CHAPTER 3

Plastic Behavior at the

Cross-Section Level

O
nce material properties have been obtained and an appropriate
stress-strain analytical model has been formulated, plastic
capacities can be calculated at the cross-section level. This is
a crucial phase of plastic analysis. The level of sophistication embraced
in the calculation of these capacities will have the foremost impact on
the resulting member and structural plastic strengths. Because plastic
analysis is generally used to compute ultimate structural capacities,
erroneous and potentially dangerous conclusions can be reached if an
overly simplistic cross-sectional model is used, thus luring the engi-
neer into a false sense of security. Hence, it is worthwhile to review
how various expressions can be derived for these cross-sectional
properties.

3.1 Pure Flexural Yielding

The basic case of pure flexural yielding should be familiar to engi-
neers who have used Limit States Design or Ultimate Strength Design.
Nonetheless, it is reviewed here because it provides some of the
building blocks necessary to understand the more complex models
presented later in the chapter.
A number of simplifying assumptions are made to calculate the
plastic moment capacity as follows:

• Plane sections remain plane; even though plastic deforma-

tions are typically larger than elastic deformations, their
overall magnitudes are still sufficiently small to satisfy this
condition.
• Structural members must be prismatic and have at least one
axis of symmetry parallel to the direction of loading.
• Membersaresubjectedtouniaxialbendingundermonotoni-
cally increasing loading (neutral axis perpendicular to axis of

111
112 Chapter Three

symmetry). Note that biaxial bending will be considered in

Section 3.7.
• Shearingdeformationsarenegligible.
• Member instability (flange local buckling, web local buck-
ling, lateral-torsional buckling) is avoided.
• Structuralmembersmustnotbesubjectedtoaxial,torsional,
or shear forces. This assumption obviously will be relaxed in
the later sections of this chapter.

The elasto-perfectly plastic model is generally used to calculate

plastic moment capacities and leads to acceptable results for most
practical problems in structural engineering (ASCE 1971). It is there-
fore used in the following derivations. However, nothing precludes
the consideration of more complex models (or a smaller number of
fundamental assumptions for that matter) because the corresponding
cross-sectional strengths and other relevant plastic properties could
be determined from the same basic concepts presented in this book.

3.1.1 Doubly Symmetric Sections

Doubly symmetric sections constitute the simplest cross-sections for
which analytical expressions of plastic strength can be developed,
because their neutral axis will always be located at their geometric cen-
troid. With an elasto-perfectly plastic material model, for a given cross-
section, the stress diagrams can be constructed directly from the strain
diagrams, as shown in Figure 3.1, for various levels of increasing strains
on an arbitrary doubly symmetric structural shape. At any point along
a given loading history (such as that shown in Figure 3.1), the cross-
section curvature, f, can be calculated by simple geometry, as:
e max
f= (3.1)
( h/2)
where emax is the maximum strain acting over the doubly symmet-
ric cross-section of depth h. If at a given curvature the yield strain,
ey, is located at a distance y* from the neutral axis, then, as shown
in Figure 3.1, by similar triangles:

e max ey  h e y
f= = y* =   (3.2)
( h/2) y *  2 e max

Once y* is known for a given curvature, it becomes a simple mat-

ter to compute the corresponding moment using simple mechanics of
materials principles:
h/2 h/2
M= ∫ s y dA = ∫ s y b( y ) dy (3.3)
- h/2 - h/2

where b is the cross-sectional width, expressed as a function of y.

 At strain state B:
y* for case B
Plastified zone
Elastic zone A B C +σy +σy +σy +σy +σy

T
D
E ∞
+ h + e

C
b –εy +εy –σy A –σy B –σy C –σy D –σy E
Generic Rectangular
Cross-sections Strain distribution Stress distributions
+σy
+σy –σy

y*
= + +

–σy +σy
–σy
I II III
Subcomponents of stress distributions

Figure 3.1 Strain and stress distributions in the plastic range for members of symmetric cross-section.

113
114 Chapter Three

For any given curvature, the cross-section can be divided into

elastic and plastic zones. Within ± y* of the neutral axis, the section is
elastic. Outside this zone, also termed the elastic core, strains exceed
ey, and the material is plastified. As curvature increases, the elastic
core progressively shrinks, and plastification progressively spreads
over the entire cross-section.
The impact of this progressive growth of the plastified zone on the
moment-curvature relationship is best understood from a case study.
For simplicity, a rectangular cross-section is used. The calculations are
directly related to the strain profiles identified in Figure 3.1, which use
a rectangular section of height h and base b. For that example, in the
elastic range, the flexural moment and curvature are given by:

 bh2  M  12  M
M = Ss =  s f= =
 6 
(3.4)
EI  bh3  E

where S is the section modulus. For strain distribution A in Figure 3.1,

theyieldstrainhasjustbeenreachedatthetopfiberofthecross-section.
Replacing s by sy in the above equation will give the yield moment, My,
and the yield curvature, fy: the transition point between purely elastic and
elasto-plastic behavior. From that point onward, any increase in curva-
ture introduces partial plastification of the cross-section, a condition
also sometimes called contained plastic flow. For example, for an arbi-
trary strain profile represented by B in Figure 3.1, the curvature is given
by Eq. (3.1) above, and the corresponding moment can be expressed as
a function of y* as follows:
y* h/2
M = 2 ∫ s ybdy + 2 ∫ s y ybdy (3.5)
0 y*

where the first term reflects the contribution of the elastic core to
moment resistance, and the second term, that of the plastified zone.
Using the relationship s/sy = sy y* valid over the elastic core, Eq. (3.5)
becomes:

 2b y* h/2  2 bh2
M= ∫
 y * 0
y 2 dy + 2 b ∫ ydy s y = b( y *)2 s y +
 3 4 y
s - b( y *)2 s y
y*

(3.6)
Interestingly, the three terms on the right side of this equation
correspond to the contributions to the moment that can be obtained
by a piecewise decomposition of the stress diagram into the simpler
subdiagrams I, II, and III, respectively, shown in Figure 3.1. Adding
and subtracting stress diagrams in this manner is a statically correct
procedure that can prove useful to simplify complicated problems or
 Plastic Behavior at the Cross-Section Level 115

to graphically verify the adequacy of analytically derived results. For

that matter, the integration approach [per Eqs. (3.5) and (3.6)] should
be avoided whenever possible and only be used as a last resort,
because integrating is more time consuming and error prone than
working directly with stress diagrams to calculate the forces and
lever arms acting on regularly shaped parts of cross-section. More
importantly, the latter approach, based on physics rather than math-
ematics,ismoreconducivetosoundengineeringjudgment.
To complete the above derivation, it is worthwhile to regroup the
terms in (y*)2 and express the results in terms of curvature. Thus,
when the following relationship is used (again, obtained through use
of the properties of similar triangles):

y* ey fy y fy
= = = (3.7)
h/2 e max y f f

the expression for the flexural moment at a given magnitude of cur-

vature (i.e., the moment-curvature relationship) becomes:

 2 2
 bh bh2  fy   (3.8)
M= - s
 4 12  f   y

Whenever possible, a solution expressed in a normalized man-

ner, that is, in terms of both (M/My) and (f/fy), should be sought. For
example, when Eq. (3.4) is used to obtain an expression for the yield
moment, the final result becomes:

 2  2
M 3 1  fy   3  1  ey 
= 1 -   = 1-  (3.9)
My 2  3  f   2  3  e max  
 

The latter part of this equation, expressed in terms of the strain-

ratio, is obtained if one takes advantage of the simple equivalence
that exists between normalized curvatures and strains.
When Eqs. (3.4) and (3.9) are used over their respective range of
validity, the entire moment-curvature relationship for this cross-
section can be calculated and plotted. This has been done in Figure 3.2
for different sections. Figure 3.1 and Eq. (3.9) show that, theoretically,
for any cross-section, full plastification and maximum flexural moment
will be reached only at an infinite curvature (Case E in Figure 3.1). For
the rectangular cross-section, this maximum moment is 1.5 times
the yield moment, as can be easily deducted from Eq. (3.9). How-
ever, for practical purposes, as seen in Figure 3.2, this maximum
moment is rapidly approached and nearly reached at only three or
116 Chapter Three

M/My

2 k = 2.0
k = 1.7
k = 1.5
k = 1.27
1
Ideal wide flange (k ≈ 1.0)
North American wide flange shape (k ≈ 1.14)

1 φ/φy

Figure 3.2 Normalized moment curvature relationship and exural shape factor, k,
for different cross-sections.

four times the yield curvature. In fact, when the maximum strain
over the cross-section approaches the onset of strain hardening of
the steel material, at approximately 10 times the yield strain value
(as mentioned in the previous chapter), Eq. (3.9) indicates that 99.7%
of the maximum moment has been reached. This demonstrates that
a fully plastified cross-section can reliably be used to calculate the
maximum moment, referred to hereafter as the “plastic moment.”
For example, for the rectangular cross-section, one can calculate
the plastic moment directly using the resulting forces and lever arms
corresponding to the stress distribution E of Figure 3.1. This gives:

 e   h   h bh2
Mp = 2 T   = 2 s y   b   = s = Zs y = 1 . 5 My (3.10)
2  2  4 4 y

where Z is the plastic section modulus, a geometrical property of any

given cross-section. Incidentally, the above algebraic expression for Z
of a rectangular section (i.e., bh2/4) is used extensively when calculat-
ing the plastic moment for structural shapes built of rectangular parts,
as will be seen later.
Another useful sectional property is given by the shape factor, k,
which is the ratio between the plastic section modulus and the elastic
section modulus. This factor, expressed by:

Mp Z
k= = (3.11)
My S
 Plastic Behavior at the Cross-Section Level 117

provides information on the additional cross-sectional strength

available beyond first yielding. The shape factors for various cross-
sections are shown in Figure 3.2. For the wide-flange sections typi-
cally used in North American steel construction (AISC 2011, CISC
2010), the shape factors typically vary from 1.12 to 1.16, with an aver-
age of 1.14. Moreover, the following normalized moment curvature
expressions could be derived:

1. When y* is located in the flange:

bd 2  bh2  1  f  
2
M f
= 1 -  + 1 -  (3.12)
My fy  6S  4S  3  fy  
 

2. When y* is located in the web:

2
M Mp  wd 2   fy 
= - 
My M y  12S   f  (3.13)

where b is the flange width, d is the total depth of the structural sec-
tion, h is the distance between flanges (or web length), w is the web
thickness, and all other terms have been defined previously. In that
case, Mp can be calculated from the individually calculated forces
resulting from the constant yield stress acting on the flanges and web,
and their respective lever arms, or by using Eq. (3.10) twice, to sub-
tract the value of Mp for a rectangle of width (b -w) and depth h from
that for a rectangle of width b and depth d. In both hand calculation
approaches, the area of the rounded corners where flanges connect to
the web is usually neglected, but it is generally included in the sec-
tional property values tabulated in design manuals.
Basic principles of mechanics of materials indicate that an ideal
wide-flange section for flexural resistance would have all its material
concentrated in flanges of infinitely small thicknesses (obviously an
impractical theoretical case). The shape factor of such an ideal section
would be unity, because the entire cross-section would reach the yield
strain simultaneously, without any possible spread of plasticity given
that no material would exist between these flanges. Hence, both the
plastic and the elastic section moduli in that case would be equal to
the area of one flange times the distance between the two flanges.

