Engineering

Journal
First Quarter 2025 | Volume 62, No. 1

1 Letter from the Editor

3 Load-Dependent Critical Temperatures for

Standard Fire Resistance of W-Shape Floor
Beam Assemblies: Experimental Validation
and Simplified Analysis
Michael M. Drury and Spencer E. Quiel

27 Generalized Elastic Lateral-Torsional

Buckling of Steel Beams
Robert S. Glauz and Benjamin W. Schafer

43 Strength Coefficients for Eccentrically

Loaded Weld Groups
Bo Dowswell

53 Errata
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 Kristin Hall

Officers
Chair
Hugh J. McCaffrey

Vice Chair
Glenn R. Tabolt

Secretary/Legal Counsel
Edward Seglias

President
Charles J. Carter, SE, PE, PhD

Senior Vice Presidents

Scott L. Melnick
Mark W. Trimble, PE

Vice Presidents
Todd Alwood
Brandon Chavel, PE, PhD
Carly Hurd, CAE
Christopher H. Raebel, SE, PE, PhD
Michael Mospan
Brian Raff

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. Published by the American
Institute of Steel Construction at 130 E Randolph Street, Suite 2000, Chicago, IL 60601. Archives: Search at [Link]/ej.
Copyright 2025 by the American Institute of Steel Construction. All rights reserved. No Article downloads are free for current
part of this publication may be reproduced without written permission. The AISC logo is a members and are available for a nominal fee
registered trademark of AISC. for non-members.
Letter from the Editor
Dear Readers,

Happy New Year! At the start of this shiny new year, I would like to recognize all the hard work
of our many reviewers, last year and every year. Their contributions are invaluable to the success
of the Journal as we continue to strive to bring you the very best articles and information in the
steel construction industry.

Is there a steel design topic you would like to see in EJ? We are always looking for ideas for
papers. Authors interested in submitting papers should visit our website at [Link]/ej for
author guidelines and submittal information.

Best wishes for a healthy and happy 2025!

Sincerely,

Margaret A. Matthew, P.E.

Editor

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 1

2 / ENGINEERING JOURNAL / FIRST QUARTER / 2025
Load-Dependent Critical Temperatures for Standard
Fire Resistance of W-Shape Floor Beam Assemblies:
Experimental Validation and Simplified Analysis
MICHAEL M. DRURY and SPENCER E. QUIEL

ABSTRACT
Comprehensive results of ASTM E119 (2019) standard fire tests (performed by AISC/AISI in 2015) are used to validate load-dependent critical
temperature relationships that conservatively predict the thermally induced loss of flexural resistance for W-shape floor beam assemblies.
The 16 tested assemblies used the same W8×28 section (coated with the same thickness of passive spray-applied fire resistive material),
supported 22 in. (64 mm) of lightweight concrete (reinforced with welded wire mesh) on a 2 in. (51 mm) corrugated metal deck, and had a
clear span of 154w in. (3.93 m) in one-way bending. Four specimen groups in the following configurations were tested with four specimens
each: restrained ends with composite slab, unrestrained ends with composite slab, restrained ends with noncomposite slab, and unre-
strained ends with noncomposite slab. The four specimens in each group were each tested with a constant applied flexural load but at a
range of magnitudes, inducing maximum bending moment from 23 to 60% of the section’s ambient nominal moment capacity. The results
of these tests clearly demonstrate a relationship between the loss of flexural resistance and the increase in steel beam temperature (particu-
larly the bottom flange temperature) as a function of applied loading. The fire-induced temperature increases in the protected steel beams
are then used to validate lumped mass thermal calculations per the AISC Specification (2022) and Part 1-2 of Eurocodes 3 and 4 (CEN, 2005,
2008), which are classified as simple analysis methods per Section A-4.2.4d of the AISC Specification. The results of this study demonstrate
that simplified thermal analysis methods can be combined with load-dependent critical temperature relationships to conservatively predict
the standard fire resistance of W-shape floor beam assemblies at the onset of flexural failure.

Keywords: standard fire resistance design (SFRD), W-shape steel floor beam assembly, composite vs. noncomposite slab, restraint of
thermal expansion, critical steel temperature at flexural runaway, hourly fire resistance ratings.

INTRODUCTION metal deck to the top flange of the beam prior to concrete
placement (see Figure 1).
Floor systems in North American steel-framed build-
Floor beams often require the application of passive fire
ing construction are often comprised of wide-flange (i.e.,
protection to meet minimum hourly fire resistance rat-
W-shape) beams that support a concrete slab. Beams that are
ings as a function of the building’s size, occupancy, and
not part of a moment-resisting frame are typically designed
purpose per the International Building Code (IBC) (ICC,
as one-way simple spans that are supported by shear con-
2020). Passive fire protection for W-shape beams in current
nections. The slab is typically reinforced with steel bars or
practice often consists of spray-applied fire-resistive mate-
welded wire reinforcement (WWR) and is cast onto cor-
rial (SFRM), which is a lightweight cementitious product
rugated or fluted light-gage metal decking. The beam and
with high thermal resistance. Due to their sheer quantity,
slab are often constructed to be composite at their interface,
floor beams can require a significant portion of the over-
thus achieving a degree of strain compatibility and ampli-
all amount of SFRM applied to all steel framing through-
fying their collective flexural stiffness and moment capac-
out the building. Hourly fire resistance ratings describe the
ity. Composite action is commonly developed by welding
time needed to exceed a thermal or structural limit criterion
headed shear studs at regularly spaced intervals through the
when the assembly is subjected to a “standard fire” heat-
ing regime, such as those provided in ASTM E119 (2019),
UL 263 (2020), or ISO 834 (2019). As shown in Figure 2,
these temperature time histories undergo a rapid rise dur-
ing the first 10 min, after which the rate of temperature
Michael M. Drury, Associate, Wiss, Janney, Elstner Associates, Inc., Princ- increase slows before approaching 2000°F (1093°C) after
eton, N.J. Email: mdrury@[Link]
3 hr. Beyond that time, these high temperatures would con-
Spencer E. Quiel, Associate Professor, Department of Civil and Environmen- tinue to gradually increase with no subsequent decay phase
tal Engineering, Lehigh University, Bethlehem, Pa. Email: seq213@[Link]
(corresponding) or any consideration of suppression from sprinklers (i.e.,
active fire protection). The initial 2 hr period of a standard
fire curve is intended to generally represent the ramp-up of
Paper No. 2023–11R temperature in a post-flashover building compartment, and

ISSN 2997-4720 ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 3

the indefinite continuation of high temperature exposure paper, but the reader is referred to several recent publica-
will ensure that the thermal and/or structural performance tions by ASCE for more guidance (ASCE, 2020; LaMalva,
criteria limits are eventually reached. 2018).
For comparison, a “natural” fire temperature time his- In a typical standard fire evaluation, a floor beam
tory would increase as a function of the geometry, fuel load, assembly (which includes a supported slab) with an
ventilation, and material characteristics of a given building SFRM-protected W-shape section is heated from below
compartment (CEN, 2009) and then eventually enter a sub- while carrying a flexural load. Standard fire tests for floor
sequent decay phase until burnout. A representative natu- beam assemblies are conducted as either restrained (i.e., the
ral fire curve from a previous study by the authors (Drury ends of the beam are fully restricted from axial thermal
and Quiel, 2023a) is plotted for comparison in Figure 2. A expansion or rotation) or unrestrained (i.e., the beam ends
natural fire curve can be used to evaluate the survivabil- are free to thermally expand and rotate). These idealized
ity and resilience of a structural assembly to a realistic fire configurations do not represent actual boundary conditions
exposure. Such considerations are outside the scope of this in a building but are intended to bracket the partial restraint

Fig. 1. Representative illustration of a composite W-shape floor beam,

including thermocouple placement for an ASTM E119 standard fire test.

2200 1204
2000 1093
1800 982
Temperature (°F)

Temperature (°C)

1600 871
1400 760
1200 649
1000 538
800 427
600 316
400 204
200 93
0 -18
0 30 60 90 120 150 180
Time (min)
R-C Test Curve
ASTM E119 / UL 263
ISO 834
Natural Fire (Drury et al., 2023a)

Fig. 2. Comparison of the R-C test’s furnace temperature time history with select standard and natural fire curves.

4 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 Table 1. Temperature and Deflection Performance Limits per Published
Standards for Standard Fire Testing of W-Shape Floor Beam Assemblies
Type of Limit Source Limit
• Midspan deflection: L / (400d)
2

Midspan ASTM E119-19 THEN

deflection • Deflection rate per minute: L2/ (9000d)
BS 476-21:1987 • Midspan deflection: L/ 20
• Average steel beam temperature from 4 thermocouples per Figure 1 (Ts,E119):
Critical 1100°F (593°C)
temperature of ASTM E119-19 AND
steel beam • Maximum steel beam temperature at a single thermocouple (Ts,MAX): 1300°F
(704°C)
• Average temperature increase above ambient at the unexposed top-of-slab
Critical
surface: 250°F (139°C)
temperature
ASTM E119-19 AND
transmission
• Maximum temperature increase above ambient at any point on the unexposed top-
through slab
of-slab surface: 325°F (181°C)

to thermal expansion and end rotation that would be pro- that of the web, Ts,web, and bottom flange, Ts,BF, calculated
vided by realistic connections and a continuous reinforced as an average of its two thermocouples because its upper
concrete floor slab. The heated assembly is monitored for surface is in contact with the supported floor slab, thus par-
its ability to sustain the applied flexural load as well as tially shielding it from heating. Ts,BF and Ts,web are often
for the temperature increase in the steel beam and thermal similar, with Ts,BF being more critical to flexural strength.
transmission through the concrete slab. The time at which For the other ASTM E119 thermal limit, a weighted average
a given criteria is met is then rounded down to the nearest temperature, Ts,E119, over a given cross section is calculated
half-hour to denote the hourly fire resistance rating for the as a simple average of all four thermocouple measurements
tested section configuration. Standard fire resistance design at that section. Ts,E119 will therefore trend hotter than a true
(SFRD), which is widely used in current practice, uses these area-weighted average value due to the placement of two
hourly ratings as a comparative indicator of ultimate capac- thermocouples on the bottom flange and only one each on
ity under a standardized fire exposure. the other plates (Drury et al., 2020).
Table 1 summarizes the thermal and structural perfor- The results of standard fire tests are compiled in catalogs
mance limit criteria that are used to signify “failure” dur- such as the UL Fire Resistance Directory (2019a). Due to
ing a standard fire test of a composite steel floor beam. The the size limitations of most available furnaces, standard fire
deflection limits per ASTM E119 (2019) or BS 476 (BSI, tests are usually conducted on floor beam assemblies with
2008, 1987) are intended to represent a loss of flexural shorter spans (10–17 ft) and use a relatively small W-shape
resistance (i.e., at or near the onset of runaway failure) and section (such as a W8×28 or W12×14). SFRM thicknesses
are calculated as a function of the clear span of the beam, needed for larger sections in actual construction can be cal-
L, and the distance between the extreme fibers of the cross culated using Appendix 4 of the AISC Specification (2022),
section in compression and tension, d, which includes the with additional guidance and examples available in ASCE
slab thickness and depth of the fluted deck for a composite 29-05 (2007) and AISC Design Guide 19 (Ruddy et al.,
section (Alfawakhiri et al., 2016; Drury and Quiel, 2023b, 2003). These resources provide semi-empirical conversion
2023a). In a standard fire test per ASTM E119, tempera- equations based on the relative ratio of cross-sectional area
tures are measured at a minimum of three sections along to fire-exposed perimeter between the actual floor beam
the length of the specimen, with four thermocouples at each and the tested specimen from the rated assembly. These
section as shown in Figure 1. The maximum temperature methods are implemented as qualification testing per of
limit per ASTM E119 is evaluated against the maximum AISC Specification Section A-4.3 and provide a straight-
reading among all thermocouples on the steel beam. forward translation of tested fire resistance to a steel floor
Because the bottom flange has the greatest exposure beam design.
to fire among these three plates, one of its thermocouples As an alternative to the prescriptive application of qual-
often indicates the section’s maximum temperature. The ification test results, IBC Section [Link] (ICC, 2020)
top flange temperature, Ts,TF, typically remains cooler than permits the use of engineering analysis to demonstrate an

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 5

equivalent fire resistance rating as required in Table 601 assembly test can be used to obtain a restrained assembly,
for an actual member or assembly. To pursue this option, unrestrained assembly, and an unrestrained beam rating
the designer would need adequate thermal and structural by applying the varied thermal and structural criteria
information about the setup and outcome of a standard limits to the single set of test data. An unrestrained beam
fire test, with which the analysis approach could be vali- rating that is extrapolated from a restrained assembly
dated for the tested specimen. Once validated, the analy- test has slightly different performance limits than that
sis approach could then be applied to a model of the actual obtained directly from a true unrestrained beam test.
assembly (with expected loading and realistic boundary Despite this, the published listings of fire resistance
conditions), and a so-called virtual standard fire test would ratings for floor beam assemblies (UL, 2019a) do not
be conducted to determine equivalency to the tested fire typically clarify which test configuration was used or
resistance rating. However, there are a few significant chal- which criterion (thermal or structural) was exceeded to
lenges when attempting to use standard fire test outcomes obtain a particular hourly rating.
in this way:
• Per AISC Specification Section A-4.3.1 (2022), demon-
• Per Section [Link] of ASTM E119 (2019), the applied strating equivalency to a standard fire resistance rating
flexural loading should represent “the maximum load is only permitted via “advanced methods of analysis”
condition allowed under nationally recognized structural as described in Section A-4.2.4c of the same document.
design criteria unless limited design criteria are These advanced methods typically use finite element (FE)
specified and a corresponding reduced load is applied.” thermal and structural models to evaluate structural fire
Appendix X7 of ASTM E119 states that fire resistance response. Expertise and project resources (namely, time
tests have been historically conducted using loads based and budget) are needed to conduct these FE analyses, and
on maximum allowable stresses. However, the exact load the benefits offered by such an approach (either in terms
used to test a floor beam assembly is rarely reported in of enhanced performance or fire protection cost savings)
the corresponding fire resistance rating. Variations in must outweigh the costs associated with the design effort.
assumed material strengths can lead to a range of loads The simple methods of analysis per Section A-4.2.4d,
being considered as “the maximum load condition.” Also, or the recently added critical temperature method per
the flexural design of floor beams at ambient conditions Section A-4.2.4e would be a more accessible tool for
is often governed by serviceability limits rather than by tailoring standard fire resistance ratings to realistic
strength. Current application of the ASTM E119 test floor beam assemblies; however, these approaches are
results in practice do not account for realistic variability not currently permitted by the AISC Specification
of design loading and flexural utilization (which can be for this application. Currently, no provisions exist in
calculated as the ratio of applied moment to nominal Section A-4.2.4e related to determining the fire resistance
ambient moment capacity, M/ Mn) for realistic floor beam of floor beam assemblies.
designs. Also, ASTM E119 does not provide any load-
In an effort to simplify the determination of equivalent
based correlation between the thermal and structural
fire resistance for W-shape floor beams, researchers with
performance limits in Table 1.
AISC and AISI performed 16 ASTM E119 standard fire
• The evolution of standard fire testing over time has tests in 2015 for which the steel beam was protected with a
produced terminology and testing procedures that can be contoured coating of SFRM (Alfawakhiri et al., 2016). The
challenging to understand if someone is not intimately results of these tests were used to develop the inaugural UL
familiar with the testing process. Per Table X3.3 of D982 assembly listing (UL, 2019b) and are reproduced in
ASTM E119 (2019), hourly ratings for steel floor beams this paper with permission from AISC. All specimens used
that support concrete slabs can be obtained from four the same W8×28 section, SFRM thickness, corrugated con-
different testing configurations: restrained assembly, crete slab characteristics, and span geometry. Four specimen
unrestrained assembly, restrained beam, and unrestrained groups in the following configurations were tested with four
beam. Assembly tests also limit thermal transmission specimens each: restrained ends with composite slab (R-C),
through the slab to its unexposed top face, and a restrained unrestrained ends with composite slab (U-C), restrained
beam test does not require the slab edge above the end of ends with noncomposite slab (R-NC), and unrestrained ends
the beam to also be restrained (i.e., only the end of the with noncomposite slab (U-NC). The four specimens in each
steel beam itself is restrained). Several classifications of group were each tested with a constant applied flexural load
hourly ratings can be obtained from a single test when but at a range of magnitudes, inducing a maximum moment,
using some of these testing configurations (Berhinig and M, ranging from 23–60% of nominal flexural strength, Mn.
Alfawakhiri, 2014; Bono, 1970; LaMalva et al., 2020; The results of those tests are used in this paper to illustrate
Ruddy et al., 2003). For example, a single restrained the relationship between flexural response and the increase

6 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

in steel beam temperature as a function of applied loading. the ratio of applied moment, M, versus the nominal flexural
The results of other quasi-standard tests in the published lit- strength due to yielding at ambient temperature, Mn:
erature will also be used to further demonstrate this relation- AISC Specification Equation A-4–23:
ship for a wider range of composite beam cross sections,
⎛M⎞
span lengths, slab configurations, and boundary conditions. Tcr = 816 − 306 ln (in °F) (1)
⎝ Mn ⎠
The results of the AISC/AISI tests were used as vali-
dation for thermo-mechanical FE models in Chapter 6 of Eurocode 3, Part 1-2 (CEN, 2005) provides a similar
the doctoral dissertation by Drury (2022). The reader is relationship for critical temperature, Tcr, of a generic steel
directed to that reference for guidance on the implemen- section as a function of initial ambient flexural utilization,
tation of advanced analysis per Section A-4.2.4c of the M/ Mn:
AISC Specification for developing equivalent standard fire Eurocode 3, Part 1-2, Equation 4.22:
resistance. This paper instead demonstrates the application
of simple methods of thermal analysis per Appendix 4 of ⎡ 1 ⎤
Tcr = 39.19 ln ⎢ − 1⎥ + 482 (in °C) (2)
the AISC Specification as well as Part 1-2 of Eurocodes 3 ⎢ ⎛M⎞
3.833 ⎥
and 4 (CEN, 2005, 2008) for predicting critical steel tem- ⎢ 0.9674 ⎥
⎣ ⎝ Mn ⎠ ⎦
peratures per the load-dependent relationship. The time at
which those critical temperatures are reached under expo- Per both specifications, the values of Tcr from Equa-
sure to a standard fire curve could then be used to indicate tions 1 and 2 should be applied to the average temperature
an equivalent standard fire resistance. of the steel cross section, Ts. The plots in Figure 3 show
that Equations 1 and 2 both provide similar relationships
LOAD-DEPENDENT CRITICAL between Tcr and M/ Mn for generic steel sections, with the
TEMPERATURE RELATIONSHIPS AISC expression providing an approximate 5% conserva-
tive value.
Existing Specifications Eurocode 4, Part 1-2 (CEN, 2008) permits the applica-
tion of Tcr from Equation 2 as a limit for either Ts or Ts,BF
Load-dependent critical temperatures for steel beams are in a floor beam that supports a composite slab. The AISC
available in current design standards but do not necessarily Specification does not currently provide an explicit critical
target composite floor beams. For example, AISC Specifi- temperature relationship for floor beams that support a con-
cation Section A-4.2.4e (2022) provides a critical tempera- crete slab; however, Table A-4.2.4 of the AISC Specifica-
ture relationship, Tcr, for flexural yielding of a continuously tion provides a flexural retention factor (kcb = Mn,T/ Mn) for
braced beam not supporting a concrete slab as a function of composite beams as a function of Ts,BF (which is intended to

1700 1700
900 900

1500 800 1500 800

Temperature (°C)
Temperature (°C)

Temperature (°F)
Temperature (°F)

