This action might not be possible to undo. Are you sure you want to continue?
Lessons Learned regarding Appropriate use of Direct Analysis Method with RAM & ETABS
To quickly summarize, here is what we have learned:
๏
The Effective Length Method for the design of steel beams and columns has been used
for years, but this method has limitations on its applicability. Thus, in the past several
years the industry has migrated to another design methodology to check for strength
and stability – The Direct Analysis Method (DA). Attached herein are two papers that
are mustreads:
• Background of DA obtained from AISC, and
• Recent paper published in Winter 2011 Engineering Journal by Shankar Nair, Jim
Malley and John Hooper on Design of Steel Buildings for EQ and Stability by AISC
360 (DA) and ASCE 7.
Note: SidePlate President Henry Gallart personally spoke with Dr. Nair and got
verbal confirmation of their conclusions that the unreduced stiffness shall be used
in the calculation of building periods, associated base shears, and checking of
building drifts.
A 1hour free webinar that Dr. Nair gave through AISC is very educational and
helpful for those who are auditory and visual learners – highly recommend it:
http://www.aisc.org/content.aspx?id=4498
๏
Both the RAM Structural System and ETABS software have been programmed in such a
way that, unless the user really understands what the program is doing and is super
careful with their bookkeeping, an engineer selecting DA may inadvertently analyze a
steel building with a global 20% reduced stiffness, resulting in a softer structure
(higher periods) than would otherwise be calculated. For many buildings, this
translates into a lower base shear, thus resulting in unconservative designs.
In an effort to assist structural engineers, we have assembled the following tips for RAM
Structural System and ETABS users:
๏
For RAM users, Allen Adams of RAM posted on the Bentley wiki an excellent write up
that describes the exact steps to follow.
• http://communities.bentley.com/products/structural/
structural_analysis___design/w/structural_analysis_and_design__wiki/6011.aspx
๏
For ETABS users, this issue was brought to their attention just recently. As such, a
detailed write up doesn’t exist today. However, using the same principles, here is one
way to utilize ETABS correctly if an engineer is using DA to design the beams and
columns:
• When starting a new model in ETABS, the program utilizes an unreduced stiffness
for beams and columns. If you never hit DESIGN, all is well. However, if Direct
Analysis was selected under PREFERENCES as the Design Analysis Method, when a
DESIGN is performed it will prompt you to overwrite the analysis. This will trigger
the 20% stiffness reduction and the original building period calculated based on an
unreduced stiffness is lost.
‣ Recommendation:
 Create ETABS model with the usual preferences like DA. Alternately,
Effective Length Method could be selected to preclude the accidental
running of the model with the DA setting which will reduce the building
stiffness by 20%.
 Analyze the building to determine actual building periods (using the
unreduced stiffness model) in order to obtain the correct static base shear
to be used for calculating the seismic base shear for both strength and
drift. Perform all key ASCE 7 code checks, except strength checks required
by AISC 360 and AISC 341.
• If performing a dynamic analysis, be sure to create a load case for both
dynamic base shears (strength and drift) using the unreduced stiffness
model so that there is a basis with which to scale later the dynamic
base shear used for strength checks.
 Upon generating a design that is satisfactory and meets all ASCE 7 code
checks, save the model (without ever activating the DESIGN feature). Now,
to check strength (stress) per AISC 360 and AISC 341 seismic provisions,
follow the principles outlined in the Wiki for RAM. It is anticipated that a
similar document from CSI will be created in the near future.
• Since some of the settings in ETABS may need to be different (like
overriding the period calculated by ETABS, etc.), we would recommend
that a separate model be used to check strength to eliminate the
possible confusion that can easily occur when switching back and forth
between a model with an unreduced stiffness and a model with a
reduced stiffness.
If you have ANY questions or would like some clarification on the content of this information,
please feel free to call our office at 9493057889.
ENGINEERING JOURNAL / THIRD QUARTER / 2011 / 199
Design of Steel Buildings for Earthquake and Stability
by Application of ASCE 7 and AISC 360
R. SHANKAR NAIR, JAMES O. MALLEY and JOHN D. HOOPER
Abstract
Design of steel buildings in the United States typically combines application of ASCE/SEI 7, Minimum Design Loads for Buildings and Other
Structures, and ANSI/AISC 360, Specification for Structural Steel Buildings. For buildings designed for seismic effects, ANSI/AISC 341, Seis
mic Provisions for Structural Steel Buildings, may also be applicable. The ASCE 7 Minimum Design Loads standard includes specific design
provisions related to stability under seismic loading which overlap and, in some instances, appear to conflict with the stability design require
ments of the AISC Specification. This paper explores the areas of overlap and apparent conflict between ASCE 7 and AISC 360 and offers
practical recommendations for seismic design incorporating the provisions of both.
Keywords: design loads, seismic design, structural stability.
D
esign of steel buildings in the United States typically
combines application of ASCE/SEI 7, Minimum Design
Loads for Buildings and Other Structures (ASCE, 2005),
and ANSI/AISC 360, Specification for Structural Steel
Buildings (AISC, 2005a; AISC, 2010). For buildings de
signed for seismic effects, ANSI/AISC 341, Seismic Provi
sions for Structural Steel Buildings (AISC, 2005b), may be
applicable in conjunction with the Specification.
ASCE 7 is used, either directly or by reference from a
building code, to define the loads for which the structure
must be designed; AISC 360 and 341 are used to design the
steel structure for those loads.
The ASCE 7 Minimum Design Loads standard, though
generally focused on loads and not on design or the response
of the structure to those loads, includes specific design pro
visions related to stability under seismic loading. These pro
visions overlap and, in some instances, appear to conflict
with the stability design requirements of the AISC Specifi
cation. (The AISC Seismic Provisions do not include stabil
ity design requirements.)
This paper explores the areas of overlap and apparent
conflict between ASCE 7 and AISC 360 and offers practi
cal recommendations for seismic design incorporating the
provisions of both.
The paper does not attempt to correlate design with ex
pected actual behavior beyond the degree of correlation im
plied by compliance with ASCE 7 and AISC 360. This is
an important limitation. It is generally recognized that ac
tual displacements in an earthquake could be much larger
than the elastic displacements due to codespecified design
loads, and the resulting secondorder effects could be quite
different from those predicted by specificationcompliant
analysis; exploration of this issue is beyond the scope of this
paper.
STABILITY DESIGN BY AISC 360
The Direct Analysis Method of design for stability was
introduced in the 2005 edition of the AISC Specification.
The 2010 edition makes that method the primary means of
design; alternative approaches have been moved to an ap
pendix. The discussion in this paper will be limited to the
Direct Analysis Method and its application in seismic design
in conjunction with ASCE 7; other stability design methods
permitted in the AISC Specification are not considered.
The rational basis of the Direct Analysis Method is ex
plained in AISCSSRC (2003); a simple introduction to
the practical application of the method is provided in Nair
(2009a). Notes on the modeling of structures for design by
the Direct Analysis Method are provided in Nair (2009b).
Design for stability by the Direct Analysis Method in
volves a secondorder analysis, use of reduced stiffness in
the analysis, consideration of initial imperfections (either
by direct modeling of the imperfections or by application of
notional loads in the analysis) under certain circumstances,
and strength check of components using an effective length
factor, K, of unity for members subject to compression.
R. Shankar Nair, Ph.D., P.E., S.E., Senior Vice President, Teng & Associates,
Inc., Chicago, IL (corresponding). Email: nairrs@teng.com
James O. Malley, P.E., S.E., Senior Principal, Degenkolb Engineers, San Fran
cisco, CA. Email: malley@degenkolb.com
John D. Hooper, P.E., S.E., Principal and Director of Earthquake Engineering,
Magnusson Klemencic Associates, Seattle, WA. Email: jhooper@mka.com
199_204_EJ3Q_2011_2010_07R.indd 199 9/13/11 2:38 PM
200 / ENGINEERING JOURNAL / THIRD QUARTER / 2011
SecondOrder Analysis
The analysis of the structure must be a secondorder analy
sis that, in the most general case, includes both PΔ effects,
which are the effects of loads acting on the displaced lo
cations of joints or nodes in the structure, and Pδ effects,
which are the effects of loads acting on the deformed shapes
of members. In tiered buildings, PΔ effects are the effects
of loads acting on the laterally displaced locations of floors
and roofs.
The 2010 AISC Specification exempts most buildings
from the need to consider Pδ effects in the analysis of the
overall structure; a PΔ–only analysis is sufficient in almost
all cases. This is an important simplification relative to the
2005 Specification, which does require inclusion of Pδ
effects in the analysis of most momentframe buildings. Re
gardless of whether Pδ effects need to be considered in the
analysis of the overall structure, the Pδ effects on individu
al beamcolumns must always be considered in the strength
check of those members.
The secondorder analysis required by the Direct Analy
sis Method of design may be performed using a computer
program formulated to provide a secondorder solution.
Alternatively, a secondorder solution may be obtained by
manipulating the results of a linear or firstorder analysis to
account for secondorder effects by application of the B
1
and
B
2
multipliers defined in the Specification.