3.1.2 Sections Having a Single Axis of Symmetry

The procedure developed above is also applicable for sections hav-
ing only a single axis of symmetry parallel to the applied load, with
the essential difference that the location of the neutral axis must
now be determined explicitly. Therefore, for any given curvature,
118 Chapter Three

the calculations must start by a determination of the neutral axis.

There are two instances when this calculation is relatively simple.
First, as long as the material is entirely linear-elastic, the location of
the neutral axis remains located at the center of geometry of the
cross-section. Second, when this same material is fully plastic, the
neutral axis must be located such that it evenly divides the cross-
sectional area. This latter case is simply a consequence of having the
entirecross-sectionsubjectedtotheyieldstress,ineithertensionor
compression. Summing the axial forces on the cross-section:

P= ∫ s dA = Atension s y + Acompression (- s y ) = 0 ⇒ Atension = Acompression

area
(3.14)
which defines the location of the neutral axis in the fully plastic condi-
tion. In the transition phase between the fully elastic and fully plastic
conditions, where plastification progresses through the cross-section
as the applied moment is increased, the neutral axis progressively
migrates. This is schematically illustrated in Figure 3.3. For these inter-
mediate cases, the development of analytical moment-curvature
expressions, although possible, can be rather complex, except for the
simplest cases.
Therefore, even though explicit moment-curvature relations can
be derived, as done in Section 3.1.1, for many doubly or singly sym-
metric cross-sectional shapes, it is sometimes more convenient to
simply calculate the moments corresponding to a large number of
curvatures and fit the data using a Ramberg-Osgood or Menegotto-
Pinto function. When using hand calculations, one can obtain each
moment-curvature point by selecting a value for the strain at the top
of the cross-section, arbitrarily choosing a location for the neutral
axis, calculating the axial force resulting from that assumed strain
diagram, and iterating by changing the position of that neutral axis
until zero axial force is obtained. By repeating that process for various

σ < σy +σy +σy +σy Acompression

C.G.

Atension

–σy –σy –σy –σy

Fully plastified condition

Figure 3.3 Migration of neutral axis in singly symmetric cross-section as applied

moment increases during progressive plastication.
 Plastic Behavior at the Cross-Section Level 119

strain values, one can plot the moment-curvature relationship for a

given cross-section. Following the same logic, moment-curvature
points can be generated through the use of computer programs. The
approximate moment-curvature relation obtained in that manner
would still be sufficiently accurate to allow precise computation of
member stiffnesses or deflections.
There exist numerous computer programs capable of developing
moment-curvature relations for arbitrary cross-sections under uniax-
ial or biaxial bending, and these are often capable of considering axial
and shear forces and other factors. Such programs can also easily be
written because their structure is generally rather simple. For exam-
ple, for an arbitrary section having a single axis of symmetry (parallel
totheappliedload)andsubjectedtouniaxialbending,itissufficient
to use a layered model of the cross-section. This consists simply of
“slicing” the cross-section into a large number of layers, say 1000, and
calculating and integrating the contributions of all layers to the flex-
ural moment at a given curvature. To structure such a program, one
could write subroutines to accomplish the following tasks:

• Performanautomaticlayeringofthecross-section(basedon
simple input of geometry characteristics).
• Initialize the stress values for all layers and establish other
initial parameters.
• Setupcontrolsfortheiterationstrategy.
• Increment the curvature for calculation of a given moment-
curvature data point.
• Estimate the location of the neutral axis, adjusting this esti-
mate in accordance with an iteration strategy (i.e., consider-
ing the results from previous iterations).
• Forgivencurvatureandneutralaxislocation,calculatestrains
for all layers.
• Calculate stresses for all layers per the assumed material
model.
• Calculatetheresultingmomentbysummingthecontribution
of all layers about the neutral axis.
• Calculate the resulting axial force on the cross-section by
summing the contribution of all layers.
• Check for convergence using, for example, a user-specified
tolerance on the axial force.
• Iterateuntiltheaxialforceisequaltozero,withinthespecified
tolerance. Convergence gives a single M and f point and
corresponding stresses at all layers of the cross-section (i.e.,
stress distribution). Repeat calculations at other curvatures to
obtain the entire M-f curve.
120 Chapter Three

Note that any material model could be implemented in such a

computer program, although the simple elasto-perfectly plastic
model is frequently sufficient. Furthermore, the above algorithm can
easily be modified to allow consideration of cyclic loading, biaxial
bending, nonzero axial forces, residual stresses, and nonsymmetric
cross-sectional shapes.

3.1.3 Impact of Some Factors on Inelastic Flexural Behavior

A large body of experimental research has confirmed that the plastic
flexural moment can indeed be developed in beams. A summary of
some of that earlier experimental work is presented in ASCE (1971). It
is understood that, for this plastic moment to develop, the constraints
set by the assumptions listed at the beginning of Section 3.1 must be
respected. However, a few additional factors that may have an impact
on this inelastic flexural behavior deserve the following brief review.

