Earls, 1999

Journal of Constructional Steel Research 49 (1999) 1–24
On the inelastic failure of high strength steel I-

shaped beams
C.J. Earls*
Department of Civil and Mechanical Engineering, United States Military Academy, West Point, New
York, USA
Received 10 November 1997; received in revised form 14 January 1998; accepted 11 March 1998
Abstract
Compactness and bracing provisions for the design of steel beams are formulated so as to
ensure that the resulting beam exhibits adequate structural ductility. The specification of such
compactness and bracing requirements involve assumptions about the constitutive nature of
the structural steel being used. Such material response assumptions are valid in designs involv-
ing most structural steel grades. However, it appears from the current research that these same
assumptions are not valid when used to predict the ductility of wide flange beams made from
the high performance steel grade HSLA80. HSLA80 wide flange beams subjected to moment
gradient loading display inelastic modes of failure, which do not lend themselves to a notional
de-coupling of so-called local buckling and lateral-torsional buckling phenomena. Rather, the
inelastic modes of failure of the HSLA80 beams tested herein display two distinct inelastic
buckling patterns at failure, both of which exhibit localized and global buckling components.
The structural ductility of the beams is very much dependent upon which of the two mode
shapes govern at failure. Cross-sectional proportions, bracing configuration, and geometric
imperfections all play a role in influencing which mode governs in the beam at failure. Cur-
rently held views as to the impact of cross-sectional compactness and bracing on structural
ductility may not apply to HSLA80 beams.  1998 Elsevier Science Ltd. All rights reserved.
Keywords: High strength steel; Finite elements; Ductility; Beams
* Corresponding author. Tel: + 1 914 938 5505; Fax: + 1 914 938 5522; E-mail: ic1506@trot-
ter.usma.edu
0143-974X/99/$—see front matter  1998 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 3 - 9 7 4 X ( 9 8 ) 0 0 2 0 4 - 1
2 C. Earls / Journal of Constructional Steel Research 49 (1999) 1–24
Notation
bf Flange width
d Depth of cross-section
d1 Distance from centerline of local buckle wave to mid-span stiffener
d2 Distance from centerline of local buckle wave to mid-span stiffener
opposite d1
Fy Steel yield stress
h Depth of cross-section between flanges, h = d − 2tf
L Span length of simply supported beam
My Moment causing the extreme cross-sectional fiber to yield
Mp Full plastic capacity of cross-section
Py Plastic resultant force in compression flange, Py = bf(tf)Fy
R Rotation capacity of cross-section
tf Flange thickness
tw Web thickness
␪y Cross-sectional rotation resulting in the yielding of the extreme fiber
1. Introduction
In the past, assumptions concerning the mechanical properties of structural steel

were viewed as being somewhat canonical in nature. The historical norm in structural
steel design has been a mild carbon steel which, when tested uniaxially, possesses
yield stresses between 220 and 450 MPa and exhibits a well-defined yield plateau
followed by a region of strain-hardening, ultimately ending in necking and rupture.
Recently, interest in the class of high strength low alloy (HSLA) steels has emerged
within the discipline of structural engineering, partially as result of the successful
employment of this type of steel in advanced applications within naval architecture.
The HSLA steels are appealing to the designer predominantly due to the high
strength-to-weight ratio of this material. However, good weldability and adequate
ductility further enhance this appeal. The material properties of the HSLA steels
tend to be dramatically different from mild carbon steel in several key areas of the
uniaxial stress–strain relationship as can be seen in Fig. 1. This figure displays a
schematical representation (based on the tests of Sooi et al. [1]) of the uniaxial stress–
strain response of HSLA80 steel superimposed over that of A36 steel. HSLA80 has
neither a well-defined yield plateau nor a substantial strain-hardening modulus as
compared with A36 steel. In general, HSLA80 steel is also somewhat less ductile
than A36 steel.
It is these characteristic differences in the mechanical properties of HSLA steels
which are at issue in light of the assumptions that the current AISC load and resist-
ance factor design (LRFD) specification [2] makes in its prediction of the ultimate
response of structural steel members. It is further noted that it may not be possible
to make strict generalizations as to the constitutive nature of these new HSLA steels
C. Earls / Journal of Constructional Steel Research 49 (1999) 1–24 3
Fig. 1. Uniaxial stress–strain response of A36 and HSLA80 steels.
due to the fact that, unlike carbon steels, HSLA steels are sold on the basis of
minimum mechanical properties, with the specific alloy content left to the discretion
of the steel producer [3]. Variations in the chemical composition of steel may pro-
foundly impact on the very mechanical properties which are deemed to be crucial
in creating a favorable overall structural response at the component and system level
within the structure.
Predicting the ultimate flexural response of wide flange beams manufactured from
HSLA80 plate has been one area of recent interest. Research along these lines has
been predominantly focused on demonstrating the ability of HSLA80 beams to resist
flexural loading in a ductile manner as quantified by rotation capacity [4–6]. This
research activity has shown, both experimentally and analytically, that wide flanges
produced from the 80 grade of HSLA steel can indeed exhibit a flexural rotation
capacity of three as required by the AISC LRFD specification [2]. This rotation
capacity of three is assumed to be adequate for accommodating moment redistri-
bution within the structural system to develop the controlling plastic collapse mech-
anism at the system-wide level [7].
The present study is restricted to the numerical testing of HSLA80 wide flange
beams subjected to a moment gradient loading. Nonlinear finite element modelling
is the vehicle by which the numerical testing is accomplished. The nonlinear finite
element modelling techniques implemented in this research are consistent with exper-
imentally verified techniques used in earlier studies by the author [4,8,9]. From the
numerical test results, it has become apparent that the notions of local buckling of
the constituent cross-sectional plate elements, and the global lateral torsional mode,
are so closely inter-related as to render them indivisible from each other in their
manifestations within the context of HSLA80 beams under moment gradient. Two
distinct modes, each possessing definite attributes of both local and global buckling,
are identified and studied. The two modes differ from each other in the manifestation
of their mode shapes, which appear to be directly related to the ability of the beam
to develop adequate ductility in the form of plastic hinge rotation capacity. Further-
more, the sensitivity of the mode shapes to inevitable imperfections in cross-sectional
dimensions are studied. Bracing stiffness requirements are also investigated, as is
the impact of a new bracing scheme for use with these HSLA80 wide flange beams.
2. Overview of modelling techniques
The commercially available multipurpose finite element software package abaqus

