-joints.
N2 N1 α = 2l0/d0
t2 t1 β = d1/d0
d2 d1 2γ = d0/t0
g ξ = g/d0
t0 l2 θ2 θ1 l1
+Nop
d0 e
l0
The configuration of uniplanar K-joints and the definition of the geometric parameters are
shown in figure 3. The geometric parameters β and 2γ analysed in van der Vegte’s numerical
study (7) are summarized in Table 1. Although the programme of twelve K-joints considered
two values of the brace angle θ (45˚ and 60˚), only the six configurations referring to θ = 45˚
are presented in Table 1.
The current study focuses on the following four K-joint geometries shown in Table 1 : β =
0.48 and 0.67, 2γ = 25.4 and 63.5 (joints K3 to K6). The K-joints with β = 0.25 are not further
studied, as the influence of the horizontal reaction forces on the ultimate load is not as
pronounced as for the other configurations due to the relatively small failure loads of K-joints
Since the current programme is limited to K-joints with positive gap-values, for some
configurations, eccentricities have been introduced to avoid overlapped joints. For each of
the K-joints with eccentricities, the gap size g was taken as 32 mm, corresponding to t1 + t2
for the thick walled joints. The values of the gap ratio ξ (= g/d0), g/t0 and e/d0 are also
presented in Table 1. The non-dimensional chord length parameter α (= 2l0/d0) is held at 16,
whereas d0 = 406.4 mm. The length of the braces is set to 5d1. The steel grade used for the
tubular members is S355 with fy = 355 N/mm2 and fu = 510 N/mm2.
In models 1 and 2, the horizontal component of the brace forces causes compressive
stresses in the left side of the chord, while in model 4, tensile stresses occur in the right side
of the chord. In model 3, the horizontal components of the brace loads are distributed
between both chord ends.
In line with the analyses of the parametric study on K-joints, for each of the configurations
considered, nine values of the external chord pre-load N0p have been analysed, giving the
following chord pre-load ratios N0p/A0fy0 : +0.9, +0.8, +0.6, +0.3 , 0.0, -0.3, -0.6, -0.8, -0.9
(positive values refer to tensile pre-load).
FE modelling aspects
The numerical analyses were carried out with the finite element package ABAQUS (8). The
joints are modelled using twenty noded solid elements employing reduced integration
(ABAQUS element C3D20R). Two layers of elements are modelled through the thickness of
each member. Due to symmetry in geometry and loading, only one half of each joint has
been analysed. The appropriate boundary conditions are applied to the nodes located in the
plane of symmetry.
For all joints, the geometry of the welds at the brace-chord intersection has been modelled.
The dimensions of the welds in the numerical model are in accordance with the
specifications recommended by the AWS (16).
Both ends of the braces and the chord ends have rigid diaphragms. The length of the
members is considered to be sufficient to exclude any influence of the end caps on the static
response of the joints.
Since the incorporation of material- and geometric non-linearity in ABAQUS requires the use
of true stress-true strain relationships, the engineering stress-strain curve is modelled as a
multi-linear relationship and subsequently converted into a true stress-true strain relationship.
The hardening rule proposed by Ramberg-Osgood has been used to describe the true
stress-true strain behaviour after the peak stress in the engineering stress strain curve is
reached.
In order to validate the numerical model, comparisons were made with experimental
evidence. In 1981, de Koning and Wardenier (3) conducted a series of static tests on
uniplanar CHS K-joints. Out of this programme, two thin walled gap K-joints (β = 0.33 and
0.65, 2γ ≈ 56, θ = 45˚, e = 0) were chosen to serve as a basis for validation of the current FE
model. As described by van der Vegte et al (7), for both geometries good agreement was
observed between the numerical and experimental load-deformation responses, not only for
the initial stiffness but also for the peak load and the post-peak behaviour.
In the first step of the numerical analyses, the chord end is pre-loaded with uniformly
distributed axial forces using the load-control method. During this first step, the chord ends
are roller supported i.e. free to move laterally.
In the second step of the loading procedure, the appropriate boundary conditions are applied
to the chord ends whereas axial brace loads are applied to the nodes of the brace tip.
Meanwhile, the axial forces at the chord ends are maintained at the same level as at the end
of the first step.
In the second step of the joints modelled with the boundary conditions of model 1, the
displacement of the tip of the compression-loaded brace is prescribed, i.e. the joint is loaded
employing the displacement-control method. In the second step of the loading history of
models 2 to 4, equal but opposite loads are applied to both the compression and tension
braces, using the Riks algorithm, enabling to monitor the load-indentation behaviour for
declining brace loads.
Failure criteria
Ultimate load is defined as the force on the compression brace first exceeding one of the
following four failure criteria :
Figure 5 shows the chord stress contours obtained for K-joints K5 and K6 (β = 0.67)
modelled with various boundary conditions. Although not presented, the contours for joints
K3 and K4 with β = 0.48 look very similar. For each joint, the ultimate load is normalized by
the corresponding capacity for N0p = 0, where N0p refers to the externally applied chord pre-
load. The diagrams shown in figure 5a display the normalized capacities versus the chord
pre-stress ratio n’ = N0p/A0fy0, while the contours in figure 5b are based on the actual chord
stress ratio n, including the horizontal brace load components. Depending on the boundary
conditions shown in figure 4, n is defined as n’-2N1,u cos θ /A0fy0 for models 1 and 2 (left side
of the chord), while n = n’+2N1,u cos θ /A0fy0 for model 4 (right side of the chord).
Since the K-joints modelled with boundary conditions 3 are statically indeterminate, the chord
stresses can not be captured by a formula and are obtained from the FE analyses. For these
joints, figure 5b presents the chord stress contours for both sides of the chord.
As presented in the section hereafter, the differences between the ultimate capacities of
models 1 and 2 are small. To avoid overlap of the chord stress contours and to enhance
clarity of the diagrams, the curves for model 1 are not shown in the diagrams of figure 5b.