3.1.3.1 Variability in Material Properties

As the plastic moment directly depends on the yield stress of the
steel, it is worthwhile to question what value can be reliably used for
its calculation.
A potentially fatal mistake would consist of using the mill test
certificate value. Steel mills typically perform a single coupon test per
batch of steel produced (the size of a batch will vary depending on
the mills, but typically consists of many tons of steel). This coupon is
tested at strain rates that usually raise the yield strength by 30 MPa
(4.4 ksi) or more and are intended only to provide confidence that the
entire batch will meet the applicable specifications. Fluctuations in
steel properties will exist within a given batch of steel produced from
the same heat, and the mill test certificate can provide, at best, only
one value for the entire tonnage of steel produced by that heat. There-
fore, using the value reported on the mill test certificate is improper
and fraught with danger; building failures have been documented
whereinsuchamistakehasbeenidentifiedasamajorreasoncontrib-
uting to collapse (e.g., Closkey 1988).
For the design of new structures, the specified yield strength is
the value that should be used. Engineers familiar with limit states
design concepts [such as the Canadian Standards Association’s Limit
States Design (LSD), or the American Institute of Steel Construction’s
Load and Resistance Factor Design (LRFD), or others] appreciate the
variability inherent in any engineering parameter, including material
properties. These variabilities are generally included in the load and
resistance factors.
However, when one is evaluating existing structures, it is impor-
tant to appreciate the cross-sectional variability of yield strength.
Extracting material from a steel member to obtain its yield stress
using a standard coupon test (such as the standard ASTM E6 test)
could give quite different results depending on whether the coupon
 Plastic Behavior at the Cross-Section Level 121

is taken from the flanges or the web. Galambos and Ravindra (1976)
reported that mean values of yield stress for coupons taken from
flanges and webs were respectively 5% and 10% larger than the spec-
ified values, with coefficients of variation of 0.11 and 0.10, respec-
tively. This is largely a consequence of the different treatment the web
and flanges receive during the rolling process: thicker plates and
members are less worked and cool more slowly, resulting in a slightly
different grain structure of the metal and weaker strength properties.
Likewise, variations exist depending on the thickness of the rolled
shapes. In fact, special alloys are added to very thick steel section
(such as W-shapes formerly known as “jumbo sections,” or AISC
Group 4 and 5 shapes, before 2005) to provide the same yield strength
as thinner sections for a given metallurgical composition. For some
types of steels, when mills do not modify the chemical composition of
the steel to compensate for this loss of strength, lower specified yield
strengths are provided for use in design (e.g., CISC 2010).

3.1.3.2 Residual Stresses

Contrary to what is commonly assumed in design, a steel member is
not stress-free prior to the application of external loads. In fact, large
internal stresses exist there in a state of self-equilibrium. These are
generally “locked in” during the rolling process and are also affected
by welding or any other heat-imparting or cold-working operations.
To understand the origin of these residual stresses, one must visual-
ize the cooling process of a rolled steel section, keeping in mind that
the modulus of elasticity of steel at high temperatures is very low,
and increases rapidly as the steel progressively cools down below
1000°F, and that steel (like other materials) shrinks as it cools.
Thus, when a rolled section is cooled by the surrounding air, the
tips of the flanges that are surrounded by air on three sides cool and
gain stiffness first. Shrinkage is essentially unrestrained by the adja-
cent softer steel. However, as cooling progresses along the flanges, the
tips of the flanges that have already partly cooled and acquired some
stiffness provide partial restraint against shrinkage of the adjacent
flange material. Hence, the tip of the flange is placed in compression,
and the adjacent cooling material, in tension. As cooling progresses
further along the flanges, the process repeats itself, and all previously
partlycooledandstiffermaterialiscompressedbytheadjacentmate-
rial that is beginning to cool. Consequently, the tips of the flanges that
have cooled first will be the most compressed, and the flange-web
core material, which is surrounded on all sides by steel and cools the
slowest,willbesubjectedtothelargestinternaltensilestresses.
Members that can cool rapidly, such as thin steel plates, will be
subjected to the largest magnitude of residual stresses, with values
occasionally reaching up to the yield stress. However, in most rolled
steel sections, the maximum residual stresses are approximately 33%
of the yield stress.
122 Chapter Three

The same logic also illustrates how welding introduces residual

stresses. As welding is accomplished, the cooling and shrinking of
both the weld metal and the steel in the heat-affected zone is restrained
by the adjacent steel material. Therefore, following welding, the
welds will be in tension. The existing residual stress pattern in the
base metal is also locally affected by the welding operations.
A schematic representation of these self-equilibrating residual
stresses is shown in Figure 3.4, using linear variations of stresses
along the flanges and web. For comparison, an actual residual stress
distribution in a steel section is illustrated in Figure 3.5. These inter-
nal stresses are in self-equilibrium because the integral of their effects
produce no resulting axial force or moment on the cross-section.
Although residual stresses can be large, they have no impact on
the plastic moment of a cross-section. For example, consider the sec-
tion of Figure 3.4 for which residual stresses are assumed to be of a
magnitude equal to half the yield stress. When that section is sub-
jected to pure axial loading in compression, the tips of the flanges
will reach their yield stress at only half the applied axial load that
would produce full yielding of the cross-section if there were no
residual stresses. From that point onward, plastification will start
spreading over the cross-section. Essentially, because every individ-
ual point along the cross-section starts from a different initial stress,

