Beruflich Dokumente
Kultur Dokumente
aa
Printed in Great Britain. Pergamon Press plc
Abstract--With the ultimate aim of improved methods for the realistic limit state design of structures,
a method has been developed for incorporating nonlinear connection response in the analysis of
three-dimensional steel structures. The method is implemented in an interactive graphics analysis and
design program which can model both geometric and material nonhnearities in framed structures. The
connection model includes nonlinear moment-rotation response for both major- and minor-axis bending.
Standardized models for different types of connections which are calibrated to existing test data and are
amenable to design are presented. Application of the method is presented for a realistic case study
consisting of a low-rise three-dimensional structure with partially restrained connections.
frames, (b) methods of implementing semi-rigid The system described above has been used in the
connections into frame analysis, (c) validation of past to study the realistic behavior and limit state
different connection models for design, (d) suggestion design of steel structures. Through interactive
of design procedures to account for connec- computer graphics, CU-PREPF provides a user-
tion flexibility, and (e) evaluation of potential friendly and efficient means of defining the structural
benefits of considering connection flexibility in de- geometry, member properties, boundary con-
sign. Most previous research focused on two- ditions, and loadings. The analysis formulation in
dimensional analysis and in-plane moment-rotation CU-STAND includes both geometric and material
characteristics of connection; few three-dimensional nonlinearities. Geometric nonlinearities are handled
frame analysis programs which account for nonlinear through the use of geometric stiffness matrices and
connection flexibility have been reported in the an updated Lagrangian method; material non-
literature. linearities are considered through a concentrated
plasticity approach assuming elastic-perfectly plastic
1.2. Interactive computer-aided analysis and design member response based on a three-parameter yield
system surface. Various solution and analysis control
With the growing availability of lower cost schemes are provided along with features for
engineering workstations, interactive computer- graphically monitoring the structural response by
aided analysis and design systems are becoming a viewing the deflected shape of the structure, load-
viable tool for design practice. One prototype system deformation response curves, etc. An example of the
for the analysis and design of three-dimensional menuing system and type of graphical information
steel frames using finite element analysis techniques which can be displayed is shown in Fig. 1. CU-
[13-l 61has been developed at Cornell University. For STAND also includes built-in features for semi-
statically loaded structures, this system consists of automated design checks and redesign based on a
two programs, CU-PREPF and CU-STAND. CU- subset of the AISC LRFD specification [6]. One of
PREPF is a program for the definition of two- the limitations in the original version of CU-STAND
and three-dimensional framed structures; CU- (version 2.0) is that all joints were modeled
STAND is a program for the elastic and inelastic as fully rigid. In this research, the ability to in-
analysis and design of three-dimensional steel framed clude nonlinear joint flexibility is incorporated into
structures. CU-STAND.
ORtlY
HINGE5 IJNBALRNCE
INFO PRINT
I .3. Objectives and scope respect to 8, the tangent stiffness can be expressed as
the following
The objective of this research is to develop and
demonstrate a tool to analyze and design two- and
three-dimensional steel structures with semi-rigid K,=; = (4 - KJ
n (*+1)/n + %- (2)
I)
connections for static loads using finite element
analysis techniques. In this paper, the following
aspects of the work are reported:
1.3.1. Development of connection model for use This equation was chosen because it can model the
in analysis and design. Based on previous research on observed experimental data well and the four par-
the behavior and modeling of semi-rigid connection, ameters are related to the physical response of the
an analytic model suitable for the objective of connection. One advantage of this model is that it
this research is developed to represent connection easily encompasses more simple models. For
behavior for design applications. example, eqn (1) becomes a linear model if K, = KP,
1.3.2. Modeling of connections in three-dimensional an elastic-perfectly plastic model if KP = 0, and a
finite element analysis. A nonlinear connection el- bilinear model when n is large.
