SteelDay Eve 2013

The Evolution of Stability Provisions in the AISC Specification

Charles J. Carter1

Introduction
Buildings and the structural systems that support them have evolved, as have the
requirements and practices upon which their design and construction are based. The
treatment of stability effects has varied significantly through the history of (and before)
the AISC Specification. Despite differences in the provisions in AISC Specifications of
differing eras, it can be generalized that stability effects are addressed in all AISC
Specifications in some combination of three features in the AISC Specification: the
column buckling equation(s), the compression and flexure interaction equation(s), and the
analysis requirements.

In this paper, the evolution of stability analysis and design provisions will be traced
through observation of the changes made in these three areas up to the most current
requirements: those included in the 2005 and 2010 versions of AISC 360 Specification
for Structural Steel Buildings (AISC, 2005; AISC, 2010). These historical developments
will be summarized by time period as follows:

1. Before the AISC Specification Existed

2. The First AISC Specification – 1923
3. Developments Leading to the 1963 AISC Specification
4. Refinements in Allowable Stress Design (1969 through 1989)
5. Load and Resistance Factor Design (1986 through 1999)
6. The 2005 and 2010 AISC Specifications

Additionally, the influence of changing characteristics of buildings and the structures

supporting them will be summarized.

A frame of reference is needed to make more sense out of the historical developments as
they are reviewed. The best frame of reference may be the current state-of-the-art, which
can be found in Chapter C and Appendix 7 in the 2005 AISC Specification, and Chapter
C and Appendices 7 and 8 in the 2010 AISC Specification. A detailed discussion is
available in the corresponding Commentary sections, but for the purposes of this paper it
is sufficient to review the list provided in Section C1.1 of factors that must be accounted
for in stability analysis and design:

1. Flexural, shear, and axial deformations – these are the member deformations and
all other component and connection deformations that contribute to displacements
of the structure;

1
Charles J. Carter, S.E., P.E., is Chief Structural Engineer at the American Institute of
Steel Construction, Chicago, IL

1
 2. Second-order effects – these are the increases that occur in forces and moments
due to displacements of the structure induced by the loads, including both P-
effects (displacements of points of intersection of members) and P- effects
(deformations of the members between points of intersection);
3. Geometric imperfections – these are the initial out-of-plumbness of the structure
and the initial out-of-straightness of the members;
4. Stiffness reductions due to inelasticity – these are the effects of residual stresses;
and,
5. Variability in component and system stiffness – these are the effects of variations
in material and cross-sectional properties of members, as well as the other effects
generally accounted for in the resistance factors (LRFD) and safety factors
(ASD).

As will be explained in greater detail when the 2005 and 2010 AISC Specification are
discussed, these factors combine to affect both the demand side (the analysis results) and
the strength side (the member strength equations). It will also be highlighted throughout
the review of historical developments which of these effects were considered and which
were not.

Griffis and White (2013) provide an excellent compilation of current state-of-the-art for
stability analysis and design. In preparing that publication, the history of column design,
interaction, and stability was explored. That work is used extensively in this paper, and
credited as Griffis and White (2008) because it was not included in the final manuscript
of Griffis and White (2013)

Before the AISC Specification Existed

Griffis and White (2008) provide an excellent summary of stability-related developments
in the pre-1923 era, including some of the very early developments:

“While columns and frames have been in use for many centuries, it was not until
1729 that van Musschenbroek published the first paper concerning the strength of
columns (Salmon, 1921). An empirical column curve was presented for a rectangular
column taking the form,

bd 2
Pk
L2

where P is the column strength, b and d are the column width and depth, respectively,
L is the column length, and k is an empirical factor. Interestingly, this equation has
remarkable similarity to those still in use today.

In 1759, Euler published his now famous treatise on the buckling of columns
(Salmon, 1921). The original buckling load determined by Euler was for a column
with one end fixed and the other free – a flagpole column. His equation took the form,

2
  2C
P
4L2

where P is the buckling strength, L is the column length, and the constant C is the
“absolute elasticity”, which was defined as depending on the elastic properties of the
material. Euler was the first to recognize that column strength could also be a
problem of stability and not just a matter of crushing the material. Euler investigated
the purely elastic phenomenon of buckling.

We know today that elastic instability of columns occurs only with very slender
columns, and the theories that define inelastic column strength began to emerge over
a century after the above developments in elasticity. Engesser published his tangent
modulus theory in 1889, and followed this in 1895 with a revised theory called the
reduced modulus theory.

The reduced modulus theory of inelastic buckling was accepted as the correct
buckling theory until 1947 when Shanley published a paper giving the buckling load
of a centrally loaded column as the tangent modulus load (Beedle, 1964). The critical
buckling stress was given by the equation,

 2 Et
Fcr  2
L
 
r

Et is the tangent modulus of the stress-strain relationship of the material at the critical
stress. Indeed, in 1924 the forerunner to the Column Research Council declared this
tangent modulus equation as the proper basis for establishing column load formulas
(Beedle, 1964).

The first discussion on problems of the stability of members that were part of
rectangular frames came in 1893 by Engesser (Timoshenko and Gere, 1961). Yet it
would not be understood until the late 1940s that the key to the tangent modulus
concept for steel column buckling was the inclusion of the effects of residual stresses
that existed in the cross-section of the column even before the application of external
load.”

Regarding the five factors that influence stability, it is not surprising that few of them
were explicitly addressed. Implicitly, however, there were compensating factors that
explain why buildings rarely experienced problems with stability. Workmanship
requirements for materials, fabrication, and erection, the customary use of heavy masonry
infill details that added uncalculated strength and stiffness, and the factors of safety used
probably served to manage P- and P- effects, member, component, and connection
deformations, and the effects of geometric imperfections in the erected structure.
Analysis methods of this era were also conservative by nature, and although the term

3
“skyscraper” had come into existence, it described buildings that were on the order of 10
stories tall.

It is interesting that the concept that would come to be known as effective length and
consideration of the impact of residual stresses date as far back as these works.
Nonetheless, the state of knowledge was focused on the column buckling equation.
Recognition had not yet been made of the role of combined compression and flexure in
column behavior; nor had the relationship between stability effects and analysis been
realized.

The First AISC Specification – 1923

In 1923, AISC published the work upon which five eminent engineers collaborated – the
Specification of the American Institute of Steel Construction, Inc. for the Design
Fabrication, and Erection of Structural Steel for Buildings (AISC, 1923). Their
deliberations resulted in nine pages of text in a document that was proposed for
acceptance by the engineering community and steel construction industry. We know
today that this document became the cornerstone upon which all else since has been built.

The allowable stress for column buckling was given as:

18,000
Fa   15,000
l2
1
18,000r 2

where l is the unsupported length of the column and r is the corresponding least radius of
gyration of the section. Thus, the 15,000 psi allowable plateau applied for values of l/r up
to 60 and the parabolic formulation controlled thereafter. It was also allowed for “short
lengths or where lateral deflection is prevented” to use Fa = 18,000 psi.

Consideration of the combination of stresses was required, but the basis of the
requirements was fairly crude: “Members subject to both direct and bending stresses shall
be so proportioned that the greatest combined stresses shall not exceed the allowed
limits.” There was no mention of second-order effects, amplification factors, effective
length factors, or overall frame behavior (Griffis and White, 2008). Thus, it remained in
this period that the impact of stability on interaction and analysis had not been
recognized.

Regarding the five factors that influence stability, it continued that few of them were
explicitly addressed. However, for most buildings, the aforementioned implicit
compensating factors continued to exist as well, albeit with one important exception.
Such landmark structures as the Empire State Building, Manhattan Tower, and Chrysler
Building were being designed to rise to heights approaching and exceeding 1,000 ft
despite a continuing lack of clear understanding of system buckling, secondary effects in
frames, and effective length factors (Griffis and White, 2013). Certainly the
compensating factors were now beginning to be tested.

4
It is also interesting to note how performance-oriented some of the text was in those
original nine pages:

1. The writers commented to the reader in their introduction that “The question of
design is all-important. It necessarily presupposes that the design is good, made
by and executed under the supervision of competent structural engineers; that
proper provision is made for secondary stresses, excentric [sic] loads, unequal
distribution of stresses on rivets, etc.; that the details are suitable and that the
workmanship is high grade.”.
2. Section 2 included general requirements that “To obtain a satisfactory structure,
the following major requirements must be fulfilled. (a) The material used must be
suitable, of uniform quality, and without defects affecting the strength or service
of the structure. (b) Proper loads and conditions must be assumed in the design.
(c) The unit stresses must be suitable for the material used. (d) The workmanship
must be good, so that defects or injuries are not produced in the manufacture. (e)
The computations and design must be properly made so that the unit stresses
specified shall not be exceeded, and the structure and its details shall possess the
requisite strength and rigidity.
3. Section 9 repeated that “Full provision shall be made for stresses caused by
excentric [sic] loads.”
4. Section 22(a) required that “The frame of all steel skeleton buildings shall be
carried up true and plumb, …”. The first AISC Code of Standard Practice, which
was also a proposed standard when first published by AISC in 1924, established
this as no greater than 1/500 for interior columns nor 1/1000 for exterior columns.

Some of these requirements may have been intended to mitigate the effects of stability,
but the effectiveness in doing so probably related more to the use of traditional
techniques than any meaning that might have been taken from these general statements.