[10] is used for all of the numerical studies reported herein. Both geometric and
material nonlinearities are considered in all of the finite element models. The sub-
sequent nature of the nonlinearities are severe in that they are associated with inelas-
tic global and local buckling phenomena. Hence, a very refined model must be used
in the incremental solution of these types of problems. As a result, all analyses are
carried out in parallel using two processors of a CRAY C90 supercomputer. The
supercomputer is appropriate for a study such as this due to the large number of
complete iterative analyses which must be conducted repeatedly during a parametric
study. abaqus is identified as the software package of choice in this research because
of its exceptional robustness in areas associated with fully nonlinear analysis. abaqus
is widely used across many disciplines with well-documented success in modelling
of highly nonlinear phenomena. Similarly, abaqus has been shown to perform quite
well in earlier studies conducted by the author [4,8,9]. These earlier studies are of
a similar nature to this one in that inelastic buckling phenomena are investigated
through the use of nonlinear shell elements. The favorable characterization of aba-
qus’ performance in these earlier studies is based upon good agreement between the
results obtained from abaqus and those obtained from equivalent physical experi-
mental tests.
HSLA80 wide flange beams are studied in the current research. The models of
these beams are constructed from a mesh of nine noded shell finite elements. The
planes of the mesh surfaces correspond with the middle surfaces of the constituent
wide flange cross-sectional plate components (Fig. 2). The mesh density used
throughout this study is considered to be adequate as evidenced by a mesh conver-
gence study which was carried out on several representative test cases.
The S9R5 nonlinear shell finite element from the abaqus element library is
employed in this research. The S9R5 shell element is shear deformable and, sub-
sequently, both reduced integration and discrete Kirchhoff theory are employed to
improve the overall thin-shell behavior of the S9R5 element. A 2 × 2 Gauss quadra-
ture is used in-plane and the discrete Kirchhoff condition is imposed at a finite num-
ber of points on the shell reference surface by way of a penalty function [10]. The
strain measures used in the S9R5 formulation are an approximation to the Koiter–
Fig. 2. Finite element mesh used to model HSLA80 wide flange beams.
Sanders theory [11]. The transverse shear strains are measured as the changes in the
projections of the shell normal, at a point on the Gaussian shell reference surface,
onto tangents to the shell reference surface. It should be noted that these transverse
shears are always treated elastically. The through-thickness integrations are
accomplished by a Simpson’s rule of user chosen order. The abaqus default order
is five, but the author has achieved improved convergence characteristics by increas-
ing this parameter to seven.
A uniaxial representation of the constitutive law used in this study appears in Fig.
3 as a plot of true stress vs. logarithmic strain. This piece-wise linear model has a
yield stress taken as 586 Mpa with fu/fy = 1.14 and ⑀u/⑀y = 17. Similarly, ⑀y/⑀st = 1
hence no well-defined yield plateau is modelled.
3. Geometries of specimens
All numerical tests reported herein are limited to simply supported beams having
a wide flange cross-sectional shape and loaded with a moment gradient which varies
linearly along the longitudinal axis in a fashion that is consistent with the structural
geometry displayed in Fig. 4. In the studies reported here, L/d varies from 13.8 to
22, where L is the total span length, and d is the cross-sectional depth. The variation
of this parameter is only an artifact of the change in specimen length since the web
depth is held constant across all tests. It is not clear at this time what impact the
moment gradient severity has on the rotation capacity of HSLA80 wide flanges, but
Fig. 3. Uniaxial stress–strain response of HSLA80 steel used in models.
Fig. 4. Schematic of beam geometry and loading condition.

