You are on page 1of 89

Lecture notes for CEE 474: Introduction

to the behavior of steel structures


C. J. Earls∗
August 14, 2007

1 Steel as a material
1.1 Steel making
The primary constituent in steel (usually more than 97 percent of its content)
is Iron. Iron is naturally occurring in the earth’s crust in combined form: most
abundantly linked with oxygen, but sulfur linkages occur as well. In the best
case, high grade ores contain as much as 50 percent iron; with the balance be-
ing impurities of various types. The primary impurities are oxygen and sulfur;
which may be removed in a process known as smelting (in which a blast fur-
nace is used to introduce coke and limestone into a molten bath made from
the heated ore.) The coke (composed almost entirely of carbon) has a greater
affinity for binding with the oxygen and the sulfur; as compared with iron
(F e2 O3 + 3CO → 2F e + 3CO2 ). As a result, the iron is liberated from the
ore during smelting due to the presence of coke. In addition, the heat from the
blast furnace converts the limestone to lime; the latter being useful in capturing
the remaining impurities in the form of a less dense mixture that floats to the
surface of the molten iron bath where it is skimmed off. A product known as
“pig iron” is the result of the smelting process. Pig iron contains about 4 to
5 percent carbon; a percentage that is far too high for steel. As we will see
later, carbon adds strength to steel, but too much of it leads to brittleness and
problems with welding.

Currently in the US, the electric arc furnace is the most commonly used next
step in the steel making process. The electric arc furnace consists of a large cru-
cible that is refractory lined (e.g. thermal shock resistant ceramic) and water
cooled with a retractable roof through which three large carbon electrodes pass.
The charge that is placed in the crucible is composed primarily (approximately
97 percent) of recycled and graded steel scrap; with the balance being made up
∗ Associate Professor, School of Civil and Environmental Engineering, Cornell University,

Ithaca, New York

1
of pig iron. Once the mixture is placed in the crucible, the lid is closed, and
a large three-phase alternating current is introduced into the electrodes; the
ensuing arcs heat and melt the metal. An oxygen “lance” is oftentimes used to
inject pure oxygen into the mixture to accelerate melting by combustion, and
also to remove excess carbon from the steel. Slag formers, such as burnt lime
and magnesium oxide, are introduced into the mix to trap oxide impurities.
These float to the top where they are skimmed from the surface of the molten
steel.

1.2 Rolling and finishing


The resulting steel from the electric arc furnace is either cast into blooms (rect-
angular parallelepiped) or billets (cube). These castings are sent to a rolling mill
where they are subjected to a staged rolling process used to create structural
sections or rolled plate.

1.2.1 Plate
Rolled plate steel is referred alternately as flat bars and plate. If the width of the
piece is less than eight inches, then the piece is designated as a flat; otherwise it
is a plate. In the case of a flat, the width dimension typically precedes the the
thickness in the dimensional designation (e.g. 6” x 1”); whereas the opposite is
true for a plate.

1.2.2 Rolled sections


Detailed geometric descriptions, and tabulated properties, can be found in Part
I of the 13th edition of the steel manual. Commonly used structural shapes go
by the designations I, C, L, or T shapes.

The most common shape is the I -shape; and within this category, the W -
shape is the most prevalent. The designation of W is used to emphasize the
wide flanges of this rolled section (as compared with the much narrower and less
efficiently proportioned flanges of the older, and less commonly used, S -sections:
so-called American Standard Sections).

W shapes come in a variety of sizes with various proportions to suit certain


application domains. In the case of columns, a sub group of wide flange sec-
tions have been designated as column sections. The distinguishing feature of
these sections is a depth that is very close in size with the width of the flanges.
Such column sections are found within the the wide flange member groupings
designated by (W8, W10, W12, and W14 ). As we will find later in the term,
the proportioning of these sections makes them quite efficient for specification
as compression elements as a result of the the comparability in the value of the
second moment of the area when taken about the principal centroidal axes of

2
Figure 1: W-shape concepts

the section.

In contrast to column sections, beam sections are proportioned such that the
flanges are not as wide as the beam is deep. In addition, the steel within these
sections is proportioned to be concentrated within the flanges. From mechanics,
we remember that this approach takes advantage of the Parallel Axis Theorem
as a means for creating a large second moment of the cross-sectional area when
taken about the principal centroidal axis.

Other I-shaped cross-sections include the smaller M, or miscellaneous, sec-


tion and the bearing pile shape HP.

The T shape is merely a W shape member cut down the longitudinal axis.
As a result, the first number is the overall depth of the T while the second
number is the weight per linear foot (each half of the value of the W section
from which it was cut).

Things to keep in mind:

• use the decimal values (and NOT fractional ones) in part I of the manual
when doing calculations
• some of the tabulated values will be slightly different than ones you may
compute by hand (e.g. slenderness values and section properties) due to
subtleties in geometry; consider the manual to be correct in these instances

3
Figure 2: Inter-atomic force in a metal lattice structure

1.3 Mechanical response


Since steel is mostly composed of iron, the details of its composition are con-
sistent with that of other metals: in that no discernible molecular structure is
present in the metallic state. Rather, the iron atoms exist in close proximity to
one another; with their valence electrons being unbound and able to move freely
about the resulting crystalline structure, without disrupting the electronic equi-
librium of the lattice. This electron cloud behavior gives steel (and all metals)
its unique elastic behavior.

When the iron atoms are assembled into a crystal lattice, the free flowing
valence electrons form a sort of electron cloud. Thus leaving the rest of the
iron atom to act a bit like a positive ion; in that it is attracted to the negative
charge of the electron cloud, but repulsed by other, like charged, iron ions. the
resulting effect can be summarized by the response described in Figure 2 and
summarized below:

• if a steel specimen is placed in tension, the attractive forces binding the


iron ions to the metal lattice dominate and the force is resisted.
• in the case of compression, the iron atoms are forced together; thus leading
to the dominance of the repulsive force arising out of the nature of like
charges.

The iron atoms in steel are configured in a Body Centered Cubic (BCC)
lattice structure. In the metallic state that we are concerned with, these crystal
lattices clump together in the form of tightly packed grains whose crystal ori-
entation changes across grain boundaries. Figure 3 summarizes and depicts the

4
Figure 3: Length-scale hierarchy in steel

various length scales that are germane to the current discussion.

It is typical for the crystallographic orientations within the metallic grains


to match the direction of rolling, in a statistical sense. In this way, we see
that actions taken at the macroscopic level have a strong influence over mate-
rial characteristics at microscopic scale. Similarly, we will observe that these
microscopic material features will profoundly affect the macroscopic mechanical
response of the steel in rolled and built-up sections. For now, we will consider
how it is that plastic deformation manifests itself in the steel microstructure.

While it is that inter-atomic forces are at work in resisting the applied loads
occurring in the elastic range, this mechanism of resistance is augmented in the
plastic range. The augmentation occurs in the form of inter-atomic slip that is
concentrated along so-called slip planes. Slip planes are locations along which
atoms within the lattice translate with respect to one another. This translation
is oriented so as to be in the plane of the active slip plane, and in a direction
roughly coinciding with the direction of loading. The actual inter-atomic slip is
manifest as jumps coinciding with integer multiples of the inter-atomic spacing
in the direction of the slip motion. Figure 4 displays a schematic representation
of these concepts.

Considering the mechanics of the slip planes forming in individual crystals, it

5
Figure 4: Slip plane activation

is expected that that the metallic crystal will have a certain strength. However,
various factors work to improve on this strength at the same time as other fac-
tors work to diminish it. Alloying atoms present in the lattice structure interfere
with the formations of slip planes and thus increase the apparent yield strength
over theoretical predictions for the pure single metal crystal lattice. Meanwhile,
dislocations (imperfections in the lattice structure) create sites of weakness that
reduce the apparent yield strength. All of these microscopic effects coalesce into
observable macroscopic mechanical response features discernible in such things
as uniaxial material tests.

1.4 Uniaxial response


Consider a steel specimen (coupon) that is machined so as a to be cylindrical,
or shaped as an elongated rectangular parallelepiped, and loaded in tension at
either end (see Figure 5). This loading condition creates a primarily principal
state of stress throughout the specimen (knows as a coupon); although stress
anamolies remain present at locations in, and around, the grips of the testing
machine used to load the coupon.

The overall mechanical response depicted in Figure 5 may be be separated


into two broad regimes: the elastic and inelastic range. In the former case, the
material response is dominated by the inter-atomic forces depicted in Figure
2, while in the latter case, the source of deformation is primarily composed
of inter-atomic slip along slip planes (as depicted in Figure 4). An important
distinction between the responses exhibited by the two regimes is the recover-
ability of the deformation. In the case of elastic response, the deformations are
fully recoverable, and thus deformations vanish upon unloading. In the case of
inelastic response, a large percentage of the observed deformation remains after
unloading (the portion due to elastic deformation is, however, fully recovered
upon unloading, but in practice, this is oftentimes a very small percentage of
the total deformation observed).

6
Figure 5: Uniaxial steel coupon response

7
It is thus important that we define the controlling deformation measure that
we will be employing in future discussions this term. For our purposes, we
will make due with the simplest definition possible: that which is known as
engineering strain;
∆L
= (1)
L0
where ∆L is the change in length as a result of the uniaxial loading, and L0 is
the original length of the coupon; in the unloaded state. The work conjugate
stress measure for engineering strain is engineering stress, defined as follows;
P
σ= (2)
A0
where P is the applied tensile loading and A0 is the original cross-sectional area
in the unloaded state. These two response measures appear on the axes of the
familiar uniaxial stress-strain response for steel, depicted in Figure 5.

Within the elastic regime, Hooke’s Law governs the response, and thus the
famous relationship between stress and strain (given in Equation 3) remains
valid until the ordered pair (yield ,σyield ) is reached. The slope of the linear
elastic portion of the stress-strain curve is measured by the modulus of elastic-
ity (also known as Young’s Modulus).

σ = E (3)
Once the deformation exceeds the yield strain, yield , inelastic response com-
mences. The initial phase of the inelastic response in mild steel coincides with
what is known as plastification. The yield plateau in the stress-strain response
occurs as a result of plastification. At the micro-structural level, the appear-
ance of plastification develops as a result of slip plane activation at various sites
within areas where local stresses in the micro-structure are high due to inclusions
or imperfections in the lattice structure. The yield plateau occurs as a result
of the activation of various slip planes at increasing strain levels. These var-
ious slip planes subsequently coalesce into larger structures known as slip bands.

The slip bands ultimately become “entangled” with one another; or become
“stuck” on the atoms of alloying metals within the lattice. This interference with
the manifestation and continuation of slip within the slip bands has the net ef-
fect of increasing the apparent strength of the steel; thus leading to a behavior
known as strain hardening. While this portion of the stress-strain response is
characterized by a nonlinear mechanical response, oftentimes an approximate
material stiffness known as the strain hardening modulus, Et , is assigned in or-
der to characterize this response regime.

Once a critical level of deformation is reached (i.e. ultimate ), a phenomenon


known as necking occurs and the specimen appears to shed load; as a result of

8
a drop off in stress. This drop off in stress is merely an artifact of the definition
of stress measure provided in Equation 2. In this equation, the original cross-
sectional area appears in the denominator term, and is thus assumed to remain
constant throughout the response history. However, in the actual specimen,
at attainment of ultimate strain, the cross-sectional area of a critical section
begins to decrease in a manner that is analogous to what occurs in a piece of
bubble gum when it is stretched a long way (i.e. it becomes thinner, more like
a filament, prior to rupture). This diminution in cross-sectional area within the
coupon continues until rupture occurs.

A ductile material will be one that displays a significant rupture strain. As


a comparison, the rupture strain of steel is 0.25, while the rupture strain for
cast iron is 0.0025; a difference of twoorders of magnitude. It easy to see that
steel is more ductile than cast iron.

1.4.1 Toughness
Steel is a relatively difficult material to fracture; it possesses a large modulus of
toughness (as defined by Equation 4 below).
Z rupture
Etoughness = σ () d (4)
0

The quantity defined by Equation 4 can be thought of as an energy dissipation


capacity (i.e. the amount of energy that can be absorbed prior to fracture of the
material). This property is very important across the spectrum of applications
within steel building construction. Toughness and energy dissipation capacity
are at the heart of connection design, seismic response, blast resistance, etc.

A specialized subset of toughness properties relates to something known as


notch toughness. Notch toughness is the material’s ability to resist the for-
mation, and subsequent propagation, of a crack. This material property is
usually gaged through the use of the Charpy V-notch test. In practice, you will
frequently see requirements related to a minimum acceptable notch toughness
presented in bridge plans. The notch toughness of most metals diminishes as
temperature drops; thus in cold climates state DOTs frequently address this
behavior through the prescription of a minimum notch toughness.

1.4.2 High temperature response


At the other extreme in temperature, steel is more ductile; to a point. Beyond
a certain temperature steel becomes compromised in terms of strength and
stiffness. Indeed, at a temperature of 1000 degrees Fahrenheit, steel exhibits
a 30 percent reduction in yield strength and elastic modulus, and almost a 50
percent reduction in ultimate strength. It is important to recognize that 1000
degrees Fahrenheit is not a terribly hot fire within a structure.

9
1.4.3 Loading rate dependency
Steel is no different from most metals in its apparent strength increase at high
strain rates. If one were to take two steel tension coupons and place them into
separate testing machines, loading one at the ASTM prescribed loading rate,
and loading the other at twice this rate, we would notice that the latter speci-
men appears noticeably stronger ; based on the measured mechanical response.
Remember, the coupons are identical in composition and proportions, and thus
it is the loading rate that accounts for this change.

However, as with anything, we can only go so far with this, and thus as
strain rates grow, we pay a penalty for the strength increase in the form of
diminished rupture strains. At very large strain rates (such as those that typi-
cally accompany blast loading), the steel exhibits no ductility and thus fails in a
brittle mode like cast iron. It is important to keep this in mind nowadays since
there is growing interest in improving on the blast resistance of structures.

As engineers, we love to look for previously solved problems and reuse the
results in a new problem, to save time and effort. The blast loading problem is
no different in this regard, and thus there are many in the profession who are
looking to try and extend the lessons that we have learned from earthquake en-
gineering to the problem of blast. Unfortunately, the salient feature describing
these two problems are dramatically different, but the notional trap comes in
the form of dissipation of energy. In earthquake engineering, we seek to dissipate
seismic energy (in the form of ground motion) within the structural system by
way of ductile member and system response (i.e. converting ground motion into
plastic deformation, and ultimately to heat) in order to avoid sudden collapse.
In blast loading, an enormous amount of energy is released; presumably in need
of dissipation. The thought is that connection design and member proportion-
ing guidelines from earthquake engineering will help in blast. Not so! The time
scale of a seismic event is on the order of 10’s of seconds, while the time scale
for the blast portion of an explosion is more like 10’s of milliseconds (a three
orders of magnitude difference!) We can easily imagine how the strain rates
during a blast might be so severe as to induce brittle response within the steel;
in a material sense. No amount of connection detailing, or increase in damping
will help with that.

Another important difference to keep in mind between blast and earthquake


is that in an earthquake, the energy imparted to the structure is spread over a
large percentage of the system. Contrary to this, in a blast, all the energy is
released in a highly localized fashion; in many cases even vaporizing the struc-
tural elements at a very close range to the blast epicenter. This last point,
however, highlights one lesson the is transferable from earthquake engineering:
the notion of redundancy.

The importance of structural redundancy within a system cannot be over-

10
stated. Indeed, the purpose of structural redundancy, as conceived of within
earthquake engineering, is completely portable to the case of blast loading. In
earthquake engineering, redundancy is introduced essentially as a form of in-
surance against any unforeseen eventuality; you never want the structure to
collapse during or after an earthquake. Redundancy hedges the designer’s bet
in this regard. In blast loading, the same imperative applies: no collapse. In
the blast context, we have members being vaporized, connections fracturing,
and structural elements being greatly distorted. While it is difficult to consider
every possible blast threat scenario individually, it is possible to adopt a de-
sign approach that maximizes the presence of redundant load paths within the
structural system.

1.5 Response to a multi-axial stress state


While uniaxial material response is the staple of steel structural design, consid-
erations of multi-axial stress states are required for certain of the specification
provisions; as well as for the practice of structural engineering at a more sophis-
ticated level.

There are two dominant theories related to the initiation of yielding in met-
als subjected to a multi-axial state of stress: the Tresca condition and the von
Mises condition. It is the latter of these that is most useful in predicting the
inelastic response of steel structures, and thus it will be the focus of the present
discussion.

The von Mises theory for the initiation of plastic response in a metallic
material (works best for ductile metals with either a Body Centered, or Face
Centered, cubic lattice structure) is somewhat empirical in nature, and very
easy to conceive of in a geometrical sense. If we consider a space spanned by
the thee principal stresses, σ1 , σ2 , and σ3 , then the von Mises failure surface
appears as a cylinder centered on a generator that coincides with the hydro-
static stress state (i.e. pure dilation, or pure compression of the material cube
- no shearing); the direction cosines of which are √13 , √13 , √13 . Figure 6 displays
a schematic representation of the von Mises failure surface. If a stress point
occurs within the interior of the cylinder, then the governing response of the
material is strictly elastic. If the stress point impinges on the circumference of
the surface, and persists there, then yielding is initiated.

Within Figure 6, the Π planes represent material states coinciding with a


given level of hydrostatic tension. Given this specific hydrostatic stress condi-
tion, a deviation may occur, within the Π plane, such that only elastic response
is exhibited. However, once the stress point reaches, and persists on, the exte-
rior of this plane then inelastic response develops.