Comparing the chord stress contours of the joints modelled with boundary conditions 1 and
2, displayed in the graphs of figure 5a, it becomes clear that the differences are small.
A better understanding of the effects of chord stress for the joints modelled with boundary
conditions 2 is gained after examining the diagrams of figure 5b, where the horizontal axis
depicts the actual chord stress ratio. Because of the applied boundary conditions, each
contour is shifted in horizontal direction towards the compression side, whereas the amount
of transition is determined by the magnitude of the equilibrium loads. From these diagrams it
becomes clear that the joints of model 2 subjected to large compressive pre-loads fail by
chord member failure. Although not shown in these chord stress contours, the joints of model
1 exhibit similar failure behaviour.
1 1
f(n)
f(n')
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
n' n
Model 4 CHS K-joints Model 4 CHS K-joints
Model 3 β = 0.67 Model 3 - right side β = 0.67
Model 2 2γ = 63.5 Model 3 - left side 2γ = 63.5
Model 1 θ = 45 Model 2 θ = 45
1.2 1.2
1 1
f(n')
f(n)
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
n' n
(a) based on chord pre-stress ratio n’ (b) based on actual chord stress ratio n
Figure 5. Chord stress contours for K-joints K5 and K6 modelled
with various boundary conditions.
Comparing the chord stress contours of the joins modelled with boundary conditions 2 and 4,
it is found that for the joints under large tensile chord pre-loads, the ultimate capacities of the
joints of model 4 are significantly lower, caused by member failure of the chord. Chord
member failures are easily detected from the chord stress contours with the actual chord
stress being displayed on the horizontal axis (diagrams displayed in figure 5b).
For the K-joints modelled with boundary conditions 4 and pre-loaded by compression or
under zero pre-load, the ultimate capacities of the joints are higher than for the
corresponding joints of model 2. The unfavourable combination of compression pre-load and
compressive reaction forces leading to member failure (i.e. chord buckling or reaching the
squash load) observed for the joints of model 2, does not occur in the joints of model 4 under
compression pre-load. For the joints under zero pre-load and modelled with boundary
conditions 4, the gap area is still subjected to tensile stresses, giving a higher ultimate
strength in comparison with the joints where the gap area is loaded in compression, as is the
case for the joints of model 2.
The chord stress contours for the joints of model 3, based on the pre-load ratio n’, show that
the effect of chord pre-load on the ultimate capacity is much smaller than for the other
models. Especially for the joints pre-loaded by compression, the reduction in strength is
much less pronounced. This may be explained by considering the chord stress distribution as
a result of both chord pre-load and reaction forces. For this purpose, K-joint K5 (θ = 45˚, β =
0.67 and 2γ = 25.4) is examined in detail.
Remarks :
- The length of each arrow is proportional to the magnitude of the force
- Open block arrow : chord pre-load
- Black arrow : brace loads and chord reaction forces
Figure 6. External pre-load, brace loads and chord reaction forces of K-joint K5
modelled with boundary conditions 3.
In figure 6, the external chord pre-load and the reaction forces of K-joint K5 modelled with
boundary conditions model 3 are schematically illustrated for large compressive (n’ = -0.9)
and tensile (n’ = 0.9) chord pre-stress. In each of the diagrams, the length of the arrows is
proportional to the magnitude of the forces.
The loading procedure considers two steps. In the first step, the chord pre-load is applied
while the chord ends are free to move horizontally. The first row of figure 6 illustrates the
external forces applied to both chord ends. At the start of the second load step, in which the
brace loads are applied, the chord is restrained in horizontal direction, causing reaction
For both chord pre-load cases, the following three stages can be distinguished when the
braces are loaded up to failure :
STEP 2A : For small brace loads, the chord reaction forces at either side of the K-joint are
almost equal in magnitude, pointing in the direction opposite to the brace loads (see second
row of figure 6). For the K-joint under pre-compression, this leads to a further increase of
compressive chord stress at the left side of the chord, while the chord stress in the right side
reduces. For the K-joint under pre-tension, the tensile stress in the right side of the chord
becomes larger. This trend continues until the left chord side of the K-joint under pre-
compression and the right side of the K-joint under pre-tension come close to yield, resulting
in the initiation of load-redistribution.
STEP 2B : For the K-joint subjected to n’ = -0.9, further loading of the braces causes the
chord reaction force at the left end to decrease and reverse. For the K-joint subjected to n’ =
0.9, a similar remark can be made for the chord reaction force at the right chord end. The
reversal of the chord reaction forces is illustrated by step 2b in figure 6.
STEP 2C : Continuous loading of the braces up to failure will then increase all chord reaction
forces. Failure of the K-joints is visualized by the “ultimate load” data points in figure 7, which
are also part of the chord stress contours displayed in figure 5b.
4000
Ultimate load
N1 [kN]
3000
1000
0
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
chord load / A0 fy0
Figure 7. Chord load at either side of the chord for K-joint K5 under compressive (n’ = -0.9)
or tensile (n’ = 0.9) pre-load as a function of brace load.
For the thick walled K-joint (2γ = 25.4) under pre-compression, the chord is subjected to such
large reaction forces at failure that the resulting chord stress in the right side of the chord has
turned tensile. For the pre-tensioned joint, a similar redistribution is observed. These aspects
also become clear from the chord stress contours based on the actual chord stress, shown in
figure 5b. While the initial chord stresses due to the external loads vary between -0.9 fy0 and
+0.9 fy0, at failure, the actual chord stresses at both sides of the chord are significantly less,
thus explaining the less pronounced effect of external chord pre-load.
A comparison between the contours based on the chord pre-load ratio n’, shown in figure 5a,
reveals no relation between the curves of model 3 and the contours of models 2 and 4.
However, after looking at the diagrams in figure 5b based on the actual chord stress, it
becomes clear that the stress contours of the “compression” (left) side of the chord of model
3 are slightly below the contours obtained for model 2, while the chord stress contours of the
“tension” (right) side of the chord of model 3 are in good agreement with the stress contours
of model 4. This means that, when the actual chord stress is considered, the contours for
boundary conditions 3 are closely related to the contours of the two other models. This
clearly confirms the need to describe the chord stress effects as a function of the actual
chord stress rather than the stress due to externally applied chord loads.