Barth and White [12] conclude that the impact is negligible for the slender web,
grade 50 steel, bridge girders which they analysed. Barth and White [12] further
mention that the effect of L/d may not be negligible on beams which have a stockier
web than those which they tested. The wide flange beams analysed in the present
study have compact webs with the h/tw ratio being held constant at 25.
Imperfection sensitivity in the development of the inelastic mode shape is also
addressed in this study. Associated with the imperfection sensitivity of the mode
shape, is the effect which these imperfections have on rotation capacity and ultimate
moment capacity. The imperfections used in the present work are limited to those
types associated with cross-sectional geometry. Fig. 5 displays the two imperfection
cases used in this work. These imperfections are based on standard mill practice as
outlined in the LRFD Manual [2], and are maintained as constant along the beam
length. Imperfections such as a sinusoidal sweep in the longitudinal direction are
not directly incorporated into the models. However, the cross-sectional imperfections
(especially imperfection B) indirectly create a tendency for the beam to move out
of the plane of loading between bracing points.
Bracing against this same out-of-plane motion of the HSLA80 wide flange beams
is required at the mid-span point of concentrated force application as well as at
support points. This is accomplished in the finite element modelling by restraining
all mid-span nodal points of the cross-section from translating in the out-of-plane
direction. Over the supports, only the bottom flange nodes are restrained against such
out-of-plane translation.
Full-depth stiffeners are incorporated on either side of the web in the model at
the support locations and under the mid-span load point. The stiffeners are tied into
the nodes of the top and bottom flanges of the cross-section. The thickness of the
stiffeners is held constant throughout the studies at 0.2 in. The stiffeners are given
an elastic modulus two orders of magnitude greater than that of steel so as to approxi-
mate perfect rigidity. No plastification of the stiffeners is permitted during the analy-
ses. While it is believed that nearly rigid stiffeners such as these influence the inelas-
Fig. 5. Cross-sectional imperfections used in models.

tic mode shapes of the beams at failure, they are used in the modelling nonetheless
so as to reduce the number of parameters that should be varied. Further study of the
influences of stiffener geometry on the overall beam failure mode shape is needed.
4. Preliminary results
Preliminary finite element studies of HSAL80 beams subjected to a moment gradi-

ent loading were conducted so as to develop a notional understanding of potential
underlying mechanisms associated with the ultimate response of such beams. The
results of these early studies were seen to run counter to expectations concerning the
influence that cross-sectional proportioning has on structural ductility as quantified by
plastic hinge rotation capacity. The definition of rotation capacity is consistent with
that given by ASCE [13] in which R = {(␪u/␪p) − 1} where ␪u is the rotation when
the moment capacity drops below Mp on the unloading branch of the M − ␪ plot
and ␪p is the rotation at which the full plastic capacity is first achieved. The influence
of flange compactness on overall rotation capacity is addressed by Table 1. The
results presented in this table seem to contradict current practical notions regarding
the role of flange compactness in wide flange ultimate response. It is seen from Table
1 that an increase in flange slenderness increases overall rotation capacity for a bf/2tf
increasing from 3 to 6. However a further increase in bf/2tf from 6 to 7 results in a
significant decrease in overall section rotation capacity. Substantial differences in
Table 1
Summary of the influence of flange compactness on HSLA80 wide flange response
Steel type b/2tf h/tw L/ry R Notes
HSLA-80 3 25 65 3.0 Failure mode: Combined LTB

and LB (flange)
Highly unsymmetric mode
shape with flange local
buckling occurring away from
stiffener. Highly unsymmetric
hinge region.
HSLA-80 6 25 65 3.4 Failure mode: LB (flange and
web) and slight LTB
Flange local buckling occurs
close to stiffener in one half-
span only. Slightly
unsymmetric hinge region.
web) and LTB
hinge region.
Table 2
Summary of the influence of bracing on HSLA80 wide flange response

and LB (flange)
hinge region.
HSLA-80 3 25 55 >>3 Rotation capacity greater than
3; Failure mode: Slight LTB
and LB. Symmetric flange
local buckling. Symmetric
hinge region.
HSLA-80 3 25 75 4.5 Failure mode: LTB and very
localized LB Symmetric
flange local buckling and
LTB.
Symmetric hinge region.
the inelastic buckling mode of all three cases presented in Table 1 are observed. It
appears that perhaps certain inelastic buckling modes are more favorable than others
as quantified by overall rotation capacity.
The influence of brace spacing on beam rotation capacity is also addressed in this
preliminary study. Tables 2 and 3 provide a summary of results relevant to a dis-
cussion of the impact that brace spacing has on such beams. The results outlined in
Table 3
Summary of the influence of bracing on HSLA80 wide flange response

web) and slight LTB
close to stiffener in one span
only. Slightly unsymmetric
hinge region.
HSLA-80 6 25 55 >>3 Failure mode: LB and LTB
Symmetric flange local
buckling. Symmetric hinge
region.
HSLA-80 6 25 75 3.3 Failure mode: LB and LTB
Symmetric flange local
buckling. Symmetric hinge
region.
Table 2 contradict practical notions concerning the role of beam slenderness on the
rotation capacity of a flexural element governed by lateral torsional buckling. It is
seen from these results that both increasing and decreasing the unbraced length can
lead to substantial improvements in the rotation capacity exhibited by a beam.
Table 3 presents results from analyses similar to those in Table 2, but for beams
having a larger flange slenderness. It appears from these results that as beam slender-
ness increases, the overall beam rotation capacity is diminished.
Tables 4 and 5 are concerned with the impact of web slenderness on the overall
response of an HSLA80 wide flange. Yet again we have a contradiction with the
accepted notions of what is typical wide flange behavior. Results presented in Table
4 provide evidence that increases in web slenderness may lead to improved rotational
behavior of the cross-section. The results of Table 5 support the conventional view
for the impact of web slenderness on beam rotation capacity. It is pointed out that
a certain degree of bracing influence is present in the Table 5 results since unbraced
lengths were not held constant.
An explanation of this unanticipated beam behavior is sought. Such an explanation
appears to be related to the manifestation of the inelastic mode shapes exhibited by
these beams. After careful examination of results obtained during graphical post-
processing, a trend emerges as to what constitutes a more favorable mode of HSLA80
inelastic beam buckling. This graphical examination of the analysis data consists of
noting the overall geometry associated with the buckling, as well as observing the
boundary in plastification, which defines the shape of the plastic hinge region at
mid-span. Furthermore, observations are made to identify possible plastic mech-
anisms that may form within the beam during large plastic deformations. Obser-
Table 4
Summary of web compactness influence on HSLA80 wide flange response