Consider now the intersection of this von Mises failure cylinder with the
σ1 -σ2 plane; denoting a 2-D state of stress. This intersection occurs in the form

11
Figure 6: Representation of von Mises failure surface in 3-D principal Stress
Space

of an ellipse on the plane of interest (as depicted in Figure 7). Like its 3-D
counterpart, the 2-D von Mises ellipse demarcates the region of elastic response
(interior of the ellipse) from the region of inelastic response (perimeter of the
ellipse). Our earlier example of a uniaxial coupon test would be represented
by motion of a stress point along one of the individual principal stress axes.
It is pointed out that this 2-D representation clearly shows that the order pair
associated with pure shear is (0.577,0.577) This result will have implications in
the AISC Specification in instances where shear behavior is being considered.
We can effectively convert the yielded state under pure shear into an equivalent
uniaxial condition, in that only 0.577σyield is required to initiate yielding when
the state of stress is pure shear. Rounding up, this becomes the 0.6 conversion
factor that occurs throughout the Specification.

2 Design philosophies and loads


There are two primary, and somewhat distinct, philosophies that are currently
promulgated by the American Institute of Steel Construction. One is somewhat
ancient in its origins, while the other takes advantage of more modern notions
of reliability theory (and is oftentimes considered to have a more rational tech-
nical foundation). For obvious reasons, we will be focusing on this latter case
(known as Load and Resistance Factor Design) in this course. The more ancient
approach (known as Allowable Stress Design) has been around notionally since
humans first started building structures; being more thoughtfully cast around

12
Figure 7: 2-D representation of the von Mises failure surface

the time when the mechanics of materials was becoming more formalized during
the 18th century.

We will begin our discussion with the more ancient of the two approaches:
Allowable Stress Design, ASD. ASD had been the design method of choice for
steel structures during most of the 20th century; with satisfactory result. Indeed,
the ASD steel specifications of the American Institute of Steel Construction,
AISC, evolved and improved as the best engineering research was presented in
various scholarly and professional venues. In other words, while its basis was
ancient, its practical form was up to date, and reflected the best engineering
theory and judgment of the day. However, at the heart of the methodology was
a serious conceptual flaw that becomes immediately apparent upon examination
of the design equation describing the the procedure:
Rn
Ra ≤ (5)

In Equation 5, Ra is known as the allowable stress. This is the stress that is
compared to what is being experienced by the structure, as obtained from equa-
tions based on engineering mechanics, etc. (as gaged using the applied loading
on the structure). It must be less than or equal to some nominal stress, Rn
divided by a factor of safety, Ω. This last point is pivotal: a single factor of
safety is used to govern the design of the component in question.

Consider two identical beams, each designed with ASD, but loaded by two
different effects. In the first beam, we have only the beam’s self-weight and

13
the weight of other permanent structural components (so called dead load ). In
the second case, we have an occupancy loading from a hallway. In the former
case, it is likely that we can say with more certainty what will be the loading
experienced by the structural member over its lifetime, within the structural
system under investigation. In the second case, we are less certain about the
loading since we could imagine an emergency evacuation, or even, say, a fork
truck driving down the hall; thus causing a difficult to quantify excursion in
the loading distribution acting on the beam. We thus see that the same beam,
proportioned with the same factor of safety, Ω, experiencing two very different
loading scenarios over the respective, hypothetical lifetimes. It seems clear that
the probability of serious overloading is more likely in the second case. However,
the design approach of ASD cannot accommodate the expression of this fact.
In effect, each of the two members, designed according to the same AISC ASD
specification provisions, will each have a different probability of failure in reality.

Partially as a result of consideration of this last point, the ASD approach


was all but abandoned by AISC in 1989; with the final printing of the ASD
Specification. However, for reasons that are not entirely clear, AISC has chosen
to reintroduce the ASD approach to design within the context of a completely
revised dual format specification; the specification that we are using in this
course. Having said that, the focus of the course will be more on the second
component of the dual format specification, the so-called Load and Resistance
Factor Design, LRFD approach.

The problem of ASD, with respect to achieving a uniform probability of


failure, was partly the motivation for the development of a more robust and
technically sound design approach known as LRFD. While it is that Equation 5
was useful in highlighting the shortcomings of ASD, so too is Equation 6 useful
in pointing out the strength of LRFD in this regard.

φRn ≥ Σγi Qi (6)

On the left hand side of the equation, φRn is known in aggregate as the design
strength; consisting of a resistance factor, φ, and a nominal strength, Rn . The
right hand side of the equation represents the load effect, and made up of the
summation of the various load effects, and their corresponding partial safety
factors, known as load factors. We immediately see that Equation 6 differs from
Equation 5 in that it has partial safety factors on both sides of the equation;
thus permitting the independent treatment of uncertainties in loading and in
strength.

2.1 Structural reliability fundamentals


In order to gain a deeper understanding of the LRFD approach to design, it
is important to recognize that the occurrence of both the load effect and the

14
Figure 8: Qualitative representation of load and resistance effects

resistances are stochastic in nature, and, it turns out, suitably described by the
so-called normal or Gaussian probability density function, PDF. Figure 8 dis-
plays a plot of both the load effect, Q, and the resistance, R, on the same real
line.

In Figure 8, we have a representation of failure depicted as the cross-hatched


region where the resistance is less than the applied load. This is a compelling
notion of failure that leads us to the formulation of a rather intuitive limit states
function:
g (R, Q) = R − Q (7)
Activation of the limit state occurs when g=0. If g ≥ 0, then the the structure
is safe from the limit state. Conversely, if g < 0, then the structure is unsafe. In
this case, the term limit state is associated with a failure on the part of the arti-
fact to perform as intended; from either a strength or serviceability standpoint.
Examples of strength limit state are: net section fracture, column buckling, for-
mation of a plastic hinge, etc. Examples of serviceability limit states include:
excessive floor vibration, failure of a coating system, excessive story drift, etc.

While it is that Figure 8 provides a useful framework for the notional in-
vestigation of the probabilistic basis of LRFD, it is not sufficiently precise to
enable a formal definition of the probability of failure related to the activation
of a particular limit state. Indeed, Figure 9 is a much better representation of
the two random variables for the purposes of computing a probability of failure.

In the case of research pertaining to the statistical evaluation of load effects


and resistance effects, it is common to know something about the expected value
(or mean value) and standard deviation of the random variable (e.g. flexural

15
Figure 9: Qualitative representation of load and resistance effects

capacity, live loading, etc.), and nothing more. Thus it is desirable that any
probabilistic basis for design require information on only these types of quanti-
ties. In Figure 9, we have just such an opportunity.
h i In this figure, the PDF is
R
depicted as a function of frequency versus ln Q . In this depiction, the loca-
tion of the the expected value is clear, but the position of the ordinate can be
shifted as a function of the reliability
h i index, β, which operates on the standard
R
deviation of the PDF for ln Q . Based on the definition of the probability of
failure, as given in Equation 8,
Z 0   
R
Pf = F ln (8)
−∞ Q
we observe that the net effect of the reliability index is to actually adjust the
probability of failure by moving the ordinate relative to the PDF. In the first
order probability theory at the heart of the LRFD, the form of the reliability
index is based solely on the expected values of the random values, and their
respective standard deviations; as can be seen from Equation 9.
h i
RM
ln Q M
β=q (9)
VR2 + VQ2

In this equation, the variables RM and QM refer to the mean value of the resis-
tance and load effect, respectively; while the quantities VR and VQ refer to the
coefficient of variation for the resistance and load effects, respectively.

It is through the combination of capacity reduction factors, φ, and load fac-


tors, γ, that a target reliability is achieved through the application of the re-
liability index, β. In the case of the LRFD specification, an initial calibration,

16
to match the design outcomes from the 1978 version of the ASD specification,
was undertaken at a live load - to - dead load ratio of three. Therefore at this
ratio in loading, the two specification approaches would yield the same design.
Slight deviations occur at other load ratios.

2.2 Load case combinations


When designing a structure, the engineer must obviously have some loading
in mind. However, it is difficult to know, with absolute certainty, what the
nature of loading will be at all points in time during the life of the structure.
In order to help with this difficulty, ASCE committee seven promulgates a so-
called loads code in the form of the publication ASCE7-02. In this document
are procedures, tables, and maps that facilitates the calculation of the service,
or nominal, loads that are used with the ASD design approach. Similarly, these
same nominal loads are used in the LRFD design approach in instances where
service limit states are being considered (e.g. floor vibration, story drift, etc.)
The nominal loads must be multiplied by load factors in order to bring them into
consonance with the design philosophy that underpins LRFD. In other words,
since it is that LRFD seeks to provide the structural proportions necessary to
guard against the manifestation of a particular limit state, then what is needed
is the required loading levels that trigger the limit state condition. So then we
much amplify these service loads found in ASCE7-02 in order that they may
reflect the loading that represents a rational estimate for the maximum lifetime
value. Toward this end, we may consider the following load case combinations:

1.4D (10)

1.2D + 1.6L + 0.5 (Lr , S, R) (11)


1.2D + 1.6 (Lr , S, R) + (0.5L, 0.8W ) (12)
1.2D + 1.6W + 0.5L + (0.5Lr , S, R) (13)
1.2D + 1.0E + 0.5L + 0.2S (14)
0.9D + 1.6W (15)
0.9D + 1.0E (16)
Within Equations 10 through 16 it is noted that parenthetical terms carry the
connotation of “or” in the sense of an exclusive disjunction (e.g. ...either this or
that...) and not the usual mathematical interpretation (e.g. ...one, or the other,
or both). Keeping this in mind, it should become clear that the load effect that
makes the load case combination sum greatest, is the one that should be used
within the design equation (i.e. Equation 6).

It is interesting to observe from Equations 10 through 16 that magnification


of the nominal loads is not universally applied to all terms. Indeed, several of
the load terms are reduced from the service condition as a result of application of

17
the various load case combination equations. This arises out of the recognition
that it is statistically unlikely that several of the load effects will simultaneously
achieve their maximum lifetime values (e.g. the maximum number of trucks
will occur on the span of a bridge at the instant that a magnitude 8.6 earth-
quake strikes in the midst of a hurricane) and thus these load factors smaller
than one are applied as a recognition of this. It seems more rational to imagine
that at the moment when a load effect reaches its maximum lifetime value, it is
probable that the other load effects will be at some average intensity. So then,
the factors that are smaller than one are use to convert the service loads into
arbitrary point in time values of the load effects they modify.

Bearing this in mind, we may assign the following usages to the individual
load case combination equations:
• Equation 10 emphasizes the effect of dead load during construction
• Equation 11 emphasizes the effects of live loading
• Equation 12 emphasizes roof live loading, or snow, or rain
• Equation 13 emphasizes wind load effects coinciding with the direction of
dead load
• Equation 14 emphasizes earthquake load effects coinciding with the direc-
tion of dead load
• Equation 15 considers overturning where the wind opposes the dead load
• Equation 16 considers overturning where the earthquake loading opposes
the dead load

3 Global response
This portion of the discussion on the behavior of steel structures has as its focus
response features that manifest themselves across characteristic length scales
that are on the order of the member length itself (in the case of individual
elements); or are system-wide in scope (as when considering the response of
a framework). The types of limit states and structural responses at issue in
this section will tend to be driven by notions of stability; a term that has a
suggestive colloquial meaning, that is useful to consider when framing our more
formal definitions, that will be provided later.

3.1 Beams
Beams are typically structural elements that resist transverse applied loading
by means of flexure induced normal and shear stresses, acting over the area of
a given cross-sectional slice. These quantities are described within elementary
flexural theory from strength of materials using the the so-called Bernoulli-Euler
Beam Theory.

18
3.1.1 Bernoulli-Euler beam theory
In this Bernoulli-Euler description, flexural normal stresses are linearly related
to cross-sectional depth as a direct result of the use of Navier’s plane section
hypothesis and Hooke’s Law for elastic material response. Equation 17 presents
the mathematical statement for the relationship between the important geomet-
ric and material parameters at work in producing flexural normal stresses.

M [y]
σ= (17)
I
In Equation 17, σ denotes the flexural normal stress; M is the applied moment
loading (assumed to act about a principal centroidal axis); I is the second mo-
ment of the area taken about the this same centroidal axis (e.g. Ix−x ); and y is
the distance along the depth dimension, as measure in the plane of bending (i.e.
along the axis of the orthogonal centroidal axis to the moment axis). Based on
the form of Equation 17, the linear quality of the flexural normal stress distri-
bution through the cross-sectional depth is evident for this case.

While it is that Navier’s plane section hypothesis would seem to indicate


an inability of the Bernoulli-Euler theory to cope with the existence of flexure-
induced shear stresses, a consideration of statical equilibrium on a beam section
under the action of a moment gradient loading leads to the form of the equation
for shear stress appearing in Equation (18).

VQ
τ= (18)
It
In Equation 18, I is once again the cross-sectional centroidal moment of inertia,
t is the thickness of the shear plane (i.e. the width of the cross-section at the
depth where the shear stresses are being computed), and Q is the first moment
of the outward cross-sectional area associated with the beam material above the
cut, of width t, where the shear stresses are being evaluated.

In the foregoing equations, certain important and fundamental assumptions


were made, the most important of which are now enumerated:
• linear elastic material response
• plane sections remain plane and orthogonal to the neutral axis
• small deformations
• cross-sections do not distort
• isotropic and homogeneous material
As it turns out, in the practical design of steel structures, we are oftentimes
interested in the response of structural elements and systems in the inelastic
range; as a result of the need to dissipate energy, redistribute loading (i.e. as

19
Figure 10: Idealized elastic-plastic constitutive response of steel

part of structural redundancy), blunt the effects of stress raisors, etc. These
interests motivate the consideration of inelastic beam response in a general
sense.

3.1.2 Inelastic flexural response of a cross-section


When treating the inelastic flexural response of beams, it is common to make
certain assumptions regarding things like beam kinematics and material re-
sponse. In the case of the latter, it is usual practice to assume an idealized
elastic-perfectly plastic uniaxial steel stress-strain response curve; as given in
Figure 10.

Such an idealization of the material behavior in a uniaxial state of stress


permits a powerful, and useful, approach to modeling the response of inelastic
beam sections, as a means for predicting relevant limit state conditions. As a
direct result of the assumed bi-linearity in the material response, two primary
response regimes present themselves: elastic, and inelastic. In the former case,
the Bernoulli-Euler beam theory from elementary strength of materials is in
force, and thus Equations 17 and 18 may be used to predict the the relation-
ship between moment and flexural normal stress; as well as the relationship
between moment gradient and horizontal shear stress within the beam section.
In addition, we may further recall from strength of materials that the governing
differential equation describing the flexural response of a Bernoulli-Euler beam
is:  2 
d y
−EI = M (x) (19)
dx2
In Equation 19, E is the elastic modulus, I is the moment of inertia, y is the
coordinate corresponding with transverse deflections of the beam, while quan-
tity x is the spatial coordinate corresponding to longitudinal position along the

20
beam centroidal axis.
2
d y
Using the usual approximation for curvature, φ from calculus ( dx 2 ≈ φ), we

may formulate a useful expression relating moment and curvature in the elastic
range as
M
φ= (20)
EI
As it is that we are interested in both elastic and inelastic flexural response,
Equation 20 will not serve our purposes by itself. Indeed it seems intuitively
obvious (due to the overt presence of E ) that Equation 20 will not be strictly
applicable in the inelastic range. fortunately, experimental testing clearly shows
that Navier’s Hypothesis holds well into the inelastic range, and thus it is very
reasonable to use kinematics based on this assumption as a means for arriving
a cross-sectional curvatures in a partially plastified flexural cross-section. From
such a consideration, we note that tan (φ) = max c , where c is the distance from
the centroidal axis to the fiber in question within the cross-section; the same
fiber where max is being measured. As a direct results of our small deflection
assumption, this result may be approximated as:

φ= (21)
y
where y is the distance, measured from the neutral axis, along the depth axis
where the strain  is being measured.

With this foundation in mind, we may consider the response of a rectangular


steel bar in flexure that takes the member into the inelastic response regime.
Figure 11 is a schematic description of what occurs in this case and the following
discussions refers to this figure and its notation.

The member depicted in Figure 11 is a subjected to a constant moment


loading and thus horizontal shear stresses do not influence the occurrence of
plasticity in the notional fibers of the cross-section. Indeed, it is the uniaixial
response of these fibers that are at issue in the present discussion. In Case 1),
the strains in the cross-section remain below the strain at first yield, and thus
Hooke’s law leads to the linear distribution of stress shown in Figure 11. In
addition, the familiar equations from the mechanics of materials (i.e. Equa-
tions 17, 18, and 20) are valid. The linearity in response can also be observed
through an examination of the moment-curvature response of the same beam
cross-section, as depicted in Figure 12.

In Figure 12 it is noted that the slope of the moment-curvature response,


emanating from the origin and intersection the response at a load level cor-
responding to the moment M1 , has a slope of EI. This quantity is known in
aggregate as the flexural rigidity, and it represents an amalgamation of a repre-
sentation of the material stiffness, E, with the geometric stiffness, I.

21
Figure 11: Inelastic response of a steel beam

Figure 12: Moment-curvature response of rectangular steel beam

22
As the moment applied to the beam grows, the strains ultimately exceed the
yield strain, y , and thus, based on the response depicted in Figure 10, the stress
reaches its idealized ceiling of the steel yield stress, σy . The internal resisting
couple, that develops in the beam to equilibrate the externally applied loading,
may be conceived of as having two parts: an elastic portion (coinciding with
the so-called elastic core proximal with the neutral axis), and a plastic portion
that is formed by the stress resultants acting at the outer fibers of the beam,
through moment arms measured back to the neutral axis.

