Engineering Structures 21 (1999) 352–364

Design guide for steel frames using advanced analysis program

a,* b
Seung-Eock Kim , Wai-Fah Chen
a
Department of Civil Engineering, Sejong University, Seoul 143-747, South Korea
b
School of Civil Engineering, Purdue University, West Lafayette, IN 47907, USA

Received 18 August 1997; received in revised form 2 October 1997; accepted 2 October 1997

Abstract

The aim of this paper is to provide a design guide of the LRFD based advanced analysis method for practical frame design. To
this end, the step-by-step analysis and design procedures for the practical advanced analysis method are presented. The proposed
advanced analysis program and input data format are discussed. Case studies are drawn to show the detailed design procedures of
the practical advanced analysis method. The case studies cover a two-bay frame and a four-story, five-bay frame. Member sizes
determined by the advanced analysis procedures are compared with those determined by the LRFD procedure, and a good agreement
is generally observed. The advanced methods achieve the LRFD design requirements without tedious separate member capacity
checks, including K-factor calculation. The methods are practical and therefore recommended for general use.  1998 Elsevier
Science Ltd. All rights reserved.

Keywords: Advanced analysis; Analysis program; Frame design; Practical method; Steel frame

1. Introduction braced and unbraced steel frames, the spatial behavior

of frames is not considered and the lateral torsional
In a recent paper [1], a practical advanced analysis buckling of members is assumed to be prevented by
method has been proposed for steel frame design. The adequate lateral braces. A compact W-section is assumed
method incorporates three models including an explicit so that sections can develop full plastic moment capacity
imperfection modeling, an equivalent notional load without local buckling. The frames are subjected to static
modeling, and a further reduced tangent modulus mode- loads, rather than earthquake or cyclic loads.
ling. The advanced analysis method can sufficiently cap-
ture the limit state strength and stability of a structural
system and its individual members. As a result, the
method can be used for practical frame design without
tedious separate member capacity checks, including K- 2. Practical advanced analysis
factor calculation.
Since the power of personal computers and engineer-
ing workstations is rapidly increasing, practical use of The important attributes which affect the behavior of
the advanced method is currently becoming feasible. In steel framed structures may be grouped into two categor-
this paper, the advanced analysis program is briefly ies: geometric and material nonlinearities. The geometric
introduced, and the design procedures using the program nonlinearity includes second-order effects associated
are presented. Case studies are drawn to show the with P-␦ and P-⌬ effects and geometric imperfections.
detailed application procedures of the practical advanced The material nonlinearity includes gradual yielding asso-
analysis method. The case studies include a two-bay and ciated with the influence of residual stresses and flexure.
a four-story, five-bay frame. The key considerations of the practical advanced analy-
Since the present study is limited to two-dimensional sis are discussed in what follows, and summarized in
Table 1 compared with the conventional AISC-LRFD
method. The proposed method has been verified for a
* Corresponding author. wide range of frames [2].

0141-0296/99/$—see front matter  1998 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 1 - 0 2 9 6 ( 9 7 ) 0 0 2 0 9 - 5
 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364 353

Table 1
Key considerations of LRFD and proposed advanced analysis method

Key considerations LRFD Proposed method

Second-order effects Column curve B1, B2 factor Stability function

Geometric imperfection Column curve Explicit imperfection modeling method
␺ = 1/500 for unbraced frame
␦c = Lc/1000 for braced frame
Equivalent notional load method
␣ = 0.002 for unbraced frame
␣ = 0.004 for braced frame
Further reduced tangent modulus method
Et⬘ = 0.85Et
Stiffness degradation associated with residual Column curve CRC tangent modulus
stresses
Stiffness degradation associated with flexure Column curve interaction equations Parabolic degradation function

冤冥
2.1. Geometric nonlinear effects ṀA
EtI
Stability functions are used to capture geometric non- ṀB =
L
linear effects, since they can account for the effect of Ṗ

冋 册
the axial force on the bending stiffness reduction of a
member. The benefit of using stability functions is that  S 22 
␩A S1 − (1 − ␩B) ␩A␩BS2
 
0
it enables only one element to accurately predict the non- S1
linear effect of each framed member [2].

2.2. Stiffness degradation associated with residual



␩A␩BS2 冋
␩B S1 −
S 22
S1
(1 − ␩A) 册 0  (2)

stresses  0 0 A/I 

冤冥
The CRC tangent modulus is employed here to ␪˙ A
account for gradual yielding effects due to residual ␪˙ B
stresses along the length of members under axial loads
ė
between two plastic hinges. When this model incorpor-
ates appropriate geometrical imperfections, it may pro- where
vide a very good comparison with the plastic zone sol- ṀA,ṀB,Ṗ = incremental end moments and axial
utions [2]. From Liew [3], the CRC Et may be written as: force, respectively
S 1, S 2 = stability functions
Et = 1.0E for P ⱕ 0.5Py (1a) Et = tangent modulus
= moment of inertia of cross-section
冉 冊
I
P P L = length of element
Et = 4 E 1− for P > 0.5Py (1b)
Py Py A = area of cross-section
␩A,␩B = scalar parameters for gradual
where Py = squash load and E = elastic modulus. inelastic stiffness reduction
␪˙ A,␪˙ B = incremental rotations at element ends
2.3. Stiffness degradation associated with flexure ė = incremental axial deformation.

