Sei sulla pagina 1di 21

SECOND-ORDER INELASTIC ANALYSIS M E T H O D S FOR

STEEL-FRAME DESIGN
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

By W. S. King, 1 D. W. White, 2 Associate Member, ASCE, and


W. F. Chen, 3 Member, ASCE

ABSTRACT: TWO simplified methods for second-order inelastic analysis of steel


frames, termed the modified plastic-hinge and the beam-column strength ap-
proaches, are presented. These analysis models are comparable to elastic-plastic-
hinge analysis in efficiency, and yet, they alleviate the problems associated with
overprediction of stiffness and strength by the usual elastic-plastic-hinge analysis
methods. The modified plastic-hinge method is based on simple modifications to
the elastic-plastic-hinge model, which account for the degradation in stiffness as
the cross-section strength is approached at critical locations along the member
length. The beam-column strength method is similar to the modified plastic-hinge
approach, except that in this model, equations for the strength of the overall
member are employed rather than expressions for the cross-section strength. The
performance of these analysis models is contrasted with results from elastie-plastic-
hinge and refined plastic-zone analysis solutions as well as results from experimental
tests. Recommendations are provided with regard to the acceptability of simplified
analysis models for use in predicting inelastic load-deflection behavior and maxi-
mum strength.

INTRODUCTION

There are a large number of possible approaches for second-order inelastic


analysis of steel frames. In general terms, these analysis methods can be
categorized into two types: plastic-hinge analysis and plastic-zone analysis.
One of the most commonly employed forms of plastic-hinge analysis is the
elastic-plastic hinge approach. Second-order elastic-plastic hinge analysis is
based on the assumptions that: (1) The members are elastic until the full
cross-section plastic strength is achieved at some location along the member
length; (2) the cross-section behavior is perfectly plastic (no strain hard-
ening) once yielding occurs; (3) the inelastic behavior is contained within
zero-length plastic hinges; and (4) the members are elastic between hinge
locations.
In contrast, second-order plastic-zone analysis explicitly follows the grad-
ual spread of yielding throughout the volume of the structure. The distri-
bution of plasticity is modeled by discretization of the members into a
number of elements as well as subdivision of the cross sections into a number
of fibers (Alvarez and Birnstiel 1969; Chu and Pabarcius 1964; White 1985).
Although plastic-zone analysis can predict accurately the inelastic structural
performance, and is generally considered to be the exact solution, it is quite
computationally intensive since a fine discretization is needed generally both
through the cross section and along the member length. Therefore, its use

'Assoc. Prof, of Civ. Engrg., Chung-Cheng Inst, of Tech., Taipei, Taiwan, R O C .


2
Asst. Prof, of Struct. Engrg., School of Civ. Engrg., Purdue Univ., West
Lafayette, IN 47907.
3
Prof. and Head of Struct. Engrg., School of Civ. Engrg., Purdue Univ., West
Lafayette, IN.
Note. Discussion open until July 1, 1992. To extend the closing date one month,
a written request must be filed with the A S C E Manager of Journals. The manuscript
for this paper was submitted for review and possible publication on July 20, 1990.
This paper is part of the Journal ofStructural Engineering, Vol. 118, No. 2, February,
1992. ©ASCE, ISSN 0733-9445/92/0002-0408/S1.00 + $.15 per page. Paper N o . 136.