Developments Leading to the 1963 AISC Specification

After the first AISC Specification was published in 1923, several revisions were issued
up to and including the 1963 AISC Specification. This period was a time of growth in
both knowledge and application, but treatment of stability evolved at a much slower pace.
Griffis and White (2008) summarize this as follows:

“By the 1936 AISC Specification, column design was based upon the following
equations:

For l/r < 120

2
l
Fa  17,000  0.485 
r

5
 For l/r ≥ 120

18,000
Fa 
l2
1
18,000r 2

Considerable research was undertaken in the late 1940s and 1950s studying the
influence of residual stresses and other factors such as initial out-of-straightness,
eccentricity of load, end fixity, transverse loads and the effect of local and lateral
buckling on column strength. This work culminated with the Column Research
Council (now Structural Stability Research Council - SSRC) publishing column
strength curves that serve as the basis for many code provisions today (CRC, 1960).

The tangent modulus concept for steel column design gave way to direct
consideration of the effects of residual stresses, which it was discovered existed in the
cross-section of the column even before the application of external load. The effects
of initial out-of-straightness, eccentricity of load, end fixity, transverse loads, and
local buckling on column strength became known.

Column buckling provisions were based upon the Johnson parabola (CRC, 1960) in
the inelastic range until it merged with the Euler curve for elastic strength.

Members subject to both axial and bending stresses were proportioned by an

interaction equation reflecting a simple combination of stresses. There was no
mention of second-order effects, amplification factors, effective length factors, or
overall frame behavior.

Although the concept of effective length was discussed in the Commentary that
accompanied the 1961 AISC Specification, it was not until the 1963 AISC
Specification that the effective length, KL, became explicit in the AISC Specification.
The now-well-known alignment chart2 was first published in a comprehensive paper
(Kavanaugh, 1962) discussing column and frame buckling.

New strength formulas were introduced for columns based upon the basic column
strength estimate suggested by the Column Research Council (CRC, 1960). An
amplification factor was introduced in one of the two interaction equations to account
for the fact that lateral displacement generates a secondary moment that must be
accounted for in the member bending stress. For the first time, stability against
sidesway of a frame was recognized in the interaction equations and design
procedure.

2
First introduced from unpublished notes in 1959 as the J & L Charts by Julian and Lawrence for
incorporation into the Boston Building Code (Griffis and White, 2013).

6
 The resulting equations for column buckling were:

For KL/r < Cc



KL 
r
2


1  2  y
F
 2C c 
Fa   
FS

For KL/r ≥ Cc

149 ,000 ,000

Fa 
KL r  2

where
5 3
FS  
KL    
r 
KL
r
3

3
3 8C c 8C c
2 2 E
Cc 
Fy

The resulting interaction equations were:

For fa/Fa ≤ 0.15

fa f
 b  1.0
Fa Fb

For fa/Fa > 0.15

fa C m Fb
  1.0
Fa  fa 
1  '  Fb
 Fe

The term (1 – fa / Fe') was the new amplification factor and Cm was a moment
modifier term.”

Regarding the five factors that influence stability, the substantial growth in knowledge in
this era resulted in the explicit inclusion of many stability factors in design provisions and
more explicit consideration of stability effects in design office practices. The treatment of
stability that evolved in this era continued to be based largely on modification made on
the design-side of the equations, however. The effects of stability on the analysis side
remained to be developed.

7
Additionally, the introduction of K was not without compromise. The methods available
to calculate K in all but the most simple of cases required a number of assumptions, few
and often none of which were actually satisfied in real structures (Kavanaugh, 1962).
Regardless, what would come to be known as the effective length method was accepted
because it did something and that was better than doing nothing.

This era also marked the beginning of changes in the way buildings were designed and
constructed, particularly at the end of it. Building systems were evolving and architecture
was demanding changes in the way structures were configured. Computational
techniques were on the verge of sophistication. Stability effects were soon to become a
by-product of these changes in methods and technologies.

Refinements in Allowable Stress Design (1969 through 1989)

There were three major revisions of the AISC Specification made in this time period:

1. The 1969 AISC Specification

2. The 1978 AISC Specification
3. The 1989 AISC (ASD) Specification3

The Column Research Council was now known as the Structural Stability Research
Council, and their work fed into the AISC Specification. None of these revisions made
any substantive changes to the column buckling equations or the interaction equations.
There were other developments, however, and Griffis and White (2008) summarize the
1969 and 1978 developments as follows:

“The 1969 Specification for the first time explicitly mentioned the word stability in
the provisions and required that “General stability shall be provided for the structure
as a whole and for each compression element”. More attention was given to overall
frame behavior as opposed to member behavior. The Commentary to Section 1.8
Stability and Slenderness Ratios focused more on determination of the effective
length factor K which was a subject of much attention among practitioners during this
period. The Commentary referred to the SSRC Guide second edition (Johnston ,1966)
which covered the subject in some detail.

It was not until the 1978 Specification (AISC, 1978) that the subject of structural
analysis and determination of secondary effects on frames was covered, albeit in the
Commentary to Section 1.8 on Stability and Slenderness Ratios. For the first time the
term “P-Δ” was used and its impact on frame behavior highlighted. Reference was
made to research at Lehigh University on the load carrying capacity of rigid
multistory frames subjected to gravity and lateral loads using second-order analysis
methods. This discussion also referenced a fairly comprehensive treatment of this
subject in the third edition of the SSRC Guide (Johnston, 1976).”

3
All AISC Specifications that predated the 1989 ASD Specification were written using the allowable stress
design method. The load and resistance factor design method was introduced in 1986 is omitted from this
section and discussed in the next section.

8
The 1989 ASD Specification was little more than an editorial reorganization of the 1978
AISC Specification to align with the format and organization of the new, alternative 1986
LRFD Specification. There were some substantive changes, but changes in stability
considerations were insignificant in spite of the significant advancements in stability
analysis and design requirements that had been made in the 1986 LRFD Specification.
This may have contributed to a mistaken perception in the engineering community that
stability was a concern when the LRFD method was used, but not when the ASD method
was used.

In spite of the advancements noted above, considerations of stability effects in design

office practice largely remained constant – and usually were limited to whatever effects
were considered by the terms that explicitly had been added to the equations in the AISC
Specification. Changes in buildings literally soared in almost every respect. The heights
of skyscrapers reached the pinnacle of the century, and every system used in building
architecture, structure, and mechanics was being developed and innovated. The curtain
wall and open floor plans with longer and longer spans dominated.

Even the marketplace was changing, with construction management firmly split as a
discipline away from the role of the architect. CM influence would further drive the
lightness and flexibility of structural systems as economic evaluations dictated that
framing be economized by more exact designs and limited use of lateral framing.

The more advanced projects like the Sears Tower, John Hancock Tower, and Standard
Oil Building in Chicago and World Trade Center Towers in New York City undoubtedly
received significant attention to stability effects. Nonetheless, building systems were
changing and this impact would manifest itself in many ways, including a heightened
need for more advanced stability analysis and design requirements.

The John Hancock Tower in Boston is perhaps the most notable example of a building
with design and construction aspects that outpaced the technology of the code in the area
of stability. A project that ultimately was beset with a number of different structural
issues, stability was cited as the reason that the majority of the panels in its all-glass
façade popped out. Other examples exist, but none are as dramatic in their illustration of
the need for further development of provisions to address stability in the design of
buildings.

Load and Resistance Factor Design (1986 through 1999)

Concurrent with the development of the load and resistance factor design method as a
replacement for allowable stress design, significant work was being done by AISC, the
Structural Stability Research Council (SSRC), and others to advance the state-of-the-art
of stability analysis and design. These efforts came together, and stability design
requirements were advanced with the release of the first LRFD Specification in 1986.

9
A new column curve was also introduced into this new Specification (Griffis and White,
2008), based on column strength curve 2P of the 4th edition of the SSRC Guide
(Galambos, 1988). The resulting equations for column buckling were:

For c ≤ 1.5

 2
Fcr  0.85 0.658c Fy 
For c > 1.5

 0.877 
Fcr  0.85  2  Fy
 c 

where

Kl Fy
c 
r E

Second-order effects were required to be included in the analysis results (that is, included
in the calculation of Pu and Mu), which simplified the resulting interaction equations:

For Pu/Pn ≥ 0.2

Pu 8  M ux M 
   uy   1.0
Pn 9  M nx M ny 

For Pu/Pn < 0.2

Pu  M M uy 
  ux    1.0
2Pn  M nx M ny 

Griffis and White (2008) summarize other significant developments as follows:

“A more comprehensive treatment of stability of frames and second-order effects was

introduced in the new LRFD Specification published in 1986 (AISC, 1986). Here, as
part of a reorganization of the Specification provisions, an entire Chapter C was
devoted to the subject of frame behavior. For the first time the Specification
specifically required that “Second-order (P-Δ) effects shall be considered in the
design of frames.” Requirements were placed on the structural analysis to include
axial deformations and the effects of frame instability under ultimate loads, a point
not always realized under the allowable stress method. … In addition, continuing a
trend toward specifying requirements for the structural analysis in order to properly

10
 address second-order effects, Chapter H on combined forces contained an
approximate second-order analysis procedure, introducing the now commonly used
B1-B2 method. The second-order amplification factor B2 was permitted to be
calculated using either the story buckling approach or the story stiffness approach.
Frame stability was recognized as a system or story buckling phenomenon.

The 1993 LRFD Specification in Chapter C on frames expanded the stability

treatment of steel frame structures – both braced and unbraced. For the first time, the
“destabilizing effects of gravity loaded columns…” were required to be considered in
the moment frame analysis and design. This important fact was frequently overlooked
in many building designs before this requirement was formally made a part of the
Specification. Emphasis was placed in the Commentary to Chapter C on acceptable
methods to calculate the effective length factor K, given the difficulty of this topic for
practicing engineers. Consideration of the leaning column effect and its use in frames
not meeting the requirements of the alignment chart commonly used in practice was
introduced. The effect of leaning columns on different versions of K factor equations
was discussed in detail. The B1 – B2 method was moved from Chapter H on combined
forces to Chapter C on frames.