In the B
1
and B
2
procedure, B
2
alters the results of a first
order analysis to account for PΔ effects; the B
2
multiplier
also accounts, in an approximate way (by application of the
factor R
M
in the calculation of B
2
), for the overall softening
of the structure’s response due to Pδ effects in individual
members. For a given vertical loading, there will be a sin
gle value of B
2
for each story and each direction of lateral
translation, applicable to the forces and moments caused by
lateral loading in all members and connections in that story.
In the unusual case where gravity load causes lateral transla
tion, B
2
is also applicable to the forces and moments caused
by the sidesway component of gravity load. While B
2
is a
story parameter, B
1
is a member parameter; a B
1
multiplier
is applied to the moments in each beamcolumn to account
for Pδ effects in that member.
Another approach, usable in the typical case where Pδ
effects do not need to be considered in the analysis of the
overall structure, is to obtain a PΔ–only secondorder solu
tion from a computer program and then to apply B
1
multipli
ers to account for Pδ effects in individual members.
Given that in secondorder analysis the effect of a load is
not proportional to its magnitude (and the principle of super
position of loads does not apply), the secondorder analysis
must be performed at the Load and Resistance Factor De
sign (LRFD) load level: For design by LRFD, LRFD load
combinations must be applied in the analysis; for design by
Allowable Strength Design (ASD), ASD load combinations
increased by a factor of 1.6 must be applied in the analysis
and the results must be divided by 1.6 to get the forces and
moments for proportioning of members and connections.
Reduced Stiffness
In the secondorder analysis, all stiffnesses in the model
ing of the structure must be reduced by applying a factor
of 0.8. An additional reduction factor, τ
b
, must be applied
to the flexural stiffnesses of all members whose flexural
stiffnesses are considered to contribute to the lateral stabil
ity of the building. Factor τ
b
is unity when the LRFDlevel
required compression strength of the member is less than
half the yield strength.
When the compression is high and τ
b
is not unity, the de
signer may still avoid the complication of calculating and
applying τ
b
by applying, instead, a small notional lateral
load (0.001 times the LRFDlevel gravity load applied at
each floor) in the analysis. This notional load will typically
be so much smaller than seismic loads that it may reasonably
be neglected when seismic loads are present. Thus, it should
be possible to take τ
b
as unity in all cases in seismic design;
what remains is the 0.8 factor, applied to all stiffnesses.
Initial Imperfections
The effect of initial imperfections must be considered in
the analysis, either by direct inclusion of the imperfections
in the analysis model or by application of notional loads, if
either (1) the load combination being considered is a gravity
only loading with no applied lateral load or (2) the ratio of
secondorder drift to firstorder drift in any story of the
building is more than 1.7.
For seismic design, there will always be applied lateral
load and condition 1 will not apply; condition 2 will also
typically not apply for buildings that satisfy the limits on
stability coefficient, θ, specified in ASCE 7. Therefore, it
will not typically be necessary in seismic design to consider
initial imperfections, either explicitly or by application of
notional loads.
Component Strength Check
For design by the Direct Analysis Method, once the appro
priate analysis has been performed, members and connec
tions are checked for strength with no further consideration
of overall structure stability. The effective length factor, K,
for members subject to compression is taken as unity (unless
a lower value is justified by rational analysis).
SEISMIC DESIGN STABILITY
PROVISIONS IN ASCE 7
Background and commentary on the seismic design provi
sions of the ASCE 7 Minimum Design Loads standard may
199_204_EJ3Q_2011_2010_07R.indd 200 9/13/11 2:38 PM
ENGINEERING JOURNAL / THIRD QUARTER / 2011 / 201
be found in FEMA (2009). Specific requirements for stabil
ity in conjunction with seismic design are presented in the
ASCE standard in a section titled “PDelta Effects” (Sec
tion 12.8.7)
*
; these requirements are included as part of the
“Equivalent Lateral Force Procedure” for seismic analysis
(Section 12.8).
The PDelta Effects section of ASCE 7 defines a stability
coefficient, indicates when Pdelta effects must be consid
ered, places limits on the stability coefficient and specifies
methods of accounting for Pdelta effects.
Stability Coefficient, θ
The stability coefficient, θ, is approximately the ratio of the
actual vertical force on a story of a building to the vertical
force that would cause elastic lateral buckling of the story.
An equation
†
is provided for the coefficient; the coefficient
is calculated with nominal, unreduced stiffnesses and for
load combinations with no individual load factor exceeding
unity.
When Must Pdelta Effects Be Considered?
Under ASCE 7, Pdelta effects need be considered only when
the stability coefficient, θ, is more than 0.10. This is roughly
equivalent to an AISC B
2
multiplier (ratio of secondorder
story drift to firstorder story drift) of 1.2, after accounting
for the fact that θ is calculated with nominal stiffnesses and
for load combinations with no individual load factor exceed
ing unity, while B
2
is calculated with reduced stiffnesses
and for LRFDlevel load combinations.
Limit on Stability Coefficient
The stability coefficient, θ, must never be higher than θ
max
,
which is variable (a function of β, the ratio of shear demand
to shear capacity of the story, and of C
d
, the deflection am
plification factor), but never higher than 0.25. A θ of 0.25
is roughly equivalent to an AISC B
2
multiplier of 1.7 after
correction for the different stiffnesses and loadings in the θ
and B
2
calculations.
Method of Analysis for Pdelta Effects
ASCE 7 specifies that when Pdelta effects are required to
be considered (i.e., when the stability coefficient, θ, exceeds
0.10), “the incremental factor related to Pdelta effects on
* ASCE 7 does not explicitly differentiate between the PΔ effects and Pδ
effects recognized by the AISC Specification; the “Pdelta” effects ad
dressed in ASCE 7 are the PΔ effects of AISC.
† There is an error in Equation 12.816, the equation for stability coef
ficient, θ, in ASCE 705. There should be I (for importance factor) in the
numerator on the righthand side of the equation. This has been corrected
in ASCE 710.
displacements and member forces shall be determined by
rational analysis.” Two types of rational analysis are envi
sioned: (1) nonlinear static (pushover) analysis and (2) non
linear response history analysis, both of which require
extensive, additional effort.
As an alternative to the rational analysis, “it is permit
ted to multiply the displacements and member forces by
1.0/(1 − θ).” This is analogous to application of the AISC B
2
multiplier with R
M
= 1 in the equation for B
2
, which amounts
to neglecting Pδ effects in the analysis. However, given that
θ is calculated with nominal stiffnesses and for load combi
nations with no individual load factor exceeding unity, while
B
2
is calculated with reduced stiffnesses and at LRFDlevel
load combinations, the AISC approach will indicate signifi
cantly higher secondorder displacements and forces.
Use of an analysis that included PΔ effects (but not Pδ
effects), and subsequent application of AISC B
1
multipli
ers to individual beam columns to account for Pδ effects,
would also satisfy the requirement of ASCE 7.
Seismic Design by Modal Response Spectrum Analysis
The preceding discussion of stabilityrelated requirements
in the seismic design provisions of ASCE 7 was based on use
of the Equivalent Lateral Force procedure. An alternative
seismic analysis procedure prescribed in ASCE 7 (Section
12.9) is the Modal Response Spectrum Analysis approach.
Design by Modal Response Spectrum Analysis is applicable
to all structures of all Seismic Design Categories; the Equiv
alent Lateral Force procedure is not permitted for certain
structures in Seismic Design Categories D through F.
Stability requirements are not presented independently in
the Modal Response Spectrum Analysis section; the same
Pdelta requirements prescribed for the Equivalent Lateral
Force procedure are incorporated by reference in the Modal
Analysis section. There is an obvious complication here in
that the Modal Analysis approach involves combining the
results of analyses for different modes, but secondorder
analyses (i.e., analyses incorporating Pdelta effects) cannot
normally be combined. A means of overcoming this diffi
culty is suggested in the following section.
COMPARISON AND RECOMMENDATIONS
Selected features of the Direct Analysis Method of design
for stability in the AISC Specification and provisions related
to seismic design and stability in the ASCE 7 Minimum De
sign Loads standard, as discussed in the preceding sections,
are summarized in Table 1. Clearly, there are areas of diver
gence between the two sets of requirements. Nonetheless, as
outlined in the following, steel buildings may be designed
for seismic effects and stability in general conformance with
both the AISC Specification and ASCE 7.
199_204_EJ3Q_2011_2010_07R.indd 201 9/13/11 2:38 PM
202 / ENGINEERING JOURNAL / THIRD QUARTER / 2011
General recommendations:
1. Observe the ASCE 7 limit on stability coefficient; θ
must not exceed θ
max
. The nominal (unreduced) stiff
ness of the structure and the ASCE 7 vertical load
(with no individual load factor greater than 1.0) may
be used in the calculation of coefficient θ.
2. In the analysis for assessing strengths, consider PΔ
effects in the analysis for all structures; also consider
Pδ effects in the analysis where required by the AISC
Specification. (Do not observe the ASCE 7 provision
that exempts from Pdelta considerations all buildings
with a stability coefficient, θ, less than 0.10.)
Recommendations specific to seismic design by the Equiva
lent Lateral Force procedure:
1. Determine the fundamental period of the building ei
ther by analysis or by use of the approximate meth
ods prescribed in ASCE 7. If determined by analysis,
use firstorder analysis (no secondorder effects), with
nominal (unreduced) stiffnesses.