and LB (flange)
hinge region.
HSLA-80 3 33 65 >3 Rotation capacity clearly
greater than 3; Failure mode:
Slight LTB and LB
Flange local buckling close to
stiffener in one span only.
Slightly unsymmetric hinge
region.
HSLA-80 3 70 65 1.3 Failure mode: LB (web) and
LTB
Tension field buckling.
Table 5
Summary of web compactness influence on HSLA80 wide flange response (with slight bracing influence)

web) and slight LTB
close to stiffener in one half-
span only. Slightly
unsymmetric hinge region.
web) and some LTB
Flange local buckling with
highly unsymmetrical mode
shape. Highly unsymmetrical
hinge region.
web) and some LTB
HSLA-80 6 53 58 ?? Failure mode: LB (flange and
web) Very slow convergence
resulted in an incomplete
analysis.
vations along these lines seem to provide evidence that a ‘more favorable’ inelastic
buckling mode involves concentrated localized buckling close to the mid-span stiff-
ener with little or no out-of-plane motion of the adjacent beam segments. An example
of this mode is displayed in an overall sense by Fig. 6. Fig. 7 provides a closer view
Fig. 6. Mode 1 inelastic buckling.

of the hinge region from Fig. 6. Both Figs. 6 and 7 are associated with the response
of an HSLA80 wide flange beam with a bf/2tf = 6, an h/tw = 25, and an L/ry = 65.
This response, however, is representative of the responses of other member geo-
metries which displayed this particular inelastic buckling mode.
From similar graphical post-processing of model data, it appears that the less fav-
orable modes are encountered when the instability is triggered at the periphery of
the hinge region. This appears to lead to a subsequent expansion and/or shifting
of the hinge’s zone of plasticity. Oftentimes this same zone of plasticity becomes
asymmetrical across the mid-span stiffener. This type of ‘unfavorable’ mode shape
is displayed in Figs. 8 and 9.
5. Manifestations of the inelastic modes
Based on the foregoing discussion concerning the preliminary finite element stud-
ies associated with this research, a subsequent comprehensive modelling program is
carried out. From this work it appears that the ultimate response of an HSLA80 wide
flange cross-section is governed by one of two distinct possible inelastic buckling
modes. These two modes will be referred to, respectively, as mode 1 and mode 2.
Figs. 6 and 7 display a typical mode 1 wide flange inelastic buckled shape, while
Figs. 8 and 9 show the mode 2 inelastic failure.
Mode 1 is characterized by a local instability of the flange, either with or without
substantial web participation, which occurs in close proximity to the mid-span stiff-
ener. The buckling wave in the flange may be either symmetric or anti-symmetric