408

J. Struct. Eng. 1992.118:408-428.


to date has been primarily for research projects, checks of the accuracy of
simplified methods, checks of experimental results, and establishing design
charts for practical purposes ("Ultimate" 1984).
Second-order elastic-plastic-hinge analysis is more practical than plastic-
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

zone methods for design use. Often, only one element per member may be
employed for this type of analysis (at least for members subjected only to
end loads), and no discretization of the cross section is required. However,
since elastic-plastic-hinge analysis only models yielding at zero-length plastic
hinges, this type of analysis does not accurately represent the degradation
in the flexural stiffness due to distributed yielding and the associated second-
order inelastic P — b effects within individual members. The member stiff-
ness between plastic hinges is modeled as elastic. Therefore, the elastic-
plastic-hinge method tends to overestimate the second-order stiffness and
strength of structures. The load-deflection curve predicted by this type of
analysis is not smooth since the plastic hinges form abruptly, i.e., the yielded
cross sections are assumed to change abruptly from the fully elastic to the
fully plastic state.
In this paper, two methods, which are comparable to the elastic-plastic-
hinge approach in efficiency and simplicity, but which do not have the
aforementioned limitations, are developed and contrasted. These rnethods
are termed the modified plastic-hinge and the beam-column strength ap-
proaches. The goal in the development of these models is to alleviate the
problems associated with the overprediction of stiffness and strength by
usual elastic-plastic-hinge methods. The present paper focuses on the anal-
ysis of frames with members of compact cross-section type, loaded in strong-
axis bending in the plane of the frame, and adequately braced in the out-
of-plane direction to prevent out-of-plane failure prior to reaching the frame
capacity.
The modified plastic-hinge method is based on simple modifications to
the basic elastic-plastic-hinge model, which account for the degradation of
member stiffness as the cross-section strength is approached at critical lo-
cations along the member length. The beam-column strength approach is
similar to the modified plastic-hinge approach, except that in this model,
equations for the strength of the entire member are employed rather than
expressions for the member cross-section strength. The member stiffness is
assumed to degrade as the overall member strength is approached. To
present the development of these two methods, first it is important to outline
the member and cross-section strength equations to be employed by the
analysis models. Also, ah equation for the onset of significant yielding must
be defined. These aspects are discussed in the next two sections.

BEAM-COLUMN STRENGTH EQUATIONS

For design of steel beam columns, the present American Institute of Steel
Construction-load and resistance factor design (AISC-LRFD) bilinear in-
teraction equations (Manual 1986) are accurate for most cases and are simple
to use, but they have some weaknesses. These equations utilize the same
functional relationships for both strong- and weak-axis strengths. As a result,
the AISC-LRFD bilinear equations are generally conservative for short
beam columns (Llry less than about 40) when weak-axis bending and axial
force are dominant. This is particularly true for beam-column members
subjected to double-curvature bending. Also, the AISC-LRFD bilinear
equations are generally conservative for biaxial loading cases. Furthermore,
409

J. Struct. Eng. 1992.118:408-428.


two separate equations are required to express the beam-column strength
as a bilinear relationship.
A unified design-interaction equation for members of class 1 cross-section
type has been proposed recently by Duan and Chen (1989a). This unified
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

equation is a single function and is applicable to all combinations of axial


force, uniaxial or biaxial bending] and specified column-slenderness ratios
(KL/r)x and (KLIr)y. In the three-dimensional force space associated with
axial force P and bending moments Mux and Muy, this equation defines a
continuous strength surface that has a vertex at its intersection with the
F-axis and a discontinuity in its gradient at its intersection with the
Mux-Muy plane at P = 0.
The Duan and Chen equation is adopted in this paper for use in the
analysis of frames subjected to strong-axis in-plane loading. For uniaxial
bending about the strong axis (i.e., the x-x axis), this equation has the form

PV M
f) wr10
+ (1>
where p = 1.3 + 0.002 {KLIr)x; Pu = the factored axial force in the beam
column; P„ = the strength of the member as a column; Mu = the maximum
second-order elastic moment within the member length; Mp = the cross-
section plastic moment capacity = FyZx; Fy = yield stress; andZx = plastic
section modulus for strong-axis bending.
The exponent p accounts for the effects of member slenderness on the
shape of the interaction curve.
If P„ is evaluated based on the AISC-LRFD column-strength expressions,
(1) gives results that are comparable to the AISC-LRFD bilinear interaction
curves for strong-axis bending cases (Duan and Chen 1989a). In the AISC-
LRFD specification, the column strength Pn is based on a single column
curve that is representative of the strength of columns with initial residual
stresses and geometric imperfections. However, a number of the numerical
examples in this paper involve comparisons to refined analytical/numerical
solutions in which the effects of residual stresses are included but the initial
geometry is assumed to be perfect. For comparison to these solutions (and
for comparison to experimental results in which the initial geometric im-
perfections are small), the Column Research Council (CRC) column-strength
expressions (Guide 1988) are more appropriate. These equations may be
written in the form

