Ej 2021 q1 Issue

Engineering
Journal
First Quarter 2021 | Volume 58, No. 1
Discussion
1 Investigation on the Performance of
a Mathematical Model to Analyze
Concentrically Braced Frame Beams
with V-Type Bracing Configurations
Charles W. Roeder, Dawn E. Lehman, Qiyang
Tan, Jeffrey W. Berman, and Andrew D. Sen
11 AISC Provisions for Web Stability under

Local Compression Applied to HSS
Fei Wei and Jeffrey A. Packer
33 Critical Temperature of Axially Loaded Steel

Members with Wide-Flange Shapes
Exposed to Fire
Ana Sauca, Rachel Chicchi, Chao Zhang,
and Lisa Choe
45 Design for Local Member Shear at Brace and

Diagonal-Member Connections:
Full-Height and Chevron Gussets
Rafael Sabelli and Brandt Saxey
Engineering
Journal
American Institute of Steel Construction
Dedicated to the development and improvement of steel construction,

through the interchange of ideas, experiences, and data.
Editorial Staff
Editor Margaret A. Matthew, PE
Managing Editor Keith A. Grubb, SE, PE
Research Editor Judy Liu, PhD
Production Editor Erika Salisbury
Officers
Jack Klimp
Chairman
Stephen Knitter
Vice Chairman
Edward Seglias
Secretary/Legal Counsel
Charles J. Carter, SE, PE, PhD
President
Scott L. Melnick
Senior Vice President
Carly Hurd, CAE
Vice President
Lawrence F. Kruth, PE
Vice President
Brian Raff
Vice President
Mark W. Trimble, PE
Vice President
The articles contained herein are not intended to represent official attitudes,
recommendations or policies of the Institute. The Institute is not responsible
for any statements made or opinions expressed by contributors to this Journal.
The opinions of the authors herein do not represent an official position of the
Institute, and in every case the officially adopted publications of the Institute
will control and supersede any suggestions or modifications contained in any
articles herein.
The information presented herein is based on recognized engineering
principles and is for general information only. While it is believed to be
accurate, this information should not be applied to any specific application
without competent professional examination and verification by a licensed
professional engineer. Anyone making use of this information assumes all
liability arising from such use.
Manuscripts are welcomed, but publication cannot be guaranteed. All
manuscripts should be submitted in duplicate. Authors do not receive a
remuneration. Guidelines for authors are printed on the inside back cover.
Engineering Journal (ISSN 0013-8029) is published quarterly. Subscriptions:
Members: one subscription, $40 per year, included in dues; Additional Member
Subscriptions: $40 per year. Non-Members U.S.: $160 per year. Foreign (Canada
and Mexico): Members $80 per year. Non-Members $160 per year. Published
by the American Institute of Steel Construction at 130 E Randolph Street, Suite Subscriptions: [email protected], 312.670.2400
2000, Chicago, IL 60601.
Copyright 2021 by the American Institute of Steel Construction. All rights Archives: Search at aisc.org/ej. Article downloads
reserved. No part of this publication may be reproduced without written are free for current members and are avaialable for a
permission. The AISC logo is a registered trademark of AISC. nominal fee for non-members.
DISCUSSION
Investigation on the Performance of a Mathematical

Model to Analyze Concentrically Braced Frame
Beams with V-Type Bracing Configurations
Paper by ALIREZA ASGARI HADAD and PATRICK J. FORTNEY
(2020, 2nd Quarter)
Discussion by CHARLES W. ROEDER, DAWN E. LEHMAN, QIYANG TAN, JEFFREY W. BERMAN, and ANDREW D. SEN
The paper “Investigation on the Performance of a Mathematical Model to Analyze Concentrically Braced Frame Beams with V-Type Bracing
Configurations” (Hadad and Fortney, 2020) addresses a difficult problem that is not well understood by engineers. While that paper does not
specifically limit its application to seismic design, the work was based upon the AISC Seismic Provisions and chevron-braced frame seismic
behavior experiments. The authors of this discussion, hereafter referred to as the responders for clarity, have also studied chevron-configured
concentrically braced frames. That research forms the basis of this discussion. In particular, this discussion aims to clarify the AISC Seismic
Provisions and prior research results, which the responders believe is misinterpreted in the paper in question. Further, in the series of contin-
uum finite element analyses on a beam with a plate discussed in the paper, it is of note that the model’s boundary conditions, loading regime,
and results are simplifications of those for beams in chevron-configured or multistory X-braced frames under seismic loading. This discussion
addresses applicability of the paper to seismic design and does not apply to any other application.
T here are several aspects of the original paper that need

to be clarified. First, the findings of some of the cited
research is misstated. For example, the paper states that
Provisions, but it is important to recognize that these are
not the forces necessary to resist the design seismic lateral
force for the system; instead, they are an idealized set of
the yielding-beam mechanism is inferior to the “nomi- demands used in the capacity-based design process.
nally elastic” chevron-beam concept, but this is not cor- In seismic design of braced frames, the braces are initially
rect. Research by the responders (Sen et al., 2016) shows designed to resist reduced forces [forces corresponding to
that chevron-braced frames with yielding beams provide the design basis earthquake reduced by the seismic reduc-
better distribution of inelastic deformation over the build- tion factor, R, as determined by ASCE/SEI 7, Minimum
ing height than those with “nominally elastic” beams, and Design Loads and Other Associated Criteria for Buildings
they develop larger inelastic deformations prior to brace and Other Structures (ASCE, 2016)]. These reduced seis-
fracture. The expected brace forces for seismic design (i.e., mic design forces are expected to correspond to demands
Equations 1, 2, and 3 in the paper in question) are used as resulting from small, frequent earthquakes, and the struc-
the basis of the analyses as specified in the AISC Seismic ture is expected to remain elastic for that seismic hazard
level. For special concentrically braced frames (SCBFs),
collapse resistance of the structure is achieved by capacity-
based design and detailing for ductility to ensure the sys-
Charles W. Roeder, Professor, Department of Civil and Environmental Engi- tem can sustain the cyclic inelastic deformation demands
neering, University of Washington, Seattle, Wash. Email: [email protected] during large earthquakes. In such events, the braces are
(corresponding)
expected to buckle in compression, yield in tension, and
Dawn E. Lehman, Professor, Department of Civil and Environmental Engineer- sustain deterioration of their post-buckling compressive
ing, University of Washington, Seattle, Wash. Email: [email protected]
resistance at large inelastic deformations. Adjacent mem-
Qiyang Tan, Visiting Graduate Student, Department of Civil and Environmental
Engineering, University of Washington, and PhD Candidate, Harbin Institute of
bers, including the beams, columns, and gusset plates, are
Technology, Harbin, China. Email: [email protected] designed using capacity-based methods to ensure the braces
Jeffrey W. Berman, Professor, Department of Civil and Environmental Engi- develop their resistance and inelastic deformation capacity.
neering, University of Washington, Seattle, Wash. Email: [email protected] The capacity-based design concept is not intended to guar-
Andrew D. Sen, Assistant Professor, Department of Civil, Construction and antee that members or connections remain elastic during
Environmental Engineering, Marquette University, Milwaukee, Wis. Email: an earthquake; they are simply intended to ensure that the
[email protected]
ductile element (the brace in this case) can develop its full
deformation capacity by designing the adjacent members to
Paper No. 2018-07RD sustain their maximum demands. A primary issue with the
ISSN 0013-8029 ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 1

stress evaluations proposed in the paper is that they will not the welds joining a midspan gusset to the chevron beam
necessarily achieve this goal. should be designed for the in-plane shear, bending moment,
The main objective of capacity-based design in SCBFs and axial force from the gusset plate, but research funded
is to develop the inelastic deformation of the brace rather by AISC has shown that this can lead to premature weld
than to design for a set of fictitious forces. The gusset plates fracture, as illustrated in Figure 2 (Swatosh, 2015). Weld
are designed for the expected brace capacities because they fracture should be prevented in seismic design because it
are directly connected to the brace. However, when braces prevents the braces and the SCBF from developing their full
are oriented for out-of-plane buckling, the gusset plates inelastic capacity. Moreover, weld fracture in some gusset
deform inelastically due to brace buckling and, as a conse- plate connection configurations can also compromise the
quence, sustain significant inelastic deformation demands, vertical load-resisting system, as shown in Figure 2.
as shown in Figure 1. The actual local stresses developed in A second issue with the research presented in the paper
gusset plates after brace buckling are distributed unevenly is the numerical simulation of the chevron beam. A finite
and are much larger than those used in gusset-plate design element analysis of the beam and midspan plate was per-
calculations; as the authors recognize, this has been well formed with the goals of improving the understanding
established even in the absence of brace buckling (Richard, of the behavior of beams in chevron-braced frames and
1986). However, gusset plates that are very conservatively supporting their proposed stress distribution at the gus-
designed will still develop significant yielding because of set plate-to-chevron beam weld. There are several issues
this out-of-plane brace deformation demand (Lehman et al., with the approach. First, the boundary conditions of the
2008). Gusset plates sustain these demands because of the beam and other aspects of the modeling approach are not
inherent ductility of steel, but welds joining gusset plates described, and these will influence the demands at that
to beams and columns are vulnerable to tearing and frac- weld. It appears that the beam was simply supported, which
ture prior to brace fracture. Therefore, such welds should be is an idealized boundary condition and not representative of
designed to develop the plastic capacity of the gusset plate, an actual SCBF. Second, a set of idealized point loads were
not the expected brace forces, because inelastic action in applied to the gusset plate. However, the demand resulting
the gusset plate is required to accommodate the deforma- from cyclic action on an SCBF does not result in a single
tion demands on the system. The User Note in AISC Seis- set of forces, and as such, this approach is not sufficient to
mic Provisions Section F2.6c.4 (AISC, 2016) is intended understand this complex stress state. Third, as a result of
to address this goal. While the gusset plates should be these two issues, the local demands are not properly simu-
designed to develop the expected brace forces, it is unwise lated, but this is not recognized in the paper. Nevertheless,
to overdesign the gusset plate because that will require a detailed discussion of the local shear forces and bending
larger and more expensive welds on the gusset plate-to- moments of the beam in the midspan gusset-plate connec-
beam or plate-to-column interface. The paper suggests that tion are provided, and it is asserted that these local effects
Fig. 1. Inelastic deformation of a gusset plate under seismic deformation.
2 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

(a) Gusset-plate connection based on expected brace forces
Fig. 2(a). Fracture of gusset-plate welds designed by stress.
ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 3

(b) Inelastic deformation of braced frames
(c) Buckled brace just prior to connection weld fracture
(d) Fractured weld
Fig. 2(b-d). Fracture of gusset-plate welds designed by stress.

control the design of the beam. This conclusion is incorrect of the beam. Further, stresses are significantly higher in
because the simulation neglects the axial load, shear force, regions that are not adjacent to the midspan gusset plate;
and bending moment demands in the beam outside the these higher stresses must be considered in design. They
gusset area. These forces and moments are larger than the are neglected in the paper, however, and this is a primary
forces and moments in the gusset area as demonstrated in concern to the responders.
Figure 3, and they form the primary design considerations
2. The paper focuses on the stress-distribution transfer
for the chevron beam.
between the midspan gusset plate and the chevron beam;
Considering these three issues, there are several concerns
to achieve this, an idealized distribution of stresses is
with the paper’s approach:
proposed, and these stresses are transferred to the beam
1. Using St. Venant’s principle indicates that the local as an external load. This approach does not represent
stresses in the beam in the vicinity of the gusset plate the stress transfer because the gusset plate is attached
are not reliable when using this calculation method. The to the beam by continuous welds (or bolts). Therefore,
principle clearly states that an analysis will produce compatibility of the strain between the beam and gusset
reasonable results at considerable distance from the load plate must be maintained at this interface, and these
if the load is in equilibrium with the true condition but are truly internal stresses. The stresses and strains
distributed differently; however, calculated stresses near in the beam and gusset plate must be the same at this
or adjacent to the applied loads will be quite different location, and the gusset plate and the beam join to resist
under these conditions. Rather than use these reliable the forces. The proposed idealization suggests that only
stresses away from the gusset plate, the analysis proposed the beam resists these forces; this is incorrect. The local
in the paper focuses on local stresses in a disturbed region, effects predicted in the paper’s proposed method do not
and first principles indicates that this is not correct. In represent the significant overall moment, shear, and axial
comparison, consider the analysis shown in Figure 4, load demands, and they do not consider that the chevron
which are the results from a nonlinear analysis of a full beam is strengthened significantly by the gusset plate in
chevron-braced frame. The figure shows that the stresses this region.
and deformations in the beam when the full system is
3. The axial load in the beam is neglected. A braced frame
analyzed are very different from those described in
is often idealized as a truss. This idealization neglects
the paper. The beams were designed for the forces of
the bending moment demands on the beams and column,
Equations 1, 2, and 3, but there is still significant yielding
but research by the responders indicates that it does
Fig. 3. Axial-load, shear-force, and bending-moment diagrams for idealized

boundary condition for capacity-based design of chevron beams.

(a) Frame with chevron beam designed to current AISC (b) Frame with chevron beam designed to current AISC
/
Seismic Provisions with tw tp of 0.8 and with stiffeners /
Seismic Provisions with tw tp of 0.8 and with no stiffeners
Fig. 4(a). Nonlinear analyses of chevron-braced frames. Fig. 4(b). Nonlinear analyses of chevron-braced frames.

provide a reliable estimate of the axial demands. As analytical results shown in Figure 4 indicate significant
such, all primary members in SCBFs must be designed stress due to shear, bending, and axial force outside the
considering the axial forces. Neglecting these forces is gusset regions. Within the midspan gusset regions, the
not an acceptable approach for chevron-beam design. The beams are more lightly stressed because the gusset plates
chevron beam must be designed to develop the combined stiffen and strengthen the beams in this location.
axial load and bending moment, not local stresses at the
gusset plate.
A BETTER WAY
4. In chevron-braced frames, the brace forces distributed
Local beam forces result from post-buckling brace forces
over the region of the gusset plate and the extreme local
and the boundary conditions of the beam, and these local
stresses shown in the paper cannot occur. The stresses at
effects simply change the direction of shear, axial force,
the interface also include stresses induced by out-of-plane
and bending moment vectors over the length of the midspan
brace buckling deformations, which are not included in
gusset plate. Figure 3 shows that the maximum moment,
the proposed model. Further, the ends of the chevron
shear, and axial load are just outside the gusset plate, shown
beam may not be simply supported. There are often
as dashed lines in Figure 3. Analysis of the frame shows the
corner gusset plates for the bracing of the story above,
von Mises stresses resulting from combined moment, shear,
and research has shown that these corner gusset plates
and axial load are smaller in the region of the midspan gus-
will provide significant end restraint to the beam and
set and largest just outside of the corner and midspan gus-
develop significant end moments. This is clearly seen in
set plates (see Figure 4). Although these forces are affected
the nonlinear analysis of Figure 4, but it is not considered
by the local stresses argued in the paper, the only effect of
in the paper. If there is no corner gusset plate, and the
these local stresses is to change direction and magnitude of
beam-to-column connection is a shear plate or similar
these forces over the short length of the gusset plate, which
connection, the end restraint and resulting moments are
is not the critical section of the beam. The actual shear and
very small, as depicted in Figure 3. Figure 3 and the
axial force must be largest in the beam section outside the
(c) Frame with chevron beam designed to current AISC

/
Seismic Provisions with tw tp of 1.0 and with no stiffeners
Fig. 4(c). Nonlinear analyses of chevron-braced frames.

gusset plates, shown by the solid lines. Note that Figure 3 however, because it is likely that the gusset plate will be
assumes a pinned condition at the beam-to-column connec- thicker than the lightest beam needed to support chevron-
tion; if the stiffness of the corner gusset is included, then the beam forces. Figure 4 illustrates the impact of stiffeners and
demands and resulting stresses are also high just outside of the gusset plate-to-beam web thickness ratio. The validity
the corner gusset plate, as shown in Figure 4. The moment of this approach is illustrated in Figure 4 for frames sub-
will be slightly larger over the length of the midspan gusset, jected to large inelastic deformations. Figure 4(b) shows a
but the gusset plate adds considerable cross-sectional area chevron-beam web that is thinner than the gusset plate, and
and depth to the beam; thus, this larger moment is resisted a significant amount of local yielding can be seen in the
by the combined capacity of the wide-flange section and beam web above the midspan gusset plate. Figure 4(a) has
gusset plate. The critical section is outside of the gusset the same beam and gusset plate, but stiffeners are applied to
plate—and, therefore, the bending moment at the edge of the beam web and midspan gusset plate. At the same inelas-
the gusset plate—needs to be checked in the beam design. It tic deformation as in Figure 4(b), the beam web adjacent to
is important to note that the beam must be designed for the the midspan gusset plate is quite clear of all yielding. Fig-
maximum combined axial, bending, and shear demands, ure 4(c) has a beam with web thickness equal to the gusset-
which occur at the circled areas of the figure. This design plate thickness and no beam-web or gusset-plate stiffeners.
requirement is not addressed in the paper. While the thinner beam web with stiffeners performed bet-
While the responders disagree with results of the paper’s ter, this frame with the thicker web and no stiffeners per-
stress-driven approach, they also recognize that additional formed acceptably.
checks are needed to evaluate the midspan gusset plate- The results indicate the stiffeners can help achieve the
to-beam connection in chevron-braced frames. The paper desired response and a web thickness less than the gusset
contends that seismic design requires a detailed stress cal- plate can be used; further research is being conducted by
culation based on forces used as aids in capacity-based the responders to provide more guidance on this aspect
design. However, these calculated stresses are not accurate, of design and detailing. While adding stiffeners might be
and they will not capture the local stress demands that actu- viewed as an undesirable expense, they serve multiple pur-
ally occur. As previously noted, the gusset-plate connec- poses here, which justify their cost. The bending moment
tion develops larger stresses than computed if the gusset in a chevron beam is largest at the midspan gusset plate,
plate is solely designed to develop the expected brace forces which means that bracing against lateral-torsional buckling
because out-of-plane brace buckling induces inelastic is also required at this location. In addition, stiffeners can
deformation demands on the gusset plate. be used to attach transverse beams or elements used for
To accommodate these demands, gusset-plate welds lateral support and aid in transferring brace forces to the
should be sized to the plastic capacity of the gusset plate as chevron beam.
stated in the User Note in AISC Seismic Provisions Section An X-braced frame with top and bottom gusset plates
F2.6c.4. If the welds are designed to meet this user note, is even more likely to require stiffeners with thin webs
capacity-based design would logically require that the beam because stiffeners will facilitate force transfer between sto-
must also be capable of developing this resistance. For cor- ries. The size of the stiffeners could be based on the size of
ner gusset plates where the beam is attached to the column the balance of the gusset-plate and beam-web thicknesses.
as part of the beam-to-column connection, the beam flange A similar approach can be used for corner gusset plates, but
could aid in developing this resistance. However, the beam this could be relaxed when the beam flange is attached to
flange is not restrained for midspan gusset plates, and hence the column because this attachment stiffens the flange so
the web of the chevron-beam must resist these demands. If that it also facilitates transfer of forces from the gusset plate
the yield stress of the gusset plate and the beam are the to the beam or column.
same, this simply means that the thickness of the beam
web, t w, should be equal to or greater than the thickness CONCLUSIONS
of the gusset plate, tgp (in multistory X-bracing, the thicker
gusset plate controls). This is a very simple check, but it will The responders recognize that this is a difficult topic and
ensure that stress and force demands required to develop more research is needed to define the optimal design pro-
the buckling capacity of the brace are resisted by the beam. cedure. However, the responders contend that the research
Further, this again illustrates why it is so important that the presented herein justifies a simpler and more reliable
gusset plate be large and strong enough to develop the brace method for evaluating beams and midspan gusset plates in
capacity, but no larger or stronger than needed. If a larger seismic design of chevron or multi-story, X-braced frames
plate is used, a thicker web is required. versus that proposed in the paper. This research clearly sug-
This simple check may be difficult to satisfy in practice, gests the following design method:

1. The brace is designed to develop the seismic design beam web, vertical stiffeners to the beam web and the
forces required by ASCE/SEI 7. gusset plate should be employed.
2. The expected tensile and compressive capacity of these
REFERENCES
selected braces are used to perform a capacity-based
design of the gusset plates. The gusset plates should be AISC (2016), Seismic Provisions for Structural Steel Build-
strong enough to develop the expected resistance of the ings, ANSI/AISC 341-16, American Institute of Steel
brace, but additional strength beyond this requirement Construction, Chicago, Ill.
is discouraged because overly strong gusset plates will ASCE (2016), Minimum Design Loads and Associated Cri-
require larger welds and adversely affect braced-frame teria for Buildings and Other Structures, ASCE/SEI 7-16,
performance. American Society of Civil Engineers, Reston, Va.
3. The welds or bolts joining the gusset plate to the beams Hadad, A.A. and Fortney, P.J. (2020), “Investigation on the
and the columns should be strong enough to develop the Performance of a Mathematical Model to Analyze Con-
resistance of the gusset plate, ensuring they are able to centrically Braced Frame Beams with V-Type Bracing
accommodate the deformation demands of the gusset Configurations,” AISC, Engineering Journal, Vol. 57,
plate due to out-of-plane brace buckling. No. 2, pp. 91–108.
4. The beams of chevron (V- and inverted V-configuration) Lehman, D.E., Roeder, C.W., Herman, D., Johnson, S., and
and multi-story, X-braced frames should be capacity Kotulka, B. (2008), “Improved Seismic Performance of
designed to develop the expected shear force, bending Gusset Plate Connections,” Journal of Structural Engi-
moment, and axial force in the beam from maximum neering, ASCE, Vol. 134, No. 6.
elastic and plastic brace forces, including deterioration of Richard, R.M. (1986), “Analysis of Large Bracing Connec-
the compressive capacity. tion Designs for Heavy Construction,” Proceedings of
the National Steel Construction Conference, AISC, June
5. Because the midspan gusset plate stiffens and strengthens
12–14, Nashville, Tenn.
the beam where they are attached, the demands should
be evaluated in the beam adjacent to the midspan gusset Sen, A.D., Roeder, C.W., Berman, J.W., Lehman, D.E., Li,
plate. No stress calculations for the beam in the midspan C.-H., Wu, A.-C., and Tsai, K.-C., (2016), “Experimental
gusset plate region are recommended. Investigation of Chevron Concentrically Braced Frames
with Yielding Beams,” Journal of Structural Engineer-
6. The midspan connection should fulfill one of two ing, ASCE, Vol. 142, No. 12.
requirements. If the thickness of the beam web is equal
Swatosh, M. (2015), “AISC SCBF3,” Report to AISC, Uni-
to or greater than the thickness of the gusset plate, no
versity of Washington, Seattle, Wash.
further design checks are required. Otherwise, when the
thickness of the gusset plate exceeds the thickness of the

AISC Provisions for Web Stability Under Local
Compression Applied to HSS
FEI WEI and JEFFREY A. PACKER
ABSTRACT
The relevant limit states for local compression loading on the webs of a rectangular HSS member are reviewed, and the 2016 AISC Specifi-
cation Chapter J provisions are adapted from their normal application to the single web of a W-shape or I-section to this case. Two recent
laboratory tests on matched-width, rectangular HSS-to-HSS cross-connections are described to illustrate the behavior of such connections
under branch axial compression. The data from these tests are supplemented by experimental results from a further 76 cross-connection
tests, with the branches being either welded plates or welded HSS. From this 78-test database, the existing provisions for local yielding of
the chord sidewalls, local crippling of the chord sidewalls, and buckling of the chord sidewalls are evaluated. Recommendations are made for
handling transverse compression loading on HSS webs in the AISC Specification, and a design example is given to illustrate the approach.
KEYWORDS: hollow structural sections, cross-connections, web yielding, web crippling, web buckling, design procedures.
INTRODUCTION cases of the concentrated compression load being applied:

(1) away from the member end (termed “interior” herein)
C oncentrated compression forces on rectangular HSS
are relatively common, especially at bearing or reac-
tion points of trusses and girders and at beam-to-column
and being free of any end effects and (2) close to the mem-
ber end (termed “end” herein). The latter would correspond
to a compression load close to an open end of an HSS mem-
moment connections. This loading situation is covered in
ber, without a cap plate. This paper evaluates transversely
AISC Specification Section K2.3 (AISC, 2016), where one
loaded HSS connections remote from the member end.
is directed to determine the connection available strength
from the applicable limit states in Chapter J.
Web Local Yielding
For loading across the full width of the HSS (or when
the branch-to-chord width ratio β = 1.0), the two webs are Local yielding of the HSS webs is a possible limit state for
loaded in compression, and yielding or instability of the both compression and tension concentrated loads, and it
chord/column webs will control the connection capacity. applies to T-, Y- and cross- (or X-) connections with β ≈
AISC Specification Section J10 (AISC, 2016) on “Flanges 1.0. The applied load, acting over a bearing length of lb, dis-
and Webs with Concentrated Forces,” which is based on perses at a slope of 2.5:1 to the “k line” and thus produces
the behavior of I-shaped sections with a single web, speci- yielding over a length of (5k + lb) for an interior connection.
fies the applicable limit states. For laterally supported HSS This load-dispersion angle of 21.8° is a classical assump-
connections these are (1) web local yielding (Section J10.2), tion throughout steel codes. The distance k, from the outer
(2) web local crippling (Section J10.3), and (3) web com- face of the flange to the web toe of the fillet for a wide
pression buckling (Section J10.5). In the following, these flange or I-section, can be taken for a rectangular HSS as
limit states are further described, applied to the case of the outside corner radius, with a conservative value of 1.5t,
HSS webs, and evaluated against test results for matched- where t is the HSS member design thickness (AISC Speci-
width, HSS-to-HSS cross-connections and plate-to-HSS fication Section J10 Commentary). The applicable connec-
connections under transverse compression. For all three tion nominal strength equations, in both wide-flange (or
limit states, the AISC Specification considers separate I-section) format and HSS format, are shown in Table 1 for
interior- and end-loading situations. In laboratory experi-
ments, this failure mode has been found to occur for short
bearing lengths (such as with plate-to-HSS connections, as
Fei Wei, Structural Engineer, Cast Connex Corporation, Toronto, Ontario,
Canada. E-mail: [email protected] shown in Figure 1) and also for stocky chord walls with
Jeffrey A. Packer, Bahen/Tanenbaum Professor of Civil Engineering, Depart-
longer bearing lengths.
ment of Civil & Mineral Engineering, University of Toronto, Toronto, Ontario,
Canada. E-mail: [email protected] (corresponding) Web Local Crippling
This limit state is defined as the crumpling of the web
Paper No. 2019-11R into buckled waves directly beneath a compression load,

occurring in more slender webs, whereas web local yield- I-shapes) is only valid for bearing lengths “…for which lb /d
ing of that same area occurs for stockier webs (AISC Speci- is approximately less than 1.” A validity range of lb /d is
fication Section J10.3 Commentary). Research by Roberts hence included in Table 1. AISC Specification Equation
(1981) on the compression of a single, slender, I-section web J10-8 originates from Newlin and Chen (1971), who showed
provided the basis for the nominal strength expressions in that their semi-empirical expression was a lower bound for
the AISC Specification. As shown in Table 1, modified web buckling failure loads achieved in a small number of
versions are provided for interior- and end-loading situa- transverse compression tests on point-loaded, wide-flange
tions. Because the overall member depth, d, is used in the sections. AISC Specification Section J10 Commentary
I-section web crippling AISC Specification Equation points out that Specification Equation J10-8 assumes
J10-4, for consistency this is replaced by the HSS over- pinned restraints at the ends of the web.
all depth, H, in the conversion shown in Table 1. This is The dimension h is defined as the clear distance between
a small and conservative difference to the presentation flanges less the fillet or inside corner radius. Thus, in the
in the 2010 AISC Specification Equation K2-10 (AISC, conversion of web compression buckling formulas to HSS
2010). This limit state is applicable to “compressive single- format, h is taken equal to (H − 3t), which represents a max-
concentrated forces” (Specification Section J10.3), hence imum height of the flat part of the chord sidewall. For long
to T- and Y-connections with β ≈ 1.0. However, this fail- bearing lengths, greater than the HSS overall depth, H, the
ure mode has not been observed in rectangular HSS con- web needs to be designed as a column member in accor-
nections, which is presumed to be because the typical H/ t dance with AISC Specification Chapter E. Treating each
values of HSS webs are below the wall slenderness require- HSS web as a column with a rectangular cross section is
ment for this failure mode to govern. [Note that the tests actually the method for handling web compression failure
reported by Roberts (1981) had overall height-to-web thick- in Eurocode 3 (CEN, 2005), CIDECT Design Guide No. 3
ness ratios ranging from 75 to 505, with very few below (Packer et al., 2009), and ISO 14346 (ISO, 2013). This fail-
100.] Nevertheless, although the scope of the study pre- ure mode has been observed experimentally for full-width
sented herein is for HSS cross-connections, the applicable HSS-to-HSS cross-connections with H/ t greater than about
connection nominal strength equations for web local crip- 15 (Figure 2). An earlier investigation on the influential
pling, in both wide flange (or I-section) format and HSS parameters affecting the web strength of HSS chords under
format, are shown in Table 1. transverse compression, by Davies and Packer (1987), indi-
cated that the bearing length parameter, lb /H, affects the
Web Compression Buckling chord sidewall slenderness, H/ t, at which failure changes
from web bearing (local yielding) to web buckling.
This limit state involves overall buckling of the entire
For web compression buckling with lb > d, or Hb / sinθ > H
web and only applies to “a pair of compressive single-
(i.e., beyond the applicable limit of Table 1), each web is
concentrated forces” (AISC Specification Section J10.5),
to be treated as a column of slenderness KL/ r, where the
hence to HSS cross-connections with β ≈ 1.0, where com-
effective length factor, K, can be taken as 1.0 (as suggested
pression force is transferred through the chord/column
by AISC Specification Appendix Section 7.2.3, considering
member. AISC Specification Section J10.5 Commen-
the main HSS through member as a non-sway frame). The
tary notes that the nominal strength expression (for W- or
column length, L, is taken as the sidewall flat dimension,
equal to (H − 3t). The radius of gyration, r, of a rectan-
gular cross-section HSS wall is t 12. Thus, the nominal
Fig. 2. Web buckling failure in a full-width

Fig. 1. Web local yielding failure in a full-width plate-to-HSS (β = 1.0) HSS-to-HSS cross-connection, with
connection, with plates in compression. /
branches in compression and H t = 23.

Table 1. Nominal (and Available) Strengths of Web Compression Limit States for
Wide Flange (I-Section) Shapes and Rectangular HSS Connections, per the AISC Specification
/
(Equation Numbers and ϕ Ω Values from the Specification)
Wide Flange or I-Section, Rn HSS-to-HSS Connection, Pn ϕ
Limit State (kips) (kips) Ω)
(Ω
Web local for lend > d for lend > H
yielding, Fyw tw (5k + lb) 2Fy t ⎛ H ⎞ 1.00
(J10-2)
interior 7.5t + b (1) (1.50)
sinθ ⎝ sin θ ⎠
Web local for lend ≤ d for lend ≤ H

1.00
yielding, Fywtw ( 2.5k + lb ) (J10-3) 2Fy t ⎛ H ⎞
end 3.75t + b (2) (1.50)
sinθ ⎝ sin θ ⎠
Web local /
for lend ≥ d 2 for lend ≥ H 2 /
crippling, ⎡ ⎛ 3Hb ⎞
⎛ l ⎞ ⎛ tw ⎞ ⎤ EFyw tf
1.5
2 0.75
interior 0.80tw2 ⎢1+ 3 b ⎥ Qf (J10-4) 1.6t ⎜ sinθ ⎟ (3)
⎝ d ⎠ ⎝ tf ⎠ ⎦ tw 1+ EFy Qf (2.00)
⎣ sinθ ⎜ H ⎟
⎝ ⎠
Web local /
for lend < d 2 / /
for lend < H 2 and Hb Hsinθ ≤ 0.2
crippling, ⎡ ⎛ 4Hb ⎞
⎛ l ⎞ ⎛ tw ⎞ ⎤ EFyw tf
1.5
2 0.75
end, and 0.40tw2 ⎢1+ 3 b ⎥ Qf (J10-5a) 0.8t ⎜
1+ sinθ⎟ EFy Qf
/
lb d ≤ 0.2 ⎣ ⎝ d ⎠ ⎝ tf ⎠ ⎦ tw
sinθ ⎜
⎝
H ⎟
⎠
(4) (2.00)

Web local /
for lend < d 2
for lend < H 2 and Hb Hsinθ > 0.2
4Hb ⎞
/ /
crippling, ⎡ ⎛ l ⎛ ⎞ ⎛ tw ⎞ ⎤ EFyw t f
1.5
end, and 0.8t 2 ⎜
0.40tw2 ⎢1+ 4 b − 0.2 0.75
⎥
0.8 + sinθ⎟ EFy Qf Qf
(5)
/
lb d > 0.2 ⎣ ⎝ d
sinθ ⎜ ⎠ ⎝ tf ⎠ ⎦
H ⎟ tw (2.00)
⎝ ⎠
(J10-5b)
Web /
for lend ≥ d 2 / /
for lend ≥ H 2 and Hb Hsinθ ≤ 1.0
compression 24tw3 EFyw 1 ⎛ 48t 3 ⎞ 0.90
buckling, Qf (J10-8) sinθ ⎝ H − 3t⎠ EFy Qf (6)
h (1.67)
interior, and
lb ≤ d
Web /
for lend < d 2 / /
for lend < H 2 and Hb Hsinθ ≤ 1.0
compression 12tw3 EFyw 1 ⎛ 24t 3 ⎞
EFy Qf (7) 0.90
buckling, Qf
h sinθ ⎝ H − 3t⎠ (1.67)
end, and
lb ≤ d
Note: lend = distance from the near side of the connecting branch or plate to end of member
flexural buckling strength of the two HSS sidewalls can be 7 in Table 1, the factor Q f (from AISC Specification Table
calculated from AISC Specification Section E3, with an K3.2) should also be included if chord compression stress
allowance for an inclined branch producing a longer web is present.
buckling length (Packer et al., 2009; IIW, 2012; ISO, 2013) A recent numerical study of welded, full-width, rectan-
by: gular HSS cross-connections by Kuhn et al. (2019) showed
that 0.25 represented a critical value for the bearing length-
KL Lc ⎛H ⎞ 1
= = 3.46 −3 (8) to-chord height ratio at which the failure mode changed
r r ⎝ t ⎠ sin θ from web yielding to web buckling. Thus, for rectangular
and, for one sidewall, a “column” cross-sectional area HSS-to-HSS cross-connections and plate-to-HSS cross-
given by Ag = (7.5t + Hb/ sinθ)t, from Equation 1 in Table 1 connections with (Hb/ sinθ)/H ≤ 0.25, web local yielding was
for lend > H, or Ag = (3.75t + Hb/ sinθ)t from Equation 2 in deemed to govern and could be predicted by a model such
Table 1 if lend ≤ H. For consistency with Equations 6 and as Equation 1. Hb represents either the HSS branch depth,

in the plane of the connection or, alternatively, the thickness EXPERIMENTS ON FULL-WIDTH
of a transverse, full-width plate. If (Hb/ sinθ)/H > 0.25, web RECTANGULAR HSS CROSS-CONNECTIONS
compression buckling was deemed to govern and could be
Two recent laboratory tests (Wei, 2019) on matched-
predicted by treating the two chord sidewalls as columns,
width, rectangular HSS-to-HSS 90° cross-connections
for which a modification of Equation 1 could be used:
are described to illustrate the behavior under branch axial
2χFy t ⎛ H ⎞ compression. These were tested to failure under displace-
Pn = 7.5t + b (9)
sin θ ⎝ sin θ⎠ ment control, in quasi-static branch compression, as shown
in Figure 4, using a 1,000-kip-capacity universal testing
where χ is a reduction factor applied to yield stress for col- machine. As can be seen from Figure 4, the branch compres-
umn buckling. For fully welded branches to either side of sion load was reacted by a steel plate, which was secured
the chord member, the end fixity of the sidewall “column” to the laboratory strong floor, and no lateral restraint was
is likely closer to fixed-fixed than pin-ended. For fixed- provided to the chord member. Displacement was captured
fixed end conditions, Kuhn et al. (2019) noticed that most at many points by a Metris K-610 3D Dynamic Laser Mea-
steel codes have a cold-formed column buckling curve that suring System together with a linear variable differential
is almost linear when plotted over a practical chord sidewall transformer (LVDT). All members were made of cold-
slenderness range; hence they advocated a simple conserva- formed HSS to either ASTM A500 Grade B/C (ASTM,
tive estimation for χ using: 2018) or CSA G40.20/G40.21 (CSA, 2013), and two chord
H 1 sizes were used: HSS 8×8×4 and HSS 8×8×a. A common
χ = 1.15 − 0.013 ≤1 (10) branch size of HSS 8×4×2 was used, oriented such that
t sin θ
β = Bb/B = 1.0 and η = (Hb/ sinθ)/B = 0.5. The branch thick-
Equation 10, for Fy ≤ 50 ksi and H/ t ≤ 50, is shown plot- ness was selected to be greater than that of the chord to be
ted in comparison to the 2016 AISC Specification column certain that local branch yielding would not occur before
buckling curve in Figure 3. The vertical axis in this figure, the chord webs failed. Measured geometric properties are
χ, is equivalent to the AISC Specification buckling stress, given in Tables 2 and 3. Mechanical properties of the two
Fcr , divided by the yield stress, Fy. An effective length fac- chord members were determined by tensile tests on cou-
tor of K = 0.65 is used as a design approximation to the pons cut from the flat regions where there was no weld
theoretical fixed-fixed factor of K = 0.5. This approach seam. Average measured values (using three coupons from
advocated by Kuhn et al. (2019) is also evaluated against each HSS) are shown in Table 4.
test data later in this paper, in addition to the current 2016 In both connection tests, sidewall buckling was the
AISC Specification method. observed failure mode, and the maximum load, Pa, was
achieved prior to the 3%B connection ultimate deformation
Fig. 3. AISC Specification column buckling curve and the linear approximation of Equation 10, for Fy = 50 ksi.

Table 2. Test Specimens and Measured Geometric Variables
Width Chord Wall-Thickness Chord Branch Fillet Weld
Ratio Slenderness Ratio Length Length Size (leg)
Ratio
Specimen β 2γγ = B t/ /
τ = tb t (in.) (in.) (in.)
X1 1.0 34.7 2.17 41.4 20.0 0.25
X2 1.0 23.6 1.47 38.1 20.0 0.22
Table 3. Average Measured Rectangular HSS Cross-Sectional Dimensions

Width Height Wall Thickness Corner Radius
Designation B (in.) H (in.) t (in.) Outer (in.) Inner (in.)
HSS 8×8×4 7.98 7.98 0.23 0.59 0.36
HSS 8×8×a 8.03 8.03 0.34 0.94 0.60
HSS 8×4×2 8.02 4.02 0.50 1.01 0.51
Table 4. Average Measured Rectangular HSS Chord Material Properties

E Fy Fu
Designation (ksi) (ksi) εy (ksi) ε rup /
Fy Fu
HSS 8×8×4 30,180 57.1 0.0039 70.3 0.308 0.81
HSS 8×8×a 28,630 56.9 0.0040 71.6 0.334 0.79
Fig. 4. Testing arrangement for rectangular HSS cross-connections, with failure by web buckling.

Table 5. Actual versus Predicted Ultimate Strengths and Failure Modes for Tests X1 and X2
Actual Predicted
Web
Ultimate Web Local Web Local Web Compression Compression
Strength Observed Yielding Crippling Buckling Buckling
Test No. (kips) Failure Mode Eq. (1) Eq. (3) Eq. (6) Eqs. (10) and (14)
Sidewall
X1 128 150.9 279.0 105.2 105.5
buckling
Sidewall
X2 244 254.2 590.6 343.5 214.3
buckling
limit state (Lu et al., 1994). Thus, the connection ultimate EVALUATION OF HSS WEB
strength was given by Pa, as shown on the load-displacement COMPRESSION LIMIT STATES
curves in Figure 5. To obtain the load-displacement curves
Although early design provisions have been evaluated
in Figure 5, connection displacement was determined from
(Packer, 1984, 1987), it is timely to apply the current
the global vertical displacement of the difference between
2016 AISC Specification rules to an expanded contempo-
light-emitting-diode (LED) targets placed slightly above
rary database of HSS experiments. Thus, aside from the
the chord face and the targets positioned at the centroid of
two laboratory tests described, an additional 76 cross-
the chord, and the branch compression load was provided
connection tests from the literature were collated. This total
by the testing machine’s load cell. Table 5 compares the
database consists of 44 tests performed at the University of
predicted ultimate strength and predicted failure mode, by
Toronto, 29 in the United Kingdom, and 5 in Spain. Perti-
the three limit states, with the observed strength and failure
nent data for all 78 tests is tabulated in Appendix G of Wei
mode. For test X1, the capacity is reasonably predicted for
(2019). The group of 78 tests covers chord sidewall slender-
the correct failure mode. For test X2, the capacity is rea-
ness ratios (H/t) from 12.6 to 56.9; bearing lengths ranging
sonably predicted but for an incorrect failure mode. Both
from 0.07H to 3.72H; chord compressive stress up to 86%
of these connections had a bearing length-to-chord height
of chord yield stress; branch angles of 45°, 60°, and 90°; and
ratio of 0.50, but different H/t ratios (34.7 and 23.6). These
three HSS production processes: cold-formed, cold-formed
results indicate that a wider review of these limit states—as
stress-relieved, and hot-formed. Measured geometric and
applied to HSS connections—is warranted.
Fig. 5. Connection load-displacement curves for tests X1 and X2.

mechanical properties of the test specimens were used. nonslender elements, as defined in Section B4.1; however,
Based on the configuration of individual connection tests, none of the cases included in Table B4.1a of the Specifica-
the database has been further divided into two categories: tion directly correspond to a laterally compressed rectan-
welded rectangular HSS to rectangular HSS connections gular HSS sidewall. Hence, the full gross cross-sectional
and welded plate to rectangular HSS connections. area, Ag, is always used in the sidewall buckling equation.
The critical stress, Fcr , is determined based on the slender-
Welded Rectangular HSS to Rectangular ness ratio of KL/r. The Specification and Commentary do
HSS Connections not clearly state which value of the effective length factor,
K, to use, and designers could adopt K = 1.0 to be conserva-
In this section, an evaluation is made of the current design
tive. Because Fcr ≤ Fy, the limit state of local yielding of
provisions against 53 welded rectangular HSS to rectangu-
the chord sidewalls is thus incorporated into the nominal
lar HSS cross-connections. When the bearing length of the
flexural buckling strength by:
load, Hb/sinθ, is less than or equal to the total depth of the
chord member, H, the three web compression limit states 2Fcr t ⎛ H ⎞
Pn = 7.5t + b Q f (13)
(local yielding of the chord sidewalls, local crippling of sin θ ⎝ sin θ⎠
the chord sidewalls, and buckling of the chord sidewalls)
are represented by Equation 1, Equation 3, and Equation 6 where Q f is a reduction factor to account for the effect of
respectively, as discussed earlier. normal stress in the chord. All 16 connection tests with
The correlation between actual experimental test results bearing length of the load greater than the total depth
and predicted connection strengths is shown in Figure 6. of the chord member were governed by the limit state of
Ultimate strengths Pn and Pa are used in the correlation flexural buckling of the chord sidewalls, represented by
plots, where Pn represents the connection theoretical capac- Equation 13, and the correlation between actual experimen-
ity calculated from the limit states and Pa represents the tal test results and predicted connection strengths, using
actual experimental test results recorded by the research- K = 1.0, is shown in Figure 7.
ers, both expressed as a force in the branch. Although the With the two bearing length situations combined, cor-
mean of this ratio is 1.37, the scatter is huge (COV = 0.45). relation between the entire group of 53 welded rectangular
A number of tests are overestimated, while some tests are HSS-to-HSS cross-connection tests is plotted in Figure 8.
significantly underestimated by the limit state of chord The mean ratio of actual/predicted capacity is 1.41 with a
sidewall buckling. The large variability shown by chord very large scatter (COV of 0.46). When the bearing length
sidewall buckling predictions indicates that the interior of the load is greater than the total depth of the chord mem-
web compression buckling equation, Equation 6—when ber, web buckling failure predictions given by Equation 13
applied to rectangular HSS-to-HSS connections—is gener- using K = 1.0 are conservative for a considerable number of
ally a poor predictor of the strength for this limit state. As the test results. In this case, 1 out of 53 tests had a ratio of
noted earlier, Equation 6 originates from point-load tests on actual/predicted capacity greater than 2.6, which did not fit
wide-flange section webs. In addition, none of the connec- into Figure 8(a). This correlation suggests that the end fix-
tion tests is governed by local crippling of the chord side- ity of the sidewall “column” is more likely to be fixed-fixed
walls, represented by Equation 3, indicating that this is not rather than pin-ended for a chord member with branches
a viable failure mode over this range of data. welded to either side, which seems logical considering the
When the bearing length of the load is greater than the large flare bevel groove welds at either end of the chord
total depth of the chord member, each chord sidewall needs member web. This implies that the effective length factor,
to be designed as a column with a slenderness ratio of KL/ r. K, can better be taken as 0.65 instead of 1.0.
As discussed earlier, instead of using Equation 6, the nomi- For these 53 welded rectangular HSS-to-HSS cross-
nal flexural buckling strength can be calculated using AISC connections, 42 of them have the failure mode recorded
Specification Section E3: (although there may be misinterpretations of the initial fail-
ure mode by some researchers). Nine of the 42 connections
Pn = Fcr Ag (11) had the failure mode incorrectly predicted when compared
against the observed actual test failure mode (Wei, 2019).
and the “column” cross-sectional area of one web can be All nine of these incorrect predictions were a result of high
calculated as: predicted sidewall buckling strength.
⎛ H ⎞
Ag = 7.5t + b t (12) Welded Plate to Rectangular HSS Connections
⎝ sin θ⎠
Unlike the welded rectangular HSS-to-HSS cross-connection
AISC Specification Section E3 applies the full gross tests, the load bearing length of all 25 welded plate to rect-
cross-sectional area, Ag, to compression members with angular HSS cross-connection tests is less than the total

(a)
(b)
Fig. 6. Correlation between 37 welded rectangular HSS to rectangular HSS

connection tests with bearing length ≤ H and the 2016 AISC Specification.

(a)
(b)
Fig. 7. Correlation between 16 welded rectangular HSS to rectangular HSS

connection tests with bearing length > H and 2016 AISC Specification, using K = 1.0.

(a)
(b)
Fig. 8. Correlation between 53 welded rectangular HSS to rectangular HSS connection tests and the AISC Specification.

depth of the chord member. Thus, the web compression [rather than the current 1.0H, as indicated below Equa-
limit states are represented by Equation 1, Equation 3, and tion 6 in Table 1]. With this modification, the capacity of
Equation 6. In this category, the web local yielding limit 37 welded rectangular HSS-to-HSS connections with bear-
state governed the predicted strength of all cases, with only ing length (Hb/ sinθ) ranging from 0.25H to H, which are
one exception. The actual-to-predicted strength distribu- originally estimated as per Table 1, can now be predicted
tion, shown in Figure 9(a), is much better compared to that by Equation 13. A second proposed modification is to
of welded rectangular HSS-to-HSS connection tests. The adopt an effective length factor of K = 0.65 instead of 1.0.
mean is slightly greater than unity (1.15) with a relatively Combined with the other 16 tests, the correlation between
low spread of data, indicated by a COV of 0.18. Figure 9(b) actual experimental test results and predicted connection
shows a somewhat more conservative correlation by taking strengths using Equation 13 is presented in Figure 10(a) to
the predicted branch capacity as simply the vertical force evaluate the effectiveness of these modifications.
component. These plots indicate the excellent applicability A clear improvement can be observed by a comparison
of the existing web local yielding model to HSS webs. of Figure 10(a) with Figure 8(a). Although the COV of 0.27
In Figure 9, the one test for which buckling governed is still not low, it is significantly reduced from the value of
had Hb/H = 0.20 and H/ t = 57. Because Hb/H < 0.25, the 0.46 obtained previously, which indicates that the sidewall
breakpoint between web yielding and web buckling estab- flexural buckling equation, Equation 13, is a better strength
lished by Kuhn et al. (2019), one might expect web yield- predictor of connections with bearing length (Hb/ sinθ) >
ing to govern. However, the 0.25 value was determined 0.25H when compared to web buckling Equation 6. More-
on the basis of numerical research on HSS up to H/t = 50 over, for connections with inclined branches (i.e., when
(Kuhn et al., 2019), so there may be less reliability in this θ < 90°), there is a trend for the connection strengths to be
breakpoint at H/t > 50. Nevertheless, in Figure 9 the actual overpredicted. Packer (1984) has already noted that the con-
strength far exceeds the predicted strength for this test and nection strength increase (measured as a force in the branch)
is conservative. is less than associated with 1/sinθ. The effect of branch mem-
ber inclination requires more study but, in the meantime, if
one takes the predicted branch capacity as simply the vertical
ALTERNATE MODELS
force component, the more-conservative correlation shown
From the evaluation of existing test results for full-width, in Figure 10(b) is the result. In Figure 10, the connection
welded rectangular HSS cross-connections, it was shown capacity prediction is based on a column-buckling model,
that the 2016 AISC Specification web crippling equation which incorporates both sidewall local yielding (squashing)
and the web buckling equation, which are based on specific and flexural buckling; hence, no legend (buckling governs/
tests on I-section webs, either never govern or result in a yielding governs) is given in this figure.
large scatter in the predicted strengths when applied to the A simple reliability analysis (Fisher et al., 1978; Ravin-
chord sidewall of rectangular HSS. One of the most influ- dra and Galambos, 1978) can be applied to the statistics
ential parameters, bearing length Hb/ sinθ, is absent from (or model parameters) in Figure 10(b) in which a resistance
the web buckling equation. Even for connections where factor, ϕ, is calculated using a target safety/reliability index
the strength prediction is governed by a sidewall flexural of 3.0 and a coefficient of separation of 0.55. Furthermore,
buckling equation, represented by Equation 13, the assump- one can introduce statistical parameters to model geomet-
tion of pinned-pinned end fixity (K = 1.0) leads to generally ric variations [as recommended by AISI (2016)] and typical
conservative estimates. Hence, the web compression limit material strength variations for ASTM A500 Grade B/C
states for rectangular HSS connections could be modified yield strength [as determined by Liu (2016)], the result of
to one of the following. which is ϕ = 0.95. Because a value of ϕc = 0.90 is used in
AISC Specification Section E1, which is lower, adequate
Model 1 safety/reliability is provided by Model 1 for welded rectan-
gular HSS-to-HSS cross-connections.
According to a recent numerical study by Kuhn et al.
For connections with a bearing length (Hb/ sinθ) ≤ 0.25H
(2019), a failure mode transition from web local yielding
(all 25 welded plate to rectangular HSS connections), the
to web buckling was observed at a critical bearing length
capacity can be predicted by Equation 1 alone since the web
(Hb/ sinθ) = 0.25H. For bearing lengths greater than 0.25H, local yielding limit state governed the predicted strength of
the sidewall compression strength was well-predicted using
all 25 cases, with only one exception. Applying this single
the column flexural buckling approach, over a practical H/ t
limit state check to connections in this bearing length range
range associated with manufactured HSS. Thus, the first
leads to a mean value of 1.14 and a COV of 0.17, which are
proposed modification is to require HSS sidewalls to be
almost identical to what was obtained in Figure 9(a). If one
considered as columns (and analyzed using AISC Specifi-
takes the predicted branch capacity as simply the vertical
cation Section E3) for bearing lengths greater than 0.25H

(a) Using Equation 1
(b) Using Equation 1 but deleting the first sinθ term (below 2Fyt)
Fig. 9. Correlation between 25 welded plate to rectangular HSS connection test and the 2016 AISC Specification.