Approximate P
fσy Without residual stresses
linear 1-D model
With residual stresses
fσy
Parabolic Py
fσy 1-D model
fσy fPy
fσy fσy

f∆y ∆y (1 + f)∆y ∆

M
fσy fσy
Mp

My Without residual stresses

With residual stresses
fMy

fσy fσy fφy φy φ

Residual stresses models Examples of effect of residual stresses

Figure 3.4 Schematic representation of self-equilibrating residual stresses in wide-

ange structural shape.
 Plastic Behavior at the Cross-Section Level 123

W 14 × 730

23 H 681

Contour lines show isostress in ksi

Figure 3.5 Two-dimensional distribution of residual stresses in rolled and

welded wide-ange structural shapes. (From L. Tall, Structural Steel Design,
2nd ed., 1974.)
124 Chapter Three

itmustbesubjectedtoadifferentmagnitudeofstrainpriortoreach-
ing the yield stress. Eventually, the flange-web core zones will be the
last points to yield, when the applied strain will be 50% larger than
would have otherwise been necessary to plastify a section free of
residual stresses. Indeed, a force-elongation diagram similar to that
shown in Figure 3.4 is usually obtained when one tests full cross-
section stub-column specimens in axial compression (or tension)
instead of standard material coupons.
Ifthesamecross-sectionweresubjectedtoflexureinsteadofaxial
force, it would start yielding at a moment equal to half of My, with
plasticity spreading from the tips of the flanges inward for the flange
in compression, and from the flange-core outward for the flange in
tension, as the respective flanges would be subjected to larger com-
pression and tension as the flexural moment increased. However, the
plasticity moment would be unchanged and remain Mp. This is dem-
onstrated in more detail in the example below.
Although residual stresses do not impact the strength of mem-
bers, the accelerated softening of the axial force versus axial deforma-
tion or moment versus curvature curves, as well as the earlier
initiation of the yielding process, will have an impact on members’
deflections and buckling resistances. Incidentally, the analytical
expressions included in steel-design codes and standards to calculate
the stability and strength of structural members (such as for columns
in compression) already take this into account.

3.1.3.3 Example: Ideal Wide-Flange Section with

Residual Stresses
An “ideal” wide-flange section for flexural resistance has flanges of
negligible thickness and area Af , a height d, and negligible web area
(Figure 3.6). The shape factor for this section is 1.0, and Mp = dAf sy. It is
assumed that the initial residual stresses introduced by the rolling pro-
cess (thus prior to the application of any external loads) have a peak
value of 0.75sy (in compression) at the tips of the flanges and 0.75sy (in
tension) at the intersection with the web and that they vary linearly in
between. The material is assumed to be elasto-perfectly plastic.
The M-f curves for the initially stress-free case, and for the case
having residual stresses, are to be drawn. For this purpose, although
accurate analytical expressions could be derived, for simplicity here,
the solution will proceed by calculating representative points to accu-
rately plot these curves.
Using the principle of stress superposition, and the knowledge
that, at any point, the newly applied stresses and strains are related as
per the elasto-perfectly plastic model, one can construct Figure 3.6. In
the resulting moment-curvature plot in this figure, the dashed line is
obtained when one considers the cross-section free of residual stresses,
using fy and ey corresponding, respectively, to the curvature and
maximum strain at the onset of yielding My (which happens to be
 M/Mp
Without residual stresses
With residual stresses
b t≈0 –0.75 σy
1.0
0.917
Aflange 0.813
+0.75 σy 0.667
Aweb = 0 0.479
d Aflange –0.75 σy 0.250

1/4 1/2 3/4 1 1 3 /4 φ/φy

+0.75 σy
A B C D
Cross section Initial residual stress distribution

A B C –1.0 σy –1.0 σy –1.0 σy –1.0 σy

D +0.50 σy +0.25 σy –0.25 σy

A B C D

–0.50 σy –0.25 σy
–εy +1/4εy +εy +0.25 σy
+1.0 σy +1.0 σy +1.0 σy +1.0 σy
Strain distribution Stresses at 0.25 φy Stresses at 0.50 φy Stresses at 1.0 φy Stresses at 1.75 φy

Figure 3.6 Plastic exural behavior of ideal structural shape with residual stresses.

125
126 Chapter Three

equal to Mp in this example). The resulting M-f curve is bilinear for

this ideal cross-section having a shape factor of unity.
Moreinteresting,however,isthecalculationoftheM-f curve for
the case having residual stresses. Stage A identifies the end of the
elastic range, because the application of strains equal to ey/4 will
uniformly add a stress of sy/4 to the flanges and bring to yield the
points for which the magnitude of residual stresses was 0.75sy.
Values of moments and curvature at multiples of this curvature are
calculated. The resulting strains, stresses, and M-f (solid line) are
shown in Figure 3.6. As an example of an intermediate calculation,
for stage C, when the curvature is f, the corresponding moment cal-
culated from the stress diagram is:

  Af 1  sy  Af  d 
Mstage -C = 2 2  s y +  + s y   = 0 . 8125 A f d s y = 0 . 8125 M p
  4 2 4 4  2 

(3.15)

3.1.3.4 Local Instabilities

Plastic curvature cannot increase indefinitely, and eventually local
buckling, or lateral torsional buckling, will occur. The problem of
member instability receives a comprehensive treatment in Chapter 14.
However, at this point, it is important to realize that local buckling
must eventually occur in any structural member compressed far into
the plastic range (such as the flange of wide-flange beams commonly
used in structures) as the strain-hardening tangent section modulus
progressively decreases when plastic strains increase (as demon-
stratedinChapter2).Morestringentwidth-thicknesslimitsimposed
by plastic design procedures simply delay local buckling to ensure
the development of large plastic deformations within the limits
expected in normal applications.