P„ = (1 - Q.25\l)Py for Xc < V2 (2a)

P„ = § for kc > \/2 (26)

(3)

These expressions may be written equivalently in terms of a column-tangent


modulus E, as
P =^Msk (4)

410

J. Struct. Eng. 1992.118:408-428.


where

for >0 5 (5fl)


*<-AETX-T) ^ ' •
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

E, = E for — < 0.5 (5b)

CROSS-SECTION STRENGTH AND INITIAL YIELD SURFACES

A cross-section strength expression is needed for any type of plastic-hinge


analysis. If Llrx is taken as zero, (1) reduces to the cross-section strength
formula

f '•£- (6)

where Py = FyA = the member squash load; and P and M = the second-
order axial force and strong-axis bending moment at the cross section being
considered. As shown by the solid line curve in Fig. 1, this equation is
convex and continuous within each quadrant of the axial force-bending
moment plane. In (Duan and Chen 1989b), this equation is shown to be in
good agreement with exact results for the cross-section strength of both
light and heavy wide-flange sections. Also, several forms of this equation
are provided in this reference for other cross-section types, including rec-
tangular and circular tubes.
In general, initial yielding within a member cross section occurs long
before the full plastic cross-section strength is reached, the extent of the
difference between initial yielding and the attainment of the cross-section

\ ^ x
0.8
M
\ = 1.0
(*)"* Mp
0.6 >
\ N.

0.4
P fM
h = 1.0\
p
y ~M7
0.2

(in i i \ \
0.0 0.2 0.4 0.6 0.8
M/Mn

FIG. 1. Cross-Section Strength Surface and Initial Yield Surface Neglecting Re-
sidual Stress Effects

411

J. Struct. Eng. 1992.118:408-428.


strength depending on the shape factor of the cross section and on the cross-
section residual stresses. If residual stress effects are neglected, an initial
yield surface may be defined as (see the dotted line in Fig. 1)
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

fM = 1.0
+ (7)
Pv M,
where / = the shape factor of the cross section.
An initial yield surface that accounts approximately for the effect of
residual stresses is shown with the cross-section limit surface in Fig. 2. The
equation for this surface is
fM
0.8P, + 0.9Mn = 1.0 (8)

An exact equation for initial yielding within any doubly symmetric cross
section can be obtained simply by subtracting the fraction (Fres)maxIFy from
the right-hand side of (7), where (Fres)max is the maximum residual stress at
the extreme fibers of the cross section. However, (8) is more appropriate
for use with the modified plastic-hinge method discussed in the next section.
This equation represents roughly the point at which the yielding within a
typical wide-flange cross section becomes significant.

0.0 0.2 0.6 0.8


M/Mn

FIG. 2. Cross-Section Strength Surface and Initial Yield Surface that Represents
Onset of Significant Yielding

M^ e, M• J| ,• e",j
FIG. 3. Two-Degree-of-Freedom Frame Element
412

J. Struct. Eng. 1992.118:408-428.


MODIFIED PLASTIC-HINGE APPROACH

The motivation for the modified plastic-hinge method stems from the fact
that in many cases, the elastic-plastic hinge model represents the structural
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

performance quite well (El-Zanaty et al. 1980; White et al. 1991; Ziemian
1990). Therefore, if this model is adjusted such that the member stiffness
is reduced gradually as the cross-section plastic strength is approached, the
resulting analysis model may be sufficient for general prediction of second-
order inelastic behavior.