2. Consider secondorder effects either by secondorder
Table 1. Comparison of Analysis and Stability Provisions in ASCE 7 and AISC 360
Subject
ASCE 7
Equivalent Lateral
Force Procedure
a
AISC 360
Direct Analysis Method
Recommendation for
Seismic Design
Limit on PΔ effect Stability coefficient θ must
not exceed θ
max
.
No limit. Observe ASCE limit, which
corresponds to PΔ multiplier of
1.33 or less.
b
Must PΔ effects be
considered?
Only when θ is greater
than 0.1.
Yes; in all cases. Always consider PΔ effects.
PΔ effects by rational
analysis?
Permitted. Permitted. Rational analysis may be used.
PΔ effects by
approximate analysis?
Permitted.
[Multiply lateral load
effects by 1/(1 – θ).]
Permitted.
[Multiply lateral load
effects by B
2
.]
The AISC method may be used;
note that 1/(1 − θ) ≈ B
2
.
Must Pδ effects be
considered in the
analysis?
Not specified. Generally yes in 2005;
generally no in 2010.
Observe AISC 2010.
c
Load in the stability
analysis
Not specified for rational
analysis. Load factor
not greater than 1.0 for θ
calculation.
LRFD load combinations
for LRFD; 1.6 times ASD
combinations for ASD.
Observe AISC: LRFD load
combinations for LRFD; 1.6 times
ASD load combinations for ASD.
Structure stiffness in
the stability analysis
Not specified. Apply factor of 0.8 τ
b
to EI; 0.8 to all other
stiffnesses.
Apply factor of 0.8 (no τ
b
) to all
stiffnesses.
Must initial
imperfections be
considered?
Not specified. Yes, with exceptions;
either model directly or
apply notional loads.
Need not consider initial
imperfections (neither direct
modeling nor notional loads).
Analysis to assess
conformance to drift
limits
Elastic stiffness; same type
of analysis as for strength.
Not specified. Use of same analysis as for
strength is permissible but may
be too conservative; see text.
Analysis to determine
period
Not specified, although
upper limits are defined.
Not specified. Use linear analysis (no PΔ
effects) with nominal, unreduced
stiffnesses.
a
See text for additional considerations for Modal Response Spectrum Analysis procedure.
b
The stability coefficient, θ, is calculated at lower load than used in AISC stability analyses; therefore, PΔ multipliers corresponding to the ASCE 7
thresholds are not strictly comparable to AISC 360 parameters.
c
Regardless of whether it is considered in the analysis of the overall structure, Pδ must always be considered in the strength check of individual
beamcolumns.
199_204_EJ3Q_2011_2010_07R.indd 202 9/13/11 2:38 PM
ENGINEERING JOURNAL / THIRD QUARTER / 2011 / 203
analysis or by application of the AISC B
1
and B
2
mul
tipliers to the results of firstorder analysis. The op
tions are:
a. Complete secondorder analysis that considers
both PΔ and Pδ effects.
b. PΔ–only secondorder analysis, followed by ap
plication of B
1
to individual beam columns (not
permissible in the unusual cases where inclusion of
Pδ effects in the overall analysis is required by the
AISC Specification).
c. Firstorder analysis, followed by application of B
2
and B
1
multipliers.
3. Perform the secondorder analysis and/or calculate
the B
1
and B
2
multipliers at LRFDlevel loads; that is,
use LRFD load combinations if design is by LRFD,
use 1.6 times ASD load combinations if design is by
ASD. (After secondorder analysis under ASD, divide
the analysis results by 1.6 for member and connection
strength checks.)
4. Apply a factor of 0.8 to all stiffnesses in the second
order analyses and in the calculation of B
1
and B
2
mul
tipliers. For load combinations that include seismic
load, the additional stiffness reduction factor τ
b
need
not be applied.
5. In the analysis for load combinations that include seis
mic load, it is not necessary to model initial imper
fections or to apply notional loads to account for the
imperfections.
6. Perform strength checks in accordance with the AISC
Specification, using an effective length factor, K, of
unity for members subject to compression (unless a
lower value is justified by rational analysis).
7. Conformance to ASCE 7 seismic drift limits may be
checked using the same analysis used for strength
checks (analysis at reduced stiffness, secondorder
analysis, secondorder effects determined at LRFD
level loads). This may be excessively conservative,
however, and it is permissible to base drift checks on
an analysis using the full unreduced stiffness of the
structure, with secondorder effects determined at the
lower loads specified by ASCE 7 for calculation of the
stability coefficient, θ. The deflection amplification
factor, C
d
, should be applied in either case.
Recommendations specific to seismic design by the Modal
Response Spectrum Analysis approach:
1. Determine modes and frequencies using firstorder
analysis, with nominal (unreduced) stiffnesses.
2. Determine member forces and moments due to gravity
load by firstorder analysis, with a factor of 0.8 applied
to all stiffnesses.
3. Use the properties of each mode and the ASCE 7 de
sign response spectrum to determine a set of lateral
forces for that mode; using these lateral forces, per
form a firstorder analysis, with a factor of 0.8 applied
to all stiffnesses, to determine member forces and mo
ments. Repeat for all modes considered.
4. Calculate a single B
2
multiplier, applicable to all
modes,
‡
for each story and each direction of lateral
translation, based on reduced stiffness (0.8 factor) and
the full LRFDlevel vertical load on the story.
§
5. Combine firstorder modal results (item 3) as speci
fied in ASCE 7 Section 12.9.3. Apply B
2
multipliers
to the combined results (member forces and moments
caused by lateral loading). Then scale the combined
results as specified in ASCE 7 Section 12.9.4. Alge
braic signs will typically be lost in modal combina
tions; the member forces and moments due to seismic
effects should, therefore, be considered reversible
(i.e., use absolute values of forces and moments in the
modal combinations and then consider the resulting
overall forces and moments due to seismic effects to
be fully reversible).
6. Combine the member forces and moment due to seis
mic effects (item 5) with the member forces and mo
ments due to gravity load (item 2), with load factors as
specified in ASCE 7.
7. Apply B
1
multipliers to the moments in beam
columns. The B
1
multiplier should be based on the full
LRFDlevel axial force in the member, including axial
forces due to lateral loading, but need be applied only
to that part of the moment in the beamcolumn that
is caused by gravity loading. (Designers may use the
conservative approximation of applying B
1
to the full
moment to avoid the obvious bookkeeping difficulties
involved in this calculation.)
8. Perform strength checks in accordance with the AISC
Specification, using an effective length factor, K, of
unity for members subject to compression (unless a
lower value is justified by rational analysis).
‡ This is an approximation. If the B
2
calculations for all stories are based
on story shears and drifts due to lateral load applied at the roof alone, the
resulting B
2
values should be reasonably accurate or conservative (high)
for all modes.
§ It should also be possible (as an alternative to the B
2
multiplier procedure
used herein) to adapt the modified geometric stiffness approach for use
with design by Modal Response Spectrum Analysis.
199_204_EJ3Q_2011_2010_07R.indd 203 9/13/11 2:38 PM
204 / ENGINEERING JOURNAL / THIRD QUARTER / 2011
9. As in the Equivalent Lateral Force procedure, confor
mance to ASCE 7 seismic drift limits may be checked
using either the same analysis used for strength checks
(convenient, but potentially very conservative) or an
analysis using the full unreduced stiffness of the struc
ture, with secondorder effects (B
1
and B
2
multipliers)
determined at the lower loads specified by ASCE 7 for
calculation of the stability coefficient, θ.
SUMMARY AND CONCLUSIONS
Provisions related to seismic design and stability in ASCE/
SEI 7, Minimum Design Loads for Buildings and Other
Structures (ASCE, 2005), and features of the Direct Analy
sis Method of design for stability in ANSI/AISC 360, Speci
fication for Structural Steel Buildings (AISC, 2005a; AISC,
2010), have been explored and compared. While there are
inconsistencies between the two sets of provisions, they are
not fundamentally incompatible. Recommendations are of
fered for the design of steel buildings for seismic effects and
stability in general conformance with both the AISC Speci
fication and the ASCE 7 standard.
REFERENCES
AISC (2005a), Specification for Structural Steel Buildings,
ANSI/AISC 360, American Institute of Steel Construc
tion, Chicago, IL.
AISC (2005b), Seismic Provisions for Structural Steel
Buildings, ANSI/AISC 341, American Institute of Steel
Construction, Chicago, IL.
AISC (2010), Specification for Structural Steel Buildings,
ANSI/AISC 360, American Institute of Steel Construc
tion, Chicago, IL.
AISCSSRC (2003), “Background and Illustrative Examples
on Proposed Direct Analysis Method for Stability Design
of Moment Frames,” Technical White Paper, AISC Tech
nical Committee 10, AISCSSRC Ad Hoc Committee on
Frame Stability, American Institute of Steel Construction,
Chicago, IL.
ASCE (2005), Minimum Design Loads for Buildings and
Other Structures, ASCE/SEI 705, American Society of
Civil Engineers, Washington, DC.
FEMA (2009), NEHRP Recommended Seismic Provisions
for New Buildings and Other Structures, FEMA P750,
Federal Emergency Management Agency, Washington,
DC.
Nair, R.S. (2009a), “Simple and Direct,” Modern Steel Con
struction, AISC, January.