In the 1999 Specification, Chapter C was expanded to include a discussion of stability

bracing of frames as well as column and beam bracing. Specific stability bracing
requirements emerged for braced frames where a minimum strength and stiffness
requirement was placed on the story or panel of a steel building. Equations for the
strength and stiffness requirements for braces in columns and beams were
introduced.”

These advancements were largely lost to the engineering community when debate about
ASD and LRFD ensued. Indeed, this duality prevailed throughout this era with LRFD
being developed further while ASD lay fallow – and without the advancements in
stability analysis and design requirements that would have been made in it had LRFD
never come into existence. It was unfortunate that this translated into the perception that
stability was an LRFD issue, not an ASD issue.

These advancements also were made within the context of the effective length method.
As a result, they merely pecked at the problems with that method that were mentioned
previously and ignored many new problems.

The 2005 AISC Specification

As the 1990s gave way to the new century, AISC established a direction to resolve the
debate between ASD and LRFD advocates. A unified specification was created and the
previously separate ASD and LRFD methods were combined, using the best of both as
the single approach for the future. At about the same time, the developmental work being
done by AISC, SSRC, and others on stability analysis and design had achieved critical
mass. This landmark work was incorporated into the 2005 AISC Specification.

11
The equations for column buckling were returned to more familiar terms. The LRFD
alternative is shown below4:

For KL/r ≤ 800/√Fy

 Fy

Fcr  0.9  0.658 Fe  Fy
 
 

For c > 1.5

Fcr  0.9  0.877 Fe 

where

 2E
Fe  2
 KL 
 
 r 

Second-order effects were required to be included in the analysis results, which

simplified the resulting interaction equations:

For Pr/Pc ≥ 0.2

Pr 8  M rx M ry 
   1.0
Pc 9  M cx M cy 

For Pr/Pc < 0.2

Pr  M rx M ry 
   1.0
2 Pc  M cx M cy 

In these equations, the numerator of each term represents the demand determined by
analysis and the denominator of each term represents the corresponding available
strength determined with the member strength provisions in the AISC Specification. All
quantities are calculated using LRFD load combinations and resistance factors, or ASD
load combinations and safety factors.

Also, for the first time, the impact of stability was addressed properly on both the demand
side (the analysis results) and the strength side (the member strength equations). The

4
The ASD alternative is similar with  = 1.67 used as a divisor instead of the  = 0.9 used as a multiplier.

12
2005 AISC Specification included three prescriptive approaches for stability analysis and
design:

1. The Direct Analysis Method (Appendix 7)

2. The Effective Length Method (Section C2.2a)
3. The First-Order Analysis Method (Section C2.2b)

Each of these methods is further illustrated in the information provided in Attachment A

(Carter and Geschwindner, 2008).

The Direct Analysis Method was new in 2005. The Effective Length Method remains
similar to the traditional approach, but has been modified. The First-Order Analysis
Method was new in 2005, and offers an approach that permits a first-order analysis but
still satisfies the requirements of the AISC Specification. These methods also are
addressed in the 2010 AISC Specification (AISC, 2010) with simplifications, refinements
and clarifications based upon usage sine the 2005 AISC Specification was published.

Direct Analysis Method

From Carter and Geschwindner (2008):

“The direct analysis method is permitted for any ratio of second-order drift, 2nd, to
first-order drift, 1st, and required when this ratio exceeds 1.5. It requires the use of:

1. A direct second-order analysis or a first-order analysis with B1-B2 amplification.

2. The nominal frame geometry with an additional lateral load of Ni = 0.002Yi,
where Yi is the total gravity load on level i from LRFD load combinations, or 1.6
times ASD load combinations.
3. The reduced stiffnesses EA* and EI* (including in B1-B2 amplification, if used).
4. LRFD load combinations, or ASD load combinations multiplied by 1.6. This
multiplier ensures that the drift level is consistent for LRFD and ASD when
determining second-order effects. The forces and moments obtained in this
analysis are then divided by 1.6 for ASD member design.

The following exceptions apply as alternatives in item 2 above:

a. If the out-of-plumb geometry of the structures is used, the notional loads can be
omitted.
b. When the ratio of second-order drift to first-order drift is equal to or less than 1.5,
the notional load can be applied as a minimum lateral load, not an additional
lateral load. Note that the unreduced stiffnesses, EA and EI, are used in this
comparison.
c. When the actual out-of-plumbness is known, it is permitted to adjust the notional
loads proportionally.

For all frames designed with this method, K = 1.0.”

13
Regarding the five factors that influence stability, it is not surprising that all of them are
explicitly addressed in this method – the direct analysis method represents the current
state-of-the-art:

1. Member, component, and connection deformations are addressed directly in the

analysis.
2. Second-order effects (both P- and P- effects) are addressed directly in the
analysis, either by rigorous second-order analysis or a first-order analysis with B1-
B2 amplification;
3. Structural out-of-plumbness is addressed with the use of notional loads (or direct
modeling of the initial out-of-plumbness).
4. Member out-of-straightness is accounted for in the column design equations for
its effect on member strength, and in the use of a reduced stiffness for its effect on
the structure stiffness.
5. Residual stresses are accounted for in the column design equations for their effect
on member strength, and in the use of a reduced stiffness for their effect on the
structure stiffness.
6. Variability in component and system stiffness is accounted for in the resistance
and safety factors for its effect on member strength, and in the use of a reduced
stiffness for its effect on the structure stiffness.

Effective Length Method

From Carter and Geschwindner (2008):

“[This] is essentially the traditional effective length method with an additional

requirement for a minimum lateral load. It is permitted when the ratio of second-order
drift, 2nd, to first-order drift, 1st, is equal to or less than 1.5, and requires the use of:

1. A direct second-order analysis or a first-order analysis with B1-B2

amplification.
2. The nominal frame geometry with a minimum lateral load (a “notional load”)
Ni = 0.002Yi, where Yi is the total gravity load on level i from LRFD load
combinations (or 1.6 times ASD load combinations). This notional load is
specified to capture the effects of initial out-of-plumbness up to the AISC
Code of Standard Practice maximum value of 1:500. In this method, Ni is not
applied when the actual lateral load is larger than the calculated notional load.
3. The nominal stiffnesses EA and EI.
4. LRFD load combinations, or ASD load combinations multiplied by 1.6. This
multiplier on ASD load combinations ensures that the drift level is consistent
for LRFD and ASD when determining second-order effects. The forces and
moments obtained in this analysis are then divided by 1.6 for ASD member
design.

When the ratio of second-order drift to first-order drift, which is given by B2, is equal
to or less than 1.1, K = 1.0 can be used in the design of moment frames. Otherwise,

14
 for moment frames, K is determined from a sidesway buckling analysis. Section
C2.2a(4) indicates that for braced frames, K = 1.0.”

Regarding the five factors that influence stability, it is slightly less obvious how the
simple addition of a notional load to the traditional effective length method could result
in all of them being explicitly addressed in this method. Nonetheless:

1. Member, component, and connection deformations are addressed directly in the

analysis.
2. Second-order effects (both P- and P- effects) are addressed directly in the
analysis, either by rigorous second-order analysis or a first-order analysis with B1-
B2 amplification;
3. Structural out-of-plumbness is addressed with the use of notional loads.
4. Member out-of-straightness is accounted for in the column design equations for
its effect on member strength.
5. Residual stresses are accounted for in the column design equations for their effect
on member strength.
6. Variability in component and system stiffness is accounted for in the resistance
and safety factors for its effect on member strength.
7. Note that K must be calculated (exceptions above noted) and used to account for
the effects on structure stiffness of member out-of-straightness, residual stresses,
and variability in component and system stiffness.

The basis in the use of K in this method means that some of the aforementioned
limitations and criticisms of the effective length method are still applicable.

First-Order Analysis Method

From Carter and Geschwindner (2008):

“The first-order analysis method is permitted when:

1. The ratio of second-order drift, 2nd, to first-order drift, 1st, is equal to or less
than 1.5.
2. The column axial force Pr ≤ 0.5Py, where  = 1.0 for LRFD, 1.6 for ASD.

This method requires the use of:

1. A first-order analysis.
2. The nominal frame geometry with an additional lateral load Ni = 2.1(/L)Yi ≥
0.0042Yi, applied in all load cases.
3. The nominal stiffnesses EA and EI.
4. B1 as a multiplier on the total moment in beam-columns.
5. LRFD load combinations, or ASD load combinations multiplied by 1.6. This
multiplier on ASD load combinations ensures that the drift level is consistent
for LRFD and ASD when determining the notional loads. The forces and

15
 moments obtained in this analysis are then divided by 1.6 for ASD member
design.

For all frames designed with this method, K = 1.0.”

Regarding the five factors that influence stability, it is even less obvious how they are
explicitly addressed in this method. This is for good reason – they are all satisfied
implicitly because the method is a mathematical manipulation of the Direct Analysis
Method based upon the characteristics of typical building frames. For this reason, this
method should be used with care to ensure that it is appropriate.