A gradual stiffness degradation of a plastic hinge is The parameter ␩ represents a gradual stiffness
required to represent the distributed plasticity effects reduction associated with flexure at sections. The partial
associated with bending actions. Herein we shall intro- plastification at cross-sections at the end of elements is
duce the hardening plastic hinge model to represent the denoted by 0 ⬍ ␩ ⬍ 1. The ␩ may be assumed to vary
gradual transition from elastic stiffness to zero stiffness according to the simple parabolic expression as:
associated with a fully developed plastic hinge. When ␩ = 4␣ (1 − ␣) (3)
the hardening plastic hinges are present at both ends of
an element, the incremental force–displacement relation- where ␣ is the force–state parameter obtained from the
ship may be expressed as [2,3]: limit state surface corresponding to the element end as:
354 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364

2.4. Geometric imperfection

Geometric imperfection models, combined with the

CRC tangent modulus model, are discussed in what fol-
lows. They are an explicit imperfection modeling, a
notional load modeling, and a reduced tangent modu-
lus modeling.

2.4.1. Explicit imperfection modeling

The AISC code of standard practice [4] limits an erec-
tion out-of-plumbness equal to Lc/500 in any story. This
imperfection value is conservative in taller frames since
the maximum permitted erection tolerance of 50 mm
(2 in) is much less than the accumulated geometric
imperfection calculated by 1/500 times building height.
In this study, however, Lc/500 is used for geometric
Fig. 1. Smooth stiffness degradation for a work-hardening plastic imperfection without any modifications. This is because
hinge based on LRFD sectional strength curve. the system strength is often governed by a weak story
which has an out-of-plumbness equal to Lc/500 [5].
P 8 M P
␣= + for ⱖ 0.2 (4a) The frame out-of-plumbness may be used for geo-
Py 9 Mp Py metric imperfections for unbraced frames but not for
braced frames. This is because the P-⌬ effect caused by
P M P out-of-plumbness is diminished by braces. As a result,
␣= + for ⬍ 0.2 (4b)
2Py Mp Py the member out-of-straightness instead of the out-of-
plumbness should be used to account for geometric
where P, M = second-order axial force and bending imperfections for braced frames. The AISC code rec-
moment at the cross-section Mp = plastic moment ommends a maximum fabrication tolerance of Lc/1000
capacity. for a member out-of-straightness. In this study, a geo-
In Fig. 1, the term ␣ of 1.0 represents the plastic metric imperfection of Lc/1000 is adopted by the cali-
strength surface and ␣ of 0.5 is assumed to be the initial bration with the plastic zone solutions. The out-of-
yield surface which is assumed to have the same shape straightness may be assumed to vary sinusoidally with
as the LRFD plastic strength surface. As the ␣ value a maximum in-plane deflection of Lc/1000 at the mid-
varies from 0 to 0.5, the element end remains in the height. Ideally, many elements are necessary in order
elastic state. When the ␣ moves from 0.5 to 1.0, the to model the sinusoidal out-of-straightness of a beam–
element stiffness changes with a parabolic degradation column member. It has been found that two elements
shape shown in Fig. 2. with a maximum deflection at the mid-height of a mem-
ber are practically adequate for reflecting the imperfec-
tion effects [2].

2.4.2. Equivalent notional load modeling

The geometric imperfections of a frame may be
replaced by equivalent notional lateral loads that are
expressed as a fraction of the gravity loads acting on a
story. In this study, the proposed equivalent notional
load for practical use is 0.002⌺Pu, where ⌺Pu is the total
gravity load in a story. The notional load should be
applied laterally at the top of each story. For braced
frames, the notional loads should be applied at mid-
height of a column since the ends of the column are
braced. In this study, appropriate notional load factor
equal to 0.004 is adopted. It may be observed this value
is equivalent to the geometric imperfection of Lc/1000.

2.4.3. Further reduced tangent modulus modeling

Fig. 2. Parabolic plastic hinge stiffness degradation function with ␣0 The idea of using the further reduced tangent modulus
= 0.5 based on LRFD sectional strength equation. method is to further reduce the tangent modulus Et to
 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364 355

account for the geometric imperfections. The reduced

tangent modulus method is capable of eliminating the
somewhat tedious work of explicit imperfection mode-
ling or notional load input. Fig. 3 shows the member
stiffness reduction curve of the further reduced tangent
modulus Et⬘. The appropriate reduction factor of 0.85 to
Et is determined by the calibration with the plastic zone
solutions [2]. The similar reduction factor of 0.85 may
be used for braced as well as unbraced frames.