ement is presented which takes into account the To allow for unloading of the connections associ-
connection flexibility using a finite element approach. ated with nonproportional loading and inelastic force
A detailed formulation for the connection element is redist~bution, the two alternative unloading curves
presented. shown in Fig. 2 are provided. One alternative is based
1.3.3. Computer implementation of semi-rigid con- on linear unloading parallel to the initial stiffness, K, ,
nections. The semi-rigid connection model is im- The other is based on a nonlinear unloading curve
plemented in the interactive analysis and design given by the following equation, where the peak
system, CU-STAND. Interactive computer graphics moments and rotations reached during loading are
is utilized for defining and editing the connection h4, and 6. (see Fig. 2)
model and for assigning connection properties to the
structural model. We-K#‘,-@
1.3.4. Design example. Capabilities and features of M=iv*-
We--K&~,- 0) ’ ‘M
the analysis and design system are demonstrated
using a three-dimensional building frame. =fo I)
-$(% - 0 (3)
2. MOMENT-ROTATION MODEL FOR CONNECTIONS
2.2. Determination of parameters
As reported by previous researchers, the most
The four parameters of the model may be deter-
significant connection effects on building frame be-
mined by several means according to the specific
havior are associated with moment-rotation behav-
needs in analysis and design. If experimental data are
ior, and therefore, only major- and minor-axis
available, the most precise representation is obtained
rotational flexibility of connections is considered in
through curve-fitting the model directly to the data.
the present work. This behavior is modeled by a
Where test data are not available, as in typically the
nonlinear equation which is calibrated to test data
case in design, the parameters may be determined
and normalized for use in design.
using analytic formulations for the connection
strength and stiffness if the connection details are
2.1. moment-rotation equation known. In design practice, however, it is usually the
A four-parameter power equation is adopted to case that the connection details may not be known
represent the moment-rotation relationship of con- until after the structural members have been sized. As
nections. This nonlinear equation was first presented described below, a third method proposed for deter-
by Richard and Abbot [ 121, and was later used by mining the parameters is based on using standardized
Kishi and Chen [I l]. With this equation, the mo- curves to provide the general shape of the response
ment-rotation (M-8) relations~p of the connection curve and analytic (design) methods to calculate the
is represented by the following expression nominal connection strength. The standardized
curves are obtained from statistical analysis of nor-
malized curves which were curve-fit to experimental
M= (1) data previously collected by Kishi and Chen [lo, 1l]
and Goverdhan [St].
Curve $tting from experimental data. An optimiz-
ation approach utilizing the conjugate-gradient
As shown in Fig. 2, K, = initial stiffness of the method is used to find a set of parameters (MO, Ke,
connection, KP = strain-hardening stiffness of the KP, and n) which gives the best curve-fit to exper-
connection, MO= a reference moment, and n = a imental response data [8]. An example of the curve-
shape parameter. By differentiating eqn (1) with fitting results for top- and seat-angle connections with
998 S.-H. Hsizu and G. G. DEIERL.EIN
o Test dala
- Curve-f i1
F I 1 4
10.0 20.0 30.0 40.0 SD.0
Relallve rolatlon (111000 red. 1
Fig. 3. Comparison between curve-fitting results and experimental results for TSAW connections.
3-D steel frames with semi-rigid connections 999
shown in Fig. 3. The curves were then normalized by 3. FINITE ELEMENT ANALYSIS OF
a value of M, equal to the moment resisted at an SEMI-RIGID FRAMES
1.4
1.2
Honenl rallo
I
1.4
+ llonenl rallo
1 I
I
I
I
I
1 .I
4 1
I
I
I
I
0
8
I
I
1 1
5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0
Relative rotallon (111000 rab.)
load step. For inelastic analyses, material nonlineari- displacements; [k,], [kg], and [k,] are elastic, geo-
ties are included through the use of element plastic metric, and plastic reduction matrices, respectively;
reduction matrices [k,,] which are based on a three- and [k,] is the resulting element tangent stiffness
parameter yield surface for modeling cross-section matrix. Depending on the type of analysis (e.g.,
plastification due to axial load and major- and minor- first-order or second-order, elastic or inelastic), [kg]
axis bending. This is a concentrated plasticity model and/or [k,] may not be included in the analysis.
approach where it is assumed that zero-length plastic The global incremental equilibrium equations are
hinges form at the end of each element. Details of the written as the following
stiffness matrices [k,], [kg], and [k,] are provided in
1161. W) = KIWI. (5)
The incremental element equilibrium equations can
be written in the following form In eqn (5), (dP} is the incremental load vector
applied on the entire structure, {dU} is the
IdsI = We1+ [&I+ tk,lW~ = [k,lIduI. (4) global incremental displacement vector, and [K,]
is the global stiffness matrix obtained by assembling
In eqn (4), {ds} is the vector of incremental element the transformed element tangent stiffness matrices,
end forces; {du} is the vector of incremental element kl.