A General Observation About the Progression of Provisions and Design Practices

As can be seen in the descriptions, the stability provisions and corresponding designs
have always been related, with one advancing slightly and affecting the other in a stair-
stepping fashion. Early in the time-span, stability design provisions were comparatively
crude, but they were acceptable then because framing systems were very conservative
and very redundant in the typical building. As provisions advanced, framing systems
changed. Stability problems sometimes influenced changes in provisions, systems, or
both.

Today’s provisions would have been unnecessary in the early days of the portal-frame
building with heavy masonry infill. Yesterday’s provisions are inadequate for today’s
economized and open buildings. Tomorrow’s buildings and provisions will undoubtedly
follow a similar trend.

Conclusions
The preceding historical summary of developments in stability analysis and design
provisions shows an extensive path of progress from the first basis of column design in
elastic buckling to the current state-of-the-art. Additionally, the key developments,
particularly those made in the AISC Specification, have been highlighted.

The most current methods – those presented in the 2005 and 2010 AISC Specifications –
have been available for several years now, and are being used in design offices and
implemented by companies that write software for use in design offices. The Direct
Analysis Method, which first appeared in the 2005 AISC Specification, offers a clean
approach that sheds the deficiencies that have been noted in the methods of the past.

References
American Institute of Steel Construction, Inc. (AISC) (1923), Specification of the
American Institute of Steel Construction, Inc. for the Design Fabrication, and
Erection of Structural Steel for Buildings, Chicago, IL
American Institute of Steel Construction, Inc. (AISC) (1936), Specification for the
Design, Fabrication and Erection of Structural Steel for Buildings, Chicago, IL
American Institute of Steel Construction, Inc. (AISC) (1961), Specification for the
Design, Fabrication and Erection of Structural Steel for Buildings, Chicago, IL

16
American Institute of Steel C onstruction, Inc. (AISC) (1963), Specification for the
Design, Fabrication and Erection of Structural Steel for Buildings, Chicago, IL
American Institute of Steel C onstruction, Inc. (AISC) (1969), Specification for the
Design, Fabrication and Erection of Structural Steel for Buildings, Chicago, IL
American Institute of Steel C onstruction, Inc. (AISC) (1978), Specification for the
Design, Fabrication and Erection of Structural Steel for Buildings, Chicago, IL
American Institute of Steel Construction, Inc. (AISC) (1989), Specification for Structural
Steel Buildings, Chicago, IL
American Institute of Steel Construction, Inc. (AISC) (1986), Load and Resistance
Factor Design Specification for Structural Steel Buildings, Chicago, IL
American Institute of Steel Construction, Inc. (AISC) (1989), Specification for Structural
Steel Buildings – Allowable Stress Design and Plastic Design, Chicago, IL
American Institute of Steel Construction, Inc. (AISC) (1999), Load and Resistance
Factor Design Specification for Structural Steel Buildings, Chicago, IL
American Institute of Steel Construction, Inc. (AISC) (2005), Specification for Structural
Steel Buildings, ANSI/AISC 360-05, Chicago, IL.
American Institute of Steel Construction, Inc. (AISC) (2010), Specification for Structural
Steel Buildings, ANSI/AISC 360-10, Chicago, IL.
Beedle, L, et al, Structural Steel Design, Ronald Press Company, New York, 1964
Bleich, F., Buckling Strength of Metal Structures, McGraw Hill Book Com pany, New
York, 1952
Carter, C. and Geschwindner, L., (2008) “A Comparison of Fra me Stability Analysis
Methods In ANSI/AISC 360-05”, 3 rd Qtr. AISC Engineering Journal, Volume 45,
No. 3, pp. 159-170.
Column Research Council (CRC) (1952), The Basic Column Formula, CRC Technical
Memorandum No. 1, May.
Column Research Council (CRC) (1960), Guide to Design Criteria for Metal
Compression Members, Column Research Council.
Griffis, L., and W hite, D., (2013) Stability of Structural Steel Buildings, AISC Design
Guide No. 28, to be published in 2013
Griffis, L., and White, D., (2008) Unpublished history of column design, interaction, and
stability, originally prepared for AISC Design Guide No. 28.
Johnston, B.G. (1966), Guide to Design Criteria for Metal Compression Members, Edited
by B. Johnston, Colum n Research Counc il, John W iley and Sons, New York, 1976
(2rd ed).
Johnston, B.G. (1976), Guide to Stability Design Criteria for Metal Structures, Edited by
B. Johnston, Structural Stability Research Council, John Wiley and Sons, New York,
1976 (3rd ed).
Kavanaugh, T. C. (1962) "Effectiv e Length of Fram ed Columns," Transactions of the
American Society of Civil Engineers, Vol. 127, pp81-101.
Salmon, E.H., Colum ns, A Treatise on the Strength and Design of Compression
Members, Oxford Technical Publications, London, 1921
Timoshenko,S. and Gere J. (1961). Theory of Elastic Stability, McGraw Hill Book
Company, New York.

17
 APPENDIX

(Carter and Geschwindner paper from 2008 AISC Engineering Journal)

18
 A Comparison of Frame Stability
Analysis Methods in ANSI/AISC 360-05
CHARLES J. CARTER and LOUIS F. GESCHWINDNER

A NSI/AISC 360-05 Specification for Structural Steel

Buildings (AISC, 2005a), hereafter referred to as the
AISC Specification, includes three prescriptive approaches
two flagpole columns (Columns D and E) and two leaning
columns (Columns C and F). This frame is used with a drift
limit of L/400 to illustrate the simplifying effect a drift limit
for stability analysis and design. Table 2-1 in the 13th can have on the analysis requirements in each method.
Edition AISC Steel Construction Manual (AISC, 2005b), Although these example frames are not realistic frames,
hereafter referred to as the AISC Manual, provides a compar- the results obtained are representative of the impact of
ison of the methods and design options associated with each. second-order elastic and inelastic effects on strength require-
A fourth approach, referred to as the Simplified Method, is ments in real frames, particularly when the number of mo-
also presented in the AISC Manual (see page 2-12) and on ment connections is reduced. The loads shown in Figures
the AISC Basic Design Values cards. These four methods are 1 and 2 are from the controlling load and resistance factor
illustrated in this paper in order to give the reader a general design (LRFD) load combination and the corresponding de-
understanding of the differences between them: signs are performed using LRFD. The process is essentially
identical for allowable strength design (ASD), where ASD
1. The Second-Order Analysis Method (Section C2.2a)
load combinations are used with α = 1.6 as a multiplier,
2. The First-Order Analysis Method (Section C2.2b) when required in each method, to account for the second-
order effects at the ultimate load level.
3. The Direct Analysis Method (Appendix 7) When it is required to include second-order effects,
the B1-B2 amplification is used with a first-order analy-
4. The Simplified Method (Manual page 2-12; AISC sis throughout this paper. A direct second-order analysis
Basic Design Values cards) is straightforward and could have been used instead of the
B1-B2 amplification.
Two simple unbraced frames are used in this paper. The
one-bay frame shown in Figure 1 has a rigid roof element THE ONE-BAY FRAME
spanning between a flagpole column (Column A) and leaning
column (Column B). Drift is not limited for this frame, which A trial shape is selected using a first-order analysis without
results in a higher ratio of second-order drift to first-order consideration of drift limits or second-order effects. There-
drift, and allows illustration of the detailed requirements in after, that trial shape is used as the basis for comparison of
each method for the calculation of K-factors, notional loads, the four methods discussed earlier.
and required and available strengths. The three-bay frame
shown in Figure 2 has rigid roof elements spanning between

Charles J. Carter is vice president and chief structural

engineer, American Institute of Steel Construction,
Chicago, IL.
Louis F. Geschwindner is vice president of special projects,
American Institute of Steel Construction, and professor
emeritus of architectural engineering at Pennsylvania State
University, University Park, PA.
Fig. 1. One-bay unbraced frame used in examples.

ENGINEERING JOURNAL / THIRD QUARTER / 2008 / 159

159-170_Carter_Geschwinder_2008_3Q.indd 159 8/18/08 11:23:38 AM

 Selection of Trial Shape Based Upon Strength Design by Second-Order Analysis (Section C2.2a)
Consideration Only
Design by second-order analysis is essentially the traditional
Based upon the loading shown in Figure 1, the first-order effective length method with an additional requirement for
axial force, strong-axis moment, and design parameters for a minimum lateral load. It is permitted when the ratio of
Column A are: second-order drift, ∆2nd, to first-order drift, ∆1st, is equal to or
less than 1.5, and requires the use of:
Pu = 200 kips Mux = (20 kips) (15 ft)
Kx = 2.0 = 300 kip-ft 1. A direct second-order analysis or a first-order analysis
Ky = 1.0 Cb = 1.67 with B1-B2 amplification.
Lx = Ly = 15 ft Lb = 15 ft
2. The nominal frame geometry with a minimum lateral
Note that Kx = 2.0, the theoretical value for a column with a load (a “notional load”) Ni = 0.002Yi, where Yi is the
fixed base and top that is free to rotate and translate, is used total gravity load on level i from LRFD load combina-
rather than the value of 2.1 recommended for design in the tions (or 1.6 times ASD load combinations). This no-
AISC Specification Commentary Table C-C2.2. The value tional load is specified to capture the effects of initial
of 2.0 is used because it is consistent with the formulation out-of-plumbness up to the AISC Code of Standard
of the lateral stiffness calculation below. Note also that the Practice maximum value of 1:500. In this method, Ni
impact of the leaning column on Kx is ignored in selecting is not applied when the actual lateral load is larger than
the trial size, although it will be considered in subsequent the calculated notional load.
sections when Kx cannot be taken equal to 1 for Column A.
Out of the plane of the frame, Ky is taken as 1.0. 3. The nominal stiffnesses EA and EI.
A simple rule of thumb for trial beam-column selection is
4. LRFD load combinations, or ASD load combinations
to use an equivalent axial force equal to Pu plus 24/d times
multiplied by 1.6. This multiplier on ASD load com-
Mu, where d is the nominal depth of the column (Geschwind-
binations ensures that the drift level is consistent for
ner, Disque and Bjorhovde, 1994). Using d = 14 in. for a
LRFD and ASD when determining second-order ef-
W14, the equivalent axial force is 714 kips and an ASTM
fects. The forces and moments obtained in this analy-
A992 W14⫻90 is selected as the trial shape.
sis are then divided by 1.6 for ASD member design.
The lateral stiffness of the frame depends on Column A only
and is: When the ratio of second-order drift to first-order drift,
3 which is given by B2, is equal to or less than 1.1, K = 1.0
k = 3EI/L
can be used in the design of moment frames. Otherwise, for
= 3(29,000 ksi)(999 in.4)/(15 ft × 12 in./ft)3
moment frames, K is determined from a sidesway buckling
= 14.9 kips/in.
analysis. Section C2.2a(4) indicates that for braced frames,
The corresponding first-order drift of the frame is: K = 1.0.
∆1st = (20 kips)/(14.9 kips/in.) For the example frame given in Figure 1, the minimum lat-
= 1.34 in. eral load based upon the total gravity load, Yi, is:
Note that this is a very flexible frame with ∆1st/L = 1.34/ Yi = 200 kips + 200 kips
(15 ft ⫻ 12 in./ft) = 1/134. = 400 kips