3. Program and input data format

This section discusses the proposed analysis program

for a two-dimensional steel frame design. This program
is developed for practical design use from the refined
plastic hinge program [3]. The program is divided into
three FORTRAN programs including DATAGEN,
INPUT, and PAAP. The first program, DATAGEN,
reads an input file P.DAT and produces the generated
data file INFILE. The second program, INPUT,
rearranges INFILE into three data files including
DATA0, DATA1, and DATA2. The third program,
PAAP (Practical Advanced Analysis Program), provides
the output files P.OUT1 and P.OUT2. The P.OUT1 file
Fig. 4. Operating procedures of the proposed advanced analysis pro-
contains an echo of the information from the input data gram.
file P.DAT. The P.OUT2 file contains detailed infor-
mation on load–displacement relationship and member
forces of the structures. The schematic diagram in Fig. 77 compiler v. 1.00 and Lahey FORTRAN 77 compiler
4 shows the operation procedures of the program. The v. 5.01. The second is in a Sun 5 machine using a Sun
programs may be executed by issuing the batch file com- FORTRAN 77 compiler. The program sizes of DAT-
mand ‘RUN’. AGEN, INPUT, and PAAP are 8 KB, 9 KB, and 84 KB,
The program has been tested in two computer respectively. The total size of the three programs is as
environments. The first is in the IBM 486 or equivalent small as 111 KB ( = 0.111 MB), and so the 3.5 inch high
personal computer system using a Microsoft FORTRAN density diskette (1.44 MB) can accommodate the three
programs and several example problems. The program
can generate the output file P.OUT in a reasonable time
period. The run time in IBM 486 PC with a memory of
640 K to get P.OUT for the four-story, five-bay frame
shown in Figs. 10 and 11, is 4 min, 10 sec and 2 min,
30 sec in real time rather than CPU time by using
Microsoft FORTRAN and Lahey FORTRAN, respect-
ively. In the Sun 5, the run time varies approximately 2–
3 min, depending on the degree of occupancy by users.
The input format in the use of the proposed program
is presented in what follows. Except considerations of
geometric imperfections and incremental loads, the input
data format of the proposed program is basically the
same as that of a usual linear elastic analysis program.
The input data consist of 13 data sets including five con-
trol data, three section property data, three element data,
one boundary data, and one load data set. The coordinate
systems both in the input data and in the output are based
on global coordinates rather than on local coordinates.
Fig. 3. CRC and reduced tangent modulus for members with geo- Since the unit system is not specified by the program,
metric imperfection. users may choose the unit in consistency at their con-
356 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364

venience. The detailed instructions of the input format

1.2D + 1.6(Lr or S or R) + (0.5L or 0.8W)
are summarized in Table 2, and presented in detail in
Ref. [2]. 1.2D + 1.3W + 0.5L + 0.5(Lr or S or R)
The proposed program may be used rigorously for
1.2D ⫾ 1.0E + 0.5L + 0.2S
practical design by taking advantage of the small pro-
gram size, short run time, and simple input format. The 0.9D ⫾ (1.3W or 1.0E)
source programs may be obtained on diskette (see Ref.
[2]). where
D = dead load
L = live load
4. Analysis and design procedures Lr = roof live load
W = wind load
4.1. Step 1: live load reduction S = snow load
E = earthquake load
The live load reduction is based on the ASCE 7-95 R = rain load.
[6] as:
4.3. Step 3: preliminary member sizing

冢
L = 0.25 +
15
√Ai o冣
L ⱖ ␣Lo (5) The preliminary member sizing is intrinsically depen-
dent on engineers’ experiences, the rule of thumb, or
some simplified analysis. For example, beam members
are usually selected assuming that beams are simply sup-
where L = reduced design live load, Ai = member influ- ported and subjected to gravity loads only. For the pre-
ence area in square feet (Ai ⱖ 36 m2, or 400 ft2), Lo = liminary sizing of column members, the overall drift
unreduced design live load, and ␣ = 0.5 for members requirements should be a good guideline to determine
supporting one floor, ␣ = 0.4, otherwise. preliminary member sizes rather than the tedious
It is important to properly carry out the application of strength checks of individual columns.
the live load reduction in analyzing a structural system.
This is because the influence area for each beam and 4.4. Step 4: modeling of element
column is generally different, and different influence
area results in a different reduction factor. In the present Columns in braced and unbraced frames may be mod-
study, the live load reduction procedures follow the work eled by two and one element, respectively. Nodal points
of Ziemian and McGuire [7]. The method is based on should be provided where concentrated loads are
the use of ‘compensating forces’ calculated by: applied. When a beam is subjected to uniform loads, it
1. applying beam live load reduction factor to the col- may be modeled by two elements. The uniform loads
umn connected beams; should be converted into these concentrated loads. The
2. applying column live load reduction factor to col- adequacy of the number of element recommended herein
umns; and has been verified in Ref. [2].
3. determining compensating forces due to different
4.5. Step 5: modeling of geometric imperfection
reduction factors between columns and beams at the
beam-to-column intersections. Geometric imperfections can be accounted for by one
The compensating forces are generally directed of the three modelings including the explicit imperfec-
upward since columns typically have a larger influence tion modeling, the equivalent notional load modeling,
area and a larger reduction factor than beams. and the further reduced tangent modulus modeling. The
explicit imperfection modeling and the equivalent
4.2. Step 2: load combinations notional load modeling require the user to input either
imperfections or notional loads, respectively. On the
The load combinations in the proposed methods are contrary, the further reduced tangent modulus modeling
based on the LRFD load combinations [4]. The member automatically reflects the imperfection effects implicitly
sizes of structures are determined from an appropriate in the proposed computer program (PAAP) [2].
combination of factored loads.
4.6. Step 6: determination of incremental loads
1.4D
The input of incremental loads rather than total loads
1.2D + 1.6L + 0.5(Lr or S or R) is required so that advanced analysis can trace nonlinear
 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364 357