Nair, R.S. (2009b), “A Model Specification for Stability
Design by Direct Analysis,” Engineering Journal, AISC,
Vol. 46, No. 1, pp. 29–37.
199_204_EJ3Q_2011_2010_07R.indd 204 9/13/11 2:38 PM
,:..
"' L 
., 
j
~
bl
~
of
July 13, 2003
Background and Illustrati ve Examples on Proposed
Direct Analysis Method for Stability Design of Moment Frames
Report 011 behalf of AISC TC 10
Prepared by: Gregory Deierleill
PREFACE
Summarized herein is background infonnation and illustrative examples for new frame stability design
provisions proposed by AISC TC 10 for the 2005 AISC Standard. In the latest AlSC ballot (July 2003),
most of the new provisions appear in a new Appendix 6, entitled "Direct Analysis Method for Moment
Frames", which provides an alternative to the frame stability provisions in Section B6 of the Standard.
The frame stability provisions of Section B6 are essentially identical to those in the 1999 (3"' edition) of
the LRFD Specification for Structural Steel Buildings, except for the addition of a minimum moment
requirement. The background and fundamental features of the standard (Section B6) and alternative
(Appendix 6) provisions are described herein. Several illustrative example problems are presented to
demonstrate and contrast the two stability design approaches.
ACKNOW LEDGM ENTS
The developments summarized in this report are the result of contributions by many individuals over the
course of deliberations by AISC TC 10 and the AlSC·SSRC Ad Hoc Comrnittee on frame stabi lity. The
AlSC·SSRC Ad Hoc Committee co·chairs (J. Yura and G. Deierlein) and members (W. Baker, J. Haliar,
T. Galambos, R. Heinge, L. Lutz, K. Mueller, S. air, C. Rex, R. Tremblay, D. White and R. Zieman)
developed the first draft of the provisions. Other noteworthy contributions include those of D. White, A.
Maleck, and R. Ziemian for their analysis of numerous benchmark validation studies and L. Lutz for
suggestions of illustrative example problems. Portions of this report are excerpts from an earlier paper
presented at the 2002 SSRC Annual Meeting, which introduced a preliminary (now superseded) draft of
the proposed provisions (Deierlein, Hajjar, Yura, White, Baker, "Proposed new provi sions for frame
stability using secood·order analysis", SSRC 2002 Annual Meeting, Seattle, WA).
INTRODUCTION
The proposed new provisions for frame stability represent the culmination of work by task committees in
AlSC and SSRC over the past four years, which incorporate concepts of second·order analysis and design
whose origins date back over twenty years. Concerted work on this began late in 1999, with the fonnation
of a joint AISC·SSRC Ad·hoc Committee whose charge was to develop improved specification provisions
for member and frame stability. The committee's goals were to develop design methods for stability that
made more effective use of modem computer analysis methods, while reducing the over· reliance on
effective buckling length procedures in the current AlSC Specifications. This ad·hoc committee was
combined with AI SC TC lOin 200 I , and the combined group developed provisions, which are proposed
for adoption in the 2005 AISC Standard. The new provisions were first balloted in March 2003 and have
since been revised to address comments raised by the AISC Specification Committee. This report reflects
the latest version of the proposed provisions for frame stability.
The July 2003 AlSC ballot outlines proposed provisions for the 2005 Standard, which will pernllt two
alternative methods to design for stability effects in moment frames . For discussion purposes, the two
approaches will be referred to as the "Effective Lellgth " and "Direct Allalysis" methods. Both approaches
"
,;0
,z,
~
."
""
require evaluation of secondorder effects and member force interaction equations. The methods differ in
their specific requirements for calculating secondorder effects and the axial strength tcrm, p., in the
member interaction equation. Requirements for the Effective Length method are contained in the proposed
Section B6 of the 2005 Standard. This method is essentially the same method as the approach used in
Chapter C of the 1999 AISCLRFD Specification. Requirements for the Direct Analysis method are
specified in a newly proposed Appendix 6 to the 2005 Standard.
This report begins with a brief review of key behavioral effects and secondorder analysis considerations,
which are relevant to stability design. Next, the two proposed approaches to framc stabi lit y are
summarized and contrasted through a design example of simplc cantilever column. This IS followed by
highlights of validation studies to evaluate the accuracy of the two proposed methods. The report
concludes with three design examples to illustrate practical application of the methods.
BEHAVIORAL EFFECTS
There are potentially many parameters and behavioral effects that influence stability of steelframed
structures. The extent to which these factors are modeled in analysis will affect the criteria that one
applies in design of the (rame, its members and connecllons. Without repeating more complete
presentations given elsewhere (Birnstiel and Imand, 1980; McGuire, 1992; White and Chen. 1993; ASCE,
1997; Deierlein & White 1998), it is helpful to review three basic aspects of behavior: geometric
nonlinearities, inelastic spreadofplasticity, and member limit states. These ultimately govern frame
deformations under applied loads and the resulting internal load effects.
Geometric Nonlinearities and Imperfections: Modern stability design provIsions are based on the
premise that the member forces are calculated by secondorder elastic analyses, where equilibrium is
satisfied on the defonned structure. When stability effects are significant, consideration must be given to
initial geometric imperfections in the structure due to fabrication and erection tolerances. For the purpose
of calibrating the stability requirements described later, initial geometric imperfection are conservatively
assumed as equal to the maximum fabrication and erection tolerances pennitted by the AI Code oj
Stalldard Practice (2000). For columns and frames, this implies a member outofstraightness equal to
UIOOO, where L is the member length between brace or framing points, and a frame outofplumb equal to
H/500, where H is the story height. The outofplumb is also limited by the absolute bounds as speCified
in the Code oj St{/l/{/ard Practice.
Inelastic Spread of Plasticity: The proposed analysis/design approaches are calibrated against inelastic
distributedplasticity analyses that account for spread of plasticity through the member crosssection and
along the member length. Thermal residual stresses in Wshape members are assumed to have maximum
values of 0.3Fy and are distributed according to the socalled Lehigh pattern  linearly varying across the
flanges and unifornl tension in the web (Deierlein & White 1998).
Member Limit States: Member strength may be controlled by one or more of the following limit states:
cross section yielding, local bUCkling, flexural buckling, and torsional flexural buckling. For structural
analyses envisioned for routine frame design, it is assumed that the analysis does not model local
flange/web buckling or torsionalflexural buckling. Therefore, these limits must be considered in separate
member design checks. For inelastic analyses, the member yield limit is incorporated directly in the
analysis; and for elastic analyses, this limit can be checked by an interaction equation that approximates
the PM yield surface. Whether or not the analysis captures inplane flexural buckling depends on the
extent to which the maximum moments are affected by distributed plasticity and member str3lghtness.
Concerns as to whether the analysis captures this effect suggest the need to apply a member check for lO
plane flexural buckling, even when an accurate secondorder analysis is used. As will be addressed later,
Page2oJI7
a key consideration for the inplane flexural buckling check relates to the assumed buckling length used in
calculating the design compression strength, ¢P".
SECONDORDER ELASTIC ANALYSIS
The AISC stability design provisions are developed for use with secondorder elastic analysis. In practice,
there are alternative approaches one can employ for conducting secondorder analyses, some of which are
more rigorous than others. For the purpose of this discussion, secondorder clastic analyses will be
categorized as "rigorous" or "approximate". The difference between these two depends on the extent to
which PO effects are modeled and whether the problem is "linearized" to expedite the solution.
Rigorous secondorder analyses are those that accurately model all significant secondorder effects.
Rigorous analyses include solution of the governing differential equation, either through stability
functions or computer frame analysis programs that model these effects (McGuire 1992; Deierlein &
White 1998). Many (but not all) modern commercial computer programs are capable of rigorous analyses,
though users should verify this. Methods that modify firstorder analysis results through secondorder
amplifiers (e.g. , B, and B: factors) are in some cases accurate enough to constitute a rigorous analysis, but
this depends on the magnitude of secondorder effects and other characteristics of the problem.
Approximate secondorder analyses are any methods that do not meet the requirements of rigorous
analyses. A common type of approximate analyses are those which only capture PtJ due to member end
translations (e.g., interstory drift) but fail to capture P oeffects due to curvature of the member relative to
its chord. Where PO effects are significant, errors arise in approximate methods that do not accurately
account for the effect of Po moments on amplification of both local member moments and the calculated
global (tJ)displacements. These errors can arise both with secondorder computer analysis programs and
with the B, and B, amplifiers. White and Maleck (2002) propose the following criteria to rule out cases
where PO effects can be safely ignored:
P, < 0. 15 P,L = 0.15(,(£IIL') (I)
where P, is the required column strength and P<L is the elastic buckling load in the plane of bending. The
alternative to this equation is to verify the accuracy of the secondorder analysis by comparisons to known
solutions for conditions similar to those in the structure. Examples of the errors one may encounter are
discussed by LeMessurier (1977) and Deierlein & White (1998).
BEAMCOLUM INTERACTION EQUATIONS
(SECTION HI OF THE 200S STANDARD)
80th the Effective Length (Section 86) and Direct Analysis (Appendix 6) stability procedures utilize the
beamcolumn interaction equations of Chapter H, albeit with differences in how the required strengths (P,
and M,) and the nominal compressive strength (p") are calculated. For reference in the later discussion, the
interaction equations for members under combined axial compression and bending are briefly reviewed.