about the web. Only very moderate out-of-plane deflections between brace points
along the beam are observed in mode 1 failures. Similarly, the plastic hinge region
is well defined and localized to the region of the beam immediately adjacent to the
mid-span stiffener. The localized flange, or flange–web buckling component of the
overall mode 1 response is most often restricted to one half-span of the beam. Thus,
strictly speaking mode 1 failure is characterized as an asymmetrical inelastic mode
about the mid-span stiffener, but it is noted that the asymmetry of mode 1 is of a
much less severe nature than that of mode 2.
Mode 2 is characterized as a highly asymmetrical inelastic mode shape where
local and global buckling are highly coupled. As can be seen in Fig. 8, the flange
buckling components, or flange–web buckling components, occur at a substantial
distance from the mid-span. This distance from the stiffener to the center of the
flange buckling wave is in general different for each half-span, but on average this
distance is roughly equal to d/2. Similarly, the degree to which the flange–web buck-
ling component of mode 2 manifests itself varies significantly between the half-
spans. Generally speaking, substantial out-of-plane deflections between brace points
occurs in the mode 2 failure. This out-of-plane deflection may be either symmetric
(M-shaped) or anti-symmetric (S-shaped) about the plane of the mid-span stiffener.
Accompanying the geometric asymmetry in the mode 2 manifestation is a corre-
sponding asymmetry in the development of the plastic hinge within the beam. The
mode 2 ‘plastic hinge’ is more aptly described as a ‘zone of plastification’, thereby
not implying the usual connotation of a tightly formed concentrated zone of plas-
ticity. On the contrary, the zone of plasticity in mode 2 is very ill defined and quite
distributed in nature. Another characteristic feature of the mode 2 inelastic mode
shape centers on the formation of a mechanism in the compression flange of the
HSLA80 beam. An example of this type of compression flange mechanism can be
seen in Fig. 10, which displays a typical top view of the compression side of an
HSLA80 wide flange beam. From this figure, it can be seen that the compression
flange behaves somewhat like a three-bar-linkage, the kinematics of which are driven
by the location of the mid-span stiffener and the linkage articulations. These articu-
lations coincide with the locations of the flange buckling component (or flange–web
buckling component) of the overall mode 2 manifestation. Beyond the clear and
pronounced geometric differences between the inelastic mode shapes of mode 1 and
mode 2, there are other more quantifiable differences in response.
Examples of measurable structural response parameters that are profoundly influ-
enced by the governing inelastic mode shape are the rotation capacity and ultimate
moment capacity. In cases where changes from mode 1 to mode 2 are triggered by
imperfections, a reduction in rotation capacities of up to 40% is possible. Similarly,
reductions in ultimate moment capacity resulting from imperfection-induced changes
from mode 1 to mode 2 lead to approximately an 8% overall reduction in capacity.
The apparent imperfection-induced shift in governing inelastic mode shape also
impacts profoundly on the shape of the moment rotation curve.
Fig. 10. Compression flange mechanism of mode 2 failure.
6. Imperfection sensitivity
Evidence from the current study of HSLA80 wide flange beams under moment
gradient suggests that a sensitivity to cross-sectional imperfections exists. The cross-
sectional imperfections used in this modelling are of the order of those that are
deemed to be of an acceptable level by the literature. Such cross-sectional imperfec-
tions are based on the Standard Mill Practice as outlined in the LRFD Manual [2]
and are displayed in Fig. 5. It is noted here that for a given beam length and cross-
sectional geometry, a variation in the governing inelastic mode can be obtained by
changing the cross-sectional imperfections only. Changes from mode 1 to mode 2,
triggered by such imperfections, impact heavily on rotation capacity and ultimate
moment capacity. This type of imperfection-induced capacity reduction is observed
in three of the beam geometries addressed in this study and presented in Tables 6–8.
It is seen in Table 6 that for the case of an HSLA80 W-shaped beam with a bf/2tf
= 6, h/tw = 25, and L/ry = 65, a 40% reduction in rotation capacity is observed when
Table 6
HSLA80 WF, d = 10 in, bf /2tf = 6, h/tw = 25, L/ry = 65
Geometry Rotation capacity Mu/My Mp/My Observations
Perfect 3.40 1.43 1.150 Mode 1

Imperfection A 3.54 1.43 1.150 Mode 1
Imperfection B 2.15 1.35 1.150 Mode 2
Table 7
Perfect 4.12 1.57 1.156 Mode 1

Table 8
Perfect 3.94 1.47 1.150 Mode 1

the transition from mode 1 to mode 2 occurs as a result of imperfection B. This

substantial reduction in rotation capacity is accompanied by a 6% reduction in ulti-
mate moment capacity. The results in Table 7 show only an 8% reduction in rotation
capacity with a corresponding 8% reduction in ultimate moment capacity, both as a
result of the imperfection-induced transition from mode 1 to mode 2 in the case of
a W-shaped beam with a bf/2tf = 3, h/tw = 25, L/ry = 57. Table 8 indicates that the
33% reduction in rotation capacity is a result of the cross-sectional imperfections
causing a transition in governing inelastic mode shape from mode 1 to mode 2 in
a beam with a bf/2tf = 6, h/tw = 25, L/ry = 57.
The imperfection-induced transition from mode 1 to mode 2 results not only in
reductions of ultimate moment and rotation capacities outlined in Tables 6–8, but
also a very significant difference in the appearances of the moment–rotation plots
themselves. The moment–rotation plots associated with the results presented in Table
6 are displayed in Fig. 11. The moment–rotation plots accompanying Tables 7 and
8 are displayed by Figs. 12 and 13, respectively. In each of the moment–rotation
plots it is seen that when the transition from mode 1 to mode 2 takes place, it is
accompanied by a shifting of the inelastic portion of the moment–rotation response.
In mode 1, the inelastic portion of the moment–rotation curve exhibits a fairly con-
stant plateau region corresponding to the ultimate moment capacity. This constant
moment plateau is then followed by a well-defined ‘knee’ region where the nominal
moment capacity diminishes quickly. Such a mode 1 moment–rotation response is
contrary to the moment–rotation response that is characteristic of mode 2 failures.
It becomes clear from the examination of Figs. 11–13 that the ultimate moment
capacity associated with a mode 2 beam failure is attained at a relatively small
rotation capacity. This peak in the moment capacity is then followed by a gradual
and constant decrease in nominal moment capacity as beam cross-sectional rotation
Fig. 11. Moment–rotation response associated with imperfection sensitivity study.
increases. There is no distinct ‘knee’ in the descending branch of the moment–

rotation response for mode 2 as there is in mode 1.
7. Strength and stiffness of mid-span bracing
Bracing a beam against premature failure due to lateral-torsional-buckling is typi-