Table 2
Input data format for the program ‘PAAP’

Data set Column Variable Description

1. Title A70 Job title and general comments

2. Design control 1–5 IGEOIM Geometric imperfection method
0: No geometric imperfection (default)
1: Explicit imperfection modeling
2: Equivalent notional load
3: Further reduced tangent modulus
6–10 ILRFD Strength reduction factor ␾c = 0.85, ␾b = 0.9
0: No reduction factors considered (default)
1: Reduction factors considered
3. Job control 1–5 NNODE Total number of nodal points of the structure
6–10 NBOUND Total number of supports
11–15 NINCRE Allowable number of load increments (default = 100); at least two or three
times larger than the scaling number
4. Total number of element type 1–5 NCTYPE Number of connection types (1–10)
6–10 NFTYPE Number of frame types (1–20)
11–15 NTTYPE Number of truss types (1–10)
5. Total number of element 1–5 NUMCNT Number of connection elements (1–50)
6–10 NUMFRM Number of frame elements (1–200)
11–15 NUMTRS Number of truss elements (1–50)
6. Connection property 1–5 ICTYPE Connection type number
6–15* Mu Ultimate moment capacity of connection
16–25* Rki Initial stiffness of connection
26–35* n Shape parameter of connection
7. Frame element property 1–5 IFTYPE Frame type number
6–15* A Cross-section area
6–25* I Moment of inertia
26–35* Z Plastic section modulus
36–45* E Modulus of elasticity
46–55* FY Yield stress
55–60 IFCOL Identification of column member, IFCOL = 1 for column (default = 0)
8. Truss element property 1–5 ITYPE Truss type number
6–15* A Cross-section area
15–25* I Moment of inertia
25–35* E Modulus of elasticity
36–45* FY Yield stress
46–50 ITCOL Identification of column member, ITCOL = 1 for column (default = 0)
9. Connection element data 1–5 LCNT Connection element number
6–10 IFMCNT Frame element number containing the connection
11–15 IEND Identification of element ends containing the connection
1: Connection attached at element end i
2: Connection attached at element end j
16–20 JDCNT Connection type number
21–25 NOSMCN Number of same elements for automatic generation (default = 1)
26–30 NELINC Element number (IFMCNT) increment of automatically generated elements
(default = 1)
10. Frame element data 1–5 LFRM Frame element number
6–15* FXO Horizontal projected length; positive for i–j direction in global x-direction
16–25* FYO Vertical projected length; positive for i–j direction in global y-direction
26–30 JDFRM Frame type number
31–35 IFNODE Number of node i
36–40 JFNODE Number of node j
41–45 NOSMFE Number of same elements for automatic generation (default = 1)
46–50 NODINC Node number increment of automatically generated elements (default = 1)
11. Truss element data 1–5 LTRS Truss element number
6–15* TXO Horizontal projected length; positive for i–j direction in global x-direction
16–25* TYO Vertical projected length; positive for i–j direction in global y-direction
26–30 JDTRS Truss type number
31–35 ITNODE Number of node i
36–40 JTNODE Number of node j
51–55 NOSMTE Number of same elements for automatic generation (default = 1)
56–60 NODINC Node number increment of automatically generated elements (default = 1)

Continued over
358 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364

Table 2
Continued

Data set Column Variable Description

12. Boundary condition 1–5 NODE Node number of support

6–10 XFIX XFIX = 1 for restrained in global x-direction
11–15 YFIX YFIX = 1 for restrained in global y-direction
16–20 RFIX RFIX = 1 for restrained in rotation
21–25 NOSMBD Number of same boundary conditions for automatic generation (default = 1)
26–30 NODINC Node number increment of automatically generated supports (default = 1)
13. Incremental loads 1–5 NODE Node number where a load applied
6–15* XLOAD Incremental load in global x-direction
16–25* YLOAD Incremental load in global y-direction
26–35* RLOAD Incremental moment in global ␪-direction
36–40 NOSMLD Number of same loads for automatic generation (default = 1)
41–45 NODINC Node number increment of automatically generated loads (default = 1)