(b) Using Equation 13 but deleting the first sinθ term (below 2Fcr t)
Fig. 10. Correlation between 53 welded rectangular HSS to rectangular HSS connection tests and Equation 13 with K = 0.65.

force component a mean value of 1.19 and a COV of 0.15 length (Hb/ sinθ) ≤ 0.25H, it is recommended for Model 2
result, which are almost identical to what was obtained in that the buckling reduction factor, χ, be taken as 1.0. As
Figure 9(b). Performing the same reliability analysis as connections within this range are very likely governed by
described above, but with model parameters of mean = 1.19 sidewall local yielding, the correlation is again very similar
and COV = 0.15, one obtains ϕ = 1.09. Because a value of to that shown in Figure 9.
ϕ = 1.00 is used in AISC Specification Section J10.2, which
is lower, adequate safety/reliability is provided by Model 1
for welded plate to rectangular HSS connections. CONCLUSIONS
To design a full-width welded rectangular HSS-to-HSS AISC Specification Section J10 (AISC, 2016) provisions
cross-connection, the AISC Specification requires design- for concentrated compression forces on webs have been
ers to check three web compression limit states: local yield- applied to the case of transversely compressed rectangu-
ing of the chord sidewalls, local crippling of the chord lar HSS members. These provisions have then been evalu-
sidewalls, and buckling of the chord sidewalls. With this ated against 78 laboratory tests, taking the form of welded
method, the predicted connection capacity can be based on interior plate-to-HSS cross-connections and welded inte-
either Equation 1 or Equation 13 with K = 0.65, depending rior HSS-to-HSS cross-connections. The web local yield-
on the connection bearing length. A recommended adjust- ing limit state, represented by Equation 1 in Table 1, has
ment to these equations, for inclined branches with θ < 90°, been found to be very applicable to HSS. The web local
is to take the predicted branch capacity as the branch force crippling limit state, represented by Equation 3 in Table 1,
vertical component. Maintaining the checks for all three has been found to never govern for the range of HSS exam-
limit states, but using the preceding recommendations, will ined (specified yield strengths up to 50 ksi and sidewall
still result in reliable predictions of connection capacity for slenderness values up to 57). It has been shown that, for
the limit state that governs. the web compression buckling limit state, represented by
Equation 6 in Table 1, greater prediction accuracy can be
Model 2 obtained if a column buckling model is used when bear-
Another simplified method to the foregoing is also possible. ing lengths are greater than 0.25 of the chord depth. It is
As discussed previously, Kuhn et al. (2019) advocated the thus recommended (as a modification to AISC Specifi-
use of Equation 10 for a reduction factor to be applied to cation Section J10.5 Commentary) that the HSS member
yield stress for column buckling, χ, as it was noticed that web be designed as a compression member, in accordance
most steel codes have a cold-formed column buckling with AISC Specification Chapter E, when lb or (Hb/ sinθ) >
curve that is almost linear when plotted over a practical 0.25H. Moreover, when doing so, the compression member
chord sidewall slenderness range (H/ t ≤ 50) for fixed-fixed (each web) can be taken to have a cross-sectional area given
end conditions. The AISC Specification buckling curve is by Equation 12 and an effective length factor of K = 0.65.
no exception, as presented in Figure 3. Thus, to simplify The influence of branch member inclination on connection
the process of calculation, the critical stress, Fcr, can be capacity is not conclusive, so it is recommended that—in
replaced by χFy to give a sidewall compression strength of: the case of inclined branches with θ < 90°—one conser-
vatively takes the predicted branch capacity, for all failure
2χFy t ⎛ H ⎞ modes, as simply the vertical force component. Table 6,
Pn = 7.5t + b Q f (14)
sin θ ⎝ sin θ⎠ which can be compared to Table 1, provides a summary
of the foregoing recommendations, applied to interior HSS
Of the 53 tests, 47 lie within the chord sidewall slen- connections. The limit state of web local crippling should
derness range of H/ t ≤ 50. The correlation between actual be redundant for normal HSS sizes, but it is included in
experimental test results and predicted connection strengths Table 6 for completeness and also in the design example
using Equation 14 is shown in Figure 11(a) and, as expected, that follows.
a similar relationship to Model 1 is obtained. The numeri- This review has studied connections that were not prone
cal research of Kuhn et al. (2019) was based only on 90° to out-of-plane stability. This will be the usual case when
connections, so it would be logical to again investigate (as an HSS main member is subject to transverse compression
in Model 1) the correlation with experiments by taking the because lateral restraint is generally provided (e.g., at reac-
predicted branch capacity as simply the vertical force com- tion or load points of trusses and beam-to-column moment
ponent. This results in the excellent correlation shown in connections). It is conceivable that lateral instability of the
Figure 11(b). In Figure 11, the connection capacity predic- chord member could arise with a lack of symmetry due
tion is based on a single limit state model; hence, no legend to misalignment or with long compression-loaded branch
(buckling governs/yielding governs) is given in this figure. members; in such cases, this should be incorporated in
For welded plate to rectangular HSS connections or modeling structural behavior of the system.
welded rectangular HSS-to-HSS connections with bearing

(b) Using Equation 14 but deleting the first sinθ term (below 2χFyt)
/
Fig. 11. Correlation between 47 welded rectangular HSS to rectangular HSS connection tests and Equation 14, with H t ≤ 50.

Table 6. Recommended Nominal (and Available) Strengths of Web
Compression Limit States for Rectangular HSS Connections
HSS-to-HSS Connection, Pn ϕ
Limit State (kips) Ω)
(Ω
Web local yielding, interior for lend > H
⎛ 1.00
Hb ⎞
2Fy t 7.5t + (15) (1.50)
⎝ sinn⎠
Web local crippling, for lend ≥ H 2/
interior ⎛ 3Hb ⎞ 0.75
1.6t 2 ⎜ sinθ ⎟
1+ EFy Qf (16) (2.00)
sinθ ⎜ H ⎟
⎝ ⎠
Web compression / /
for lend ≥ H 2 and Hb Hsinθ ≤ 0.25
0.90
buckling, interior, and ⎛ 48t 3 ⎞
lb ≤ 0.25H EFy Qf (17) (1.67)
⎝ H− 3t ⎠
Web compression / /
for lend ≥ H 2 and Hb Hsinθ > 0.25
0.90
buckling, interior, and Use AISC Specification Equations E3-1, E3-2, and E3-3 with K = 0.65,
(1.67)
lb > 0.25H /
Lc r from Equation 8, and Ag (for each sidewall) from Equation 12
DESIGN EXAMPLE
Given:
Determine the adequacy of the welded rectangular HSS-to-HSS 90° cross-connection shown in Figure 12 subjected to the
loads indicated. The branch members are oriented such that the chord is loaded across its full width, and the loads shown con-
sist of 50% dead load and 50% live load. Assume the welds are noncritical and that there is zero force in the chord member.
From AISC Manual (AISC, 2017) Table 2-4, the material properties are as follows:
All members
ASTM A500 Grade C
Fy, Fyb = 50 ksi
Fu , Fub = 62 ksi
From AISC Manual Table 1-11 and Table 1-12, the HSS geometric properties are as follows:
HSS 8×8×a
A = 10.4 in.2
B = 8.00 in.
H = 8.00 in.
t = 0.349 in.
HSS 8×4×2
Ab = 9.74 in.2
Bb = 8.00 in.
Hb = 4.00 in.
tb = 0.465 in.

Solution:
Required strength (expressed as a force in the branch)
From ASCE/SEI 7 (ASCE, 2016) Chapter 2, the required strength of the connection is:
LRFD ASD
Pu = 1.2 (50.0 kips) + 1.6 (50.0 kips) Pa = 50.0 kips + 50.0 kips
= 140 kips = 100 kips
The strength of a matched-width (β = 1.0), welded, rectangular HSS to rectangular HSS cross-connection, under branch axial
compression, can be determined from the limit states of web local yielding, web local crippling, and web compression buckling.
8.00 in.
β= = 1.00
8.00 in.
Limit State of Web Local Yielding

From Equation 15 in Table 6,
⎛ H ⎞
Pn = 2Fy t 7.5t + b (15)
⎝ sin θ⎠
⎡ 4.00 in. ⎤
= 2 ( 50 ksi ) ( 0.349 in.) ⎢7.5 ( 0.349 in.) +
⎣ sin 90° ⎥⎦
= 231 kips
Fig. 12. Rectangular HSS-to-HSS cross-connection subjected to branch axial compression.

By applying the resistance factor of ϕ = 1.00, and the safety factor of Ω = 1.50, for this limit state (AISC Specification Section
J10.2), the available strength (ϕPn or Pn/ Ω) is:
LRFD ASD
ϕPn = 1.0 ( 231 kips ) Pn 231 kips
=
= 231 kips 1.50
= 154 kips
231 kips > 140 kips o.k.
Limit State of Web Local Crippling

From Equation 16 in Table 6,
⎛ 3H b ⎞
⎜ ⎟
Pn = 1.6t 2 ⎜1+ sin θ ⎟ EFy Q f (16)
⎜ H ⎟
⎝ ⎠
Q f = 1.0 for a chord with no load, or a tension force, in accordance with AISC Specification Table K3.2.
⎡ 3 ( 4.00 in.) ⎤
2⎢ ⎥
Pn = 1.6 ( 0.349 in.) ⎢1+ sin 90° ⎥ ( 29,000 ksi ) ( 50 ksi ) (1.0 )
⎢ 8.00 in. ⎥
⎢⎣ ⎥⎦
= 587 kips
By applying the resistance factor of ϕ = 0.75, and the safety factor of Ω = 2.00, for this limit state (AISC Specification Section
J10.3), the available strength is:
LRFD ASD
ϕPn = 0.75 (587 kips) Pn 587 kips
=
= 440 kips Ω 2.00
= 294 kips
Limit State of Web Compression Buckling

H b sin θ = 4.00 in. > 0.25H = 0.25 (8.00 in.) = 2.00 in.
Hence, from Table 6, the member webs will be designed as compression members in accordance with AISC Specification
Chapter E, using K = 0.65.
Critical Buckling Stress, Fcr

Calculate the effective slenderness ratio (L c / r) using Equation 8, with K = 0.65, to determine applicable equation:
KL Lc ⎛H ⎞ 1
= = 3.46K −3 (8)
r r ⎝ t ⎠ sin
⎛ 8.00 in. ⎞ 1
= 3.46 ( 0.65) −3
⎝ 0.349 in. ⎠ sin 90°
= 44.8

E 29,000 ksi
4.71 = 4.71
Fy 50 ksi
= 113.4
KL E
Because < 4.71 , AISC Specification Equation E3-2 applies:
r Fy
⎛ Fy ⎞
Fcr = ⎜0.658 Fe ⎟ Fy (Spec. Eq. E3-2)
⎝ ⎠
where
π 2E
Fe = 2
⎛ Lc ⎞
⎝r⎠
π 2 ( 29,000 ksi )
=
( 44.8 )2
= 143 ksi
Hence,
50 ksi
⎛ ⎞
Fcr = ⎜ 0.658143 ksi ⎟ ( 50 ksi)
⎝ ⎠
= 43.2 ksi
Flexural Buckling of the Chord Sidewalls

The nominal compressive strength of the two sidewalls, by flexural buckling, is given by AISC Specification Equation E3-1:
Pn = Fcr Ag(Spec. Eq. E3-1)
where Ag for one sidewall is given by Equation 12. For two sidewalls,

⎛ H ⎞
Ag = 2 7.5t + b t (12)
⎝ sin θ ⎠
⎡ 4.00 in. ⎤
= 2 ⎢7.5 ( 0.349 in.) + ( 0.349 in.)
⎣ sin 90° ⎥⎦
2
= 4.61 in.
Therefore, the nominal strength of the two sidewalls in flexural buckling is:
Pn = ( 43.2 ksi ) (4.61 in.2 )
= 199 kips

By applying the resistance factor of ϕc = 0.90, and the safety factor of Ω c = 1.67, for this limit state (AISC Specification Section
E1), the available strength is:
LRFD ASD
ϕcPn = 0.90 (199 kips) Pn 199 kips
=
= 179 kips Ωc 1.67
= 119 kips
As expected, because the bearing length is greater than 0.25H, the connection resistance by web compression buckling governs.
The connection shown in Figure 12 has an identical configuration to Specimen X2, which, as indicated in Table 5, failed by
sidewall buckling.
ACKNOWLEDGMENTS Hb Overall height of rectangular HSS branch member or

plate, perpendicular to the plane of the connection,
Financial support for this project was provided by the Natu-
in.
ral Sciences and Engineering Research Council of Canada
(NSERC). Hollow structural sections used for two experi- K Effective length factor
ments reported herein were donated by Atlas Tube, Harrow,
Lc Effective length of member, in.
Ontario.
P Axial force, kips
SYMBOLS PD Axial force due to dead load, kips
A Cross-sectional area of rectangular HSS chord PL Axial force due to live load, kips
member, in.2
Pa Actual connection ultimate load, kips; required axial
Ab Cross-sectional area of rectangular HSS branch strength using ASD load combinations, kips
member, in.2
Pa,X1 Actual connection ultimate load of specimen X1,
Ag Cross-sectional area of element, in.2 kips
B Overall width of rectangular HSS chord member, Pa,X2 Actual connection ultimate load of specimen X2,
perpendicular to the plane of the connection, in. kips
Bb Overall width of rectangular HSS branch member or Pn Nominal connection strength, kips
plate, perpendicular to the plane of the connection,
Pu Required axial strength in tension or compression,
in.
using LRFD load combinations, kips
COV Coefficient of variation
Qf Chord-stress interaction parameter
E Modulus of elasticity of HSS member, ksi
Rn Nominal strength, kips
Fcr Critical stress of HSS chord member, ksi
d Full nominal depth of member, in.
Fu Ultimate tensile strength of HSS chord member, ksi
h Clear distance between flanges less the fillet or
Fub Ultimate tensile strength of branch member, ksi corner radius, in.
Fy Yield stress of HSS chord member, ksi k Distance from outer face of flange to web toe of
fillet for I-section, in.; outside corner radius for
Fyb Yield stress of branch member, ksi
rectangular HSS section, in.
Fyw Yield stress of web material, ksi
lb Bearing length of the load, measured parallel to the
H Overall height of rectangular HSS chord member, axis of the HSS member, in.
perpendicular to the plane of the connection, in.
lend Distance from the near side of the connecting branch
or plate to end of member, in.

t Wall thickness of rectangular HSS chord member, in. CSA (2013), CSA G40.20-13/G40.21-13, General Require-
ments for Rolled or Welded Structural Quality Steel/
tb Wall thickness of rectangular HSS branch member,
Structural Quality Steel, Canadian Standards Associa-
in.
tion, Toronto, Canada.
tf Thickness of flange, in. Davies, G. and Packer, J.A. (1987), “Analysis of Web Crip-
tw Thickness of web, in. pling in a Rectangular Hollow Section,” Proceedings
of the Institution of Civil Engineers, Part 2, Vol. 83,
Ω Safety factor pp. 785–798.
β Ratio of branch width to chord width, perpendicular Fisher, J.W., Ravindra, M.K., Kulak, G.L., and Galambos,
to the plane of the connection T.V. (1978), “Load and Resistance Factor Design Crite-
ria for Connectors,” Journal of the Structural Division,
χ Reduction factor for (column) buckling
ASCE, Vol. 104, No. 9, pp. 1,427–1,441.
εrup Strain at rupture IIW (2012), Static Design Procedure for Welded Hollow
εy Strain at yield Section Joints—Recommendations, 3rd Ed., IIW Doc.
XV-1402-12, International Institute of Welding, Paris,
τ Ratio of branch wall thickness to chord wall France.
thickness
ISO (2013), Static Design Procedure for Welded Hollow Sec-
γ Chord slenderness; the ratio of one-half the width to tion Joints—Recommendations, ISO 14346, International
wall thickness for rectangular HSS Organization for Standardization, Geneva, Switzerland.
η Load length parameter; the ratio of the length of Kuhn, J., Packer, J.A., and Fan, Y. (2019), “Rectangular Hol-
contact of the branch with the chord in the plane of low Section Webs under Transverse Compression,” Cana-
the connection to the chord width dian Journal of Civil Engineering, Vol. 46, pp. 810–827.
Liu, J. (2016), “Updates to Expected Yield Stress and Tensile
ϕ Resistance factor
Strength Ratios for Determination of Expected Member
θ Acute angle between the branch and chord, degrees Capacity in the 2016 AISC Seismic Provisions,” Engi-
neering Journal, AISC, Vol. 53, No. 4, pp. 215–227.
REFERENCES Lu, L.H., de Winkel, G.D., Yu, Y., and Wardenier, J. (1994),
AISC (2010), ANSI/AISC 360-10, Specification for Struc- “Deformation Limit for the Ultimate Strength of Hollow
tural Steel Buildings, American Institute of Steel Con- Section Joints,” Sixth International Symposium on Tubu-
struction, Chicago, Ill. lar Structures, Melbourne, Australia, pp. 341–347.
AISC (2016), ANSI/AISC 360-16, Specification for Struc- Newlin, D.E. and Chen, W.F. (1971), “Strength and Stabil-
tural Steel Buildings, American Institute of Steel Con- ity of Column Web in Welded Beam-to-Column Connec-
struction, Chicago, Ill. tions,” Fritz Engineering Laboratory Report No. 333.14,
Lehigh University, Bethlehem, Pa.
AISC (2017), Steel Construction Manual, 15th Ed., Ameri-
can Institute of Steel Construction, Chicago, Ill. Packer, J.A. (1984), “Web Crippling of Rectangular Hol-
low Sections,” Journal of Structural Engineering, ASCE,
AISI (2016), AISI S100-16, North American Specification
Vol. 110, No. 10, pp. 2,357–2,373.
for the Design of Cold-Formed Steel Structural Members,
American Iron and Steel Institute, Washington, D.C. Packer, J.A. (1987), “Review of American RHS Web Crip-
pling Provisions,” Journal of Structural Engineering,
ASCE (2016), Minimum Design Loads and Associated Cri-
ASCE, Vol. 113, No. 12, pp. 2,508–2,513.
teria for Buildings and Other Structures, ASCE/SEI 7-16,
American Society of Civil Engineers, Reston, Va. Packer, J.A., Wardenier, J., Zhao, X.-L., van der Vegte, G.J.,
and Kurobane, Y. (2009), Design Guide for Rectangu-
ASTM (2018), ASTM A500/A500M-18, Standard Specifi-
lar Hollow Section (RHS) Joints under Predominantly
cation for Cold-Formed Welded and Seamless Carbon
Static Loading, CIDECT Design Guide No. 3, 2nd Ed.,
Steel Structural Tubing in Rounds and Shapes, American
CIDECT, Geneva, Switzerland.
Society for Testing and Materials, West Conshohocken,
Pa. Ravindra, M.K. and Galambos, T.V. (1978), “Load and
Resistance Factor Design for Steel,” Journal of the Struc-
CEN (2005), Eurocode 3: Design of Steel Structures—Part
tural Division, ASCE, Vol. 104, No. 9, pp. 1,337–1,353.
1-8: Design of Joints, EN 1993-1-8:2005, European Com-
mittee for Standardization, Brussels, Belgium.

Roberts, T.M. (1981), “Slender Plate Girders subject to Edge
Loading,” Proceedings of the Institution of Civil Engi-
neers, Part 2, Vol. 71, No. 3, pp. 805–819.
Wei, F. (2019), “Experimental Study of Laterally Offset RHS
X-Connections in Branch Axial Compression,” M.A.Sc.
Thesis, University of Toronto, Toronto, Canada.