As it is that that an idealized stress-strain response is adopted in this ex-


ample, the material stiffness associated with the yielded portions of the outer
fibers in the beam vanishes, and thus the moment curvature response begins to
soften, as depicted in Figure 2 for Case 2). in the limit, as strains grow very
large, the inelastic outer region migrates in toward the neutral axis, and the
elastic core shrinks to nothing (e.g. Case 3 in Figures 11 and 12). Notionally,
this condition is referred to as plastification of the cross-section. At this load,
it is assumed that a plastic hinge forms in the member (i.e. a condition where
the persistence of the load level at Mp leads to a continued rotation of the beam
slice, much like a rusty hinge pushing back with a constant load while moving
through an ever increasing angle.)

3.1.3 Plastic Collapse


This conceptualization of the “rusty hinge” proves to be useful when designing
statically indeterminate structural systems. As an illustration of this, consider
the example of a propped cantilever acted upon by a uniformly distributed
transverse loading; as depicted in Figure 13. We recognize that the formation
of a plastic hinge in a flexural member is certainly worthy of consideration as a
flexural limit state. However, this characterization is context dependent: if the
beam in question were simply supported, then the formation of a single plastic
hinge in the span would transform the beam into a mechanism that is unable
to maintain its current load level - definitely a limit state. However, change
the boundary conditions of this beam, and you have the example in Figure 13;
where we will show that the formation of the first hinge is not a concern at all
in terms of well behaved system-wide response.

In Figure 13, we observe an evolution in response wherein the load W grows


to an intensity W1 at which the first hinge occurs at the wall. At this point in
the load history, the plastic hinge acts to reduce the number of redundants in the
system by one and thus the structure behaves as if it is statically determinate as
additional load is added. Thus in the case of first hinge formation in our exam-
ple, the structure remain stable; although its overall flexural stiffness is reduced
somewhat. Beyond the load level W1, the structure may resist increasing load
intensity until the load level W2 is reached. At W2, the second hinge forms and
the structure becomes a mechanism and fails. We may summarize this response
in terms of the overall beam length, L, and the member cross-sectional moment

23
Figure 13: Example of moment redistribution in a statically indeterminate sys-
tem

24
capacity, Mp :
• intensity of distributed loading at incipient 1st hinge formation → W1 =
8Mp
L2
11.67Mp
• load intensity at collapse, when 2nd hinge forms → W2 = L2

Based on this example it seems clear that a system-wide perspective is use-


ful (and economical) when considering when a strength limit state has actually
been reached. In the case of Figure 13, the structure was able to resist an addi-
tional ≈ 46 percent increase in loading above what was needed to form the first
hinge. Indeed, one way to conceive of the behavior displayed by this example
is through a re-distribution of moments. At load level W1, the moment at the
wall attains its theoretical maximum value (Mp ), and thus mantains this load
level as the increasing load level forces a increase in the in-span moments; until
the largest of these attains Mp , at which point a mechanism forms and collapse
ensues.

It is important to recognize that in order for this redistribution to occur


within the structure, the plastic hinge at the wall must maintain Mp , while
absorbing the required inelastic rotation to effect the moment redistribution
needed for mechanism formation. We will see later in the course that this need
for plastic hinge rotation capacity is one of the considerations that drives the
proportioning of any given flexural cross-section. This requirement on inelastic
rotation capacity at a plastic hinge motivates consideration of cross-sectional
compactness and adequate bracing. Cross-sectional compactness is achieved
when the proportions of the the plate components that comprise a given flexu-
ral cross-section are sufficient to delay the manifestation of local plate buckling
modes that may interfere with the ability of the member to maintain Mp while
providing sufficient plastic hinge rotation required for moment redistribution.
Similarly, adequate bracing is required to maintain the alignment of the beam
longitudinal axis while the cross-section is becoming increasingly plastified. Re-
call from our discussion of the behavior presented in Figure 13 that once a fiber
within the cross-section yields, it is assumed that it no longer possesses any ma-
terial stiffness, and thus its contribution to the in-plane flexural rigidity, EIx−x ,
vanishes. The same holds true for the contribution of the yielded fiber to the
out-of-plane flexural rigidity, EIy−y .

As plastification of the cross-section progresses during the formation of a


plastic hinge, the member’s ability to remain straight is compromised due to
the reductions in material stiffness. This can lead to a phenomenon known as
inelastic lateral torsional buckling; a topic that will be treated more carefully
later in the course. For now, consider the compression flange of the beam to be a
column that is somewhat braced by the tension flange that it is attached to. As
the out-of-plane flexural rigidity diminishes, then the likelihood that the com-
pression flange will buckle out-of-plane grows: lateral-torsional buckling. Since
the compression side of the beam is tied to, and stabilized by, the tension flange

25
Figure 14: Schematic of lateral-torsional beam deformation

(which does not want to move out-of-plane), it experiences twisting within the
plane of the cross-section: lateral-torsional buckling (Figure 14 displays this
deformation schematically).

It is logical to imagine that a restraint against lateral-torsional buckling


should be present in the regions of the beam adjacent to the hinge locations.
Such restraint is provided by bracing members that either restrict: lateral deflec-
tions, twist, or both of these. Appendix 6 of the Specification provides guidance
on how to proportion these bracing members for stiffness and strength; a topic
that we will discuss more fully later in the course. For this discussion, we will
merely consider the placement of such members.

In the case where a given flexural member is only required to develop a


plastic hinge, without there being a large demand for plastic rotation, then AISC
prescribes a maximum unbraced length as follows (see Specification Chapter F):

Lb ≤ Lp (22)
s
E
Lp = 1.76ry (23)
Fy

26
Equation 22 makes the statement that the maximum unbraced length adja-
cent to the plastic hinge region must be smaller than Lp . Equation 23 defines
Lp in terms of the elastic modulus, E, the yield stress, Fy , and the weak-axis
radius of gyration, ry . Equation 23 can be arrived at through a somewhat
conservative application of the classical equation governing the elastic lateral
torsional buckling of a beam; by solving for the unbraced length that results in
a critical moment that coincides with the full plastic capacity of the member
cross-section. This approach is rational in light of the assumption regarding
the material response of steel being elastic-perfectly plastic. In such a case, the
equations describing elastic lateral-torsional buckling may considered to be valid
up until incipient buckling (which is assumed to occur at the attainment of Mp ).

While Equations 22 and 23 help in alleviating the threat of out-plane-


instability interfering with the formation of a plastic hinge in a particular region
of a structural element, these bracing provisions are too liberal for cases in which
significant moment redistribution is required. In order to accommodate the type
of moment redistribution that is consistent with the behavior depicted in Figure
13, it is necessary to specify more restrictive spacing limits for bracing members.
Specifically, Appendix 1 of the Specification treats the case of inelastic analysis
and design (formerly known as plastic analysis and design). In this appendix,
a different equation is promulgated for the maximum bracing spacing required
to develop a collapse mechanism:
    
M1 E
Lpd = 0.12 + 0.076 ry (24)
M2 Fy
where ry is the minor axis radius of gyration and M1 and M2 are the smaller
and larger end moments, respectively, acting on the unbraced beam length. This
last term is positive in cases when the moment loading acting on the unbraced
length induces reverse curvature; and negative in cases where such moment in-
duces single curvature. The form of Equation 24 was developed as a result of
the research of Bansal and Yura at the University of Texas, Austin.

3.1.4 Lateral-torsional buckling


As we embark on a more formal description of what comprises the limit state of
lateral torsional buckling it is important to recognize that the rolling and fabri-
cation processes of steel members have an impact on the manifestation of this
mode of failure. Specifically, when considering that hot-rolled members leave
the rolling stands at relatively high temperatures (greater than 1450 degrees F),
the finished shape, left to air cool, may do so unevenly as a result of the nature
of the standard heat conduction problem (i.e. the flange-web junction has more
thermal mass, the flange tips have more surface area through which to cool, etc.)

Such uneven cooling leads to a residual stress field that is distributed in


fashion similar to that depicted in the inset of Figure 15. In this figure, only

27
Figure 15: Residual stresses in a flexural cross-section

the residual stress distribution through the top flange and web are shown. It is
pointed out that compression stresses tend to form at the flange tips and web
mid-height as a result of the greater heat transfer occurring at these locations;
thus leading to their earlier hardening. Once these locations have hardened,
they become effective in resisting stresses and, consequently, as the flange-web
junctions cool, they try to shrink the size of these previously hardened regions;
thus creating a compressive loading. Likewise, these later cooling portions of the
cross-section harden with this restraint from adjacent cross-sectional portions
still intact, and thus these later cooling sections experience a tensile residual
stress effect. The residual stresses in a hot-rolled section may be as high as
20ksi; a high percentage of the nominal yield strength of the steel. The net
effect of this self-equilibrating residual stress field is that a given flexural cross-
section may experience an earlier than expected nonlinear response as a result
of certain portions of the beam reaching yield earlier than they theoretically
should. Such response is depicted in the moment-curvature schematic in Figure
15.
With this background, we may now lay out the nature of the response regime
through which lateral-torsional buckling manifests itself in steel flexural mem-
bers. Figure 16 is a depiction of evolution of the lateral-torsional buckling limit
state as it spans the extremes in beam response: from plastic hinging in beams
with closely spaced bracing, to elastic lateral-torsional buckling in long beams.
It is pointed out that Figure 16 depicts a prediction of the governing type of
lateral-torsional buckling failure mechanism, for a given cross-section, as the un-
braced length is varied: short beams can develop Mp , while long beams cannot
(experiencing instead a manifestation of elastic lateral-torsional buckling.)

28
Figure 16: Prediction of global beam failure modes for a given section as a
function of Lb

We may more formally describe what is meant by the term lateral-torsional


buckling through a consideration of equilibrium formulated on a deformed struc-
tural configuration that is consistent with the qualitative description provided
earlier. The development that follows will be a simplified treatment of the topic
since it is not practical for us to develop all of the apparatus of non-uniform
torsion theory needed to properly support this derivation. Throughout this
development, it will be assumed that the material remains linear-elastic, the
deformations are small, and the major axis moment of inertia is much larger
than the minor axis moment of inertia (i.e. Ix−x >> Iy−y ).

Figure 17 depicts a beam subjected to concentrated end moments having a


magnitude Mo . In this figure, a cross-sectional slice a distance z from the ori-
gin is considered. The deformation of this cross-section in the lateral-torsional
mode is displayed in the right portion of the figure. The word mode is entirely
appropriate in the present discussion since we describing a type of buckling,
in this case lateral-torsional buckling. When this type of buckling manifests
itself, the in-plane configuration of the beam becomes unstable and thus un-
avoidable imperfections in loading and initial geometry lead the beam onto a
stable secondary post-buckling path. We thus formulate internal equilibrium
using a cross-sectional displacement that is consistent with this path (note: the
cross-section is assumed to be undistorted - it merely experiences a rigid body
motion).

We may now consider the deformation of the compression (top) flange result-
ing from the manifestation of the lateral torsional mode under investigation. An
important component of the governing deformation mode involves the lateral
translation, -u, of the centerline of the flange, at the flange-web junction. Figure

29
Figure 17: Cross-sectional kinematics in lateral-torsional mode

18 depicts this motion, in addition to the angle change between the longitudinal
Z-axis and the internally consistent longitudinal axis that is tangent to the the
curved arc that describes the elastic curve of the compression flange, the Z’-axis.

Based on the deformation depicted in Figure 18, we may employ small angle
theory to see that the externally applied moment, Mo , is resolvable into two
components, consistent with the deformation. The first of these components
represents the projection of the externally applied moment onto the major prin-
cipal centroidal axis of the cross-section under investigation (i.e. the X’-axis).
This component is given by Equation 25.
  
du
Mx0 = Mo cos − ≈ −Mo (25)
dz

Similarly, the second component of the externally applied moment may be


resolved as:
  
du du
Mz0 = Mo sin − ≈ −Mo (26)
dz dz
The moment provided in Equation 26, Mz0 , acts about the Z’-axis. This axis
is the instantaneous longitudinal axis at a cross-section; by virtue of its tangency
with the top flange elastic curve. As a result of the foregoing, we recognize Mz0
to be acting as a torque within the context of the current deformation mode.

30
Figure 18: Compression flange deformation accompanying lateral-torsional
buckling

Thus, we may consider this component of the externally applied moment as the
forcing function that drives the differential equation for non-uniform torsion.
This equation is rigorously developed in CEE779, requiring a significant portion
of the class time. For our purposes, this equation simply appears as equation
27 in the present discussion; suitably modified with our torque.

dφ d3 φ
Mz0 = GJ − ECw 3 (27)
dz dz
In Equation 27, G denotes the shear modulus (sometimes approximated as
0.385E for steel), J is the polar moment of inertia (also sometimes St. Venant’s
Torsion Constant), and Cw is known as the warping constant. It is pointed out
that tabulated values for both J and Cw can be found in Part I of the Manual.
It is also pointed out that in Equation 27 the term GJ dφ dz is associated with
3
the uniform torsional response of the beam while the term ECw ddzφ3 is asso-
ciated with the nonuniform torsional response; the latter also being known as
warping torsion, and the former sometimes being known as St. Venant’s torsion.

We may now make a substitution for Mz0 in Equation 27 in order that we


may describe the response in terms of the directly applied end moment (i.e. we
will use the relationship Mz0 = − du
dz Mo ) and thus re-express Equation 27 as:

du dφ d3 φ
− Mo = GJ − ECw 3 (28)
dz dz dz

31
Equation 28 may now be differentiated one time to obtain:

d2 u d2 φ d4 φ
− M o = GJ − ECw (29)
dz 2 dz 2 dz 4
It is recalled that curvature (in this case out-of-plane) may be expressed ap-
proximately as:
d2 u My 0 Mo φ
= = (30)
dz 2 EIy−y EIy−y
The last expression in Equation 30 comes about from a consideration of Figure
17. We may resolve the applied moment, Mo , into components acting in the
X’-axis and the Y’-axis; the latter being applied in Equation 30. Substituting
the result from Equation 30 into Equation 29 leads to the governing differential
equation for lateral-torsional buckling:

d4 φ d2 φ Mo2 φ
ECw − GJ − =0 (31)
dz 4 dz 2 EIy−y
The general solution of Equation 31 is:

φ = A1 enz + A2 e−nz + A3 cos (qz) + A4 sin (qz) (32)


q p q p GJ
In equation 32, n = ± α + β + α2 , q = ± −α + β + α2 , α = 2ECw ,
Mo2
and β = E 2 Cw Iy−y . The constants A1 through A4 may be obtained through a
consideration of the boundary conditions. In our case, we will be considering a
simply supported beam and thus our boundary conditions are given as:
• φ = 0 at z = 0 and z = L (i.e. no twist over the simple end supports)
d2 φ
• dz 2 = 0 at z = 0 and z = L (i.e. no warping restraint at end supports)
This leads to a form of the Sturm-Liouville boundary value problem; the solu-
tion of which leads to an eigenvalue problem that yields the moment expression
appearing in Equation 33.
s 2
π πE
Mcr = Cw Iy−y + EIy−y GJ (33)
L L
As it is that Equation 33 is valid only for load levels below those that yield
any portion of the cross-section, an additional theory is needed to treat cases
where the beam is inelastic at the time of lateral-torsional buckling. In fact, the
nature of such a theory was implied in the form of Figure 16; where a line was
used to span the region of inelastic lateral-torsional buckling. It is observed that
this linear expression must connect the ordered pairs: (Mp , Lp ) and (Mr , Lr )
and thus we may use an equation of the form:
  
Lb − Lp
Mn = Mp − (Mp − Mr ) (34)
Lr − Lp

32
Despite its simple form, Equation 34 actually works quite well in predicting the
nominal moments associated with the limit state of inelastic lateral-torsional
buckling.

It should be observed that in the forgoing treatment of lateral torsional buck-


ling (both elastic and inelastic), the external loading case considered is that of
constant moment. Indeed, the very form of the boundary value problem leading
to the critical moment expressed in Equation 33 is predicated on an assumed
loading configuration. If this loading were to change, then the entire problem
would have to be re-framed and then solved again. This is the case for each
slight variation in the applied loading and boundary conditions. Fortunately,
there is an approach that AISC promulgates that involves the modification of
the constant moment solution to account for such variations; thus enabling the
extension of Equation 33 for use in other cases. The modification of Equation 33
is effected through the use of an amplification term, referred to as the moment-
gradient coefficient, Cb . This equation has evolved over the years, in the various
editions of the steel specification. Currently it takes the form:
12.5Mmax
Cb = Rm ≤ 3.0 (35)
2.5Mmax + 3MA + 4MB + 3MC
In Equation 35, Mmax is the maximum moment occurring anywhere within the
unbraced beam length under investigation. In addition, MA , MB , and MC are
the moment at the quarter, half, and three quarter points, respectively, within
the unbraced length. Rm is known as the mono-symmetry parameter, and is
related to the fact that the application of a moment gradient amplification ap-
proach may be un-conservative when used on singly symmetric members bent
in double curvature. Thus this factor is other than unity only in the latter case
(this is most closely related to the case of bending of TEEs in reverse curvature).

Application of equation 35 can be illustrated with a brief example. Consider


the simply supported beam shown in Figure 19. In this case, Mmax and MB
coincide, Mmax = MB = P4L and MA = MC = P8L . Thus leading to:

12.5 P4L

Cb =  (1.0) = 1.3158 ≈ 1.32 (36)
2.5 P4L + 3 P8L + 4 P4L + 3 P8L
  

It is pointed out that all moments appearing in Equation 35 are to be in the


form of absolute values (i.e. do not preserve the sign when using Equation 35).