*Indicates the real value (‘F’ or ‘E’ format) to input, otherwise integer value (‘I’ format).

load–displacement behavior. The incremental loads can in the resulting load carrying capacity. The resistance
be determined by dividing the combined factored loads factors are selected as 0.85 for axial strength and 0.9 for
by the recommended scaling number of 10–50. flexural strength corresponding to AISC-LRFD specifi-
The simple incremental solution method is cation [4]. The adequacy of the structural system can
implemented here for practical use. The errors in using be directly evaluated by comparing the resulting load
the simple incremental solution method can be minim- carrying capacity ␾Rn with the applied factored loads
ized with an automatic scaling of the incremental loads, ⌺␥iQi in Step 2. Separate member capacity checks are
when changes in the element stiffness parameter exceed not necessary since the predicted ultimate load of a
a predefined tolerance in the program. structural system will not violate individual member
capacity encompassed by the AISC-LRFD equations [2].
4.7. Step 7: preparation of input data
4.10. Step 10: serviceability check
Based on the input data format in Table 2, the input
data file P.DAT can be made by users. The input data According to the ASCE Ad Hoc Committee report [8],
format is basically similar to that of the linear elastic the normally accepted range of overall drift limits for
analysis with proper considerations of geometric imper- buildings is 1/750–1/250 times the building height H
fections and incremental loads. with a typical value of H/400. The general limits on the
interstory drift are 1/500–1/200 times the story height.
4.8. Step 8: program execution Based on the studies by the Ad Hoc Committee] [8] and
Ellingwood [9], the deflection limits for girder and story
The program execution can be done simply by typing are selected as:
in a batch file command ‘RUN’ on the screen. The pro-
1. floor girder deflection for service live load: L/360,
gram continues analysis for increased loads, and stops
2. roof girder deflection: L/240,
when it reaches its ultimate state.
3. lateral drift for service wind load: H/400, and
4. interstory drift for service wind load: H/300.
4.9. Step 9: ultimate load carrying capacity check
The serviceability of a structural system should be
The proposed methods are based on the limit state checked also to ensure the adequacy of the system and
approach to strength design. The limit state format may its member stiffness at service loads. The first-order dis-
be written as: placements at service loads can be obtained by either
using the proposed analysis program or a usual commer-
⌺␥iQi ⱕ ␾Rn (6) cial program. The proposed analysis program is rec-
ommended since the input data prepared for the ultimate
where ␥i = load factors, Qi = nominal design loads, ␾ = strength checks may also be used for the serviceability
resistance factors and Rn = nominal resistances. checks without major change. Using the proposed analy-
From the output file P.OUT, the ultimate load carrying sis program, the first-order displacements can be
capacity, i.e. the ultimate stregth ␾Rn of a structure, can obtained by setting the number of load increments equal
be obtained. The resistance factors ␾ are built in the to 1 (NINCRE = 1), and then by multiplying the
analysis program so that they are automatically included resulting displacements by the scaling number. This pro-
 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364 359

cedure is based on the fact that the structural system When the yield stress is equal to 250 MPa (36 ksi),
behaves linearly and elastically under a small incremen- M1/M2 is − 0.5, and the radius of gyration is assumed
tal load. The adequacy of the system and its member to be approximately 50mm (2 inches), the permissible
stiffness can be evaluated by comparing the resulting unbraced length L results in 3.5 m (11.6 ft). Since the
displacements with the limited displacements. unbraced length of 3.5 m (11.6 ft) is within the range
of a typical story height of 3.0–3.7 m (10–12 ft), lateral
4.11. Step 11: ductility check torsional buckling is usually not a governing factor for
the ductility of beam–columns in typical building fram-
Adequate inelastic rotation capacity is required for es.
members in order to develop their full plastic moment
capacity. The required rotation capacity may be achieved 4.12. Step 12: adjustment of member sizes
when members are adequately braced and their cross-
sections are compact.
If the conditions of Steps 9–11 are not satisfied, appro-
Compact sections are capable of developing the full
priate adjustments of member sizes should be made.
plastic moment capacity Mp and sustaining large hinge
Adjustment of member sizes can be carried out referring
rotation before the onset of local buckling. The compact
to the sequence of plastic hinge formation shown in the
section in the LRFD specification is defined as
‘P.OUT’. For illustration, if the resulting load carrying
1. Flange: capacity of a structural system is less than the applied
factored loads, the member containing the first plastic
bf 65 hinge should be first replaced with a stronger member.
ⱕ (7)
2tf √Fy On the contrary, if the loading carrying capacity of a
structural system exceeds the applied factored load sig-
where bf = width of flange, tf = thickness of flange, and nificantly, members without plastic hinges may be
Fy = yield stress in ksi; and replaced with lighter members. If lateral drift of a struc-
tural system does not meet the drift requirement, beam
2. Web: and column sizes should be increased for serviceability,