Two-Degree-of-Freedom Plane Frame Element


A beam-column element that contains only two end-rotational degrees
of freedom is shown in Fig. 3. The incremental stiffness relationships for
this element may be written as

8M, Ku Kv
8M, sS;j w
In the present work, the first three terms in the Taylor series expansion of
the elastic stability functions are retained for computation of stiffness coef-
ficients. For an elastic beam column with compressive axial load, the re-
sulting terms in (9) are (Goto and Chen 1987):
AEI 2PL 44P2L3
Ka = K„
V/ = —L + - 15
r r + 25,000£7
2 26p2L3
K =K = M_lk„ (m
y K
" ' L 30 25,000£/ '
where P is taken as negative for compression. In the modified plastic-hinge
approach, the element stiffness is assumed to degrade as described in the
following after (8) has been breached by the element end forces. When the
full plastic cross-section strength is reached, the section at the plastic hinge
is modeled as a real hinge with a constant moment, Mpc, and the incremental
equations are adjusted to account for the effect of the hinge on the member
stiffness. The term Mpc is the plastic moment capacity as reduced by the
effect of axial force. This term may be computed by solving (6) for M based
on the value of P at the hinge location.

Plastic Hinge at Node i Only


If a plastic hinge has formed at node i of element ij, the element moment-
rotation relationships may be written as
8M,- = KitbQ, + ^ 8 6 ; = 0 (11)
based on the assumption of zero change in the moment at the plastic hinge,
and
hMj = KjibQt + K,fiQ, (12)
Solving (11) for 88, and substituting into (12) gives

8My = CKff - K,Mtoj (13)

413

J. Struct. Eng. 1992.118:408-428.


and the corresponding element stiffness equations become

(SM\ 0 0 (be, (14)


Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

\?MJ L° K'A {be,

K'H = Kt, - K *2 (15)


11 i)

To represent a gradual transition from the elastic stiffness at the onset of


yielding to the inelastic stiffness associated with a full plastic hinge at node
i, the element stiffness matrix can be written as

(KB - KM) (Kif - Ktf,,)


K' = (16)
[_(*„ - Kjfil) IK, - K^Kg/Kufc

where p, = a scalar parameter that varies between zero and one.


If this equation is employed to describe the elastic-plastic hinge approach,
it can be stated that the parameter p,- jumps abruptly from 0 to 1 when a
plastic hinge forms at node i. In other words, for M, less than Mpc, p, is
equal to zero, and for M, equal to Mpc, p, is equal to one. Correspondingly,
(16) reduces to the stiffness matrix given by (9) for p, = 0, and to (14) for
P, = 1.
In the proposed modified plastic-hinge method, the parameter p,- is defined
as

o * P/ = M> - M^ 1 (17)
Mpc - Myc
where M,- = the moment at node i; and Myc = the value of M obtained
from (8) for the current value of P (i.e., the moment at the onset of yielding
associated with the current value of P, as shown in Fig. 2). In this approach,
p ; is equal to zero and the stiffness is fully elastic when M, is less than Myc.
However, when M, is equal to or greater than Myc, the stiffness reduces
linearly as Mt approaches Mpc. At the limit of full cross-section plastification,
the stiffness matrix reduces to the extreme case of (14). The yield moment
Myc is taken as zero when PIPy is greater than or equal to 0.8. This simple
linear relationship for the stiffness degradation results in significant im-
provement in analysis predictions for frames that have beam columns with
slenderness ratios KLIr roughly between 50 and 100 and/or axial loads roughly
greater than about 0.4F,,. Relationships similar to (14)—(16) are obtained
for the case where a plastic hinge forms only at end /.