Fig. 2. Three-bay unbraced frame used in examples.

160 / ENGINEERING JOURNAL / THIRD QUARTER / 2008

159-170_Carter_Geschwinder_2008_3Q.indd 160 8/19/08 1:03:52 PM

 Ni = 0.002 Yi Kx* = Kx(1 + ΣPleaning / ΣPstability)2
= 0.002 (400 kips) = 2.0(1 + 1)2
= 0.8 kips = 2.83
Because this notional load is less than the actual lateral load, The amplified axial force (Equation C2-1b) and associated
it need not be applied. For a load combination that did not design parameters for this method are:
include a lateral load, the notional load would need to be
Pr = Pnt + B2Plt
included in the analysis.
= 200 kips + 1.21(0 kips)
For Column A, using first-order analysis and B1-B2 amplifi- = 200 kips
cation: Kx* = 2.83, Ky = 1.0
Pnt = 200 kips, Plt = 0 kips
Lx = Ly = 15 ft
Mnt = 0 kip-ft, Mlt = 300 kip-ft
The amplified moment (Equation C2-1a) and associated
For P-δ amplification, since there are no moments associated
design parameters for this method are:
with the no-translation case, there is no need to calculate B1.
For P-∆ amplification, the first-order drift ratio is determined Mrx = B1Mnt + B2Mlt
from the calculated drift of 1.34 in. Thus, = (0 kip-ft) + 1.21 (300 kip-ft)
= 363 kip-ft
∆1st /L = (1.34 in.)/(15 ft ⫻ 12 in./ft)
= 0.00744 Cb = 1.67
For moment frames, Rm = 0.85 and from Equation C2-6b Lb = 15 ft
with ∆H = ∆1st and ΣH = 20 kips, Based upon these design parameters, the axial and strong-
ΣPe2 = Rm ΣH/(∆1st /L) axis available flexural strengths of the ASTM A992 W14×90
= 0.85 (20 kips)/(0.00744) are:
= 2,280 kips Pc = φc Pn
= 721 kips
For design by LRFD, α = 1.0 and ΣPnt is the sum of the
gravity loads. Thus, Mcx = φb Mnx
= 573 kip-ft
αΣPnt /ΣPe2 = 1.0 (200 kips + 200 kips)/2,280 kips
= 0.175 To determine which interaction equation is applicable, the
ratio of the required axial compressive strength to available
From Equation C2-3, the amplification is: axial compressive strength must be determined.
1
B2 = ≥1 Pr 200 kips
α Σ Pnt =
1− Pc 721 kips
Σ Pe 2
= 0.277
1
= ≥1
1 − 0.175 Thus, because Pr /Pc ≥ 0.2, Equation H1-1a is applicable.
= 1.21 ≥ 1.0
= 1.21 Pr 8 ⎛ M rx ⎞ 8 ⎛ 363 kip-ft ⎞
+ ⎜ ⎟ = 0.277 + ⎜
Pc 9 ⎝ M cx ⎠ 9 ⎝ 573 kiip-ft ⎟⎠
Because B2 = 1.21, the second-order drift is less than 1.5
= 0.840
times the first-order drift. Thus, the use of this method is
permitted. Because B2 > 1.1, K cannot be taken as 1.0 for
The W14×90 is adequate because 0.840 ≤ 1.0.
column design in the moment frame with this method. Thus,
K must be calculated, including the leaning-column effect.
Design by First-Order Analysis (Section C2.2b)
Several approaches are available in the AISC Specifica-
tion Commentary to include this effect. A simple approach The first-order analysis method is permitted when:
that uses the ratio of the load on the leaning columns to the
1. The ratio of second-order drift, ∆2nd, to first-order drift,
load on the stabilizing columns had been provided in previ-
∆1st, is equal to or less than 1.5.
ous Commentaries and is used here (Lim and McNamara,
1972): 2. The column axial force αPr ≤ 0.5Py, where α = 1.0 for
ΣPleaning /ΣPstability = (200 kips)/(200 kips) LRFD, 1.6 for ASD.
=1

ENGINEERING JOURNAL / THIRD QUARTER / 2008 / 161

159-170_Carter_Geschwinder_2008_3Q.indd 161 8/19/08 1:07:39 PM

 This method requires the use of: This moment must be amplified by B1 as determined from
Equation C2-2. The Euler buckling load is calculated with
1. A first-order analysis.
K1 = 1.0. Thus,
2. The nominal frame geometry with an additional lat- Pe1 = π2EI/(K1L)2
eral load Ni = 2.1(∆/L)Yi ≥ 0.0042Yi, applied in all load = π2(29,000 ksi)(999 in.4)/(1.0 × 15 ft × 12 in./ft) 2
cases. = 8,830 kips

3. The nominal stiffnesses EA and EI. The moment on one end of the column is zero, so the mo-
ment gradient term is:
4. B1 as a multiplier on the total moment in beam- Cm = 0.6 – 0.4(M1/M2)
columns. = 0.6 – 0.4(0/394 kip-ft)
= 0.6
5. LRFD load combinations, or ASD load combinations
multiplied by 1.6. This multiplier on ASD load com- From Equation C2-2,
binations ensures that the drift level is consistent for α Pr /Pe1 = 1.0(200 kips)/(8,830 kips)
LRFD and ASD when determining the notional loads. = 0.0227
The forces and moments obtained in this analysis are Cm
then divided by 1.6 for ASD member design. B1 = ≥1
αPr
1−
For all frames designed with this method, K = 1.0. Pe1
= 0.6 ≥ 1.0
For the example frame given in Figure 1, the additional lat-
eral load is based on the first-order drift ratio, ∆/L, and the 1 − 0.0227
total gravity load, Yi. Thus, with ∆ = ∆1st, = 0.614 ≥ 1.0
∆1st /L = (1.34 in.)/(15 ft × 12 in./ft) = 1.0
= 0.00744 The axial force and associated design parameters for this
Yi = 200 kips + 200 kips method are:
= 400 kips Pr = 200 kips
Kx = Ky = 1.0
Ni = 2.1(∆1st /L)Yi ≥ 0.0042Yi
Lx = Ly = 15 ft
= 2.1(0.00744)(400 kips) ≥ 0.0042(400 kips)
= 6.25 kips ≥ 1.68 kips The amplified moment and associated design parameters for
= 6.25 kips this method are:
It was previously determined in the illustration of design by Mrx = B1Mu
second-order analysis example that the second-order drift is = 1.0 (394 kip-ft)
less than 1.5 times the first-order drift. Additionally, = 394 kip-ft
αPr = 1.0(200 kips) Cb = 1.67
= 200 kips Lb = 15 ft
And for a W14×90, Based on these design parameters, the axial and strong-axis
0.5Py = 0.5Fy Ag available flexural strengths of the ASTM A992 W14⫻90
= 0.5(50 ksi)(26.5 in.2) are:
= 663 kips Pc = φcPn = 1,000 kips
Because ∆2nd < 1.5∆1st and αPr < 0.5Py, the use of this method Mcx = φbMnx = 573 kip-ft
is permitted. To determine which interaction equation is applicable, the
The loading for this method is the same as that shown in Fig- ratio of the required axial compressive strength to available
ure 1, except for the addition of a notional load of 6.25 kips axial compressive strength must be determined.
coincident with the lateral load of 20 kips shown, resulting Pr 200 kips
in a column moment, Mu, of 394 kip-ft. =
Pc 1, 000 kips
= 0.200

162 / ENGINEERING JOURNAL / THIRD QUARTER / 2008

159-170_Carter_Geschwinder_2008_3Q.indd 162 8/19/08 1:08:13 PM

 Thus, because Pr /Pc ≥ 0.2, Equation H1-1a is applicable. Thus, the notional load can be applied as a minimum lateral
load, and that minimum is:
Pr 8 ⎛ M rx ⎞ 8 ⎛ 394 kip-ft ⎞
+ = 0.200 + ⎜
Pc 9 ⎜⎝ M cx ⎟⎠ 9 ⎝ 573 kiip-ft ⎟⎠
Yi = 200 kips + 200 kips
= 400 kips
= 0.811
Ni = 0.002Yi
The W14×90 is adequate since 0.811 ≤ 1.0. = 0.002(400 kips)
= 0.8 kips
Design by Direct Analysis (Appendix 7) Because this notional load is less than the actual lateral load,
The Direct Analysis Method is permitted for any ratio of it need not be applied. For a load combination that does not
second-order drift, ∆2nd, to first-order drift, ∆1st, and required include a lateral load, the notional load would need to be
when this ratio exceeds 1.5. It requires the use of: included in the analysis.