冉 冊
or a braced structural system may be considered instead.
h 640 2.75Pu Pu
ⱕ 1− for ⱕ 0.125 (8a)
tw √Fy ␾bPy ␾bPy
5. Design example 1: two-bay frame
h
tw
ⱕ
冤 191
√Fy 冉
2.33 −
Pu
␾bPy冊ⱖ
253
√Fy 冥 for
Pu
␾bPy
> 0.125 Fig. 5 shows a fixed-supported two-bay portal frame
subjected to gravity and lateral loads. Detailed pro-
cedures for selecting appropriate member sizes are
(8b) presented in what follows. For comparison purposes, the
LRFD member size is also presented. All members are
where h = clear distance between flanges and tw = thick- assumed to be laterally braced, and wide flange sections
ness of web. of W16 and W14 of A36 steel are used for beams and
columns, respectively.
In addition to the compactness of section, the lateral
unbraced length of a member is also a limiting factor
for the development of the full plastic moment capacity
of members. The LRFD specification provides the limit
on spacing of braces for beam–columns as:

[3600 + 2200 (M1/M2)]ry

Lpd ⱕ (9)
Fy

where
Lpd = unbraced length of the compression flange when
using plastic analysis
ry = radius of gyration about y-axis
Fy = yield strength in ksi
M1, M2 = smaller and larger end moment
M1/M2 = positive in double curvature bending Fig. 5. Configuration and load condition of the two-bay frame.
360 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364

Fig. 6. Explicit imperfection modeling for the two-bay frame.

Fig. 8. Further reduced tangent modulus for the two-bay frame.

5.1. Design by the proposed method

5.1.1. Step 1: live load reduction

The live load reduction is assumed to be made in the each column is calculated by 0.2% times the column
load combination in Fig. 5. height of 6 m (20 ft) as shown in Fig. 6. In the equivalent
notional load method, the notional load of 1.36 kN
5.1.2. Step 2: load combinations (0.306 kips) results from 0.2% times the total gravity
Herein, the critical load combination is assumed load of 681 kN (153 kips) and is added to the lateral load
shown in Fig. 5. shown in Fig. 7. In the further reduced tangent modulus
method, the program automatically accounts for geo-
5.1.3. Step 3: preliminary member sizing metric imperfection effects shown in Fig. 8. Users can
Preliminary member sizes are selected as W16 × 77 choose one of the three modelings for generating their
and W14 × 53 for the beams and the columns, respect- input data.
ively.

5.1.4. Step 4: modeling of element 5.1.6. Step 6: determination of incremental loads

Each column is modeled as one element, and the Herein, the incremental loads are determined as
beams are modeled as three and two elements for the 8.896 kN (2 kips) by dividing the factored concentrated
left bay and right bay, respectively. The element and load of 227 kN (51 kips) by the scaling number of 25.5
node numbering are shown in Figs. 6–8. shown in Figs. 6–8.

5.1.5. Step 5: modeling of geometric imperfection 5.1.7. Step 7: input data

In the explicit imperfection modeling method, the Based on the input data format in Table 2, three kinds
geometric imperfection of 0.012 m (0.04 ft) at the top of of input data file P.DAT in Table 3 corresponding to
three different geometric imperfection methods can be
made by users, but only one input data among these three
is required for analysis and design.

5.1.8. Step 8: program execution

Program execution is simply done by typing in
‘RUN’.

5.1.9. Step 9: ultimate load check

From the output file P.OUT, the ultimate load carrying
capacities of the structure are obtained as 229 kN
(51.4 kips) by the explicit imperfection modeling, and
the equivalent notional load method, and 231 kN
(52.1 kips) by the further reduced tangent modulus
method. All three load carrying capacities are slightly
greater than the applied factored load of 227 kN
Fig. 7. Equivalent notional load modeling for the two-bay frame. (51 kips).
 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364 361

Table 3 order analysis is calculated as 1/155 which does not meet

Input data for the two-bay frame the drift limit of H/400. As a result, the member sizes
Two-bay frame, explicit imperfection modeling should be increased.
1 1
9 3 100 5.1.11. Step 11: ductility check
0 2 0 The compactness and lateral unbraced length of Beam
0 8 0
1 20.00 723.0 115.0 29 000.0 36.0
BC and Column BE are checked in what follows:
2 26.20 1300.0 175.0 29 000.0 36.0
1. check for compactness of Beam BC (W16 × 77):
1 0.48 240.0 1 1 4
2 0.48 240.0 1 2 7
3 0.48 240.0 1 3 9 bf
4 90.0 0.0 2 4 5 = 6.8
2tf
5 180.0 0.0 2 5 6
6 90.0 0.0 2 6 7 65 65
7 300.0 0.0 2 7 8 2 = = 10.8 > 6.8
1 1 1 1 3 √Fy √36
4 2.00
5 ⫺2.00 2 therefore, the flange is compact.
8 ⫺2.00
Two-bay frame, equivalent notional load modeling
2 1 h
9 3 100 = 31.2
tw
0 2 0
0 8 0
1 20.00 723.0 115.0 29 000.0 36.0 Pu 21.8
= = 0.0298 ⱕ 0.125
2 26.20 1300.0 175.0 29 000.0 36.0 ␾bPy (0.9) (22.6) (36)
1 0.00 240.0 1 1 4