CONCLUSIONS
Numerical analyses have been carried out into the strength of four axially loaded uniplanar
CHS gap K-joints subjected to axial chord pre-loading, with the main variables being the
geometric parameters β and 2γ and various sets of boundary conditions. Based on the
results of this selected set of K-joints, the following conclusions can be drawn :
a. In line with the data obtained for uniplanar X-joints, it is found that compressive chord
stresses have a detrimental effect on the ultimate capacity of axially loaded uniplanar K-
joints. For K-joints under tensile chord pre-load, the capacity of the joints either
increases or decreases compared to the ultimate strength of the corresponding joints
under zero pre-load, dependent on the value of β and 2γ, the amount of pre-load and
the boundary conditions.
b. The influence of boundary conditions on the ultimate capacity of K-joints can be
significant. The differences between the chord stress contours obtained for the joints
with the tensile brace end being pinned (model 1) and the joints where the tensile brace
is roller-supported to enable equal brace loads (model 2) are negligible. For K-joints
with both chord ends being pinned (model 3), the reduction in strength due to chord
pre-load diminishes, contrary to the behaviour exhibited by the joints modelled with the
other sets of boundary conditions, for which the strength reducing effects due to chord
pre-load are more pronounced.
c. The chord stress contours generated for K-joints clearly show that a better
understanding of the effects of chord pre-load is attained by considering the maximum
chord stress as the governing variable, rather than the chord stress due to external pre-
loads.
d. In future publications, the FE data generated in this study will be combined with
available data on other types of CHS and RHS joints and proposals for new chord
stress functions will be made.
ACKNOWLEDGEMENT
The first author would like to thank the Centennial Anniversary Foundation of the Faculty of
Engineering, Kumamoto University, Japan for the opportunity to carry out the research
reported herein.
(1) Vegte, G.J. van der and Makino, Y., (2001). The Effect of Chord Stresses on the Static
Strength of CHS X-Joints. Memoirs of the Faculty of Engineering, Kumamoto University,
Vol. 46, No. 1.
(2) Kurobane, Y. and Makino, Y., (1965). Local Stress in Tubular Truss Joints. Research
Report, Kyushu Branch of Architectural Institute of Japan, No. 4, pp. 75-80 (in
Japanese).
(3) Koning, C.H.M. de and Wardenier, J., (1981). The Static Strength of Welded CHS K-
Joints. TNO-IBBC Report BI-81-35/63.5.5470, Stevin Report 6-81-13, Delft, The
Netherlands.
(4) Dier, A.F. and Lalani, M., (1998). New Code Formulations for Tubular Joint Static
Strength. Proc. 8th International Symposium on Tubular Structures, Singapore, pp. 107-
116.
(5) Pecknold, D.A., Ha, C.C. and Mohr, W.C., (2000). Ultimate Strength of DT Tubular
Joints with Chord Preloads. Proc. 19th International Conference on Offshore Mechanics
and Arctic Engineering, New Orleans, U.S.A.
(6) Pecknold, D.A., Park, J.B. and Koppenhoefer, K.C., (2001). Ultimate Strength of Gap K
Tubular Joints with Chord Preloads. Proc. 20th International Conference on Offshore
Mechanics and Arctic Engineering, Rio de Janeiro, Brazil.
(7) Vegte, G.J. van der, Makino, Y. and Wardenier, J., (2002). The Effect of Chord Pre-load
on the Static Strength of Uniplanar Tubular K-joints. Proc. 12th International Offshore
and Polar Engineering Conference, Kitakyushu, Japan, Vol. IV, pp. 1-10.
(8) ABAQUS/Standard, (2000). Version 6.1, Hibbitt, Karlsson & Sorensen, U.S.A.
(9) Connelly, L.M. and Zettlemoyer, N., (1989). Frame Behaviour Effects on Tubular Joint
Capacity. Proc. 3rd International Symposium on Tubular Structures, Lappeenranta,
Finland, pp. 81-89.
(10) Bolt, H.M., Seyed-Kebari, H. and Ward, J.K., (1992). The Influence of Chord Length and
Boundary Conditions on K-Joint Capacity. Proc. 2nd International Offshore and Polar
Engineering Conference, San Francisco, U.S.A., Vol. IV, pp. 347-354.
(11) Healy, B.E., (1994). A Numerical Investigation into the Capacity of Overlapped Circular
K-joints. Proc. 6th International Symposium on Tubular Structures, Melbourne, Australia,
pp. 563-571.
(12) Bjornoy, O.H., (1993). Static Strength of Tubular Joints, Phase II, Analyses and Tests of
Gap and Overlap K-Joints. Veritec Report No 91-3393, AS Veritec.
(13) Dexter, E.M., Lee, M.M.K. and Kirkwood, M.G., (1994). Effect of Overlap on Strength of
K-joints in CHS Tubular Members. Proc. 6th International Symposium on Tubular
Structures, Melbourne, Australia, pp. 581-588.
(14) Liu, D.K., Yu, Y. and Wardenier, J., (1998). Effect of Boundary Conditions and Chord
Preload on the Strength of RHS Uniplanar Gap K-Joints. Proc. 8th International
Symposium on Tubular Structures, Singapore, pp. 223-230.
(15) Liu, D.K., Yu, Y. and Wardenier, J., (1998). Effect of Boundary Conditions and Chord
Preload on the Strength of RHS Multiplanar Gap K-Joints. Proc. 8th International
Symposium on Tubular Structures, Singapore, pp. 231-238.
(16) AWS American Welding Society, (1992). Structural Welding Code. AWS D1.1-92.
(17) Lu, L.H., Winkel, G.D. de, Yu, Y. and Wardenier, J., (1994). Deformation Limit for the
Ultimate Strength of Hollow Section Joints. Proc. 6th International Symposium on
Tubular Structures, Melbourne, Australia, pp. 341-347.