3.2. Modeling of semi-rigid connections tion or for including additional connection degrees-
of-freedom). On the other hand, a ~~dv~~ge of
Zero length connection elements are used to permit this approach is that it increases the total number of
relative flexural rotations between connected mem- degrees-of-freedom in the global system of equations.
bers; the connections do not allow for relative tor- However, this disadvantage is becoming less signifi-
sional rotation or translational displacements. When cant with the continuing improvement of computer
a semi-rigid connection is specified at one end of a hardware.
member, the global rotational degrees-of-freedom at Example formulation. An example is presented to
the corresponding structural node are associated with demonstrate the formulation and transformation of
the connection element. The corresponding local the beam-column element stiffnesses with nonlinear
rotational degrees-of-freedom between the member connection ~exibility. A portion of a three-dimen-
end and connection are treated as additional sional structure discretized in a global coordinate
global unknowns of the structural system and are system with orthogonal axes X, Y and Z is shown in
included in the global equilibrium equations [i.e., Fig. 6. The connection elements Aa and Bb connect
eqn (5)]. Condensation is not used here because it the beam ab to the nodes A and B. For clarity, the
is not as efficient for nonlinear analyses in which connection elements are shown ‘exploded’ to a finite
stifiness matrices are updated many times. In length but are actually zero-length. The local (or
the formulation, the additional rotational degrees- element) coordinate system of the beam ab with
of-freedom described above are always measured orthogonal axes x, y and z is also shown. In general,
with respect to the local member coordinates even the local coordinate axis (x, y, z) are arbitrarily ori-
though they are treated as global unknowns in ented with respect to the global coordinate axes
eqn (5). X Y, Z).
To introduce the local degrees-of-fr~dom into In Fig. 6, the degr~s-of-fr~dom {DOFs) at
the global solution system, the conventional nodes A and B are represented using the displace-
element transformation matrices are modified for ment components in the global coordinates, and
the elements with semi-rigid connections. The the DOFs of beam ab are shown in terms of
modified matrices are used to transform the local coordinates at both ends of the beam. In the
element stiffness matrices from the local to the displacement components, AV and (&, the first
global coordinate system with some of the un- subscript i refers to the global or local com-
knowns retaining their local coordinate reference ponent axes and the second subscript j refers to
axes. the global or member node designation. These
A key advantage of this approach is that the subscripts are also used in developing the ex-
existing nonlinear formulation for the beam-column pressions for the forces and moments, 41 and M,,
elements is unaffected. This avoids the difficulty respectively.
(particularly for three-dimensional problems) of di- In the example considered, it is assumed that
rectly formulating the connection flexibility into the the connection element Aa includes both major-axis
nonlinear element stiffness matrices, [Jr,] and [k,]. The and minor-axis rotational flexibility while the connec-
use of separate connection elements also facilitates tion Bb is rigid. Aside from major- and minor-axis
further modifications to the connection model (for flexure, no other deformations are considered in the
example, to account for the finite size of the connec- connection Aa.
Beam ab
Matrix notations. The following matrix notations Since, in this particular case, no connection flexi-
are used for the modeling formulation bility is considered at end b of the beam, the displace-
ment and force transformations are aiven in the
[y13X3= the conventional rotation matrix of beam following standard forms
ab
=Pl
r1 AXB
ArB
(9)
I J
eYA
e ZA
PI I2 x 14 =
[Gll,x,
0
0
[G’&., 1 (11)
’
AXA
A YA
-I
F“,l,., 0 AZA In eqn (15) Kbh X14= Kblf, X12M12 XdrbL2 x l4r
e XA
[k,] = element tangent stiffness matrix of the beam ab
[eqn (4)], and {Fl}, {F2}, {D 1}, and {02} are defined
e YA
in eqns (7x10). Note that e,, and oz, in {Dl} are
0 PI2x2 e ZA local rotations at end a of the beam and Mya and M,
e
N
in {F 1) are the corresponding member end moments
9 in
measured in local coordinates.