For bisymmetric beamcolumns under combined axial compression and uniaxial bending, the 2005
Standard introduces a new interaction equation for checking outofplane (lateraltorsional) II1stability,
which is separate from the check for inplane (flexural buckling) instability. These separate equations are
introduced since they provide more accurate predictions of inplane and outofplane limit states, which
tests and theory show are independent phenomena. The separate equations reduce the conservatism in the
Page J oj17
·;:1 •
.::>
'::J
,
.:0
..J
current (I999 A1SCLRFD) provisions, which combine the two limit state checks into one equation, by
combining the most severe combinations of inplane or outofplane limits for P ./;p" and M,/,pM".
Shown here for illustration are the interaction equations in LRFD fornlat for bisymmetric beamcolumns
subjected to axial compression and uniaxial bending. For members subjected to compression and minor
axis bending. only the inplane check applies; whereas for columns under compression and major axis
bending, both checks apply.
The limit state of inplane flexural buckling is checked using the following equations, which have the
same format as those in the 1999 A1SCLRFD Specification:
p
for " ;, 0.2
,p, p"
(2a)
(2b)
where p" and M. are the required strengths, calculated from secondorder analysis under the design loads;
and p" and M" are Ole nominal compression and bending strengths, calculated in the plane of the frame.
For the Effective Length method, p" is detemlined using the effective buckling length KL in the plane of
bending, whereas in the proposed Direct Analysis method, p" is calculated using K=/ (KL=L) in the plane
of bending. For compact member sections, M" for the inplane check is equal to M".
The outofplane lateraltorsional limit state is checked by the following equation:
(3)
Here the required strengths p. and M" are the same as for Eq. 2a and 2b, and p" and M" are calculated
using the unbraced length in the outofplane direction. These outofplane nominal strengths would
typically be evaluated on the same basis for the Effective Length and Direct Analysis methods.
EFFECTIVE LE GTH METHOD
(SECTION 86 OF 2005 STANDARD)
The Effective Length (or critical load) approach for assessing member axial compressive strength has
been used in various forms in the A1SC Specification since 1961. The provisions proposed for Section 86
of the 2005 Standard are essentially the same as those from the 3'd edition (1999) of the AISCLRFD
Specification, with the exception of a new minimum moment requirement. The approach is based on
calculating effective column buckling lengths, KL, which have their basis in elastic (or inelastic) stability
theory. The effective buckling length KL, or aiternatively the equivalent elastic column buckling load, P,
= 1iEII(KL/, is used to calculate an axial compressive strength, p", through an empirical column curve
that accounts for member geometric imperfections, yielding, and residual stresses. This column strength
is then combined with the design moment strength, ,pM", and secondorder member forces, p. and M., in
the beamcolumn interaction equations.
Differences between the Effective Length and Direct Analysis approaches lie mainly in the inplane
check. Figure la shows a plot of the inplane interaction equation for the Effective Length approach,
Page 4 of 17
where the anchor point on the vertical axis, p.
KL
, is determined using an effective buckling length factor.
Also shown in this plot is the same interaction equation with the first term is based on the squash load. P,.
The loaddeformation response of a typical member, obtained from secondorder spreadofplasticity
analysis and labeled "actual response," indicates the maximum axial force, p., that the member can
sustain prior to the onset of instability. The loaddeflection response of a secondorder elastic analysIs, as
would be done in design practice, is also shown. The "actual response" curve reveals larger moments than
the secondorder elastic curve due to the combined effects of partial yielding and geometric imperfections,
which are not included in the secondorder elastic analysis. The intersection of the secondorder elastic
curve with the p.
KL
interaction curve represents the design strength. The plots in Fig. 2a show how the
effective length procedure has been calibrated to give a resultant axial strength, P" consistent with the
actual response. For slender columns, accurate assessment of the effective length (and p.
KL
) is critical to
achieving an accurate solution.
While the effective length approach is calibrated to accurately predict the resultant member strength, one
consequence of the procedure is that it underestimates the actual internal moments under the factored
loads (see Fig. I a). This is inconsequential for the beamcolumn (since the p.
KL
reduces the effective
strength in the correct proportion), but the reduced moment can affect design of the beams and
connections, which provide rotational restraint to the column. This is of greatest concern when the
calculated moments are small and axial loads are large, where PLl moments induced by column outof
plumb can be significant. As a safeguard for these cases, the Effective Length procedure in Section B6
includes a new minimum required moment strength for beams and connections, which restrain the column
ends. This requirement is specified through the following equation (Eq. B63 in the July 2003 Ballot),
'fM. > O.OI'f.P.L (4)
where EMu is the minimum required strength, P
u
is the required strength (axial compression force) in the
columns being restrained, and L is the column length.
P
P,
p ..
Pu
Pi
P
'"R
jWl
P,
0 .. ': jWl
elastic 2
nd
..order
p.
elastic 2
nd
order (D.A.)
actual response
actual response
(a)
Mp M
(b)
Mp M
Fig. I  Comparison of beamcolumn interaction checks for (a) the
effective length approach and (b) direct analysis approach
DIRECT ANALYSIS METHOD
(APPENDIX 6 TO 2005 STANDARD)
The Direct Analysis approach has been developed with the goal to more accurately model frame stability
effects in analysis, and thereby. eliminate the need for calculating effective buckling length factors for
Page 5 of 17
columns. As summarized below, the new provIsions in Appendix 6 of the 2005 Standard Involve
reducing the nominal elastic stiffness and applying a notional load to the frame. Some aspects of the
proposed provisions are similar to socalled "notional load·' methods found in steel standards in olher
countries, e.g., Canadian and Australian Standards and the Eurocode, however, many aspects of the
proposed provisions are unique 10 the AISC Standard and address known shoncomings of conventional
notional load approaches in other standards.
Like the Effective Length procedure, the Direct Analysis melhod begins with a basic requirement to
calculale internal member forces using a secondorder elaslic analysis. As will be shown later in the
examples, the Direct Analysis method places a greater reliance on the secondorder analysis (primarily in
the accurate calculation of secondorder moments, M.), and for this reason, the method stipulales
requirements to ensure accuracy of the second order analysis. Analysis rigor is most important where
secondorder amplifications are large, one measure of which is given by the ratio of member aXial
compression forces to their elastic buckling strengths (see Eq. I). Two additional requirements for Direct
Analysis are as follows:
• A notional load of N, = 0.002 Y, is to be applied in combInalion with other factored loads. where
N, is the notional lateral load applied at floor i and Y, IS the gravity load (from strength load
combinations) acting at floor i . The notional load is applied to represent the destabilizing effect of
a geometric imperfections and olher effects (yielding. nonideal boundary and loading condItIons,
elc.). The notional load magnitude of 0.002 corresponds to a frame outofplumb equal to HlSOO
(where H is the story height).
• The nominal elastic flexural stiffness assumed in the secondorder elastic analysis is equal 10
0.8tEI , where t is calculated as follows:
For members where p. S O.SPy: r = I
For members where p. > O.SP, : r = 4[P/ P, (J P/ P,)}
Alternatively, where p. > 0.5P, for any members in the frame. r = I provided that an additional
notional load of N, = 0.001 Y, is applied to the frame.
There are two reasons for imposing the reduced stiffness for analysi s. For frames with slender members,
where the limit state is governed by elastic stability, the 0.8 factor on stiffness results in a system deSign
strength equal to 0.8 times the elastic stability limit. This is roughly equivalent to the margin of safety
implied by design of slender columns by the effective length procedure where the design strength ;p.
0.9(0.877)P, = 0.79P. where P, is the elastic critical load, 0.9 is the specified resistance factor, and 0.877
is a reduction factor in the column curve equation. For frames with inlermediate or stocky columns, the
0.8t factor reduces the stiffness to account for inelastic softening prior to the members reaching their
design strength. The t is similar to the inelastic stiffness reduction factor implied in the column curve 10
account for loss of stiffness under high compression loads (P. > 0.5P,). and the 0.8 factor accounts for
additional softening under combined axial compression and bending. It is a fonuitous coincidence that the
reductions coefficients for the slender and stocky columns are close enough, such that the single reductIOn
faclor ofO.8t works over the full range of slenderness.
The reduced stiffness and notional load requirements only penain 10 analysis of the strength lImit stale.
and they do not apply to analysis of other serviceability conditions for excessive deflections, vibration,
etc. For ease of application in design practice, the reduction on EI can be applied by modifying E In Ihe
analysis; however, in doing so, one should consider whether the possible sideeffects of reducing EA.
Moreover, for computer programs that do semiautomated design, one should be sure that the reduced E IS
Page 6 of 17
'.
,;:p
,;:p
only applied for the secondorder analysis, The elastic modulus should not be reduced in design
equations, which involve E to evaluate the design strength (e.g., M, for laterally unbraced beams).
As shown in Fig. Ib, the net effect of modifying the analysis in the manner just described is to amplify the
secondorder moments to be closer to the actual internal moments in the member. It is for this reason that
the beamcolumn interaction for inplane flexural buckling is checked using an ax.ial strength P,
calculated from the column curve using the actual unbraced member length L, i.e., with K I . In fact ,
arguments have been made to use p). in the interaction equation. but this would require recalibration
of the analysis adjustments, including additional adjustments to account for member outofstraightness
(sweep). After considering alternative strategies, TC 10 decided to use the proposed method (with p.
based on L) as a pragmatic and conservative approach for practical design.