ABSTRACT
INTRODUCTION
Gusset plates can be found in almost any type of steel building. As hollow sections have
become more popular due to their exceptional properties in compression and torsion, the
combination of both gusset plates and hollow sections can be found in numerous
applications. These gusset plate connections can be used to splice hollow section members
or to connect web members to the chords in roof trusses (see Figure 1). Three possible
fabrication details are shown in Figure 2. Slotting the hollow section (Figures 2 (b) and (c))
is the most common version of this connection type. Various failure modes are possible
under tension loading with shear lag being one of them. Shear lag of the hollow section can
occur as the circumference of the hollow section is connected to the gusset plate only at two
points on opposing sides. The unconnected
circumference of the hollow section is not fully engaged
and contributes only in part to the resistance of the
member. In addition, local stress peaks at the slot ends
can cause initiating cracks that may result in an early
failure.
LITERATURE STUDY
40000
Of the failure modes that 90
can occur in gusset plate θ
connections, shear lag is 30000 0 180
one of the most ill-defined.
Strain [ µ m]
For tension loaded connections, Eurocode 3 (1) provides shear lag provisions that are only
applicable for bolted connections. For connections with welds, no specific design method is
provided in the Eurocode. Otherwise, shear lag is mentioned only in connection with locally
introduced shear loads causing bending moments in longitudinally stiffened plated
structures.
The North American specifications address shear lag for welded connections under tension
loading. Unfortunately, the American and Canadian specifications (AISC 2, CSA 3) differ in
their design methods for this limit state. In addition, changing formulae in old and new
specifications (e.g. AISC 4 versus AISC 5) indicate a lack of certainty with this connection
type. The Japanese specification (6) excludes shear lag by providing minimum connection
lengths. The following paragraphs briefly introduce the different design methods of the
various specifications.
D
with: x = for round hollow sections (2)
π
The effective net area is then calculated as:
Ae = An · U (3)
centre of gravity
of top half
w
tsl
x
D
tp
t
Lw wp
The current LRFD Specification (AISC 4, Equation B3-2) uses the gross area Ag of the
member to calculate the effective net area Ae (Ae = Ag · U) which can result in considerably
different design strengths for gusset plate connections where the hollow section is slotted
(An = Ag - 2tp·t). In practice, the slot width tsl is usually greater than tp to allow ease of
fabrication, and in such cases An = Ag - 2tsl·t.
Recently, a general examination of the AISC LRFD shear lag design provisions has been
made by Kirkham and Miller (8). Based on recent studies, it was concluded that the existing
design approaches are overly conservative and further research was necessary. The draft
version (5) of the upcoming AISC Specification in 2005 now uses the net area of the
member, An, in its formula but no longer has an upper limit of 0.9 for the eccentricity factor U
(see Equation 1). For round HSS with Lw ≥ 1.3D, the factor U becomes equal to unity.
Connection lengths Lw less than D are not covered.
The current Canadian Standard (CSA 3) addresses shear lag in elements connected by a
pair of welds parallel to the load by calculating the "effective net area" (Clause 12.3.3) based
on an efficiency factor that depends on the ratio of the distance between the welds around
the hollow section perimeter, w, and the connection length, Lw (see Figure 4). The efficiency
factor given is:
1.0 for Lw/w ≥ 2.0;
0.5 + 0.25 Lw/w for 2.0 > Lw/w ≥ 1.0;
0.75 Lw/w for Lw/w < 1.0.
A similar approach based on the former Canadian standard CAN/CSA-S16.1-94 (CSA 9), as
well as research done by Korol et al. (10), is given in the design guide for hollow structural
sections by Packer and Henderson (11). The recommended efficiency factor there is:
1.0 for Lw/w ≥ 2.0;
0.87 for 2.0 > Lw/w ≥ 1.5;
0.75 for 1.5 > Lw/w ≥ 1.0;
0.62 for 1.0 > Lw/w ≥ 0.6.
The Japanese recommendations for the design and fabrication of tubular truss structures in
steel (6) exclude shear lag by providing a minimum connection length of Lw ≥ 1.2D for gusset
plate connections. To account for uncertainties in fabrication of these connections, the
connection capacity is restricted to 90% of the unslotted (gross) member strength. Also, AIJ
avoid use of the net area of the slotted tube by means of a specific fabrication detail. Table
1 gives an overview of all the above shear lag provisions.
A number of studies on gusset plate connections have been carried out in the last few years.
The latest study, on a special type of gusset plate connection, the so-called hidden joint
connection, by Willibald (12) showed that shear lag was not critical for square HSS in this
specific connection type but can become critical for rectangular HSS. The results of the
parametric study supported the use of the American specifications (2, 4, 5) but indicated a
generally overly conservative approach in all current design methods. In an experimental
study by British Steel (Swinden Laboratories 13) on slotted end plate connections for
circular, square and rectangular hollow sections, 13 of the 24 specimens failed by shear lag.
The results of an experimental as well as numerical investigation on shear lag failure for
slotted circular hollow sections were given by Cheng et al. (14). Nine tests on gusset plate
connections to CHS tension members were performed, but none of the specimens failed by
shear lag. However, the experimental and numerical investigations showed that
considerable stress concentrations occur at the slot ends. Comparing the results of the study
with the then current Canadian (Cheng et al. 15) as well as American specifications (Cheng
and Kulak 16), it was shown that neither code accurately represented the behaviour of
slotted circular hollow section connections. In contrast to the specifications, Cheng and
Kulak (16) concluded that shear lag failure was not critical for round HSS if the connection
A study on shear lag in slotted square and rectangular hollow sections has been performed
by Korol et al. (10). A total of 18 specimens was tested under tensile loading with seven
specimens failing by shear lag. The authors concluded that for six of the seven specimens
that failed by shear lag, all with Lw/w ≈ 1.0, the connection capacity was nearly equal to the
tensile capacity of the hollow section, Nu = An · Fu. One specimen, where Lw/w = 0.61, failed
very prematurely due to shear lag. For specimens with Lw/w-ratios smaller than 0.6, base
metal shear failure of the hollow section governed. The influence of the eccentricity x on
the connection capacity was found to be only minor. Based on the results of the earlier
study, Korol (17) proposed a slightly modified approach for the calculation of the effective
shear lag net section area. Instead of using the efficiency factors as given in the Canadian
or American specifications, less conservative formulae were provided:
α = 1.0 for Lw/w ≥ 1.2 (net/gross section failure governs) (4a)
α = 0.4 + 0.5 Lw/w for 1.2 > Lw/w ≥ 0.6 (shear lag failure governs) (4b)
non-applicable for Lw/w < 0.6 (block shear tear-out governs).