Stiffness matrix of connection Aa. Given K, = the
(7) tangent stiffness of connection Aa for major-axis
(z-axis) bending, KY = the tangent stiffness for minor-
Similarly, equilibrium gives the following transform- axis (y-axis) bending, and the relative connection
ation of forces from the local to the mixed global rotations are (6iA-6lo) and (fJ,,-e,,), respectively, the
system equilibrium equations for connection element Aa are
FXA given as the following
F YA
‘[T,l,.e 0 F ZA
MXA
MYA
0 m2x MzA
eYA
Mya e
M,, =k14x4 e:,: . (13)
ii
=[G%.,Wh.,. (8) I8 TO
J
3-D steal framea with semi-rigid connections 1003
AX.4
1
A YA
8 A Extensionfor other situations. The transformation
uA =tT*l*Yb e; (14)
11e L4 procedure demonstrated above in eqns (6)+7) can
e be modified to cover connections with only major-
e; or minor-axis rotational flexibility and for beam-
column elements with connections at both ends.
All of these cases are included in the computer
implementation in CU-STAND.
Note that the first three columns of [T,] are null for
the present case of i~tesim~ joint size and non- 3.3. Computer ~ple~~t~tio~
eccentric member ends. In addition to implementing the connection model
Adding 13~~and 6, to eqn (14), and using a similar noted above, control menu’s were added in CU-
transformation for forces, the displacement and force STAND for definition of the connection parameters
tr~fo~ations at the connation are given by eqns and to locate connations by graphically attaching
(15) and (16) them to specific members. One of the connection
editor menus is shown in Fig. 7 through which the
AXA user can interactively assign the four parameters
which define the shape of the connection model and
A YA
the nominal connection strength, M,. Based on the
F212xt5 AZA user input, a plot of the moment rotation curve for
0 XA the connection is shown in the viewport in the upper
B YA left portion of the screen. In the program, the four
connection parameters can be either specified directly
8 ZA
2 by the user, chosen from a library of values for
~.
8Ye standard connection types, or generated from mo-
e II? ment-rotation data using the built-in curve-fitting
routine. The nominal connection strength, M, can be
(15) specified either as an absolute value or as some
fraction of the plastic moment of the member, Mti,
to which the connection is attached. The latter option
is particularly suited to an iterative design process
FXA where connection and members properties are un-
!
FYA
known at the outset and updated in the course of
design.
FZA
MXA 4. APPLICATION TO DESIGN
MYA
A case study is presented to demonstrate appli-
MZA
cation of the compu~r-~ded analysis system for
MY# investigating the design and behavior of a low-rise
Ma, three-dimensional steel frame with partially re-
strained connections. The frame geometry, member
(16) sizes, and loading am shown in Fig. 8. Partially
restrained TSAW (top- and seat-angle with double
web angles) beam-column connections are used for
Using the transformation matrix [fc] from eqns framing in both directions and the column base plate
(IS) and (16), the equilibrium equations for connec- details are modeled as EEPS (extended end plate with
tion Aa in the global system are given as the follow- stiffeners) connections. The member sixes shown were
ing, where Kclr Xr = V’CIS’~ #cl, x,V’cl~x8 and designed based on the AISC-LRFD specification
[kc] = element stiffness matrix of connection AP provisions with members forces calculated by the
measured in local coordinates nonlinear analysis described below. In the analysis
CA.5 41/s-I
1004 S.-H. Hsrtxand G. G. DEJERLEIN
model, beams were discretized into four elements and assumed above (i.e., ML = Mcd= 0.65 M,), The re-
gravity loads were applied at the quarter points of duction factor of 0.65 was chosen based on a re-
beams in the Noah-South (N-S) direction to ap- liability analysis described in (171. As will be
proximate loads applied through the one-way joist demonstrated, the reduced value of the connection
floor system. In this case, the strength limit state for strength provides a conservative estimate of the over-
the frame was controlled by the factored gravity load all strength and stiffness of the frame. Based on this
combination, 1.2 DL + 1.6 LL. approach, the following values of Ml,, were used in
In the analysis, all beam-column connections and the analysis: TSAW (N-S), M;,, = 0.26 Mph; TSAW
the column base plate connections were modeled (E-w), M:, = 0.56 Mph; and EEPS column base,
using the procedure described previously. The TSAW M& = 1.OMph. For the subsequent inelastic analyses
beam-column connections were modeled using the described below, resistance (reduction) factors were
set of normalized average parameters presented in also applied to the member yield surfaces to reflect
Table 1 (i.e., TSAW-Ave). The connection design the design, rather than the nominal, member cross-
strength was assigned as a fraction of the plastic section strengths (e.g., the design plastic moment was
moment capacity of the connected members. For taken as 0.9 M,,). In this way, resistance factors are
framing in the N-S direction which carried most of included in the inelastic analyses based on an ap-
the gravity load, the nominal connection strength, proach suggested by Ziemian et al. 1191which follows
M,, was taken as 40% of the plastic moment ca- the philosophy of the AISC-LRFD specification.