冉 冊
2 0.00 240.0 1 2 7
3 0.00 240.0 1 3 9 640 2.75Pu 640
1− = (1 − (2.75) (0.0298))
4 90.0 0.0 2 4 5 √Fy ␾bPy √36
5 180.0 0.0 2 5 6
6 90.0 0.0 2 6 7 = 98 > 31.2
7 300.0 0.0 2 7 8 2
1 1 1 1 3
4 2.012 where Pu = 21.8 kips is obtained from the output file
5 ⫺2.00 2 P.OUT2. Thus, the web is compact;
8 ⫺2.00
Two-bay frame, further reduced tangent modulus modeling 2. check for unbraced length of Beam BC (W16 × 77):
3 1
9 3 100 [3600 + 2200 (M1/M2)]ry
0 2 0 Lpd =
0 8 0 Fy
1 20.00 723.0 115.0 29 000.0 36.0 1
[3600 − 2200 (2752/3648)] (2.47)
2 26.20 1300.0 175.0 29 000.0 36.0 =
1 0.00 240.0 1 1 4 36
2 0.00 240.0 1 2 7
3 0.00 0.0 1 3 9 = 133 inches ( = 11 ft)
4 90.0 0.0 2 4 5
5 180.0 0.0 2 5 6 where M1 = 2752 kip/inch and M2 = 3648 kip/inch are
6 90.0 0.0 2 6 7
7 300.0 0.0 2 7 8 2
obtained from the output file P.OUT2. Therefore, Beam
1 1 1 1 3 BC should be laterally braced at the mid-span and less
4 2.00 than a 11 ft interval;
5 ⫺2.00 2
8 ⫺2.00 3. check for compactness of Column EB (W14 × 53):

bf
= 6.1
2tf
5.1.10. Step 10: serviceability check
The lateral displacement of the first-order analysis
corresponding to the load 2 kips (NINCRE = 1) is equal 65 65
= = 10.8 > 6.8
to 1.53 mm (0.06062 inches) at Node 9. The resulting √Fy √36
displacement is computed as 40 mm (1.55 inches) ( =
0.06062 inches × 25.5). The overall drift of the first- therefore, the flange is compact.
362 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364

h
= 30.8
tw

Pu 96.6
= = 0.191 ⱖ 0.125
␾bPy (0.9) (15.6) (36)

191
√Fy 冉
2.33 −
Pu
␾bPy
=
191
√36 冊
(2.33 − 0.191) = 68 > 30.8

where Pu = 96.6 kips is obtained from the output

P.OUT2. Thus, the web is compact; and
4. check for unbraced length of Column EB (W14 × 53):

[3600 + 2200 (M1/M2)] ry

Lpd =
Fy
[3600 + 2200 (1111/2511)] (1.92)
= Fig. 10. Typical plan of the four-story, five-bay frame.
36
5.3. Comparison of results
= 244 inches ( = 20 ft)
The three proposed methods result in identical mem-
where M1 = 1111 kip/inch and M2 = 2511 kip/inch are ber sizes, and they are one or two sizes smaller than
obtained from the output P.OUT2. Since the column those determined by the LRFD method as shown in Fig.
height is equal to 20 ft, no lateral intermediate brace is 9. The proposed procedures possess the benefit of inelas-
required. tic moment redistribution that leads to a reduction of
steel weight.
5.1.12. Step 12: adjustment of member sizes
Since the lateral drift of Step 10 does not satisfy the 6. Design example 2: four-story, five-bay frame
limitation, the beam and column sizes should be
increased. Herein, the iteration procedures are not 6.1. Description of frame
presented, since they are the same as those presented
above. The AISC frame is provided in the 1991 Lecture Ser-
ies. The building has four stories with a one-story pent-
5.2. Design by the LRFD method house on the center of the top story. The story height is
3.7 m (12 ft) except for the first story of 4.3 m (14 ft)
In the LRFD method, two first-order analyses are per- and the fifth story of 3.0 m (10 ft). There are five bays
formed, one for the sway case and the other for the non- in the east–west direction with spacing of 9 m (30 ft).
sway case. The design procedure of the LRFD is usual The three bays in the north–south direction consist of
and thus is not presented herein. The resulting member the 11 m (36 ft) exterior bays and the 8.4 m (28 ft)
sizes shown in Fig. 9 are obtained from Ref. [2]. interior bay. The frame is supported by exterior moment
frames in the east–west direction and laterally braced in
the north–south direction. All other columns except
those moment frames are simply supported. The typical
plan and section of the frame are shown in Figs. 10
and 11.