3.1.3.5 Strain Hardening

Models that neglect strain hardening, such as the elasto-perfectly
plastic model, are effectively much simpler to use and particularly
useful for hand calculation. Although the consideration of strain
hardening is always possible, at the cost of additional computational
efforts, it has been proven to be generally conservative to neglect the
effectofstrainhardeninginbeamssubjectedtomomentgradient,as
long as strength is the primary concern. The influence of strain hard-
ening on plastic rotation calculations can be more significant. For the
specialcaseofbeamssubjectedtouniformmoments(inwhichplasti-
fication occurs simultaneously over the entire uniform moment
region), the consideration of strain hardening would bring few ben-
efits as local buckling typically promptly occurs at the onset of strain
hardening in such beams.
 Plastic Behavior at the Cross-Section Level 127

3.1.4 Behavior During Cyclic Loading

Amajorapplicationofplasticanalysisconceptsisfoundinthedesign
of earthquake-resistant structures. Consequently, it is important to
examine the effect of cyclic loading on a partly or fully plastified
cross-section. The key to understanding the plastic behavior of a
cross-section is to consider it as a series of layers of material. All lay-
ers of a cross-section must abide by the same material model rules,
but each layer is strained differently, so its stress history will also
differ. For example, in Figure 3.7, the points A to F are assigned to

+εy +σy σ D E F
F
E C
D
C B
B
A A
ε

–σy
–εy
+σy σ < –σy M
Case A
Unloading My
+ = EI EI

Permanent φ
–σy σ > +σy
plastic offset
+σy –2 σy –σy
M

Case B
Load reversal My
+ = EI EI 2 My

φ
–σy +2 σy +σy

+σy –2 σy –σy M
+Mp
Case C
Full plastic My
load reversal + = 2 My
EI EI

EI
φ
+σy +2 σy +σy 2 My

–Mp

Figure 3.7 Example of cyclic cross-sectional plastic behavior.

128 Chapter Three

various layers along the height of a rectangular cross-section that

hasbeensubjectedtoamomentlargerthanitsyieldmoment,M y
,
but less than its plastic moment, M p. When the applied moment
is removed, the cross-section is unloaded to its point of zero
applied moment on the M-f diagram, and all layers unload elas-
tically according to the elasto-plastic element model. In fact, a
layer will always unload elastically as long as its stress doesn’t
reach the opposite yield level of -s y. Therefore, one must first try
to remove the applied moment elastically, as shown in Case A of
Figure 3.7.
If none of the stresses in the resulting stress diagram exceed
the yield stress, the solution is deemed acceptable. Note that
although the externally applied load has been removed, an inter-
nal residual-stress diagram in self-equilibrium has been created.
Moreover, as can be seen from the M-f diagram, a residual curva-
ture remains. One can obtain the magnitude of this curvature by
first calculating the maximum curvature that was reached during
the initial loading phase [for the rectangular section used, Eq. (3.9)
derived previously can be used for this purpose] and subtracting
from this value the curvature that would correspond to the removal
of this moment, assuming elastic response [using Eq. (3.4), that is,
f = M/EI]. Because the two values are quite different, a residual
curvature must remain.
Following the above logic, one can observe, as in Case B of
Figure 3.7, that for any cross-section that was first stressed above
My, it is possible to remove elastically a moment as large as 2My
without having any stresses exceed the yield value in the resulting
stress diagram. In fact, the section can be subjected to reversed
loading histories within this range of 2My without producing any
new plastification. This is equivalent to having a new elastic range
shifted upward by the difference between M and My. However, as
soon as loading exceeds these bounds, new yielding occurs, and
the stress-strain history of each layer must be followed to deter-
mine the actual stress distribution throughout the cross-section. To
determine this stress profile, the procedure followed in the exam-
ple on residual stresses in the previous section can be used, but
one must account for the offset produced by the residual curvature
at zero moment. A maximum moment of -Mp will eventually be
reached, as shown in Case C of Figure 3.7. The procedure can be
repeated by reversing the loading and cycling repeatedly, thus
producing M-f hysteretic curves.
Generally, the ductile behavior of steel members will develop
until a local instability occurs, because of either excessive straining
under noncyclic loading or fracture under alternating plasticity (e.g.,
low-cycle fatigue under cyclic inelastic loading), as discussed in the
previous chapter.
 Plastic Behavior at the Cross-Section Level 129

3.2 Combined Flexural and Axial Loading

In Section 3.1, it was assumed that no externally applied axial load
acted on the cross-section. However, in many instances this will not
be the case, and it is necessary to investigate how the plastic moment
is affected by the presence of axial force. The same fundamental con-
cepts and modeling previously presented still apply: strains linearly
distributed across the member’s cross-section are related to stresses
using an elasto-plastic model, and the moments and axial forces are
obtained by integrating the stresses acting on the cross-section (or,
more simply, using the stress-resultant forces). Figure 3.8 illustrates
how the stress diagram changes as the applied moment is progres-
sively increased for a given axial force.
Calculation of the reduced plastic moment, Mpr, in the fully plas-
tified state for a given axial load, P, is a straightforward operation.
Directly, equilibrium of the horizontal forces acting on the cross-
section gives:

P = C - T = A compression s y - Atension s y = ( Acompression - Atension )s y (3.16)