The net effect of using Equation 35 is to amplify the nominal moment ca-
pacity computed for a simply supported beam segment subjected to constant
moment loading (i.e. worst case) so as to be in closer agreement with the ob-
served response of beams with other loadings and end conditions. Of course,
such amplification is limited to the full plastic capacity. Thus, while circum-
stances will arise in which the moment-gradient coefficient will lead to predic-
tions of nominal moment greater than Mp ; the engineer must be on the lookout

33
Figure 19: Moment gradient loading in a simple beam

for such erroneous predictions and always ensure that the moment prediction
from any design equations does not exceed the full plastic capacity. Figure 20
illustrates this notion by showing that the the net effect of the application of
the moment-gradient coefficient, Cb , is to shift the moment-unbraced length re-
sponse up, for a given cross-section; at the expense of the length in the inelastic
transition region of the response.

3.1.5 Simple analysis example


The following example couples what we have learned, thus far, with design aids
that are available in Part 3 of the Manual. Specifically, the beam curves in Ta-
ble 3-10 are employed as a means for quickly arriving at the nominal moment
capacity for a beam possessing a specific cross-section and unbraced length.
These curves are, in principle, similar in form to what was presented in Figure
16. Since this is the case, we may simply find the relevant beam curve in Table
3-10 matching the cross-section under investigation. The unbraced length may
then be used to move along the relevant beam curve until the governing design
moment is identified. Figure 21 presents a schematic description of the example
now considered.

In this example, a simply supported W18x50 A992 hot-rolled member is


subjected to three concentrated transverse loads imposed in a fashion consis-
tent with the presentation in Figure 21. It is assumed that lateral-torsional
bracing is present at each of the three loading points, as well as at the support
locations at the ends. The resulting shear and moment diagrams appear in Fig-
ure 21. It is the latter of these that will be considered in the determination of
the adequacy of this member.

34
Figure 20: Illustration of moment gradient amplification

Figure 21: Beam analysis example

35
Figure 22: Table of moments for beam example

Figure 22 displays a table comprised of the analysis variables at issue in this


example. The first row indicates the critical applied moment (i.e. the demand)
that is acting within the first unbraced length of the example beam (e.g. the
first 18’ beam section immediately to the right of the left-most support). Each
of the four columns of numerical data correspond with the ordering of the un-
braced lengths appearing in Figure 21. The second row contains the nominal
moment capacity. The third row contains the moment gradient coefficient asso-
ciated with the unbraced length corresponding to the particular column in the
table. The second to last line provides the design moment (e.g. the supply). It
is noted that the LRFD capacity reduction factor, φ, takes a value of 0.9 for all
flexural limit states.

As was pointed out earlier, the values contained in the second row of the
table in Figure 22 are obtained from Tables 3-10 in the Manual. It is important
to observe that the W18x50 line is dashed for the unbraced lengths at issue in
this context. If this were a design problem, that would be an indication to use
that a stronger section of equal, or lesser, weight is available. This section may
be identified by continuing up the moment axis at our given unbraced length
until we arrive at the first solid line. The section associated with the first solid
line will be stronger, and often lighter, as compared with the section whose re-
sponse curve is dashed in the region of the unbraced length at issue.

3.1.6 Shear resistance of beams


Chapter G in the Specification provides a framework for the treatment of shear
resistance in hot-rolled shapes. We may consider several of these as they apply
to the case of our simple example outlined in Figure 21. In the case of this ex-
ample, a W18x50 is considered. This particular member experiences a critical
shear loading of 12kips; assuming this loading to have resulted from the con-
trolling load case combination, we are then able to call this the ultimate shear
loading, Vu .

In the shear theory espoused by AISC, it is only the web region that is
effective in resisting the transverse shear loading within the section, and thus

36
we can frame the AISC nominal shear resistance as:

Vn = φ (0.6) Fy Aw Cv (37)
In Equation 37, φ = 1.0, Fy is the minimum specified yield strength, and Aw
is the area of the web. It is observed that, in the flexural theory assumed valid
in steel beam response, it is rational to assume that only the member web is
effective in resisting shear since it is presumed that as any flexural limit state
is reached, the flange participation in load resistance will be significant. Thus,
based our assumed elastic-plastic material response, the material stiffness in
the flange regions will go to zero; and thus these portions of the cross-section
will not be effective in providing shear resistance. In addition, in the vast
majority of currently hot-rolled I-shaped cross-section, the web shear coefficient,
Cv , is unity. This latter condition is justifiable when web shear buckling does
not exert a significant influence over the governing shear response; a situation
accompanying a web proportioned to satisfy the following relation:
s
h E
≤ 2.24 (38)
tw Fy

It is also pointed out that the factor 0.6 appearing in Equation 37 is a direct
result of application of the von Mises failure criterion discussed in section 1.5
of the present discussion.

Considering now the example from Figure 12, we note that Vu = 12k. This
ultimate load must be resisted by an adequate internal shear response in the
member. This condition may be checked using Equations 37 and 38. Equa-
tion 38 is satisfied for our example and thus Equation 37 may be applied with
Cv = 1. It is pointed out that Equation 38 is a local web buckling check to
determine if the web is susceptible to shear buckling prior to the manifestation
of the yielding limit state described by Equation 37. While our W18x50, A992
beam is not susceptible to shear buckling, in members where such a limit state
is a concern, reference should be made to equations (G2-3) through (G2-5) in
Chapter G of the Specification.

Continuing on with our example, we note that minimum specified yield stress,
Fy , for A992 steel is 50 ksi. In addition, Aw may be computed as the product
of d = 18.0” and tw = 0.355”; thus Aw = 6.39in2 . Equation 37 subsequently
yields a design shear capacity of φVn = 191.7 kips. This result agrees with the
result provided by the design aid in Part 3 of the Manual, Table 3-2, where
φVn is given as 192 kips on p. 3-17.

3.2 Columns
Columns are structural elements that develop internal stresses in a fashion that
permits the resistance of an external loading that tries to make the member

37
shorter in the direction of its longitudinal axis. As with beams, the two primary
limit states in a column member involve: 1) cross-sectional yielding and 2) a
global instability. Also as in the case of flexural members, there are a series of
local plate buckling limit states that must be investigated; but discussion on this
is delayed until later in the course when local behavior is considered formally.
For the purposes of the present discussion, global failure modes will be treated.

3.2.1 Yielding
It seems intuitive to imagine that as axial compressive loading on a steel struc-
tural element grows, an upper limit to response may be reached. Such a ceiling
would be akin to the plastic moment, Mp , but would have to be consistent with
the uniform stress field present in a concentrically loaded axial compression
member. Taking the gross member cross-sectional area, Ag , and multiplying it
by the yield stress, Fy , seems to be a rational candidate for such an upper limit.
Indeed, the load corresponding to Ag Fy is exactly the notional upper limit on
strength that we are after. You will sometimes hear this load referred to as the
squash load.

We recognize that the manifestation of such a limit state as that which


governs at the squash load, can only occur under circumstances in which either
the column is very stocky (i.e. displaying a small ratio of length, L, to minimum
radius of gyration, rmin ), or when the column is very well braced. Absent these
two conditions and we can easily imagine that a given column would more
likely fail due to some form of global buckling. There are three primary global
buckling modes exhibited by hot-rolled steel members in compression: 1) flexural
buckling; 2) torsional buckling; and 3) flexural-torsional buckling. We begin our
discussion on column stability with a treatment of flexural buckling.

3.2.2 Flexural buckling


Consider the differential column length presented in Figure 23. This differen-
tial column segment may be considered as a free body since all of the internal
forces needed to maintain equilibrium with the missing column portions are de-
picted. Consideration of statical equilibrium on the differential column length
will subsequently lead to the identification of the governing differential equation
for the flexural buckling response of a column. Just as in the case of lateral-
torsional buckling in a beam, the development of the equations governing the
failure response of a column in flexural buckling can only be arrived at through
a consideration of equilibrium as it is formulated on the deformed configuration
of the structural element.

It is observed that the formulation presented in Figure 23 admits the possi-


bility of a moment gradient being present along the differential column segment,
and thus transverse shear forces must be considered in the formulation. Equilib-

38
Figure 23: Differential column length

rium in the direction of the y-axis leads to the following equilibrium equation:
+
→ ΣFy = 0 = (V + dV ) − V ⇒ dV = 0 (39)

While consideration of moment equilibrium about the B-end of the differential


segment leads to:
dy dM
(CCW +) ΣMB = 0 = P dy + V dx − dM → P − +V (40)
dx dx
Differentiating Equation 40 one time with respect to x yields:

d2 y d2 M dV
P 2
− 2
+ =0 (41)
dx dx dx
We may now use the result from Equation 39 to show that the last term on
the left hand side of Equation 41 is zero. In addition, we may recall from
d2 y
Section 3.1.2, Equation 19 that M (x) = −EI dx 2 thus leading to the governing

differential equation for flexural buckling in a column:

d4 y d2 y
EI + P =0 (42)
dx4 dx2
Equation 42 is a fourth order linear, homogeneous ordinary differential equation;
the solution of which appears below as Equation 43.

y (x) = Asin (kx) + Bcos (kx) + Cx + D (43)


q
P
In Equation 43, k = EI . We observe that the solution has four unspecified
constants: A, B, C, and D. These are the constants of integration required for
the four integrations needed to solve Equation 42. The integration constants

39
Figure 24: Euler column

are computed during the solution of the boundary value problem that special-
izes the general solution (i.e. Equation 43) for use in the specific problem under
investigation.

As a starting point for our discussion, we will consider the classical pinned-
pinned case; often referred to as the Euler column. This column case is depicted
in Figure 24. It is pointed out that while the convention is to refer to columns,
such as the one depicted in Figure 24, as pinned-pinned, it is clear that the
loaded end must actually be a roller ; otherwise, the column would never receive
load. Nonetheless, the convention is to refer to columns, like that of Figure 24,
as pinned-pinned.

It is important to enumerate the various assumptions that will underpin the


development that follows. These assumptions are now listed for later reference:
• linear elastic material
• Navier’s hypothesis
• prismatic member
• initially straight member
• concentric loading
The boundary conditions associated with the Euler column are framed math-
ematically as:
• y (0) = 0

40
• y (L) = 0
d2 y(0)
• M (0) = 0 ⇒ dx2

d2 y(L)
• M (L) = 0 ⇒ dx2

We may now take derivatives of our general solution in Equation 43 to facilitate


the evaluation of integration constants:
dy
• dx= (Ak) cos (kx) − (Bk) sin (kx) + C
2
d y 2
 
• dx2 = −Ak sin (kx) − Bk 2 cos (kx)
d3 y
 
• dx3 = −Ak 3 cos (kx) + Bk 3 sin (kx)
2
Using the boundary conditions d dxy(0)
2 = 0, we recognize that B = 0. Subse-
quently applying the condition that y (0) = 0 results in the vanishing of D. We
now observe that our general solution has become considerably simplified:

y (x) = Asin (kx) + Cx (44)

We may now apply our remaining boundary conditions to realize that our bound-
ary value problem has become an eigen-problem. To facilitate the consideration
of the remaining boundary value problem we may take derivatives of Equation
44:
dy
• dx= (Ak) cos (kx) + C
2
d y 2

• dx2 = −Ak sin (kx)
2
From the moment boundary condition, d dx
y(L)
2 = 0, we get that:

−Ak 2 sin (kL) = 0



(45)

Similarly, from the boundary condition y (L) = 0, we obtain:

Asin (kL) + CL = 0 (46)

For convenience, we will now combine Equations 45 and 46 into matrix form:
    
sin (kL) L A 0
= (47)
−k 2 sin (kL) 0 C 0

The trivial solution is easily identified as the case wherein A = C = 0. Unfor-


tunately, this is an uninteresting result, and thus we must go deeper.

Specifically, we are looking for a solution other than the trivial solution
for our system of linear algebraic equations in Equation 47. Such a solution
only exists if our coefficient matrix is singular (thus implying that it cannot be

41
inverted). A gage of singularity in a square coefficient matrix such as the one
in Equation 47, is the determinant. As you recall, the definition of a matrix
inverse, for an arbitrary square matrix [Ω] can be given by:

−1 adj [Ω]
[Ω] = (48)
det [Ω]

In Equation 48, you will recall that adj [Ω] is the adjoint of Ω (i.e. the transpose
of the matrix of cofactors of Ω), and det [Ω] simply denotes the determinant of Ω.

From Equation 48 it is clear to see that if the determinant vanishes then


the matrix inverse is no longer defined; and we then have a singular matrix.
Consistent with this notion, we now solve the following determinant:

sin (kL) L
det
=0 (49)
−k 2 sin (kL) 0

Equation 49 may be easily evaluated using the approach provided in Equation


50:
sin (kL) (0) − −k 2 sin (kL) L = 0

(50)
Simplification of Equation 50 leads to:

k 2 Lsin (kL) = 0 (51)

Granted, Equation 51 could be solved by having k = 0. This solution, however,


is as uninteresting as the trivial solution since this k = 0 implies no load on the
column! What we are left with, that is interesting, is the eigen-problem:

sin (kl) = 0 (52)

the eigenvalues of which correspond to kL = nπ, n = 1, 2, 3, ...∞. The case


where n = 1 is the case requiring the least energy and thus is the case that
nature picks; thus leading to:

π 2 EI
PEuler = (53)
L2
Other boundary conditions may be treated through the formulation, and subse-
quent solution, of new boundary value problems. As in the case of the moment
gradient coefficient, Cb , described in Section 3.1.4, AISC has a simplified ap-
proach to the treatment of various boundary conditions within the context of
flexural buckling.

3.2.3 Effective length


It is possible to re-apply the Euler buckling solution to cases of columns with
other than pinned ends through the application of an effective length factor,
K. The notional approach behind the effective length factor is to identify the

42
Figure 25: Effective column lengths - theoretical

equivalent length of Euler column resulting in the same critical buckling load
as the specific case under investigation. This approach is most readily achieved
through an investigation into the locations of the inflection points within the
buckling mode (i.e. elastic curve assumed by the buckled member) of the mem-
ber under consideration.

Figure 25 schematically depicts such a consideration on four common col-


umn cases. It is pointed our that in Figure 25, all columns are assumed to have
the same original length and buckle under the action of a concentrically applied
loading at the column tops. In addition, it is noted that the effective length
factors presented in Figure 25, are theoretical values that are not normally rec-
ommended for design. Rather slight variations on these values are promulgated
by AISC in order to account for such practical considerations as the difficulty
in producing a truly fixed end condition, etc. For actual design applications,
it is recommended that the effective lengths presented in Table C-C2.2 in the
commentary of the Specification be applied. The AISC recommended values
for the effective length factors corresponding to the individual cases presented
in Figure 25 (in the order of appearance, from left to right) are K=1.0, 0.65,
0.8, and 2.10. For the case where multiple members intersect at a joint (i.e. as
within a multi-story framework), the AISC alignment charts may be used.

The alignment charts embody multiple solutions to the governing differential


equations as they apply to the case of a column with elastically restrained ends;
as opposed to the idealized end conditions appearing in Figure 25. The form
of the solution to the governing equations is in the graphical form known as a
nomograph. These nomographs appear in Chapter C of the commentary to the
Specification; immediately following the previously referenced Table C-C2.2. To

43
apply the alignment charts, the designer ascertains the values of the end points
to the nomographs: GA and GB using the expression (for each end: A and B)
ΣEI

L
G= ΣEI
c (54)
L g

In Equation 54, the subscript c denotes columns while the subscript g denotes
girders. Therefore, we observe that the numerator of Equation 54 is made
up of the sum of all column member stiffness contributions converging on the
structural joint under investigation; while the denominator applies in a similar
fashion for all of the girder converging on a given structural joint. Once the
values of G for the A-end and B-end of a given column have been obtained, a
corresponding effective length factor, K can be read of the read from the align-
ment charts by using GA and GB as the end points of a line that occur on the
vertical marked with their respective headings on the right and left sides of the
alignment charts. A value for the corresponding effective length factor, K may
then be obtained by simply “connecting the dots”, and reading off the value of
K where the line connecting GA and GB intersects the K-axis. It is pointed
out that there are two alignment charts provided in commentary Chapter C :
sidesway inhibited and sidesway uninhibited. The first case (sidesway inhibited )
is perhaps more commonly thought of as a braced frame, but the salient point
here is that there is relatively little lateral translation of the upper story joints,
relative to the lower story joints assumed in this case. This is contrast to the
second case (sidesway uninhibited ) in which there is expected to be consider-
able inter-story drift. The sidesway uninhibited case is frequently applied to
unbraced, or rigid frames.

It is pointed out that while useful in many instances, the solutions provided
in the form of the alignment charts is predicated on the satisfaction of a number
of very restrictive assumptions, now given:
• behavior is purely elastic
• all members are prismatic
• all joints are rigid
• for sidesway inhibited frames, rotations at opposite ends of beams are
equal in magnitude; thus producing single curvature
• for sidesway uninhibited frames, rotations at opposite ends of the restrain-
ing beams are equal in magnitude; thus producing reverse curvature bend-
ing (i.e. the beam inflection point occurs at the beam mid-span under the
action of purely lateral loading)
q
P
• the column stiffness measure L EI is identical for all columns in a story

• joint restraint is distributed to the column above and below the joint in
proportion to LI of the two columns

44
Figure 26: Initially crooked Euler column

• all columns buckle simultaneously


• no significant axial compression force exists in the girders
As a result of these restrictive assumptions, it is important to remember that
the alignment charts may only be applied to the case of regular framing in civil
structural applications. The solutions contained in the alignment charts would
not be applicable to the case of the framing of a seating gallery in stadium for
instance. Also as a result of the considerable number of restrictive assumptions
made in the formulation of the alignment charts, they have fallen out of favor
in the more advanced practice of structural engineering. More sophisticated
alternatives to there use will be explored later in the course and in CEE779.