pacity of the connected member (i.e., 0.4 M,,), and Once the member sizes were selected based on the
for framing in the E-W direction, M, was taken as analysis and design assumptions described above,
0.8 Mph. The column base plate connections were additional nonlinear (second-order inelastic) analyses
modeled as EEPS connections connections with were made to investigate the limit strength of the
M En= 1.5 Mph. The normalized parameters for the frame and its sensitivity to the assumed connection
EEPS connections are equal to the following values properties. The connection models used in the five
based on [17]: k:= 185, Kj = 1.6, MA= 1.05 and cases are summarized in Table 2. In case A, which is
n = 1.54. the basic case, the connection properties are the same
For strength design purposes, a reduced (design) as those used in the o~~na1 analysis used to deter-
connection strength, MC,,, is used in the analysis mine the member sizes. In cases B and C, the
which is equal to 0.65 times the nominal strengths parameters are modified to reflect the upper and
IIII [U-STAND
1 EDITOR
“’
1..
, ‘S’
“0’
w
b 81 83
11 H-t i -t
b
s 61 83 83
65 UYP) I
a )(I H-t -I---
b 82 84 84
@ ‘!I H-t i- t
L
North-South F’raming
A
I
IB
TSAW-Ave
TSAW-Upper
0.26 Mpb
0.26 Mpb
TSAW-Ave
TSAW-Upper
0.52 Mph
0.52 Mpb
EEPS-Ave
EEPS-Upper
1.0 Mpb
1.0 Mpb
‘0.2
II :
I.0
:
2.0
;
3.0
;
4.0
;
5.0
;
6.0 1.0
IO.0
of the full factored gravity load combination. For
example, for case A the drift at the full factored load
(applied load ratio = 1.O) was approximately 0.6 in.
which corresponds to a drift index of H/767 where H
is the building height. However, at the limit point the
drift for this frame equaled 8.1 inches which corre-
Fig. 9. Load-deformation response under factored gravity sponds to a drift index of H/57. The large drifts at the
loads. limit point demonstrate that even for low rise build-
ings, where partially restrained connections are used,
lower bound connection response curves shown pre- second order (geometrically nonlinear) effects should
viously in Fig. 5. In case D, the unreduced nominal be considered.
connection strength (i.e., without the reduction factor The location of plastic hinges at the inelastic limit
of 0.65) is used in the analysis. Note that the differ- points for the partially restrained frame (case A) is
ences between cases A, B, and C is mainly with regard shown in Fig. 10. In general, all of the partially
to the shape of the response curve, whereas in case D restrained frames (cases A-D) had a similar pattern
the strength is varied. Finally, in case E, all connec- of hinges. As indicated in Fig. 9 and Table 3, the first
tions are treated as perfectly rigid. As noted pre- hinges formed at 94-105% of the full factored load;
viously, the design strength was governed by the this is expected since the original AISC-LRFD design
gravity load combination and the limit state analyses for case A was based on the most critical members
were conducted for this loading. reaching their plastic limit strength at an applied load
The overall load-deformation behavior for each ratio of approximately 1.0. As shown in the inset of
case is shown in Fig. 9 where the applied load ratio Fig. 9, for the partially restrained frames, the first
is plotted versus the lateral roof drift in the N-S hinges occurred due to large minor-axis bending
direction, Note that since the frame is not symmetric moments in the top story columns along column line
in the N-S direction, drift occurred under the gravity A (columns A.2-A.5). In the rigid frame, the first
Table 3. Applied load ratios from inelastic analysis for gravity loads
A B C D E
A B C D E
A B C D E
A B C D
hinges formed at approximately the same applied Differences between the connection response par-
load ratio but at different locations; in case E the first ameters caused a large variation in calculated lateral
hinges formed due to negative bending at the interior drifts at full factored loads, but relatively little differ-
ends of beams B4 adjacent to column line C. ences in drifts at the limit points. Also, the differences
In general, the difference in the inelastic limit in parameters had little influence on the calculated
strength due to variations in the normalized connec- vertical beam deflections. As reported in Table 4,
tion parameters (K:; K;, Mh and n) was not signifi- under the full factored loads, the calculated roof drift
cant. As reported in Table 3, the applied load ratios for case B was 67% greater and for case C was 26%
at the inelastic limit for cases B and C were within less than for case A. The 50% increase in connection
1.6% of case A. By comparing cases A and D, it was strength in case D decreased the drift by 23% com-
evident, however, that the connection strength used pared to case A, and for the rigid frame (case E) the
in the analysis had a greater effect on the response drift was 69% less than case A. At the limit point,
where a 50% increase in connection strength (l/0.65) however, the drifts for cases A, B, and C were all
increased the overall system strength by approxi- within lo%, and the drift for case D was 15% less
mately 5%. While this is not a major increase in than case A. Overall, the rigid frame experienced
strength, it is somewhat significant if considering much less drift and at the limit point the drift was
that, for the case with perfectly rigid connections 90% less than that of the partially restrained frames.