CANTILEVER EXAMPLE
To illustrate an application of the two stability design methods, consider the design of the cantilever beam
column shown in Fig. 2. The cantilever is subjected to the vertical and proportional hori zontal load
shown, such that the design is controlled by the combined p. and M. at the base of the column.
Maximum strengths are calculated for three different column lengths, with slenderness ratios of Ur 20,
40 and 60 (equivalent to KUr 40, 80, and (20). Bending is about the major axis, and the column has
full outofplane (lateral) restraint. The design checks are based on the inplane interaction check (Eq. 2a
b). ate that the checks were made using a resistance factor of
in compression (consistent with the 1999 AJSCLRFD Specification),
so the results would be slightly different if made with the revised value
of as proposed for Chapter E of the 2005 Standard.
Shown Fig. 3ac are plots of the axial load versus moment at the column
base for the three column lengths, deternlined according to the Direct
Analysis (DA) and Effective Length (EL) methods. Notice that the
internal moments Mil increase much faster with P
II
for the Direct
Analysis method, due to the reduced stiffness (0.8tEI) and added
notional loads. Most of the stiffness adjustment is due to the 0.8 factor,
since t only affects the column with Ur 20 (Fig. 4a,) where the
maximum load p. > 0.5P, 440 kips. Overlaid on these forcepoint
traces are the beamcolumn strength interaction diagrams, where the P
n
anchor point for the Effective Length method P,,KL is based on
oJ
0
It)
.:
Cl
r:i
<0
)(
0
.....
and for the Direct Analysis method P,L is based on L. Fig. 2  Canti lever Example
The calculated strengths, as determined by the two methods, are sununarized in Table I in tenns of the
max.imum vertical load p. (shown in bold). Net strengths for the two methods are within 10%, even
though the interim results are quite different. For example, as shown in Figs. 3ac and summarized in
Table 1, the maximum internal moments at the strength limit point are much larger for the Direct Analysis
method; whereas the P ,/¢P" ratios, which indicated the relative significance of the axial load and moment
terms in the governing interaction equation, are consistently larger for the Effective Length method.
Moreover, the P ,/¢>P, ratios for the Effective Length procedure do not change much with increasingly
slenderness, because this procedure relies to a much greater degree on capturing stability effects in the P,
term. Conversely, in the Direct Analysis procedure the P '/¢P, contribution decreases and the moment
term dominates the solution for cases with increasing slenderness.
Page 7 of 17
800
700
600
_ 500
a.
~ 400
"
D.
300
200
100
0
0
700
600
500
E: 400
~
"
300
D.
200
100
0
0
700
600
500 J
E: 400
~
"
300
D.
200
100
0
0
1000
1000
(e)
1000
DALIr=20
.....ELLIr·20
Interaction PnKL
~ Interaction PnL
2000
Mu (kin)
3000
DALir='O
EL Lir='O
Interaction PnKl
_Interaction PnL
2000
Mu (kin)
3000
DALir = 60
M EL Lir = 60
Interaction PnKL
Interaction PnL
2000
!tIu (kin)
300D
Fig. 3  Comparison of PM interaction curves for cantilever column example
(a) short Ur = 20, (b) medium Ur = 40, (c) long Ur = 60
Page 8 of 17
. .0
'.J
a e
 T bl I R esu ts or anti ever C ' 1 CI o umn E a m ~ l e
[frecti n Len th Method Direct Analys is "Iethod
Ur P.(kips) M.(kill) MI M, P/ (I',
P.(kips) M,lkiII) MI M, P / (1',
20 562 580 1.17 0.85 578 777 1.27 0.80
40 345 988 1.63 0.74 371 1680 2.16 0.56
60 193 1007 1.98 0.74 213 2376 3.53 0.37
This comparison highlighlS Ihe pros and cons of each method. Compared to Direct Analysis procedure,
the Effective Length method has the advantage of being less sensitive to the accuracy of the secondorder
analysis. On the other hand, the method requires calculation of effective column buckling lengths (KL),
which can be difficult for complicated structures. Direct Analysis eliminates the need to calculate
effective buckling lengths and provides more accurate measures of the true secondorder moments. This
latter point is important for the design of members and connections, which restrain the beamcolumn. For
example, in the cantilever column example, the base moments from the Direct Analysis procedure take
into account initial outorplumb and inelastic secondorder effects, which are not captured in the
Effective Length procedure. Referring to the secondorder moments reported in Table I, the difference in
moments between the two methods can be quite large. Subject to the assumed geometric imperfections
(outofplumb) and residual stresses, validation studies have shown that the momenlS calculated by the
Direct Analysis procedure are generally conservative and closer to the true values. Observations of the
type described here about the underestimation of design moments the Effective Length method, led to the
new minimum moment requirement (Eq. 4) for the Effective Length method. One should recognize,
however, that this minimum does not address cases such as shown in this cantilever example, where the
calculated moments in the Effective Length procedure are above the minimum of O.O/PL, but still less
than the actual values, which are calculated more accurately by the Direct Analysis procedure.
VALIDATI ON STUDI ES
Over the course of developing the proposed stability provisions, hundreds of validation analyses have
been investigated by members of the SSRCAISC Ad Hoc Committee. Some of the early investigations
(e.g., Maleck 200 I, Maleck and White 2003) he lped guide development of the provisions, and two recent
papers by Maleck and White (2002) and MartinezGarcia (2002) provide selected case studies to validate
the final version of the proposed design methods. These two studies investigated twentyfive frame
configurations under multiple load cases, representing several hundred analyses with about 150
comparison points between the two design approaches and refined nonlinear analyses. These studies focus
on the limit state of combined axial load and bending in the beamcolunms and do not specifically address
design checks in restraining beams and connections.
Examples of the frame configurations considered in the benchmark srudies by Maleck and White (2002)
are shown in Fig. 4. These twocolunm portal frames and individual colunm structures provide rigorous
test cases of nonredundant systems of varying slenderness, levels of axial compression, and leaning
column effects. Other multistory and multibay frames investigated by MartinezGarcia (2002) embody
attributes of realistic structures that pose particular challenges in evaluating stability, three of which are
presented in the next session of JIIustrative Examples. The problems investigated for the benchmark
studies are ones where secondorder effects are large and where errors between the stabi lity design
methods and more exact methods are accentuated. In this sense, these benchmark studies represent
extreme cases, which tend to exaggerate the differences one would typically encounter in design practice.
Page 9 of 17
;0
.... .)
Symmetric Frame
LeanedColumn
Frame
P
w_+
6
0
B.
E
b
• Ib ,Lb
Eel Ie ,Le
,
,
,
Lb » Lc
PinnedPinned
BeamColumn
it
,
B.
\
2W

;t
2P
+
f
2L,
(EI /L), = 0 and 3
(EI/L)b
(Llr), = 40
601L,= 0.002
BoiL, = 0 and 0.001
f" = 0 and 0.3F,
(EIIL),
= 0 and 1
(EI/L)b
(Llr), = 40
601L, = 0.002
BoiL, = 0 and 0.001
f" = 0 and 0.3F,
(2L1r), = 80
BoI2L, = 0 and 0.001
f" = 0 and 0.3F,
straight line and sine curve
Fig. 4  Test structures used for validation study (Maleck 2001)
Detailed analysis solutions based on secondorder spreadofplasticity analyses are used as benchmarks
against which the proposed design methods were validated. These benchmark solutions incorporate the
effects of gradual yielding, initial geometric imperfections, and residual stresses, as outlined previously in
the section of this report on Behavioral Effects. Thus, they represent the stateofart in simulating inelastic
stability of beamcolumns and frames. Material properties (E and Fy) in the spreadofplasticity analyses
were reduced using a resistance factor of 0.9, such that the maximum strength calculated in these anal yses
corresponds to the structural "design strength"  as opposed to a "nominal strength".
Overall, the two studies (Maleck and White 2002 and MartinezGarcia 2002) confirm that both the
Effective Length and Direct Analysis methods are sufficiently accurate (relative to current methods) for
design and that the errors (relative to the spreadofplasticity solutions) are comparable for the two
methods. Maleck and White report that on average the two approaches give strengths within 2% to 7% of
those obtained by refined analyses. The maximum discrepancies they observed for the Direct Analysi s
approach, relative to the refined analyses, range from  6% (unconservative) to +13% (conservative) for
members subjected to strongaxis bending and  13% to + 15% for members subjected to weakaxi s
bending. For the Effective Length approach, the maximum discrepancies range from  8% to + 18% for
strongaxis bending and  17% to + 17% for weakaxis bending. These errors are based on design checks
made with rigorous secondorder elastic analyses. Maleck and White caution that the Direct Analysis
design checks based on approximate P6 analyses can be up to 23% (unconservative) for members
subjected to weak axis bending. This is an example of why the Direct Analysis provisions specifY the
need for a rigrous analysis when secondorder effects are large. Maleck and White further note that for
frames where Pu < 0. 15 (71' ElfL' ), the maximum unconservative errors associated with approximate P
Page 100JI7
'.0
J>.
analyses for the Direct Analysis approach are limited to the maximum errors present in existing Effective
Length procedures.