1300 700 1300 700

1100 600 1100 600

900 500 900 500

400 400
700 700
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Utilization (M/Mn) Utilization (M/Mn)
(a) Ts,BF (b) Ts,E119

AISC Tcr,BF
Bletzacker-FC Bletzacker-NC Zhao and Kruppa
AISC Tcr
Wang et al. Jiang et al. Kordosky et al.
EC3 Tcr
Choe et al.
ASTM E119 Tcr

Fig. 3. Critical steel beam temperatures from the existing experimental literature as a function of initial ambient flexural utilization.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 7

represent the maximum temperature of the steel section and Ts,BF and calculate Ts,E119 at the time of flexural failure per
govern its flexural resistance). The values in Table A-4.2.4 the criteria defined in each study, and the results are plot-
can be reframed as a function of critical bottom flange tem- ted as a function of initial ambient flexural utilization in
perature, Tcr,BF, versus flexural utilization, M/ Mn, and this Figure 3. The experimentally measured values of critical
relationship is plotted in Figure 3 for comparison. As would Ts,BF in Figure 3(a) shows slightly greater dispersion than
be expected, the values for the AISC Tcr,BF per Table A-4.2.4 the corresponding critical Ts,E119 in Figure 3(b), which sug-
are greater than those for Tcr from Equations 1 and 2 up to gests that a more consistent result may be obtained when
a flexural utilization of ∼75%. These critical temperature web and top flange temperatures are also incorporated in
curves are compared against the results of quasi-standard the critical temperature evaluation. Both plots clearly show
fire testing from the published literature in the following that the existing flat critical temperature limits per ASTM
section. E119 are unable to capture the test results across varia-
tions in applied loading. Both AISC Specification Equation
Comparison with Published Results of A-4-23 (Equation 1) and the Eurocode critical temperature
Quasi-Standard Fire Tests per Equation 2 provide a good correlation to critical Ts,E119
across all tests in Figure 3(b). In particular, AISC Specifi-
Numerous studies have used a quasi-standard fire test to
cation Equation A-4-23 provides a lower bound to most of
demonstrate a load-dependent relationship between steel
the test data, with the only exception being the Zhao and
beam temperature and one-way flexural performance of
Kruppa (1997) results at a flexural utilization ratio of 0.72
composite and noncomposite floor assemblies, thus support-
(which is above a realistic upper bound of 0.6 for flexural
ing the notion that ultimate fire resistance can be correlated
utilization at service levels). The curves for Tcr per Equa-
to a critical temperature. Table 2 lists seven experimental
tions 1 and 2 provide an unnecessarily conservative predic-
programs from 1967 to 2020 that tested one-way composite
tion of critical Ts,BF in Figure 3(a); however, the Tcr,BF based
floor beams under a heating regime that either matched or
on Table A-4.2.4 of the AISC Specification provides a rea-
closely resembled a standard fire curve. All tests used var-
sonably conservative best prediction across all data points.
ied shear stud layouts to induce a broad range of compos-
In the next section, these curves will be further evaluated
ite action. Test parameters included both flat and profiled
against the results of the aforementioned 16 standard fire
slabs (with varying compressive strengths and concrete unit
tests by Alfawakhiri et al. (2016).
weights), protected and unprotected steel beams (with vary-
ing steel grade, based on their construction era), lengths
ranging from 11 to 40 ft (3.35–12.2 m), and varying degrees STANDARD FIRE TESTING
of axial and rotational end restraint (due to variation in the
available support conditions for a particular furnace). Floor Beam Specimen Design
Though limited in some cases by laboratory-specific
As shown in Figure 4, test specimens consisted of a
constraints to prevent damage to the furnace, the flexural
hot-rolled, W8×28 [50 ksi (345 MPa) yield strength per
failure criteria used for each study was generally simi-
ASTM A992/A992M (2020)] that supported 2.5 in.
lar to that shown in Table 1 and were intended to capture
(64 mm) of lightweight concrete (LWC) on a 2 in. (51 mm)
the moment at which the beam would no longer be able to
deep galvanized, fluted metal deck [0.037 in. (0.9 mm)
carry the superimposed load. Flexural loading for each test
thick]. The steel beams were cut to a total length of 163 in.
was applied via multiple point loads to induce a maximum
(4.14 m), and the ends were capped with welded steel plates
bending moment that equaled between 26 and 80% of the
[12 in. × 8 in. × 0.38 in. (305 mm × 203 mm × 9.7 mm)].
ambient (unfactored) moment capacity. Where nominal val-
The beam ends were positioned on steel angle bearing sup-
ues were absent in the test reporting, reasonable strength
ports as shown in Figure 4, leaving a clear span of 154.75 in.
assumptions were made based on the reported properties
(3.93 m). The slab had a total width of 47 in. (1.19 m), and
from material testing. For reference, expected flexural utili-
the flutes were oriented perpendicular to the beam span.
zation in practice will range from 20 to 60% of the nominal
The LWC had a nominal dry density of 115 pcf (1840
moment capacity, with an upper limit of 70% being reached
kg/m3) and a specified minimum compressive strength
in rare design circumstances (Newman, 1999).
of 3 ksi (20.7 MPa). A single layer of 6×6 W1.4×W1.4
Nearly all test programs reported a full thermal profile
welded wire reinforcement (WWR) [65 ksi (450 MPa)
of Ts,TF, Ts,BF, and Ts,web in the steel beam at failure, with the
yield strength per ASTM A1064 (2022)] was placed at
exception of tests CB-SP and CB-DA-SC per Choe et al.
mid-thickness of the 2.5 in. concrete topping. The slabs
(2020) (which only reported a bottom flange or maximum
were cured for ∼8 months prior to testing, and the relative
temperature due to thermocouple malfunctions during
humidity of the concrete was measured to be less than 70%
testing). The reported temperatures were used to identify
on the day of testing per ASTM E119 (2019).

8 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 Table 2. Summary of Previous Quasi-Standard Fire Tests on One-Way Floor Beam Assemblies
Flexural Failure
Source Span Specimen Design End Conditions Protection Fire Exposure Initial M/Mn Criteria
Beam: 12W×27 d in. (22 mm) SFRM “…when the
Unrestrained
4 in. (102 mm) on the beam, 2 Constant for assembly could
Bletzacker 16.8 ft Slab: (3 tests), various
flat slab in. (13 mm) on ASTM E119 each test at no longer sustain
(1967) (5.12 m) restraint levels
Fully composite (2 tests), underside of deck 41–60% the vertical
(9 tests)
Noncomposite (10 tests) (1.5–2.5 hr rating) design load.”
IPE300 Midspan
Beam:
(roughly W12×22) Unprotected (1 test); concrete
Zhao and Simply supported Constant for
16 ft 4.7 in. (119 mm) 1 in. (25 mm) mineral crushing (3 tests);
Kruppaa Slab: beam with ISO 834 each test at
(4.88 m) flat slab wool on the beam midspan yielding
(1997) unrestrained slab 26–72%
(3 tests) of steel section
Fully composite (4 tests) (1 test)
Built-up
Beam:
(roughly W12×26)
Wang Simply supported Constant for Maximum
17.3 ft 3.9 in. (99 mm) 0.43 in. (11 mm)
et al. Slab: beam with ISO 834 each test at deflection of
(5.27 m) flat slab SFRM on the beam
(2017) unrestrained slab 50–57% L 30
/
Fully composite (1 test),
50% composite (1 test)
Built-up
Beam:
Jiang (roughly W12×26)
Simply supported
et al. 16.7 ft 3.9 in. on 3 in. 0.39 in. (10 mm) Constant at Maximum
beam with ISO 834
Test SB2 (5.09 m) Slab: (99 mm on 76 SFRM on the beam 80% deflection of L 20
/
unrestrained slab
(2017) mm) profiled deck
Fully composite (1 test)
Beam: W12×26 Maximum
Unprotected
3.25 in. on 2 in. Partially deflection of
Kordosky (1 test); 2 hr rated Constant for
11 ft Slab: (83 mm on 51 restrained simple L2 (400d) and
/
et al. SFRM on the beam ASTM E119 each test at
(3.35 m) mm) profiled deck connection with maximum
(2020) (1 test); connections 32%
unrestrained slab deflection rate of
24% composite (2 tests) protected for both
L2 (9000d)
/
Beam: W18×35 Restrained simple
3.25 in. on 3 in connections with SFRM at 2 hr rating Parametric fire
Choe Constant for
40 ft Slab: (83 mm on 76- slab restrained on the beam, that initially Maximum
et al.b each test at
(12.2 m) mm) profiled deck (2 tests) or 3 hr rating on the emulated deflection of L 20
/
(2020) c 45%
unrestrained connections ASTM E119
82% composite (4 tests) (2 tests)
a
Zhao and Kruppa (1997) Test 5 was partially composite but was halted prematurely due to loading equipment failure; it is therefore excluded.
b
Choe et al (2020) Test CB-DA was automatically unloaded prior to reaching significant displacement or a collapse state; it is therefore excluded.
c
Constant 3791 Btu/s (4000 kW) heat release rate was targeted—actual temperatures were 266°F (130°C) hotter than ASTM E119 beyond 15 min.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 9

 For the eight noncomposite specimens, a single pud- width of 39.7 in. (1010 mm) (neglecting contributions from
dle weld was made in every deck flute to the top flange. the WWR). The W8×28 section is compact at ambient con-
The nominal ambient plastic moment strength, Mn, of ditions, and the top flange is assumed to have continuous
the noncomposite specimen was, therefore, 1,360 kip-in. bracing from the slab.
(153.7 kN-m), which accounts for the contributions of the As shown in Figure 5(a), a gap of 1.5 in. (38 mm) was left
steel beam only. For the eight composite specimens, a between the furnace’s restraining support and the welded
cluster of four shear studs [each w in. (19 mm) in diam- end plate of each specimen. For the restrained configura-
eter and 3.5 in. (90 mm) in length] were welded through tion in Figure 5(b), the fluted deck was positioned such that
each deck flute [see Figures 5(b) and 5(c)] to the top flange concrete would be placed directly against the furnace sup-
at 12 in. (305 mm) on longitudinal centers, resulting in a port. Also, a groove was cut into the deck so that wet con-
fully composite design per the 2010 AISC Specification crete would infill the gap between the beam’s end plate and
(AISC, 2010; Vinnakota et al., 1988). The nominal moment the furnace support. This configuration thus provided full
strength, Mn, of the composite specimens was calculated bearing restraint to both the beam and slab against out-
to be 2,540 kip-in. (287 kN-m) based on an effective slab ward thrust from thermal expansion and rotation, though

Fig. 4. Side view schematic of the floor beam test specimen, support conditions, and applied loading.

(a) Typical cap-plated beam end (b) Restrained configuration (c) Unrestrained configuration
prior to deck installation

Fig. 5. Plan-view photos of the beam end conditions before concrete placement (provided by Farid Alfawakhiri, courtesy of AISC).

10 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

no restraint would be provided if the specimen were to pull four equally spaced locations along its length (for a total
away from the furnace support. For the unrestrained con- of 16 thermocouples per beam). Additional thermocouples
figuration in Figure 5(c), the deck was trimmed such that were placed throughout the furnace to monitor its inter-
the concrete placement would stop short of the furnace nal temperature as well as the temperature increase of the
wall, with a wood rail used to set the unrestrained edge slab at a few locations on the top and bottom surfaces and
of the slab. Fiberglass fill was placed into gap between the within. The average of those thermocouples for the R-C test
beam’s end plate and the furnace support during concrete group is plotted in Figure 2 and closely follows the ASTM
placement to prevent accidental infill. Both the wood rail E119 standard fire curve. The average furnace temperature
and fiberglass fill were removed prior to testing, preserving for all other test groups were similarly consistent with the
the gap and allowing unrestrained expansion and rotation at ASTM E119 fire curve and are not plotted for brevity.
the ends of the beam and slab. Constant flexural loading beyond the specimen
As shown in Figure 6(a), lightweight SFRM with a nomi- self-weight was applied to each beam using concentrated
nal dry density of 15 pcf (240 kg/m3) was applied to the forces at the three locations in Figure 4 via hydraulic cylin-
steel beams at a uniform thickness of 1 in. (25 mm). The gap ders and bearing plates as shown in Figure 6(b). Each of the
between the top flange and underside of the deck between four specimens in a simultaneous furnace environment was
flutes was also filled with SFRM, as is common practice. loaded such that the maximum moment at midspan would
No other SFRM was applied to the underside of the fluted reach a different percentage of Mn, ranging from 23 to 60%
deck other than some minor overspray just beyond the edges (see Table 3). These moment calculations assumed simple
of the top flange. It should be noted that this SFRM appli- end supports (since significant rotational restraint would
cation did not formally target a UL-listed assembly hourly not occur at the beam ends under pure flexural loading at
rating and was applied solely to enable comparison between ambient conditions) and an effective span length of 159 in.
test specimens on the basis of their parametric variation. (i.e., the average of the 163 in. total length and the 154w in.
clear span). It should be noted that the 57 to 60% flexural
Test Setup and Loading utilization level was referred to as the “100% load intensity”
by the AISC test team (Alfawakhiri et al., 2016) since it
Four specimens were simultaneously tested at a time,
represented a maximum expected or service load condition
side-by-side, in the same furnace using the same heat expo-
per ASTM E119 (2019) and allowable strength design per
sure. As shown in Figure 6(b), the unbonded interface
the AISC Specification (2022). For simplicity, tests herein
between the longitudinal slab edges of neighboring spec-
are labeled using the following nomenclature: restrained or
imens were covered from above with ceramic wool blan-
unrestrained (R/U), composite or noncomposite (C/NC),
kets to mitigate the escape of heat during testing. Prior to
and initial flexural utilization percentage. For example,
SFRM application, four thermocouples were installed on
R-C-60 refers to the restrained, composite beam tested at
the steel beam cross section per the pattern in Figure 1 at
an initial ambient flexural utilization of 60%.

(a) Typical SFRM-protected beam (showing the (b) Load application

longitudinal seam between slabs of adjacent specimens)

Fig. 6. Pre-test photos of the experimental setup (provided by Farid Alfawakhiri, courtesy of AISC).

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 11

 Table 3. Summary of Flexural Loading and Corresponding Critical Temperatures per Load-Dependent
Relationships in the AISC Specification (2022) and Eurocodes 3 and 4 (CEN, 2005, 2008)
Initial Flexural Tcr,BF [°F (°C)] Tcr [°F (°C)]
Utilization, Load P AISC Specification EC4 Part 1-2, AISC Specification EC3 Part 1-2,
M/ Mn [kips (kN)] Table A-4.2.4 Section [Link].3(4) Equation A-4-23 Equation 4.22
60% 17.8 (79.2) 1100 (593) 1030 (554) 972 (522) 1030 (554)
Composite

48% 14.1 (62.7) 1209 (654) 1096 (591) 1041 (560) 1096 (591)

36% 10.5 (46.7) 1313 (712) 1177 (636) 1129 (609) 1177 (636)

24% 6.82 (30.3) 1429 (776) 1288 (698) 1253 (678) 1288 (698)

57% 8.75 (38.9) 1130 (610) 1047 (564) 990 (532) 1047 (564)
Noncomposite

45% 6.90 (30.7) 1232 (667) 1113 (600) 1058 (570) 1113 (600)

34% 5.05 (22.5) 1331 (721) 1193 (645) 1146 (619) 1193 (645)

23% 3.20 (14.2) 1448 (787) 1303 (706) 1270 (688) 1303 (706)

Before any heat was applied, simultaneous testing for four Thermal Response
side-by-side specimens was initiated by slowly applying all
The temperatures measured in the flanges and web of the
loads for several minutes until equilibrium was achieved.
four steel beams with the highest loading percentages are
Heating via the ASTM E119 standard fire curve was then
plotted in Figure 7 for demonstration—all other specimens
initiated, and the loading was held constant until the spec-
with lower loading have very similar steel temperature time
imen was deemed to be no longer capable of sustaining
histories as those shown here and are thus not provided for
the applied loads. Generally, this meant that the midspan
brevity. The solid “Avg” curves for Ts,TF, Ts,BF, and Ts,web
deflection had surpassed the corresponding ASTM E119
represent the mean value of all thermocouples on a cor-
criteria per Table 1:
responding plate over the length of the beam. The dashed
• Composite beam: curves represent the maximum and minimum recorded
temperatures at any single thermocouple on that plate over
° Deflection limit = L / [400(dbeam+dslab)] = 4.81 in.
2
the length of the beam. The average temperature in each
(122 mm)
plate is very consistent between specimens. The bottom
° Deflection rate limit = L / [9000(dbeam+dslab)] =
2
flange experiences the greatest temperature increase due to
0.214 in/min (5.44 mm/min) its large amount of fire-exposed surface. The web also has
a significant amount of exposed surface and is thinner than
• Noncomposite beam:
the flanges, and it therefore develops temperatures that are
° Deflection limit = L / (400dbeam) = 7.51 in. (191 mm)
2
up to ∼200°F (93.3°C) lower than the bottom flange. The
difference between maximum and minimum recorded tem-
° Deflection rate limit = L / (9000dbeam) = 0.334 in/min
2
perature for the bottom flange and web typically does not
(8.48 mm/min)
exceed ∼100°F (37.8°C) because they are more uniformly
Upon reaching these limits, the applied loading for that heated. The top flange has a greater range between its min-
specimen would be removed, although active heating would imum and maximum value and develops a lower average
continue until all four specimens reached the deflection limit temperature because it is in contact with the slab and has
states. The midspan deflection of each beam was measured less surface area exposed to fire. The top surface of the top
as the increase in distance between the top-of-slab and the flange conducts heat to the slab, which has significant ther-
underside of the overhead loading frame. These measure- mal mass and remains cooler than the steel beam through-
ments were initiated after the beam was fully loaded and out the test. Note that the thermal transmission through
before heating was applied, such that the small amount of the slab is not considered to be within the scope of this
deflection in the loading frame due to the reactions from load paper and is therefore not presented; the reader is instead
application would be neglected. referred to the paper by Alfawakhiri et al. (2016) for that
information.

12 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 The time at which each specimen reached flexural run- Ts,TF curve of the other R-NC specimens did not show any
away is also marked in each plot in Figure 7. The tempera- noticeable signs of SFRM delamination. Likewise, the steel
ture time history for U-NC-57 in Figure 7(d) shows a sudden temperature curves for the composite specimens in Fig-
increase in Ts,TF at this point, indicating that the SFRM on ure 7(a-b) are relatively smooth throughout heating and also
that plate began to delaminate when the beam lost its flex- indicate no obvious disruption to the integrity of the SFRM.
ural resistance. In particular, the maximum temperature for Figure 8 shows good overall consistency among all
Ts,TF converges to the fire curve near the end of the heat 16 specimens regarding their average values of Ts,BF and
application around ∼160 min. As will be shown in the next Ts,E119, calculated across all pertinent thermocouple loca-
section, U-NC-57 experienced the largest deflection among tions for each specimen. The curves for Ts,BF in Figure 8(a)
the four specimens plotted here, which likely enabled the are especially consistent due to the preservation of SFRM
SFRM delamination. It should be noted that U-NC-45 and integrity on the bottom flange throughout heating, even past
U-NC-34 also showed similar signs of top flange SFRM the onset of flexural runaway. Due to late onset of SFRM
delamination at the onset of flexural runaway but at a later delamination from the top flange, the three gray curves for
time due to their lower level of loading. The Ts,TF curve for Ts,E119 in Figure 8(b) from U-NC specimens with 57%, 45%,
R-NC-57 in Figure 7(c) also shows indications of minor and 34% flexural utilization trend higher beyond 120 min
SFRM delamination (via a sudden uptick in maximum Ts,TF) than those for all other specimens.
but not until 30 min after the loss of flexural resistance. The

2000 1093 2000 1093

1800 982 1800 982
1600 871

Temperature (°C)
Temperature (°F)
1600 871
Temperature (°C)
Temperature (°F)

1400 760 1400 760

1200 649 1200 649
1000 538 1000 538
800 427 800 427
Runaway

600 316 600 316

Flexural

Runaway
Flexural
400 204 400 204
200 93 200 93
0 -18 0 -18
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time (min) Time (min)
(a) R-C-60 (b) U-C-60

2000 1093 2000 1093

1800 982 1800 982
1600 871
Temperature (°C)

1600 871
Temperature (°F)

Temperature (°C)
Temperature (°F)

1400 760 1400 760

1200 649 1200 649
1000 538 1000 538
800 427 800 427
Runaway

Runaway

600 316
Flexural

600 316
Flexural

400 204 400 204

200 93 200 93
0 -18 0 -18
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time (min) Time (min)
(c) R-NC-57 (d) U-NC-57

BF Avg TF Avg Web Avg

BF Min/Max TF Min/Max Web Min/Max

Fig. 7. Measured time histories of Ts,BF, Ts,web , and Ts,TF for indicated
specimens (experimental data provided by Farid Alfawakhiri, courtesy of AISC).