(case E) the total strength increase was only 10% As shown in Table 5, the beam deflections for cases
compared to case A. B and C were all within 3-8% of those for case A, and
Fig. 10. Hinge locations for partially restrained frame at the inelastic limit point under gravity loads.
1008 S.-H. HSIEHand G. G. DHERLEIN
the deflections for case D were ll-13% less than for evaluating the realistic performance of partially
those for case A. Beam deflections in the rigid frame restrained frames.
were 50-75% less than those for case A.
The maximum connection rotations for each case Acknowledgements-The work reported in this paper was
are reported in Table 6. As shown, at the full factored supported by the National Center for Earthquake Engin-
eering Research (Project 88-1005A and 87-1008) and the
load the maximum rotations (roughly 0.017-0.018 National Science Foundation (Grant MM-8908870). The
radian for cases A-C and 0.014 for case D) were less contributions of faculty, staff, and student colleagues are
than 0.020 radian. This lends validity to the analysis gratefully acknowledged, in particular the advice of Pro-
since the nominal connection capacity was defined as fessors J. F. Abel and W. McGuire of Cornell University
the moment resisted at a rotation of 0.020 radian. At and Professor T. W. Lin of National Taiwan University,
and the many individuals who contributed to the develop-
the limit point, the maximum rotations were between ment of the computer programs utilized in this work.
0.043 and 0.051 radian. In general, these rotations Finally, the assistance of Professors D. W. White and
seem to be in a permissible range based on the W. Chen of Purdue University who provided the data bank
deformation capacities reported from tests. It should of connection tests used in this work is appreciated. Any
opinions, findings, and conclusions or recommendations
be noted that in other examples which the authors expressed in this paper are those of the authors and do not
have studied, connection rotations calculated in the necessarily reflect the views of the sponsors.
analysis were sometimes in excess of 0.050 radian,
particularly where local beam mechanisms form prior
to reaching the inelastic limit point.
REFERENCES
14. G. G. Deierlein, W. McGuire, J. F. Abel and Report 89-13, School of Civil and Environmental
S. Srivastav, Some interactive graphics and parallel pro- Engineering, Cornell University, Ithaca, NY (1989).
cessing for earthquake engineering. In Computer Utiliz- 17. Y. J. Shen, Computer aided design of steel frames with
ation in Structural Engineering (Edited by J. K. Nelson). partially restrained connections. M.S. thesis, School of
ASCE Structures Congress, San Francisco (1989). Civil and Environmental Engineering, Cornell Univer-
15. W. McGuire, G. G. Deierlein, T. K. Sooi and Y. T. sity, Ithaca, NY (1990).
Zhao, Illustrated Primer to CU-PREPF, CU-STAND 18. W. McGuire and R. H. Gallagher, Matrix Structural
and CU-QUAND. Struct. Engrg. Report 89-12, School Analysis. John Wiley NY (1979).
of Civil and Environmental Engineering, Cornell Uni- 19. R. Ziemian, D. W. White, G. G. Deierlein and
versity, Ithaca, NY (1989). W. McGuire, One approach to inelastic analysis and
16. S. H. Hsieh, G. G. Deierlein, W. McGuire and J. F. design. Proceedings of 1990 National Steel Construction
Abel, Technical manual for C&STAND. Struct. Engrg. Conference, AISC, Chicago, 19-1 to 19-19 (1990).