1. A direct second-order analysis or a first-order analysis For Column A, using first-order analysis and B1-B2 amplifi-
with B1-B2 amplification. cation:
Pnt = 200 kips, Plt = 0 kips
2. The nominal frame geometry with an additional lateral Mnt = 0 kip-ft, Mlt = 300 kip-ft
load of Ni = 0.002Yi, where Yi is the total gravity load
on level i from LRFD load combinations, or 1.6 times To determine the second-order amplification, the reduced
ASD load combinations. stiffness, EI*, must be calculated.
αPr = 1.0(200 kips)
3. The reduced stiffnesses EA* and EI* (including in = 200 kips
B1-B2 amplification, if used).
and
4. LRFD load combinations, or ASD load combinations 0.5Py = 0.5Fy Ag
multiplied by 1.6. This multiplier ensures that the drift = 0.5(50 ksi)(26.5 in.2)
level is consistent for LRFD and ASD when determin- = 663 kips
ing second-order effects. The forces and moments ob-
tained in this analysis are then divided by 1.6 for ASD Thus, because αPr < 0.5Py, τb = 1.0 and
member design. EI* = 0.8τbEI
= 0.8EI
The following exceptions apply as alternatives in item 2:
For P-δ amplification, since there are no moments associated
a. If the out-of-plumb geometry of the structures is used, with the no-translation case, there is no need to calculate
the notional loads can be omitted. B1. For P-∆ amplification, the reduced stiffness EI* must be
used to determine the first-order drift. Because EI* = 0.8EI,
b. When the ratio of second-order drift to first-order drift the first-order drift based upon EI* is 25% larger than that
is equal to or less than 1.5, the notional load can be calculated previously. Thus,
applied as a minimum lateral load, not an additional
lateral load. Note that the unreduced stiffnesses, EA ∆1st = 1.25(1.34 in.)
and EI, are used in this comparison. = 1.68 in.
The first-order drift ratio is determined from the amplified
c. When the actual out-of-plumbness is known, it is per- drift of 1.68 in.
mitted to adjust the notional loads proportionally.
∆1st /L = (1.68 in.)/(15 ft × 12 in./ft)
For all frames designed with this method, K = 1.0. = 0.00933
It was previously determined in the illustration of design by For moment frames, RM = 0.85 and from Equation C2-6b
second-order analysis example that the second-order drift is with ∆H = ∆1st and ΣH = 20 kips,
less than 1.5 times the first-order drift (note that this check is ΣH
ΣPe2 = RM
( )
properly made using the unreduced stiffnesses, EA and EI).
∆ 1st / L
20 kips
= 0.85
(0.00933)
= 1,820 kips

ENGINEERING JOURNAL / THIRD QUARTER / 2008 / 163

159-170_Carter_Geschwinder_2008_3Q.indd 163 8/19/08 1:08:39 PM

 For design by LRFD, α = 1.0 and ΣPnt is the sum of the Thus, because Pr /Pc ≥ 0.2, Equation H1-1a is applicable.
gravity loads. Thus,
Pr 8 ⎛ M rx ⎞ 8 ⎛ 384 kip-ft ⎞
+ = 0.200 + ⎜
αΣPnt /ΣPe2 = 1.0(200 kips + 200 kips)/1,820 kips Pc 9 ⎜⎝ M cx ⎟⎠ 9 ⎝ 573 kiip-ft ⎟⎠
= 0.220
= 0.796
From Equation C2-3, the amplification is:
The W14×90 is adequate since 0.796 ≤ 1.0.
1
B2 = ≥1
α Σ Pnt
1− The Simplified Method
Σ Pe 2
This method is provided in the AISC Basic Design Values
1
= ≥ 1.0 Cards and the 13th Edition Steel Construction Manual
(1 − 0 .220 ) (AISC, 2005b), and excerpted as shown in Figure 3. This
= 1.28 ≥ 1.0 simplified method is derived from the effective length
= 1.28 method (Design by Second-Order Analysis; Section C2.2a)
It is worth noting that use of the reduced axial stiffness, using B1-B2 amplification with B1 taken equal to B2. Note
EA* = 0.8EA, in members that contribute to lateral stability that the user note in Section C2.1b says that B1 may be taken
is also required in this method. However, due to the char- equal to B2 as long as B1 is less than 1.05. However, it is also
acteristics of the structures chosen for this example, there conservative to take B1 equal to B2 any time B1 is less than
are no axial deformations that impact the stability of the B2. Although it cannot universally be stated that B1 is always
structure. equal to or less than B2, this is the case for typical framing. It
is left to engineering judgment to confirm that this criterion
The amplified axial force (Equation C2-1b) and associated is satisfied when applying the simplified method.
design parameters for this method are: This method is permitted when the ratio of second-order
Pr = Pnt + B2Plt drift, ∆2nd, to first-order drift, ∆1st, is equal to or less than
= 200 kips + 1.28(0 kips) 1.5 as with the Design by Second-Order Analysis method.
= 200 kips It allows the use of a first-order analysis based on nomi-
nal stiffnesses, EA and EI, with a minimum lateral load
Kx = Ky = 1.0
Ni = 0.002Yi, where Yi is the total gravity load on level i from
Lx = Ly = 15 ft LRFD load combinations or ASD load combinations. The
1.6 multiplier on ASD load combinations is not used at this
The amplified moment (Equation C2-1a) and associated de-
point but its effect is included in the determination of the
sign parameters for this method are:
amplification multiplier upon entering the table.
Mrx = B1Mnt + B2Mlt The ratio of total story gravity load (times 1.0 in LRFD,
= (0 kip-ft) + 1.28(300 kip-ft) 1.6 in ASD) to the story lateral load is used to enter the ta-
= 384 kip-ft ble in Figure 3. The second-order amplification multiplier
Cb = 1.67 is determined from the value in the table corresponding to
the calculated load ratio and design story drift limit. While
Lb = 15 ft linear interpolation between tabular values is permitted, it
Based upon these design parameters, the axial and strong- is important to note that the tabular values have, in essence,
axis available flexural strengths of the ASTM A992 only two significant digits. Accordingly, the value deter-
W14×90 are: mined should not be calculated to more than one decimal
place. The tabular value is used to amplify all forces and
Pc = φcPn = 1,000 kips moments in the analysis.
Mcx = φbMnx = 573 kip-ft When the ratio of second-order drift to first-order drift is
To determine which interaction equation is applicable, the equal to or less than 1.1, K = 1.0 can be used in the de-
ratio of the required axial compressive strength to available sign of moment frames. Otherwise, for moment frames, K is
axial compressive strength must be determined. determined from a sidesway buckling analysis. For braced
frames, K = 1.0.
Pr 200 kips
=
Pc 1, 000 kips
= 0.200

164 / ENGINEERING JOURNAL / THIRD QUARTER / 2008

159-170_Carter_Geschwinder_2008_3Q.indd 164 8/19/08 1:16:59 PM

 For the example frame given in Figure 1, the minimum The amplified axial force (with the full axial force am-
lateral load is: plified by B2) and associated design parameters for this
Yi = 200 kips + 200 kips method are:
= 400 kips Pr = 1.3Pu
= 1.3 (200 kips)
Ni = 0.002Yi
= 0.002(400 kips) = 260 kips
= 0.8 kips Kx* = 2.83, Ky = 1.0
Because this notional load is less than the actual lateral load, Lx = Ly = 15 ft
it need not be applied. For a load combination that does not
The amplified moment (with the full moment amplified by
include a lateral load, the notional load would need to be
B2) and associated design parameters for this method are:
included in the analysis.
Mrx = 1.3Mu
The actual first-order drift of the trial frame corresponds to a drift = 1.3(300 kip-ft)
ratio of L/134 and the load ratio is: = 390 kip-ft
1.0 × (200 kips + 200 kips)/(20 kips) = 20 Cb = 1.67
Entering the table in the column for a load ratio of 20, the Lb = 15 ft
corresponding multiplier for a drift ratio of H/134 is 1.3
(determined by interpolation to one decimal place). This Based on these design parameters, the available axial com-
multiplier is less than 1.5; thus, ∆2nd < 1.5∆1st and the use of pressive strength and strong-axis available flexural strength
this method is permitted. However, because the multiplier is of the ASTM A992 W14×90 are:
greater than 1.1, K cannot be taken as 1.0 for column design Pc = φc Pn = 721 kips
in the moment frame with this method. Thus, K must be cal- Mcx = φb Mnx = 573 kip-ft
culated, including the leaning column effect. Using the same
approach as previously discussed (Lim and McNamara, To determine which interaction equation is applicable, the
1972): ratio of the required axial compressive strength to available
axial compressive strength must be determined.
ΣPleaning /ΣPstability = (200 kips)/(200 kips)
=1 Pr 260 kips
=
Pc 721 kips
Kx* = Kx (1 + ΣPleaning /ΣPstability)½
= 2.0(1 + 1)½ = 0.361
= 2.83

Fig. 3. Simplified method from AISC basic design values cards.