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 13

Flexural Response • The corresponding ASTM E119 deflection limits for
composite and noncomposite specimens are marked as
The measured time histories of vertical midspan deflection
a dashed horizontal line in Figure 9, while an “×” is
for all 16 tests are plotted in Figure 9, with the ASTM E119
used to mark the time at which the deflection rate limit
criteria from Table 1 implemented as follows:
is met. Eleven of the 16 specimens met both criteria;

2000 1093
1800 982 2000 1093
1800 982

Temperature (°C)
Temperature (°F)

1600 871

Temperature (°C)
Temperature (°F)
1400 760 1600 871
1200 649 1400 760
1200 649
1000 538
1000 538
800 427
800 427
600 316 600 316
400 204 400 204
200 93 200 93
0 -18 0 -18
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time (min) Time (min)
Test Data 1LM(BF)-EC4 Test Data 1LM-AISC 1LM-EC3 3LM-EC4
(a) Ts,BF (b) Ts,E119

Fig. 8. Comparison of steel temperature time histories from all 16 test specimens (calculated
as an average of all longitudinal measurement locations on each beam) against lumped mass
temperature predictions (experimental data provided by Farid Alfawakhiri, courtesy of AISC).

R-C-24 U-C-24 R-C-36 U-C-36 R-NC-23 U-NC-23 R-NC-34 U-NC-34

R-C-48 U-C-48 R-C-60 U-C-60 R-NC-45 U-NC-45 R-NC-57 U-NC-57

(a) Composite (b) Noncomposite

Fig. 9. Measured time histories of vertical midspan deflection for standard fire
test specimens (experimental data provided by Farid Alfawakhiri, courtesy of AISC).

14 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 restrained specimens R-C-60, R-C-48, R-NC-57, R-NC- exceeded at different times according to the applied load
45, and R-NC-34 exceeded the deflection rate, but these level. Figures 10(a) and 11(a) clearly show that the hogging
tests were stopped before reaching the deflection limit moment at the ends of the restrained specimens caused sig-
to protect the testing equipment from the rapid onset of nificant transverse cracking in the top of the composite slab
runaway deflection. The end result for these specimens [Figure 10(b)] as well as plastic hinging at the end of the
was deemed to sufficiently signify the loss of flexural beam [Figures 10(c) and 11(b)]. Despite lacking a true con-
resistance. nection, the restrained specimens were able to develop these
hogging moments at their ends due to the compressive reac-
• The time at which the ASTM E119 thermal limits for
tion of the end plates against the furnace supports to resist
Ts,MAX or Ts,E119, whichever came first, are exceeded is
axial thermal expansion. As shown previously in Figure 7,
marked with a diamond on each curve in Figure 9. The
the bottom flange and web undergo a larger temperature
temperatures used to calculate the limits are reflected in
increase and would therefore experience more restraint of
the gray curves in Figure 8. All specimens in this test
thermal expansion than the top flange and slab. Larger hori-
program were governed by the average temperature
zontal compressive reactions near the bottom of the beam
criteria, Ts,E119. Ts,MAX was typically exceeded in the
would produce an eccentricity at the contact of the end plate
bottom flange 5–10 min later than Ts,E119.
against the support, thereby inducing the hogging moment
The ASTM E119 thermal limit was reached for all speci- (i.e., inducing upward rotation) that is observed at the ends
mens between 100 and 110 min, and Figure 9 shows that of the restrained specimens.
this flat threshold does not adequately describe the loss of For the unrestrained specimens in Figures 10(d) and
flexural resistance when variations in applied loading are 11(c), the gap between the furnace support and the ends
considered. In fact, U-C-60 was the first specimen to exceed of the specimens allowed thermal expansion and end rota-
both deflection criteria at ∼95 min, even though its TE119 was tion; therefore, there was no noticeable post-test transverse
still ∼50°F (28°C) lower than the corresponding thermal cracking or plastic hinging near the unrestrained ends. The
limit. R-C-24 reached flexural runaway just past 175 min, final parabolic deflected shapes of the unrestrained speci-
roughly 1 hr after the thermal limit was exceeded. The time mens in Figures 10(f) and 11(d) resemble the anticipated
needed to reach the thermal and structural limits generally deflected shapes for idealized simply supported boundary
showed closer agreement at higher flexural utilizations, but conditions intended by the standard test setup. The U-NC
the deflection limits were reached at significantly longer specimens in Figure 11(c) show very little slab cracking due
times when flexural utilization was reduced. to the relatively low engagement of the noncomposite slab;
As expected, the restrained specimens exhibited a stiffer however, increased deflections due to lower stiffness caused
initial flexural resistance response in Figure 9 than their some observable SFRM delamination, particularly at the
unrestrained counterparts at equivalent loads and, as a top flange [see Figure 11(d)]. This observation supports the
result, achieved slightly longer times to flexural runaway. onset of rapid temperature increase in the top flange after
All specimens show relatively similar initial deflection flexural runaway as shown previously for U-NC-57 in Fig-
down to a value of ∼1 in., after which the deflection rate ure 7(d). Conversely, Figures 10(c) and 10(f) show that none
for restrained specimens becomes shallower due to the of the composite beam specimens experienced noticeable
growing influence of restraint forces and hogging moment SFRM loss prior to the loss of flexural resistance.
at the beam ends. After they have stiffened and developed As shown in Figure 10(e), the U-C specimens developed
additional restraining stresses, however, the restrained large lengthwise longitudinal cracks just beyond the width
specimens then develop a rapid increase in deflection rate of the bearing plates for point load application. No trans-
toward runaway failure at lower magnitudes of deflection verse support was provided to the slab edges in these tests,
than in the unrestrained cases. The unrestrained specimens and it should be noted that longitudinal cracking would not
exhibit a more gradual overall deflection response as their be expected in an actual building with a continuous slab
temperatures increases, with the deflection rate accelerat- that spanned transversely to the next parallel floor beam.
ing toward runaway only in the last few minutes before the In these tests, the longitudinal cracking of the U-C speci-
loading was stopped. mens was likely caused by differential thermal expansion
between the beam and unrestrained composite slab (which
Observations from Post-Test Inspection was significantly cooler than the beam due to its larger ther-
mal mass, lower thermal conductivity, and lower ratio of
Photos from the post-fire inspection of each group of speci-
heated area to overall volume). The hotter beam is, there-
mens are shown in Figures 10 and 11. Photos taken from
fore, longitudinally restrained by the cooler slab due to
above show that all four specimens in each group exhib-
their composite interface. The unrestrained slab, as a result,
ited similar ultimate deflected shapes regardless of initial
develops nonuniform transverse stress distribution, and the
flexural utilization, even though flexural resistance was

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 15

 (a) R-C specimens from above (d) U-C specimens from above

(b) R-C slab edge transverse cracking (e) U-C longitudinal slab cracking

(c) R-C specimens from below (f) U-C specimens from below

Fig. 10. Post-test photos of the composite specimens (provided by Farid Alfawakhiri, courtesy of AISC).

16 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

longitudinal cracks indicate a shear plane in the transition loading on structural resistance to standard fire exposure.
between the highly composite center portion and the less The best correlation for those limits is observed at 48% ini-
restrained outer portions of the slab. tial flexural utilization, which is consistent with the fact
that the 1100°F (593°C) critical temperature limit for Ts,E119
Critical Temperature Verification correlates to an approximate 50% reduction in steel yield
strength (AISC, 2022; CEN, 2005). The AISC Specifica-
The plots in Figure 12 are similar to those in Figure 3, but the
tion flexural retention factor provides a conservative lower
data points from previous experimental studies have been
bound across all critical Ts,BF in Figure 12(a). AISC Speci-
replaced with those derived from the standard fire tests of
fication Equation A-4-23 (Equation 1) and the Eurocode
the 16 specimens. The following critical temperatures are
critical temperature per Equation 2 (which are intended to
taken for each specimen at the time at which the specimen
represent the temperature of the entire steel member) logi-
reached the ASTM E119 deflection rate limit (again, marked
cally provide a closer correlation to critical Ts,E119 across all
with an “×” in Figure 9): (a) maximum Ts,BF measured at
tests in Figure 12(b) but provide an unnecessarily conserva-
any location along the length of the beam and (b) maxi-
tive prediction of critical Ts,BF in Figure 12(a).
mum Ts,E119 at any one of the three cross-section locations
It should be noted that the noncomposite specimens con-
along the length of the beam, as shown in Figure 1(b). Both
sistently exhibited ∼100°F (37.8°C) higher critical tempera-
plots again show that the ASTM E119 flat critical tempera-
ture at the onset of flexural runaway than their composite
ture limits do not accurately capture the effects of applied

(a) R-NC specimens from above (c) U-NC specimens from above

(b) R-NC specimens from below (d) U-NC specimens from below

Fig. 11. Post-test photos of the noncomposite specimens (provided by Farid Alfawakhiri, courtesy of AISC).

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 17

 Table 4. Summary of Standard Fire Resistance Times (in minutes) Based on Thermal and Structural Test Results
Limit: ASTM E119 Limit: Limit:
Initial Flexural Deflection Rate AISC Tcr,BF AISC Tcr
Utilization, R-Test U-Test R-Test U-Test R-Test U-Test
M/ Mn Deflection Deflection Ts,BF Ts,BF Ts,E119 Ts,E119
60% 112 98 93 89 96 93
Composite

48% 116 107 104 100 101 101

36% 137 131 123 116 118 115

24% 174 171 161 141 144 140

57% 129 114 95 92 93 90

Noncomposite

45% 139 123 104 102 99 96

34% 151 144 124 117 114 108

23% 163 167 154 139 138 132

counterparts. It is likely that the noncomposite slab made Table 4 summarizes the times at which each specimen
a non-negligible contribution to the flexural resistance of reached the ASTM E119 deflection rate limit. Also shown
those beams; however, the loading applied to each noncom- are the times at which the recorded average Ts,BF and Ts,E119
posite beam was calculated as a percentage of the beam’s from each specimen (plotted previously in Figure 8) reach
plastic moment capacity only. Excess capacity relative to their corresponding AISC critical temperature from Table 3.
the assumed level of initial flexural utilization would there- The relationship between these fire resistance times is visu-
fore translate into a higher critical temperature at the onset alized in Figure 13 as a correlation plot, with thermally
of flexural runaway. This design concept indicates that the predicted values on the vertical axis and deflection-based
design flexural utilization (as plotted) is often a conservative values on the horizontal axis. Points that fall below the 1:1
estimate of noncomposite strength, meaning that critical line indicate that the thermal prediction of fire resistance is
temperature relationships as a function of flexural utiliza- lower (or conservative) relative to the deflection-based resis-
tion for composite beams can be used to conservatively esti- tance. The banded lines indicate a percentage of increase or
mate noncomposite critical temperature relationships. decrease in thermally predicted fire resistance versus the
deflection-based resistance.

950
1700
900
1600
850
Temperature ( oC)
Temperature ( oF)

1500

Temperature (°C)
Temperature (°F)

800
1400 750
1300 700
1200 650
1100 600

1000 550
500
900
0.2 0.3 0.4 0.5 0.6 0.7
Utilization (M/Mn)
(a) Ts,BF (b) Ts,E119

AISC Tcr,BF EC3 Tcr R-C U-C

AISC Tcr ASTM E119 T cr R-NC U-NC

Fig. 12. Critical steel temperatures for flexural failure of standard fire test specimens as a function of initial ambient flexural utilization.

18 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 Figure 13 shows that imposing the AISC Specification method outlined in Section A-4.2.4c. Typically, this analy-
critical temperature limits on the experimental measure- sis would use a 2D or 3D finite element (FE) mesh (Drury
ments of Ts,BF and Ts,E119 provides similarly conservative and Quiel, 2023b; Franssen and Gernay, 2017; Selden and
fire resistance predictions relative to the onset of flex- Varma, 2016) of the composite or noncomposite beam cross
ural runaway. Fire resistances based on critical tempera- section. Rather than resorting to such advanced analysis
tures for noncomposite specimens are more conservative, methods (which require greater effort and expertise by the
again because the calculation of initial flexural utilization analyst), the temperature increase in a protected W-shape
neglected any flexural contributions from the slab. Fire floor beam cross section can instead be predicted using
resistance times based on critical temperatures for the com- lumped mass (LM) thermal analysis techniques, which are
posite specimens show good agreement with flexural run- classified as simple analysis methods in Section A-4.2.4d.
away (with most no more than 15% conservative), while The W-shape beam section can be represented as a single
those for noncomposite specimens are 15–30% conserva- LM (AISC, 2022; Buchanan and Abu, 2017; CEN, 2005;
tive. As shown in Figure 12(b), the Eurocode 3 critical tem- Gamble, 1989), or multiple LMs can be used to represent
peratures are slightly greater than the AISC Specification the flange and web plates (CEN, 2008; Drury et al., 2020,
Equation A-4-23 curve and would, therefore, be expected to 2021; Ghojel and Wong, 2005). These methods can be
provide even closer predictions of flexural runaway under implemented in spreadsheets or other simple mathemati-
standard fire exposure when used with Ts,E119. Altogether, cal solvers and have been shown to provide conservatively
these results suggest that these critical temperature limits accurate predictions of thermal behavior in experimental
can provide reasonably conservative estimates of one-way testing of composite floor beams (Drury et al., 2020, 2021;
flexural resistance to standard fire exposure for this type of Drury and Quiel, 2023b). The simplified LM calculations
floor beam assembly. presented herein are demonstrated as capable tools for pre-
dicting critical temperature for floor beams while simul-
taneously providing a means of calculating equivalence to
SIMPLIFIED METHODS FOR CRITICAL
ASTM E119 hourly fire resistance ratings.
TEMPERATURE ANALYSIS
There are currently no provisions in the AISC Specification Single Lumped Mass Methods
for determining critical temperature per Section A-4.2.4e
The following calculations from the AISC Specification and
for floor beams. Also, there are no direct means per Sec-
Part 1-2 of Eurocode 3 represent the steel beam section as
tion A-4.3.1 for achieving equivalence to a standard fire
a single lumped mass (1LM) that has uniform temperature
resistance rating other than through an advanced analysis

180
Fire Resistance (min): Thermal Limits

170 R-C R-NC

160 × U-C + U-NC
150
140
130
120
110
100
90
80
80 90 100 110 120 130 140 150 160 170 180
Fire Resistance (min): Deflec!on Rate

Thermal Data: Ts,BF ; Limit: AISC Tcr,BF

Thermal Data: Ts,E119 ; Limit: AISC Tcr

Fig. 13. Correlation plot of standard fire resistance times based on thermal and structural test results.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 19

Ts and is uniformly coated with a contoured constant thick- The value of D for both formulations is calculated using
ness of fire protective insulation: the inner perimeter of the contoured fire protection per
AISC Specification Equation C-A-4-7 (U.S. units): Eurocode 3 (CEN, 2005). For a W-shape floor beam sec-
tion that supports a slab, D is calculated as follows:
k p,i ⎡⎢ T f,i − Ts,i−1 ⎤⎥
Ts,i = Ts,i−1 + ⎢ ⎛ W ⎞ cp,i ρp,i d p ⎥ Δt (3) D = 2d + 3b f − 2t w (5)
dp ⎢cs,i + ⎥
⎣ ⎝ D⎠ 2 (144 ) ⎦ In this study, the thermal properties of the steel and SFRM
are realistically considered as a function of their increas-
Eurocode 3, Part 1-2, Equation 4.27 (metric units):
ing temperature. To avoid the need for iteration at each
k p,i ⎡⎢ T f,i − Ts,i−1 ⎤⎥ time step, these properties can be calculated using tem-
Ts,i = Ts,i−1 + ⎢ Δt (4a)
dp ⎛ ⎞ + p,i p,i p ⎥
W c ρ d peratures obtained at the previous time step (Drury et al.,
⎢cs,i ⎥ 2020; Gamble, 1989), as long as the time step remains suffi-
⎣ ⎝ D⎠ 3 ⎦
ciently small. The following calculations are made to obtain
− ( eϕi /10 − 1) ( T f,i − Tf,i−1 ) the thermal properties of the steel and SFRM at each time
step i:
cp,i ρp,i d p
ϕi = (4b)
• cs,i for the steel section is calculated as a function of
cs,i ⎛ ⎞
W
⎝ D⎠ Ts,i−1 using the temperature-dependent relationship in
Eurocode 3, Part 1-2 (CEN, 2005).
where
As = total cross-sectional steel area, ft2 (m2) • cp,i, kp,i, and ρp,i for the fire protection are calculated as a
function of the assumed temperature of the fire protection
D = heated perimeter, in. (mm)
material, Tp,i, taken as Tp,i = (0.9Ts,i−1 + 0.1Tf,i−1 ) using the
Tf,i = temperature of the fire at time step i, °F (°C) mean value of the empirical relationship per Khorasani
Ts,i = temperature of the steel section at time step i, et al. (2015) for standard, low-density SFRM (with a
°F (°C) density of approximately 15 pcf (240 kg/m3)].
W = weight (mass) per unit length, lb/ft (kg/m) = ρsAs
For this study, all LM thermal calculations are performed
W
= section factor for the fire exposed surfaces using Δt = 30 sec in accordance with Eurocode 3, Part 1-2
D (CEN, 2005), which was sufficiently small to obtain conver-
cs,i = steel specific heat at time step i, Btu/lb-°F
gent solutions as well as good agreement with the experi-
(J/kg-K)
mentally measured Ts,E119 [see Figure 8(b)]. The use of
cp,i = fire protection specific heat at time step i,
weighted average Tp,i = ( 0.9Ts,i−1 + 0.1Tf,i−1) to calculate the
Btu/lb-°F (J/kg-K)
temperature-dependent SFRM properties was also deter-
dp = fire protection thickness, in. (m) mined based on good agreement with the experimental data.
i = time step number The stronger weighting of the steel temperature reflects the
kp,i = fire protection thermal conductivity at time step presumed shape of the thermal gradient that develops over
i, Btu/ft-sec-°F (W/m-K) the thickness of the SFRM when exposed to fire. Specifi-
Δt = time step increment (sec) cally, the high temperature at the fire-exposed outer SFRM
ρp,i = fire protection density at time step i, lb/ft3 surface would exhibit an approximately exponential decay
(kg/m3) over the SFRM thickness toward the protected inner sur-
face of the steel section. More research is needed to demon-
ρs = steel density = 490 lb/ft3 (7,850 kg/m3)
strate a broad applicability of this approach for calculating
Note that the original form of Equation 4 from Eurocode 3 temperature-dependent SFRM thermal properties in these
has been reformatted to be similar to Equation 3 (which is 1LM methods. Figure 8(b) shows that the simpler AISC for-
taken from the AISC Specification) to facilitate a direct com- mulation provides a slightly closer prediction of the experi-
parison. The two 1LM calculation approaches are similarly mentally measured Ts,E119 than the EC3 formulation when
based on the assumption that the outer surface of the pro- using the same material inputs.
tection layer is equal to the fire temperature, and the inner
surface of the protection layer is equal to the steel tempera- Multiple Lumped Mass Method
ture. Eurocode includes an additional reduction in ∆Ts,i at As shown in Figure 7, the bottom flange, top flange, and
every time step as a function of the change in applied fire web will realistically develop different temperatures due to
temperature ∆Tf,i during that step. In this way, the Eurocode differences in fire-exposed perimeter; the assumption of a
formulation slightly reduces ∆Ts,i when the fire temperature single uniform Ts for the entire cross section is therefore a
is growing at a rapid rate.