The eccentricity factor U was then calculated by:
x
U = 1.0 – 0.4 L (5)
w
The effective shear lag net section was then given by:
Ae = An · α · U (6)
EXPERIMENTAL STUDY
Scope of testing
The experimental study comprised of six gusset plate connections for round hollow sections.
The specimens had varying fabrication details (see Figure 2): slotted versus unslotted HSS,
slot end welded versus no weld return. A further parameter in the test series was the weld
or connection length Lw with the Lw/w-ratio varying between 0.66 and 0.88 (with
w = 0.5 · π · D - tsl or w = 0.5 · π · D – tp). Standard cold-formed 168 x 4.8 mm Class C
hollow sections with a specified yield stress of 350 MPa (CSA 18) were used. The gusset
plate was made out of 1" (25.4 mm) Grade 300W steel. Table 2 shows the dimensional, and
Table 3 the material, properties of the tested specimens. The welds connecting the hollow
section and the gusset plate were standard 10 mm fillet welds using E480XX electrodes
(CSA 19). Each specimen was equipped with 10 linear strain gauges measuring the
longitudinal strain distribution on the hollow section (see Figure 5). The displacement of the
connection was measured by four LVDTs (Linear Variable Differential Transformers). The
results of these measurements were later on used to verify the numerical models of the
tested specimens.
22.5o
50 50 50 22.5o
8
7 22.5o
6 4 2
10 5 3 1 22.5o
9
Test results
Failure of all six specimens was caused by either shear lag (specimens 3, 4, 10, 11)), tear-
out of the HSS base material along the weld (specimen 2) or both failures taking place at the
same time (specimen 1, see Figure 6). Shear lag failure causes the HSS to fail circum-
ferentially while block shear tear-out happens along the weld. Table 4 shows the ultimate
connection strength, the failure mode as well as the predicted failure loads according to
current design methods. Generally, the specimens with the shorter connection lengths
(specimens 1, 2 and 10) had a reduced connection capacity. Before failure, all specimens
showed ovalization in the hollow section, especially pronounced in the specimens with the
unslotted tube and the slotted plate (see Figure 2 (a)). In all specimens, the strain gauge
readings showed very high stresses at the strain gauges closest to the weld at the beginning
of the connection (strain gauge 5, see Figures 3 and 5). The strain gauge furthest away
from the weld (strain gauge 8) reported either negligible or even negative strains (specimens
10 and 11) at ultimate load. The strain distribution along the weld and beyond could be
Considerable displacement took place in all tested specimens (see Figure 7). The
specimens having an unslotted tube and a slotted plate showed the largest deformation
before reaching their ultimate loads. Considering these sizable displacements, it might be
necessary to define a deformation limit which, for some connections, will then govern their
capacity. Having comparable connection lengths, specimens 1,2 and 10 as well as
specimens 3, 4 and 11 have only slightly different capacities. This indicates that the
fabrication detail only has a little influence on the connection strength. Currently most codes
do not specify the use of a certain detail but provide one design method to cover all three
cases. However, with increasing gusset plate thickness, the difference in connection
strength between the various fabrication details might be more pronounced. Generally, the
design methods found in the American specifications (AISC 4,5) show the best agreement
with the tests but further research seems necessary.
NUMERICAL STUDY
For further study of gusset plate connections, a numerical study has been started. The final
goal of this numerical study will be a parametric study concentrating on the influences of
several variables: weld length, hollow section diameter, hollow section wall thickness, the
eccentricity of the top or bottom part of the HSS, and fabrication details.
The Finite Element program ANSYS 5.7 (Swanson Analysis System Inc. 20) has been used
for the numerical study. A geometric and material non-linear analysis was performed for all
specimens. 8-noded, large strain solid elements (solid45) with reduced integration and
hourglass control were used throughout. The material properties were input as a multi-linear
curve with the engineering stress and strain converted to the true stress and strain values.
To simulate cracking in the models, a maximum equivalent plastic strain limit was used. The
so-called "birth and death" elements allow the user to significantly reduce the stiffness of the
elements, or "kill them", if an equivalent plastic strain is reached.
Due to the symmetry of the connection it was only necessary to model an eighth of the
connection (see Figure 8). The welds were fully modelled. A gap between the gusset plate
1400
1400
1200
1200
Connection Load [kN]
1000 1000
0 0
0 2 4 6 8 10 12 14 16 0 5000 10000 15000 20000 25000 30000
Displacement [mm] Strain [µm]
Figures 9a and 9b. Comparison between test and numerical results.
and the hollow section was modelled to prohibit any direct stress transfer between the plate
and the hollow section thus forcing load transfer to occur only via the welds. Symmetry
boundary conditions were employed along the planes of symmetry (translations normal to
the plane of symmetry were fixed) and the nodes at the HSS end were fixed. The finite
element models were then loaded by displacing the nodes at the end of the gusset plate.