Given that the sum of Acompression and Atension must equal the total
cross-sectional area, A, one can directly solve for the location of the
neutralaxis;thestress-resultantforces,C and T;andthecorrespond-
ing reduced plastic moment, Mpr. By repeating the process for axial
forces varying from zero to the axial plastic load (= Asy), one can plot
an interaction diagram for a given cross-section. Alternatively, for the
simplest cross-sections, closed-form solutions may be developed.
Some of these closed-form solutions are provided here, taking
advantage of the fact that the fully plastified stress diagram for com-
bined flexural-axial response can be divided into a pure moment con-
tribution and a pure axial contribution, as shown in Figure 3.9. For
convenience, that figure is developed using the same arbitrary neutral
axislocationforvariouscross-sections;theimplication isthat,forthis
generic state of full plasticity, different corresponding axial loads and
moments would be obtained for each cross-section. The basic princi-
ple, nonetheless, remains the same: when a location for the neutral
axis is assumed, expressions for the corresponding applied axial force
and reduced plastic moment can be developed, which are valid over
all or some depths of the cross-section. Through algebraic manipula-
tions, it is possible to develop equations for interaction diagrams that
express the applied axial force as a function of the reduced plastic
moment, although this sometimes proves to be a tedious process.
These equations are developed hereunder for some simple dou-
bly symmetric sections, for a neutral axis generically located at a dis-
tance yo above the geometric center of these sections.
130
<+σy +σy +σy +σy +σy
T
Plastic neutral axis
et
Elastic neutral axis P
Center of gravity
C =

ec M

–σy A –σy B –σy C –σy D –σy E

Cross section Stress distributions

Figure 3.8 Stress diagrams as plasticity progresses in a cross-section subjected to combined exural and axial loading.
 t
+σy +σy
–σy

h yo
= + C
d h w or or
yo
M

–σy –σy
b b
Cross sections Stress distributions

Figure 3.9 Fully plastied condition for an arbitrary location of neutral axis (corresponds to different set of P and M for the different cross-
sections shown).

131
132 Chapter Three

3.2.1 Rectangular Cross-Section

If one divides the stress diagram of Figure 3.9 into pure flexural and
axial contributions, the resulting axial force is:

 P  (P = 2 byo s y )  2 yo (3.17)

 P  =  (P = hbs )  = h
 y   y y  

where Py is the capacity of the cross-section fully plastified axially,

and all geometric parameters are defined in Figure 3.9.
The expression for the reduced plastic moment can be developed
if one subtracts the plastic moment of a section of depth 2yo (i.e., that
portion of the cross-section assumed to resist the axial load, P) from
the plastic moment (i.e., the flexural strength in absence of axial load),
making it possible to obtain:

 bh2   b(2 yo )2   b( h2 - 4 yo2 )

Mpr = Z s y - Zy s y =   -    s = sy (3.18)
o
 4   4  y 4

In a normalized format, and substituting the result of Eq. (3.17):

Mpr
=
( )
b h2 - 4 yo2 s y  4 
= 1 -
4 yo2
= 1 -
 P
2
(3.19)
Mp 4  bh2 s  h2 P 
 y  y

The resulting interaction diagram is plotted in Figure 3.10.

Note that, for a rectangular cross-section, a significant difference
exists between the elastic interaction curve for which no strain is
allowed to exceed the yield strain, and the plastic interaction curve
derived above.

3.2.2 Wide-Flange Sections: Strong-Axis Bending

For wide-flange sections (i.e., those having constant flange thickness),
the closed-form solution will vary depending on whether the neutral
axis falls in the web (yo ≤ h/2) or the flange (h/2 < yo ≤ d/2). For the
former case:

 P  (P = 2 yo ws y )  2 wyo Aw
 = = ≤ (3.20)
 Py   (Py = As y )  A A
 Plastic Behavior at the Cross-Section Level 133

1.0

0.9

0.8

0.7
Plastic
0.6
P/Py

0.5
Elastic
0.4

0.3

0.2

0.1

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
M/Mp

Figure 3.10 Elastic and plastic M-P normalized interaction diagram for a
rectangular cross-section.

where A is the total cross-sectional area (= 2bt + wh), A w is the web

area (= wh), and:

  w(2 yo )2  
Mpr = Zs y - Zy s y = Z -    s = (Z - wyo2 )s y (3.21)
o
 4  y

Then, to obtain a normalized M-P interaction curve, divide the

above equation by Mp and substitute the result of Eq. (3.20):

Mpr  Zy   wy 2 
2
 P  A2 P Aw
= 1 -  o  = 1 -  o  = 1 -   for ≤ (3.22)
Mp Z Z  Py  4wZ Py A

When the neutral axis falls in the flange, the following expression
is obtained using a similar procedure:

Mpr  P  A P  1  P Aw
= A 1 -  d -  1 -     for > (3.23)
Mp  Py   2b  Py   2 Z Py A
 
134 Chapter Three

Note that, until the axial force exceeds 30% of the axial plastic
value, the reduction in moment capacity is typically less than 10% for
these structural shapes. For most commonly available wide-flange
sections, the normalized M-P interaction curve is a function of the
ratio of the web area to total cross-sectional area (which can be alter-
natively expressed as the web area to sum of the flange areas, or many
other variations). This could be demonstrated, as an example, by
expansion of the non-normalized term of Eq. (3.22) as follows:

A2 A2
=
4wZy  wh2 A f (d - t)
4w  + 
4 2
A2 (3.24)
=
Aw2 + 2 A f ( Aw + tw)