20 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

significant simplification. As an alternative, the steel sec- Figure 8(b) shows that Ts,AVG per the 3LM approach in
tion can be subdivided into multiple lumped masses for Equation 9 provides an even closer prediction of the experi-
which a thermal calculation can be made at each time step mentally measured Ts,E119 than the 1LM approaches.
(Drury et al., 2020, 2021; Ghojel and Wong, 2005). Part 1-2
of Eurocode 4 adapts the equations from Part 1-2 of Euro- Sensitivity to SFRM Thermal Properties
code 3 such that each flange and the web are considered as
It should be emphasized that the accuracy of these LM pre-
separate LMs:
dictions is dependent on the temperature-dependent rela-
Eurocode 4, Part 1-2, Equation 4.8 (metric units):
tionships used to obtain thermal properties of the steel and
k p,j,i ⎡ T f,i − Ts,j,i−1 ⎤ fire protection materials. For comparison, two additional
Ts,j,i = Ts,j,i−1 + ⎢ ⎥ Δt (6a)
d p,j ⎢ ⎛ W j ⎞ c p,j,i ρp,j,i d p,j ⎥ predictions of Ts are made using Equation 3 from the AISC
⎢cs,j,i ⎜ ⎟ + ⎥ Specification but with two alternative temperatures used to
⎣ ⎝ Dj ⎠ 3 ⎦ calculate the SFRM thermal properties at every time step
− (eϕ j,i /10 − 1) ( T f,i − T f,i−1 ) via the Khorasani et al. (2015) mean value relationships for
cp,i, kp,i, and ρp,i : Tp,i = 500°C (as permitted by the AISC
c p,j,i ρp,j,i d p,j Specification as a simplification), and Tp,i = Ts,i-1 (repre-
ϕ j,i = (6b)
⎛ Wj ⎞ senting a lower-bound simplification). Figure 14 shows
cs,j,i ⎜ ⎟ that these alternate approaches can provide predictions of
⎝ Dj ⎠
Ts,E119 that are either slightly higher or lower, respectively,
where versus the experimental data or the 1LM-AISC prediction
j = plate component designation (BF, TF, web) of Ts using Tp,i = (0.9Ts,i-1 + 0.1Tf,i-1) (reproduced from Fig-
Ts,j,i = temperature of plate component j at time step i, °C ure 8). The user should exercise caution and potentially
seek opportunities for preliminary experimental valida-
Wj
= section factor for the fire exposed surfaces of plate tion when applying these methods in practice to calculate
D j component j equivalent standard fire resistance. Also note that some of
the gray curves representing the experimental data begin to
In the Eurocode 4 approach, the section factor Wj / D j for
each flange is calculated using the following inputs:
Bottom flange: DBF = 2t f + 2bf (7a)
As,BF = bf t f (7b) 2000 1093
1800 982
Top flange: DTF = 2t f + bf (7c)
1600 871
As,TF = bf t f (7d) 1400 760

Temperature (°C)
Temperature (°F)

1200 649
Again, note that the original form of Equation 6 from
1000 538
Eurocode 4 has been reformatted to be similar to Equa-
800 427
tion 3 for comparison. Figure 8(a) shows very good agree-
ment between the experimental measurements of Ts,BF and 600 316
that predicted by the Eurocode 4 approach for the BF as 400 204
a single lumped mass (1LM). If the beam depth does not 200 93
exceed 20 in. (500 mm), Eurocode 4 notes that the tempera- 0 -18
ture of the web can be taken as equal to that of the bottom 0 20 40 60 80 100 120 140 160 180
flange for simplification: Time (min)
Ts,web,i ≈ Ts,BF,i (8) Test Data Tp,i = (0.9Ts,i-1+0.1Tf,i-1)

The average temperature of the W-shape cross section Tp,i = 500°C Tp,i = Ts,i-1
can then be calculated as an area-weighted average among
the three plate lumped masses (3LM): Fig. 14. Comparison of time histories for Ts,E119 from all
16 specimens (calculated as an average of all longitudinal
3 ⎛ As,j ⎞
Ts,AVG,i = ∑ Ts,j,i ⎜ (9) measurement locations for each beam) against 3 iterations
j=1 ⎝ As ⎟⎠ of the AISC Specification 1LM prediction for Ts (with
different values of Tp used to determine the SFRM thermal
properties at each time step) (experimental
data provided by Farid Alfawakhiri, courtesy of AISC).

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 21

 Table 5. Comparison of Standard Fire Resistance Times (in minutes): Flexural Performance of
Tested Specimens Versus Thermal LM Predictions of Load-Dependent Critical Temperature.
Limit: ASTM E119 Limit: Limit: Limit:
Initial Flexural Deflection Rate AISC Tcr,BF AISC Tcr EC3 Tcr
Utilization, R-Test U-Test 1LM 1LM 1LM 3LM
M/ Mn Deflection Deflection EC4 Ts,BF AISC Ts EC3 Ts EC4 Ts,AVG

60% 112 98 93 91 91 101

Composite

48% 116 107 105 98 98 110

36% 137 131 119 108 107 122

24% 174 171 149 123 122 146

Noncomposite

57% 129 114 96 93 92 104

45% 139 123 108 100 100 113

34% 151 144 122 110 109 125

23% 163 167 151 126 124 148

trail upward beyond 120 min, due to likely SFRM loss after the standard fire resistance of W-shape floor beam assem-
the onset of flexural runaway. blies at flexural runaway. In practice, a designer would be
able to iteratively perform these calculations to determine
Critical Temperature Predictions the thickness and material properties of SFRM needed to
achieve an equivalent targeted hourly fire resistance rating
The load-dependent critical temperature relationships from
Table 3 can be applied to the various LM predictions of Ts to
predict the standard fire resistance. Table 5 compares these
predictions against the fire resistance times correspond- 180
LM Predicted Fire Resistance (min)

ing to the onset of flexural runaway in each test (which are 170 Restrained
reproduced from Table 4). Similar to Figure 13, the relation- 160 Unrestrained
ship between these fire resistance times is visualized in Fig-
150
ure 15 as a correlation plot, with LM thermal predictions on
the vertical axis and deflection-based values from the tests 140
on the horizontal axis. Points that fall below the 1:1 line 130
indicate that the LM prediction of fire resistance is lower
120
(or conservative) relative to the deflection-based resistance.
Due to the enhanced accuracy of the Eurocode 4 LM pre- 110
dictions of steel temperature (see Figure 8), the predicted 100
fire resistance times based on those calculations (plotted
90
in red and blue in Figure 15) are similarly conservative as
those based on the experimentally measured temperatures 80
in Figure 13. Fire resistances based on the 1LM methods 80 90 100 110 120 130 140 150 160 170 180
(plotted in green and gold in Figure 15) are slightly more Test Fire Resistance (min): Deflec!on Rate
conservative than those in Figure 13 because their predic-
tions of Ts are more conservative relative to the experimen- Predic!on: EC4 Ts,BF ; Limit: AISC Tcr,BF
tally measured Ts (see Figure 8). Predic!on: AISC Ts ; Limit: AISC Tcr
Predic!on: EC3 Ts ; Limit: EC3 Tcr
CONCLUSIONS Predic!on: EC4 Ts,AVG ; Limit: EC3 Tcr
The results of this study demonstrate that simplified ther-
Fig. 15. Correlation plot of standard fire resistance
mal analysis methods can be combined with load-dependent
times from structural test results versus thermal LM
critical temperature relationships to conservatively predict predictions of load-dependent critical temperature.

22 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

for a given floor beam section. The design of floor beams steel and fire protection materials at each time step.
is often governed by deflection-based serviceability crite- Simplifying approximations were shown to be capable of
ria rather than strength criteria; the load-dependent critical conservatively predicting the steel temperature increase
temperature relationships per the AISC Specification and under standard fire exposure.
Part 1-2 of Eurocodes 3 and 4 therefore offer the ability to
tailor an equivalent standard fire resistance to the actual
ACKNOWLEDGMENTS
flexural utilization of the member.
The following conclusions can be drawn from this study: Funding for this project was provided by the Ameri-
can Institute of Steel Construction (AISC) via the Milek
• Based on the results of standard fire testing by AISC/
Faculty Fellowship, of which Professor Quiel is the 2016
AISI in 2015 (as well as other quasi-standard fire test
recipient. Additional funding to support Dr. Drury during
results in the published literature), the critical value of
his involvement in this research for his doctoral disserta-
bottom flange temperature, Ts,BF, at the onset of flexural
tion was provided by Lehigh University via a 2021 Faculty
runaway can be conservatively predicted using the
Innovation Grant. All experimental data and photos pre-
load-dependent values of Tcr,BF in AISC Specification
sented in this paper are reproduced with expressed written
Table A-4.2.4 (2022). Likewise, the critical value of
consent of AISC and AISI for the purposes of this study.
Ts,E119 (i.e., the average steel section temperature based
Special thanks to Dr. Devin Huber (AISC) and Dr. Farid
on the thermocouple placement during an ASTM E119
Alfawakhiri (AISI, retired) for their support of this project
standard fire test) at the onset of flexural runaway can
and for furnishing all data and documentation pertaining to
be conservatively predicted using the load-dependent
the tests. All opinions, findings, and conclusions expressed
relationships of Tcr per Equation 4.22 from Eurocode
in this paper are the authors’ and do not necessarily rep-
3, Part 1-2 (CEN, 2005) as well as AISC Specification
resent the policies and views of AISC, AISI, or Lehigh
Equation A-4-23.
University.
• Based on the standard fire test data presented in this
paper, these load-dependent critical temperature REFERENCES
relationships for standard fire resistance are robust for
AISC (2010), Specification for Structural Steel Buildings,
SFRM-protected W8×28 floor beam assemblies that are
ANSI/AISC 360-10, American Institute of Steel Con-
restrained or unrestrained to thermal expansion, with
struction, Chicago, Ill.
composite or noncomposite reinforced concrete slabs
(cast on corrugated metal decking). The quasi-standard AISC (2022), Specification for Structural Steel Buildings,
test results in the published literature indicate that these ANSI/AISC 360-22, American Institute of Steel
critical temperature relationships may also be robust for Construction, Chicago, Ill.
section depths ranging from W8 to W18, the inclusion of Alfawakhiri, F., Carter, C.J., Berhing, R.M., Zeeveld, P.,
shear tab connections (rather than unrealistic end-plate Hervey, F.E., and Woods, L.C. (2016), “The Effects of
bearing connections, which are used in standard fire Load Intensity and Restraint on the Fire Resistance of
tests), the implementation of flat slabs (rather than those Steel and Composite Beams,” 9th International Confer-
cast on corrugated metal decking), the material property ence on Structures in Fire SiF’16, Princeton University,
details of the slab’s concrete, unprotected steel beam N.J.
sections (i.e., with no applied passive fire protection), ASCE (2007), Standard Calculation Methods for Structural
variation in steel beam yield strength (with nominal Fire Protection (ASCE/SEI/SFPE 29-05), American
ranging from 36–50 ksi), and span lengths ranging from Society of Civil Engineers, Reston, Va.
11 to 40 ft.
ASCE (2020), Performance-Based Structural Fire Design:
• The 3LM predictions of steel temperature for each Exemplar Designs of Four Regionally Diverse Buildings
flange and the web under standard fire exposure per Using ASCE 7-16, Appendix E, American Society of Civil
Eurocode 4 slightly outperformed 1LM predictions of Engineers, Reston, Va., doi: 10.1061/9780784482698.
Ts for the entire section per Eurocode 3 and the AISC ASTM (2019), Test Methods for Fire Tests of Building
Specification. However, all LM predictions provided Construction and Materials, ASTM E119-19, ASTM
reasonably accurate and conservative estimations of International, West Conshohocken, Pa., doi: 10.1520/
standard fire resistance compared to those corresponding E0119–19.
to flexural runaway of the AISC specimens tested
ASTM (2020), Standard Specification for Structural Steel
to ASTM E119. The quality of these LM predictions
Shapes, ASTM A992–20, ASTM International, West
depends on the temperature-dependent relationships
Conshohocken, Pa., doi: 10.1520/A0992_A0992M-20.
that are used to calculate thermal properties of the

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 23

ASTM (2022), Standard Specification for Steel Wire and Drury, M.M. (2022), Critical Temperature Relationships
Welded Wire Reinforcement, Plain and Deformed, for for the Performance-Based Design of Realistically-
Concrete, ASTM A1064-22, ASTM International, West Restrained Steel Composite Floor Assemblies, Ph.D.
Conshohocken, Pa., doi: 10.1520/A1064_A1064M-22. Thesis, Lehigh University, Bethlehem, Pa.
Berhinig, R.M. and Alfawakhiri, F. (2014), “New Fire Drury, M.M., Kordosky, A.N., and Quiel, S.E. (2020),
Resistance Test for Loaded Unrestrained Beams,” “Structural Fire Resistance of Partially Restrained,
ASCE/SEI Structures Congress, American Society of Partially Composite Floor Beams, II: Modeling,” Journal
Civil Engineers, Boston, Mass., pp. 1175–1184, doi: of Constructional Steel Research, Vol. 167, p. 105946,
10.1061/9780784413357.106. doi: 10.1016/[Link].2020.105946.
Bletzacker, R.W. (1967), “Fire Resistance of Protected Drury, M.M., Kordosky, A.N., and Quiel, S.E. (2021),
Steel Beam Floor and Roof Assemblies as Affected by “Robustness of a Partially Restrained, Partially
Structural Restraint,” Symposium on Fire Test Methods— Composite Steel Floor Beam to Natural Fire Exposure:
Restraint & Smoke 1966 (ASTM STP 422), ASTM Novel Validation and Parametric Analysis,” Journal of
International, West Conshohocken, Pa., pp. 63–90., doi: Building Engineering, Vol. 44, p. 102533, doi: 10.1016/j
10.1520/STP41306S. .jobe.2021.102533.
Bono, J.A. (1970), “New Criteria for Fire Endurance Drury, M.M. and Quiel, S.E. (2023a), “Standard versus
Tests,” Fire Test Performance (ASTM STP 464), ASTM Natural Fire Resistance for Partially Restrained
International, West Conshohocken, Pa., pp. 106–126, doi: Composite Floor Beams: 1—Testing,” Journal of
10.1520/STP44715S. Constructional Steel Research, Vol. 202, p. 107768, doi:
BSI (1987), Fire Tests on Building Materials and Structures— 10.1016/[Link].2022.107768.
Part 21: Methods for Determination of the Fire Resistance Drury, M.M. and Quiel, S.E. (2023b), “Standard versus
of Load-Bearing Elements of Construction, BS 476- Natural Fire Resistance for Partially Restrained
21:1987, British Standards Institute (BSI), London, UK. Composite Floor Beams: 2—Analysis,” Journal of
BSI (2008), Fire Tests on Building Materials and Structures. Constructional Steel Research, Vol. 202, p. 107767, doi:
Part 10. Guide to the Principles, Selection, Role, and 10.1016/[Link].2022.107767.
Application of Fire Testing and Their Outputs., BS 476- Franssen, J.-M. and Gernay, T. (2017), “Modeling
10:2009, British Standards Institute (BSI), London, UK. Structures in Fire with SAFIR®: Theoretical Background
Buchanan, A.H. and Abu, A. (2017), Structural Design for and Capabilities,” Journal of Structural Fire
Fire Safety, 2nd Ed., John Wiley & Sons Inc., Chichester, Engineering, Vol. 8, No. 3, pp. 300–323, doi: 10.1108/
West Sussex, UK. JSFE-07–2016–0010.
CEN (2005), EN 1993-1-2:2005 Eurocode 3: Design of Steel Gamble, W.L. (1989), “Predicting Protected Steel Member
Structures—Part 1-2: General Rules—Structural Fire Fire Endurance Using Spread-Sheet Programs,” Fire
Design, EN 1993–1-2:2005, European Committee for Technology, Vol. 25, No. 3, pp. 256–273, doi: 10.1007/
Standardization, Brussels, Belgium. BF01039782.
CEN (2008), EN 1994–1-2:2005 Eurocode 4: Design of Ghojel, J.I. and Wong, M.B. (2005), “Three-Sided Heating
Composite Steel and Concrete Structures—Part 1-2: of I-Beams in Composite Construction Exposed to Fire,”
General Rules—Structural Fire Design, European Journal of Constructional Steel Research, Vol. 61, No. 6,
Committee for Standardization, Brussels, Belgium. pp. 834–844, doi: 10.1016/[Link].2004.11.006.
CEN (2009), EN 1991–1-2:2002 Eurocode 1: Actions on ICC (2020), IBC: International Building Code, 2021,
Structures—Part 1-2: General Actions—Actions on International Code Council, Country Club Hills, Ill.
Structures Exposed to Fire, European Committee for ISO (2019), Fire Resistance Tests—Elements of Building
Standardization, Brussels, Belgium. Construction—Part 11: Specific Requirements for
Choe, L., Ramesh, S., Grosshandler, W., Hoehler, M., the Assessment of Fire Protection to Structural Steel
Seif, M., Gross, J., and Bundy, M. (2020), “Behavior Elements, ISO 834–11:2014, International Organization
and Limit States of Long-Span Composite Floor Beams for Standardization, Geneva, Switzerland.
with Simple Shear Connections Subject to Compartment Jiang, S.-C., Ranzi, G., Chen, L.-Z., and Li, G.-Q. (2017),
Fires: Experimental Evaluation,” Journal of Structural “Behaviour and Design of Composite Beams with
Engineering, Vol. 146, No. 6, p. 04020088, doi: 10.1061/ Composite Slabs at Elevated Temperatures,” Advances in
(ASCE)ST.1943–541X.0002627. Structural Engineering, Vol. 20, No. 10, pp. 1451–1465,
doi: 10.1177/1369433216682507.