Specimens 2 and 3, which have a slotted HSS with the slot end welded (fabrication detail
(c), see Figure 2) have been numerically modelled and show very good agreement with the
tests. The predicted ultimate loads for these two specimens are both within 2% of the actual
ultimate loads for each test (see Table 5). Figure 9a compares the load-displacement
curves for specimens 2 and 3 with the respective results of the numerical models. For both
specimens the agreement is very good up to peak load. Unfortunately, the numerical
models had problems converging beyond a certain point. At this load step, a high number of
elements are “killed” which causes a sudden change in stiffness thus causing severe
convergence problems. Yet, due to the high number of lost elements it is safe to assume
that the ultimate load of the FE-model has been reached and subsequent calculation steps
would result in a lower connection load. The comparison of the most critical strain gauge
(strain gauge 5, see Figure 5) also shows good agreement between test and FE model (see
CONCLUSIONS
An experimental, numerical as well as literature study on shear lag in round hollow section
gusset plate connections under tension loading has been carried out. The experimental
study showed that shear lag can indeed become critical in gusset plate connections. The
connection length had the largest effect on the connection capacity, whereas the fabrication
detail of the connection (see Figure 2) only had a minor influence on the capacity. For some
specimens, large displacements could be observed before failure, which could become
critical if deformations are restricted by a deformation limit.
The numerical study showed that it is possible to generate very good finite element models
of these connections. The numerical study will soon be extended to do further parametric
studies to finally provide suitable design methods against shear lag failure. The design
methods that can be found in current international specifications have been introduced in the
literature study and have been evaluated against the experimental research carried out. At
present, the American specification (AISC 5) seems to be best suited to design against
shear lag failure under quasi-static tension loading, but all design methods are overly
conservative and additional research is still necessary.
For the future, further research is currently planned on shear lag in gusset plate connections
under cyclic loading, as can be found in earthquake situations. With these connections,
special attention will be paid to the fabrication and refined connection details will be
considered.
ACKNOWLEDGEMENTS
Financial support for this project has been provided by CIDECT (Comité International pour le
Développement et l’Etude de la Construction Tubulaire) Programme 8G and NSERC
(Natural Sciences and Engineering Research Council of Canada). IPSCO Inc. and Walters
Inc. (Hamilton, Canada) generously donated steel material and fabrication services,
respectively.
NOTATION
REFERENCES
(1) Eurocode 3, (1993). Design of steel structures - General rules - Part 1-8: Design of
joints. Draft version. British Standards Institute, London, England.
(2) AISC, (2000). Load and Resistance Factor Design Specification for Steel Hollow
Structural Sections. American Institute of Steel Construction, Chicago, USA.
(3) CSA, (2001). Limit States Design of Steel Structures. CAN/CSA-S16-01. Canadian
Standards Association, Toronto, Canada.
(4) AISC, (1999). Load and Resistance Factor Design Specification for Structural Steel
Buildings. American Institute of Steel Construction, Chicago, USA.
(5) AISC, (2003). Specification for Structural Steel Buildings. Draft (December 1, 2003)
version of the forthcoming (2005) Specification. American Institute of Steel
Construction, Chicago, USA.
(6) AIJ, (2002). Recommendations for the Design and Fabrication of Tubular Truss
Structures in Steel. (in Japanese) Architectural Institute of Japan, Japan.
(7) Chesson E., Jr., and Munse, W.H. (1963). Riveted and bolted joints: Truss type
tensile connections. Journal of the Structural Division, ASCE, 89(1), 67-106.
(8) Kirkham, W.J., and Miller, T.H. (2000). Examination of AISC LRFD shear lag design
provisions. Engineering. Journal, AISC, 3rd Quarter, 83-98.
(9) CSA, (1994). Limit States Design of Steel Structures. CAN/CSA-S16.1-94. Canadian
Standards Association, Toronto, Canada.
(10) Korol, R.M., Mirza, F.A., and Mirza, M.Y. (1994). Investigation of shear lag in slotted
HSS tension members. Proceedings, 6th International Symposium on Tubular
Structures, Melbourne, Australia, 473-482.
(11) Packer, J. A., and Henderson, J. E. (1997). Hollow structural section connections and
trusses - A design guide. 2nd Ed., Canadian Institute of Steel Construction, Toronto,
Canada. ISBN: 0-88811-086-3.
(12) Willibald, S. (2003). Bolted Connections for rectangular hollow sections under tensile
loading. PhD thesis, University of Karlsruhe, Germany.
(13) Swinden Laboratories, (1992). Slotted end plate connections. Report No.
SL/HED/TN/22/-/92/D. British Steel Technical, Rotherham, England.
(14) Cheng, J.J.R., Kulak, G.L., and Khoo, H. (1996). Shear lag effect in slotted tubular
tension members. Proceedings, 1st CSCE Structural Specialty Conference,
Edmonton, Alberta, Canada, 1103-1114.
(15) Cheng, J.J.R., Kulak, G.L., and Khoo, H. (1998). Strength of slotted tubular tension
members. Canadian Journal of Civil Engineering, Vol. 25, 982-991.
(16) Cheng, J.J.R., and Kulak, G.L. (2000). Gusset plate connection to round HSS tension
members. Engineering Journal, AISC, 4th Quarter, 133-139.
(17) Korol, R.M. (1996). Shear lag in slotted HSS tension members. Canadian Journal of
Civil Engineering, Vol. 23, 1350-1354.
ABSTRACT
This note reviews and compares four sets of tubular connection design criteria
for axially loaded circular tubes. The four criteria are AWS D1.1, API RP2A,
ISO/WD 15-1.2, and ANSI/AISC 360-05.
INTRODUCTION
The existing American design codes for welded tubular connections are AWS D1.1 Structural
Welding Code (1990 thru 2002) and the substantially identical AISC Specification for the
Design of Steel HSS (1997), the basis of which is documented in the author’s book (1).
Three new sets of design criteria are in the works. They are:
(1) Proposed update to API RP2A, Design... of Fixed Offshore Platforms, based on research
conducted by Prof. Pecknold at the University of Illinois, and sponsored the API Offshore
Tubular Joint Research Consortium. A lengthy Commentary is included to self-document
these criteria. Extensive nonlinear finite element analyses were used to extend the
experimental data base, particularly in the areas of overlapped K-joints, a moment-free
baseline for T-joints, and chord stress interaction for a wide variety of joint types and
loadings. In view of reduced scatter compared to existing criteria, a reduced WSD safety
factor of 1.6 is proposed. This update has been approved for publication in the 22nd edition
of RP2A, and is in the final stages of editing.