Critical Temperature of Axially Loaded Steel Members
with Wide-Flange Shapes Exposed to Fire
ANA SAUCA, RACHEL CHICCHI, CHAO ZHANG, and LISA CHOE
ABSTRACT
This paper presents closed-form equations that were developed to evaluate critical temperatures of structural steel compression and ten-
sion members exposed to fire. The deterministic approach involved a parametric study using finite element simulations in order to iden-
tify influencing factors—for example, mechanical properties of steel, member slenderness, and axial load ratios. Statistical models were
employed to develop closed-form equations representing the best fit of numerical results. A comparison with experimental column test data
indicates that the proposed equation for compression members provides a conservative lower bound (16% lower on average) relative to the
test data at load ratios greater than 0.3. A sensitivity study was also performed to further explore uncertainty in predicted critical tempera-
tures due to variability of axial load ratios. For both compression and tension members, the ambient-temperature yield stress of steel, Fy,
has a great impact on determination of axial load ratios, subsequently influencing the overall accuracy of the critical temperature estimated
by the proposed equations. The applicability of the proposed equations is limited to wide-flange steel members that are simply supported,
concentrically loaded, and exposed to uniform heating.
Keywords: critical temperature, structural steel, compression, tension, fire.
INTRODUCTION levels, semi-rigid support conditions, and both member and

section slenderness.
Background Prescriptive methods have provided little information
regarding the high-temperature strength and associated
I n the United States, fire resistance design of load-carrying
steel members (beams and columns) in steel-framed
buildings is mainly achieved through compliance with
failure modes of steel members exposed to fire. As an alter-
native engineering approach, AISC Specification Appen-
dix 4 (AISC, 2016b) provides high-temperature member
prescriptive provisions in the International Building Code
strength equations for the limit states of flexural buck-
(ICC, 2009). In this approach, fireproofing insulation is
ling and lateral-torsional buckling. To calculate member
applied to exposed steel so that the steel does not exceed
strengths at elevated temperature, users need to define the
the critical temperature under standard fire conditions for
temperature of interest as an input, which must be greater
a minimum specified duration (known as a fire-resistant
than 392°F (200°C), based on heat transfer analyses or
rating). According to the American Society for Testing
engineering judgments. These equations are less practical
and Materials (ASTM) E119 standard (ASTM, 2019), the
for solving the critical temperature at which the member
critical temperature of exposed steel members in a stan-
demand exceeds its capacity because iteration with increas-
dard fire test is 1000°F (538°C) for columns and 1100°F
ing temperatures is required (Sauca et al., 2019).
(593°C) for beams, determined as the average temperature
In Europe, the evaluation of critical temperatures of axi-
of all measurement points. However, these limiting tem-
ally loaded steel members was of interest beginning in the
peratures seldom account for the effects of imposed load
late 1970s. Kruppa (1979) defined “critical” or “collapse”
temperature as the temperature at which the structure
cannot assume its function and proposed a critical tem-
perature equation for steel columns using the temperature-
Ana Sauca, Post-doc Research Engineer, Danish Institute of Fire and Security dependent axial stress and buckling coefficient. Rubert and
Technology, Denmark. Email: [email protected] Schaumann (1988) used finite element models for calculat-
Rachel Chicchi, Assistant Professor, University of Cincinnati, Cincinnati, Ohio. ing critical temperature of steel columns. The analytical
Email: [email protected]
results were compared with 50 full-scale column tests and
Chao Zhang, Guest Researcher, National Institute of Standards and Technol-
showed good correlation at temperatures in the range of
ogy, Gaithersburg, Md. Email: [email protected]
390°F (200°C) to 1300°F (700°C) and utilization (demand-
Lisa Choe, Research Structural Engineer, National Institute of Standards and
Technology, Gaithersburg, Md. Email: [email protected] (corresponding) to-capacity) ratios of 0.2 to 0.6.
Neves (1995) further explored the critical temperature
of restrained steel columns analytically, with three column
Paper No. 2019-21 slenderness values (40, 80, and 120) and eccentricity of the

applied load. Due to the variety of parameters being con- critical temperatures were evaluated, including various
sidered, a critical temperature equation was not proposed. axial load levels, steel grades, and section compactness
Similarly, Franssen (2000) applied an arc-length numerical and member slenderness at ambient temperature. The use
technique to calculate the collapse temperature of columns. of proposed equations presented herein should be limited
Wang et al. (2010) evaluated the critical temperature of to wide-flange steel members simply supported, concen-
restrained steel columns using a finite element ABAQUS trically loaded, and exposed to uniform heating. Future
model (Smith, 2009) with two-dimensional beam elements. work will include the effects of thermal restraints as well
Their study indicated that the section geometry had very as thermal gradients through the section depth and along
limited effects on the column critical temperature, and the the member length.
critical temperature of a restrained column can be obtained
by making a reduction in corresponding values of columns
NUMERICAL ANALYSES
without axial restraint.
The European standards provide critical temperature
Test Bed
equations or tabulated data for steel members. For steel
members “without instability phenomena” (e.g., tension or The critical temperature of axially loaded steel columns
flexural yielding), the critical temperature is only a function with wide-flange rolled shapes was evaluated using the
of a utilization ratio for fire conditions (CEN, 2005). This finite element method (FEM). In this study, a total of 900
equation is very similar to an inverse of the temperature- FEM models were analyzed in combination with various
dependent yield strength of structural steel. For steel col- ranges of parameters summarized in Table 1. Five differ-
umns, however, only tabulated forms (e.g., Vassart et al., ent wide-flange rolled shapes, including W8×31, W10×68,
2014; BSI, 2005) are available to evaluate critical tempera- W14×22, W14×90, and W14×211, were used in this study.
tures, depending upon the member slenderness and utiliza- With the exception of the W14×22, all other shapes are com-
tion ratio. Despite all the limitations (i.e., applicability only pact for compression at ambient temperature. In addition,
under standard fires, uniform distribution of temperatures two American standard grades of structural steel shapes,
across the section and length, and simplified boundary con- including Fy = 50 ksi and Fy = 36 ksi, are considered, where
ditions), the critical temperature method would remain as a Fy is the minimum specified yield stress. Effective slen-
useful tool to evaluate the fire resistance of load-carrying derness ratios, L c/ r, range from 20 to 200, and applied
steel members (Milke, 2016). load ratios vary from 0.1 to 0.9. The load ratio is defined
as the axial demand at elevated temperatures, Pu, normal-
Objectives, Scope, and Limitations ized by the nominal capacity at ambient temperature, Pna.
The demand for fire condition can be determined from the
The significance of the critical temperature method lies in
load combination for extraordinary events, 1.2 × dead load
its simplicity and the useful information obtained about a
+ 0.5 × live load + AT, where AT is the force and defor-
structural member exposed to varying temperatures dur-
mation induced by fire effects (ASCE, 2016). In this study,
ing a fire event. To date, however, a critical temperature
all investigated members were assumed to be simply sup-
method is not available in AISC Specification Appendix 4
ported, concentrically loaded, and exposed to uniform heat-
(AISC, 2016b). The objective of the study presented herein
ing; therefore, the magnitude of AT was assumed to be zero.
was to develop closed-form solutions that can be used to
The nominal capacity at ambient temperature, Pna, can be
evaluate critical temperatures of axially loaded steel mem-
calculated using AISC Specification Section E3.
bers exposed to fire. The methodology adopted in this study
Numerical models of columns were developed using
included (1) a parametric study using 900 finite element
three-dimensional shell elements. Each model was dis-
models to identify the influencing variables for determina-
cretized into 50 elements along the member length and 8
tion of critical temperatures of steel members at elevated
elements each for the flange and the web. The FEM solution
temperatures, (2) three-dimensional regression analyses to
with this element size was converged with the maximum
develop a closed-form equation that represents the best fit
error of about 2%, based on the mesh density study presented
of numerical results with given ranges of the parameters
in Sauca et al. (2019). Linear kinematic constraints were
considered in this study, (3) comparison of the critical tem-
applied to both the flanges and web at each end in order to
perature predicted using the proposed equation with test
enforce rigid planar behavior. The column ends were sim-
data in literature, and (4) a sensitivity study to estimate
ply supported. An axial force was applied to the centroid
uncertainty in critical temperatures computed using pro-
of the end section. An initial displacement at midspan was
posed equations.
taken as the 1/1000 of the column length to simulate global
The scope of this study focused on the critical tempera-
imperfections (initial sweep). Local geometrical imperfec-
ture of structural steel tension and compression members
tions were implemented by scaling a sinusoidal deformation
with wide-flange rolled shapes. The parameters influencing
of the cross sections using elastic buckling analyses. The

Table 1. Test Parameters Used in Numerical Analyses
Shape Fy /
Lc r /
Pu Pna
W8×31
W10×68
36 ksi (250 MPa) 20 to 200 0.1 to 0.9
W14×22
50 ksi (350 MPa) (increment: 20) (increment: 0.1)
W14×90
W14×211
scaled value was taken as the larger of a web out-of-flatness • Applied load level: The critical temperature is affected
equal to the ratio of the section depth over 150 (Kim and by the magnitude of applied loads. The reduction in
Lee, 2002) or a tilt in the compression flanges taken as the critical temperature can reach nearly 80% between the
ratio of the flange width over 150 (Zhang et al., 2015). No load ratio of 0.1 and 0.9 and 20% on average at each
residual stresses were applied because their influence is increment of 0.1. Larger scatter of the results is observed
limited at elevated temperature (Vila Real et al., 2007). The for the models with the load ratio between 0.5 and 0.8, as
Eurocode 3 (CEN, 2005) temperature-dependent stress- shown by the error bars in Figure 1(a), due to variation
strain relationship was employed, whereas no thermal creep in member slenderness. The critical temperature versus
model was incorporated explicitly. applied load relationship shows a very good linear fit,
In order to estimate critical temperatures of columns similar to an empirical relationship presented in Choe et
using FEM models, an axial load as a fraction of Pna was al. (2011).
applied at ambient temperature, and then the member tem- Figure 2 shows critical temperatures of steel columns
perature was increased monotonically until force equi- relative to load ratio with (1) all five shapes and two dif-
libriums could not be achieved. The maximum value of ferent steel grades and (2) W14×22 and W14×90 columns
temperature achieved from each FEM model was defined with Fy = 50 ksi. Both graphs considered the slenderness
as a critical temperature. ratios of 20, 40, and 100. Some discussions on the effect of
the ambient yield stress, Fy, and the section compactness
Numerical Results are as follows.
Figure 1 shows the critical temperature, Tcr , of steel col- • Ambient yield strength: The variation in critical
umns predicted using the finite element models with Fy = temperatures predicted using two different steel grades
50 ksi (350 MPa), where the dotted lines indicate the lin- (36 ksi versus 50 ksi) is about 1% on average. This is to
ear regression of these predicted results. Figure 1(a) shows be expected as the buckling behavior of columns with the
the average critical temperature of columns as a function slenderness ratio greater than 40 (i.e., medium-length to
of a load ratio. The error bars indicate the standard devia- slender columns) is mainly affected by low strain levels
tion of the results varying with five different shapes and all (less than 0.05% strain) and temperature-dependent
slenderness ratios (L c/ r = 20 to 200) at the same load level. elastic modulus (Choe et al., 2017).
Figure 1(b) shows the relationship of the average critical
temperature of all five columns versus the slenderness ratio • Section geometry: Between two different wide-flange
at four different load ratios (Pu/Pna) of 0.1, 0.3, 0.6, and 0.9. shapes, the variation in critical temperatures is over 10%
As shown, the critical temperature appears to be linearly for short columns subjected to large axial loads (i.e., a
decreasing with both increasing load ratios and increasing slenderness ratio less than 60 and a load ratio greater
slenderness ratios. However, the critical temperature is less than 0.6). The critical temperature variation for slender
sensitive to the member slenderness at the same load level. columns subjected to small axial loads is below 5%.
Some statistical results and discussions on the effect of
member slenderness and applied load levels are as follows.
PROPOSED CLOSED-FORM EQUATION
• Member slenderness: The reduction in critical
temperatures with increasing slenderness ratios is Compression Members
influenced by the applied load level. At load ratios
smaller than 0.5, the critical temperature is reduced The numerical results from 900 finite-element models were
by about 10% between the slenderness ratio of 20 and used to develop a closed-form equation that predicts criti-
200. At higher load ratios, the critical temperature can cal temperatures of steel columns as a function of member
reduce by 30% to 60% for the L c/r ratio of 20 to 200. This slenderness and load ratio. The three-dimensional linear
reduction is not proportional to load ratios. polynomial model, as shown in Figure 3, was employed

based on the results from the parametric study presented Lc P
previously. Equations 1 and 2 show the resulting best linear Tcr = 1580 − 0.814 − 1300 u in °F (2)
r Pna
fit equation in °C and °F, respectively, with the R-square
value of 0.97. Figure 4 shows a comparison of critical temperatures cal-
Lc P culated using the proposed equation with those estimated
Tcr = 858 − 0.455 − 722 u in °C (1) using various methods, including FEM models, the ASTM
r Pna E119 limiting temperature of columns, and the AISC Speci-
fication Appendix 4 equation. In Figure 4(a), the results of
900 1652 900 1652

Tcr (°F) = -1303.7(Pu/Pna) + 1488.7 0.1 0.3 0.6 0.9
800 R2 = 0.9925 1472 800 1472
700 1292 700 1292
600 1112 600 1112
Tcr (°C)
500 932 Tcr (°F)
Tcr (°C)
Tcr (°F)
500 932
400 752 400 752
300 572 300 572
200 392 200 392
100 212 100 212
0 32 0 32
0 0.2 0.4 0.6 0.8 1 0 30 60 90 120 150 180 210
Pu /Pna Lc /ry
(a) (b)
Fig. 1. Average critical temperatures for columns predicted using FEM models of five shapes
/
with Fy = 50 ksi as a function of (a) load ratio (P u Pna) and (b) member slenderness (Lc r y). /
900 1652 900 1652

36 ksi W14×22
800 1472 800 1472
50 ksi
W14×90
700 1292 700 1292
600 1112 600 1112
Tcr (°C)
Tcr (°C)
Tcr (°F)
Tcr (°F)
500 932 500 932
400 752 400 752
300 572 300 572
200 392 200 392
100 212 100 212
0 32 0 32
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Pu / Pna Pu / Pna
(a) All five shapes with Fy = 36 ksi and 50 ksi (b) W14×22 and W14×90 shapes with Fy = 50 ksi
Fig. 2. Predicted critical temperatures of columns with slenderness ratios of 20, 40, and 100.

Fig. 3. A three-dimensional linear curve fit of 900 FEM models of columns.
900 1652 900 1652

FEM (mean-std) AISC eqn
800 FEM (mean+std) 1472 800 Proposed eqn 1472
Proposed eqn
700 1292 700 1292
E119
600 1112 600 1112
Tcr (°C)
Tcr (°C)
Tcr (°F)
500 932 500 932 Tcr (°F)
400 752 400 752

300 572 300 572
200 392 200 392

100 212 100 212
0 32 0 32
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Pu /Pna Pu /Pna
(a) FEM results and ASTM E119 limiting temperature (b) AISC Specification Appendix 4 equation results
Fig. 4. Comparisons of the proposed column equation.

FEM models are presented with two lines: the upper bound boundary conditions and were concentrically loaded (i.e.,
as mean values plus standard deviations (std) and the lower an eccentricity of axial loading was less than the 1/1000
bound as mean values minus standard deviations. The stan- of the column length) at elevated temperatures. In this data
dard deviation incorporates the total variation in the FEM set, the ambient-temperature yield stress ranged from 32 ksi
data resulting from the range in parameters described in (220 MPa) to 60 ksi (400 MPa), and effective slenderness
Table 1 at each load level. The error bars plotted with the ratios varied from 30 to 137.
critical temperature predicted using Equation 1 indicate the Figure 5 shows a comparison of the column test data with
standard deviation due to slenderness ratio ranges from 20 predicted critical temperatures using Equation 1 and with
to 200. Overall, the proposed equation compares reason- the linear regression of the data itself. Overall, the proposed
ably well with the FEM results. With this equation, the equation provides a conservative lower bound of the test
load-bearing capacity of steel columns is approximately results. For the specimens with load ratios greater than 0.3,
40% of the ambient capacity at the ASTM E119 limiting the calculated critical temperatures are approximately 16%
temperature of 1000°F (540°C). lower than the measured values on average. For load ratios
Figure 4(b) gives a comparison with critical temperatures less than 0.2, Equation 1 slightly overestimates the critical
estimated using the AISC Specification Appendix 4 flex- temperature by 4%.
ural buckling strength equation, Equation A-4-2. A detailed
description of computation methods, which required an iter- Tension Members
ation process, is presented in Sauca et al. (2019). The error
bars in this figure indicate the standard deviation resulted Critical temperatures of uniformly heated steel members in
from a variety of steel shapes and slenderness ratios con- tension have a dependency of high-temperature mechanical
sidered in this study. For columns with load ratios less than properties, such as temperature-dependent yield stress and
0.6, the proposed equation also adequately predicts critical ultimate tensile strength. This paper also suggests a critical
temperatures, with 2% difference on average. At load ratios temperature equation for tensile yielding in gross sections of
equal to or greater than 0.6, however, the proposed equa- a steel member as a function of imposed tension loads, Tu, at
tion may overestimate critical temperatures estimated using elevated temperature normalized by the nominal capacity,
AISC Specification Equation A-4-2. Tna , at ambient temperature. As shown in Figure 6, the criti-
The efficacy of Equation 1 was examined by compar- cal temperature equation is an inverse relationship of the
ing predicted critical temperatures with observed critical AISC Specification temperature-dependent retention fac-
temperatures from previous experimental studies (Franssen tors for yield stress, k y , essentially the same as the Eurocode
et al., 1996; Ali et al., 1998; Choe et al., 2011) of steel col- 3 (CEN, 2005) retention factors. The logarithmic regres-
umns that had similar properties used for the present study. sion model was employed similar to the Eurocode 3 criti-
Test data used for this comparison included 36 wide-flange, cal temperature equation for members “without instability
hot-rolled column specimens that had simply supported phenomena.” Equations 3 and 4 show the best fit equation
900 1652
800 1472
700 1292
600 1112
Tcr (°C)
Tcr (°F)
500 932
400 752
300 572
Franssen et al. (1996)
200 Ali et al. (1998) 392
Choe et al. (2011)
100 Linear (Proposed eqn) 212
Linear (All test data)
0 32
0 0.2 0.4 0.6 0.8 1
Lc/r
Fig. 5. A comparison of critical temperatures of columns calculated using Equation 1 with experimental test data.

Table 2. Statistical Data for Uncertainties (Takagi and Deierlein, 2007)
Variable Mean CV Std
Fy 50 ksi (350 MPa) 0.10 5 ksi (35 MPa)
E 29,000 ksi (200 GPa) 0.06 1,740 ksi (12 GPa)
DL 103% unfactored 0.10 a
LL 25% unfactored 0.60 b
a: The standard deviation for DL is taken as the mean load × 1.025 × 0.10.
b: The standard deviation for LL is taken as the mean load × 0.25 × 0.60.
in °C and °F, respectively, with the R-square value of 0.99. Although uncertainty in geometric properties are present
For the use of these equations, the load ratio, Tu/ Tna, must in the proposed equation, such as column length, L c, and
be greater than or equal to 0.01. the radius of gyration, r, this effect was neglected with the
assumption that compliance of standard fabrication toler-
⎛ Tu ⎞
Tcr = 435 − 170 ln in °C (3) ances specified in the AISC Code of Standard Practice
⎝ Tna ⎠ for Steel Buildings and Bridges (AISC, 2016a) would not
result in notable critical temperature changes. A compari-
⎛ Tu ⎞
Tcr = 816 − 306ln in °F (4) son of the influence of each parameter (Fy, E, DL, and LL)
⎝ Tna ⎠ on the variation in the critical temperature was calculated
by considering reasonable upper and lower bounds of each
variable. Each parameter was evaluated at the mean ±1
ESTIMATED UNCERTAINTY OF standard deviation (std) that represents 68% confidence
CLOSED-FORM EQUATIONS intervals. The mean ±2 standard deviations (to represent a
95% confidence interval) were also reported. A normal dis-
Compression Members tribution of each variable was assumed.
Because the proposed closed-form solution was developed Statistical properties of the investigated variables are
using a deterministic approach, which does not account for summarized in Table 2, based on work from Takagi and
uncertainty in estimation of applied load ratios, Pu/Pna, sen- Deierlein (2007), who proposed the member strength equa-
sitivity was examined with variability in mechanical prop- tion for gravity columns at elevated temperature in AISC
erties of steel (Fy and elastic modulus, E) and the magnitude Specification Appendix 4. The mean values and coeffi-
of design loads (e.g., dead load, DL, and live load, LL). cients of variation (CV) were determined from statistical
data obtained by Ellingwood et al. (1980). The percentages
1400 2552
ky
1200 Proposed eqn 2192
1000 1832
800 1472
Tcr (°C)
Tcr (°F)
600 1112
400 752
200 392
0 32
0 0.2 0.4 0.6 0.8 1
Pu/Pna
Fig. 6. Critical temperature versus load ratio relationship of tension members.

for DL and LL were obtained from load surveys using prob- represents the critical temperatures determined using the
abilistic load models. They represent the mean values of the proposed closed-form equation [Equation (1)]. The dashed
unfactored design loads for dead and live loads relative to lines represent the critical temperatures calculated with Fy
the nominal design loads in the American National Stan- adjusted by a positive and negative standard deviation. The
dard A58. The standard deviation (std) for each variable uncertainty in the critical temperature estimated using the
was calculated as the mean times the coefficient of varia- propose equation is more pronounced at lower L c/r ratios
tion (CV), as shown in Table 2. Ambient temperature values and at higher load ratios where Euler buckling does not
of Fy and E were used to calculate the mean and CV values likely occur. At higher L c/r levels, where elastic buckling
due to a lack of statistical data on their high-temperature of the column would dominate, the impact of a change in
values. Fy, appears to be minimal and becomes negligible for L c/r
A range of columns used in this study (W8×31, W14×90, ratios of 120 and greater. At a load ratio (Pu/Pna) of 0.6, the
and W14×211 with Fy = 50 ksi) were examined for sensitiv- uncertainty in estimated critical temperatures is about 20%
ity. The change in critical temperature due to uncertainty at L c/r = 40 and about 10% at L c/r = 80 due to ±1 std of Fy.
of 1 standard deviation is consistent across all compact These percentages represent the ratio of change in critical
column shapes, so the results presented represent all of the temperature due to uncertainty relative to the closed-form
compact shapes listed above. Figure 7 shows the change in proposed equation without uncertainty.
critical temperature for the W14×211 column with L c/r = 40 Figure 8 shows the variation in estimated critical temper-
and L c/r = 80 due to uncertainty in Fy. The solid line ature for the W14×211 column with L c/r = 40 and L c/r = 120
900 1652 900 1652

800 ± 1 std of Fy 1472 800 ± 1 std of Fy 1472
W14×211 W14×211
700 Lc/r = 40 1292 700 1292
Lc/r = 80
600 1112 600 1112
Tcr (oC)
Tcr (oC)
Tcr (oF)
Tcr (oF)
500 932 500 932

400 752 400 752
300 572 300 572
200 392 200 392
100 212 100 212
0 32 0 32
0.5 0.6 0.7 0.80.2 0.3 0.4 0.2 0.3 0.4 0.5 0.6 0.7 0.8
P u / Pna Pu / Pna
(a) Lc r = 40 / /
(b) Lc r = 120
Fig. 7. Sensitivity of calculated critical temperatures of a W14×211 column due to uncertainty in Fy.
900 1652 900 1652

800 ±1 std of E 1472 800 ±1 std of E 1472
700 W14×211 1292 W14×211
Lc/r = 40 700 1292
Lc/r = 120
600 1112 600 1112
Tcr (oC)
Tcr (oF)
500 932
Tcr (oC)
Tcr (oF)
500 932
400 752 400 752
300 572 300 572
200 392 200 392
100 212 100 212
0 32
0 32
0.5 0.6 0.7 0.8 0.2 0.3 0.4
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Pu / Pna Pu / Pna
Lc r = 40
(a) / /
(b) Lc r = 120
Fig. 8. Sensitivity of calculated critical temperatures of a W14×211 column due to uncertainty in E.

due to uncertainty in the elastic modulus, E, in the calcula- 50 psf for the composite slab plus 15 psf for superimposed
tion of Pna. The uncertainty in estimated critical temper- dead loads such as ceilings and ductwork and piping for
ature is most pronounced at both higher slenderness and utilities. The live load values of 50 psf and 100 psf repre-
higher load ratios where elastic buckling likely governs. In sent average and high levels of live loading, respectively.
this study, the maximum uncertainty is observed for slen- According to ASCE/SEI 7 (2016), 50 psf represents live
der columns (L c/r ≥ 120) and the applied load ratio of 0.8. loads for office spaces, while 100 psf represents lobbies
For these columns, the uncertainty in critical temperatures and other assembly areas. The final DL/LL ratio that was
can be as large as 30%. However, for stockier columns used was 0.33. This ratio is given in the AISC Specification
(L c/r ≤ 40), this uncertainty in critical temperatures associ- Section A1 Commentary (AISC, 2016b) as the ratio that
ated with ±1 std of E becomes very minor, less than 3%. results in the same reliability between the ASD and LRFD
Sensitivity due to uncertainty in applied loads under fire design methods. Using these ratios, the dead and live loads
conditions (Pu) was determined by considering three differ- on the column were determined by assuming that the
ent DL/LL ratios selected based on engineering judgment. demand-to-capacity ratio for each column at ambient con-
The first DL/LL ratio was 0.65, which was determined by ditions is equal to 1.0 for the ambient load combination,
assuming a dead load of 65 psf and a live load of 100 psf. 1.2DL + 1.6LL. Converting to the fire load combination
The second DL/LL ratio of 1.3 was calculated using the (1.2DL + 0.5LL), this equates to a Pu /Pna ratio of approxi-
same dead load of 65 psf but a live load of only 50 psf. The mately 0.4, 0.5, and 0.6 for DL/LL ratios of 0.33, 0.65, and
65-psf dead load was selected based on the assumption of 1.3, respectively. Figure 9(a) shows the change in critical
(a) Dead load (DL)
(b) Live load (LL)
/
Fig. 9. Sensitivity of the change in critical temperature due to uncertainty. Note: ΔTcr is presented (not Tcr ); ΔTcr (°F) = 9 5[ΔTcr (°C)].

temperature due to uncertainty in dead load, while Fig- a total of 900 FEM models were analyzed in combination
ure 9(b) represents the change in critical temperature due with various ranges of parameters, including five different
to live load uncertainty. These results show that critical wide-flange rolled shapes made of two American standard
temperatures are more influenced by a higher DL/LL ratio grades of structural steel, member slenderness ratios from
for dead load variability and a lower DL/LL for live load 20 to 200, and applied load ratios varying from 0.1 to 0.9.
variability. These critical temperature changes (ΔTcr) are Load ratios represent the axial demand at elevated tempera-
independent of the L c/ r ratio of the column. The maximum tures, Pu, normalized by the nominal capacity at ambient
change in critical temperature due to uncertainty of 1 stan- temperature, Pna.
dard deviation in DL and LL is 59°F and 44°F, respectively. The parametric study indicates that the most influential
parameters for critical temperature of columns are member
Tension Members slenderness and applied load ratios. A closed-form equa-
tion predicting critical temperatures of steel columns with
The same variables (Fy, DL, and LL) were studied for ten-
these two factors is proposed based on curve-fitting of the
sion members to determine the sensitivity of the closed-
FEM results using the three-dimensional linear polyno-
form equation. There is no sensitivity in the equation to a
mial model. With this equation, the load-bearing capac-
change in modulus of elasticity (E). A W14×22 shape was
ity of steel columns is approximately 40% of the ambient
chosen to demonstrate the sensitivity. Figure 10 summarizes
capacity at the ASTM E119 limiting temperature of 1000°F
the sensitivity by showing the change in critical temperate
(540°C). At load ratios less than 0.6, the proposed equation
for ±1 std and ±2 std of each parameter, estimated using CV
accurately predicts critical temperatures determined using
values in Table 2. The same DL/LL ratios of 0.33, 0.65, and
the high-temperature flexural buckling strength equation
1.3 were also used. This comparison shows that the great-
in AISC Specification Appendix 4, whereas it may overes-
est change in critical temperatures is due to a change in
timate critical temperatures (10% difference or greater) at
the yield stress of the material. At 1 standard deviation, the
load ratio greater than or equal to 0.6. The proposed equa-
change in temperature is −32°F to 29°F, and at 2 standard
tion also provides a conservative lower bound (16% lower
deviations, it is −68°F to 56°F. The variation in DL with a
on average) of the published test data for the specimens
high DL/LL ratio produces the second highest sensitivity.
with load ratios greater than 0.3. This result considers col-
umn failure by flexural buckling at elevated temperature.
SUMMARY AND CONCLUSIONS A critical temperature equation for tension members is
also proposed using the logarithmic regression model for
This paper presents the development of closed-form solu-
the case with tensile yielding only. This equation is essen-
tions to evaluate critical temperatures of axially loaded
tially the same as an inverse relationship of the AISC
steel members exposed to fire. For compression members,
Fig. 10. Sensitivity of the change in critical temperature of tension members due to
/
uncertainty in parameters. Note: ΔTcr is presented (not Tcr ); ΔTcr (°F) = 9 5[ΔTcr (°C)].