cally achieved by restraining the out-of-plane translation and twisting of the cross-
section at discreet points along the longitudinal axis of the beam. In the case of a
beam subjected to a moment gradient loading, the bracing point is often chosen as
the point of maximum moment. Once the bracing point location is known, strength
and stiffness requirements for the bracing member must be specified.
The strength of the bracing members used in this study are quantified as a percent-
age of the resultant force, which is present in the compression flange of the beam
at attainment of the full plastic cross-sectional capacity. In other words, percentages
of the total flange plastic capacity are given by the quantity Py defined below as
Py = bf(tf)Fy
where bf is the flange width, tf is the flange thickness, and Fy is the yield stress of
the particular grade of steel used. In this research, four different percentages of Py
are considered. These are 5%, 2%, 1%, and 0.5% of Py, respectively. The analyses
conducted with the bracing force corresponding to 0.5% of Py resulted in very slowly
converging finite element solutions. As a result of this slow convergence, the compu-
tational expense associated with the continued inclusion of this bracing stiffness in
the remainder of the study was deemed too high. Hence, the case of 0.5% Py is only
partially studied herein. It is felt, however, that a bracing force this low is somewhat
unrealistic in any case. Besides bracing strength, the other parameter at issue in this
portion of the study is bracing stiffness.
The bracing stiffness used in this research is obtained by considering a reasonable
out-of-plane deflection at the mid-span bracing point of the beam. The value of L/500
is used as a rational assumption for an acceptable out-of-plane deflection as suggested
by Winter [14]. The bracing stiffness for each case is then obtained by taking the
particular fraction of the flange plastic resultant force Py and dividing it by the
deflection value of L/500. The subsequent stiffness value is then input into the
abaqus model as the spring stiffness associated with the SPRING1 finite element.
The SPRING1 elements are used to simulate the required bracing characteristics
which act at the mid-span of the beams considered. The SPRING1 elements are
present at both the tension and compression flanges of the stiffened mid-span region
of the beams as outlined in the section view given by Fig. 14.
The beam geometry selected for this bracing study possesses a bf/2tf ratio equal
to 6, an h/tw ratio equal to 25, and an L/ry ratio equal to 57. This geometry is chosen
since it was observed to exhibit an imperfection-induced change in inelastic mode
shape during earlier analyses. Subsequently, both perfect geometries and geometries
exhibiting imperfection A are studied. In the perfect geometry case, bracing stiff-
nesses of 5%, 2%, 1%, and 0.5% are considered. Only 5%, 2%, and 1% of Py are
considered in the case associated with imperfection A.
Fig. 15 displays the normalized moment–rotation response as influenced by brac-
ing force for the case of a beam with perfect geometry at a bf/2tf ratio equal to 6,
an h/tw ratio equal to 25, and an L/ry ratio equal to 57. It is noted that while only
the response of the case with a bracing stiffness of 5% Py is traced into unloading
below Mp due to slow convergence rates exhibited by the other two cases, these
other cases display very similar moment–rotation responses over their respective
cross-sectional rotations. What is meant here is that the normalized moment–rotation
responses of all the bracing cases associated with the perfect geometry essentially
overlap each other to the extent that each curve proceeds into the inelastic range.
Upon examining the deformed configurations of the beams, it appears that none of
the perfect geometry, variable bracing force, cases deviate from mode 1 inelastic
Fig. 14. Section view of mid-span stiffener and bracing configuration.

Fig. 15. Moment–rotation response associated with bracing study.
buckling. This observation is consistent with the shape of the moment–rotation plots,
which display the characteristic ‘knee’ region in the unloading portion of the
moment–rotation response indicative of mode 1 failures. Only the 5% bracing case
is traced into the unloading region below the load level Mp as a result of compu-
tational expense.
The moment–rotation response of a wide flange beam with a bf/2tf ratio equal to
6, an h/tw ratio equal to 25, and an L/ry ratio equal to 57 associated with imperfection
A is plotted in Fig. 16. In this case, an imperfection induced transition from mode
1 to mode 2 inelastic buckling is observed at a bracing force of 1%. Only mode 1
failure is present for the 5% and 2% cases. This transition of inelastic mode shape
is observed in the graphically post-processed results from these analyses. Similarly,
the behavior of the moment–rotation response supports this conclusion since the
moment–rotation ‘knee’ is present in the 5% and 2% cases, but not the 1% Py brac-
ing case.
From the foregoing it appears that mid-span bracing properties can have a very
significant effect on the inelastic mode shape manifestation, and subsequent rotation
capacity, within HSLA80 wide flange beams subjected to moment gradient loading.
In light of the role which the inelastic mode shape plays on the plastic deformation
capacity of these types of beams, bracing point location along the longitudinal axis
seems important. If it is possible to locate bracing points at points along the beam
longitudinal axis where the out-of-plane deflections associated with a mode 2 failure
are most likely to occur, then perhaps the bracing may become more effective in
increasing ultimate moment capacity and ductility.
Fig. 16. Moment–rotation response associated with bracing study.
8. Alternative bracing scheme
In this portion of the research, alternative bracing configurations are sought based
on the quantification of dimensions associated with mode 2 inelastic buckling mani-
festations. A series of measurements are made on the post-processed geometry of
the HSLA80 wide flange beams exhibiting mode 2 failure so that trends in the
geometry of the mode shape may be identified. The measurements consist of noting
the distance from the mid-span stiffener to the centerline of the localized buckling
wave present at the compression flange articulations of the mode 2 failures. The
results of this study are summarized in Table 9. In this table, the distances ‘d1’ and
‘d2’ are the respective distances to the centerline of the localized buckling mode as
measured on either side of the mid-span stiffener. These distances are normalized
with respect to the depth of the HSLA80 wide flange (d = 10.17 in). It is observed
from the results in Table 9 that the location of the compression flange articulations,
coinciding with the localized buckling of the mode 2 failures, are seldom of equal
distance from the mid-span stiffener. Hence, it may be concluded that a high degree
of asymmetry is present in the mode 2 failures. The ‘NA’ designation present in
Table 9 indicates that these cases experienced mode 1 failures.
Upon averaging the distances ‘d1’ and ‘d2’ in Table 9, a value of approximately
d/2 arises. It is this value of d/2 that is assumed to be the average distance on either
side of the mid-span stiffener where one might expect to find the localized buckling
of the compression flange associated with mode 2 failures. Hence, it is at these
locations where bracing is placed in the alternative bracing scheme carried out in
this portion of the research. As outlined in Fig. 17, the bracing is placed at the top
Table 9
Dimensions of inelastic mode shape manifestations
Proportions Geometry d1/d d2/d
bf/2tf = 3, h/tw = 25, L/ry = 65 Perfect 0.61 0.54