(2) Static Strength Procedure for Welded Hollow Section Joints, IIW doc XV-E-03-279,
based on CIDECT research. This is also on the fast track to becoming an international
standard as ISO/WD 15-1.2, with IIW commission XV as secretariat. The immediate purpose
of this note was to provide comments for a Sept. 2003 meeting of IIW s/c XV-E.
(3) ANSI/AISC 360-05, Standard Specification for Structural Steel Buildings (draft of August
20, 2003), Chapter K, HSS Connections, being prepared by an ad hoc task group under the
direction of Larry Kloiber. This is essentially the same as the CIDECT-based IIW document,
except that it gives the characteristic ultimate strength without hiding a partial safety factor
therein. Separate safety factors are then given for LRFD and ASD.
Principal results of this review are shown in the Tables and Figures. Table 1 gives a side-by-
side tabulation of the design criteria for different types of circular joints. The square bracket
term is Qu in API, and simply written out in IIW and AISC. AWS criteria have been converted
to this format for comparison.
Allowable Pn sin θ = vp to L1 + 2 vw tw L2
In Figure 1, f1 and f2 shown for 45º N-joint with β < 0.5. These values were obtained
graphically from a scale layout of joints with varying degrees of overlap.
1.0
f1(q)
0.5
0.0
0.0 0.5 1.0 L2
q
L
1
1.0
q(βD)
f2(q)
0.5
0.0
0.0 0.5 1.0
q
Figure 1. Layout and functions f1 and f2 for overlapping 45º N-joints with β=0.5.
It may be noted from Table 1 that the AWS criteria cover a wider variety of design situations
than the others, with particular reference to the general multi-planar case. The proposed
AISC criteria cover the least, in a deliberate effort to minimize complexity.
Figures 2-5 show parametric comparison of the criteria for T&Y joints and X joints, which was
performed on an Excel spreadsheet. The base case is beta of 0.5, tau of 0.5, and gamma of
20. The upper plots show the effect of varying beta, with the other parameters kept at the
base case. The author’s 1969 and 1975 criteria were subsequently shown to be un-
conservative for X-joints, but they are very close to the latest OTJRC results for T&Y joints.
The lower plots show the effect of varying gamma; there is no effect if the T2Fy format tells
the whole story. There was no effect of varying tau in any of these cases.
40
30
Qu
20
10
0
0 0.2 0.4 0.6 0.8 1 1.2
20
15
Qu
10
0
0 5 10 15 20 25 30 35
40
30
Qu
20
10
0
0 0.2 0.4 0.6 0.8 1 1.2
20
15
Qu
10
0
0 5 10 15 20 25 30 35
γ
K & N joints
40
30
Qu
20
10
0
-1 -0.5 0 0.5 1
(overlap) g/D (gap)
Figure 6. Effect of gap/overlap for 45º N-joint with β=0.5, τ=0.5, and γ=20.
Figures 7 and 8 show the effect of varying the parameters β, γ, and τ at 60% overlap and
0.1D gap, respectively. Here we see that the proposed AISC (and IIW/CIDECT) criteria
completely miss the strong effect of tau in the overlap region, as predicted by the AWS
strength-of-materials approach and confirmed by the API finite element studies. They also
appear to under-predict the beneficial effect of large beta.
80
60
Qu
40
20
0
0 0.2 0.4 0.6 0.8 1 1.2
overlap joints
50
40
30
Qu
20
10
0
0 5 10 15 20 25 30 35
overlap joints
40
30
Qu
20
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
80
60
Qu
40
20
0
0 0.2 0.4 0.6 0.8 1 1.2
gap joints
30
20
Qu
10
0
0 5 10 15 20 25 30 35
gap joints
30
20
Qu
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
1.2
0.8
0.4
0
-1 -0.5 0 0.5 1
DISCUSSION
A. When comparing existing AWS-AISC criteria for circular tubular connections to CIDECT,
both in 1992 and today, neither criteria appear to have significantly different errors on the
unsafe side. Thus, one may ask the following questions:
“Why churn the Code by adopting essentially similar criteria but in a different format?”
“Why not look at new API results having a more significant impact on reliability?”
The issue is not simply whether or not to maintain the American status quo. It is important to
keep the Codes evergreen in the sense that they reflect the latest data, with researchers still
Q. There are obviously parameters that are treated very differently (tau overlapped and
overlap) and others that vary much less between the standards. Why?
A. Tau: All the criteria capture the effect of tau (t/T) in the same way, but without explicit
expression in point load criteria for T, Y, and X connections. This is not to say that tau is
unimportant; indeed, its primary importance was more obvious to the user in the old AWS
punching shear format. When one gets called in after the fact on structural failures and
tubular projects in trouble, one of the first things to look for is excessive tau ratios.
Tau effect in overlapped connections: In the old AWS-API-AISC criteria, this is captured by
mechanistic consideration of both punching at the partial footprint (L1) and shear in the
overlap weld (L2). In Pecknold’s new API criteria, this is based on an extensive inelastic
finite element parameter study, with totally separate Qg expressions for gap and overlap
joints. Pecknold’s indicated higher strength for large beta and large tau is also consistent
with the unexpectedly good performance of same-size overlapped K-bracing in Hurricane
Andrew. In CIDECT-IIW-ISO criteria, overlap is treated as an extension of the behaviour of
gap connections in which tau has no effect; a single Kg expression for both is curve-fit to the
smaller empirical test data base. This makes computerized design easier, but gives the
designer no insight into the physical mechanism of load transfer.