ILLUSTRATIVE EXAMPLES
Three example problems are presented next to illustrate practical application of the proposed stability
design methods. The first two examples are [rames with heavy gravity loads and large secondorder
amplification factors. The third example is a stiffer sixstory frame, which is more representative of
multistory building frames. Each example includes a comparison of results from the Effective Length
and Direct Analysis approaches and more rigorous spreadofplasticity solutions. The spreadofplasticity
solutions have been independentl y reported by Maleck (2001) and MartinezGarcia (2002), and the design
solutions have been prepared by multiple members of A1SC TC 10 and the A1SCSSRC Ad Hoc
Committee. All design checks are based on the LRFD approach and load combinations. Resistance
factors used in the design checks are ¢.=0.9 for bending and ¢,=0.85 for compression, the latter of whi ch
is slightly smaller than the proposed change to ¢,=0.9 in Chapter E of the 2005 Standard.
LowRise Industrial Example
The first example, see Figure 5, is a framing bent from a large floor plan single story industrial building,
such as an automobile plant. With heavy material handling equipment hung from the roof and a small
wind exposure, such structures are dominated by gravity loads wilh large secondorder effects
(Springfield, 199 1). Loading shown in Figure 5 represents an eleven bay configuration with ten leaning
columns (only two of which are shown) and two lateralload resi sting columns. The concentrated load P
has a tributary roof area of 35 ft x 35 ft , and the wind load W = 5.12 kips.
The member sizes sati sfy a drift limit of Hl400 for the service load wind of 0.7W, and the design strength
of the frame exceeds the minimum requirement of the LRFD strength load combinations. Based on
refined spreadofplasticiry analyses, the frame has a desi gn strength ratio 17% larger than the required
strength for gravity (¢A.UQ. /.6L=1.I 7, where ~ = 0 . 9 ) and 20% larger than required under gravity plus wind
(¢A/m.o.5L /.6,.,=1. 20). Using the equation B55 (from the July 2002 ballot of Chapter B for the 2005
Standard), the secondorder amplification factor under design gravity loads is B,= 2.41. Under the design
gravity plus wind loading combination, B,= 1.74.
4P 4P
Frame spadng = 35'·0"
W)
W27 x 84 W27 x 84
Fy = 50 ksi
'"
9
..,.
x E = 29,000 ksi
00
0 ~
~
~
J
\.
,
l
3 @ 35'0" = 105'0"
DL 80 psI Load Combinations:
'1
LL 40 psI 1.2DL + 1.6LL
Wind 16.25 psI 1.2DL + 0.5LL + 1.6WL
Figure 5  SingleStory Industrial Building
Page II ofl7
• .£.'
' . .,

Axial column forces and maximum moments under the factored load combinations are summarized in
Table 2. The Effective Length results are from a secondorder analysis of a model based on the ideal
geometry of the frame under the factored load combinations (no geometric imperfections or notional loads
are introduced). The Direct Analysis results incorporate initial geometric imperfections through the
notional load of 0.2% times the factored gravity loads (1.20 + 1.6L for the first combination and 1.20 +
0.5L for the second combination); and stiffness degradation is incorporated by reducing the flexural
stiffness of all framing members by to 0.8EI. Since the axial load ratio PIP, <0.5, no additional tfactor
stiffness adjustments are required. The "spread of plasticity" results are from a secondorder inelastic
analysis, which models gradual yielding through the member cross sections and along their length due to
the combined effects of thermal residual stresses and the applied loads.
Table 2: Member Effects Under Factored Load Combinations
L R' I d . I E I or ow Ise n ustna xample
Analysis/Design l\ l elhod
Load Case
Check
SI)read of Effecti ve Direct
Plasticitv Allah'sis
p,. (kip) 215 216 218
1.2D+1.6L
M"" (kin) 930 407 1220
Mbm (kin) 8660 8410 8690
P"" (kip) 154 158 160
1.2D+0.5L+1.6W
M" (kin) 1310 1040 1550
Mbm (kin) 6490 6360 6630
Referring to Table 2, the Effective Length and Direct Analysis methods both predict the maximum beam
moments and axial column forces within about 4% of those from the spreadofplasticity analysis. On the
other hand, there are significant differences in the column moments, particularly for the gravity load case
(1.2D + 1.6L). The Direct Analysis method predicts the column moments on average about 25% higher
than the spreadofplasticity solution, and the Effective Length method predicts column moments on
average about 40% smaller than the spreadofplasticity solution. These differences are also reflected in
the calculated displacements. This small moments calculated according to the Effective Length method
illustrate the need for the newly proposed minimum connection moment requirement (Eq. 4, 'f.M. >
0.0I''i.P"L). Without this minimum requirement, the comlection would be underdesigned for the second
order moment induced by the combined effects of gravity load and column outofplumb.
Using the member forces from Table 2, the columns are checked using the interaction formula for inplane
or outofplane (torsional flexural) failure, and the resulting interaction ratios are summarized in Table 3.
For the Effective Length method, the inplane checks are based on a column strength of ¢P = 236 kips,
obtained with an effective length factor of K = 2.3 using Eq. CC26 of A1SC (1999). Inplane checks for
the modified stiffness and notional load methods are based on ¢p ...
t
= 511 kips, and outofplane checks
Table 3: Interaction Values for LowRise Industrial Example
Analysis/ Design 1\ l ethod
Load Case
Member
Check Effeclive Direcl
Length Analvsis
1.2D+1.6L
Inplane 1.05 0.83
Outofplane 0.62 0.81
1.2D+0.5L+I.3W
Inplane 1.0 I 0.82
OutofpJanc 0.58 0.77
Page 120[17
are all based on ;p ..
L
361 kips. The column design moment is ;Mp 2718 kin. Inplane interaction is
checked using Eqs. 2a & 2b, and the outofplane check is made using Eq. 3.
Referring to Table 3, both the Effective Length and Direct Analysis checks are governed by the inplane
strength (shown shaded). The Effective Length method is more conservative, as evidenced by a larger
interaction value as compared to the direct analysis method. The inplane checks can be compared to
inelastic limit load ratios of ;),."0 and ;),."D.0.J/. >1.61V obtained from the spreadof
plasticity analyses. The inverse of these limits (0.85 and 0.84 for gravity and gravity+wind, respectively)
help to gauge the conservatism in the methods, where larger interaction checks would be conservative and
smaller checks unconservative. Compared to these values (0.85 and 0.84) the inplane checks for the
Effective Length method about 15% conservative, whereas the Direct Analysis results appear slightly
unconservative (e.g., 0.83 < 0.85 and 0.82 < 0.84). However, since the member forces vary nonlinearly
with load (due to second order effects), this simple comparison is approximate and a more accurate
comparison would be obtained by scaling up the loads in the Direct Analysis to the point that the inplane
interaction check is equal to 1.0. Scaling the loads in this way results in a limit load for
the Direct Analysis which is about 6% smaller (conservative) as compared to the inplane limit from the
spreadofplasticity solution. Thus, this case demonstrates that the Direct Analysis is conservative (safe)
and provides the potential for a more efficient design as compared to the Effective Length method.
Exa mple Connection Design for Effective Length Method:
For the gravity load case, the minimum beamcolumn
connection strength provision of the Effective Length
procedure (Eq. 4) provides for a minimum required connection
strength of 1:M, > 2330 kin. This is calculated using the
vertical load in one lateral load resisting column and one of the
leaning columns (a total of EP, 1078 kips). Comparing thi s
to the more exact required column moment from the spreadof
plasticity solution (M, 930 kin) indicates that the minimum
required by the Effective Length method is quite conservative
in this case. The required strength using the Direct Analysis
method would be 1220 kin. To help gauge the impact of these
provisions, a connection designed for the Effective Length
method moment of 2330 kin is shown in Fig. 6. This
connection would require eight  I inch diameter bolts, which
is not excessive for the connected members.
Grain Storage Bin
_.
Figure 6: Example Connection Design
(81" dia. bolts)
The second example is the support rack for a grain storage bin with the dimensions and loading shown in
Figure 7. This is a case where calculation of the column effective lengths is not obvious, and where the
Direct Analysis method offers a clear benefit. Colunms are assumed to be braced outofplane and the
crossbeams and bracing are pinconnected to the colunms. For the diagonal bracing, oneinch diameter
round bars are assumed. Using an elastic critical load analysis, the secondorder amplification factors are
B, 2.75 and B, 2.20 for the gravity and wind load combinations, respectively. The spreadofplasticity
analyses predict inelastic limit load ratios of 1.13 and ;),. I1G 16W 1.07.
As in the previous example, results for the Effective Length method are calculated for the ideal geometry
and stiffness; whereas the Direct Analysis method is based on a reduced stiffness (0.8EI) and with
notional loads applied in combination with the design loads. Like the previous example, no additional t
adjustment of stiffness properties is required for Direct Analysis since the axial load ratio PIP, < 0.5.