ENGINEERING JOURNAL / THIRD QUARTER / 2008 / 165

159-170_Carter_Geschwinder_2008_3Q.indd 165 8/18/08 11:23:39 AM

 Thus, because Pr /Pc ≥ 0.2, Equation H1-1a is applicable. The lateral stiffness of the frame depends on Columns D and
E only, and based on a classical stiffness derivation with the
Pr 8 ⎛ M rx ⎞ 8 ⎛ 390 kip-ft ⎞
+ = 0.361 + ⎜ given end conditions, it is calculated as follows:
Pc 9 ⎜⎝ M cx ⎟⎠ 9 ⎝ 573 kiip-ft ⎟⎠
k = 2 × 3EI/L3
= 0.966 = 2 × 3(29,000 ksi)(I)/(15 ft × 12 in./ft)3
= 0.0298(I)
The W14×90 is adequate since 0.966 ≤ 1.0.
With the service level lateral load on the frame of 10 kips:
Summary for the One-Bay Frame 0.0298(I) ≥ (10 kips)/(0.300 in.)
All methods illustrated in the foregoing sections produce Thus, Ireq = 1,120 in.4 and an ASTM A992 W14×109 is
similar designs. The results are tabulated here for compari- selected as the trial shape with Ix = 1,240 in.4
son, where the result of the beam-column interaction equa-
tion is given for each method. A lower interaction equation The actual lateral stiffness of the frame is:
result for the same column shape signifies a prediction of k = 2 × 3EI/L3
higher strength. = 2 × 3(29,000 ksi)(1,240 in.4)/(15 ft × 12 in./ft)3
= 37.0 kips/in.
Method Interaction Equation
Second-Order 0.840 The corresponding first-order drift of the frame under the
LRFD lateral load of 15 kips is:
First-Order 0.811
∆1st = (15 kips)/(37.0 kips/in.)
Direct Analysis 0.796
= 0.405 in.
Simplified 0.966
The first-order axial force, strong-axis moment, and design
In this example, the direct analysis method predicts the high- parameters for Columns D and E are:
est strength, while the simplified method predicts the lowest Pu = 150 kips Mux = (15 kips)(15 ft)/2
strength. This would be expected because the Direct Analysis Kx = 2.0 = 113 kip-ft
Method was developed as the most accurate approach while Ky = 1.0 Cb = 1.67
the simplified method was developed to produce a quick yet Lx = Ly = 15 ft Lb = 15 ft
conservative solution.
The designs compared here are based on strength with no Note that Kx = 2.0, the theoretical value for a column with
consideration of drift limitation, except to the extent that the a fixed base and pinned top, is used rather than the value of
actual drift impacts the magnitude of the second-order effects. 2.1 recommended for design in the AISC Specification Com-
The usual drift limits of approximately L/400 will necessitate mentary Table C-C2.2. The value of 2.0 is used because it is
framing members and configurations with more lateral stiff- consistent with the formulation of the lateral stiffness calcu-
ness than this frame provides. Hence, the designer may find lation that follows. Note also that the impact of the leaning
that a frame configured for drift first will often require no in- column on Kx is ignored in selecting the trial size, although
crease in member size for strength, including second-order ef- it will be considered in subsequent sections when Kx cannot
fects. This will be explored further with the three-bay frame. be taken equal to 1.0 for Column A. Out of the plane of the
frame, Ky is taken as 1.0.
THE THREE-BAY FRAME
For the frame shown in Figure 2, a trial shape is selected Design by Second-Order Analysis (Section C2.2a)
using a first-order drift limit of L/600 under a service level For the example frame given in Figure 2, the minimum
lateral load of 10 kips. Thereafter, that trial shape is used as lateral load is:
the basis for comparison of the four methods used previously
Yi = 75 kips + 150 kips + 150 kips + 75 kips
for the one-bay frame.
= 450 kips
Selection of Trial Shape Based on the Drift Limit Only Ni = 0.002 Yi
= 0.002(450 kips)
For the dimensions shown in Figure 2:
= 0.90 kips
L/600 = (15 ft × 12 in./ft)/600
= 0.300 in. Because this notional load is less than the actual lateral load,
it need not be applied.

166 / ENGINEERING JOURNAL / THIRD QUARTER / 2008

159-170_Carter_Geschwinder_2008_3Q.indd 166 8/19/08 1:49:55 PM

 For Columns D and E, using first-order analysis and B1-B2 The amplified moment (Equation C2-1a) and associated de-
amplification: sign parameters for this method are:
Pnt = 150 kips, Plt = 0 kips Mrx = B1Mnt + B2Mlt
Mnt = 0 kip-ft, Mlt = 113 kip-ft = (0 kip-ft) + 1.09 (113 kip-ft)
= 123 kip-ft
For P-δ amplification, because there are no moments associ-
ated with the no-translation case, there is no need to calcu- Cb = 1.67
late B1. For P-∆ amplification, the first-order drift ratio is
Lb = 15 ft
determined from the calculated drift of 0.405 in. Thus,
∆1st /L = (0.405 in.)/(15 ft × 12 in./ft) Based on these design parameters, the available axial com-
= 0.00225 pressive strength and strong-axis available flexural strength
of the ASTM A992 W14×109 are:
For moment frames, Rm = 0.85 and from Equation C2-6b
Pc = φcPn = 1,220 kips
with ∆H = ∆1st and ΣH = 15 kips,
Mcx = φbMnx = 720 kip-ft
ΣH
ΣPe2 = RM
(
∆ 1st / L ) To determine which interaction equation is applicable, the
ratio of the required axial compressive strength to available
15 kips axial compressive strength must be determined.
= 0.85
(0.00225)
Pr 150 kips
= 5,670 kips =
Pc 1, 220 kips
For design by LRFD, α = 1.0 and ΣPnt is the sum of the
gravity loads. Thus, = 0.123

αΣPnt /ΣPe2 = 1.0(75 kips + 150 kips + 150 kips + Thus, because Pr /Pc < 0.2, Equation H1-1b is applicable.
75 kips)/5,670 kips Pr M rx 0.123 123 kip-ft
= 0.0794 + = +
2 Pc M cx 2 720 kip-ft
From Equation C2-3, the amplification is: = 0.232
1
B2 = ≥1 The W14×109 is adequate because 0.232 ≤ 1.0.
⎛ α Σ Pnt ⎞
⎜ 1 − ΣP ⎟
⎝ e2 ⎠ Design by First-Order Analysis (Section C2.2b)
1 For the example frame given in Figure 2, the additional lat-
= ≥ 1.0
(1 − 0.0794) eral load (with ∆ = ∆1st) is:
= 1.09 ≥ 1.0 ∆1st /L = (0.405 in.)/(15 ft × 12 in./ft)
= 1.09 = 0.00225
Because B2 = 1.09, the second-order drift is less than 1.5 times Yi = 75 kips + 150 kips + 150 kips + 75 kips
the first-order drift. Thus, the use of this method is permitted. = 450 kips
Because B2 < 1.1, K can be taken as 1.0 for column design in Ni = 2.1(∆1st /L)Yi ≥ 0.0042Yi
the moment frame with this method. = 2.1(0.00225)(450 kips) ≥ 0.0042(450 kips)
The amplified axial force (Equation C2-1b) and associated = 2.13 kips ≥ 1.89 kips
design parameters for this method are: = 2.13 kips
Pr = Pnt + B2Plt It was previously determined in the illustration of design by
= 150 kips + 1.09(0 kips) second-order analysis example that the second-order drift is
= 150 kips less than 1.5 times the first-order drift. Additionally,
Kx = Ky = 1.0 αPr = 1.0(150 kips) = 150 kips
Lx = Ly = 15 ft and for the ASTM A992 W14×109,
0.5Py = 0.5Fy Ag
= 0.5(50 ksi)(32.0 in.2)
= 800 kips