3.2.4 Design for flexural buckling


We are now in a position to use what we have learned in Section 3.2.1 through
Section 3.2.3 to frame a practical design approach for the proportioning of safe
and economical column members. We will use as our point of departure the
behaviors described in an ideal way in Section 3.2.1 and 3.2.3. In the latter
case, it was assumed the the members was initially straight at the outset of
compressive loading. This, of course, is not achievable in practice and thus in
a design context, we must explore what the effect of initial imperfections are
on the strength of columns. In pursuit of this, we may modify Equation 41
to include a term on the right hand side that accounts for the initial displaced
shape of the column. While we will not pursue this case mathematically in the
present discussion, we may consider the effect of initial imperfections on column
behavior in a schematic sense, for an Euler column, in Figure 26.

It is observed from Figure 26, that the effect of the initial imperfection is
to simplify the response of the column member from that of an eigen-problem
to a response that is simply one more classical in terms of a functional rela-
tionship between load and displacement. Indeed, in the imperfect column case,

45
Figure 27: Initially crooked Euler column

we observe that the resulting equilibrium path asymptotically approaches the


bifurcation response of the perfect column. The word bifurcation is applied as
a modifier related to the perfect column response since we observe a fork in the
equilibrium response at the critical load. There are three mathematically dis-
tinct possibilities for load and displacement in the neighborhood of the critical
load: 1) the column could remain straight; 2) the column could buckle to the
left; or 3) the column could buckle to the right. We can now understand how the
lack of existence of an inverse for the coefficient matrix in Equation 47 makes
sense in light of the present discussion on behavior. The existence of an inverse
for the coefficient matrix in Equation 47 would have implied a unique functional
relationship at the buckling load; something that we do not have in the case of
the perfect column, but do have in the case of the imperfect column. The effect
of a non-concentric loading on the load-deflection response is similar in nature
to what is observed for the initially crooked case. In both of these instances,
the load at which relatively large lateral displacements occur is somewhat less
than the critical buckling load computed using idealized column geometries.

While initial column crookedness and presence of non-concentric loading is


important to consider in practical column design approaches, so too is the occur-
rence if material nonlinearities. Indeed, the presence of residual stresses create
conditions where inelastic buckling response dominates for most practical col-
umn lengths encountered in engineering practice. In light of these mitigating
influences, it is useful to create an idealized column design paradigm and then
modify it to account for such practical matters as those that have been discussed
thus far.

If one did not think of such practical matters initially, one may have arrived
at a reasonable envelope for column designs as being one that is demarcated

46
by the squash load (in red) and the Euler solution (in blue). While justifiable
in a theoretical sense, such an approach proves inadequate when one superim-
poses experimental results from column tests, performed by various researchers
around the world, on the idealized curved presented in red and blue in Figure
27. Instead of the individual test results clustering around the red and blue
lines, the data instead cluster around the black line which essentially represents
a curve fit by AISC to the experimental results. The design equations (E3-1)
to (E3-3) describe the curve shown in black in Figure 27. In all fairness, these
equations are not pure curve fits to the experimental data. Rather they are
semi-empirical in nature; taking a theoretical basis as a point of departure for
the curve fitting (e.g. the modified solution of Equation 47 alluded to earlier).

A fairly simple example of this last notion presents itself through a consid-
eration of equations E3-3,4 in the Specification; reproduced below as Equations
55 and 56:
Fcr = 0.877 (Fe ) (55)
π2 E
Fe = (56)
KL 2

r
In the form of this presentation of the design design equation governing the
elastic flexural buckling of a hot-rolled column member, it is straight forward to
observe that the classical Euler stress given by Equation 57,
π 2 (EI)
s
PEuler 2 I π2 E
FEuler = = L →r= →= (57)
Ag Ag Ag L 2

r

is precisely what appears in Equation 56 (except for the addition of the effective
length factor, K, appearing in the design equation for cases that are other than
pinned-pinned. The nominal strength given by Equation 55 is simply a linear
reduction of the Euler stress; in this case, to account for an initial imperfection.
The imperfection field assumed by AISC (thus leading to the 0.877 reduction)
is that of a sine wave scaled such that the maximum out-of-straightness occurs
L
at mid-height of the column; assuming a value of 1000 at that location.

3.2.5 Torsional buckling


We now take up the case where global buckling manifests itself in a fashion
that preserves the length and straightness of the longitudinal axis and thus the
deformation occurs as a twisting about the straight longitudinal axis. An illus-
tration of this mode is provided in Figure 28 as it pertains to a column with a
cruciform cross-sectional geometry.

In considering the case depicted in Figure 28, we will build from knowledge
that we have gained in treating lateral-torsional buckling in beams, and flexural
buckling in columns. Specifically, we recall that in the case of a column, we may

47
Figure 28: Torsional buckling in a column with cruciform cross-section

48
reproduce the column buckling equation in terms of moment:

d2 M d2 y
EI + P =0 (58)
dx2 dx2
This form of the column buckling equation may also be thought of as simply
being the differential equation of flexure wherein the “transverse loading func-
d2 y
tion” is of the form −P dx 2 . We may now imagine that a differential strip along

the Z-axis, of our cruciform column, behaves in an analogous manner to the


strut described by Equation 58 in that a sort of transverse force is acting at
the location of the differential material element under consideration in Figure
28 (employing the relationship: y = rφ). This elemental transverse force acts
through a moment arm, r. We can describe this mathematically as:

d2 (rφ)
Ptransverse = − (σz tdr) (59)
dz 2
rd2 φ
dmz = − (σz trdr) (60)
dz 2
where Ptransverse is the fictitious transverse load acting along a notional strut in
the cross-section buckling about the longitudinal axis, and dmz is the equivalent
distributed torque developed by the fictitious transverse load acting about the
longitudinal axis.

We may now sum all such contributions, in order to arrive at an aggregate


torque associated with the given length of column, by integrating over the cross-
sectional area at the Z-axis location under investigation:

d2 φ
Z
mz = −σz 2 dz r2 tdr (61)
dz A

As it is that the expression of torque in Equation 61 represents a torque-per-


differential length, dz, we may re-express this notional distributed applied torque
as a differential:
dMz = mz (62)
This observation proves useful when we recall the form of the governing equation
for torsion that was presented in Section 3.1.4, as Equation 27. If we differentiate
Equation 27 one time we obtain:

dMz d2 φ d4 φ
= GJ 2 − ECw 4 (63)
dz dz dz
As a result of the appearance of the torsional deformation in Figure 28, a minus
sign must be used when substituting the result from Equations 61 and 62 into
Equation 63, thus resulting in:

d2 φ d2 φ d4 φ
Z
σz 2 r2 tdr = GJ 2 − ECw 4 (64)
dz A dz dz

49
Equation 64 may be placed in a slightly more convenient form (to facilitate
solution):
d4 φ 2
2d φ
+ ψ =0 (65)
dz 4 dz 2
where, s R
σz A r2 tdr − GJ
ψ= (66)
ECw
The solution Equation 65 can be expressed as:

φ = Asin (ψz) + Bcos (ψz) + Cz + D (67)

We may now take derivatives of our general solution in Equation 67 to facilitate


the evaluation of integration constants:

• dz = (Aψ) cos (ψz) − (Bψ) sin (ψz) + C
d2 φ
 
• dz 2 = −Aψ 2 sin (ψz) − Bψ 2 cos (ψz)
d3 φ
 
• dz 3 = −Aψ 3 cos (ψz) + Bψ 3 sin (ψz)
The boundary conditions pertaining to the case depicted in Figure 28 my be
expressed as:
• φ (0) = 0; twist at the column top is restrained
• φ (L) = 0; twist at the column bottom is restrained
d2 φ(0)
• dz 2 ; warping is unrestrained at the column top
d2 φ(L)
• dz 2 ; warping is unrestrained at the column bottom
From the first and third of these, we obtain the result that B = D = 0 and
thus our solution is simplified to facilitate further consideration of boundary
conditions:
φ = Asin (ψz) + Cz (68)
Consideration of Equation 68, in conjunction with the two remaining boundary
conditions leads us to the evaluation of the eigenvalue buckling problem that
involves the evaluation of determinant below:

sin (ψL) L
−ψ 2 sin (ψL) 0 = 0 (69)

Evaluation of this determinant leads to the result:

ψ 2 Lsin (ψL) = 0 ⇒ sin (ψL) = 0 ⇒ ψL = nπ (70)

50
Consideration of Equation 70 in light of Equation 66 leads to the following
expression for the buckling stresses (eigenvalues):
ECw n2 π 2
+ GJ
σzeigenvalues = R2
L
(71)
2
r tdr
A

Recognizing that
R that the critical condition arises out of n = 1, and also rec-
ognizing that A r2 tdr represents the polar moment of inertia about the shear
center of the cross-section, Ip , we may write:
 2 
π ECw 1
σzcritical = 2
+ GJ (72)
L Ip

We recall that the polar moment of inertia, J, is analogous to Ip except that


is is integrated about the centroid, as opposed to the shear center. It is noted
that in the present development, the shear center and centroid coincide.

Recognizing that Ip = Ix + Iy , and switching our variable name for stress,


we are able to reproduce equation (E4-4) from the Specification:
" #
π 2 ECw 1
Fe = 2 + GJ I + I (73)
(Kz L) x y

Within the denominator of the first term within the braces of Equation 73, an
effective length factor appears as Kz . In this application, the effective length
factor modifies the column length considered; so as to account for differences in
warping restraint. If the ends are free to warp, then Kz = 1.0. If both ends are
fixed with regard to warping then Kz = 0.5. Alternately, if one end can warp
and the other cannot, then Kz = 0.7. Within the specification, an equation de-
scribing torsional buckling in cases of single symmetry (where the shear center
and centroid do not coincide) appears as equation (E4-5).

In order to put what have learned into context, it is important to point out
that for most building and bridge applications, involving hot-rolled structural
elements, it is flexural buckling that is most common. However, there are a few
cross-sections in the manual that are susceptible to torsional buckling under the
right circumstances (i.e. the torsional buckling load is lower than the flexural
buckling load, for the same length column). Even in these instances, however,
the torsional buckling load is within a few percent of the critical flexural load,
and thus a design based on the flexural buckling mode would have been fine
(albeit by fiat). Another important example to consider is a wide flange member
whose unbraced length in torsional buckling is greater than its out-of-plane
unbraced length. For example it is not difficult to imagine a case where counters
are used to brace a 20 foot W16x26 column by passing through the web at mid-
height of the column. While it is that such a bracing scenario would be effective
in restraining the motion associated with flexural buckling at the mid-height, the

51
Figure 29: Arbitrary unsymmetrical cross-section

same is not true with regards to the deformation associated with the torsional
mode. Indeed, the unbraced length in torsion would be the full 20’. The net
result is that the torsional buckling load would approximately 85 percent of
the weak axis flexural mode. The moral of the story is: you must be aware
of the nature of bracing used in structural systems. Bracing that is useful in
resisting one mode of buckling may be nearly completely ineffectual in resisting
another mode. Many a failure and death has resulted from failure to consider
such important details related to stability and bracing.

3.2.6 Flexural-torsional buckling


In the foregoing discussion on flexural and torsional buckling, the scope of the
members considered was essentially that of doubly and singly symmetric cross-
sections acting as compression members. When the cross-section under con-
sideration is unsymmetrical, a more general case of flexural-torsional buckling
must be investigated (singly symmetric cross-sections may also experience this
mode of buckling.)

Figure 29 displays a schematic view of an admissible deformation scenario


related to an unsymmetrical cross-section. It is observed that any such deforma-
tion would likely involve displacements of the shear center that are resolvable
in terms the coordinate axes shown. However, an additional displacement of
the centroid will arise due to cross-section rotations occurring about the shear
center :
u + yo φ (74)
and
v − xo φ (75)
Considering now the implication of of these kinematic considerations on the
nature of the differential equation governing flexural buckling (appearing as

52
Equation 42 in the foregoing) we now obtain two equations:
d2 u
EIy + P (u + yo φ) = 0 (76)
dz 2
d2 v
EIx + P (v − xo φ) = 0 (77)
dz 2
We may make a similar modification to Equation 59 for torsional buckling by
realizing that the torsional deformation of the cross-sectional strip considered
when deriving Equation 59 now become more complicated expressions describing
the motion in the x- and y-directions respectively:
u + (yo − y) φ (78)
v − (xo − x) φ (79)
As a result of Equations 78 and 79, we now replace Equation 59 by the more
general expressions below (as part of our development of the equations governing
flexural-torsional buckling).
d2
− (σtds) [u + (yo − y) φ] (80)
dz 2
d2
− (σtds) [v − (xo − x) φ] (81)
dz 2
Consideration of Equations 80 and 81 lead to the following expression for the
contribution to the twist per unit length obtained from a differential cross-
sectional element (analogous to Equation 60):
 2
d2 φ
 2
d2 φ
 
d u d v
dmz = − (σtds) (yo − y) + (y o − y) +(σtds) (xo − x) − (xo − x)
dz 2 dz 2 dz 2 dz 2
(82)
We may now integrate this equation to obtain the following expression for twist
per unit length:
d2 v d2 u Ip d2 φ
 
mz = P xo 2 − yo 2 − P (83)
dz dz Ag dz 2
Considering now an extension of Equation 63, we obtain our final equation in
the form that governs flexural-torsional buckling:
d4 φ
 2
d2 v d2 u

Ip d φ
ECw 4 − GJ − P − P xo + P y o =0 (84)
dz Ag dz 2 dz 2 dz 2
where Ag is the gross cross-sectional area of the member.

We now have our governing equations for flexural-torsional buckling: Equa-


tions 76, 77, and 84. These three simultaneous differential equations describe
the flexural torsional mode.

Consideration of simple boundary conditions at the members ends (i.e. at


Z = 0 and Z = L), leads to the following:

53
• u (0) = 0, x-direction translation is zero at top
• v (0) = 0, y-direction translation is zero at top
• φ (0) = 0, z-direction rotation is zero at top
• u (L) = 0, x-direction translation is zero at bottom

• v (L) = 0, y-direction translation is zero at bottom


• φ (L) = 0, z-direction rotation is zero at bottom
d2 u(0)
• dz 2 = 0, moment about the y-axis is unrestrained at the column top
d2 v(0)
• dz 2 = 0, moment about the x-axis is unrestrained at the column top
2
d φ(0)
• dz 2 = 0, warping is free at the column top
d2 u(L)
• dz 2 = 0, moment about the y-axis is unrestrained at the column bottom
d2 v(L)
• dz 2 = 0, moment about the x-axis is unrestrained at the column bottom
d2 φ(L)
• dz 2 = 0, warping is free at the column bottom
These boundary conditions will be satisfied if solutions to Equations 76, 77, and
84 take the forms of Equations 85, 86, and 87 below:
 πz 
u (z) = Asin (85)
L
 πz 
v (z) = Bsin (86)
L
 πz 
φ (z) = Csin (87)
L
We may now take derivatives of the solution elements presented in Equations
85 though 87:
• du π πz

dz = A L cos L

• dv π πz

dz = B L cos L

• dφ π πz

dz = C L cos L
2
π 2
• ddzu2 = −A L sin πz
 
L
2
π 2
• ddzv2 = −B L sin πz
 
L

d2 φ π 2 πz
 
• dz 2 = −C L sin L

54
Substitution of the first and fourth result above, into Equation 76, leads to:
 π 2  πz   πz 
−EIy A sin + P Asin + P yo C = 0 (88)
L L L
which may be further simplified:
  π 2 
−EIy + P A + P yo C = 0 (89)
L
Similar consideration of Equations 77 and 84 lead to the following:
  π 2 
−EIx + P B + P xo C = 0 (90)
L

π2
 
Ip
P yo A − P xo B − ECw 2 + GJ − P C=0 (91)
L Ag
As before, we seek nontrivial solutions to the system of linear homogeneous alge-
braic equations 89 through 91 and thus will need to solve an eigenvalue problem.

In order the simplify the treatment of the eigenvalue buckling problem asso-
ciated with our treatment of flexural-torsional buckling, we may introduce the
following expressions for use in framing the subsequent determinate:

π 2 EIx
Px = (92)
L2
π 2 EIy
Py = (93)
L2
π2
 
Ag
Pφ = GJ + ECw 2 (94)
Ip L
It is pointed out that Equations 92, 93, and 94 represent, respectively, the flex-
ural buckling load about the x-axis, the flexural buckling load about the y-axis,
and the torsional buckling load.