Specification temperature-dependent retention factors for AISC (2016b), Specification for Structural Steel Buildings,
yield stress. ANSI/AISC 360-16, American Institute of Steel Con-
A sensitivity study was performed to estimate the uncer- struction, Chicago, Ill.
tainty in critical temperatures predicted using the proposed ASCE (2016), Minimum Design Loads and Associated Cri-
equations due to the variability in axial load ratios. The teria for Buildings and Other Structures, ASCE/SEI 7-16,
results show that these critical temperatures depend on the American Society of Civil Engineers, Reston, Va.
ambient temperature Fy and E as well as design loads (DL
ASTM (2019), Standard Methods of Fire Test of Building
and LL). The variation in Fy is the most influential factor
Construction and Materials, ASTM E119-19, ASTM
among other uncertain variables for critical temperatures
International, West Conshohocken, Pa.
of both compression and tension members. The influence
of Fy uncertainty is apparent in stout columns with a low BSI (2005), UK National Annex to Eurocode 3. Design of
slenderness ratio. All results show that variations in critical Steel Structures. General Rules. Structural Fire Design,
temperature are relatively minor for uncertainty of 1 stan- BS NA EN 1993-1-2, United Kingdom.
dard deviation, particularly for typical columns, which are Choe, L., Varma, A.H., Agarwal, A., and Surovek, A. (2011),
assumed to have load ratios of approximately 0.6 and L c/r “Fundamental Behavior of Steel Beam Columns and Col-
ratios of approximately 40 to 60. Consideration of material umns under Fire Loading: Experimental Evaluation,”
sensitivity should be implemented for load ratios beyond Journal of Structural Engineering, Vol. 137, pp. 954–966.
0.6. Choe, L., Zhang, C., Luecke, W.E., et al. (2017), “Influence
The findings and equations from this study are limited of Material Models on Predicting the Fire Behavior of
to the range of parameters included in the numerical evalu- Steel Columns,” Fire Technology, Vol. 53, pp. 375–400,
ation. Future studies will be conducted to further incorpo- https://doi.org/10.1007/s10694-016-0568-4.
rate probabilistic analyses into the current deterministic
CEN (2005), Eurocode 3: Design of Steel Structures—Part
approach, accounting for the effects of thermal restraints
1-2: General Rules—Structural Fire Design, Standard
as well as thermal gradients through the section depth and
EN 1993-1-2, European Committee for Standardization,
along the member length.
Luxembourg.
Ellingwood, B., Galambos, T.V., MacGregor, J.G., and
ACKNOWLEDGMENT Cornell, C.A. (1980), “Development of a Probability-
Based Load Criterion for American National Standard
Valuable comments and input on this work were provided
A58,” National Bureau of Standards Special Publication
by the AISC Committee on Specifications Task Committee
No. 577, Washington, D.C.
8, Design for Fire.
Franssen, J.M., Schleich, J.B., Cajot, L.G., and Azpiazu,
W. (1996), “A Simple Model for the Fire Resistance of
DISCLAIMERS Axially-Loaded Members—Comparison with Experi-
Certain commercial entities, equipment, products, software, mental Results,” Journal of Construction Steel Research,
or materials are identified in this paper in order to describe Vol. 37, pp. 175–204.
a procedure or concept adequately. Such identification is Franssen, J.M. (2000), “Failure Temperature of a System
not intended to imply recommendation or endorsement by Comprising a Restrained Column Submitted to Fire,”
the National Institute of Standards and Technology, nor is Fire Safety Journal, Vol. 34, No. 2, pp. 191–207.
it intended to imply that the entities, products, software, ICC (2009), International Building Code, International
materials, or equipment are necessarily the best available Code Council, Falls Church, Va.
for the purpose.
Kim, S. and Lee, D. (2002), “Second-Order Distributed
Plasticity Analysis of Space Steel Frames,” Engineering
REFERENCES Structures, Vol. 24, pp. 735–744.
Ali, F.A., Shepherd, P., Randall, M., Simms, I.W., O’Connor, Kruppa, J. (1979), “Collapse Temperature of Steel Struc-
D.J., and Burgess, I. (1998), “The Effect of Axial Restraint tures,” Journal of the Structural Division, ASCE,
on the Fire Resistance of Steel Columns,” Journal of Con- Vol. 105, No. ST9, September.
struction Steel Research, Vol. 46, pp. 305–306. Milke, J.A. (2016), “Analytical Methods for Deter-
AISC (2016a), Code of Standard Practice for Steel Build- mining Fire Resistance of Steel Members,” In:
ings and Bridges, ANSI/AISC 303-16, American Institute Hurley M.J. et al. (eds.), SFPE Handbook of Fire Pro-
of Steel Construction, Chicago, Ill. tection Engineering, Springer, New York, N.Y., https://doi
.org/10.1007/978-1-4939-2565-0_53.

Neves, I.C. (1995), “The Critical Temperature of Steel Col- Vassart, O., Zhao, B., Cajot, L.G., Robert, F., Meyer, U., and
umns with Restrained Thermal Elongation,” Fire Safety Frangi, A. (2014), “Eurocodes: Background & Applica-
Journal, Vol. 24, pp. 211–227. tions Structural Fire Design,” JRC Science and Policy
Rubert, A. and Schaumann, P. (1988), “Critical Tempera- Reports, European Union.
tures of Steel Columns Exposed to Fire,” Fire Safety Vila Real, P.M.M., Lopes da Silva, N.L.S., and Franssen,
Journal, Vol. 13, pp. 39–44. J.M. (2007), “Parametric Analysis of the Lateral—
Sauca, A., Zhang, C., Seif, M., and Choe, L. (2019), “Axi- Torsional Buckling Resistance of Steel Beams in Case of
ally Loaded I-Shaped Steel Members: Evaluation of Criti- Fire,” Fire Safety Journal, Vol. 42, pp. 461–524.
cal Temperature Using ANSI/AISC-360 Appendix 4 and Wang, P., Wang, Y.C., and Li, G.Q. (2010), “A New Design
Finite Element Model,” Proceedings of the Annual Stabil- Method for Calculating Critical Temperatures of
ity Conference, Structural Stability Research Council, St. Restrained Steel Column in Fire,” Fire Safety Journal,
Louis, Mo., April 2–5. Vol. 45, pp. 349–360.
Smith, M. (2009), ABAQUS/Standard User’s Manual, Ver- Zhang, C., Choe, L., Seif, M., and Zhang, Z. (2015), “Behav-
sion 6.9, Simulia, Providence, R.I. ior of Axially Loaded Steel Short Columns Subjected
Takagi, J. and Deierlein, G.G. (2007), “Collapse Perfor- to a Localized Fire,” Journal of Constructional Steel
mance Assessment of Steel-Frames Buildings under Research, Vol. 111, pp. 103–111.
Fires,” John A. Blume Earthquake Engineering Center
Technical Report No. 163.

Design for Local Member Shear at Brace and
Diagonal-Member Connections: Full-Height and
Chevron Gussets
RAFAEL SABELLI and BRANDT SAXEY
ABSTRACT
Large local member shear forces develop in beams in chevron-braced frames due to the delivery of brace forces to beam flanges, which
are at a distance from the beam centerline (Fortney and Thornton, 2015, 2017; Hadad and Fortney, 2020). Using the “lower bound theorem”
(Thornton, 1984), Sabelli and Arber (2017) developed design methods to address this local member shear by optimizing the internal stress
distribution and thus maximizing the resistance utilized in design. This paper further develops those design methods for chevron beams and
extends them to gusset connections at columns.
Keywords: gusset plates, braced frames, truss connections.
INTRODUCTION including reinforcement and proportioning for (chevron)

gussets within the beam span and for full-height gussets at
T he “chevron effect” is a term used to describe local
beam forces in the gusset region of a chevron (also
termed inverted-V) braced frame. These local forces are
beam-column-brace connections.
V-braced frames (and their variants) are commonly used
in steel structures and are commonly termed “chevron-
not captured by beam analysis methods that neglect con-
braced frames.” Figure 1 shows three chevron configura-
nection dimensions. Fortney and Thornton (2015, 2017) and
tions: the inverted-V-braced frame (a), in which two braces
Hadad and Fortney (2020) have shown methods of analy-
connect to the bottom of the beam at its midpoint; the
sis for these forces. This study adds design solutions for
V-braced frame (b), in which two braces connect to the top
addressing high member shear in the connection region,
of the beam at its midpoint; and the two-story, X-braced
frame (c), in which four braces connect to the beam at its
midpoint, two from above and two from below.
The beams and columns of these frames are typically
Rafael Sabelli, Director of Seismic Design, Walter P Moore, San Francisco, designed using centerline models, and equilibrium is
Calif. Email: [email protected] (corresponding)
addressed in the design at the “workpoint” (the intersection
Brandt Saxey, Technical Director, CoreBrace, West Jordan, Utah. Email:
[email protected]
of member centerlines). In typical design, a substantial gus-
set plate is provided at brace connections, and force trans-
fer is accomplished over the length of that plate. Figure 2
Paper No. 2020-01R shows a frame with such gusset plates. Similar connections
(a) Inverted-V-braced frame (b) V-braced frame (c) Two-story, X-braced frame
Fig. 1. Chevron-braced frame configurations.

are used in truss construction with web-vertical wide-flange transferred by the gusset to the column), or with a connec-
chords (Figure 3). tion similar to a single-plate connection (also known as a
Work by Fortney and Thornton (2015, 2017) and Hadad “shear tab”), which minimizes these flexural forces. Adap-
and Fortney (2020) highlights the importance of analysis of tation of these methods to beam-column-brace connections
chevron braced-frame connections. In particular, Fortney with traditional gussets (Figure 5) is beyond the scope of
and Thornton derive expressions for the local beam shear this paper.
and moment that result from the distribution of brace forces The first part of the paper derives the design equa-
over the gusset-plate length. These beam forces (in particu- tions employing statics and two models of stress distribu-
lar, the shear) can result in the need to supplement the beam tion along the gusset-flange interface: the Uniform Stress
web with a doubler plate. An example of such a condition is Method, based on Fortney and Thornton (2015), and the
shown in the second edition AISC Seismic Design Manual Concentrated Stress Method, based on Sabelli and Arber
(AISC, 2012). In the third edition Seismic Design Manual (2017). The former model is simpler, but if that model indi-
(AISC, 2018), the example connection utilizes some of the cates that reinforcement is required, significant economy
relationships developed by Sabelli and Arber (2017) to can be realized by using the latter. The second portion of
eliminate the need for reinforcement. the paper is a brief design example that addresses both
This study builds on the work of Sabelli and Arber, apply- methods for the design of a chevron connection.
ing the same concepts developed by Fortney and Thornton, This study addresses both member shear and member
with the aim of providing methods for the design of con- moment caused by the local connection forces as these dif-
nections that do not require reinforcement. The methods fer from the shears and moments from a simple, centerline
presented in this paper rely heavily on the “lower bound model of members. In the authors’ experience, the local
theorem” as presented by Thornton (1984) for similar con- member shear often controls the connection design (such
nections, demonstrating adequate strength through inves- as by necessitating a minimum gusset length), but the addi-
tigation of an advantageous internal stress distribution in tional member moment caused by the local connection
a ductile connection and examining forces at gusset edges forces does not.
and at critical sections. The design equations derived here are based on the static
This study also extends and generalizes the equations equilibrium of the gusset plate based on the brace axial
developed for chevron connections for use in other condi- forces (and beam reactions for the column connection). As
tions, such as columns with full-height gussets (also called such, they are equally applicable to frames designed as part
“mega-gussets”) in which the gusset extends through the of a ductile seismic system (in which brace forces typically
beam depth and the beam connects to the gusset rather than correspond to the brace capacity), and those designed for
to the column (see Figure 4). In addition to transferring brace wind or other cases that do not involve capacity design.
forces, full-height gusset connections transfer beam forces Additional considerations for seismic design, such as deter-
to the column. Such connections may be accomplished with mination of the appropriate brace force level for which
welded beam flanges (as shown in Figure 4), which provide beam yielding should be precluded, are beyond the scope
flexural continuity (and thus additional flexural forces to be of this paper.
Fig. 2. Typical braced frames with gussets. Fig. 3. Truss with gussets.

Fig. 4. Full-height gusset brace connection at column.
Fig. 5. Traditional gusset at column.

SYMBOLS, NOMENCLATURE, Mmax Maximum member moment (within connection
AND CONVENTIONS region) due to brace forces, kip-in. (N-mm)
This study employs the following symbols and terms: MTot Total moment acting on beam due to Mf1 and Mf2,
kip-in. (N-mm)
Dclip Diagonal dimension of reduced critical-diagonal-
section length due to Yclip, in. (mm) NBm Beam axial force transferred to gusset at column
connection, kips (N)
Dcrit Length of critical diagonal section of gusset, in.
(mm) Ng Normal force on a gusset section transverse to the
member axis, kips (N)
Fi,j Brace axial force for brace “j” connecting to gusset
“i,” kips (N) (sign conventions are per the figures) Pn Nominal member or element axial strength, kips (N)
FN Gusset concentrated force at member flange, Ru Required strength, kips (N)
transverse to member axis, kips (N) (compression is
Rz Normal force from moment transfer for
positive)
Concentrated Stress Method, kips (N)
FV Gusset shear component parallel to member axis at
VBm Beam connection shear transferred to gusset at
interface with flange, kips (N)
column connection, kips (N)
FXcrit Force parallel to the member axis acting on critical
VCh Chevron shear, equal to the sum of member shear
diagonal section of gusset, kips (N)
and gusset shear transverse to member axis, kips (N)
FYcrit Force transverse to member axis acting on critical
Vef Effective member shear strength (deducting demands
diagonal section of gusset, kips (N)
other than brace connection forces), kips (N)
Fy Specified minimum yield stress, ksi (MPa)
VefTot Effective member shear strength considering the
Lbeam Beam length (column centerline to centerline), in. effects of unbalanced forces from gussets on both
(mm) sides of the member, kips (N)
Lg Gusset length, in. (mm) Vg Shear on a gusset section transverse to the member
axis, kips (N)
Lw Length of weld, in. (mm)
VM Member shear due to loading other than from braces,
MBm Beam moment transferred to gusset at column
kips (N)
connection, kip-in. (N-mm)
Vma Member shear (outside connection region), kips (N)
MCh Chevron moment at face of member due to force FV
(equal and opposite to the distributed moment MFV Vmc Member shear (within connection region), kips (N)
for concentric workpoints), kip-in. (N-mm)
Vn Nominal member or element shear strength, kips (N)
Mcrit Moment acting on critical diagonal section of gusset,
W Width of brace-to-gusset connection used to locate
kip-in. (N-mm)
critical diagonal section, in. (mm)
Mf Moment at gusset-to-flange-interface due to brace
Xcrit Dimension parallel to the member axis used to locate
forces, kip-in. (N-mm)
critical diagonal section, in. (mm)
MFV Moment in the connection due to force FV,
Yclip Dimension transverse to member axis of gusset
distributed along gusset length and eccentric to
corner clip, in. (mm)
workpoint (equal and opposite to the chevron
moment, MCh, for concentric workpoints), kip-in. dm Member depth, in. (mm)
(N-mm)
dg Gusset dimension transverse to member axis, in.
Mg Moment on a gusset section transverse to the (mm)
member axis, kip-in. (N-mm)
ecrit Location of force FYcrit with respect to intersection of
MM Member moment at workpoint due to loading other critical diagonal section and gusset edge, in. (mm)
than from braces, kip-in. (N-mm).
eg Eccentricity parallel to member axis of gusset
Mn Nominal member or element flexural strength, kips midpoint from workpoint (e.g., beam centerline at
(N) column connection), in. (mm)

em Transverse eccentricity from member flange to the corresponding subscript. The subscript “Tot” refers to
workpoint, typically equal to half the member depth, total forces, combining those from gusset “1” and gusset “2.”
in. (mm) Brace axial forces have two subscripts. The first pertains
to which gusset the brace connects to (“1” or “2”). The sec-
ez Length of moment arm between centroids of z
ond pertains to which of the two braces is indicated. Sign
regions, in. (mm)
conventions match the figures such that positive brace axial
k Distance from outer face of flange to web toe of forces F1,2 and F2,1 correspond to compression and positive
fillet, in. (mm) brace axial force F1,1 and F2,2 correspond to tension. Forces
and angles pertaining to each brace carry the same designa-
ru Required strength per unit length, kips/in. (N/mm)
tion subscript.
tg Gusset thickness, in. (mm) The design equations are presented in a general form
such that they can be used for both column and beam gus-
tw Member web thickness, in. (mm)
sets. To permit this, certain general terms are used, such as
w Weld size, in. (mm) “member” in lieu of “beam” or “column.” This approach
carries through to the symbols.
x Distance from gusset midpoint along member axis,
Brace-force components acting on the gusset–member
in. (mm)
interface are described as (gusset) “shear” or “normal”
z Length of concentrated stress region at ends of forces. Gusset shear forces, F V, are parallel to the member
gusset, in. (mm) axis (horizontal for the chevron beam and vertical for the
column); normal forces on the connection, FN, are perpen-
γ Brace angle from member longitudinal axis, deg
dicular to the member axis (vertical for the chevron beam
ϕb Resistance factor for bending (0.9) and horizontal for the column).
The term “workpoint” refers to the intersection of brace
ϕc Resistance factor for compression (0.9)
centerlines with each other or with the column centerline.
ϕn Resistance factor for nonductile limit states such as This workpoint is typically also at the beam centerline.
web crippling and weld rupture (0.75) Figure 6 shows dimensions noted on beam and gusset-
plate diagrams. Braces may occur above the beam, below
ϕt Resistance factor for tension (0.9)
the beam, or both. The diagram shows a symmetrical con-
ϕw Resistance factor for web local yielding (1.0) dition, but the connection calculations apply for asymmet-
rical cases. (Beam shear and moment require adjustment
ϕv Resistance factor for shear (1.0)
for asymmetrical applications.) Figure 7 shows dimensions
Subscripts are employed in some equations to distin- noted on column and gusset-plate diagrams. Braces may
guish actions and dimensions related to one gusset or one occur in various combinations, and the column may con-
brace from another. Gussets are designated “1” and “2,” and tinue up past the connection or may terminate as shown in
dimensions and forces associated with each gusset are given the upper diagram. The diagram shows a full-height gusset:
a gusset plate that comes between the beam and the column.
Fig. 6. Chevron gusset geometry.

STATICS connecting members are known. The analysis and design of
braces and truss diagonals is typically based on their ideal-
Although the methods developed in this paper are intended
ization as pin-pin members. In some cases, this idealization
to facilitate the design of connections of multiple mem-
could be modified to permit reduction of the moment Mf
bers (such as is shown in Figure 7), in essence the methods
(which causes the panel-zone shear in the connection); this
simply provide designers with the means of designing an
introduces design moments for both the main member and
attachment to a wide-flange member (such as is shown in
the diagonals, and thus requires an integration of member
Figure 8) for a set of known, in-plane forces, converting
design and connection design.
these to a normal force transverse to the member axis, FN ,
Figure 9 shows free-body diagrams of the gusset plate at
a shear force parallel to the member axis, F V, and a moment
the beam midspan; Figure 10 shows the same at the column.
in the plane of the web, Mf . The design of this connection
Both figures convert a known set of in-plane forces acting
includes evaluation of local limit states within the member,
on the gusset plate from connecting members into three
including web local yielding, web crippling, and local shear.
forces at the midpoint of the gusset-flange interface: normal
This local shear (in the gray zone in the center diagram in
force transverse to the member axis, FN, shear force paral-
Figure 8) is essentially panel-zone shear, and determination
lel to the member axis, F V, and moment in the plane of the
of the effective depth of the force couple is central to the
web at the face of the member, Mf. The brace forces used
design methods presented. The right-hand diagram in Fig-
for the connection design typically do not include moments,
ure 8 illustrates that while the member flange and bracket
although these could be included in determining the gusset
rotate in unison, the panel-zone section of the member web
forces.
can undergo large shear strains while the bracket remains
elastic, and thus the member can yield in shear even if there
is a substantial bracket present. The shear strength of the LOCAL MEMBER FORCES
member is not increased by the addition of the bracket shear (DERIVATION FOR TWO BRACES)
strength; a thicker or wider bracket would not preclude
For simplicity, only two braces are considered in the subse-
panel-zone shear yielding. Instead, the shear demand on the
quent derivation: those on “side 1” of the connection. A later
member panel zone can be reduced by using a bracket that
section shows the procedure for the inclusion of the effects
extends further along the flange, thus increasing the height
of an additional gusset on the far side (side 2). These braces
of the panel zone.
may be at equal angles (as is typical for the beam case) or at
The same reconstitution of forces is applicable to braced-
unequal angles (as happens frequently for the column case
frame and truss connections, assuming the forces in the
Fig. 7. Column gusset geometry.

and, on occasion, in the beam case). The design equations at the member face. This moment may be conceptualized
are presented in general terms applicable to both the beam by considering the forces from the two braces as applied
and the column condition. Minor adjustments to the equa- point loads at the locations where their centerlines inter-
tions are required for the column case due to the additional sect the member flange. If the brace forces are decomposed
forces from the connecting beams; these are noted. into components parallel to and normal to the member axis
The forces on the gusset-to-flange interface are statically (Figure 11), the chevron moment can be determined from
determined. For clarity, brace forces are separated into the normal force components and their eccentricities along
normal, FN, and shear, F V, components. For the column the member axis:
connection, these forces are combined with forces from
⎡ em ⎤
the beam: VBm, NBm, and MBm, as shown in Figure 10. (For MCh1 = ⎡⎣F1,1 sin ( γ 1,1 )⎤⎦ ⎢ ⎥ (4)
the case of a beam chevron connection, these forces are all ⎣ tan ( γ 1,1 ) ⎦
zero.) Assuming two braces with forces F1,1 and F1,2, the ⎡ em ⎤
shear force is: + ⎡⎣F1,2 sin ( γ 1,2 )⎤⎦ ⎢ ⎥
⎣ tan ( γ 1,2 )⎦
FV1 = F1,1 cos γ 1,1 + F1,2 cos γ 1,2 + VBm1 (1)
which reduces to:
The normal force is:
MCh1 = FV1em (5)
FN1 = F1,2 sin γ 1,2 − F1,1 sin γ 1,1 + NBm1 (2)
The chevron moment is opposed by a moment, MFV1, that
For the column connection, the collector force NBm1 should
corresponds to the parallel components, F V1, multiplied by
be determined from an analysis consistent with brace forces
the eccentricity of the flange from the centerline (2dm):
used in the connection design.
In addition to these normal and shear forces, there is a d m FV1d m
MFV1 = ⎡⎣F1,1 cos ( γ 1,1 ) + F1,2 cos ( γ 1,2 )⎤⎦ = (6)
moment (required for static equilibrium). While the moment 2 2
due to the brace forces is zero at the workpoint, at the flange
the moment is: which can be simplified to:
M f 1 = MCh1 − MBm1 (3) FV1d m

MFV1 = (7)
2
The first term in Equation 3, MCh1, is the “chevron moment”
Fig. 8. Free-body diagram of bracket.