Fig. 9. Comparison of member sizes of the two-bay frame. Fig. 11. Typical section of the four-story, five-bay frame.
 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364 363

6.2. Load condition

The load condition is determined by BOCA 1990 [10]

as follows:
1. roof loads
dead: 1440 N/m2 (30 psf)
live: 1000 N/m2 (21 psf)

2. floor loads
dead: 3260 N/m2 (68 psf)
live: office: 3590 N/m2 (75 psf)
lobby and penthouse: 4790 N/m2 (100 psf)

3. cladding: 720 N/m2 (15 psf)

Fig. 12. Comparison of member sizes of the four-story, five-bay
4. wind load: 130 Km/h (80 mph), Exposure C. frame.
The following three load combinations govern the
member sizes of the frame:
6.5. Lateral drift
1. 1.2D + 1.6L + 0.5S
2. 1.2D + 1.6S + 0.5L The overall drift of the first-order analysis is calcu-
3. 1.2D + 1.3W + 0.5(L + S) lated as H/366 for the wind loads of 1.0W. This does
not meet the drift limit of H/400. The maximum
where D = dead load, L = live load, S = snow load and interstory drift is equal to H/357 which satisfies the drift
W = wind load. limit of H/300. The column sizes should be increased to
The live load reduction is considered in the calculation meet the overall drift requirements. If the interior col-
of loads. umn sizes in the first and second stories are increased
from W14 × 68 to W16 × 67, the overall drift limit is
6.3. Analyses found to be H/404, which is less than the drift limit.

The moment frame with leaning columns in the east–

west direction was analyzed and shown in Fig. 11. The 7. Conclusions
yield stress was selected as 250 MPa (36 ksi). The col-
umn sizes were assumed to change every two stories. A design guide for steel frames using advanced analy-
The analysis and design processes were carried out by sis is provided. The step-by-step analysis and design pro-
the proposed three advanced methods. The geometric cedures of the practical advanced analysis methods are
imperfection was assumed to be ␺ = 1/500 and the equi- presented. The two cases studied are the two-bay frame
valent notional load was equal to 0.002⌺Pu. The further and the four-story, five-bay frame. The three practical
reduced tangent modulus was adopted as 0.85Et. modelings used are: (1) an explicit imperfection mode-
ling; (2) an equivalent notional load modeling; and (3)
6.4. Member sizes a further reduced tangent modulus modeling. Member
sizes determined by the proposed advanced analysis pro-
Most member sizes of the frame are governed by the cedures are compared with those determined by the
wind load combination except that the upper girders are LRFD procedures, and a good agreement is generally
governed by the gravity load combinations. All three observed. These comparative studies should provide
methods result in identical member sizes. The resulting further confirmation of the validity and simplicity of the
member sizes of the moment frame are compared with proposed procedures for a routine LRFD design. Since
those by the conventional LRFD method shown in Fig. the proposed procedures strike a balance between the
12 where the LRFD member sizes are taken directly requirement for realistic representation of actual
from Ref. [5]. The member sizes by the proposed pro- behavior and failure mode of a structural system and the
cedures are generally one or two sizes smaller than those requirement for simplicity in use without the calculation
by the LRFD method. This is because the AISC frame of K-factor, it is considered that in both these respects,
is highly redundant and therefore possesses much benefit all three modelings are satisfactory, but the further
of inelastic moment redistribution that leads to a reduced tangent modulus modeling appears to be the
reduction of steel weight. simplest and is therefore recommended for general use.
364 S.-E. Kim, W.-F. Chen / Engineering Structures 21 (1999) 352–364

The proposed program may be used rigorously for prac- Factor Design, Manual of Steel Construction, Vols 1 and 2, 2nd
tical design by taking advantage of the small program ed., Chicago, 1994.
[5] Maleck AE. Practical application of advanced analysis in steel
size, short run time, and simple input format. design. Structural Steel, Proceedings of the 4th Pacific Structural
Steel Conference 1995;1:119–26.
[6] ASCE 7-95 American Society of Civil Engineers Minimum
References Design Loads for Buildings and Other Structures, ASCE, New
York, 1995.
[1] Kim SE, Kim WF. Practical advanced analysis for steel frame [7] Ziemian RD, McGuire W. A method for incorporating live load
design, in Special Proceeding Volume on Analysis and Compu- reduction provisions in frame analysis. Engrg J AISC
tation, ASCE Structural Congress XIV, Chicago, 1996: pp. 19–30 1992;29(1):1–3.
[2] Chen WF, Kim SE. LRFD Steel Design Using Advanced Analy- [8] Structural serviceability: a critical appraisal and research needs.
sis, CRC Press, New York, 1997. J Struc Engrg ASCE 1986;112(12):2646–64.
[3] Liew JYR. Advanced Analysis for Frame Design, PhD thesis, [9] Ellingwood. Serviceability guidelines for steel structures. Engng
School of Civil Engineering, Purdue University, West Lafayette, J AISC 1989;26(1):1–8.
IN, 1992. [10] BOCA The BOCA (Building Officials and Code Administrations)
[4] American Institute of Steel Construction Load and Resistance National Building Code, 11th ed., BOCA, Chicago, 1990.