Using Equations 92 through 94 within Equations 89 through 91, we may


formulate our determinate needed to arrive at the buckling loads in the flexural-
torsional mode:
(P − Py )

0 P yo
(P − Px ) −P xo

0 =0 (95)
Ip
P yo −P xo (P − P φ)
Ag

Expansion of the determinate in Equation 95 leads to the following cubic equa-


tion; the lowest root of which is the critical buckling load for the member:
Ip
(P − Py ) (P − Px ) (P − Pφ ) − P 2 yo2 (P − Px ) − P 2 x2o (P − Py ) = 0 (96)
Ag

55
The result from Equation 96 may be cast in terms of stress; yielding a form
identical to equation (E4-6) in the Specification (reproduced as Equation 97
below - for an unsymmetrical cross-section):
 2  2
xo yo
(Fe − Fex ) (Fe − Fey ) (Fe −Fez )−Fe2 (Fe − Fey ) −Fe2
(Fe − Fex ) =0
r¯o r¯o
(97)
where Fex and Fey are the Euler buckling stress considering the major and minor
centroidal axes, respectively. Additionally, xo , yo , and r¯o are the distance from
the centroid to the shear center and the polar moment of inertia about the shear
center, respectively. Fe is the flexural-torsional buckling stress being sought.
A specific example where flexural-torsional buckling governs over other global
buckling modes occurs in the case of a WT8x28.5, wherein we can compute that
for flexural buckling, Fey = 50.9ksi, for torsional buckling, Fez = 131ksi, and for
flexural-torsional buckling, Fe = 45.3ksi.

3.2.7 Built-up compression members


Built-up compression members are columns comprised of two or more rolled
sections; attached to one another through the use of some form of shear con-
nection. Recall from elementary mechanics that the moment of inertia for to
beams, of a given I, resting on one another is 2I. Conversely, if the beams are
bonded together at their interface, this number increases to 4I. Since column
buckling depends very strongly on cross-sectional stiffness, it is vitally important
to carefully consider the efficiency of shear transfer between the rolled elements
making up the built-up column when determining column capacity

The approach adopted by AISC in this regard is a combination of empirical


and analytical considerations; with the end result being equations (E6-1) and
(E6-2) in the Specification. These two equations have become the center of a
bit of controversy as of late, but this topic is beyond the scope of the present
discussion. For our purposes, we will recognize equations (E6-1) and (E6-2) as
a somewhat empirically based approach to considering the less than perfect ef-
ficiency in shear transfer occurring within built-up members. More specifically,
this discussion will focus on the most common built-up column: the double an-
gle column.

In any application involving built-up columns, it is vitally important to en-


sure that surface preparation of the faying surfaces (i.e. contact surfaces) at
the built-up column ends are carefully considered (as per Section E6.2 within
the Specification). Satisfactory surface preparation of the column ends enables
the application of the following modified effective length expressions; resulting
in column slenderness values greater than that obtained from assuming perfect
shear transfer between members. Now it turns out that equation (E6-1) in the
Specification is highly empirical, and based on only a single study reported in
the literature. Recently (i.e. 2006), another empirical study has essentially cast

56
significant doubt on the validity of (E6-1); showing instead that equation (E6-
2) is valid for both cases: snug-tight and fully tensioned fasteners. As a result,
we will only be considering equation E6-2 in our current discussion.

Equation (E6-2) furnishes a modified column slenderness for the case of


built-up columns (a slenderness greater than that associated with the presence
of perfect shear transfer):
s 2  2
α2
 
KL KL a
= + 0.82 (98)
r m r o (1 + α2 ) rib

In Equation 98, the following variable definitions apply:


• KL

r m ≡modified column slenderness

• KL

r o ≡column slenderness assuming perfect shear transfer

• a ≡distance between connectors (in.)


• ri ≡absolute minimum radius of gyration of individual component (in.)
• rib ≡minimum radius of gyration of individual component relative rela-
tive to its centroidal axis parallel to the overall built-up member axis of
buckling (in.)
h
• α ≡separation ratio≡ 2rib

• h≡distance between centroids of individual components perpendicular to


the member axis of buckling (in.)
As can be surmised from the foregoing, there are certain important geometrical
considerations that must be checked to ensure proper the proper behavior of
the built-up column member

One such critical geometric parameter is related to the distance between


connectors, a. A minimum number of connectors must be present in order to
guarantee that the shear transfer assumed within the context of Equation (98)
(i.e. AISC Specification equation (E6-2)) is available. At least as important
as this previous consideration, is that concerning the occurrence of interactive
buckling. Interactive buckling is a condition that arises when two different buck-
ling modes are activated in a nearly simultaneous fashion within the compression
member under consideration. In such circumstances, the net effect is an ulti-
mate capacity of the column member that is significantly eroded; as compared
with the lesser of the two capacities involved in the interaction buckling (a bit
like a negative synergy). The simplified check that AISC promulgates, as a
means for avoiding this unfortunate effect, is the geometrical limit of section
E6.2 : where the slenderness ratio, Ka ri , of the individual component elements
making up the built-up column is limited to be less than three quarters of the

57
Figure 30: Double angle column example - 2L 6x6x5/8, A36 (assume a 3/8”
gap)

governing column slenderness for the entire built-up column, in aggregate. It is


noted that an approach such as this does not strictly preclude to the possibility
of interactive buckling, but it is a positive step in the correct direction.

3.2.8 Column analysis example


Consider the 8 foot long double angle column below. Compute the column de-
sign strength for this member.

We can start off by listing some useful section properties that will be used
in subsequent calculations:
• Ag = 14.3in2 from T1-15 in the Manual
• rx = 1.84in from T1-15 in the Manual
• ry = 2.65in from T1-15 in the Manual
• rib = 1.84in from T1-7 in the Manual
• r¯o = 3.52in from T1-15 in the Manual

58
• H = 0.840 from T1-15 in the Manual
• ȳ = 1.72in from T1-7 in the Manual
• J = 0.955in4 from T1-7 in the Manual
As this is a singly symmetric built-up section, it will be important to consider
all three of our global buckling modes when evaluating the design strength of the
member. We begin along these lines with a consideration of flexural buckling:
x-x buckling:  
KL 1.0 (96in)
= = 52.17 (99)
r x 1.84in
y-y buckling (reference Equation 98 above):
v 2 2
   2 
KL u 1.0 (96in) (1.0367) 96in/4
u
=t + 0.82   = 37.2
r m y 2.65in 1 + (1.0367)
2 1.84in
(100)
In equation 100, the quantity α was employed. It was defined as follows:
h 2 (1.72in) + 3/8in
α= = = 1.0367 (101)
2rib 2 (1.84in)
Based on a consideration of the foregoing, it is clear that buckling about the x-x
axis controls the case of flexural buckling and thus we proceed according to the
following:
x-x flexural buckling: Evaluate the nature of the flexural buckling using
Section E3 of the Specification.
s
E
52.17 < 4.71 = 133.68 (102)
Fy

thus, inelastic flexural buckling about the x-x axis governs for the flexural buck-
ling strength and thus we apply Specification equation E3-2 :
h 36ksi
i
Fcrx = 0.658 105.2ksi 36ksi = 31.2ksi (103)

Within Equation 103 the Euler buckling stress used in the exponential is com-
puted as:
π2 E π2 E
Fe = 2 = 2 = 105.2ksi (104)
KL

r
(52.17)
and thus our design compressive strength for the case of flexural bucklingis:

(φPn )y−f lexural = 0.9 (31.2ksi) 14.3in2 = 401.5kips



(105)

We now consider the remaining two global modes in turn.

59
torsional buckling:

GJ 11200ksi 2 · 0.955in4
Fcrz = = 2 = 120.73ksi (106)
Ag r̄o2 14.3in2 (3.52in)

and thus our design compressive strength for the case of torsional bucklingis:

(φPn )z−torsional = 0.9 (120.73ksi) 14.3in2 = 1553.8kips



(107)

flexural-torsional buckling:
 " s #
Fcry + Fcrz 4Fcry Fcrz H
Fcr−f t = 1− 1− 2 (108)
2H (Fcry + Fcrz )

Based on the form of Equation 108 it is clear that despite the fact that x-x
flexural buckling governs for the flexural mode, the y-y buckling stress in the
flexural mode is nonetheless needed for use in computing the flexural-torsional
buckling strength, and thus we proceed as:
"  # s
KL E
= 37.2 < 4.71 = 133.68 (109)
r y Fy
m

thus, inelastic flexural buckling is germane when considering the y-y axis re-
sponse, and so we apply Specification equation E3-2 :
h 36ksi
i
Fcry = 0.658 206.8ksi 36ksi = 33.47ksi (110)

Within Equation 110 the Euler buckling stress used in the exponential is com-
puted as:
π2 E π2 E
Fe = = 2 = 206.83ksi (111)
KL 2

r
(37.2)
substituting...
" s #
33.47ksi + 120.73ksi 4 (33.47ksi) (120.73ksi) (0.84)
Fcr−f t = 1− 1− 2 = 31.67ksi
2 (0.840) (33.47ksi + 120.73ksi)
(112)
and thus our design compressive strength for the case of flexural-torsional buck-
ling is:

(φPn )f lexural−torsional = 0.9 (31.67ksi) 14.3in2 = 407.6kips



(113)

We may compare this result with the value listed in the column load tables
on p. 4-119 in table T4-8 in the Manual, where φPn is given as 400kips; a value
that compares quite well with our design capacity of 401.5 kips.

60
3.3 Beam - columns
When considering the strength of a steel member under the combined effects
of axial thrust and flexure, two general approaches have historically been used:
design charts and tables of safe moment-thrust combinations; and so-called in-
teraction expressions that are, in principle, based on the formulaic representa-
tion:   !
Prequired Mrequired
f , ≤ 1.0 (114)
Pprovided Mprovided X,Y

Where P denotes axial load, and M denotes moments about the X-axis (ma-
jor principal centroidal axis or strong-axis) and the Y-axis (minor principal
centroidal axis or weak-axis). It has been the latter approach, the interaction
equation method, that has emerged as the dominant approach for design within
the context of modern building specifications though out the world.

As an initial approach in the formulation of a reasonable base-line interac-


tion expression, Equation 114 may be restated in a commonly accepted form;
serving as the point of departure for essentially all specifications that employ
the interaction approach (see Equation 115).

Prequired Mrequired
+ ≤ 1.0 (115)
Pprovided Mprovided

In Equation 115, Pprovided denotes the pure column strength of the member be-
ing considered (i.e. its axial capacity in the absence of moment) and Mprovided
denotes the capacity of the same member under pure bending (i.e. its flexural ca-
pacity in the absence of axial loading). The quantities Prequired and Mrequired
denote the respective applied loads acting in combination on the member in
question. The capacity of a given member may be obtained from experimental
testing, but frequently such an approach is cumbersome due to the fact that an
extremely large potential design space must be explored; involving such vari-
ables as: member cross-sectional proportions, moment-thrust ratios, unbraced
length, etc.

In practice, it has been commonplace to employ numerical means for the


identification of failure in members loaded by a combination of axial force and
moment. Indeed, raw cross-sectional capacity may be treated through the con-
sideration of a fiber analysis. A fiber analysis is a means by which a zero-length
member, of given cross-sectional geometry, is studied by assigning a finite num-
ber of uniaxial fibers to fixed coordinates within the cross-section. These fibers
are assigned idealized stress-strain response histories; as would be consistent
with coupon testing in a universal testing machine. With this type of idealiza-
tion, moment and axial force can be varied (in the presence of initial residual
stresses) until an unstable combination of loading is achieved - a failure point
on the interaction line / surface. This technique can be extended to the con-
sideration of members of finite lengths wherein inelastic global instability limit

61
states may also be treated. In addition to inelastic cross-sectional strength and
member stability issues, the amplification of the primary applied moment as a
result of so-called second-order effects (due to the axial force of the member
acting over a moment arm emanating from the lateral deflection induced by
the primary moment) can trace its familiar form in specifications to the case
wherein Equation 115, above, is modified as:
1st
Prequired Mrequired
+   ≤ 1.0 (116)
Pprovided P
Mprovided 1 − required
Pe

1st
where Mrequired is the first-order, or primary, moment applied directly to the
member, and Pe is the Euler load for the member when buckling in the plane of
primary moment is enforced (k=1.0 ). The additional denominator term appear-
ing in Equation 116, as compared with Equation 115, represents a consideration
of member level second order effects; so called P − δ effects in the Specification.

Within the context of beam-columns, it is critically important to investigate


the severity of second order effects arising out of the loading on the member. In
fact, we will see later in the present development that the Specification requires
the consideration of such effects when evaluating the interaction equations pro-
mulgated by AISC. It is important, then, to provide some background into how
these amplification factors arise in practice.

Consider the beam-column member depicted in Figure 31. If we sum mo-


ments about point B in free body depicted in the figure, we can obtain the
following second order ordinary differential equation:

d2 y wx2 wLx
EI 2
+ P y = − (117)
dx 2 2
q
P
Substituting k = EI into Equation 117, we obtain:

d2 y wx2 wLx
2
+ k2 y = − (118)
dx 2EI 2EI
the solution of which is obtained in two parts: the complementary part and the
particular part.
y = ycomplementary + yparticular (119)
The solution for the complementary part is given in the well known form involv-
ing the superposition of two transcendental functions:

ycomplementary = Asin (kx) + Bcos (kx) (120)

while the particular solution is obtained from the method of undetermined co-
efficients.

62
Figure 31: Simply supported beam-column of length L acted on by a transverse
uniformly distributed loading

63
Since the right hand side of Equation 117 is a polynomial, we can assume
the yparticular is a polynomial of the form:

yparticular = C1 x2 + C2 x + C3 (121)

where C1 , C2 and C3 are unknown constants to he treated as follows. We take


our assumed particular solution, and differentiate in accordance with Equation
117:
dyparticular
= 2C1 x + C2 (122)
dx
and
d2 yparticular
= 2C1 (123)
dx2
Substituting these results back into Equation 118, and collecting terms, we
obtain:
d2 yparticular
+ k 2 yparticular = C1 k 2 x2 + C2 k 2 x + 2C1 + C3 k 2
  
(124)
dx2
Upon completion of a term-by-term comparison with the original equation
(Equation 118) we observe that:
w
C1 k 2 = (125)
2EI
wL
C2 k 2 = − (126)
2EI
2C1 + C3 k 2 = 0 (127)
and hence we may solve for the constants:
w
C1 = (128)
2EIk 2
wL
C2 = − (129)
2EIk 2
2C1 w
C3 = − =− (130)
k2 EIk 4
and so,
w wL w
yparticular = x2 − x− (131)
2EIk 2 2EIk 2 EIk 4
Thus, the general solution to Equation 118 appears as:
w wL w
y = Asin (kx) + Bcos (kx) + x2 − x− (132)
2EIk 2 2EIk 2 EIk 4
It is now possible to apply the boundary conditions in order to arrive at the
integration constants A and B.

64
We may consider the kinematic conditions associated alternately with the
end and symmetry conditions:
y (0) = 0 (133)
and,
dy L2

=0 (134)
dx
from which, we obtain
w
B= (135)
EIk 4
and  
w kL
A= tan (136)
EIk 4 2
Thus the solution to our problem may be stated as:
 
w kL w w wL w
y= tan sin (kx) + cos (kx) + x2 − x−
EIk 4 2 EIk 4 2EIk 2 2EIk 2 EIk 4
(137)
Making the following substitution,
kL
µ= (138)
2
we may restate Equation 137 in a more convenient form:

wL4 wL2
     
2µx 2µx
y= 4
tan (µ) sin + cos − 1 − x (L − x) (139)
16EIµ L L 8EIµ2

Hence the variation in moment along the beam-column longitudinal axis follows
as:
d2 y wL2
     
2µx 2µx
m (x) = −EI 2 = tan (µ) sin + cos − 1 (140)
dx 4µ2 L L

We may now use these results to arrive at amplification factors (such as that
employed in Equation 116) arising out of so called second order effects.

We may consider Equation 139, as applied at the mid-span, and after some
simplification arrive at the following expression for the maximum transverse
deflection within our beam-column (Figure 31):
" #
5wL4 12 2sec (µ) − µ2 − 2
 
L
ymax = y = (141)
2 384EI 5µ4

which can be thought of as:


" #
12 2sec (µ) − µ2 − 2
ymax = yo (142)
5µ4

65
In Equation 142, we recognize that the term yo is none other than the mid-span
deflection of a beam acted upon by a uniformly distributed transverse loading.
We can think of the term, yo , as being the first order deflection response. The
subsequent action of the axial compression is to then amplify that first order
deflection; as can be seen from the bracketed term in Equation 142. May recast
this second order amplification term (considering member-level, or p−δ, effects)
through the consideration the following power series expansions:

x2 5 61 6 277 8
sec (x) = 1 + + x4 + x + x + ··· (143)
2 24 720 8064
5 4 61 6 277 8
2sec (µ) = 2 + µ2 + µ + µ + µ + ··· (144)
12 360 4032
Substituting the results from Equations 143 and 144 into the amplification term
within the brackets in Equation 142 yields the following new form of the ampli-
fication term:
5 4 61 6 277 8
  
12 2 + µ2 + 12 µ + 360 µ + 4032 µ + · · · − µ2 − 2
(145)
5µ4
Or, after simplification:

1 + 0.4067µ2 + 0.1649µ4 + · · ·

(146)

So, then
ymax = yo 1 + 0.4067µ2 + 0.1649µ4 + · · ·

(147)
We may now re-expand the term µ as follows:
r s r
kL P L π P π P
µ= = = 2 EI = (148)
2 EI 2 2 πL 2
2 Pe

Using this result in Equation 147, we obtain:


"    2 #
P P
ymax = yo 1 + 1.003 + 1.004 + ··· (149)
Pe Pe

leading to the approximate expression:


"    2 #
P P
ymax ≈ yo 1 + + + ··· (150)
Pe Pe

Or equivalently, " #
1
ymax ≈ yo (151)
1 − PPe
The term in brackets within Equation 151 is the new form of the amplification
factor in relation to the existence of member level second-order effects, P − δ.