ENGINEERING JOURNAL / THIRD QUARTER / 2008 / 167

159-170_Carter_Geschwinder_2008_3Q.indd 167 8/19/08 10:01:53 AM

 Because ∆2nd < 1.5∆1st and αPr < 0.5Py, the use of this method To determine which interaction equation is applicable, the
is permitted. ratio of the required axial compressive strength to available
axial compressive strength must be determined.
The loading for this method is the same as shown in Figure
2, except for the addition of a notional load of 2.13 kips co- Pr 150 kips
=
incident with the lateral load of 15 kips shown, resulting in a Pc 1, 220 kips
moment Mu of 128 kip-ft in each column. = 0.123
This moment must be amplified by B1 as determined from
Thus, because Pr /Pc < 0.2, Equation H1-1b is applicable.
Equation C2-2. The Euler buckling load is calculated with
K1 = 1.0. Thus, Pr M rx 0.123 128 kip-ft
+ = +
2
Pe1 = π EI / (K1L) 2 2 Pc M cx 2 720 kip-ft
= π2(29,000 ksi)(1,240 in.4)/(1.0 × 15 ft × 12 in./ft) 2 = 0.239
= 11,000 kips
The W14×109 is adequate because 0.239 ≤ 1.0.
Because the moment on one end of the column is zero, the
moment gradient term is: Direct Analysis Method (Appendix 7)
Cm = 0.6 – 0.4(M1/M2)
It was previously determined in the illustration of design by
= 0.6 – 0.4(0/128)
second-order analysis example that the second-order drift is
= 0.6
less than 1.5 times the first-order drift (note that this check
From Equation C2-2, is properly made using the unreduced stiffness EI). Thus, the
αPr /Pe1 = 1.0(150 kips)/(11,000 kips) notional load can be applied as minimum lateral load, and
= 0.0136 that minimum is:
Cm Yi = 75 kips + 150 kips + 150 kips + 75 kips
B1 = ≥1 = 450 kips
αPr
1−
Pe1 Ni = 0.002Yi
0.6 = 0.002(450 kips)
= ≥ 1.0 = 0.9 kip
1 − 0.0136
= 0.608 ≥ 1.0 Because this notional load is less than the actual lateral load,
= 1.0 it need not be applied.
The axial force and associated design parameters for this For Columns D and E, using first-order analysis and B1-B2
method are: amplification:
Pr = 150 kips Pnt = 150 kips, Plt = 0 kips
Kx = Ky = 1.0 Mnt = 0 kip-ft, Mlt = 113 kip-ft
Lx = Ly = 15 ft To determine the second-order amplification, the reduced
The amplified moment and associated design parameters for stiffness, EI*, must be calculated.
this method are: αPr = 1.0(150 kips)
Mrx = B1Mu = 150 kips
= 1.0(128 kip-ft) and for the ASTM A992 W14×109,
= 128 kip-ft
Cb = 1.67 0.5Py = 0.5Fy Ag
Lb = 15 ft = 0.5(50 ksi)(32.0 in.2)
= 800 kips
Based on these design parameters, the available axial com-
pressive strength and strong-axis available flexural strengths Thus, because αPr < 0.5Py, τb = 1.0 and
of the ASTM A992 W14×109 are: EI* = 0.8τbEI
Pc = φcPn = 1,220 kips = 0.8EI
Mcx = φbMnx = 720 kip-ft For P-δ amplification, because there are no moments as-
sociated with the no-translation case, there is no need to
calculate B1. For P-∆ amplification, the reduced stiffness
EI* must be used to determine the first-order drift. Because

168 / ENGINEERING JOURNAL / THIRD QUARTER / 2008

159-170_Carter_Geschwinder_2008_3Q.indd 168 8/19/08 9:28:17 AM

 EI* = 0.8EI, the first-order drift based on EI* is 25% larger Cb = 1.67
than that calculated previously. Thus, Lb = 15 ft
∆1st = 1.25(0.405 in.) Based on these design parameters, the available axial com-
= 0.506 in. pressive strength and strong-axis available flexural strengths
The first-order drift ratio is determined from the amplified of the ASTM A992 W14×109 are:
drift of 0.506 in. Pc = φcPn = 1,220 kips
∆1st /L = (0.506 in.)/(15 ft × 12 in./ft) Mcx = φb Mnx = 720 kip-ft
= 0.00281 To determine which interaction equation is applicable, the
For moment frames, RM = 0.85 and from Equation C2-6b ratio of the required axial compressive strength to available
with ∆H = ∆1st and ΣH = 15 kips, axial compressive strength must be determined.
ΣH Pr 150 kips
ΣPe2 = RM =
( ∆ 1st /L ) Pc 1, 220 kips
15 kips = 0.123
= 0.85
(0.00281) Thus, because Pr /Pc < 0.2, Equation H1-1b is applicable.
= 4,540 kips
Pr M rx 0.123 125 kip-ft
For design by LRFD, α = 1.0 and ΣPnt is the sum of the + = +
2 Pc M cx 2 720 kip-ft
gravity loads. Thus,
= 0.235
αΣPnt /ΣPe2 = 1.0(75 kips + 150 kips + 150 kips +
75 kips)/4,540 kips The W14×109 is adequate because 0.235 ≤ 1.0.
= 0.0991
The Simplified Method
From Equation C2-3, the amplification is:
For the example frame given in Figure 2, the minimum lat-
1
B2 = ≥1 eral load is:
⎛ α Σ Pnt ⎞
⎜ 1 − Yi = 75 kips + 150 kips + 150 kips + 75 kips
⎝ Σ Pe 2 ⎟⎠ = 450 kips
1 Ni = 0.002Yi
= ≥ 1.0
(1 − 0.0991) = 0.002(450 kips)
= 1.11 ≥ 1.0 = 0.9 kips
= 1.11 Because this notional load is less than the actual lateral load,
It is worth noting that use of the reduced axial stiffness, EA* it need not be applied.
= 0.8EA, in members that contribute to lateral stability is also The 15-kip lateral load produces slightly less drift than that
required in this method. However, due to the characteristics corresponding to the design story drift limit because the
of the structures chosen for this example, there are no axial W14×109 has I = 1,240 in.4 (versus the 1,120 in.4 required to
deformations that impact the stability of the structure. limit drift to L/400). The lateral load required to produce the
The amplified axial force (Equation C2-1b) and associated design story drift limit is:
design parameters for this method are: 15 kips × (1,240 in.4)/(1,120 in.4) = 16.6 kips
Pr = Pnt + B2Plt The load ratio is then:
= 150 kips + 1.11(0 kips)
= 150 kips 1.0 × (75 kips + 150 kips + 150 kips + 75 kips)/
(16.6 kips) = 27.1
Kx = Ky = 1.0
Entering the table in the row for H/400, the corresponding
Lx = Ly = 15 ft multiplier for a load ratio of 27.1 is 1.1 (determined by in-
The amplified moment (Equation C2-1a) and associated de- terpolation to one decimal place). Because this multiplier is
sign parameters for this method are: less than 1.5, ∆2nd < 1.5∆1st and the use of this method is
permitted. Additionally, because the multiplier is equal to
Mrx = B1Mnt + B2Mlt
1.1, K can be taken as 1.0 for column design in the moment
= (0 kip-ft) + 1.11(113 kip-ft)
frame with this method.
= 125 kip-ft

ENGINEERING JOURNAL / THIRD QUARTER / 2008 / 169

159-170_Carter_Geschwinder_2008_3Q.indd 169 8/19/08 9:30:17 AM

 The amplified axial force (with the full axial force ampli- CONCLUSIONS
fied by B2) and associated design parameters for this method
The following conclusions can be drawn from the foregoing
are:
examples:
Pr = 1.1Pu
= 1.1(150 kips) 1. If conservative assumptions are acceptable, the easiest
= 165 kips method to apply is the Simplified Method, particularly
when the drift limit is such that K can be taken equal
Kx = Ky = 1.0 to 1.
Lx = Ly = 15 ft
2. None of the analysis methods in the AISC Specification
The amplified moment (with the full moment amplified by B2) are particularly difficult to use. The First-Order Analy-
and associated design parameters for this method are: sis Method and Direct Analysis Method both eliminate
Mrx = 1.1Mu the need to calculate K, which can be a tedious process
= 1.1(113 kip-ft) based upon assumptions that are rarely satisfied in real
= 124 kip-ft structures. Nonetheless, those who prefer to continue
to use the approach of past specifications, the Effective
Cb = 1.67 Length Method, can do so, provided they incorporate the
Lb = 15 ft new requirement of a minimum lateral load in all load
combinations.
Based on these design parameters, the axial and strong-axis
flexural available strengths of the ASTM A992 W14×109 3. Second-order effects and leaning columns have a signifi-
are: cant impact on strength requirements, but usual drift lim-
Pc = φc Pn = 1,220 kips its such as L/400 sometimes can result in framing that re-
Mcx = φb Mnx = 720 kip-ft quires no increase in member size for strength. For frames
with little or no lateral load and/or heavy floor loading,
To determine which interaction equation is applicable, the
it is more likely that stability will control, regardless of
ratio of the required axial compressive strength to available
the drift limits. This should not be taken as a blanket in-
axial compressive strength must be determined.
dication that the use of a drift limit eliminates the need
Pr 165 kips to consider stability effects. Rather, it simply means that
=
Pc 1, 220 kips drift-controlled designs may be less sensitive to second-
= 0.135 order effects because the framing is naturally stiffer and
provides reserve strength. Drift limits also result in sig-
nificant simplification of the analysis requirements when
Thus, because Pr /Pc < 0.2, Equation H1-1b is applicable.
the increased framing stiffness allows more frequent use
Pr M rx 0.135 124 kip-ft of the simplifications allowed in the various methods,
+ = +
2 Pc M cx 2 720 kip-ft such as the use of K = 1.
= 0.240
REFERENCES
The W14×109 is adequate because 0.240 ≤ 1.0.
AISC (2005a), Specification for Structural Steel Buildings,
Summary for the Three-Bay Frame ANSI/AISC 360-05, American Institute of Steel Con-
struction, Chicago, IL.
As before, all methods produce similar designs. The result of
the beam-column interaction equation for each method is: AISC (2005b), Steel Construction Manual, 13th ed., Ameri-
can Institute of Steel Construction, Chicago, IL.
Method Interaction Equation Geschwindner, L.F., Disque, R.O. and Bjorhovde, R. (1994),
Second-Order 0.232 Load and Resistance Factor Design of Steel Structures,
First-Order 0.239 Prentice Hall, Englewood Cliffs, NJ.
Direct Analysis 0.235 Lim, L.C. and McNamara, R.J. (1972), “Stability of Novel
Simplified 0.240 Building System,” Structural Design of Tall Buildings,
In this example, the interaction equations predict values that Vol. II-16, Proceedings, ASCE-IABSE International Con-
are so close to each other that there is no practical difference ference on the Planning and Design of Tall Buildings,
in the results. Bethlehem, PA, pp. 499–524.

170 / ENGINEERING JOURNAL / THIRD QUARTER / 2008

159-170_Carter_Geschwinder_2008_3Q.indd 170 8/19/08 9:31:00 AM