Imperfection A 0.61 0.46
Imperfection B 0.54 0.54
bf/2tf = 6, h/tw = 25, L/ry = 65 Perfect NA NA
Imperfection A NA NA
Fig. 17. Alternative bracing scheme for HSLA80 beams subjected to moment gradient loading.
of the compression web at a distance d/2 on either side of the mid-span stiffener.
By imposing this bracing configuration on the models, a profound impact is observed
on the inelastic deformation capacity of the HSLA80 beams tested. Table 10 shows
that for the three cases investigated here, the new bracing scheme eliminated the
occurrence of the mode 2 failure. Correspondingly, the ultimate moment capacity
and rotation capacity of the beams tested increased substantially.
9. Conclusions
HSLA80 wide flange beams subjected to a moment gradient loading may behave
differently at failure than similarly proportioned steel beams made from more tra-
ditional types of structural steel. The impacts that cross-sectional slenderness and
unbraced length have on the plastic deformation capacity of HSLA80 beams seem
to contradict conventional views about the influence of such parameters on plastic
deformation capacity. Investigations into the source of these apparent discrepancies
leads to the conclusion that the phenomena of ‘local buckling’ and ‘lateral-torsional’
buckling may not be satisfactorily de-coupled for the case of HSLA80 beams as it
can in the design of mild carbon steel beams. Similarly, two distinct inelastic flexural
buckling modes emerge as the dominant modes occurring at failure in HSLA80
beams under moment gradient. Mode 1 is characterized as exhibiting little out-of-
plane deformation accompanying localized buckling, which occurs close to the mid-
span stiffener. Mode 1 results in favorable inelastic flexural deformation capacity.
Mode 2, conversely, exhibits a great deal of out-of-plane deflection associated with
the regions near localized buckling in the compression flange. Mode 2 is highly
asymmetric in nature and results in relatively poor flexural deformation capacity in
the inelastic range.
It appears that various cross-sectional imperfections may trigger a transition from
one mode to another. Such imperfections need only be of the order of those which
are judged to be within the practical tolerances of steel mills. Such imperfection-
induced transitions may be minimized, or even eliminated, through the use of bracing
along the beam’s longitudinal axis.
Table 10
Influence of bracing geometry on plastic deformation capacity
Proportions Geometry Bracing at mid-span Bracing at d/2 from mid-span
Mu/My R Mode Mu/My R Mode
bf/2tf = 6, h/tw = 25, Perfect 1.35 2.15 Mode 2 1.40 3.11 Mode 1
L/ry = 65
bf/2tf = 6, h/tw = 25, Imperfection A 1.36 2.66 Mode 2 1.43 3.42 Mode 1
L/ry = 57
bf/2tf = 6, h/tw = 25, Imperfection B 1.36 2.74 Mode 2 1.42 3.25 Mode 1
L/ry = 57
The stiffness of such bracing plays a significant role in the nature of the inelastic
mode shape manifested at failure in HSLA80 beams. If a mid-span brace point is
used in conjunction with a moment gradient loading, it appears that a bracing force
of at least 2% of the compression flange plastic resultant force should be used with
an out-of-plane deflection limit of L/500 in the design of bracing members such that
the required brace stiffness is 10Py/L.
An alternative to the conventional mid-span bracing scheme is obtained as a result
of the current research. It appears that two bracing members attached to the top of
the compression web at a distance of d/2 on either side of the mid-span stiffener
constitutes a more efficient bracing alternative to the single mid-span brace case.
Such a bracing scheme as this d/2 alternative actually inhibits the formation of the
mode 2 failure. Thus, a superior rotation capacity is obtained from a beam geometry
which would otherwise provide poor ductility as a result of the mode 2 failure.
References
[1] Sooi TK, Green PS, Sause R, Ricles JM. Stress–strain properties of high-performance steel and
the implications for civil-structure design. Proceedings of the International Symposium on High-
Performance Steel for Structural Applications, Cleveland (Ohio, USA): 1995:35–43.
[2] American Institute of Steel Construction, Load and Resistance Factor Design, 2nd edn. Chicago (IL,
USA): American Institute of Steel Construction Inc., 1994.
[3] Davis JR. ASM specialty handbook, carbon and alloy steels. Materials Park (OH, USA): ASM
International, 1996:29.
[4] Galambos TV, Hajjar JF, Earls CJ, Gross JL. Required properties of high-performance steels. Techni-
cal report NISTR 6004. Gaithersburg (MD, USA): United States Department of Commerce, National
Institute of Standards and Technology, 1997.
[5] Ricles JM, Sause R, Green PS. High-strength steel: implications of geometric and material character-
istics on rotation capacity. Engineering Structures 1997; (to appear).
[6] Green PS, Sooi TK, Ricles JM, Sause R, Kelly T. Inelastic behavior of structural members fabricated
from high performance steel. Proceedings of Structural Stability Research Council, Bethlehem
(Pennsylvania, USA): Lehigh University, 1994:435–55.
[7] Yura JA, Galambos TV, Ravindra MK. The bending resistance of steel beams. Journal of the Struc-
tural Division ASCE 1978;104(ST9):1355–70.
[8] Earls CJ, Galambos TV. Design recommendations for single angle flexural members. Journal of
Constructional Steel Research 1997;43(1):65–85.
[9] Earls CJ. On the use of finite element analysis techniques to model structural steel angle response.
Ph.D. Dissertation, Minneapolis (MN, USA): University of Minnesota, 1995.
[10] abaqus Theory Manual, Pawtucket (RI, USA): Hibbitt, Karlsson and Sorensen, Inc., 1994.
[11] Budiansky, B., Sanders, J.L., On the ‘best’ first-order linear shell theory. In: Progress in Applied
Mechanics—the Prager Anniversary Volume. New York (NY): The Macmillan Company, 1963.
[12] Barth KE, White DW. Finite element evaluation of pier moment–rotation characteristics in continu-
ous-span steel I-girders. Engineering Structures 1997; (to appear).
[13] American Society of Civil Engineers, Plastic design in steel, a guide and commentary. New York
(NY): American Society of Civil Engineers, 1971:80.
[14] Winter G. Lateral bracing of columns and beams. Transactions ASCE 1960;125:807–45.