Effect of overlap amount: Both Pecknold (new API) and CIDECT agree that there is a
significant, but nearly constant, beneficial effect beyond g/D of -0.1. The old AWS-API
approach (dating back to 1975) was apparently intuitively appealing but wrong, while
remaining on the safe side for moderate amounts of overlap. Pecknold gives no equations
for |g/D| smaller than 0.05, and suggests interpolation in this region. In older codes, this was
a prohibited zone, due to concern over creating a weak hard spot and awkward welding
conditions. The smooth transition shown in the CIDECT strength criteria (and in Efthymiou’s
SCF criteria) may simply be an artefact of curve-fitting. This issue needs to be re-examined,
seeking data in the prohibited zone. Welded connections with tensile strains over 2% in
inelastic finite element solutions may be considered vulnerable to fracture.
Effect of chord loading: The 1972 AWS Code included a modest Qf penalty for compressive
chord loads, based on Japanese data. Existing AWS-AISC criteria reflect further effects of
gamma and chord load type, based on Yura’s X-joint data. CIDECT criteria are simpler, with
Kp close to being on the safe side of Yura. Pecknold’s criteria, based on extended finite
element results, reflect effects of both chord and brace load type (but not gamma). This is
reasonably consistent with CIDECT for K and N connections, but its better prediction for
other types of connections has a significant effect on reliability, prompting a modest reduction
in working stress design safety factor in API. One common design case in which both
CIDECT and AWS-AISC are significantly on the unsafe side is equal size X-braces with
equal but opposite loads and no joint can. A caveat for this case is urgently needed.
Q. Do we need more data to choose the right direction for the parameters that vary widely?
A. Pecknold’s API data base includes both the CIDECT physical tests and his extended
finite element results. It could be readily compared to CIDECT-IIW-ISO criteria, to quantify
the reliability consequences of adopting these into AISC. Gathering additional research data
must be left to the future. In order to meet its ambitious publication schedule, AISC should
select one set of criteria and stick with it – no mix-and-match. For circular connections at this
point in time, Pecknold’s criteria have not been as widely vetted as CIDECT, while CIDECT
does not have the extensive set of worked examples and familiarity to American designers
as the existing AWS-AISC criteria.
A. The differences are not always that great. D/t of 48 is quite typical for offshore structures,
so punching at the material shear strength rarely governs. Most CIDECT tests are in the D/t
range of standard weight structural tubing, which ranges from 6 to 34. European bridge
designers seem to favor the 3 to 13 D/t range of double extra strong. American designers
use D/t up to 120 for fabricated large diameter chord members, with much thicker joint cans
at the nodes. None of the design criteria show an adverse size or thickness effect on static
strength, although fracture mechanics and some of the tension test data suggest one. AWS
criteria give no bonus for tension.
A. People generally want to know the impact on cost and safety before they change a
design code. AISC should get a feel for this while re-working all the example problems and
tabulated results in the HSS manual. Bridges and offshore structures are also influenced by
fatigue, so the impact of a static strength change is muted.
CONCLUSION
ACKNOWLEDGEMENTS
The foregoing Q&A discussion was prompted by thoughtful review of an earlier draft of this
note from Tom Schlafly at AISC. Drafting of the figures has been performed by Dakang Liu
at TU Delft. The author is grateful to the conference organizers for accommodating this
paper.
REFERENCE
CONCLUSIONS
24 individual countries
56 attendees
EU 32
America (N&S) 16
Asia 5
Others 3
STRUCTURAL CONNECTIONS
conception
conception
design calculation
design calculation
code check
code check
process
fabrication
fabrication
erection
erection
maintenance
maintenance
1960
tubular connections
steel-to-concrete connections
composite connections
2004
1960
plastic design
non-linear analysis
numerical
calculations 3-D analysis
component method
1960
ISO ; IIW ; CEN
statistical calibration
detailed rules
2004
1960
gas shielded metal-arc welding
punching
off-site production
health & safety
requirements
friction grip bolting
injection bolts
weathering steel
environmental
requirements
stainless steel
aluminium
2004
fabrication
fabrication
tools
products
erection
erection
maintenance
maintenance
STRUCTURAL CONNECTIONS
conception
conception
fabrication
fabrication innovation
tools
products
erection
erection
maintenance
maintenance
- 3 Papers on the behaviour of the equivalent T-stub in tension (steel and aluminium)
All proposing methodologies for the assessment of this component
Special emphasis on ductility
under monotonic (2 papers) and
cyclic loading conditions (2 papers)
Despite the fact that much research has been carried out on bolted connections
in shear, the papers addressed several new topics.
- The influence of connection behaviour is a major topic of research. It was
shown that connection deformation may induce serviceability problems.
- The negative influence of hot dip galvanising and punching of holes on fatigue
behaviour can be overcome by applying friction grip bolts.
- A new type of bolt with a rounded head was investigated for use in repair and
renovation of historical buildings.
- A proposal was presented for a single harmonised equation for block shear in
coped beams.
- Tests on single plate connections were reported to investigate the influence of
moments in combination with shear.
- The influence of stiffness and ductility aspects for the distribution of forces in
bolted shear connections was demonstrated once again
- 1 Paper on partial joint penetration welding for connections under seismic loads
Experimental research project.
The possibility of brittle fracture was investigated and it was concluded that brittle
fracture is unlikely. Ultimate strength can be predicted by simple plastic analysis.
The papers provide not only information on topics not yet included in the
recommendations and codes but also cover topics which are still in
discussion for the revision of the IIW/Cidect recommendations.
STRUCTURAL CONNECTIONS
3
25 conception
conception
5
knowledge design calculation
design calculation 1
2 4
codes code check
code check education
1
1 fabrication
fabrication innovation
tools
products
erection
erection
1
maintenance
maintenance
1 1
fabrication
fabrication innovation
tools
products
erection
erection
1
maintenance
maintenance
conception
fabrication
fabrication innovation
tools
products
erection
erection 0
maintenance
maintenance
June 2004 ECCS/AISC workshop : Connections in Steel Structures V 23
A good connection
is better than a joint !
Zhang X. 211
Zupančič E. 131
Zygomalas M.D. 287
webplates 27
welded connections 37
welded joints 65
welded tubular connections 457
welding 381
welds 1, 177
wide flange column 211