1
= 2
 Aw   2 A f Aw   2 A f tw 
  +   + 
A A2   A2 

where Af is the total flange area (i.e., sum of the areas of both flanges).
Simple observation, or trial calculation, reveals the very small signifi-
cance of the third denominator term on the resulting normalized M-P
interaction curve. This term is the only nonconstant value for a given
ratio of web-to-flange area. As a result, normalized interaction curves
can be conveniently expressed as a function of the flange-to-web area
ratio, as shown in Figure 3.11a.
It can be observed from Figure 3.11a that the normalized M-P
interaction curves of wide-flange sections that have the lowest ratio
of flange-to-web areas are closest to the curve obtained previously for
the rectangular cross-section, as is logically expected. Although this
observation is useful to demonstrate the relative physical behavior of
various cross-sections, it should be remembered that steel rectangu-
lar shapes inefficiently use material and are not desirable, in spite of
their more extensive plastic range. Finally, note that wide-flange sec-
tions are usually rolled with a relatively constant ratio of flange-to-
web area, normally between 2 and 3, approximately corresponding to
the shaded area in Figure 3.11a. This has made possible the develop-
ment of a convenient and reliable M-P design interaction curve for
the plastic strength of wide-flange cross-sections in strong-axis bend-
ing, as shown in Figure 3.11b, and expressed by:

M  P
= 1 . 18 1 -  ≤ 1 . 0 (3.25)
Mp  Py 
 Plastic Behavior at the Cross-Section Level 135
1.0
Rectangle
bt
0.5 =
wh

bt
1.0 =
wh

Most bt
sections 1.5 =
P wh
0.5
Py
b

w
h d
df
bt
2.0 =
wh

0
0 0.5 1.0
Mpc
Mp
(a)

1.0

0.9

0.8

0.7

0.6
P
0.5
Py
0.4

0.3

0.2

0.1

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
M
Mp
(b)

Figure 3.11 Plastic M-P normalized interaction diagrams: (a) for wide-ange
structural shape, strong-axis bending; (b) simplied design interaction curve
for wide-ange structural shape, strong-axis bending; (c) for wide-ange
structural shape, weak-axis bending. (Figures a and c reprinted from ASCE
Manual #41: Plastic Design in Steel: A Guide and Commentary, 2nd ed.,
with permission of American Society of Civil Engineers.)
136 Chapter Three

1.0
bt
0.5 =
wh

Rectangle 1.0

Most sections
P
0.5 Af
Py
x
2.0
y y 1.5

x
Af
Aw

0
0 0.5 1.0
Mpc
Mp
(c)

Figure 3.11 (Continued )

3.2.3 Wide-Flange Sections: Weak-Axis Bending

Normalized M-P interaction equations can be obtained for wide-
flange sections in weak-axis bending following an approach similar
to that presented above for strong-axis bending. Two cases must be
considered, depending on whether the neutral axis falls in the web
(yo ≤ w/2) or outside the web (w/2 < yo ≤ b/2). For the former:

2
Mpr  P  A2 P wd
= 1-   for ≤ (3.26)
Mp  Py  4dZy Py A

whereas, for the latter case:

Mpr 4bt  P  P   A2  P wd

= - 1 -   1 -   for > (3.27)
Mp  A  Py    Py   8tZy  Py A
 

Again, for most commonly available wide-flange sections, the

normalized M-P interaction curve is a function of the ratio of the
flange area to the web area. Normalized interaction curves expressed
as a function of that ratio are shown in Figure 3.11c. This time, because
 Plastic Behavior at the Cross-Section Level 137

a wide-flange shape in weak-axis bending is simply two rectangular

cross-sections joined at their centroids by a thin member, the wide-
flange sections that have the largest ratio of flange-to-web areas will
have the interaction curves closest to those for a rectangular cross-
section, as can be observed from Figure 3.11c. Again, the relatively
narrow shaded area in that figure represents the range of flange-to-
web area ratios corresponding to most wide-flange sections rolled
today. Based on this observation, the following design interaction
curve expression for the plastic strength of wide-flange cross-sections
in weak-axis bending has been proposed:

M   P  2
= 1 . 19 1 -    ≤ 1 . 0 (3.28)
Mp   Py  

In some instances, however, simpler but more conservative equa-

tions have also been used in some design codes (e.g., CSA 2009).

3.2.4 Moment-Curvature Relationships

If the complete moment-curvature relationship for a given axial load
and cross-section is desired, the use of numerical techniques is rec-
ommended, even though some closed-form solutions can be devel-
oped for the simplest cross-sections. In fact, to account for combined
flexural and axial loading, only minor changes are necessary to the
moment-curvature algorithm described in the previous section.

3.3 Combined Flexural and Shear Loading

The interaction between axial force and moment could easily be con-
sidered in the previous section because they both produce axial
strains in a structural member. However, before one can consider the
combined effect of flexural and shear loading on the plastic moment
of a cross-section, the interaction between axial and shear stresses
when yielding is reached must first be described. Based on experi-
mental observations, two mechanics-of-materials rules have been
formulated to describe this fundamental material behavior: the Tresca
andtheVonMisesyieldconditions(Popov1968).Thelatterhasbeen
most widely used to describe aspects of the behavior of steel struc-
tures.TheVonMisescriterioncanbeexpressedasfollows:

s 2 + 3t 2 = s 2y (3.29)

where s is the axial stress, t is the shear stress, and sy is the yield
stress in uniaxial tension.
According to that criterion: (1) no shear stress, t, can be applied
when the axial stresses reach yield, and (2) in absence of axial stresses,