Page 13 oj 17
'.0
J
w
M
X
WI2x26
"' iI"""
H
12'0"
H
Loading:
IV  \.75 kops (10 psQ
G  360 kips
Load Combinations:
1;>
1.4G
1.2G + 1.6W
Materials:
1;> Fy  SOksi
'"
E  29.000 ksi
Figure 7  Grain Bin Support Frame
Maximum column forces and moments (required strengths) are summarized in Table 4 and the interaction
checks are summarized in Table 5. As in the previous examples, there is not much difference in axial
loads between the methods, but there are large variations in the calculated moments. This is particularly
prevalent for the gravity load case, where the calculated column moments for the Effective Length method
are essentially zero, and the moments for the Direct Analysis method are about 36% larger than those in
the spreadofplasticity analysis. Under the lateral load case, the differences are less, with moments for
the Effective Length method about 24% less (unconservative) than the spreadofplasticity results and
those for Direct Analysis about 38% larger (conservative).
a e em er eets T bl 4 M b G or ram S torage III xample
Member
AnalYsis/ Desig n Met hod
Load Case
Check
Spread of Effecti ve Direct
Pl astici ty Analysis
P,.,,,,,Jkip) 233 247 249
lAG
P,.""dkin) 255 252 257
M (kin) 161 2 220
P , (kip) 203 217 220
1.2G+\.6W
P. "
kin) 224 225 230
M, (kin) 380 288 526
T bl a e 5: Interactton Va ues or ram
G . S
torage B' E 10
Member
Analysis/Design Method
Load Case
Check
Effecti ve Direcl
Length Anal\'sis
\.4GL
Top Column 1.07 0.84
Bot. Column 1.04 0.85
1.2GL+1.6W
Top Column 1.12 0.96
Bol. Column 1.11 0.99
Page 14 of 17
• .:,
.::>
.:::>
.....
. ..:>
The interaction checks, shown in Table 5, are based on the following inplane column strengths: critical
load method ;p.KL .. op = 232 kips (K = 2.4), .... = 243 kips (K = 2.9); and the direct analysis method,
;p.
L
.• "" = 355 kips, = 366 kips. The K factors for the Effective Length procedure are based on an
elastic critical load analysis of the structure under gravity loads. The results in Table 5 show that strength
interaction checks based on the Effective Length method are roughly 20% more conservative than the
Direct Analysis method. Using the spreadofplasticity analysis as a benchmark of the actual behavior,
interaction values larger than 0.89 (for lAG) and 0.93 (for 1.2G+I.6W) are conservative. The Effective
Length interaction values (1.07 and 1.12) exceed these and are about 20% conservative. The Direct
Analysis method is sli ghtly conservative for the gravity plus wind case (0.99 > 0.89) and appears slightly
unconservative for the gravity load case (0.85 < 0.93). However, as mentioned in the industrial frame
example these linear comparisons are only approximations. When the gravity loads are scaled in the
direct analysis to provide an interaction value of 1.0 for the bottom column, the limit load ratio was ;)., ' Ii
= 0.99, which exceeds the value of ;)."G= 1.07 from the spreadofplasticity solution. This indicates that
the Direct Analysis is, in fact , 9% lower (conservative) as compared to the spread of plasticity solution.
Multistory Frame Example
The final example is the multi story frame shown in Figure 8. One load case is investigated (J .OG +
I .OW, where the specified loads are already factored), and member forces and interaction checks are
presented for the three columns in the first story. Unlike the previous examples , this frame is fairly stiff
with B1 = 1.10 for the first story. The secondorder spreadofplasticity analysis predicts an inelastic
design strength ratio of ;)., ' Ii = 1.06 for this frame, which combined with the low B 1 indicates that it is
dominated more by yielding than secondorder effects. The center columns of the first two stories have
high axial forces and are subject to the t  factor adjustment in the Direct Analysis method. As di scussed
earlier, when PIP, > 0.5 for any column, the t  factor adjustments can be used or an additional notional
load of 0.001 Y, can be used. In thi s example, both approaches are presented and compared.
ell
IPf2"O
w
x
IPOOO
1ii
ffi
x
JPEJOO
N
ffi
x
IPf330
ffi
x
IPOW
m
w
x
IPf400
(il
:;:
CI2
w
x
2 0 6.Om  12m
CJ3
l.IwIl
Gravity: "9. 1 kN/ m (11oor)
31.7 kN/ m (root)
Wind: 20 ..... kN (store 1 • 5)
10.23 kN (root)
F, _ 235 N/ mml'
E • 205 kN/mml
Figure 8  Multistory frame
Page 150[1 7
The first floor column forces, summarized in Table 6, reveal that differences between the three methods
are much smaller than in the previous examples. This follows from the fact that the secondorder
amplification is small er in this example, which is more typical of most multistory frames than the prior
examples.
T bl 6 M b a e em er eelS or F u t, story rame
E xample
Location and
Anal"sis/Desie" Method
[(ferr
Spread of EffeClivt
Direct Direct
Analysis "ilh Analysis with t
(t.OG+ t.OW)
Plasticity Length
Notional Load
Reduction
P,,, (kN) 683 672 659 662
P " (kN)
1720 1770 t770 1770
P,,, (kN) 92t 884 897 894
M ,(kNml 67 48 58 59
M,,(kNm) t t5 11 8 143 135
Mu (kN·rn) 99 87 96 96
Results of the beamcolumn interaction checks (Table 7) show that the Effective Length method is sli ghtly
less conservative than the Direct Analysis method, which is in contrast to the previous two examples
where the opposite was true. Based on the AlSC (1999) alignment charts, the effective buckling length
factors for the first story columns are K = 1.35, assuming the AlSC suggested value of G = 1.0 for the
foundation support . Accordingly, the inplane interaction checks for the Effective Length procedure are
based on design compression strengths of $P,KL" = $P,KL IJ = 1580 kN and $P,KL"  2140 kN. The in
plane Direct Analysis checks (with K = J) are based design strengths of "  IJ  1680 kN and
,P,L 12 = 2230 kN. All outofplane checks are based on K = I, with = IJ = 1460 kN and $P'L
12 = 2010 kN; and moment strengths of = = 175 kNm and  275 kNm are used
throughout. As summarized in Table 7, the resulting interaction checks were all close, with the Direct
Analysis solutions about 2% to 3% conservative, relative to the Effective Length method. With reciprocal
of the inelastic limit load factor equal to 0.94, the average interaction values ranging from 0.94 to 0.99
indicate that all of the stability design methods are conservative in this ca e.
Ta bl e 7: Interachon VI a ues MI ' F or u ttstory ' rame E xampte
Location
Anahsis/Design Method
(t.OGL+t.OW)
Eerective Direct Analysis Direct Anll)sis
Length with Notional Load with t Reduction
Ctl m·olane 0.67 0.69 0.69
CII  oUIofpt'ne 0.54 0.56 0.57
Ct2 Inplane I.2t 1.25 1.23
Ct2 1.06 t.t5 1.12
cn  Inptane 1.00 1.02 1.02
Cl3 outofolanc 0.85 0.92 0.9t
A VCT3SlC inolanc 0.96 0.99 0.98
Page 16 of 17
REFERENCES
A1SC (1999), Load alld Resistallce Factor Specificatioll for Structural Steel Buildillgs, Arner. Insl. of
Steel Constr., Chicago, IL.
ASCE (1997), "Effective Length and Notional Load Approaches for Assessing Frame Stability", ASCE,
New York, NY.
A1SC (2000), Code ofStalldard Practicefor Steel Bllildillgs alld Bridges, American Institute of Steel
Construction, Chicago, IL.
Bimstiel, C., Imaod, 1. S. B. ( 1980), "Factors Influencing Frame Stability," JI. of the Struct. Div .. ASCE,
I 06(ST2), 491504.
Deieriein, G.G., White, D. W. (1988), "Chapter 16  Frame Stability," Gllide to Stability Desigll Criteria
for Metal Structures, Ed. T.V. Galambos, Fifth ed., Wiley, 1998.
Deierlein, Hajjar, Yura, White, Baker, "Proposed new provisions for frame stability using secondorder
analysis", Proceedillgs 0fSSRC 2002 Allllllal Meetillg, Seattle, WA
LeMessurier, W. 1., (1977), "A Practical Method of SecondOrder Analysis  Part 2 Rigid Frames," Ellgr.
JI., A1SC, 14(2), 4967.
Maleck, A. E. & White, D. W. (2002), "Direct Analysis Approach for the Assessment of Frame Stability:
Verification Studies," SSRC Annual Technical Meeting, Baltimore, MD, SSRC, 17 pgs.
Maleck, A. E. & White, D. W. (2003), "Alternative Approaches for Elastic Analysis and Design of Steel
Frames: Part I Overview, PartH Verification Studies, " JI. ofStruct. Ellgrg., ASCE, submitted for
review.
MartinezGarcia, J.M. (2002), "Benchmark Studies to Evaluate New Provisions for Frame Stability Using
SecondOrder Analysis," M.S. Tllesis, supervised by R.D. Ziemian, Bucknel1 University, 241 pgs.
McGuire, W. (1992), "Computeraided analysis," COllst.1 Steel Desigll Allillti. Gllide, Dowling, et al.
(eds.), Elsevier, NY 915932.
Springfield, J. (1991), "Limits on SecondOrder Elastic Analysis," Proc. SSRC Annu. Tech. Sess., SSRC,
Bethlehem, PA, 8999.
White, D. W., Chen. W.F. (eds.) (1993), Plastic Hillge Based Methodsfor Adv. Allalysis alld Desigll of
Steel Frames, SSRC.
Page 170JI7