24 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

Khorasani, N.E., Gardoni, P., and Garlock, M. (2015), Selden, K.L. and Varma, A.H. (2016), “Composite Beams
“Probabilistic Fire Analysis: Material Models and under Fire Loading: Numerical Modeling of Behavior,”
Evaluation of Steel Structural Members,” Journal of Journal of Structural Fire Engineering, Vol. 7, No. 2,
Structural Engineering, Vol. 141, No. 12, p. 04015050, pp. 142–157, doi: 10.1108/JSFE-06–2016–011.
doi: 10.1061/(ASCE)ST.1943–541X.0001285. UL (2019a), Fire Resistance Directory, Underwriters
Kordosky, A.N., Drury, M.M., and Quiel, S.E. (2020), Laboratories Inc., Northbrook, Ill.
“Structural Fire Resistance of Partially Restrained, UL (2019b), Fire Resistance Directory: Design No. D982,
Partially Composite Floor Beams, I: Experiments,” Underwriters Laboratories Inc., Northbrook, Ill.
Journal of Constructional Steel Research, Vol. 167,
UL (2020), Standard for Fire Tests of Building Construction
p. 105945, doi: 10.1016/[Link].2020.105945.
and Materials, UL 263, Underwriters Laboratories, Inc.,
LaMalva, K., Bisby, L., Gales, J., Gernay, T., Hantouche, E., Northbrook, Ill.
Jones, C., Morovat, A., Solomon, R., and Torero, J. (2020),
Vinnakota, M.R., Foley, C.M., and Vinnakota, S. (1988),
“Rectification of ‘Restrained vs Unrestrained’,” Fire and
“Design of Partially or Fully Composite Beams with
Materials, Vol. 44, No. 3, pp. 341–351, doi: 10.1002/
Ribbed Metal Deck Using LRFD Specifications,”
fam.2771.
Engineering Journal, AISC, Vol. 25, No. 2, pp. 60–78.
LaMalva, K.J. (Ed.). (2018), Manual of Practice 138:
Wang, W., Engelhardt, M.D., Li, G., and Huang, G. (2017),
Structural Fire Engineering, American Society of Civil
“Behavior of Steel–Concrete Partially Composite Beams
Engineers, Reston, Va.
Subjected to Fire—Part 1: Experimental Study,” Fire
Newman, G. (1999), “The Cardington Fire Tests,” Technology, Vol. 53, No. 3, pp. 1039–1058, doi: 10.1007/
Proceedings of the North American Steel Construction s10694–016–0618-y.
Conference, AISC, Toronto, Canada, pp. 28.1–28.22.
Zhao, B. and Kruppa, J. (1997), Fire Resistance of Composite
Ruddy, J., Marlo, J., Ioannides, S., and Alfawakhiri, F. Slabs with Profiled Steel Sheet and of Composite Steel
(2003), Fire Resistance of Structural Steel Framing, Concrete Beams, Part 2: Composite Beams Final
Design Guide 19, AISC, Chicago, Ill. Report, Office for Official Publications of the European
Communities, Luxembourg.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 25

26 / ENGINEERING JOURNAL / FIRST QUARTER / 2025
Generalized Elastic Lateral-Torsional
Buckling of Steel Beams
ROBERT S. GLAUZ and BENJAMIN W. SCHAFER

ABSTRACT
A concise review is provided of the classical elastic lateral-torsional buckling moment for steel beams as utilized in the AISC Specification
(2022). Rather than make assumptions regarding the cross-section properties, the derivation is provided in its general form for an arbitrary
steel beam—that is, one that may be asymmetric and may include any manner of varying geometry, thickness, or cross-section shape. The
cross-section properties that underpin the calculation are fully detailed. The assumptions that are inherent in the classical derivations (no
shear, no cross-section distortion, etc.) are also fully detailed. The manner in which the generalized lateral-torsional buckling formula may
be simplified for particular sections (e.g., a singly symmetric channel) with no loss of accuracy is explained. Adaptations and approximations
utilized in the 2022 edition of the AISC Specification for elastic lateral-torsional buckling moment of specific sections (e.g., mono-symmetric
I-section, angles, etc.) are assessed against the actual elastic solution, and the accuracy and clarity of the assumptions utilized are evalu-
ated. The generalized formula, consistent with current assumptions but applicable to all structural steel cross sections, is recommended for
future reference in the main body of the AISC Specification.

Keywords: lateral-torsional buckling, elastic buckling, steel beams.

INTRODUCTION The AISC Specification for Structural Steel Buildings,

hereafter referred to as the AISC Specification, has evolved
Steel beams with large unbraced spans are susceptible to
significantly over its history. The first edition (AISC, 1923)
lateral-torsional buckling. This instability is manifested by
protected against lateral-torsional buckling in a simple, but
simultaneous lateral translation and twisting of the mem-
effective and safe manner, treating the compression flange
ber, as shown in Figure 1. Upon lateral-torsional buckling,
as a column expressing strength as a function of slenderness
the beam is unable to carry additional load due to the for-
(L/ b), where L is the beam length and b is the compression
mation of plastic mechanisms in the cross section triggered
flange width. Column stability was reasonably well under-
by the buckling that ultimately result in localized loss of
stood based on the early work of Euler (1744), although the
stiffness and the potential for collapse. Slender members
mathematics of beam elastic lateral-torsional buckling was
with narrow cross sections and long unbraced lengths are
not formalized until later by Timoshenko (1936). As shown
more susceptible to lateral-torsional buckling.
in Figure 2, the design bending stress Fb was reduced as L/ b
An early example of lateral-torsional buckling failure,
increased from 15 to an upper limit of 40. This upper limit
before this mechanism was better understood, was the col-
ensured sufficient stability to prevent elastic buckling fail-
lapse of the Dee Bridge in Chester, England. Built in 1846
ure. The 1936 edition of the AISC Specification continued
for rail transportation, the main girders were 45 in. deep
using the same approach, allowing higher stress consistent
cast iron with 24-in.-wide bottom flanges but only 7-in.-
with the higher yield strength of A9 structural steel at that
wide top compression flanges. Lateral-torsional instability
time.
(among other factors) led to the collapse during the cross-
By the 1946 edition (AISC, 1946), the mechanics of elas-
ing of a multi-car passenger train in 1847 (Commissioners
tic lateral-torsional buckling were better understood. The
of Railways, 1848). Bridge engineers learned from this and
two components of torsional stiffness for a doubly sym-
other failures that the compression flange must be stabi-
metric I-section can be approximated in terms of (L/ b)2 for
lized or increased in size to prevent buckling.
the warping resistance and Ld/ bt for the pure torsion resis-
tance (Salmon and Johnson, 1980). For the vast majority
of I-sections used in construction at the time, the flanges
were thick enough for the pure torsion resistance to domi-
Robert S. Glauz, Owner, RSG Software, Inc., Lee’s Summit, Mo. Email: glauz@
nate. Therefore, the lateral-torsional buckling design stress
changed to a form using slenderness Ld/ bt, where d is the
[Link] (corresponding)
Benjamin W. Schafer, Professor, Johns Hopkins University, Baltimore, Md.
Email: schafer@[Link] depth of the section and t is the thickness of the flange. As
shown in Figure 2, this significantly increased the design
stress for many I-sections and the upper limit on slender-
Paper No. 2024–08 ness was removed.

ISSN 2997-4720 ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 27

 The 1961 edition (AISC, 1961) continued with the same into account its torsional rigidity about its longitudinal axis
design stress expression using Ld/ bt. Due to the introduction as well as the bending stiffness of its compression flange,
of A36 steel and the availability of different steel grades, are too complex for general office use.” Tables of torsional
this was the first edition to incorporate the material yield properties did not appear in the AISC Steel Construction
strength Fy into the provisions. This edition also recognized Manual, hereafter referred to as the AISC Manual, until the
that deeper I-sections with thinner flanges could achieve eighth edition (AISC, 1981).
higher strength due to the warping resistance. Therefore, It wasn’t until the 1986 LRFD edition (AISC, 1986) that
the design stress was permitted to be the larger of the 1946 both components of torsional stiffness were combined to
formula and a new inelastic buckling expression utilizing correctly calculate the critical buckling moment for an
(L/ r)2. Instead of the compression flange width b, this for- I-section, rather than the larger of the two components. As a
mula used r, defined as the radius of gyration of a tee sec- theoretically accurate calculation, its use was later extended
tion comprising the compression flange plus 6 of the web to singly symmetric channel sections and may actually be
area, which is approximately 0.27b for a rectangular flange. used for any section bending about an axis of symmetry.
This generalization accommodated the use of flange geom- This edition also provided approximations to the theoretical
etries other than rectangular. The 1961 edition also included solution for other sections that are not symmetrical about
a Commentary that acknowledged the existence of more the axis of bending. These and other approximations unique
accurate calculation methods stating: “Rational expressions to particular cross sections are still in use in the current edi-
for the elastic buckling strength of the beam, which take tion of the AISC Specification (2022).

Fig. 1. Lateral-torsional buckling—translation and twist.

Fb
(ksi)
(Fy =30 ksi)
(Fy =33ksi)
(Fy =33 ksi)
(Fy =36 ksi) W14×68
(Fy =36 ksi)

L/b

Fig. 2. AISC history of design stress for lateral-torsional buckling.

28 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 Figure 2 also illustrates that the 1986 strength permitted end translation and twist are restrained (u = v = ϕ = 0), and
moments up to full plasticity, where the allowable design end moments and bimoments are zero (u″ = v″ = ϕ″ = 0).
stress is shown as Mp /1.67Sx. A new inelastic buckling The result is three simultaneous equations solved by equat-
strength was implemented using a linear transition between ing the determinant of the coefficients to zero:
the fully plastic moment and the elastic buckling moment. ⎡ P −P ⎤ (4)
0 − Pyo + M x
This paper focuses on determination of the elastic ⎢ ey ⎥
lateral-torsional buckling moment and does not attempt to ⎢ 0 Pex − P Px o − M y ⎥
⎢ ⎥
address inelastic buckling strength. The purpose is to show ⎢ − Py + M Px o − M y ( Pt − P )ro2 − β x M x − β y M y ⎥
o x
that the elastic buckling moment calculation can be applied ⎣ ⎦
in a consistent way for all cross sections without the approx- ⎧u ⎫ ⎧ 0 ⎫
imations that can result in unacceptable error, thus provid- ⎪ ⎪ ⎪ ⎪
ing a unified approach for all cross sections. ⎨ v ⎬=⎨ 0 ⎬
⎪ϕ ⎪ ⎪ 0 ⎪
⎩ ⎭ ⎩ ⎭
ELASTIC THEORY where Mx (Pey) and My (Pex) are the end moments produced
The classic approach to the elastic buckling solution of a by the axial eccentricities, and Pex, Pey, and Pt are the axial
linear prismatic member is to consider only end forces. loads at which elastic buckling occurs about the x-axis,
The application of axial load P at eccentricities ex and about the y-axis, and in torsion, respectively:
ey as shown in Figure 3 provides a general solution for a π 2 EI x
beam-column undergoing any combination of axial load Pex = (5)
L2
and uniform moments.
Assuming small displacements such that the longitudi- π 2 EI y
nal stresses remain constant throughout the member length Pey = (6)
L2
(first-order analysis), the three equations of equilibrium as
expressed by Timoshenko and Gere (1961) are given as: 1 ⎛ π 2 ECw ⎞
Pt = 2 ⎜GJ + ⎟ (7)
EI y uiv + Pu″ + P ( yo − e y ) ϕ″ = 0 (1) ro ⎝ L2 ⎠

EI x v iv + Pv ″ + P ( x o − ex ) ϕ″ = 0 (2) The case of interest for lateral-torsional buckling is bend-

ing about the major principal axis. Using x as the major
ECwϕiv − ( GJ − β x Pe y − β y Pex − Pro2 ) ϕ″ (3) principal axis, the eccentricity ey is increased as the axial
load P approaches zero. In the limit, P = 0, My = 0, and Mx
+ P ( yo − e y ) u″ − P ( x o − ex ) v ″ = 0
is the critical moment for buckling about the x axis given
by the equation:
where x and y are the principal axes of the cross section, u
and v are translational displacements in the x and y direc- M x2 + β x Pey Mx − ro2 Pey Pt = 0 (8)
tions, ϕ is twisting displacement, E is the modulus of elas-
ticity, G is the shear modulus, and the other variables are The general solution to this quadratic has two roots rep-
geometric properties of the cross section (Ix, Iy, Cw, J, resenting the critical moments for positive and negative
xo, yo, ro, βx, βy). To solve these differential equations for bending (±ey). For beams bending about a non-principal
warping-free pinned end conditions, the displacements (u, x-axis, the elastic lateral-torsional buckling moment is
v, ϕ) are assigned sinusoidal forms of one half-wavelength, given by the same formula as shown in Glauz (2017), with

Fig. 3. Beam-column elastic buckling problem.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 29

a more generalized expression for the flexural buckling The buckling solution (Equation 8) is also based on a uni-
component. form moment induced by equal and opposite end moments.
The elastic buckling moment for unequal end moments has
M x2 + β x Pey Mx − ro2 Pey Pt = 0 (9)
historically been handled in the AISC Specification by
π 2 EI y ⎛ 2
I xy ⎞ π 2 E ⎛ I x I y − I xy
2
⎞ π 2 EI x I y approximation using a multiplicative bending coefficient
Pey = 2 ⎜
1− ⎟= 2 ⎜ ⎟= (10) Cb, based on the ratio of the end moments as given by Equa-
L ⎝ IxIy ⎠ L ⎝ Ix ⎠ L2 I x
tion 11, where M2 is the larger end moment and the ratio
where Ix and Iy are the moments of inertia about any orthog- M1/M2 = −1 for uniform moment.
onal x- and y-axes (often the geometric axes of the section), M1 ⎛M ⎞
2
and Ixy is the product of inertia. As the x-axis approaches Cb = 1.75 + 1.05 + 0.30 ⎜ 1 ⎟ ≤ 2.3 (11)
the principal x-axis, the product of inertia Ixy approaches
M2 ⎝ M2 ⎠
zero and Pey becomes Pey. This Cb coefficient increases the buckling moment by up
to 2.3 times the uniform moment case, where the resulting
OTHER CONSIDERATIONS buckling moment is the magnitude at the M2 end. Further,
Equation 11 for Cb assumes a linear moment diagram; how-
The preceding development assumes small displacements ever, the primary application for beams involves transverse
and the results agree with numerical elastic buckling anal- loading and thus a nonlinear moment diagram (e.g., para-
yses based on first-order internal stresses. In reality, dis- bolic for uniform load). Many codes use the refined bend-
placements prior to buckling (second-order effects) alter the ing coefficient shown in Equation 12 to more accurately
buckling response. For bending about the x-axis, deflection approximate the buckling moment for unbraced spans with
in the y-direction prior to buckling provides a stabilizing transverse loads. It is a slight variation of the expression
effect and enables a higher buckling moment. The closer the developed by Kirby and Nethercot (1979), which uses the
ratio Iy/Ix approaches 1, the greater the increase in buckling absolute value of applied moments at the quarter points
moment. For Iy > Ix, the x-axis is no longer the major axis, (M A, MB, MC) and the maximum moment (Mmax) within the
and lateral-torsional buckling is unlikely to occur, although unbraced span. This multiplier ranges from 1.0 to 5.0, and
still possible for unsymmetrical sections with large βx mag- the resulting buckling moment is the magnitude at the loca-
nitude, low torsional stiffness, and shear center in tension. tion of Mmax.
In addition, the preceding solution assumes no distortion
occurs within the cross section as the member undergoes 12.5M max
Cb = (12)
lateral-torsional buckling. If the web is slender enough, 2.5Mmax + 3M A + 4M B + 3MC
destabilization of the flange during lateral-torsional buck-
ling may result in section distortion—often referred to as Other bending coefficient expressions have been devel-
elastic lateral-distortional buckling (Bradford, 1992). This oped to approximate the buckling moment for some spe-
mode of buckling may be especially likely when bracing cific cases, such as Wong and Driver (2010) for doubly
is only present for the compression flange and not the full symmetric I-shape beams, Helwig et al. (1997) for singly
member depth. Further, elastic local plate buckling in the symmetric I-shape beams, and Yura (1995) for interior span
section (another form of potential distortion in the section) of I-shape beams with top flange lateral restraint.
and its potential interaction with lateral-torsional buckling The application of transverse loads also produces shear
are not covered here. stresses in the member. The elastic buckling solution lead-
The provided buckling solution assumes warping-free ing to Equation 8 does not consider shear stresses, which for
pinned end conditions—that is, translation and twisting are slender beams are minor compared to longitudinal stresses;
restrained, and rotation and warping are free at both ends. however, Liang et al. (2022) have shown cases where shear
Other end conditions where flexural and twisting wave- stresses are important to consider and provided modifica-
lengths align will yield the same solution, except with L tions to the classic formula. The location of transverse load
replaced by the half-wavelength KL, where K is an effec- application can also influence the lateral-torsional buck-
tive length factor. As an example, fixing the rotation and ling behavior. For a vertical load applied with a horizon-
warping at both ends corresponds to K = 0.5. For bound- tal offset from the shear center, torsional forces are applied
ary conditions where flexural and twisting wavelengths to the member. The resulting pre-buckling torsional dis-
do not align, the solution is more complex. However, the placements can adversely affect the torsional and flexural
critical moment can be approximated using different KL components of the lateral-torsional buckling response.
values for the flexural component, Pey, and the torsional For a vertical load applied with a vertical offset from the
component, Pt. shear center the load application point can either increase

30 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

or decrease the buckling moment. For a downward load cross section. These integrations correspond to the familiar
applied to a beam above the shear center, small rotation terms area, A, moments of inertia about the x- and y-axes, Ix
of the member prior to buckling will be amplified by the and Iy, the product of inertia, Ixy, the polar moment of iner-
additional torque induced by the load location and decrease tia about the centroid, Ic, and the radius of gyration about
the buckling moment. On the other hand, a downward load the centroid, rc. The angle to the principal axes, α, and the
applied below the shear center will counteract any small principal axis moments of inertia, Ix and Iy, are also given.
rotation of the member prior to buckling and increase the A = ∫A dA (13)
buckling moment. Some design codes such as Eurocode 3
(CEN, 2005) provide additional coefficients to account for I x = ∫ A y 2 dA (14)
the load height effect, and the AISC Specification provides
additional references in its Commentary. I y = ∫ A x 2 dA (15)
All of these considerations are excluded from the formu-
las presented in this article, except that the commonly used Ixy = ∫ A xy dA (16)
bending moment gradient coefficient Cb is included for con-
sistency and clarity. I c = ∫ A (x 2 + y 2 ) dA = I x + I y (17)

rc = I c A = I x A + I y A = rx2 + ry2 (18)

SECTION PROPERTIES
The elastic buckling solution is applicable to a member of 1 −2I xy
α = arctan (19)
any cross section. The section properties required to deter- 2 Ix − Iy
mine the buckling moment about the x-axis are the moments 1 1 2
of inertia, Ix and Iy, the product of inertia, Ixy, the torsional I x ,I y = ( I x + I y ) ± ( I x − I y ) + 4I xy2 (20)
2 2
warping constant, Cw, the St. Venant torsional constant, J,
the polar radius of gyration about the shear center, ro, and Analysis of the stresses resulting from torsion first
a unique asymmetry property, βx. Figure 4 shows a gen- require determination of the torsion axis, or shear center
eral unsymmetrical cross section with x- and y-axes pass- of the cross section. The shear center is the point (xo, yo) in
ing through the centroid, c, along with the location of the the cross section where an applied shear force in any trans-
shear center, o, and asymmetry point, a. The centroid is the verse direction produces no torsion. The polar moment of
location where axial loads produce no moments, the shear inertia, Io, and radius of gyration, ro, about the shear center
center is the location where transverse loads produce no tor- are similar to those about the centroid, but greater due to
sion, and the asymmetry point is the offset from the shear the offset of the shear center. The warping constant, Cw, is a
center that produces torsional geometric stiffness due to measure of torsional stiffness due to warping, analogous to
flexure. The vector from the shear center to the asymmetry the moment of inertia, I, and bending stiffness due to flex-
point is half of β, and the component of that vector perpen- ure. The St. Venant torsional constant, J, is a measure of
dicular to the x-axis is βx / 2. torsional stiffness due to pure torsion.
Analysis of the longitudinal stresses resulting from com- The distribution of torsional stresses can be difficult to
pression and flexure require integration over the area of the determine for a general cross section. Numerical methods