66
As it is that this amplification factor is essentially identical to the amplification
factor appearing in the denominator of the second term in Equation 116, we
may be inclined to believe that the amplification applied for moment as well.
In this assumption, we would be correct; as the following development demon-
strates.

Recognizing that the maximum moment occurs at the mid-span of our beam-
column, we may consider a modified form of Equation 140:
wL2 wL2 2 (sec (µ) − 1)
   
L
mmax = M = (sec (µ) − 1) = (152)
2 4µ2 8 µ2
Equation 152 may also be though of as:
 
2 (sec (µ) − 1)
mmax = mo (153)
µ2
where mo represents the first order moment from a simply supported beam
acted up on by a uniformly distributed loading. The so-called p − δ moments
are arrived by applying an amplification; as embodied by the bracketed term
within Equation 153.

We may now reapply our power series expansion for 2sec (µ) from Equation
144:
wL2
 
5 61 4 277 6
mmax = 1 + µ2 + µ + µ + ··· (154)
8 12 360 4032
which can be expressed as:
mmax = mo 1 + 0.4167µ2 + 0.169µ4 + 0.0687µ6 + · · ·
 
(155)
q
Re-expressing µ as π2 PPe , we subsequently obtain the following:
"    2  3 #
P P P
mmax = mo 1 + 1.028 + 1.031 + 1.032 + ··· (156)
Pe Pe Pe

We may expand Equation 156 and apply a slight approximation to obtain:


"    "    2 ##
P P P
mmax ≈ mo 1 + 1.028 1+ + + ··· (157)
Pe Pe Pe

thus leading to
"    " ##
P 1
mmax ≈ mo 1 + 1.028 (158)
Pe 1 − PP e

which can be written as


  
P
1 + 0.028 Pe
mmax ≈ mo  P
 (159)
1− Pe

67
or, finally " #
1
mmax ≈ mo (160)
1 − PPe
Of course, this is exactly what we observe as the amplification factor appearing
in the second term of Equation 116.

It is pointed out that in amplification factors appearing Equations 151 and


160 were developed for the case of a simply supported beam column subjected to
axial compression and transverse loading from a uniformly distributed loading.
As was the case with effective lengths in columns, a modification term is available
for extension of these amplification factors for use in a wider range of problems:
cm
B1 = ≥ 1.0 (161)
1 − PPe1
r

in which the nomenclature of the AISC Specification Chapter C has been adopted:
• B1 → member-level amplification factor for second-order effects
• Pr → required compressive strength
• Pe1 → elastic critical buckling capacity within the plane of bending under
consideration and using K=1.0
Within equation 161, cm is used to treat boundary and loading conditions that
differ from those assumed in the foregoing development. For the case where
the transverse loading is uniformly distributed, but the end conditions are fixed
(instead of pinned), cm assumes a value of 0.85. In the case where there is no
transverse loading, but instead concentrated end moments, then the following
equation applies:  
M1
cm = 0.6 − 0.4 (162)
M2
where the moments, M1 and M2 are the larger and smaller member end moments
arrived at using a first order analysis (the ratio is positive when the moments
induce reverse curvature). In situations that are beyond what is described here,
one may refer to the Commentary of the Specification; specifically Table C-C2.1

We are now in a position where we can actually apply a set of interaction


equations to the evaluation of a rolled steel member. Specifically, we will con-
sider the interaction expressions promulgated by AISC in Chapter H of the
Specification (reproduced below as Equations 163 and 164) for the case if axial
compressive loading in the presence of flexure.
Pr
for Pc ≥ 0.2,
 
Pr 8 Mrx Mry
+ + ≤ 1.0 (163)
Pc 9 Mcx Mcy

68
Figure 32: AISC interaction equations - bending about a single axis depicted

Pr
for Pc < 0.2,
 
Pr Mrx Mry
+ + ≤ 1.0 (164)
2Pc Mcx Mcy
where,
• Pr → required axial compressive strength, kips
• Pc → available compressive strength, kips
• Mr → required flexural strength, kip-in
• Mc → available flexural strength, kip-in
• x, y → subscripts defining major and minor principal axes, respectively
It is instructive to plot the bi-linear failure envelopes described by the AISC
interaction equations reproduced in Equations 163 and 164 (see Figure 32).

We are now in a position where we can work an example. Consider the beam
column depicted in Figure 33. We may tabulate useful information to support
the solution of this problem:
• Fy = 50ksi

69
• Fu = 65ksi
• A = 51.8in2
• Zx = 320in3

• Zy = 163in3
• rx = 6.43in
• ry = 4.02in
Computation proceeds with the use of the column load tables to arrive at the
available column capacity, φPc = 2050kips. We are now in a position to use the
axial loading parameters to identify the interaction equation that governs this
analysis problem:
Pr
= 0.683 ≥ 0.2 → (H1 − 1a) (165)
Pc
We now need to address the need for amplification of the member moments
due to the presence of the member-level second order effects. As a beginning
point, it is observed that no transverse loading is present; however end moments
are present. Noting that the end moments present induce a bi-axial bending
condition, with reverse curvature about each axis, we compute cm factors for
both principal centroidal axes as:
 
M1
cm = 0.6 − 0.4 = 0.6 − 0.4 (1.0) = 0.2 (166)
M2

and the subsequent Euler buckling loads within the respective planes of flexure
are:
2 
π2 E π 2 (29000ksi) (6.43in) 51.8in2
(Pe )x = h i2 A = h  i2 = 21719kips (167)
L in
rx 14f t 12 ft

2 
π2 E π 2 (29000ksi) (4.02in) 51.8in2
(Pe )y = h i2 A = h  i2 = 8489kips (168)
L in
ry 14f t 12 ft

We are now in a position to compute the moment amplification factors ap-


plicable to each of the two axes of of flexure within our example:
cm 0.2
B1x = = = 0.214 < 1.0 ⇒ B1x = 1.0 (169)
1− P
Pe
r
1 − 1400kips
21719kips

cm 0.2
B1y = Pr
= 1400kips
= 0.239 < 1.0 ⇒ B1y = 1.0 (170)
1 − Pe 1 − 8489kips

70
Figure 33: AISC interaction equation example problem (length = 14 feet)

The implication of the results from Equations 169 and 170 is that there are
no significant member-level second order effects present in this member. Thus
Mrx = 200kip − f t and Mry = 70kip − f t.

When considering the flexural capacity of this member, it is noted that the
the member length, L, is less that the maximum permissible unbraced length to
for the attainment of the full plastic capacity of the cross-section and thus:

φFy Zx 0.9 (50ksi) 320in3


φMcx = in
= = 1200kip − f t (171)
12 f t 12 in
ft

φFy Zy 0.9 (50ksi) 163in3


φMcy = in
= = 611kip − f t (172)
12 f t 12 in
ft

Substituting these results into Equation 163 results in:


 
1400kips 8 200kip − f t 70kip − f t
+ + = 0.973 ≤ 1.0 (173)
2050kips 9 1200kip − f t 611kip − f t

Thus we conclude that our W14x175, A992 beam-column is adequate for the
loading depicted in Figure 32.

In the case of axial tension with flexure, the same interaction expressions
apply, but instead of axial compressive loading in the first interaction term, we
now use expressions for required tensile capacity and provided tensile capacity.
Furthermore, the axial tensile forces stave off the occurrence lateral-torsional
buckling due to flexure. AISC accounts for this tensile strengthening effect
within the moment term in the interaction expression by amplifying the moment

71
Figure 34: Tension yielding illustration

gradient factor, Cb , using the following:


s
Pr
1+ (174)
Pey
where,
π 2 EIy
Pey = (175)
L2b

3.4 Tension member yielding


Tension member yielding is the limit state that involves the excessive elongation
of a structural element as a result of plastic flow occurring in the presence of
axial tension. This limit state is assumed to control in regions that are located
away from connections (a location that is usually governed by the limit state of
net section fracture.) See Figure 34 for an example of this failure mode.

The design expression for tension member yielding can be found in Chapter
D of the Specification; reproduces below as Equation 176:
Pn = Fy Ag (176)
where, Pn is the nominal tensile capacity considering the tension yielding limit
state, Fy is the minimum specified yield stress, and Ag is the gross cross-sectional
area (i.e. the area reported in Part I of the Manual ). It is noted that the ca-
pacity reduction factor, φ, associated with this limit state is 0.9

We are now finished with our treatment of global limit states (limit states
involving behavior that manifests itself over significant portions of the structural
element under consideration.)

72
Figure 35: Tension rupture illustration

4 Local response
Local failure modes are those that manifest themselves at smaller scales within
the structural element; oftentimes interacting with and affecting other global
limit states within the same member.

4.1 Net section fracture


The net section fracture limit state is a localized limit state that frequently
occurs within the connection region of steel tension members. Since connection
regions frequently contain stress raisers, it is the maximum strength of the
material (i.e. the ultimate strength) that is considered in the AISC design
equation; reproduced here as Equation 177 (see Figure 35 for an example):

Pn = Fu Ae (177)

In Equation 177, Ae is known as the effective area for the tension member. The
definition of the effective area is provided as:

Ae = U An (178)

wherein An is the net section area, and U is the shear lag coefficient.

The net section area represents the gross section area minus the area of any
bolt holes, etc. It is noted that consideration must be given to the different
areas associated with differing fracture trajectories; considering the interesting
effect of bolt stagger.

Bolt stagger effects arise from the beneficial effects that increased fracture
trajectory length has on the net section fracture capacity of a given tension
member. The bolt stagger effect is accounted for using an empirical approach
that has its origins in a study from 1922, wherein the fracture trajectory is
lengthened by an amount:
s2
stagger → (179)
4g

73
Figure 36: Bolt stagger illustration

where s is the bolt pitch (i.e. distance between fasteners measured along the
direction of the line of action of the tensile force) and g is the bolt gage (i.e.
distance between fasteners in the direction perpendicular to the line of action
of the tensile force). This approach is illustrated in in Figure 36 where it is
observed that the net area associated with fracture plane AB is:
  
1
(An )AB = width − 2 dhole + in thickness (180)
16

and the net area associated with fracture plane CD is:

s2
   
1
(An )CD = width − 2 dhole + in + thickness (181)
16 4g

In general, then, we can say:

s2
An = Ag − Aholes + Σ (thickness) (182)
4g
where,   
1 1
Aholes = db + in + in (n) thickness (183)
16 16
It is pointed out that the first parenthetical term in Equation 183 represents
that hole diameter (i.e. the diameter of the bolt plus a small oversize to per-
mit the bolt’s passage through the hole without interference). Equation 183
assumes that the hole was punched, and thus there is an addition sixteenth of
an inch that is notionally removed for the area calculation. This additional
oversize effect accounts for the material proximal to the hole that is assumed
to be ineffective as a result of damage incurred during punching. In the case of
drilled holes, this additional oversize is not requires when computing the area
of the holes.

Returning now to the shear lag factor, we may illustrate the notion of shear
lag through a consideration of Figure 37.

74
Figure 37: Shear lag illustration

We can think of shear lag as a phenomenon related to the St. Venant effect
(wherein localized stress anomalies are smoothed out as one moves away from
the initiation point of the anomaly). In the case of shear lag, the stress condition
is usually tensile, and the context is partial attachment of the cross-sectional
components. In the case illustrated in Figure 37, we have a gusset plate joining
two I-shaped members through their webs only. As a result, the stress trajec-
tories must traverse a certain length along along the longitudinal axis of the
tension members before that cross section can become fully effective (in the
sense that a uniform tensile stress filed across the entire section is converged
to.) To arrive at the shear lag coefficient, U, We can either apply the general
expression for it (see section D3.3 in the Specification:

U =1− (184)
L
where x̄ is the connection eccentricity (defined as the perpendicular distance, in
inches, from the connection interface to the centroid of the connected member
under consideration), and L is the length (measured along the line of action of
the tension force) of load transfer within the connection. As an alternative to
Equation 184, Table D3.1 in the Specification may be applied.

4.2 Plate buckling


Since we may conceive of rolled and built-up structural shapes as being com-
prised of individual plate components, acting in concert within a structural
cross-section, it is important to understand the behavior of such constituent
plate elements. While a rigorous development of the plate buckling theory is
beyond the scope of this class, we may leverage what we have learned in column

75
buckling as a way to facilitate a notional understanding of the approach used
in arriving at plate buckling formulae.

Consider a flat plate loaded along two opposite edges by a uniform dis-
tributed edge loading. The plate boundary conditions along the loaded edge
may be conceived of as as being consistent with the hinge of a door; in the
sense that rotation about an axis parallel to the edge is unrestrained, while all
other rotations are restrained. We may consider combinations of other types
of boundary conditions along the unloaded edges (which are opposite to one
another).

Given an instance of such combinations of boundary conditions along the


plate edges, in conjunction with the applied uniform edge loading, we may
imagine that at some critical loading, an initially flat plate (of thickness, t) may
deform into a plate buckling mode that is consistent with the subject boundary
conditions. Equilibrium may then be formulated on a free body taken from the
plate; thus resulting in the governing differential equation:

∂4w ∂4w ∂4w ∂2w ∂2w ∂2w


 
1
+2 2 2 + = Nx 2 + 2Nxy + Ny 2 (185)
∂x4 ∂x ∂y ∂y 4 D ∂x ∂x∂y ∂y

where w (x, y) is the plate deflection function, Nx and Ny are the plate edge
loading aligned with the x and y axes, respectively; while Nxy is the in-plane
shearing load. In addition, D is the plate flexural rigidity defined as:

Et3
D= (186)
12 (1 − ν 2 )

We may use the biharmonic operator to express Equation 185 in a shorter form
that you sometimes may see in the literature:

∂2w ∂2w ∂2w


 
4 1
∇ w= Nx 2 + 2Nxy + Ny 2 (187)
D ∂x ∂x∂y ∂y

Now that we have the governing differential equation, we may arrive at a general
solution to it.

Such classical solutions exist; going by names such as Navier’s Solution or


Levy’s Solution - depending on the nature of the unloaded edge boundaries. In
either case, advantage is taken of a Fourier Series’ ability to match certain types
of functions arbitrarily closely; depending on the number of terms taken in the
series. As this ends up being a type of Sturm-Liouville Problem, an egenvalue
problem results from this approach. Solution of the eigenproblem yields the
eigenpair: that is the buckling load and its corresponding buckling mode.

As in the case of column buckling, it is possible to solve a certain generic


plate buckling problem, and the to extend the solution to other cases involving

76
Figure 38: Schematic representation of buckling behavior in rectangular plates

different boundary conditions on the unloaded edges. However, in the case of


plate buckling the situation becomes more complex, in that the aspect ratio of
the plate also comes into play.

Consider the generic plate depicted in Figure 38. The loaded edges are sim-
ply supported and the four curves depicted correspond to cases with various
unloaded boundary conditions. These four cases will be useful in later discus-
sions in the notes. What is provided in Figure 38 are the various plate buckling
coefficients that are required to extend a general buckling solution, to individ-
ual cases with the various unloaded edge boundary conditions and aspect ratios
depicted. This general, and extendable, solution is given as:

π2 E
σcr = k (188)
b 2

12 (1 − ν 2 ) t

in which k is the plate buckling coefficient, E is the elastic modulus, and ν is


Poisson’s ratio.

4.3 Local buckling in steel sections


In structural steel members, the cross-sectional geometries are such that por-
tions of the cross-section may be conceived of as individual plate components.
As might be surmised, for the entire cross-section to be relied upon in strength
calculations associated with global limit states, the individual plate components

77
must remain effective; so as to participate with the whole in resisting the for-
mation of the global limit state.

As a result of such considerations, there are various checks that need to


be made to ensure the validity of the assumed cross-sectional response to the
various global limit states that we have treated thus far in the course. For the
present discussion, it is convenient to separate these concerns along the lines of
flexural compression and pure compression. In the former case, it is the notion
of compactness that is an integral part of a flexural cross-section’s ability to ex-
hibit required structural ductility in conjunction with the formation of a plastic
hinge. In the case of pure compression it is the ability of each of the constituent
plate components to achieve the yield strain of the material, prior to localized
plate buckling, that is at issue. Failure to achieve these requirements leads to
compromised capacities in flexure and compression; thus precipitating the need
for modified theories to treat these more complex conditions. Alternatively, in
the case of design, simply selecting a member with more stocky cross-sectional
plate components will also suffice.

4.3.1 Flexure
It can be recalled from our discussion in Section 3.1.3 of these notes that in or-
der to take advantage of the moment redistribution required for the formation
of a plastic collapse mechanism in a structure, adequate plastic hinge rotation
capacity must be exhibited. In Figure 16 we discussed the need to furnish ad-
equate bracing in our designs in order that the occurrence of lateral-torsional
buckling did not erode the structural ductility (i.e. rotation capacity) of the
member. An analogous condition exists for the case of local buckling of the
cross-sectional plate components in the flexural member. Indeed, the occur-
rence of localized buckling in the compression flange, or the portion of the web
in compression, can lead to unfavorable behavior. As a result, the notion of
cross-sectional compactness is formulated.