Fig. 9. Force conventions for chevron connection.
Fig. 10. Force conventions for column connection.

This moment does not affect the gusset or the weld but is additive to the brace forces as discussed later.) Because
is necessary for member equilibrium and affects member the effect of the beam moment MBm1 is to reduce the total
moment. If the workpoint is at the member centerline (as demand, designers should consider how much of this ben-
shown in Figure 9), the distance from the flange to the eficial effect can be relied on, and a range of this moment
workpoint, em, is: could be considered.
dm For the case of an asymmetric column gusset (such as
em = for the workpoint at the member centerline (8) with only one brace, as shown in the upper diagram of Fig-
2 ure 7), the eccentricity between the gusset midpoint and the
If this is the case, the two brace-shear-component- beam centerline contributes to the moment:
induced moments are equal and opposite (MCh1 − MFV1 = 0), M f 1 = FV1em − M Bm1 + FN1eg (9)
and there is effectively no moment at the workpoint location
on the member centerline due to F V1. If the workpoint is off A similar adjustment can be made for chevron beams if
the member centerline (as shown in Figure 11), the braces the gusset midlength and the workpoint are not aligned
induce a moment in the member at the centerline location vertically.
aligned with the workpoint; this moment causes shear and In addition to the “chevron moment,” there is a “chevron
bending both within and outside of the connection region. shear.” The chevron shear (VCh1) is resisted by the gusset
Moving the workpoint from the main member centerline and the beam in combination:
toward the gusset reduces the chevron moment while intro- VCh1 = Vg1 + Vmc1 (10)
ducing a moment into the main member. In general, such an
approach can be economical but requires coordination with This chevron shear can be determined using static equilib-
member design and building analysis. Similarly, the con- rium on either half segment of the gusset:
nection analysis could assign only a portion of this chev- 1
ron moment to the main member and apply the remainder VCh1 = F1,1 sin γ 1,1 + FN1 (11)
2
to the braces. This might reduce the connection demands
but would introduce additional design moments into both
1
the main member and the braces, and this member flexure VCh1 = F1,2 sin γ 1,2 − FN1 (12)
could, in principle, affect the building response to lateral 2
loads. The former approach (modifying the work point) can
Figure 12 shows a free-body diagrams of beam and gus-
be integrated with the methods presented in this paper, but
set segments at the connection, including a transverse sec-
the latter approach (assigning counterbalancing moments to
tion through the beam and gusset showing the sharing of
the main member and the braces) is separate.
the chevron shear (VCh1) between the beam shear (Vmc1) and
For the column connection, the beam moment, MBm1,
the gusset transverse shear (Vg1). Note that the location of
affects the moment at the column flange, Mf1. This moment
the centroid of the transverse forces from the gusset seg-
(MBm1) should be consistent with the brace forces, and the
ment to the beam flange (Vmc1 ± 2FN1) is not specified in
minus sign reflects the direction of forces in Figure 10.
the free-body diagrams of the gusset segments; that loca-
Typically, both the beam moment and the brace force are
tion may be selected (within certain constraints) in the con-
due to the lateral drift. In such cases, the signs are consis-
nection design by the use of a stress distribution model,
tent with those shown in Figure 10; when the brace is in
such as the Elastic Method, the Plastic Method, or the
compression, the corresponding beam flange is in tension,
Optimized Plastic Method as described in Section 8 of the
and vice versa, and thus Mf1 in Equation 3 is the differ-
AISC Steel Construction Manual (2017). Once this location
ence rather than the sum. (The corresponding column shear
Fig. 11. Brace forces at flange.

is set, the division of VCh1 between Vmc1 and Vg1 is statically Member Shear
determined. (Moment at the vertical gusset section is zero,
Stresses at the member-to-gusset interface are assumed to
regardless of the stress distribution, for symmetrical condi-
be distributed uniformly using the full length for the normal
tions with equal brace forces, as shown in the subsequent
and shear forces and a plastic-section-modulus approach for
section, “Gusset Design: Mid-Length Transverse Section.”)
the moment (Fortney and Thornton, 2015). Following this
The same relationship applies to full-height column gus-
approach, the member shear within the connection region is
sets. As will be shown in the following sections, the sharing
described by the following equation:
of this chevron shear between the gusset and the beam or
column can be controlled by a combination of the gusset 2M f 1 4M f 1 F
V1 (x ) = − + x + N1 x + VM (13)
dimension selected and the force distribution assumed at Lg1 Lg12 Lg1
the gusset-to-flange interface. In this way, the design meth-
ods presented can reduce or eliminate any required web The first two terms are the shear due to the gusset
reinforcement. moment, which includes the chevron moment, MCh, plus
any other moment transmitted by the gusset per Equation 3.
The third term in the equation is the shear from the unbal-
UNIFORM STRESS METHOD anced normal force, and the value at the gusset end (FN1/ 2)
The Uniform Stress Method is the simplest model for is the shear in the member outside of the gusset region for
addressing the chevron effect, both for discussion and the typical, symmetrical case with VM = 0.
design purposes. While the treatment of this method here In this equation, x is the distance from the gusset mid-
is general, it does not specifically address conditions such point, as shown in Figure 13. This differs from Fortney
as beams with workpoints not at the beam midspan; Fort- and Thornton but is presented in this manner to facilitate
ney and Thornton (2017) provide a more thorough treat- combination of forces from gussets of different lengths on
ment. [The term “Uniform Stress Method” is not employed opposite flanges of the main member. The member shear is
by Fortney and Thornton. Sabelli and Arber use this term; additive for connections with gussets on opposite sides for
Hadad and Fortney refer to it as the “Chevron Effects (CE) the typical braced-frame case (with forces as shown in Fig-
Method.” The term “Uniform Stress Method” has been ures 9 and 10), although the member shear VM should only
used in practice and so is used here.] be added once. (Hereafter, it is assumed that the member
Fortney and Thornton (2015, 2017) employ the Uniform shear VM is zero in the connection region.)
Stress Method for the transfer of forces over the length of The maximum shear in the connection region occurs at
a chevron gusset. In this method, the moment transfer is the gusset midpoint (x = 0) and is equal to:
achieved through two blocks of principal stress in the gus- 2M f 1
set, each with a length equal to half that of the gusset, as Vmc1 = (14)
Lg1
in the “plastic method stress distribution” described in the
AISC Steel Construction Manual. Figure 13 shows the This shear, Vmc1, is not equal to the chevron shear, VCh1, nor
Uniform Stress Method applied to the column and chevron to a trigonometric component of either of the brace forces
connections.
Fig. 12. Transverse section of beam and gusset showing chevron shear.

(e.g., F1,1sinγ1,1 or F1,2sin γ1,2); it may be greater or smaller region should be considered by reducing the available shear
than the trigonometric component, depending on the geom- strength or by engaging the gusset to distribute the load
etry of the connection. The difference between the two is over the gusset length along with the normal force, FN1.
the shear carried by the gusset, Vg1, as indicated in Equa- The member shear is the result of both the eccentricity
tion 10. The longer the gusset plate, the lower the shear in (typically a function of the member depth) and the gusset
the main member due to the chevron moment, Vmc1, and length. These can be adjusted (within practical bounds) to
thus the greater the shear resisted by the gusset, Vg1. This provide a member that does not require web strengthening.
is similar to a moment connection in which beam haunches Following this approach, the minimum gusset length to
can be used to engage a larger column panel-zone height. eliminate web strengthening is:
In this sense, the gusset plate can be used as external shear 2M f 1
reinforcement for the beam, although in this method, it is Lg1 ≥ (15)
ϕ vVn
the length of the gusset that permits it to provide a larger
arm for transfer of the chevron moment (and thus reduce the which is equivalent to:
force imposed on the main member) rather than a simple
addition of member and gusset shear strength. 2FV1em
Lg1 ≥ (16)
Note that this member shear in Equation 14, Vmc1, is due ϕ vVn
only to the force components parallel to the member axis
A subsequent section addresses member selection to avoid
(shear on the connection). The unbalanced normal compo-
web reinforcement.
nent does cause shear in the member, but this unbalanced-
component shear becomes zero at the workpoint and thus is
Concentrated Forces
not considered in conjunction with shear from the balanced
component (i.e., the shear from Equation 14). Figure 14 The limit states of web local yielding and web crippling
shows a shear diagram for brace-induced shears in a typical typically can be satisfied without reinforcement at chevron
pin-end beam consistent with Equation 13. connections, especially those designed using the Uniform
Note that the maximum member shear occurs at the gus- Stress Method. These limit states can be evaluated consid-
set midpoint (where the member shear neglecting connec- ering two concentrated forces (Ru), each acting on a bearing
tion effects is zero), and thus the member shear outside of the length of 2Lg:
connection does not affect the maximum shear in the con- FN1 2Mf 1
nection region. Concentrated loads within the connection Ru1 = ± (17)
2 Lg1
Fig. 13. Uniform Stress Method (after Fortney and Thornton).

Moment 2MCh1 2MCh1 MFV1 (18)
M1 (x) = − x+ 2 x x− x
Shear forces in the member have an effect on member Lg1 Lg1 Lg1
moment. While this effect is generally small, Fortney and ⎛L Lg1 x2 ⎞
Thornton (2015) describe conditions in which the beam − FN1 ⎜ beam − −
⎝ 4 8 2Lg1⎟⎠
moment determined using these assumptions (if not con-
sidered in design) may necessitate reinforcement using the
Note that Equation 18 includes more than the integral of
Uniform Stress Method. Hadad and Fortney (2020) show
the member shear formula (Equation 13). It also includes a
that in finite element analyses, the beam moments are sub-
distributed moment due to the applied force parallel to the
stantially lower than those calculated using the Uniform
member axis, F V1, at the gusset–flange interface (which, for
Stress Method. While the authors do not propose evalua-
simplicity, is assumed to be uniformly distributed along the
tion of the member moment within the connection region as
length of the gusset):
necessary, examination of the effects that contribute to the
moment may aid in understanding of the method. MFV1 FV1em
= (19)
Lg1 Lg1
Beam Moment
Thus:
In the case of a chevron beam, braces are often considered to
MFV1 MCh1
be a support point for the beam for wind design. For seismic = (20)
design of ductile systems, the expected behavior typically Lg1 Lg1
entails brace yielding with resulting transverse loading of
The applied force parallel to the member axis, F V1, thus
the beam causing beam shear and bending. [See, for exam-
has two equal and opposing effects: the transverse stress
ple, AISC Seismic Provisions Section F2.3(b).] The beam
resulting from MCh, which causes member shear and
is evaluated for these forces (adding any gravity loading
moment, and the distributed moment corresponding to
effects) in combination with the axial force resulting from
MFV. These counteracting effects produce zero moment at
the components of brace force parallel to the member axis.
the ends of the connection region and at the gusset mid-
The combined effect of the moment Mf1 and the brace force
length; at other locations, some moment may result based
components parallel to the member axis, F V1, produces no
on the differing rates of accumulation over length within
shear or flexure outside of the connection region.
the connection region, corresponding to the assumed trans-
For the chevron beam connection, MBm = 0, and thus
verse stress distribution and the assumed distribution of F V1
Mf = MCh. Beam moments within the connection region are
over the gusset length.
described by the following equation:
Fig. 14. Brace-induced shears in pin-end beam (Uniform Stress Method).

Equation 18 simplifies to: shear forces outside the connection region are not generally
large, especially for frames with pin-ended beams.
FV1em ⎛ 2 ⎞ ⎛L Lg1 x ⎞ 2
M1 (x) = −
⎜x − x x ⎟ − FN1 ⎜ beam − − ⎟
Lg1 ⎝ Lg1 ⎠ ⎝ 4 8 2Lg1⎠
CONCENTRATED STRESS METHOD
(21)
In the preceding discussion, the Uniform Stress Method
Nonuniform distributions of transfer of the shear force F V stress distribution was assumed to determine the member
from the gusset to the beam are also admissible, including forces in the connection region. The calculated member
distributions that minimize or eliminate the local moment shear may be reduced by selecting a more favorable dis-
effect MCh (x) − MFV (x). However, the authors have found tribution. The Concentrated Stress Method maximizes the
such approaches unnecessary for demonstrating beam ade- moment arm within a given gusset length and thus mini-
quacy and, at times, uneconomical for the gusset weld. mizes the corresponding force caused by the moment. This
A simplified equation can be used to provide a liberal method is based on the Optimized Plastic Method (AISC,
estimate of the maximum brace-induced moment in the 2017), modified to optimize only for moment resistance
beam: (rather than both moment and normal force) and to allow
FV1em ⎛L Lg1 ⎞ for incorporation of design limits based on both gusset yield
Mmax1 ≤ + FN1 ⎜ beam − − MM
+ (22)
8 ⎝ 4 8 ⎟⎠
and member limit states.
In the Concentrated Stress Method, the moment Mf1 is
where MM is the member moment neglecting brace forces assumed to be transferred at the ends of the gusset over
(typically due to gravity). This equation simplifies the deter- lengths z1. The remaining segment in the middle of the
mination of moment, providing a liberal upper bound by gusset does not participate in transmitting the flexure; it
combining two maxima: the beam moment corresponding is assumed to resist the unbalanced force FN1. Figure 15
to the local shear, Mf1/ 8, which occurs at the gusset quar- shows this stress distribution.
ter point, and the midspan moment due to the unbalanced
normal force (the second term) and any other beam loading, Member Shear and Minimum Gusset Length
MM. (The idealization of the unbalanced normal force as a The Concentrated Stress Method converts the moment Mf1
point load rather than distributed over the gusset length also into a normal force couple Rz1 with a moment arm of ez1.
slightly overestimates the moment.) For the typical braced- This normal force Rz1 is distributed over a length z1. The
frame case (with forces as shown in Figures 9 and 10), the values of Rz1 and ez1 are determined such that Rz1 does not
moments from the two gussets are additive, with MM being exceed the force that would cause shear yielding of the
added only once. member.
The first term in Equation 22, Mf1/ 8, is a local effect Figure 16 shows a shear diagram corresponding to this
of the connection geometry and is typically small, corre- stress distribution for a chevron connection. Note that the
sponding to a small eccentricity for the axial force in the maximum beam shear in the Concentrated Stress Method
beam (which is typically F V1/ 4 at the gusset quarter point). does not occur at the beam midpoint (as it does for the Uni-
form Stress Method), and thus the beam shear outside of
Column Moment the connection affects the maximum shear within the con-
The column is not required to span to resist the unbal- nection region.
anced brace forces. The column moment in the connection The moment arm ez1 is:
region for the Uniform Stress Method is similar to that from ez1 = Lg1 − z1 (24)

Equation 22:
FV1em The normal force from the moment transfer is thus:
Mmax1 ≤ + MM (23)
8 Mf 1
Rz1 = (25)
ez1
The effect of any moment MBm on the member moment is
captured in the term MM, conservatively taken at its full
Mf1
value at the quarter point (where the effect of F V is at its Rz1 = (26)
maximum). Lg1 − z1
The column moment is typically permitted to be neglected This normal force causes shear in the member. For
in capacity-design calculations for seismic loads per AISC beams, the maximum shear is a combination of the shear
Seismic Provisions D1.4a (AISC, 2016a). For other braced- due to the unbalanced force and the shear due to delivery of
frame cases this moment is typically very small. The design

the chevron moment. The maximum shear is given by the The shear VM is typically zero adjacent to (and within) the
following equation: connection region for beams.
Vmc1 = Vma1 + Rz1 (27) If the gusset is long enough, the total connection shear
Vmc may be set less than or equal to the design shear strength
The shear outside of the connection region, Vma1, is due to of the member in order to preclude the need for shear rein-
net normal force, FN1, and the member shear from gravity forcement. For a given gusset length, the maximum moment
or other sources, VM: transfer can be achieved by the highest concentration of
stress at the ends. For a minimum gusset length, stiffeners
1
Vma1 = FN1 + VM (28) at the gusset edges may be used to create a moment arm
2
Fig. 15. Stress distribution for the Concentrated Stress Method.
Fig. 16. Brace-induced shear in pin-end beam with Concentrated Stress Method.

equal to the gusset length Lg1, similar to a moment connec- This length is minimized by taking the maximum normal
tion in which beam flanges deliver moment to the face of a force Rz1 that the member can resist based on its effective
column. For the more typical chevron-moment transfer via shear strength per Equation 30.
the gusset plate, the concentrated stress may be limited by Considering combined shear and tension in the gusset,
web local yielding, web crippling, or the gusset yielding. the minimum gusset thickness corresponding to the mini-
If the gusset length is minimized (without stiffeners), mum length selected using Equation 34 and the maximum
the concentrated stress will be maximized such that the normal force Rz1 (equal to Vef1) is determined using the von
full member shear strength is utilized. Unlike the Uniform Mises yield criterion, with shear stress over the full gusset
Stress Method, in the Concentrated Stress Method the max- length and the moment delivered by a force couple.
imum member shear is maintained over a significant por- A gusset that satisfies this criterion is required:
tion of the connection length, and thus (for beams) occurs 2 2
at locations that also have shear induced by the unbalanced ⎛ FV1 ⎞ ⎛ Rz1 ⎞
tg1 ≥ ⎜ ⎟ +⎜ ⎟ (35)
normal force from braces. Considering that some of the
⎝ ϕv 0.6Fy Lg1⎠ ⎝ ϕt Fy ( Lg1 − ez1 )⎠
shear strength is utilized in resisting this unbalanced force,
the remaining member shear strength that can be utilized which is equivalent to:
for the moment transfer is: 2 2
⎛ FV1 ⎞ ⎡ Rz1 ⎤
Vef 1 = ϕ vVn − Vma1 (29) tg1 ≥ ⎜ ⎟ +⎢ ⎥ (36)
⎝ ϕ v 0.6Fy Lg1 ⎠ ⎢ ⎛ Mf1⎞ ⎥
⎢ ϕt Fy Lg1 − ⎥
For designs with a gusset on the opposite flange, both the ⎣ ⎝ Rz1 ⎠ ⎦

design shear strength (ϕVn) and the net shear outside the
connection (Vma1 − Vma2) can be apportioned between the The center zone may be similarly examined, although
two gusset designs. This is addressed in a later section. generally this zone is much less stressed:
The Concentrated Stress Method is derived such that 2 2
the maximum shear from Equation 26 does not exceed the ⎛ FV1 ⎞ ⎛ FN1 ⎞
tg1 ≥ ⎜ ⎟ +⎜ ⎟ (37)
effective shear strength from Equation 29: ⎝ ϕ v 0.6Fy Lg1⎠ ⎝ ϕt Fy ( Lg1 − 2z1 )⎠

Rz1 ≤ Vef 1 (30)
which is equivalent to:
The minimum gusset length possible corresponds to the 2 2
use of stiffeners (flanges) to transfer the moment Mf1. Thus ⎛ FV1 ⎞ ⎡ FN1 ⎤
tg1 ≥ ⎜ ⎟ +⎢ ⎥ (38)
as z1 approaches zero, Equation 26 (combined with Equa- ϕ
⎝ v 0. 6Fy g1 ⎠
L ⎢ϕ F ⎛ 2M f1 ⎞ ⎥
− Lg1
tion 30) gives the minimum gusset length dimension: ⎢⎣ t y ⎝ Rz1 ⎠ ⎥⎦

Mf1
Lg1 = for z1 = 0 (31) If the required gusset thickness is excessive, a longer gus-
Vef 1 set may be employed. The gusset length required for a given
Without such stiffeners, there is a finite length of gusset gusset thickness is the root of a fourth-power polynomial,
z1 over which the force Rz1 is transferred to the beam. The for which various solution methods are available, including
minimum length z1 may be governed by the limit states of trial-and-error and computer solvers. A closed-form solu-
web local yielding, web crippling, or yielding of the gusset. tion may also be derived using Ferrari’s formula (Euler,
For web local yielding, AISC Specification Equation J10-2 1765). A simple approximate formula can be obtained,
(AISC, 2016b) can be rearranged to solve for the minimum however, if the shear, F V1, is neglected and Rz1 is set equal
bearing length z1: to Vef1:
Rz1 Mf 1 Vef 1
z1 ≥ − 5k (32) Lg1 > + (39)
ϕ w wFy t w Vef 1 t g1ϕ t Fy

The corresponding minimum gusset length based on the As in the design example, a length slightly greater than that
web local yielding limit state is: indicated by Equation 39 is generally satisfactory.
Because of the complexity of the equations for web crip-
Lg1 ≥ ez1 + z1 (33) pling, that limit state is not integrated into the equations
Combining Equations 24, 32, and 33: for minimum gusset length but may be evaluated after the
gusset length is determined, as shown in the subsequent
M f1 Rz1 section.
Lg1 ≥ + − 5k (34)
Rz1 ϕw Fy t w

Gussets Longer than the Minimum Length corresponding to the larger of the values from Equations
40 and 41.
In many cases, the gusset length will exceed the minimum
The minimum value of z1 is the largest of the values from
from Equation 34, due to design considerations such as the
Equations 40, 41, and 42. Larger values of z1 may be used up
required gusset thickness or the brace-to-gusset attach-
to a maximum value of z1 limited by the minimum moment
ment. For gussets longer than the minimum, the designer
arm corresponding to the maximum transverse force:
has some flexibility in selecting a stress distribution that
transfers the moment. Maximizing the length z decreases Mf 1
z1 ≤ Lg1 −
the moment arm and thus increases the force to be resisted; Vef (43)
it also leaves less weld length for the forces that are resisted
in the center zone. The authors have found that the total Above this value of z1, the length of the moment arm ez
weld volume tends to be minimized by minimizing the is insufficient and the transverse force required to transmit
length z1 for optimized weld lengths and sizes. However, in the moment will exceed Vef . If this maximum value is nega-
many cases, the weld size in the center zone is controlled tive the gusset is too short to transmit the moment regard-
by a minimum weld size, or a proportioning requirement to less of how concentrated the force delivery can be.
ensure deformation compatibility is used to size the center- In principle, the maximum value of z1 may also be lim-
zone weld or the minimum length for z1; in such cases, ited by stresses in the center region. Use of the additional
maximizing the length z1 may be more economical. bearing length of 5k (as opposed to 2.5k) in Equation 40
The minimum length z1 is determined considering web distributes some of the force Rz1 into the center region. For
local yielding, web crippling, and gusset yielding, consid- cases with high unbalanced load (or very small dimension
ering the normal force Rz1 acting over the length z1 (and ez ), 2.5k may be used in Equation 40, or the following eval-
combined with a shearing force for the gusset evaluation). uation can be made based on the total transverse force Rz1 +
The force Rz1 corresponding to the length z1 is determined FN1 acting on a length Lg1 − z1:
by Equation 26 and is bounded by Equation 30. Rz1 + FN1
Considering web local yielding, the minimum length z1 z1 ≤ Lg1 − + 5k (44)
ϕ w Fy t w
is:
Similarly, if the normal force FN1 is large (or the center
Lg1 Lg12 M f1
z1 ≥ − − − 5k (40) zone is very short), the gusset stress in the center region
2 4 ϕ w Fy t w should also be considered. Using the von Misses yield crite-
rion gives this maximum:
The von Mises yield criterion is used to determine the
minimum length z1 that, for the design loads and a given ⎡ FN1 ⎤
1⎢ ϕt Fy tg1 ⎥
gusset thickness and length, will result in effective stresses z1 ≤ ⎢Lg1 − (45)
2 2 ⎥
at the yield limit in the gusset. The shear stress is due to FV1 ⎢ ⎛ FV1 ⎞ ⎥
and the normal stress is due to the moment Mf1. The thick- ⎢ 1 − ⎜ ϕ 0.6F t L ⎟ ⎥
⎣ ⎝ v y g1 g1 ⎠ ⎦
ness should satisfy Equation 36. The length z1 correspond-
ing to the gusset length and thickness selected is obtained
by combining Equations 24, 26, and 36:
Concentrated Stress Method Validation
Mf 1
Lg1 Lg12 ϕt Richards et al. (2018) analyzed a number of braced-frame
z1 ≥ − − (41) beams using finite element models and compared inter-
2 4 ⎛ ⎞
2
2 F nal forces with those obtained from the Uniform Stress
( Fy tg ) − ⎜ V1
⎟
⎝ ϕ v 0.6Lg1 ⎠ Method and the Concentrated Stress Method (as presented
by Sabelli and Arber, 2017). Figure 17 shows the results
Considering web crippling, the minimum length z1 can be of one such analysis (from Figure 3.32 from Richards et
determined by rearranging AISC Specification Equation al.). The finite element analysis results (FE) are shown
J10-4: along with the shears determined using the Uniform Stress
Method (USM) and the Concentrated Stress Method (CSM,
⎡ Rz1 tw ⎤ ⎛ d ⎞ ⎛ tf ⎞ 1.5
z1 ≥ ⎢ − 1⎥ m (42) using the modified method as presented in this paper); the
⎢⎣ ϕ n 0.80t w2 EFy tf ⎥⎦ ⎝ 3 ⎠ ⎝ tw ⎠ beam shear strength, ϕVn, is also indicated. The values of

maximum shear are reasonably consistent between the
Due to the number of terms of Equation 42, it is con- finite element analysis and the Concentrated Stress Method
venient to use the maximum value of Rz1 = Vef1 per Equa- for this example, and while the total shear in the finite
tion 30 rather than solving for z1 using Equations 26 and element model is less than the beam shear capacity, the
42. Alternatively, web crippling can be evaluated using Rz1 finite element analysis indicated local yielding in the web.