Fig. 4. Properties of general unsymmetrical cross section.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 31

utilizing finite elements of the cross section are useful for Ux I y − Uy Ixy
irregular shapes. Computer software is available for this ya = (32)
2 ( I x I y − I xy
2
)
task such as MSASect2 (2023), ShapeBuilder (2023), and
Sectionproperties open-source code (2024). For thin-walled Uy I x − Ux Ixy
xa = (33)
open sections, these properties are more readily calculable 2 ( Ix I y − Ixy
2
)
using sectorial coordinates (ωc, ωo, ωn) and integrating over
distance s along the midlines of the cross-section elements Ux = ∫ A y3 dA + ∫A x 2 y dA (34)
of thickness t. These integrations are given as follows:
Uy = ∫ A x 3 dA + ∫ A y 2 x dA (35)
I y ∫0l ω c yt ds − Ixy ∫0l ω c xt ds
xo = 2 (21)
Ix I y − I xy For bending about a geometric axis, the properties
required to calculate Mcr may be calculated using the pre-
I x ∫0l ω c xt ds − Ixy ∫0l ω c yt ds ceding formulas without transforming coordinates to the
yo = 2
(22)
Ix I y − I xy principal axes. If the geometric axes are the principal axes,
Ixy = 0 and the formulas for xo, yo, xa, and ya are simplified.
Io = ∫ A ⎡⎣( x − x o)2 + ( y − yo) ⎤⎦ dA = I c + Ax o2 + Ayo2
2
(23) For bending about a principal axis that is not the geometric
axis, the necessary section properties may either be calcu-
ro = Io A = rc2 + x o2 + yo2 (24) lated as given above and the coordinates xo, yo, xa, and ya
transformed to the rotated principal axes, or all the prop-
Cw = ∫0l ω 2nt ds = ∫0l ω 2ot ds − Aω 2o (25) erties could be calculated using rotated principal axis x
and y coordinates. Note, for a doubly symmetric section,
ω c = ∫0s ρc ds (26) the asymmetry point and the shear center coincide with the
centroid, thus all the properties xo, yo, xa, ya, βx, and βy are
ω o = ∫0s ρo ds (27) equal to 0, and the formulas greatly simplify.
1 l
ω n = ωo − ω o = ω o − ∫ ωo t ds (28)
APPLICATION
A 0
1 The solution to the quadratic given by Equation 8 is shown
J = ∫0l t 3 ds (29) below, where Mcr is the elastic critical moment about the x
3
(major principal)-axis.
The variables ρc and ρo are the perpendicular distances to ⎡ 2 2 ⎤
β ⎛ β x ⎞ ro Pt
the element midline, measured from the centroid and shear Mcr = Cb Pey ⎢− x ± + ⎥ (36)
center, respectively. The details of the sectorial coordinate ⎢ 2 ⎝ 2⎠ Pey ⎥
⎣ ⎦
calculations are given in several texts such as Timoshenko
and Gere (1961), Yu (2000), and Liu et al. (2018). An expanded form for Mcr is also shown in Equation 37,
Torsional equilibrium given by Equation 3 shows where Pt and Pey have been replaced by their respective
that the coefficient on ϕ″ is the pure torsion elastic stiff- expressions.
ness GJ counteracted by the torsion geometric stiffness
π 2 EI y ⎡ β β
2
GJL2 Cw ⎤
(Mx βx + M y β y + Pro2), which is produced by the distribution Mcr = Cb ⎢− x ± ⎛ x ⎞ + 2 + ⎥ (37)
of normal stresses over the cross section resulting from the L2 ⎢⎣ 2 ⎝ 2⎠ π EI y I y ⎥⎦
applied forces Mx, My, and P. Based on the work by Glauz
(2017), it can be shown that Mx βx + My β y = Mx βx + My β y; Equation 36 has the advantage of compactness and com-
therefore, the β properties associated with the geometric monality with variables used for column buckling; Equa-
x- and y-axes may be used with the moments about these tion 37 explicitly reveals the influence of unbraced length L
axes. The properties βx and βy are distances given by Equa- and eliminates the ro property.
tions 30 and 31, equal to twice the orthogonal distances Two Mcr solutions are given by the positive and nega-
from the shear center (xo, yo) to the asymmetry point (xa, tive roots, where the positive root corresponds to a positive
ya). The coordinates of the asymmetry point are calculated moment about the x axis—that is, compression on the posi-
using integrations over the cross-sectional area without the tive y side of the x-axis (ey > 0, top flange in compression).
need for thin-walled assumptions. Design codes typically provide the magnitude of the buck-
β x = 2 ( ya − yo ) (30) ling moment as a positive number; therefore, the negative
root should be negated by multiplying through by −1. This
β y = 2( xa − xo ) (31) makes the radical additive and changes the sign on the βx/ 2

32 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

term to positive for negative bending. The following form ⎡ 2 ⎤
Cb π 2 EI y ⎢ β y CbGJ ⎛ β x CbGJ ⎞ C
handles this with the Cs sign coefficient having a value of 2
L = Cs + + ⎜Cs + ⎟ + w⎥
+1 for negative moment (ey < 0, bottom flange in compres- M cr ⎢ 2 2M cr ⎝ 2 2M cr ⎠ Iy ⎥
⎣ ⎦
sion) and −1 for positive moment (ey > 0, top flange in com-
(41)
pression); resulting in:
⎡ β 2 2 ⎤ The AISC Specification presumes that transitions in
⎛ β x ⎞ ro Pt strength may occur when Mcr reaches a given moment—for
Mcr = Cb Pey ⎢Cs x + + ⎥ (38)
⎢⎣ 2 ⎝ 2⎠ Pey ⎥⎦ example, Mr. Therefore, the length Lr where Mcr = Mr may
be found by simple substitution into Equation 41:
or
⎡ 2 ⎤
Cb π 2 EI y ⎢ β y CbGJ ⎛ β x CbGJ ⎞ Cw ⎥
π EI y ⎡⎢ β x
2 2
GJL2 Cw
⎤ L2r = Cs + + C
⎜ s + ⎟ +
⎛ βx ⎞ ⎥ Mr ⎢ 2 2M r ⎝ 2 2M r ⎠ Iy ⎥
Mcr = Cb C s + + + (39) ⎣ ⎦
L2 ⎢⎣ 2 ⎝ 2⎠ π 2 EI y I y ⎥⎦
(42)
For bending about a non-principal x-axis, the elas- Calculating Mcr requires the section properties Iy, J, Cw,
tic lateral-torsional buckling moment is determined from and βx. The Iy property is readily available for standard
Equation 9 with Pt and Pey replaced by their expressions shapes and easily calculated for others, whereas the other
from Equations 7 and 10, respectively: properties are used less frequently and can be more dif-
π 2 E ( I x I y − I xy
2
) ficult to calculate. The following subsections discuss these
M cr = Cb (40) properties for different cases. Formulas are provided for
L2 I x many of these properties using midline dimensions and
⎡ ⎤ thin-walled assumptions, ignoring fillets. Full integration
⎢C β x + ⎛ β x ⎞ + ( GJL + π ECw ) I x
2 2 2
⎥ including fillets is available in software tools such as those
⎢ s ⎝ 2⎠ ⎥
⎢ 2 π 2 E ( I x I y − I xy
2
) ⎥ previously mentioned and could be predetermined and tab-
⎣ ⎦
ulated for standard shapes as is done currently for J and Cw
As will be discussed, Equation 40 applies to every cross in the AISC Manual (2023).
section utilized in the AISC Specification. There is only one
formula necessary for predicting the elastic lateral-torsional Bending about Axis of Symmetry
buckling moment of all steel sections under the assump- Figure 5 illustrates several common shapes bending about
tions previously stated. the axis of symmetry, which is both the geometric x-axis
It can be useful to determine the unbraced length L cor- and the principal x-axis (Ixy = 0). This symmetry results
responding to any critical moment Mcr. Solving Equa- in βx = 0 because the shear center and asymmetry point lie
tion 39 (principal axis bending) for L2 gives the following on the axis of symmetry (yo = ya = 0). The elastic buckling
expression:

Fig. 5. Sections bending about the axis of symmetry.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 33

equation then simplifies to the following, which is also The βx / 2 value is negative when the flange is on the posi-
given in a User Note of the AISC Specification (2022) Sec- tive y-side of x-axis for tee sections with typical proportions
tion F2: (t1 > t w, b1 < 2dc). The shear center is at the intersection of
the flange and web, therefore Cw can be taken as 0.
π ⎛ π2 ⎞
Mcr = Cbro Pey Pt = Cb EI y ⎜ GJ + 2 ECw⎟ (43)
L ⎝ L ⎠ Double Angle

For doubly symmetric I-sections (W, M, S, HP) and chan- Back-to-back angle sections are similar to tee sections but
nels (C, MC), the torsional properties J and Cw are given in with an optional spacing between the vertical legs. Fig-
AISC Manual (2023) tables. For rectangular HSS, tee (WT, ure 9 shows general back-to-back angles with dimensional
MT, ST), double-angle sections, and equal-leg angles, J is parameters and formulas for the required section proper-
given in the AISC Manual tables, whereas Cw is much less ties. The βx / 2 value is negative when the flanges are on the
significant and ignored in the AISC Specification. Thick positive y-side of x-axis for typical double angle propor-
flanges and large fillets increase Cw, which can be com- tions (2sc < bc < dc). The shear center is at the intersection
puted using numerical methods. of the flanges and axis between the vertical legs. For close
spacings, Cw is small and can be taken as 0.
Box Sections
Single Angle
Figure 6 shows a hollow rectangular box section with
dimensional parameters and formulas for the required sec- A common application for a structural angle is loading in
tion properties. This section is doubly symmetric, thus βx = the plane of the web with bending about the geometric axis
0, and Equation 43 is applicable. perpendicular to the web. The more general form of the Mcr
calculation given by Equation 40 is required, which incor-
Mono-Symmetric I-Section porates the term (IxIy − Ixy2)/Ix in place of Iy.
Figure 10 shows a general angle section with dimen-
Figure 7 shows a general mono-symmetric I-section with sional parameters and formulas for the required properties,
dimensional parameters and formulas for the required sec- including a direct formula for the expression (IxIy − Ixy2)/Ix.
tion properties. The Cs sign coefficient corresponds to the A simple formula is provided for βx / 2, which is negative
direction of the applied moment, but the sign of βx / 2 is also when the flange is on the positive y-side of the x-axis. The
important and must be calculated properly. The βx / 2 value shear center is at the intersection of the angle legs; there-
is negative when the larger flange is on the positive y-side fore, Cw can be taken as 0 (again assuming mid-line dimen-
of the x-axis for I-sections with typical proportions (t1 = t 2 > sions and thin-walled assumptions, thus ignoring fillets and
t w, b2 < b1 < dc). secondary warping that result in quite small, but non-zero
Cw).
Tee Section
Figure 8 shows a general tee section with dimensional
parameters and formulas for the required section properties.

A = ( bc + t w )( dc + tf ) − ( bc − t w )( dc − t f )

1 3 1 3
Ix = ( bc + tw )( dc + t f ) − ( bc − tw)( dc − tf )
12 12
1 1
Iy = ( dc + tf )( bc + tw )3 − ( dc − tf )( bc − tw )3
12 12

2bc2dc2
J=
dc t w + bc t f

bc dc (dc t w − bc t f )
ω n,max =
4 (dc t w + bc t f )

1
Cw = Aω 2n,max
3

Fig. 6. Box section properties.

34 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 A = t1b1 + t2b2 + t w dc

y1 = ( t2 b2 dc + 1 tw dc2 ) A y2 = dc − y1
2
1
I x = t1b1 y12 + t2b2 y22 + t w ( y13 + y23 )
3
1 3 1
Iy = t1b1 + t2b23 = I y1 + I y2
12 12
1
J= ( b1t13 + b2t 23 + dc tw3 )
3

yo = ( t1b13y1 − t2b23 y2 ) 12I y

⎡⎢ t b y ( b 2 + 12y 2 ) − t b y ( b 2 + 12y 2 )⎤⎥

1 1 1 1 1 2 2 2 2 2 1
ya = ⎢ ⎥
⎢ +3t w ( y14 − y24 ) ⎥ 24Ix
⎣ ⎦

β x 2 = ya − yo

Cw = t1b13t2b23d c2 144I y = I y1I y2 dc2 I y

Fig. 7. Mono-symmetric I-section properties.

t w dc2
y1 = y2 = dc − y1
2 ( t1b1 + t w dc )

Ix = t1b1y12 + 13 t w ( y13 + y 23 )

1
I y = 12 t1b13

J = 13 (b1t13 + dc t w3 ) Cw = 0

yo = y1

ya = ⎡⎣t1b1y1 ( b12 + 12y12 ) + 3t w ( y14 − y24 )⎤⎦ 24Ix

β x 2 = ya − yo

Fig. 8. Tee section properties.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 35

 tdc3 ⎛ 4bc + dc ⎞
Ix = ⎜ ⎟ J = 23 ( bc + dc ) t 3
6 ⎝ bc + dc ⎠

t 3
Iy = ⎡bc + 3bc ( sc + bc )2 + 3dc sc2 ⎤
6⎣ ⎦

dc2
yo = Cw = 0
2 ( bc + dc )

bc2 ( 2bc + 3sc )( bc + dc ) − 3bc2dc2

ya =
2dc ( bc + dc )( 4bc + dc )

βx ( 2bc2 − 2bc dc − dc2 )( bc + dc ) + 3bc2sc

= ya − yo =
2 2dc ( 4bc + dc )

Fig. 9. Double-angle section properties.

tdc3 ⎛ 4bc + dc ⎞ tbc2dc2

Ix = ⎜ ⎟ Ixy =
12 ⎝ bc + dc ⎠ 4 ( bc + dc )

tbc3 ⎛ bc + 4dc ⎞
Iy = ⎜ ⎟ J = 13 ( bc + dc ) t 3
12 ⎝ bc + dc ⎠

2
Ix I y − I xy tb3 ⎛ b + dc ⎞
= c ⎜ c ⎟ Cw = 0
Ix 3 ⎝ 4bc + dc ⎠

dc2 bc ( bc2 + 3dc2 )

yo = ya = −
2 ( bc + dc ) 8dc ( bc + dc )

βx 4d 2 − bc dc + bc2
= ya − yo = − c
2 8dc

Fig. 10. Single-angle section properties.

36 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 For bending about the principal x-axis, the properties DISCUSSION OF AISC PROVISIONS
Iy and βx/ 2 are required, which are determined as follows: The elastic lateral-torsional buckling formulas in the AISC
1 1 2 Specification (2022) are substantially based on Equa-
I y = ( I x + I y) − ( I x − I y ) + 4Ixy2 (44)
2 2 tion 39; however, many different formulas are given—
uniquely customized and approximated for all the different
1 −2Ixy types of standard sections. The accuracy of these elastic
α= arctan (45)
2 Ix − I y buckling formulas and approximations is examined in this
section. Note, the objective of the AISC Specification is
βx βx βy
= cos α − sin α = (46) to provide a reliable nominal moment; however, here only
2 2 2 the elastic lateral-torsional buckling portion of the calcula-
⎛ 4d 2 − bc dc + bc2 ⎞ ⎛ 4bc2 − bc dc + dc2 ⎞ tion is assessed, and thus, the differences do not necessarily
−⎜ c ⎟ cos α − ⎜ ⎟ sin α
⎝ 8dc ⎠ ⎝ 8bc ⎠ equate to meaningful strength reliability and only to accu-
racy of the elastic expressions.
For the special case of an equal leg angle, Iy = tbc3/12, α =
−π/4, and βx/ 2 = 0. The major principal axis is an axis of Bending about Axis of Symmetry
symmetry and the simpler form of Equation 43 can be used.
AISC Specification Equation F2–4 is shown here as Equa-
tion 47 converted from stress to moment, applicable to dou-
General Built-Up Section
bly symmetric I-shapes and channels.
For any general section built up of other shapes, most sec- 2
tion properties can be calculated from the properties of the Cb π 2 ES x Jc ⎛ L ⎞
Mcr = 1+ 0.078 ⎜ ⎟ (47)
individual shapes. Figure 11 shows an example built-up sec- ( L rts )2
Sx ho ⎝ rts ⎠
tion and the method of calculating many of the properties.
Calculating the torsion constant J as the summation of I y Cw
rts2 = (48)
the individual J values is conservative and often accurate Sx
enough, but built-up sections that form fully enclosed hol-
low regions will have much larger J values. This and other ho I y
c = 1 for I-shapes, c = for channels (49)
torsional properties (shear center and warping constant) are 2 Cw
more difficult to calculate for a general section. Software
Equation 47 is identical to Equation 43 but restructured
tools that utilize numerical methods are recommended
to use new variables rts and c, which are themselves defined
for determining these properties. As with the single angle
as functions of Iy, Cw, Sx, and ho. When the expressions for
section, bending about the non-principal x-axis would use
rts and c are substituted into this equation, the variables Sx
Equation 40.

I x = ∫A y 2 dA = ∑ I xi + ∑ Ai yi2

I y = ∫A x 2 dA = ∑ I yi + ∑ Ai xi2

Ixy = ∫ A xy dA = ∑ I xyi + ∑ Ai xi yi

J = ∑ Ji (conservative)

Ux I y − Uy I xy
ya =
2 ( I x I y − I xy
2
)

Ux = ∫A y3 dA + ∫ A x 2 y dA

Uy = ∫ A x 3 dA + ∫ A y 2 x dA

Fig. 11. Built-up unsymmetrical section properties.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 37

and ho factor out entirely. These additional variables are AISC Specification Section F5 also applies to I-shapes
unnecessary and even misleading in suggesting what prop- but with slender webs. Specification Equation F5-4 is shown
erties control lateral-torsional buckling. It should also be here as Equation 51, which is the same as Equation 50 with
noted that lateral-torsional buckling for other section types the exclusion of torsional constant J, and is therefore a more
bending about an axis of symmetry as shown in Figure 5 conservative approximation of Mcr.
are not addressed in Chapter F.
Cb π 2 ES x
Mcr = (51)
Mono-Symmetric I-Section ( L rt )2
AISC Specification Equation F4-5 is shown here as Equa-
tion 50 converted from stress to moment, applicable to Square and Rectangular HSS and Box Sections
mono-symmetric I-shapes.
AISC Specification Equation F7-9 is shown here as Equa-
2 2 tion 52, applicable to closed doubly symmetric rectangular
Cb π ES x J ⎛ L⎞
M cr = 1+ 0.078 ⎜ ⎟ (50) sections—that is, hollow structural sections (HSS) and box
( L rt )2 Sx ho ⎝ rt ⎠
sections. This is equivalent to Equation 43 where substi-
This is similar to Equation 47 but restructured to use tutions have been made for G = E/ 2.6, Iy = Ar y2, Cw = 0,
the additional variable rt, which is defined as a function and π 2.6 has been rounded up to 2. The assumption that
of dimensional parameters of the section (i.e., the flange Cw = 0 (i.e., negligible) is accurate for square HSS of uni-
and a portion of the web). The square of rt is intended to form thickness, although Cw increases with aspect ratio for
approximate Iy Cw Sx combined with the influence of βx. rectangular HSS. Torsional stiffness is dominated by pure
However, this approximation can lead to significant error torsion (GJ) for common HSS and box sections; hence, it
compared to Equation 39 as illustrated in Figure 12. The is generally appropriate to ignore the negligible influence
use of rt (and rts for doubly symmetric I-shapes) reverts to of Cw.
the pre-1986 concept of treating the compression flange as a JA
column, which is no longer necessary given the availability Mcr = 2ECb (52)
L ry
of the relevant section properties or methods of calculating
them. Further, the subtle difference between rt and rts can
lead to additional error/confusion. Tees and Double Angles
The AISC Specification Commentary (2022) includes
the theoretically based Equation C-F4-3 incorporating Cw AISC Specification Equation F9-10 is shown here as Equa-
and βx and is equivalent to Equation 39. An accurate for- tion 53, applicable to tees and double angles. Note that the
mula for Cw and an approximation for βx are provided as bending coefficient Cb is not utilized in this equation as
functions of section depth and flange moments of inertia. explained in the AISC Specification Commentary. For these
Figure 12 shows this approach is more accurate than Spec- sections, Cw is typically negligible and can be ignored. The
ification Equation F4-5, although the βx approximation new variable, B, in the AISC formula is a dimensionless
increases the error for larger flanges. value intended to represent (π βx 2L ) EI y GJ by (crudely)

1.5
Mcr,AISC
Mcr

1
b1
Eq. F4-5(+)
0.5
b1/b2=2 dc Eq. F4-5(-)
tf /tw=2 Eq. C-F4-3(+)
L/dc=20 b2 Eq. C-F4-3(-)
0
0 0.5 1 1.5 2
b1 / dc

Fig. 12. AISC Mcr equation accuracy for mono-symmetric I-sections.