Compactness is the quality of a flexural cross-section with regard to its ability


to resist compressive strains, approaching those associated with strain hardening
in steel, without the occurrence of local buckling. Guidelines for proportioning
compact cross-sections are contained in Chapter B of the Specification. These
compactness limits represent maximum cross-sectional plate slendernesses that
can be tolerated before structural ductility is eroded. Specifically, the measure
of ductility at issue in this context is what is known as rotation capacity. A
definition of rotation capacity is furnished in Figure 39. AISC requires that
all flexural members that are part of a continuous structural system, that is to
take advantage of moment redistribution resulting from plastic hinging, must be
comprised of members that are compact: that is able to achieve, and maintain
the full plastic capacity, Mp , without suffering from excessive unloading due to
the occurrence of local buckling in the compression plate components. Adequacy
is gaged in this regard by having a rotation capacity of no less than three, for

78
Figure 39: Definition of rotation capacity in flexural member

all members that are experiencing moment loading by virtue of framing into
the continuous system under investigation. The requirements for compactness
also extends to simple beams when it is that a plastic hinge is to be formed
somewhere in the span. This minor notional inconsistency is tolerated in return
for simplicity in design (i.e. strictly speaking no rotation capacity is required of
a simple beam since incipient hinge formation is also incipient collapse within
the flexural member.)

4.3.2 Compression
In the case of column members, we had made assumptions regarding the nature
of the governing global response. Specifically, Figure 27 was used to depict the
three regimes parsing the global buckling limit states in compression members:
1) squash load; 2) inelastic buckling; and 3) elastic buckling. In order for this
assumed response to remain in force, and for all of the subsequent theoreti-
cal underpinnings to remain useful, the effectiveness of the entire compression
cross-section must be ensured. Meaning that all constituent cross-sectional plate
components must be relied on to form the effective cross-section; assumed to
be valid when developing capacities associated with the governing column limit
states.

Within the context of this discussion, it is important to realize that, unlike


the case of flexural compression, significant straining is not germane to the
consideration of global column limit states. Since issues such as plastic hinge
rotation capacity are not at issue in column members, we may focus on the

79
development of the yield strain, rather than the strain hardening strain treated
in beams.

4.3.3 Design including local buckling effects


When considering design approaches useful for the consideration of local buck-
ling vis − á − vis global buckling conditions, it is instructive to understand how
it is that AISC approaches this important design imperative. The subsequent
discussion follows this thread; thus lending insight into how the American Spec-
ification works, while at the same time providing context for the understanding
of other design specifications used throughout the world.

AISC takes as a point of departure the general classification of plate com-


ponents as being either stiffened, or unstiffened. In the latter case, only one of
the unloaded edges experiences some sort of restraint. Within the context of
what was presented in Figure 38, the unstiffened case corresponds to the lower
two curves in the plot (i.e. k = 0.435 and 1.277, respectively). Conversely,
the stiffened case encompasses those plate instances where restraint is provided
along both unloaded edges(i.e. k = 4.00 and 6.97). Practical examples of these
classifications would be the flange outstand of an I-shapes member as an un-
stiffened component, while the web of the same member would be an example
of a stiffened cross-sectional plate component. The current discussion will make
use for these definitions supporting plate classification for the purposes of design.

We begin our practical discussion of plate response in compression with a


useful definition of non-dimensional plate slenderness:
Fcr 1
= 2 (189)
Fy λc

Applying Equation 188 to the foregoing, results in a useful expression for


plate slenderness: r
b 12Fy (1 − ν 2 )
λc = (190)
t π 2 Ek
Using the plate slenderness defined in Equation 190, we may study the be-
havior of plates in compression, with the help of Figure 40.

As in the case of the global response of column members, we are once again
able to divide the plate compression response into three regimes associated with
a fully yielded condition, a classical elastic buckling solution, and an inelastic
transition region that connects the two. In the development depicted in Fig-
ure 40, a residual stress state is assumed with a peak intensity of 0.5Fy . For
the case of a flat plate, these three response regimes become: strain hardening,
inelastic buckling, and elastic post-buckling. These three designations emanate
from a recognition of the important response feature to consider at the ultimate
load of the plate. In the instance of strain hardening, the plate is sufficiently

80
Figure 40: Summary of behavior for plates in compression

stocky to enable the development of strains exceeding strain hardening at incip-


ient buckling. At greater slendernesses, elastic plate buckling governs, and thus
advantage may be taken of the practical post-buckling strength inherent in the
flat plate.

If we consider the response depicted by the solid lines in Figure 40 as the


response to be considered for design, then we may cast the cross-sectional plate
buckling design imperative as:

(Fcr )cross−sectional−plate−component ≤ (Fcr )global−buckling−response (191)

In support of this design model, the following three cross-sectional classifications


are employed by AISC :
• compact → λ ≤ λp

• non-compact → λr > λ > λp


• slender compression element section → λr ≤ λ
where λ is the plate slenderness ratio defined by the width-to-thickness ratio b/t.

For the case of flexural compression, all three of the foregoing cross-sectional
classifications have relevance. For the case of a columns, only the final of the
three is germane. This particular plate slenderness classification corresponds

81
with the case where the plate slenderness is such that incipient plate buckling
occurs at no less than the yield stress, which can be expressed as:

Fcr ≥ Fy (192)

kπ 2 E
≥ Fy (193)
b 2

12 (1 − ν 2 ) t
Since we are considering steel, then we may adopt the usual values for Pois-
son’s ratio (i.e. ν = 0.3) and the elastic modulus (i.e. E = 29000000psi)
within Equation 195 to obtain an expression useful in defining cross-sectional
slenderness limits: s
b k
≤ 5120 (194)
t Fy
It is pointed out that Equation 194 takes the yield stress, Fy , to be in psi. It
is observed that this point should theoretically correspond to the ordered pair
(1,1) on the plot in Figure 40. However, in practice the unavoidable occurrence
of residual stresses and initial out-of-flattness leads to some erosion in plate
capacity. As a result of these influences, AISC arbitrarily decided select a point
to the left of the point (1,1), that is roughly midway between this point and the
point at which incipient strain hardening commences (i.e. λ = 0.7) Thus we may
arrive at the AISC slender plate element limits from the following expression of
plate slenderness: s s
b k k
≤ 0.7 (5120) = 3580 (195)
t Fy Fy
Once again, we are using a yield stress given in psi in Equation 195. We may
now consider various practical examples of slender compression element limits:
• unstiffened element: single angle leg → using k = 0.425 in Equation 195
2340 2400
yields √ ≈√ = √ 76
Fy Fy Fy−ksi

• unstiffened element: flange → using k = 0.70 (arbitrarily selected to be


one third of the way between the pinned condition and full fixity along
3000
the restrained, unloaded edge) in Equation 195 yields √ = √ 95
Fy Fy−ksi

• unstiffened element: stems of WT → using k = 1.277 in Equation 195


4050 4000
yields √ ≈√ = √ 127
Fy Fy Fy−ksi

• stiffened element: webs → using k = 5.00 (arbitrarily selected to be one


third of the way between the pinned condition and full fixity along the
8010 8000
restrained, unloaded edges) in Equation 195 yields √ ≈√ = √ 253
Fy Fy Fy−ksi

Returning now to the case of flexural compression, we may treat the case
the case of compactness in the constituent cross-sectional plate elements com-
prising a beam element. In this case, research by Yura, Galambos, and Leigh

82
has indicated that in order to attain a rotation capacity of three (identified
as the target for structural ductility in non-seismic applications as a result of
Yura’s work), compressive strains in the cross-sectional plate components must
be resisted, without local local buckling, up to levels approaching the strain
hardening strain for mild carbon steel. As a result, it is λ values in the vicinity
of 0.5 that are of interest in the present discussion.

Focusing first on the case related to unstiffened elements, the plate slender-
ness of interest is λc = 0.46; thus we obtain:
s
b k
≤ 2350 (196)
t Fy

where Fy is in psi. Letting k = 0.45; it’s lowest value (as depicted in Figure 38),
we obtain:
b 48.5
≤p (197)
t Fy−ksi
In actuality, AISC somewhat arbitrarily liberalizes the strict requirement
from Equation 197 to acknowledge the fact that flexural compressive strains
may, in reality, be slightly less than the worst case strain hardening values.
As a result, AISC promulgates the following equation in Chapter B of the
Specification: s
b 65 E
≤p = 0.38 (198)
t Fy−ksi F y

In the case of stiffened elements in flexural compression, the λ of interest is 0.58


and thus we have: s
b k
≤ 2965 (199)
t Fy
where Fy is in psi. Letting k assume its lowest value of four (which is very
conservative since the flange-web junction is not a pinned edge) we thus obtain
the AISC equation: s
b 187 E
≤p ≈ 1.12 (200)
t Fy−ksi Fy
We now have developed the important plate slenderness limits, as contained in
the Specification.

5 Simple connections
Properly functioning connection elements are critical to the satisfactory per-
formance of any structural system. As an acknowledgment of this importance,
the AISC promulgates design approaches for connection regions that result in
a higher reliability index, β, as compared with that of individual structural

83
elements within the system. This higher target reliability is adopted as a recog-
nition of the vital importance that connections have, as a place where structural
elements are joined together, in ensuring the overall integrity of the structural
systems. Generally connections employ one of two types of fastening approach:
1) bolts and 2) welds. We begin our discussion on connections with a treatment
of these joining techniques.

5.1 Bolts
The two most commonly encountered bolts in civil structural applications are
the ASTM A325 bolt and the ASTM A490 bolt. The former is more common
in building applications, while the latter is more commonly applied to bridge
applications. The two bolts differ in composition and subsequent strength:
• A325 - quenched and tempered medium carbon steel
• A490 - quenched and tempered alloy steel with higher carbon content
than A325 bolts
As a result of the different steel compositions, it is possible to galvanize A325
bolts, but not A490 bolts since in the latter case a phenomenon known as
hydrogen embrittlement becomes a concern. In the case of both bolt grades, an
ASTM A563 nut is specified with the same dimension as the head of the bolt.
However, in the case of an A325 bolt, Grade C is specified for the nut, while
Grade DH is specified for use with A490 bolts. The use of hardened washers
is required under the the turning portion of the fastener (i.e. the head or the
nut) so as to provide a hard smooth surface for the fastener to turn against.
When A490 bolts are used to join lower grade base metal parts, washers may
be required under both the nut and the head.

5.1.1 Bearing connections


Bearing connections get their name from the fact that the bolts making up
the connection bear against the connected piece in order to effect shear transfer
within the connection region. As a result, it is important to check two governing
limit states in this type of connection: bolt shear strength and bearing strength
in the connected pieces (see Figure 41).

The design equation predicting the nominal bolt shear strength is given as:

φRn = φFnv Ab (shear − plane − number) (bolt − number) (201)

In Equation 201, φ is 0.75, Fnv is the nominal bolt shear strength from TJ3.2 in
the Specification and Ab is the bolt cross-sectional area. in addition, the number
of plane and bolts refer to the number of connection plies within the connection,
and the number of total bolts used to connect these plies, respectively.

84
Figure 41: Bearing connection schematic

Requirements for proper spacing of bolts and proper edge distances are con-
tained in Specification Sections 3, 4, and 5. In addition, the bearing strength
at holes is treated in a relatively empirical way by AISC, as can be seen in
Specification Section J10.

5.1.2 Slip critical connections


Oftentimes slip critical connections are employed in instances where the con-
nection being designed might be susceptible to fatigue failure. Since slip critical
connections rely on bolt pre-tension force to clamp the connection plies together,
shear forces are developed to transfer service loads across the connection inter-
face (see Figure 42). This is helpful in a connection that may be susceptible
to fatigue since it permits the loading / un-loading cycles to be resisted by a
friction-induced shear transfer across the interface; rather than by reversing the
bearing state of stress on the bolt (a condition that could lead to premature
bolt failure).

It is important to realize that in slip critical connections, preparation of the


connection surface and adequate bolt pre-tensioning are critically important for
the proper functioning of the connection. It is also important to note that while
slip critical connections behave quite differently from bearing connections at ser-
vice load levels, the difference evaporates at ultimate and the two connections
behave identically. In other words, at ultimate the connection will have slipped,
and thus will be functioning as a bearing connection and thus the connection
must be designed for this case.

An important requirement for the proper performance of slip critical con-


nections is the attainment of proper bolt pre-tension forces. There are four
commonly used methods for achieving this pretensioning:

85
Figure 42: Slip critical connection schematic

• turn of the nut→ number of turns past snug tight


• calibrated wrench→ think of a torque wrench
• direct tension indicator → a washer with embossed bumps that get squeezed
flat when the bolt is tight
• twist-off stub→ a special stub that the wrench is attached to gets sheared
off when the bolt is tight
The design equations governing the slip critical connection are contained in
Specification Section J8.

5.2 Welds
Modern welding techniques for building and bridge applications universally in-
volves arc welding. Arc welding, as a joining process, is essentially comprised
of two components: heating and deposition. Both of these are byproducts of
the method by which arc welding gets its name: the use of a powerful electric
current.

The heating results from the passage of electric current (i.e. a spark that
is referred to as an arc) into the components to be joined. As with any elec-
trical circuit, there is a power source (the transformer / generator known as
the welder ) and legs of the circuit (known as welding leads). In the case of
structural welding, the components to be joined are also part of the circuit in
question with the circuit being completed by having an arc jump from an elec-
trode to the base metal components to be joined; this is critical to the deposition

86
portion of the process. The electrode from which the arc springs is a consumable
in the process: meaning, the electrode gets smaller as the weld is made, as a
result of it being ionized and re-deposited on the molten base metal portions
being joined. The heat required to melt the electrode and base metal arises out
of the natural resistance of the steel to the flow of current.

Generally speaking, in any of the arc welding techniques, it is important to


protect the molten metal within the newly joined connection from atmospheric
contamination during the critical solidification phase of the the welding process.
Specifically, nitrogen and oxygen can contribute to the formation of detrimental
compounds in the cooling weld metal. It is the job of the flux coating on the
electrode to help alleviate this contamination issues.

The electrode used in arc welding is typically concentrically coated with a


clay-like flux coating that is vaporized during welding. The vapor serves three
purposes:
• form a hard slag layer over recently deposited weld metal
• form a gaseous cloud around the arc to exclude impurities from the molten
metal pool
• and to provide useful chemical components for absorption into the molten
metal pool
While the concentric flux coating is used to facilitate the welding process, it is
the core of the electrode that serves as the filler metal to complete the joining
process. As can be imagined, there are a multitude of arc welding strategies
available to modern industry. However, there are two particular methods that
find wide usage within the domain of civil structural applications: Shielded
Metal Arc Welding (SAW) and Submerged Arc Welding (SAW). Both are typ-
ically employed with Direct Current (DC) power sources; with the electrode
being present on the anode of the circuit.

Shielded Metal Arc Welding (SMAW) is most commonly employed as the


manual procedure of choice to effect field welds. In this particular method, flux
coated electrodes are the staple for the process and the maximum fillet weld
size deposited in a single pass is usually 5/16 inch. A fillet weld is the most
common of the field welds (it comprises about 80 percent of all welds); involv-
ing the deposition of a line of filler metal possessing a cross-section that closely
resembles an equilateral triangle. In SMAW, eye protection is required during
the deposition process

Submerged Arc Welding (SAW) is an automated procedure typically used


for shop welding. In SAW the maximum weld size deposited in a single pass is
1/2 inch. In this particular method, a granular flux material is used to surround
the arc (thus the name) and thus no eye protection is needed during deposition.

87
Figure 43: Assumed fillet weld failure planes

As it is that fillet welds are so common in building and bridge construction,


the current discussion will focus on this weld type. As with all welds, fillet
welded connections must consider the two important strength limit states of
weld strength and base metal strength. An interesting peculiarity to the con-
sideration of weld strength in fillet welds is that the failure plane that develops
as the weld ultimate load is attained is assumed to be uniformly oriented at
forty five degrees from each of the weld legs (the legs are dimensioned with the
designation a in Figure 42). This assumption is not far from the truth, and
also results in a greatly simplified design approach with weld groups in connec-
tions such as that which is schematically illustrated in Figure 43. Based on the
sketch in Figure 43, and the stated assumption in the foregoing, it seems clear
that each line of weld making up the weld group in Figure 43 will fail along the
forty five degree failure plane illustrated in the upper right inset to the figure.
This uniform orientation is assumed valid irrespective of whether the weld line
is parallel or perpendicular to the line of action of the tensile force.

As would be expected from consideration of Mohr’s circle, the theme uni-


fying the weld failure that are parallel and perpendicular to the line of action
of the connection force is that of shear failure. The plane of the fracture that
forms in the weld during shear rupture defines a plane known as the throat of the
weld. From a consideration of elementary trigonometry, it is easily recognized
that this the throat dimension that accompanies√
an particular cross-sectional
2
slice of the rupturing fillet weld is given as 2 a.

AISC takes advantage of this approximation in filled weld strength when

88
specifying fillet weld strength in Specification Table J2.5. in this table, the
capacities for tensions normal to the weld axis and shear are both given as
0.60FEXX ; where the term 0.6 emanates from our earlier consideration of the
von Mises failure criterion where we used 0.577 as the conversion that takes a
uniaxial tensile yield strength and converts it to a yield strength in pure shear.
In addition the quantity FEXX refers to the electrode classification number,
within which the XX portion refers to the strength in ksi (e.g. E70 refers to a
70 ksi electrode). Based on these quantities, we can specify the fillet weld design
strength as:
√ !
2
φRn = 0.75 (0.6FEXX ) a Lw (202)
2
where a is the fillet weld size and Lw is the total length of fillet weld in the weld
group having that particular size.

We may arrive an equivalent expression for the base metal design strength
to round out our discussion on fillet welds:

φRn = 0.9 (0.6Fy ) (t) Lw (203)

In Equation 203 the quantity t is the plate thickness of the base metal and Fy
is the minimum specified yield strength for the base metal steel. In addition, in
Specification Section J2.2b requirements are promulgated for weld sizing that
are not based strictly on the foregoing strength requirements from Equations
202 and 203.

89

You might also like