Earls, 1999

Uploaded by

Copyright:

Available Formats

Earls, 1999

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Earls, 1999

Uploaded by

Copyright:

Available Formats

Journal of Constructional Steel Research 49 (1999) 1–24

On the inelastic failure of high strength steel I-

Keywords: High strength steel; Finite elements; Ductility; Beams

In the past, assumptions concerning the mechanical properties of structural steel

Fig. 1. Uniaxial stress–strain response of A36 and HSLA80 steels.

2. Overview of modelling techniques

The commercially available multipurpose finite element software package abaqus

Fig. 3. Uniaxial stress–strain response of HSLA80 steel used in models.

Fig. 4. Schematic of beam geometry and loading condition.

Fig. 5. Cross-sectional imperfections used in models.

Preliminary finite element studies of HSAL80 beams subjected to a moment gradi-

Steel type b/2tf h/tw L/ry R Notes

HSLA-80 3 25 65 3.0 Failure mode: Combined LTB

Steel type b/2tf h/tw L/ry R Notes

HSLA-80 3 25 65 3.0 Failure mode: Combined LTB

Steel type b/2tf h/tw L/ry R Notes

HSLA-80 6 25 65 3.4 Failure mode: LB (flange and

Steel type b/2tf h/tw L/ry R Notes

HSLA-80 3 25 65 3.0 Failure mode: Combined LTB

Steel type b/2tf h/tw L/ry R Notes

HSLA-80 6 25 65 3.4 Failure mode: LB (flange and

Fig. 6. Mode 1 inelastic buckling.

Fig. 7. Mode 1 inelastic buckling.

5. Manifestations of the inelastic modes

Fig. 8. Mode 2 inelastic buckling.

Fig. 9. Mode 2 inelastic buckling.

Fig. 10. Compression flange mechanism of mode 2 failure.

Geometry Rotation capacity Mu/My Mp/My Observations

Perfect 3.40 1.43 1.150 Mode 1

Geometry Rotation capacity Mu/My Mp/My Observations

Perfect 4.12 1.57 1.156 Mode 1

Geometry Rotation capacity Mu/My Mp/My Observations

Perfect 3.94 1.47 1.150 Mode 1

the transition from mode 1 to mode 2 occurs as a result of imperfection B. This

Fig. 11. Moment–rotation response associated with imperfection sensitivity study.

Fig. 12. Moment–rotation response associated with imperfection sensitivity study.

increases. There is no distinct ‘knee’ in the descending branch of the moment–

Fig. 13. Moment–rotation response associated with imperfection sensitivity study.

7. Strength and stiffness of mid-span bracing

Bracing a beam against premature failure due to lateral-torsional-buckling is typi-

Fig. 14. Section view of mid-span stiffener and bracing configuration.

Fig. 15. Moment–rotation response associated with bracing study.

Fig. 16. Moment–rotation response associated with bracing study.

8. Alternative bracing scheme

Proportions Geometry d1/d d2/d

bf/2tf = 3, h/tw = 25, L/ry = 65 Perfect 0.61 0.54

Proportions Geometry Bracing at mid-span Bracing at d/2 from mid-span

Mu/My R Mode Mu/My R Mode

You might also like