38 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

approximating βx as 0.9y2. The error introduced by this J = At 2/ 3, and Cw = 0. Values for βx are provided in AISC
approximation is shown in Figure 13 for tee sections and Specification Commentary Table C-F10.1 as βw for com-
Figure 14 for double-angle sections, as compared to Equa- mon angle sizes independent of thickness. These values are
tion 39. For tee sections with b1/dc > 1.5 and double angles accurate for b/ t = 16, where b is the longer leg. For b/ t other
with b/d > 1, the error is significant. The error for double than 16, some small error is introduced.
angles increases as the spacing between them increases.
⎡ 2 ⎤
Cb EAryt ⎢ β x ry ⎛ β x ry ⎞
( )
1.95E M cr = 1.125 ±4.4 + ⎜ 4.4 ⎟ + 1 ⎥ (55)
Mcr = Iy J B + 1+ B 2 (53) L ⎢ Lt ⎝ Lt ⎠ ⎥
Lb ⎣ ⎦
y2 Iy AISC Specification Equations F10-5a and 5b are shown
B = ±2.3 (54)
L J here as Equation 56, applicable to equal leg angles bend-
ing about the geometric x-axis. This is equivalent to Equa-
tion 40, where substitutions have been made for G = E/ 2.6,
Single Angle (IxIy − Ixy2)/ Ix = 2tbc3/15, J = 2bct 3/ 3, Cw = 0, βx / 2 = −bc/ 2,
AISC Specification Equation F10-4 is shown here as Equa- b = bc+t/ 2, and t = b/16. For b/ t other than 16, some small
tion 55, applicable to single angles bending about their error is introduced. AISC Specification Section F10 has no
major principal axis. This is equivalent to Equation 39, provisions for an unequal leg angle bending about a geo-
where substitutions have been made for G = E/ 2.6, Iy = Ary2, metric axis.

Fig. 13. AISC Mcr equation accuracy for tee sections.

Fig. 14. AISC Mcr equation accuracy for double-angle sections.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 39

 ⎡ 2 ⎤ The recommendation is to provide Equation 57, along
Cb Eb4t ⎢±1+ 0.88 ⎛ Lt2 ⎞ + 1 ⎥
M cr = 0.58 (56) with the simplified formulas for the special cases associ-
L2 ⎢⎣ ⎝b ⎠ ⎥⎦ ated with Equations 58 and 59. For all the common sec-
tions covered in AISC Specification (2022) Chapter F,
closed-form equations have been provided herein for cal-
Unsymmetrical Sections culating all the necessary section properties (Ix, Iy, Ixy, J,
AISC Specification Section F12 offers no formula to deter- Cw, yo, ya, βx / 2), and most of these properties are tabulated
mine Mcr for an unsymmetrical section and must therefore in the AISC Manual (2023). The addition of βx / 2 in the
be determined by analysis. It is important to note that Equa- torsion property tables would fulfill the section properties
tion 40 is valid for unsymmetrical sections, so the task is requirements. For other sections, formulas can be derived,
simply to determine the required properties of the cross or software tools utilizing numerical methods may be used.
section. Although this can be difficult for the torsional
properties, section property software tools are available to CONCLUSIONS
assist if necessary. Then a more rigorous elastic beam anal-
ysis is not necessary. The elastic lateral-torsional buckling moment, Mcr, for a
linear prismatic beam is a single formula applicable to any
Recommendations cross section. Properties of the cross section are required
to calculate this buckling moment, including the St. Venant
The AISC Specification (2022) presents several different torsional constant, J, and warping constant, Cw. In addition,
formulas for calculating Mcr , with the use of additional the less common asymmetry property βx is required and
variables and approximations. The authors assert that it involves some care in calculation. For sections symmetric
would be more useful and instructive to provide one gen- about the axis of bending, βx is zero, and the lateral-torsional
eral formula to the designer, applicable to any cross section, buckling moment calculation is greatly simplified.
in the main body of the Specification—and provide design This article provides formulas for directly calculating
aids, or user notes where appropriate, for how the expres- βx for many standard shapes using centerline dimensions
sion simplifies under certain assumptions (cross-section and thin-wall assumptions. This includes mono-symmetric
types). Specifically, we recommend that the Specification I-sections, tee sections, double-angle sections, and
provide Equation 40 for the elastic lateral-torsional buck- single-angle sections. For sections bending about a non-
ling moment bending about the geometric x-axis, repeated principal axis, such as single angles or unsymmetrical
here for clarity: built-up sections, the buckling formula is essentially the
same except that the minor axis moment of inertia must be
π 2 E ( Ix I y − I xy
2
)
Mcr = Cb (57) generalized to incorporate the product of inertia.
L2 Ix The AISC Specification provisions for lateral-torsional
⎡ ⎤ buckling consist of different formulas for the various types
⎢C β x + ⎛ β x ⎞ + ( GJL + π ECw ) I x
2 2 2
⎥ of standard shapes, where additional variables and formulas
⎢ 2
s
⎝ 2⎠ π 2 E ( Ix I y − I xy
2
) ⎥ are used to approximate βx rather than calculating the actual
⎣ ⎦ values. These approximations introduce errors in the calcu-
For sections where the geometric axes coincide with lation of Mcr that can be unacceptably large. The removal of
the principal axes, Ixy = 0 and the buckling formula can be these approximations is recommended. The preferred alter-
expressed as: native is to include the βx property in all the torsional prop-
erty tables in the next edition of the AISC Manual. Such
π 2 EI y ⎡ β 2 2 ⎤ tabulated values could be more accurately determined using
Mcr = Cb ⎢Cs x + ⎛ β x ⎞ + GJL + Cw ⎥ (58)
L2 ⎢ 2 ⎝ 2⎠ π 2 EI y I y ⎥ direct integration of the cross-section properties to account
⎣ ⎦ for fillets and other features not incorporated in the typical
mid-line, thin-walled properties. The section property for-
If the x-axis is also an axis of symmetry, the shear cen-
mulas provided herein would also be helpful to include in
ter and asymmetry point lie on the x-axis (yo = ya = 0), thus
a future edition of the AISC Specification Commentary or
βx / 2 = 0 and the buckling formula further simplifies to:
the AISC Manual.
π ⎛ π2 ⎞ For unsymmetrical or built-up sections, the calculation
Mcr = Cb EI y ⎜ GJ + 2 ECw⎟ (59) of torsional properties and βx are more difficult. Software
L ⎝ L ⎠
tools using numerical methods are available to calculate

40 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

these properties. In addition, though not detailed herein, Glauz, R.S. (2017), “Elastic Lateral-Torsional Buckling
finite element software can be used to model and calculate of General Cold-Formed Steel Beams under Uniform
Mcr, not only for unique cross sections, but also for other Moment,” Thin-Walled Structures, Vol. 119.
conditions that affect elastic stability, such as pre-buckling Helwig, T.A., Frank, K.H., and Yura, J.A. (1997), “Lateral-
displacements, nonuniform moment, unusual boundary Torsional Buckling of Singly-Symmetric I-Beams,”
conditions, shear stresses, and load height. Journal of Structural Engineering, ASCE, Vol. 123.
Kirby, P.A. and Nethercot, D.A. (1979), Design for Structural
REFERENCES
Stability, John Wiley & Sons, Inc., New York, N.Y.
AISC (1923), Standard Specification for Structural Steel Liang, C., Reichenbach, M.C., Helwig, T.A., Engelhardt,
Buildings, American Institute of Steel Construction, New M.D., and Yura, J.A. (2022), “Effects of Shear on the
York, N.Y. Elastic Lateral Torsional Buckling of Doubly Symmetric
AISC (1936), Specification for the Design, Fabrication and I-Beams,” Journal of Structural Engineering, Vol. 148.
Erection of Structural Steel Buildings, American Institute Liu, S.W., Ziemian, R.D., Chen, L., and Chan, S.L. (2018),
of Steel Construction, New York, N.Y. “Bifurcation and Large-Deflection Analyses of Thin-
AISC (1946), Specification for the Design, Fabrication and Walled Beam-Columns with Non-Symmetric Open-
Erection of Structural Steel Buildings, American Institute Sections,” Thin-Walled Structures, Vol. 132.
of Steel Construction, New York, N.Y. MSASect2 (2023), Matrix Structural Analysis for Arbitrary
AISC (1961), Specification for the Design, Fabrication and Cross-Sections, Version 1.0.6, Liu, S.W. and Ziemian,
Erection of Structural Steel Buildings, American Institute R.D. [Link]
of Steel Construction, New York, N.Y. Salmon, C.G. and Johnson, J.E. (1980), Steel Structures:
AISC (1981), Manual of Steel Construction, 8th Ed., Design and Behavior, 2nd Ed., Harper & Row, New York,
American Institute of Steel Construction, Chicago, Ill. N.Y.
AISC (1986), Specification for Structural Steel Buildings— ShapeBuilder (2023), Geometric and Structural Properties
Load and Resistance Factor Design, American Institute of Cross Sections, Version 14.0, IES, Inc. [Link]
of Steel Construction, Chicago, Ill. .[Link]/sb.
AISC (2022), Specification for Structural Steel Buildings, Sectionproperties (2024), Open Source Python Package for
ANSI/AISC 360–22, American Institute of Steel the Analysis of Arbitrary Cross-Sections Using the Finite
Construction, Chicago, Ill. Element Method, Version 3.2.1, van Leeuwen, R. https://
AISC (2023), Steel Construction Manual, 16th Ed., [Link].
American Institute of Steel Construction, Chicago, Ill. Timoshenko, S.P. (1936), Theory of Elastic Stability, 1st Ed.,
Bradford, M.A. (1992), “Lateral-Distortional Buckling of McGraw-Hill Book Co., New York, N.Y.
Steel I-Section Members,” Journal of Constructional Timoshenko, S.P. and Gere, J.M. (1961), Theory of Elastic
Steel Research, Vol. 23. Stability, 2nd Ed., McGraw-Hill, New York, N.Y.
CEN (2005), Eurocode 3: Design of Steel Structures, EN Wong, E. and Driver, R. (2010), “Critical Evaluation of
1993, The European Committee for Standardization, Equivalent Moment Factor Procedures for Laterally
Brussels, Belgium. Unsupported Beams,” Engineering Journal, AISC,
Commissioners of Railways (1848), Report of the Vol. 47, No. 1, pp.1–20.
Commissioners of Railways, London, England. Yu, W.W. (2000), Cold-Formed Steel Design, 3rd Ed., John
Euler, L. (1744), Methodus Inveniendi Lineas Curvas, Wiley & Sons, Inc., New York, N.Y.
Lausannæ & Genevæ, Apud Marcum-Michaelem Yura, J.A. (1995), “Bracing for Stability—State-of-the-
Bousquet & Socios. Art,” Proceedings of the ASCE Structures Congress XIII,
Boston, Mass., ASCE, New York, N.Y.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 41

42 / ENGINEERING JOURNAL / FIRST QUARTER / 2025
Strength Coefficients for Eccentrically
Loaded Weld Groups
BO DOWSWELL

ABSTRACT
When AISC Manual Tables 8-4 through 8-11 are used to calculate the strength of eccentrically loaded weld groups with FEXX ≠ 70 ksi, the
values are multiplied by an electrode strength coefficient, C1. The C1 values are dependent on the filler metal classification strength; how-
ever, they are not proportional to the filler metal classification strength ratio when FEXX ≥ 80 ksi. To consider the potential effect of reduced
ductility, the C1 values include reduction factors of 0.85 and 0.90 for higher-strength welds.

To investigate the accuracy of the electrode strength coefficients, the ductility of high-strength welds was evaluated using the data from 93
experimental tests from three existing research projects with FEXX > 70 ksi. The data was used to plot the weld metal tensile strength ver-
sus the normalized rupture deformation of both longitudinal and transverse fillet welds. The analysis showed that, when FEXX ≤ 120 ksi, the
C1 values can be based solely on the filler metal classification strength ratio, FEXX/ 70 ksi, without the reduction factors of 0.85 and 0.90 for
higher-strength welds.

Keywords: eccentrically loaded weld groups, electrode strength coefficient, weld ductility.

INTRODUCTION with an empirical load-deformation curve for short seg-

ments within the weld group.
AISC Steel Construction Manual (2023) Tables 8-4
Because the ICR method is iterative, considerable design
through 8-11 are used to calculate the strength of eccentri-
effort is required to calculate the strength of a weld group
cally loaded weld groups. The tables were developed using
using this method. AISC Manual Tables 8-4 through 8-11
the instantaneous center of rotation (ICR) method with a
provide a simpler, noniterative design method by listing the
filler metal classification strength, FEXX, of 70 ksi. For other
appropriate coefficients for several different weld group
filler metal strengths, Table 8-3 provides electrode strength
geometries. The tables were developed using the equations
coefficients, C1, that are used with Tables 8-4 through 8-11.
in AWS D1.1 Subclause [Link]. These equations are also
The C1 values are dependent on the filler metal classifica-
shown on AISC Manual pages 8-13 and 8-14. AWS D1.1 is
tion strength; however, they are not proportional to the filler
based solely on ASD. The AISC Manual provides the nom-
metal classification strength ratio when FEXX ≥ 80 ksi. This
inal strength equations, which can be used with ASD and
results in a significant strength reduction for higher-strength
LRFD. The weld group strength is the sum of the strengths
welds, which is not required in either the AISC Specifica-
of each element within the group. The nominal stress in
tion (2022) or AWS D1.1 (2020a). Based on experimental
Element i is calculated with AISC Manual Equation 8-3.
tests, the accuracy of the electrode strength coefficients
0.3
will be determined and revised values will be proposed. Fnwi = 0.60FEXX (1.0 + 0.50sin1.5 θi ) ⎡⎣ pi (1.9 − 0.9 pi )⎤⎦
(Manual Eq. 8-3)
AISC MANUAL
where 0.60FEXX is the nominal stress for concentrically
Part 8 of the AISC Manual discusses three methods to ana- loaded fillet welds from AISC Specification Table J2.5.
lyze eccentrically loaded weld groups: The ICR method, The first term in parentheses is the directional strength
the elastic method and the plastic method. Both the elastic increase factor, which is identical to AISC Specification
and plastic methods were developed theoretically. The ICR Equation J2-5 if θi is replaced with θ. The second bracketed
method was developed using a semi-empirical approach, term is a strain compatibility factor, which is dependent on
the deformation ratio, pi. The ratio of Element i deformation
to its deformation at maximum stress is
Δi
Bo Dowswell, PE, PhD, Principal, ARC International, LLC, Birmingham, Ala. pi = (Manual Eq. 8-4)
Email: bo@[Link]
Δ mi

Paper No. 2024-11

ISSN 2997-4720 ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 43

 The deformation of Element i at maximum stress is ICR method will be briefly reviewed, followed by the
−0.32 implementation history of the electrode strength coeffi-
Δ mi = 0.209w ( θi + 2 )
cients in the AISC Manual.
(Manual Eq. 8-6)
The deformation of Element i at an intermediate stress Instantaneous Center of Rotation Method
level is Butler et al. (1972) developed the ICR method based on the
ri empirical load-deformation curves from Butler and Kulak
Δ i = Δ ucr (Manual Eq. 8-5) (1971), who tested concentrically loaded fillet welds at
rcr
angles of 0°, 30°, 60°, and 90° from the loading direction.
where Δucr is the deformation of the weld element with the The tests by Butler and Kulak as well as the tests on eccen-
minimum Δui /ri ratio at ultimate (rupture) stress. The defor- trically loaded weld groups by Butler et al. used 60 ksi elec-
mation of Element i at ultimate (rupture) stress is trodes and 4 in. fillet welds.
The ICR equations in both AWS D1.1 Subclause [Link]
Δ ui = 1.087w ( θi + 6 )−0.65 < 0.17w and the AISC Manual were primarily developed by Lesik
(Manual Eq. 8-7) and Kennedy (1990). Lesik and Kennedy used linear regres-
sion to formulate the load-deformation curves with the data
where
from Miazga and Kennedy (1989), who tested concentri-
FEXX = filler metal classification strength, ksi
cally loaded 70 ksi fillet welds with varying load angles,
rcr = distance from the instantaneous center of rotation θ, from 0 to 90° in 15° increments. These equations were
to the element with the minimum Δui /ri ratio, in. used to plot the strength ratio, Fnw/ (0.6FEXX), versus nor-
ri = distance from the instantaneous center of rotation malized deformation, Δ/ w, curves for θ = 0°, 30°, 60°, and
to Element i, in. 90° in Figure 1. An increase in θ results in a nonlinear
w = weld leg size, in. strength increase based on the directional strength increase
θi = angle between the longitudinal axis of weld ele- factor and a decrease in ductility based on AISC Manual
ment i and the direction of the resultant force act- Equation 8-7.
ing on the element, degrees Lesik and Kennedy (1990) originally developed the
strain compatibility factor as a polynomial function; how-
BACKGROUND ever, their equation was replaced by the simpler empiri-
cal approximation in AISC Manual Equation 8-3. Also, an
This section of the paper will document background infor- upper limit of 0.17w was added to their original equation for
mation related to the electrode strength coefficients. The Δut, resulting in AISC Manual Equation 8-7.

1.6 θ = 90°

1.4
θ = 60°
1.2
θ = 30°
Strength Ratio

1.0
θ = 0°
0.8

0.6

0.4

0.2

0.0
0.00 0.05 0.10 0.15 0.20
Δ /w

Fig. 1. Fillet weld strength ratio versus normalized deformation as a function of load angle, θ.