Nevertheless, the point of maximum shear is not identical The beam shear in the center region is:
between the finite element analysis and the Concentrated FN1
Stress Method, indicating that the Concentrated Stress V1 ( x ) = −Rz1 + x for |x| ≤ Lg1/ 2 − z1 (46)
Lg1
Method, while useful for design, is not a perfect representa-
tion of the internal stresses. In the z1 region, the beam shear is:
It should be noted that both Richards et al. (2018) and
Hadad and Fortney (2020) found that the Uniform Stress ⎛ Lg1 x ⎞ FN1 x
V1 ( x ) = − Rz1 − + (47)
Method is generally representative of the beam shear at ⎝ 2z1 z1 ⎠ 2 x
levels of loading that do not result in web shear yielding. for |x| ≥ Lg1/ 2 − z1
At higher levels of loading, the work of Richards et al.
Similar to the Uniform Stress Method, the distributed
(2018) indicates adequate performance of gussets meeting
moment MFV may be assumed to be transferred over the
the minimum required length for the Concentrated Stress
length Lg1 using Equation 19. The brace-induced moment
Method and exceeding the length required for the Uniform
in the beam is:
Stress Method. Additional comparisons of finite-element
analyses from Richards et al. with the two design models MFV1 ⎛L Lg1 x2 ⎞
M1 (x) = − Rz1x + x − FN1 ⎜ beam − − (48)
are presented in Sabelli et al. (2020). Lg1 ⎝ 4 8 2Lg1 ⎟⎠
Moment for |x| ≤ Lg1/ 2 − z1
Shear, such as shown in the three analyses represented in Similar to the Uniform Stress Method, the two equal and
Figure 17, implies moment. The authors do not propose opposing effects of the applied force parallel to the member
evaluation of the member moment within the connection axis, F V1, are included in Equation 48: the transverse stress
region as necessary but present the equations for moment to resulting from MCh , which causes member shear in addi-
facilitate understanding of the Concentrated Stress Method. tion to moment, and the distributed moment corresponding
Hadad and Fortney (2020) show that in finite element anal- to MFV, which does not affect the shear. The shape of the
yses, the beam moments are substantially lower than those shear diagram in the Concentrated Stress Method results in
calculated using the Uniform Stress Method. a somewhat smaller moment within the connection region
than that corresponding to the Uniform Stress Method.
Beam Moment A liberal estimate may be made by computing the
connection-induced moment and combining with the mid-
The beam moment is the combination of the integral of the span moment due to overall beam flexure:
beam shear and the distributed moment MFV (Equation 19).
Fig. 17. Concentrated Stress Method (CSM) and Uniform Stress Method (USM) analysis of beam from Richards et al. (2018).

where
MCh1 ⎡ z1 ⎛ z1 ⎞ ⎤
2
⎛L L ⎞
⎢
Mmax1 ≤ − ⎥ + FN1 beam − g1 + MM MTot = M f 1 + M f 2 (53)
2 ⎢ Lg1 ⎝ Lg1 ⎠ ⎥ ⎝ 4 8 ⎠
⎣ ⎦
(49) Other methods of apportionment are admissible, but this
method allows for design of the gusset connections based
The connection-induced moment due to MCh never on member forces established prior to gusset design and
exceeds MCh/ 8 (the value for z1 = Lg1/ 2, which corresponds without additional interdependence.
to the Uniform Stress Method distribution). The moment
is additive with the moment from gusset 2 for the typical Uniform Stress Method
braced-frame case.
In the Uniform Stress Method, the full member shear
Column Moment strength generally may be utilized:
VefTot = vVn (54)
Similar methods can be applied to calculate moment in the
column within the connection region. Column moments Gusset plates may be of different lengths, but for simplic-
due to frame behavior, which reverse over the connection ity, they may be set to be equal. If equal-length gussets are
depth, are typically additive to the effect of MCh. used (Lg1 = Lg2), Equation 15 for the minimum gusset-plate
2 length to preclude the need for reinforcement can be modi-
MCh1 ⎡ z1 ⎛ z1 ⎞ ⎤
Mmax1 ≤ ⎢ − ⎜ ⎟ ⎥ + MM (50) fied thus:
2 ⎢ Lg1 ⎝ Lg1 ⎠ ⎥
⎣ ⎦ 2M Tot
Lg ≥ (55)
The effect of any moment MBm on the member moment is VefTot
captured in the term MM, conservatively taken at its full
value. Similar to the beam moment, the column moment
Concentrated Stress Method
due to the chevron effect tends to be small.
In the Concentrated Stress Method, the effective beam
shear strength is reduced:
COMBINATION OF FORCES
FOR TWO GUSSET PLATES FN1 FN 2
VefTot = ϕvVn − − − VM (56)
2 2
The member forces derived are for braces on one side with
opposite forces (one brace in tension and the other in com- If equal-length gussets are used, Equation 34 for the
pression). These forces may be combined with gravity- minimum gusset-plate length to preclude the need for rein-
induced forces and with shear due to flexural restraint for forcement can be modified thus:
frames with moment connections. While the diagrams show
M Tot VefTot
the left brace in tension and the right brace in compression, Lg ≥ + − 5k (57)
forces corresponding to the opposite case are easily deter- VefTot ϕw Fy tw
mined by using negative values for the brace forces.
Similarly, Equation 39 based on gusset yielding can be
For a configuration with braces on both sides of the
modified thus:
member (such as a two-story X-configuration in a beam),
brace-induced shears and moments will be additive for the MTot VefTot
Lg1 > + (58)
typical case in which the story shears are in the same direc- VefTot tg1ϕt Fy
tion. The effective web shear resistance may be apportioned
between the two gussets, considering the relative magni-
tudes of their moments, Mf , to permit independent design MEMBER SELECTION
of the two gussets:
The procedures described earlier allow for the design of
Mf 1 a connection based on design forces and the strength of
Vef 1 = VefTot (51)
M Tot a member already selected. Economy in steel construc-
tion can often be achieved by consideration of connection
and
requirements in member selection. Equations for required
Mf 2 beam strength, rather than required gusset length, can be
Vef 2 = VefTot (52)
M Tot derived from the methods presented. Fortney and Thornton
(2015) suggest a preliminary assumption of a gusset length
of one-sixth of the beam span for chevron connections; this
value can be used to facilitate member selection.

The chevron shear at the connection is due mainly to reinforcement, albeit with possible moderate adjustment in
the chevron moment, Mch, which is proportional to the gusset length. (Use of 75% of the value from Equation 62
eccentricity, em , typically half the member depth. Selection requires ez1∼qLg1; use of 60% requires ez1∼yLg1.)
of a shallower member reduces the eccentricity and thus
reduces the chevron shear. Because the shear capacity is
GUSSET AND WELD DESIGN
also proportional to the depth, the member depth appears in
both the demand and capacity terms, and the required web The stress distributions assumed in the Uniform Stress
thickness is not a function of member depth. Method and the Concentrated Stress Method impose dif-
ferent demands on gusset plates and welds. The design of
Uniform Stress Method those elements should be compatible with each other (and
with the checks on member local limit states, such as web
Using the Uniform Stress Method equation for minimum
local yielding, web crippling, and panel-zone shear), or the
gusset length for a given member strength (Equation 15),
connection may not be able to resist the applied forces. For
the minimum shear strength is
example, if the design for local limit states is based on the
ϕ vVn = ϕ v 0.6Fy dm t w (59) Uniform Stress Method, but the gusset thickness is sized
using the optimized plastic stress method (which is implicit
2M f 1 2M f 2 in the interaction Equation 9-1 in the AISC Steel Construc-
ϕ vVn ≥ + (60)
Lg1 Lg2 tion Manual), the member may be subject to a combina-
tion of web local yielding and gusset plate yielding prior to
FV1d m FV 2d m developing the required strength. It is recommended that
ϕ vVn ≥ + (61) the method used for member local limit states be carried
Lg1 Lg2
through the design of the gusset and the weld.
Note that the member depth appears both in the demand,
Mf1, and in the resistance, ϕVn, and thus cancels out in Gusset Design: Section Parallel to Member Axis
Equation 62 for the minimum member web thickness:
Uniform Stress Method
FV1 FV 2
Lg1 + Lg2 For the Uniform Stress Method, the gusset section at the
tw ≥ (62)
ϕ v 0.6Fy interface with the flange can be evaluated using an interac-
tion method such as the von Mises yield criterion and solv-
For beams or columns with small moments due to unbal- ing for the required thickness:
anced normal forces, a shallow member meeting this
2 2
requirement may be economical. Note that the optimal gus- ⎛ 4M f 1 FN1 ⎞ ⎛ FV1 ⎞
set length may be a function of member depth. tg ≥ ⎜ 2 + ⎟ +⎜ ⎟ (63)
⎝ ϕt Fy Lg1 ϕt Fy Lg1⎠ ⎝ ϕ v 0.6Fy Lg1 ⎠
Concentrated Stress Method
Use of a member with web thickness less than that required Concentrated Stress Method
by Equation 62 necessitates either reinforcement of the For the Concentrated Stress Method, the gusset section at
web or use of a greater moment arm to deliver the moment the interface with the flange is implicitly designed by use of
Mf1 than is assumed in the Uniform Stress Method (such a length z conforming to Equations 36 and 38.
as ez1 in the Concentrated Stress Method). There is not a
corresponding simple equation for minimum web thickness Gusset Design: Mid-Length Transverse Section
using the Concentrated Stress Method. However, the mini-
mum gusset length based on beam shear strength with z1 = Statics require that certain forces be transferred across the
0 (Equation 31) represents a limiting value. This minimum midpoint of the gusset. Figure 18 shows free-body diagrams
gusset length corresponds to a moment arm ez1 equal to Lg1 of half of a gusset for both the Uniform Stress Method (a)
(rather than 2Lg1, as corresponds to the Uniform Stress and the Concentrated Stress Method (b).
Method and Equation 62), and thus, if there are no other The normal force on the gusset transverse section (i.e.,
member shear demands to consider, the required web thick- the force parallel to the member axis) for both models is:
ness for this limiting case is half of that from Equation 62. 1
A web thickness of 60% to 75% of that given by Equation 62 Ng1 = FV1 − F1,1 cos γ 1,1 (64)
2
(based on an assumed gusset length) generally permits use
of the Concentrated Stress Method without the need for web

which is equivalent to: The gusset moment (for the Concentrated Stress Method)
1 is:
Ng1 = ( F1,2 cos γ 1,1 − F1,1 cos γ 1,2 ) (65)
2 ⎛ Lg1 z1 ⎞ FN1 ⎛ Lg1 z1 ⎞
Mg1 = Rz1 − − − (70)
⎝ 2 2⎠ 2 ⎝ 4 2⎠
Uniform Stress Method ⎛ d g1 ⎞ FV1
+ Ng1 em + − em
⎝ 2 ⎠ 2
The shear on this gusset section transverse to member axis
(for the Uniform Stress Method) is: which simplifies to:
2M f 1 FN1 ⎛ d g1 ⎞ FN1Lg1 FN1z1
Vg1 = F1,1 sin γ 1,1 − + (66) Mg1 = Ng1 em + − + (71)
Lg1 2 ⎝ 2⎠ 8 4
The gusset moment (for the Uniform Stress Method) is: Note that the terms related to Mf1 and F V1 cancel out in
both Equations 68 and 71. Thus, the gusset moment is only
⎛ 2M f 1 FN1 ⎞ Lg1 ⎛ d g1 ⎞ FV1
Mg1 = ⎜ − ⎟ + Ng1 ⎜ em + ⎟− em due to the unequal brace force components transverse to the
⎝ Lg1 2 ⎠ 4 ⎝ 2 ⎠ 2 member axis (resulting in an unbalanced transverse force
(67) FN1) and to unequal brace force components parallel to the
member axis (resulting in a force transfer Ng1 from one half
which simplifies to:
of the gusset to the other), with those two effects offsetting
⎛ d g1 ⎞ FN1Lg1 each other.
Mg1 = Ng1 ⎜em + ⎟− (68)
The gusset should be evaluated for the interaction of
⎝ 2⎠ 8
these shear, normal, and moment forces. This may be done
using von Mises yield criteria or other methods as discussed
Concentrated Stress Method in the AISC Steel Construction Manual (AISC, 2017).
The gusset shear transverse to member (for the Concen-
Gusset Design: Diagonal Section (Concentrated
trated Stress Method) is:
Stress Method)
FN1
Vg1 = F1,1 sin γ 1,1 − Rz1 + (69) Brace-to-gusset connections are typically evaluated for
2
the limit state of block shear without consideration of the
subsequent load path through the gusset. This may not
(a) Uniform Stress Method (b) Concentrated Stress Method
Fig. 18. Free-body diagrams of half gusset.

present any significant inconsistency for the Uniform Stress Note that this length is somewhat greater than the length
Method, which assumes a uniform stress over each half of of one of the shear areas used for the block-shear rupture
the gusset, but may for the Concentrated Stress Method, check in the gusset plate. In some cases, a simplified check
in which high stresses are assumed at the gusset zones z1. with that portion of the block-shear area suffices.
In order to ensure that the gusset has sufficient strength to The forces acting to the left of that section are:
transfer the force Rz1 to the region z1 in the gusset-to-beam X crit
connection, the gusset should be evaluated along a diagonal FX crit = FV (77)
Lg
section, as shown in Figure 19. That section is aligned with
the outside shear area used in the block-shear calculation If Xcrit is less than or equal to z:
and projected to the beam–gusset interface.
The intersection of the diagonal section with the gusset X crit
FYcrit = Rz (78)
edge occurs at a point defined by the dimension Xcrit: z
Lg e W
X crit = − m − (72) X crit
2 tan γ 2sin γ ecrit = (79)
2
thus,
⎛ D cos γ ⎞ D sin γ
Lg d m cos γ + W Mcrit = FYcrit ecrit − crit + FXcrit crit (80)
X crit = − (73) ⎝ 2 ⎠ 2
2 2sin γ
If Xcrit is greater than z1:
The length of the section is determined using the dimen-
sions indicated in Figure 19: X crit − z
FYcrit = Rz + FN (81)
Lg − 2 z
Dclip
= X crit tan γ − (dg − Yclip ) (74)
sin γ Note that on one side of the gusset, the two terms of Equa-
tion 81 will be additive; that is the more critical condition.
X crit
Dcrit = − Dclip (75) ( Xcrit − z )2
cos γ Rz ( X crit − z 2 ) + FN
2 ( Lg − 2z )
ecrit = (82)
FYcrit
Dcrit = X crit cos γ + ( dg − Yclip ) sin γ (76)
Fig. 19. Critical gusset section.

⎛ D cos γ ⎞ D sin γ the design of the weld (which need not exceed the size cor-
Mcrit = FYcrit ecrit − crit + FXcrit crit (83) responding to the gusset-plate strength).
⎝ 2 ⎠ 2
The forces are transformed to act on the diagonal section: Concentrated Stress Method
Vcrit = FXcrit cos γ + FYcrit sin γ (84) The Concentrated Stress Method inherently addresses non-
uniform stress in the gusset and may indicate stresses in the
Ncrit = FXcrit sin γ − FYcrit cos γ (85) z region much higher than indicated by the Uniform Stress
Method. As such, the increase to address nonuniform stress
The gusset should be evaluated for these forces using an is not proposed for this method.
interaction method such as the von Mises yield criterion or For designs employing the Concentrated Stress Method,
interaction Equation 9-1 in the AISC Steel Construction stresses may redistribute along the weld due to beam inelas-
Manual. ticity. As such, the stress distribution corresponding to the
instantaneous center of rotation method may be impossible
Weld Design to achieve with the beam web strength provided. The weld
The design of welds should provide adequate strength to in the z1 zones should be evaluated for the force normal to
transfer forces across the gusset-to-beam interface (F V, the member axis, Rz1. For welds in the center region (Lg1 −
FN, and Mf ) and adequate ductility to achieve the assumed 2z1), the normal force is FN1. The shear force (parallel to
stress distribution. The weld size for a double fillet need the member axis) in both regions may be taken as F V1/Lg1.
not exceed s of the gusset thickness (for an adequately Often the weld size required in the z regions will be sub-
sized gusset), as discussed for single-plate connections in stantially greater than that required in the center region.
the AISC Steel Construction Manual (2017); this weld size To address strain compatibility of the linear weld group
permits yielding of the gusset before weld rupture. (This consisting of a larger weld size in the z regions and a
proportioning rule implicitly accepts use of a resistance smaller size in the center, two measures are proposed. First,
factor greater than 0.75.) Weld sizes greater than s of the the weld size in the z region may be sized to develop the
gusset thickness are not effective in developing their full strength of the gusset plate (e.g., a double fillet weld of at
force because they are limited by the gusset capacity (and least s of the gusset thickness); this ensures that the defor-
thus also indicate an inadequate gusset thickness); however, mation required of that zone of the joint may be provided
increasing a weld that requires less than s of the gusset by the gusset, and thus the full strength of both that region
thickness up to this value allows for yielding of the gusset and the center region can be achieved.
before weld rupture and thus permits stress redistribution. Second, the two welds may be proportioned so that their
strains are consistent with the design strength utilized. This
Uniform Stress Method may be done by analyzing the deformation of the differ-
ent weld elements, as in the instantaneous center of rota-
Under the Uniform Stress Method, the weld adequacy tion method. The authors have found satisfactory designs
should be evaluated using methods from the AISC Steel by proportioning the weld group with the weld in the center
Construction Manual, such as the instantaneous center of zone being s of the size of the welds in the z region and
rotation, which represents both weld strength and the limits then extending the larger z-region weld to the 4 point of
on weld ductility. As a minimum, the weld should be large the gusset at each end. It is expected that, with more study,
enough to resist the local stress consistent with the Uniform those minima could be reduced or eliminated. Alterna-
Stress Method. The required strength per unit length is: tively, the weld size selected for the z region may be used
2 2 for the entire gusset length.
⎛ 4M f 1 FN1 ⎞ ⎛ FV1 ⎞
ru = ⎜ 2 + ⎟ +⎜ ⎟ (86)
⎝ Lg1 Lg1 ⎠ ⎝ Lg1 ⎠
PROPOSED DESIGN PROCEDURE
The AISC Seismic Design Manual (2018) utilizes the The design of chevron and full-height gussets may be gov-
25% increase related to gussets at beam-column-brace con- erned by design considerations other than the local forces
nections to promote ductility per the AISC Steel Construc- addressed in this paper. In such cases, the material effi-
tion Manual and Hewitt and Thornton (2004). Hadad and ciency of the Concentrated Stress Method cannot be real-
Fortney (2020) determined a ratio of maximum to average ized, and the Uniform Stress Method (which is simpler
stress of approximately 3 (including a standard deviation) to implement) may be convenient. The following design
for weld stresses in their finite element analyses of chevron procedure may be used to minimize the complexity of the
connections. They suggest that the factor of 3 be applied in required calculations:

1. Establish parameters. length and this minimum length. If this length
is excessive, consider reinforcing the member
1.1 Determine the forces F V, FN, and Mf , acting on the
web (or using a different member).
gusset-member interface.
3.1.2. Determine the required gusset thickness per
1.2 Determine the optimal gusset-plate length based
Equations 36 and 38. Revise gusset length if
on the brace-to-gusset connection (and any other
necessary.
considerations). If desired, determine the optimal
gusset thickness.
3.2. Analyze connection and check member
1.3 For connections with gussets on opposite flanges,
3.2.1. Determine the length of the zone z (Equa
determine shear-strength apportionment for the two
tions 40, 41, and 42). Use the maximum length
gussets per Equations 51 and 52.
from these three equations.
2. Try the Uniform Stress Method. 3.2.2. Determine the concentrated force Rz (Equa-
tion 26).
2.1. Check if the optimal gusset-plate length exceeds the
minimum length required for the Uniform Stress 3.2.3. Determine Vmc (Equation 27); check member
Method using Equation 15 or 55. If so: shear.
2.2. Check member.
3.3. Design gusset.
2.2.1. Determine Vmc (Equation 14); check shear.
3.3.1. Check the gusset section at the interface
2.2.2. Evaluate web local yielding and web crippling with the flange in the center zone. (Gusset
Equation 17. horizontal section at the interface with the
flange in the z zone is implicitly checked by
2.3. Design gusset. the required thickness calculation.)
2.3.1. Design the gusset section parallel to the 3.3.2. Check the transverse gusset section for the
member axis using the Uniform Stress Method forces from Equations 65, 69, and 71. (Any
Equation 63. procedure in the AISC Manual may be used.)
2.3.2. Check the transverse gusset section for the 3.3.3. Check the diagonal gusset section for the
forces from Equations 65, 66, and 68. (Any forces from Equations 80 (or 83), 84, and 85.
procedure in the AISC Manual may be used.) (Any procedure in the AISC Manual may be
used.)
2.4. Design the gusset–member interface weld. (Design
for peak stress using the Uniform Stress Method 3.4 Design weld.
distribution; apply appropriate ductility factor or size
3.4.1. Design zone z weld.
to develop the gusset plate strength.)
3.4.2. Design center-zone weld.
3. If the Uniform Stress Method design is unsatisfactory,
try the Concentrated Stress Method. Note that this recommended design procedure implicitly
checks member shear in the connection region for both
3.1 Select gusset length.
the Uniform Stress Method and the Concentrated Stress
3.1.1. Check minimum length required for the Method. It does not include a check of the member for
Concentrated Stress Method (Equation 34). combined axial and bending forces within the connection
Use the maximum of the optimal gusset-plate region based on the authors’ experience and judgment.

Table 1. Brace Forces
Brace Axial Force Shear Component Normal Component
F F cos(γγ ) F sin(γγ )
(kips) (kips) (kips)
F1,1 568 364 436
F1,2 653 418 502
F2,2 511 327 393
F2,1 588 376 451
DESIGN EXAMPLE
The connection shown in Figure 20 will be designed following the recommended procedure, proceeding from the Uniform
Stress Method to the Concentrated Stress Method developed in this study to eliminate reinforcement.
Given:
The brace design forces are presented in Table 1. All brace angles are 50.2° from horizontal. To facilitate subsequent calcula-
tions, the shear and normal components of the brace forces are determined and presented in the table.
Both beam and gusset are Grade 50 material. The beam is a W24×94 (ϕVn = 375 kips; A=27.7 in.2; Z=254 in.4 ), 25 feet long.
The workpoint is at the beam centerline:
dm
em = (8)
2
= 12.15 in.
The beam moment due to loading other than from braces, MM, is 80 kip-ft.
Based on the brace-to-gusset connection (not shown), the minimum gusset length is 48 in. For the brace-to-gusset connection
design, a w-in.-thick gusset is optimal, and the depth required is 21 in.
Fig. 20. Design example.

Table 2. Connection Forces
Combination
Equation Gusset 1 Gusset 2 (total or difference)
F V (i) (kips) 1 782 703 78.5
FN (i) (kips) 2 65.5 58.9 6.6
Mf (i) (kip-in.) 3 9500 8550 18000
/
Vef (i) VefTOT 51; 52 0.526 0.474 1.0
Solution:
Design of Gusset 1
1. Establish Parameters

Optimal gusset dimensions have been given. The forces acting on the flange from each of the gussets is shown in Table 2,
which also shows the apportionment factors for beam effective shear strength.
2. Try the Uniform Stress Method

The minimum gusset length is determined from:
2MTot
Lg ≥ (55)
VefTot
18,000 kip-in.
=
375 kips
= 96.1 in.
For 48-in. gussets both above and below the beam, the Uniform Stress Method requires a web thickness of:
FV1 FV 2
Lg1 + Lg2
tw ≥ (62)
ϕ v 0.6 Fy
F +F
= V1 V 2
ϕ v 0.6 Fy Lg
782 kips + 703 kips
=
(1.0)0.6 ( 50 ksi )( 48 in.)
= 1.03 in.
This would require a W24×250. (Using the same gusset length, a W21×248 or a W18×211 would also be suitable.) Alterna-
tively, an increase in effective shear strength of 96.1/ 48 = 2.0 could be achieved by a web doubler of w × 18 in.:
2MTot
Lg ≥ (55)
VefTot
2MTot
=
(ϕ vVnbeam + ϕvVn doubler )
2 (18,000 kip-in.)
=
[375 kips + 1.0 ( 0.6)( 0.75 in.)(18 in.)(50 ksi )]
= 46.2 in.