44 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 Table 1. Electrode Strength Coefficient, C1
FEXX 60 70 80 90 100 110
C1 0.857 1.00 1.03 1.16 1.21 1.34
FEXX
0.857 1.00 1.14 1.29 1.43 1.57
70 ksi

C1
⎛ FEXX ⎞ 1.00 1.00 0.90 0.90 0.85 0.85
⎜ ⎟
⎝ 70 ksi⎠

Electrode Strength Coefficient made from electrodes other than E60 and E70 by ascertain-
ing the load-deformation response for these welds.”
The values in AISC Manual Tables 8-4 through 8-11 were
An accurate solution relies on sufficient ductility of the
calculated using FEXX = 70 ksi. The strength of weld groups
critical segment for load redistribution without rupture.
with other weld metal strengths can be calculated by adjust-
The reduction factors were likely implemented in the AISC
ing the tabulated values by the electrode strength coeffi-
Manual because higher-strength welds are less ductile than
cient, C1 in Manual Table 8-1.
E60 and E70 welds. However, in 1986 when these reduc-
The 6th Edition AISC Manual (1967) was the first to
tion factors were first published, experimental strength ver-
provide information on eccentrically loaded weld groups.
sus deformation data was unavailable for welds with FEXX
The elastic method was used to develop design tables with
greater than 70 ksi.
FEXX = 60 ksi. The weld group strengths for other filler
metal classification strengths were calculated with the filler
metal classification strength ratio, FEXX/ 60 ksi. The 7th Edi- DUCTILITY OF HIGH-STRENGTH WELDS
tion Manual (AISC, 1973) used elastic design with FEXX =
To investigate the accuracy of the electrode strength coef-
70 ksi; therefore, the weld group strength for other filler
ficients, the ductility of high-strength welds will be evalu-
metal classification strengths was calculated with the filler
ated. After the modified coefficients were first published in
metal classification strength ratio, FEXX/ 70 ksi.
1986, a significant amount of experimental strength-versus-
The 8th Edition Manual (AISC, 1980) was the first to
deformation data has become available for welds with FEXX
publish design tables that were based on the ICR method.
greater than 70 ksi.
The development of these tables, which were also published
Because transverse fillet welds have significantly less
in the 9th Edition Manual (AISC, 1989), was discussed by
deformation capacity than longitudinal fillet welds, the duc-
Tide (1980). The table coefficients were calculated with
tility of transverse high-strength welds are the primary con-
FEXX = 70 ksi, and C1 was used to calculate the weld group
cern. In weld groups with both longitudinal and transverse
strength for other filler metal classification strengths, where
welds, the longitudinal weld strength will be limited by
C1 = FEXX/ 70 ksi.
the ductility of the transverse weld. According to the AWS
For the 1st Edition LRFD Manual (AISC, 1986) and the
D1.1 equation for rupture deformation (which is identical to
13th Edition combined ASD/LRFD Manual (AISC, 1992),
AISC Manual Equation 8-7), the normalized rupture defor-
as well as all later editions, the tables were based on the
mations for longitudinal and transverse welds are Δui/w =
ICR method with FEXX = 70 ksi. However, the value of C1
0.17 and Δui/w = 0.056, respectively.
included a reduction factor equal to either 0.90 (for 80 and
Figure 2 shows a plot of the weld metal tensile strength
90 ksi welds) or 0.85 (for 100 and 110 ksi welds). These val-
versus the normalized rupture deformation, Δu/w, of fil-
ues are shown in Table 1.
let welds. The data are from the 93 experimental tests on
The background of these reduction factors is ambigu-
high-strength longitudinally and transversely loaded fillet
ous, and communication with members of past AISC Man-
welds by Collin and Johansson (2005), Bjork et al. (2012),
ual Committees (Thornton, 2020; Tide, 2020) revealed
and Sun et al. (2019). The red × data points represent trans-
no further information. According to Butler et al. (1972),
verse welds, and the blue hollow circles represent longitudi-
who originally developed the ICR method based on FEXX =
nal welds. The red and blue vertical dashed lines represent
60 ksi, “Because E60 and E70 electrodes have specified
the AWS normalized rupture deformations for longitudinal
ultimate elongations nearly the same, it is felt that these
and transverse welds, respectively. It can be observed that,
results could be applied to connections made using E70
for tensile strengths less than 120 ksi, the AWS D1.1 equa-
electrodes by proper consideration of the increase in elec-
tion for rupture deformation (which is identical to AISC
trode strength. The method could be used for fillet welds

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 45

Manual Equation 8-7) generally provides conservative esti- limited comparisons with experimental results showed “an
mates of the normalized rupture deformation. acceptable fit.” The ultimate (rupture) force and normal-
The average normalized deformations from this data are ized deformation are calculated with Equations 1 and 2,
listed in Table 2. The data for 60 ksi welds from Butler and respectively.
Kulak (1971) are also listed. A comparison of the rupture
deformations shows that, for longitudinal welds, the rupture 1+ 15sin 2 α d
Ru = σ tu wL (1)
deformation of high-strength welds is 68% of that of 60 ksi 6 (1+ 7sin 2 α d )
welds; however, the rupture deformation of transverse
welds is independent of strength. Because the shape of the 3
δu = εu (2)
load-deformation curves for high-strength welds is similar 2 (1+ 7sin 2 α d )
to that of 60 ksi welds, high-strength longitudinal welds in
weld groups will reach a higher proportion of their rupture The load-deformation curve can be plotted with Equa-
load compared to 60 ksi welds. The average transverse-to- tions 3 through 5.
longitudinal normalized deformation ratio for lap joints is fi
0.103/ 0.284 = 0.363, which is similar to the value calculated Ri = Ru (3)
fu
with AISC Manual Equation 8-7: 0.056/ 0.17 = 0.33.
e −25δi + e−75δi
Load-Deformation Curves fi = 1 − (4)
2
An evaluation of the load-deformation curves can provide
further information on the behavior of high-strength fillet e −25δ + e −75δu
fu = 1 − (5)
welds. The equations developed by Neis (1985) explicitly 2
compensate for the effect of reduced weld metal ductility on
As a conservative estimate, the authors noted that the true
the behavior. Neis used plasticity theory to derive the ulti-
tensile rupture stress can be calculated with Equation 6.
mate strength and maximum displacement of fillet welds.
Although several simplifying assumptions were required, σ tu = σ u (1+ 0.75εu ) (6)

200
Measured Weld Metal Tensile Strength (ksi)

180

160

140

120

100

80

60
Longitudinal Tests
40 Longitudinal Rupture (AWS)
Transverse Tests
20
Transverse Rupture (AWS)
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
Δu /w

Fig. 2. Weld metal tensile strength versus normalized rupture deformation.

46 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 Table 2. Average Normalized Deformation
FEXX = 60 ksi (Butler High Strength
and Kulak, 1971) (FEXX ≈ 80 to 180 ksi)
Average Number of Average
Joint Type Δ u/ w Specimens Δ u/ w
Longitudinal 0.420 26 0.284
Transverse (total) — 67 0.0966
Transverse lap joints 0.104 36 0.103
Transverse T-joints — 31 0.0889

Table 3. Minimum Elongation for All-Weld-Metal Tension Tests (%)

Welding Process
FEXX ksi SMAW GMAW FCAW SAW
60 17 to 22 — 22 22
70 17 to 25 19 to 24 20 to 22 22
80 17 to 24 17 to 24 19 20
90 17 to 24 16 to 18 16 to 17 17
100 16 to 20 16 15 to 18 16
110 15 to 20 15 15 15
120 11 to 18 14 to 15 14 14

Equation 7 provides an approximate value of the angle A5.20 (2015), A5.23 (2011), A5.28 (2020b) and A5.29
between the weld longitudinal axis and the weld displace- (1998) are summarized in Table 3. Generally, weld metals
ment direction. exceed these requirements. For example, the average elon-
tan θ gation measurements in Dowswell et al. (2021) Table 3.7
tan α d = (7) are 40 to 50% higher than the required minimum values in
4
Table 3. These measurements are from all-weld-metal ten-
where sile tests with FEXX = 70, 80 and 100 ksi.
Ri = strength at deformation Δi, kips The values in Table 4 are appropriate lower bounds
L = weld length, in. for analyses with the Neis (1985) equations. The strength
ratios, σtu/σu, in Table 4 are between 1.11 and 1.17. These
αd = angle between the weld longitudinal axis and the
values are similar to the constraint factor by Miazga and
weld displacement direction
Kennedy (1989), which is 1.14 when θ = 90°.
Δi = deformation at an intermediate strength level, in. The Butler and Kulak (1971) curves were scaled up from
Δu = ultimate (rupture) deformation, in. 60 ksi to 70 ksi and plotted in Figures 3 and 4 for longitu-
δi = Δi/w dinal and transverse welds, respectively. These normalized
δu = Δu/ w load versus normalized deformation curves are for 70 ksi
εu = uniaxial engineering tensile rupture strain electrodes. The figures also include the AWS D1.1 and Neis
(1985) equations. The figures show that the Neis curves
σtu = true tensile rupture stress, ksi
provide a close approximation of the shape of the empiri-
σu = uniaxial engineering tensile rupture stress, ksi cal curves of Butler and Kulak, while also resulting in rup-
The elongation requirements for carbon and low-alloy ture loads that are similar to the AWS D1.1 equations. Also,
steels for shielded metal arc welding (SMAW), gas metal arc the Neis equations explicitly compensate for the effect of
welding (GMAW), flux cored arc welding (FCAW) and sub- reduced weld metal ductility on the behavior. Therefore, the
merged arc welding (SAW) welding processes from AWS Neis curves can be used as a baseline to project the behavior
A5.1 (2012), A5.5 (2014), A5.17 (2019), A5.18 (2017), of higher-strength weld metals.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 47

 Table 4. Variables for Neis (1985) Equations
FEXX ksi εu σtu ksi σtu/ σu
70 0.22 81.6 1.17
80 0.19 91.4 1.14
90 0.17 101 1.12
100 0.16 112 1.12
110 0.15 122 1.11
120 0.14 133 1.11

80
Butler & Kulak
Normalized Load, R/(wL), (ksi)

70 Neis
60 AWS
AWS rupture limit
50

40

30

20

10

0
0.00 0.05 0.10 0.15 0.20 0.25

Δ /w

Fig. 3. Normalized load versus normalized deformation for 70 ksi longitudinal fillet welds.

80
Normalized Load, R/(wL), (ksi)

70

60

50

40

30 Butler & Kulak

Neis
20
AWS
10 AWS rupture limit

0
0.00 0.05 0.10 0.15 0.20 0.25

Δ /w

Fig. 4. Normalized load versus normalized deformation for 70 ksi transverse fillet welds.

48 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

 For both the AWS D1.1 and Neis (1985) equations, the SUMMARY AND CONCLUSIONS
normalized load versus normalized deformation curves are
When AISC Manual Tables 8-4 through 8-11 are used to
plotted in Figures 5 and 6 for 70 ksi and 120 ksi electrodes,
calculate the strength of eccentrically loaded weld groups
respectively. Generally, the AWS D1.1 curves are higher
with FEXX ≠ 70 ksi, the values are multiplied by an electrode
than the curves for transverse welds and lower than the Neis
strength coefficient, C1. The C1 values are dependent on
curves for longitudinal welds. Because the AWS D1.1 equa-
the filler metal classification strength; however, they are not
tions predict a similar, but more conservative, proportion of
proportional to the weld metal tensile strength ratio when
the longitudinal strength at the transverse rupture load, it
FEXX ≥ 80 ksi.
can be concluded that the AWS D1.1 curves are conserva-
tive for both 70 ksi and 120 ksi electrodes.

80
Longitudinal (Neis)
Normalized Load, R/(wL), (ksi)

70 Longitudinal (AWS)
Longitudinal Rupture (AWS)
60
Transverse (Neis)
50 Transverse (AWS)
Transverse Rupture (AWS)
40

30

20

10

0
0.00 0.05 0.10 0.15 0.20 0.25

Δ /w

Fig. 5. Normalized load versus normalized deformation for 70 ksi fillet welds.

80
Normalized Load, R/(wL), (ksi)

70

60

50

40
Longitudinal (Neis)
30 Longitudinal (AWS)
Longitudinal Rupture (AWS)
20 Transverse (Neis)
10 Transverse (AWS)
Transverse Rupture (AWS)
0
0.00 0.05 0.10 0.15 0.20 0.25

Δ /w

Fig. 6. Normalized load versus normalized deformation for 120 ksi fillet welds.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 49

 Table 5. Current and Proposed Values for the Electrode Strength Coefficient
FEXX 60 70 80 90 100 110
Current 0.857 1.00 1.03 1.16 1.21 1.34
Proposed 0.857 1.00 1.14 1.29 1.43 1.57
Current
1.00 1.00 0.90 0.90 0.85 0.85
Proposed

An accurate solution relies on sufficient ductility of the AISC (2023), Steel Construction Manual, 16th Ed.,
critical segment for load redistribution without rupture. American Institute of Steel Construction, Chicago, Ill.
In 1986, when these modified electrode strength coeffi- AWS (1998), Specification for Low-Alloy Steel Electrodes
cients were first published, experimental strength-versus- for Flux Cored Arc Welding, AWS A5.29/A5.29M,
deformation data were unavailable for welds with FEXX > American Welding Society, Miami, Fla.
70 ksi. To consider the potential effect of reduced ductility,
AWS (2011), Specification for Low-Alloy Steel Electrodes
the C1 values included reduction factors of 0.85 and 0.90 for
and Fluxes for Submerged Arc Welding, AWS A5.23/
higher-strength welds.
A5.23M, American Welding Society, Miami, Fla.
To investigate the accuracy of the electrode strength
coefficients, the ductility of high-strength welds was evalu- AWS (2012), Specification for Carbon Steel Electrodes
ated. A significant amount of experimental strength-versus- for Shielded Metal Arc Welding, AWS A5.1/A5.1M,
deformation data is now available for welds with FEXX > American Welding Society, Miami, Fla.
70 ksi. The data from the 93 experimental tests from three AWS (2014), Specification for Low-Alloy Steel Electrodes
existing research projects were analyzed. The data were for Shielded Metal Arc Welding, AWS A5.5/A5.5M,
used to plot the weld metal tensile strength versus the nor- American Welding Society, Miami, Fla.
malized rupture deformation of both longitudinal and trans- AWS (2015), Specification for Carbon Steel Electrodes for
verse fillet welds. The analysis showed that when FEXX ≤ Flux Cored Arc Welding, AWS A5.20/A5.20M, American
120 ksi, the C1 values can be based solely on the filler metal Welding Society, Miami, Fla.
classification strength ratio, FEXX/ 70 ksi, without the reduc-
AWS (2017), Specification for Carbon Steel Electrodes and
tion factors of 0.85 and 0.90 for higher-strength welds. Both
Rods for Gas Shielded Arc Welding, AWS A5.18/A5.18M,
the current and proposed values for the electrode strength
American Welding Society, Miami, Fla.
coefficient are listed in Table 5.
AWS (2019), Specification for Carbon Steel Electrodes and
REFERENCES Fluxes for Submerged Arc Welding, AWS A5.17/A5.17M,
American Welding Society, Miami, Fla.
AISC (1967), Manual of Steel Construction, 6th Ed.,
AWS (2020a), Structural Welding Code—Steel, AWS D1.1,
American Institute of Steel Construction, Chicago, Ill.
American Welding Society, Miami, Fla.
AISC (1973), Manual of Steel Construction, 7th Ed.,
AWS (2020b), Specification for Low-Alloy Steel Electrodes
American Institute of Steel Construction, Chicago, Ill.
and Rods for Gas Shielded Arc Welding, AWS A5.28/
AISC (1980), Manual of Steel Construction, 8th Ed., A5.28M, American Welding Society, Miami, Fla.
American Institute of Steel Construction, Chicago, Ill.
Bjork, T., Toivonen, J., and Nykanen, T. (2012), “Capacity of
AISC (1986), Manual of Steel Construction—Load and Fillet Welded Joints Made of Ultra High-Strength Steel,”
Resistance Factor Design, 1st Ed., American Institute of Welding in the World, Vol. 56.
Steel Construction, Chicago, Ill.
Butler, L.J. and Kulak, G.L. (1971), “Strength of Fillet Welds
AISC (1989), Manual of Steel Construction—Allowable as a Function of Direction of Load,” Welding Research
Stress Design, 9th Ed., American Institute of Steel Supplement, pp. 231–234.
Construction, Chicago, Ill.
Butler, L.J., Pal, S., and Kulak, G.L. (1972), “Eccentrically
AISC (1992), Manual of Steel Construction—Volume II Loaded Welded Connections,” Journal of the Structural
Connections, ASD 9th Ed./LRFD 1st Ed., American Division, ASCE, Vol. 98, No. ST5.
Institute of Steel Construction, Chicago, Ill.
Collin, P.P. and Johansson, P.B. (2005), “Design of Welds in
AISC (2022), Specification for Structural Steel Buildings, High Strength Steel,” Proceedings of the Fourth European
ANSI/AISC 360-22, American Institute of Steel Conference on Steel and Composite Structures.
Construction, Chicago, Ill.

50 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

Dowswell, B., Cox, C., Gallow, M.S., and Fouad, F.H. (2021), Sun, F.F., Ran, M.M., and Wang, Y.B. (2019), “Mechanical
Fillet and PJP Welds, Final Report, American Institute of Behavior of Transverse Fillet Welded Joints of High
Steel Construction, Chicago, Ill. Strength Steel Using Digital Image Correlation
Lesik D.F. and Kennedy, D.J.L. (1990), “Ultimate Strength Techniques,” Journal of Constructional Steel Research,
of Fillet Welded Connections Loaded in Plane,” Canadian Vol. 162.
Journal of Civil Engineering, Vol. 17, No. 1, pp. 55–67. Thornton, W.A. (2020), Personal Communication.
Miazga, G.S. and Kennedy, D.J.L. (1989), “Behavior of Fillet Tide, R.H.R. (2020), Personal Communication.
Welds as a Function of the Angle of Loading,” Canadian Tide, R.H.R. (1980), “Eccentrically Loaded Weld Groups—
Journal of Civil Engineering, Vol. 16, No. 4, pp. 583–599. AISC Design Tables,” Engineering Journal, AISC,
Neis, V.V. (1985), “New Constitutive Law for Equal Leg Vol. 17, No. 4, pp. 90–95.
Fillet Welds,” Journal of Structural Engineering, Vol. 111,
No. 8.

ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 51

52 / ENGINEERING JOURNAL / FIRST QUARTER / 2025
 ERRATA
Torsion of Rectangular Connection Elements
BO DOWSWELL
Vol. 56, No. 2, 2019

In Example 1 on page 79, the nominal torsional strength is corrected to:

2
( 0.6 )( 50 ksi )( 0.750 in.) (15 in.) ⎡ 15 in. ⎤
Tp = ⎢1+ ⎥
2 ⎣ ( 2.4 ) (10 in. ) ⎦
= 206 kip-in.

And the available torsional strengths are updated to:

LRFD ASD
ϕTp = ( 0.90 ) ( 206 kip-in.) Tp Ω = 206 kip-in. 1.67
= 185 kip-in. = 123 kips

On page 80, the interaction calculations are revised to:

For LRFD, interaction according to Equation 61 is:
4 ⎡ 0.59
⎛ 60.0 kip-in. ⎞ ⎤
2 2 1.7 1.7
⎛ 30 kips ⎞ ⎛ 42.9 kip-in. ⎞ ⎛ 60.3 kips ⎞ ⎛ 600 kip-in. ⎞
⎜ ⎟ ⎜+ ⎟ ⎜+ ⎟ + ⎢ ⎜ ⎟ + ⎜ ⎟ ⎥ (61)
⎝ 507 kips ⎠ ⎝ 185 kip-in. ⎠ ⎝ 338 kips ⎠ ⎢⎣⎝ 1,900 kip-in. ⎠ ⎝ 94.5 kip-in. ⎠ ⎥⎦
= 0.800 <1.0 o.k.

For ASD, interaction according to Equation 61 is:

0.59
⎛ 20 kips ⎞ ⎛ 28.6 kip-in. ⎞ ⎛ 40.2 kips ⎞ ⎡ ⎛ 400 kip-in. ⎞ ⎛ 40.0 kip-in. ⎞ ⎤
2 2 4 1.7 1.7

⎜ +
⎟ ⎜ ⎟ ⎜+ ⎟ + ⎢ ⎜ ⎟ + ⎜ ⎟ ⎥ (61)
⎝ 337 kips ⎠ ⎝ 123 kip-in. ⎠ ⎝ 225 kips ⎠ ⎢⎣ ⎝ 1.260 kip-in. ⎠ ⎝ 62.9 kip-in. ⎠ ⎥⎦
= 0.802 < 1.0 o.k.

ISSN 2997-4720 ENGINEERING JOURNAL / FIRST QUARTER / 2025 / 53

54 / ENGINEERING JOURNAL / FIRST QUARTER / 2025

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 matthew@[Link].
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.
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 [Link]/ej.
Smarter. Stronger. Steel.
American Institute of Steel Construction
130 E Randolph St, Ste 2000, Chicago, IL 60601
312.670.2400 | [Link]/ej