The minimum extents of the doubler can be determined using Equation 13.
The Uniform Stress Method would require a 96-in. gusset without reinforcement, a much heavier beam, or significant rein-
forcement to permit a gusset on the order of the optimal 48-in. length. As none of these is desirable, the design will proceed
with the Concentrated Stress Method.
3. Try the Concentrated Stress Method

For the Concentrated Stress Method, the effective shear strength, Vef TOT, must be reduced considering the net unbalanced force:
FN1 FN 2
VefTot = ϕ vVn − − (56)
2 2
6.6 kips
= 375 kips −
2
= 372 kips
The lower gusset will be designed to utilize no more than 52.6% of the available member shear strength per Equation 51.
Mf 1
Vef 1 = VefTot (51)
MTot
9,500 kip-in.
= (372 kips)
18,000 kip-in.
= 196 kips
Minimum Gusset Length and Corresponding Thickness

Assuming (for preliminary design) that the transverse force R z1 is equal to this effective shear strength, the minimum gusset
length is:
Mf 1 Vef 1
Lg1 > + − 5k (34)
Vef 1 ϕ w Fy t w
9,500 kip-in. 196 kips
= + − 5 (1.38 in.)
196 kips (1.0 )( 50 ksi )( 0.515 in.)
= 49.2 in.
The approximate length that corresponds to a w-in. gusset is:

Mf 1 Vef 1
Lg1 > + (39)
Vef 1 ϕt Fy t g1
9,500 kip-in. 196 kips
= +
196 kips ( 0.9 )( 50 ksi )( 0.75 in.)
= 54.3 in.
As this length does not include the effect of the shear, a slightly larger value will be used, and the effect of shear addressed in
the determination of the minimum length z1. A 56-in. effective gusset length will be investigated. (The length also results in an
economical weld design, which is presented later in the example.) The detailed length is 58 in., recognizing that the weld will
not extend to the very end of the gusset.

The minimum bearing length z1 based on the limit states of web local yielding and web crippling is:
2
Lg1 Lg1 Mf 1
z1 ≥ − − − 5k (40)
2 4 w Fy t w
56.0 in. ( 56.0 in.)2 9,500 kip-in.

= − − − 5 (1.38 in.)
2 4 (1.0 )( 50 ksi )( 0.515 in.)
= 0.73 in.
⎡ Vef 1 tw ⎤ ⎛ d m ⎞ ⎛ t f ⎞ 1.5
z1 ≥ ⎢ − 1⎥ (42)
⎢⎣ ϕ n 0.80t w2 EFy t f ⎥⎦ ⎝ 3 ⎠ ⎝ t w ⎠
⎡ 196 kips 0.515 in. ⎤ ⎛ 24.3 in.⎞ ⎛ 0.875 in.⎞ 1.5
=⎢ − 1⎥
⎢⎣ ( 0.75) ( 0.80 ) ( 0.515 in.)2 ( 29,000 ksi ) ( 50 ksi ) ( 0.875 in.) ⎥⎦ ⎝ 3 ⎠ ⎝ 0.515 in.⎠
= −3.87 in.
The low value from Equation 40 and the negative value from Equation 42 indicate that the force being developed, Rz, does not
require a significant bearing length to satisfy the limit states of web local yielding and web crippling.
The minimum length z1 corresponding von Mises yield criterion for stresses in the gusset is obtained from Equation 41:
Mf 1
Lg1 Lg12 ϕt
z1 = − − (41)
2 4 ⎛ FV1 ⎞
2
2
(Fytg1) − ⎜ ⎟
⎝ ϕ v 0.6 Lg1 ⎠
9,500 kip-in.
56.0 in. ( 56.0 in.)2 0.9
= − −
2 4 ⎛ 782 kips ⎞
2
[(50 ksi )( 0.75 in.)]2 − ⎜ (1.0 ) ( 0.6 ) ( 56 in. ) ⎟
⎝ ⎠
= 7.38 in.
The maximum length z1 is:

Mf 1
z1 ≤ Lg1 − (43)
Vef1
9,500 kip-in.
= 56.0 in. −
196 kips
= 7.51 in.
The value from Equation 41 will be used. The corresponding transverse force is:
Mf 1
Rz1 = (26)
Lg1 − z1
9,500 kip-in.
=
56.0 in. − 7.38 in.
= 195 kips
The limit states of web local yielding, web crippling, and gusset combined tension and shear yielding are implicitly checked
by the gusset-length selection and this length z1 determined above (Equations 40, 42, and 41). This value may also be used
to check gusset stress using Equation 36; however, gusset stress is implicitly checked by the selection of a dimension z1 that
complies with Equation 41.

The beam shear is evaluated considering 52.6% of the shear due to the total unbalanced force:
Vmc1 = Vma1 + Rz1 (27)
⎛ M f 1 ⎞ FN1 − FN 2
= + Rz1
⎝ M Tot ⎠ 2
⎛ 6.6 kips ⎞
= 0.526 + (195 kips )
⎝ 2 ⎠
= 197 kips
Vmc1 197 kips
=
ϕvVn 375 kips
= 0.525 ≤ 0.526 o.k.
This is consistent with the apportionment of available beam shear strength between the two gussets established in Table 2.
Beam Moment and Axial Force in Gusset Region

For completeness, the combined effects of the internal beam moment and axial force are evaluated. (The proposed design
procedure does not include this evaluation.)
Although Equation 49 permits a more precise calculation of beam moment, the moment is typically small, and the upper-bound
value is used here for convenience for gusset 2, as that design has not been performed. Adapting Equation 49 to include the
effect of two gussets gives:
2
MCh1 ⎡ z1 ⎛ z1 ⎞ ⎤ MCh2 ⎡ z2 ⎛ z2 ⎞ 2 ⎤ ( FN1 − FN 2 ) L beam
Mmax = ⎢ −⎜ ⎥+ ⎢ −⎜ ⎟ ⎥+ + MM
2 ⎢ Lg1 ⎝ Lg1 ⎟⎠ ⎥ 2 ⎢⎣ Lg2 ⎝ Lg2 ⎠ ⎥⎦ 4
⎣ ⎦
For gusset 1:
1 ⎡ z1 ⎛ z1 ⎞
2⎤
1 ⎡ 7.38 in. ⎛ 7.38 in.⎞ 2⎤
⎢ − ⎥= ⎢ − ⎥
2 ⎢ Lg1 ⎝ Lg1 ⎠ ⎥⎦ 2 ⎢⎣ 56.0 in. ⎝ 56.0 in.⎠ ⎥⎦
⎣
= 0.0572
For gusset 2, the dimension z2 has not been determined. In this example (with Mf 2 < Mf1), z2 could reasonably be assumed to
be less than or equal to z1 if Lg2 = Lg1. The general limit is:
1 ⎡ z2 ⎛ z2 ⎞
2⎤
1
⎢ − ⎥≤
2 ⎢ Lg2 ⎝ Lg2⎠ ⎥⎦ 8
⎣
⎛ 9,500 kip-in.⎞ ⎛ 8,550 kip-in.⎞ ( 65.5 kips − 58.9 kips ) ( 25 ft )

Mmax ≤ 0.0572 + 0.125 + + 80 kip-ft
⎝ 12 in./ft ⎠ ⎝ 12 in./ft ⎠ 4
≤ 255 kip-ft
The axial force is conservatively taken as the maximum at the end of the connection region, assuming a symmetric distribution
of collector forces:
FV1 FV 2
Pu = −
2 2
782 kips 703 kips
= −
2 2
= 39.5 kips

Assuming the section is fully braced at this location, the full section strength is used:
ϕc Pn = ϕc As Fy
= ( 0.9 )(27.7 in.2 )(50 ksi)
= 1,250 kips
ϕb Mn = ϕb ZFy
= ( 0.9 )(254 in.3 )(50 ksi)
= 11,400 kip-in.
= 953 kip-ft.
Pu
= 0.03
ϕc Pn
Mu
= 0.27
ϕb Mn
The interaction check from AISC Specification Equation H1-1b is used:

1 Pu Mu 1
+ = ( 0.06 ) + 0.27 (from Spec. Eq. H1-1b)
2 ϕc Pn ϕb M n 2
= 0.30 o.k.
Gusset Selection
A w-in.-thick, 21-in.-deep, and 58-in.-long (56-in. effective length) gusset will be investigated.
t g1 = 0.75 in.
d g1 = 21.0 in.
Lg1 = 56.0 in.
Gusset Check at Section Parallel to the Member Axis at Beam Flange

The Concentrated Stress Method implicitly checks the gusset over the lengths z1 for combined stresses in determining the
minimum length z1 (Equation 41). For the center zone between the lengths z1, Equation 38 gives the interaction ratio:
2 2
FV1 FN1
+ (38)
ϕ v 0.6Fyt g1Lg1 ϕt Fy tg1 ( Lg1 2z1 )
2 2
⎡ 782 kips ⎤ ⎧ 65.5 kips ⎫
= ⎢ ⎥ +⎨ ⎬
⎣ 0.6 ( 50 ksi ) ( 0.75 in.) ( 56.0 in.)⎦ ⎩ ( 0.9 ) ( 50 ksi ) ( 0.75 in.)[ 56.0 in. − 2 ( 7.38 in.)] ⎭
= 0.622
Gusset Check at Mid-Length Transverse Section

The adequacy of the gusset depth is verified examining a section of the gusset transverse to the member axis. (See Figure 18.)
The gusset force in direction of member is:
1
Ng1 = ( F1,2 cos γ 1,2 − F1,1 cos γ 1,1 ) (65)
2
1
= ( 418 kips − 364 kips )
2
= 27 kips
The axial resistance of the gusset is:
ϕt Pn = 0.9Fy d g1t g1
= ( 0.9 ) ( 50 ksi ) ( 21.0 in.) ( 0.75 in.)
= 709 kips
The gusset moment is:

⎛ d g1 ⎞ FN1 ⎛ Lg1 z1 ⎞
Mg1 = Ng1 em + − − (71)
⎝ 2 ⎠ 2 ⎝ 4 2⎠
⎛ 21.0 in.⎞ 65.5 kips ⎛ 56 in. 7.38 in.⎞
= ( 27 kips ) 12.15 in. + − −
⎝ 2 ⎠ 2 ⎝ 4 2 ⎠
= 280 kip-in.
The flexural resistance of the gusset is:

2
d g1t g1
ϕb Mn = 0.9Fy
4
( 21.0 in.)2 ( 0.75 in.)
= 0.9 ( 50 ksi )
4
= 3,720 kip-in.
Gusset shear transverse to member:

FN1
Vg1 = F1,1 sin γ 1,1 − Rz1 + (69)
2
65.5 kips
= 436 kips − 195 kips +
2
= 274 kips
The shear resistance is:

ϕ vVn = 1.00 ( 0.60Fy ) dg1t g1
= 1.00 ( 0.60 ) ( 50 ksi ) ( 21.0 in.) ( 0.75 in.)
= 473 kips
Using the von Mises interaction equation:

2 2 2 2
⎛ Mg1 Ng1 ⎞ ⎛ Vg1 ⎞ ⎛ 280 kip-in. 27 kips ⎞ ⎛ 274 kips ⎞
+ + = + +
⎝ ϕt M n ϕt Pn ⎠ ⎝ ϕ vVn ⎠ ⎝ 3,720 kip-in. 709 kips ⎠ ⎝ 473 kips ⎠
= 0.590
Gusset Check along Diagonal Section

The gusset will be checked along the critical diagonal section (Figure 19). The bolt gage, W, is 8 in. and the transverse dimen-
sion Yclip is 8.0 in.
Lg1 dm cos γ + W
X crit = − (72)
2 2sin γ
56.0 in. ( 24.3 in.) cos ( 50.2° ) + 8 in.
= −
2 2sin ( 50.2° )
= 12.7 in.

This value is greater than z1.
The diagonal section length is:
Dcrit = X crit cos γ + ( d g1 − Yclip) sin γ (75)
= (12.7 in.) cos ( 50.2°) + ( 21.0 in. − 8 in.) sin (50.2°)
= 18.1 in.
The forces acting on the section are:

X crit
FX crit = FV1 (77)
Lg1
12.7 in.
= (782 kips)
56.0 in.
= 177 kips
X crit − z1
FYcrit = Rz1 + FN1 (81)
Lg1 − 2z1
12.7 in. − 7.38 in.
= 195 kips + ( 65.5 kips )
56.0 in. − 2 (7.38 in.)
= 204 kips
⎛ z⎞ ( X − z )2
Rz1 X crit − 1 + FN1 crit 1
⎝ 2⎠ 2 ( L g − 2z1 )
ecrit = (82)
FYcrit
(12.7 in. − 7.38 in.)2
(195 kips ) ⎛195 kips −
7.38 in.⎞
+ ( 65.5 kips )
⎝ 2 ⎠ 2 [ 56.0 in. − 2 ( 7.38 in.)]
=
204 kips
= 8.72 in.
⎛ D cos γ ⎞ D sin γ
Mcrit = FYcrit ecrit − crit + FXcrit crit (83)
⎝ 2 ⎠ 2
⎡ (18.1 in.) cos ( 50.2°) ⎤ (18.1 in.) sin ( 50.2°)
= ( 204 kips ) ⎢8.72 in. − ⎥ + (177 kips )
⎣ 2 ⎦ 2
= 1,825 kip-in.
Vcrit = FXcrit cos γ + FYcrit sin γ (84)
= 270 kips
Ncrit = FXcrit sin γ − FYcrit cos γ (85)
= (177 kips) sin (50.2°) − ( 204 kips ) cos (50.2°)
= 5 kips
The gusset is evaluated for these forces using the von Mises yield criterion:
2 2
⎛ 4 M crit N crit ⎞ ⎛ Vcrit ⎞
⎜ ϕ F t D 2 + ϕ F t D ⎟ + ⎜ ϕ 0.6F t D ⎟
⎝ t y g1 crit t y g1 crit ⎠ ⎝ v y g1 crit ⎠
2 2
⎡ 4 (1,825 kip-in.) 5 kips ⎤ ⎡ 270 kips ⎤
= ⎢ 2 + ⎥ +⎢ ⎥
⎣ ( 0.9 ) ( 50 ksi ) ( 0.75 in.) (18.1 in.) ( 0.9 ) ( 50 ksi ) ( 0.75 in.) (18.1 in.) ⎦ ⎣1.00 ( 0.60 ) ( 50 ksi ) ( 0.75 in.) (18.1 in.)⎦
= 0.94 ≤ 1.0 o.k.

Gusset Weld (z-Region)
The weld along the length z1 must deliver a normal force equal to Rz1; it must also deliver a shear force proportional to its length:
Nweld = Rz1
= 195 kips
z1
Vweld = Fv1
Lg
7.38 in.
= (782 kips)
56.0 in.
= 103 kips
The weld in this zone therefore resists a force at an angle:

Pu = Nweld 2 + Vweld 2
= (195 kips ) + (103 kips )

2 2
= 221 kips
The angle θ is tan-1(195/103) = 62° from the weld axis. Using AISC Specification Equation J2-5:
Pu ≤ ϕRn
2
= ϕ n 0.6FEXX (1.0 + 0.5sin1.5 θ) wL w
2
Pu
w≥
2
ϕn 0.6FEXX (1.0 + 0.5sin1.5 θ ) ( 2z1)
2
221 kips
w≥
( 0.75) 0.6 ( 70 ksi ) ⎡⎣1.0 + 0.5sin1.5 (62.2°)⎤⎦ 2 ( 7.38 in.)
= 0.474 in.
A double-sided 2-in. fillet weld will be used. The weld size need not exceed s of the gusset plate thickness:
5
w ≤ tg1
8
5
= (0.75 in.)
8
= 0.469 in.
This weld must include the z region; a 14-in. length will be used to extend to the gusset 4 points. Because this weld fully devel-
ops the gusset strength, deformation compatibility is inherently addressed.
Gusset Weld (Center Region)

The weld in the center region must be checked. The required strength is based on:
Lg1 − 2z1 = 56.0 in. − 2 (7.38 in.)
= 41.2 in.
Nweld = FN1
= 65.5 kips

Lg1 − 2 z1
Vweld = FV1
Lg1
56.0 in. − 2 (7.38 in.)
= ( 782 kips)
56.0 in.
= 576 kips
Pu = Nweld 2 + Vweld 2
= ( 65.5 kips) + ( 576 kips)

2 2
= 579 kips
The angle θ is tan-1(65.5/ 576) = 6.5°. Using AISC Specification Equation J2-5:
Pu
w≥
2
ϕn 0.6FEXX (1.0 + 0.5sin1.5 θ ) 2 (Lg1 − 2 z1 )
2
579 kips
=
( 0.75) 0.6 ( 70 ksi ) ⎡⎣1.0 + 0.5sin1.5 ( 6.5°)⎦⎤ 2 (41.2 in.)
= 0.310 in.
A pair of the c-in. fillet welds will be used.

The weld group consisting of the z-region and center welds conforms to both of the deformation compatibility recommenda-
tions: the z-region welds develop the gusset strength and extend to the gusset 4 points, and the center region welds are s of
the size of the welds in the z-regions.
Design Summary
Figure 21 shows the design based on the calculations above. The 2-in. fillet welds in the z-regions are presented as x-in. fillet
welds over the c-in. full-length fillet welds.
A similar design is required for gusset 2. If the same method is followed, the beam shear resulting from the two gussets (each
designed for a portion of Vef ) combined with the net unbalanced load will not exceed the beam shear capacity.
Fig. 21. Gusset design.

CONCLUSIONS AISC (2018), Seismic Design Manual, 3rd Ed., American
Institute of Steel Construction, Chicago, Ill.
This study provides equations that can be used in the design
of bracing connections to eliminate the need for web rein- Euler, L. (translated by Hewlett, J.) [published 1765; 1822
forcement. Recommendations are made for the selection translation], “Of a New Method of Resolving Equations
of braced-frame beams and columns to facilitate connec- of the Fourth Degree,” Elements of Algebra, Longman,
tion design. The design method allows engineers to use the Hurst, Rees, Orme, and Company, London.
gusset plate to limit the shear demand on the member web. Fortney, P.J. and Thornton, W.A. (2015), “The Chevron
These equations can be used to assess the effects of mem- Effect—Not an Isolated Problem,” Engineering Journal,
ber depth and gusset length on the required member shear AISC, Vol. 52, No. 2, pp. 125–164.
strength in order to optimize member selection and gusset Fortney, P.J. and Thornton, W.A. (2017), “The Chevron
design. The Concentrated Stress Method presented allows Effect and Analysis of Chevron Beams—A Paradigm
for significantly smaller gusset plates than the Uniform Shift,” Engineering Journal, AISC, Vol. 54, No. 4,
Stress Method for an unreinforced section. For cases in pp. 263–296.
which the Uniform Stress Method requires an undesirably
Hadad, A.A. and Fortney, P.J. (2020), “Investigation on the
large gusset or the use of a web doubler, the Concentrated
Performance of a Mathematical Model to Analyze Con-
Stress Method may permit a more economical design.
centrically Braced Frame Beams with V-Type Bracing
Configurations,” Engineering Journal, AISC, Vol. 57,
ACKNOWLEDGMENTS No. 2, pp. 91–108.
The authors would like to recognize their debt to the pio- Hewitt, C.M. and Thornton, W.A. (2004), “Rationale Behind
neering work of William Thornton and the late Pat Fortney. and Proper Application of the Ductility Factor for Brac-
Leigh Arber provided invaluable assistance in the review ing Connections Subjected to Shear and Transverse Load-
of this paper. ing,” Engineering Journal, AISC, Vol. 41, No. 1, pp. 3–6.
Richards, P., Miller, B., and Linford, J. (2018), Finite Ele-
REFERENCES ment Evaluation of the Chevron Effect in Braced Frames,
Brigham Young University Report No. SSRP-2018/02.
AISC (2012), Seismic Design Manual, 2nd Ed., American
Institute of Steel Construction, Chicago, Ill. Sabelli, R. and Arber, L. (2017), “Design of Chevron Gusset
Plates,” 2017 SEAOC Convention Proceedings.
AISC (2016a), Seismic Provisions for Structural Steel Build-
ings, ANSI/AISC 341-16, American Institute of Steel Sabelli, R., Saxey, B., and Richards, P. (2020), “The Chevron
Construction, Chicago, Ill. Effect in Web Shear at Midspan Gussets,” Proceedings of
the 17th World Conference on Earthquake Engineering,
AISC (2016b), Specification for Structural Steel Buildings,
Sendai, Japan.
ANSI/AISC 360-16, American Institute of Steel Con-
struction, Chicago, Ill. Thornton, W.A. (1984), “Bracing Connections for Heavy
Construction,” Engineering Journal, AISC, Vol. 21,
AISC (2017), Steel Construction Manual, 15th Ed., Ameri-
No. 3, pp. 139–148.
can Institute of Steel Construction, Chicago, Ill.

Guide for Authors
Scope Engineering Journal is dedicated to the improvement and
advancement of steel construction. Its pages are open to all who
wish to report on new developments or techniques in steel design,
research, the design and/or construction of new projects, steel
fabrication methods, or new products of significance to the uses of
steel in construction. Only original papers should be submitted.
General Papers intended for publication should be submitted by email
Margaret Matthew, editor, at [email protected].
The articles published in the Engineering Journal undergo
peer review before publication for (1) originality of contribution;
(2) technical value to the steel construction community; (3) proper
credit to others working in the same area; (4) prior publication of the
material; and (5) justification of the conclusion based on the report.
All papers within the scope outlined above will be reviewed by
engineers selected from among AISC, industry, design firms, and
universities. The standard review process includes outside review by
an average of three reviewers, who are experts in their respective
technical area, and volunteers in the program. Papers not accepted
will not be returned to the author. Published papers become the
property of the American Institute of Steel Construction and are
protected by appropriate copyrights. No proofs will be sent to
authors. Each author receives three copies of the issue in which his
contribution appears.
Manuscripts Manuscripts must be provided in Microsoft Word format. Include a
PDF with your submittal so we may verify fonts, equations and figures.
View our complete author guidelines at aisc.org/ej.
Smarter. Stronger. Steel.
American Institute of Steel Construction
130 E Randolph St, Ste 2000, Chicago, IL 60601
312.670.2400 | aisc.org/ej

Ej 2021 q1 Issue

Uploaded by

Copyright:

Available Formats

Ej 2021 q1 Issue

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ej 2021 q1 Issue

Uploaded by

Copyright:

Available Formats

Engineering

11 AISC Provisions for Web Stability under

33 Critical Temperature of Axially Loaded Steel

45 Design for Local Member Shear at Brace and

Dedicated to the development and improvement of steel construction,

Investigation on the Performance of a Mathematical

T here are several aspects of the original paper that need

ISSN 0013-8029 ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 1

Fig. 1. Inelastic deformation of a gusset plate under seismic deformation.

2 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

Fig. 2(a). Fracture of gusset-plate welds designed by stress.

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 3

(c) Buckled brace just prior to connection weld fracture

(d) Fractured weld

Fig. 2(b-d). Fracture of gusset-plate welds designed by stress.

4 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

Fig. 3. Axial-load, shear-force, and bending-moment diagrams for idealized

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 5

6 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

(c) Frame with chevron beam designed to current AISC

Fig. 4(c). Nonlinear analyses of chevron-braced frames.

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 7

8 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 9

INTRODUCTION cases of the concentrated compression load being applied:

ISSN 0013-8029 ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 11

Fig. 2. Web buckling failure in a full-width

12 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

Web local for lend ≤ d for lend ≤ H

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 13

14 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

Table 3. Average Measured Rectangular HSS Cross-Sectional Dimensions

Table 4. Average Measured Rectangular HSS Chord Material Properties

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 15

Fig. 5. Connection load-displacement curves for tests X1 and X2.

16 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 17

Fig. 6. Correlation between 37 welded rectangular HSS to rectangular HSS

18 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

Fig. 7. Correlation between 16 welded rectangular HSS to rectangular HSS

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 19

20 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 21

22 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 23

24 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 25

26 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

Limit State of Web Local Yielding

Fig. 12. Rectangular HSS-to-HSS cross-connection subjected to branch axial compression.

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 27

Limit State of Web Local Crippling

Limit State of Web Compression Buckling

Critical Buckling Stress, Fcr

28 / ENGINEERING JOURNAL / FIRST QUARTER / 2021

Flexural Buckling of the Chord Sidewalls

where Ag for one sidewall is given by Equation 12. For two sidewalls,

ENGINEERING JOURNAL / FIRST QUARTER / 2021 / 29

ACKNOWLEDGMENTS Hb Overall height of rectangular HSS branch member or

M f 1 = MCh1 − MBm1 (3) FV1d m

Moment for |x| ≤ Lg1/ 2 − z1