Plastic Hinges at Nodes i and j


For the general case in which plastic hinges may form at both ends of an
element, the element stiffness matrix may be written in the form

bM, K'a K'ij 86,-


(18)
bM, K
ii K
'n. 89,

where

414

J. Struct. Eng. 1992.118:408-428.


K'u = U„ - K„ j * p,j(l - p,) (19a)
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

K'v = K„ = KV{1 - p,)(l - P;) .'. (196)

K^(K„- Kj,j* p,)(l - Py) : (19c)

Eqs. (18) and (19) account for the effect of plastification at both ends of
the member. For example:

1. When p,- = py = 0, both ends are in a fully elastic state. Eq. (18) reduces
to (9) with the stiffness coefficients defined by (10).
2. When p, = 1 and p; = 0, a plastic hinge has formed at node i, but node
/is still elastic. Eq. (18) reduces to (14).
3. When p,- = 0 and p;- = 1, a plastic hinge has formed at node /, but node
i is still elastic. Eq. (18) then reduces to a form similar to (14).
4. When p, = 1 and p; = 1, plastic hinges are formed at both ends of the
member, and the stiffness matrix in (18) becomes

0 0 (20)
K' = 0 0
5. When 0 < p, < 1, and 0 < p; < 1, partial plastification within the
member cross section is accounted for at both ends of the member.

Six-Degree-of-Freedom Plane Frame Element


By making use of the fundamental kinematic and equilibrium relationships
for a beam-column element (Chen and Lui 1987), (18) can be transformed
to the following stiffness relationships of a six-degree-of-freedom plane frame
element:

S
u

V
n/
V
AE -AE
0 0 0 0
L L
(K;, + 2K;i + K^ p (K'u + K$ - ( K i + 2if;+iCJ) m+K^
0 o
L2 L L L2 L
(Kh + K'n) -(Kk + Kli)
0 K'i, 0 Kit
L L
-AE AE
0 0 0 0
L L
-{K't + lKlj + Kfi P -(KJ+KJ) -(K'a + Kji)
0 0
L1 L L L2 L
W + K'll) -(Kli + K'„)
0 K'i, 0 K/j
L L

415

J. Struct. Eng. 1992.118:408-428.


J I (2D
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

\d6,

where P is taken as negative for compression.

BEAM-COLUMN STRENGTH APPROACH

The modified plastic-hinge approach models the overall member strength


indirectly based on cross-section strength. However, use of the cross-section
strength by itself is not, in general, sufficient to represent the actual second-
order inelastic behavior of a beam-column member. In contrast, the beam-
column strength approach seeks to bypass the use of cross-section strength
and to utilize an expression for the overall member strength as a direct part
of the analysis. That is, the strength of each member is considered directly
in defining the current beam-column stiffness and force state.

Basic Concept
In the beam-column strength approach, (1) is employed as a limit surface
for the member strength. The axial strength P„ employed in this equation
is computed based on K = 1. The member initial yield surface is assumed
to be the same shape as the limit surface based on K = 1, but it is taken
as 70% of the size of this surface. This definition is somewhat arbitrary,
but the analysis results are relatively insensitive to the precise size and shape
of the initial yield surface.
Therefore, the flexural stiffness of the beam column is assumed to start
to reduce when the strength ratio of the member reaches 0.7 (i.e., when
the member initial yield surface is contacted). The strength ratio is defined
as the coefficient, 7, obtained by evaluation of the beam-column strength
equation

-(£)'•£ <*>
Degradation of the Beam-Column Stiffness
Eqs. (18), (19), and (21) are used for the member stiffness also in the
beam-column strength approach. However, in this approach, only one p
parameter is utilized to represent the state of yielding throughout the entire
beam-column member and its effect on the member stiffness. This p pa-
rameter is defined as
p= 0 for 7 < 0.7 (23A)

(y - 0.7) s
P= 0 3 1 for 7 > 0.7 (236)

Therefore, (19a)-(19c) become

K'n = K},= (K„ - Kifj* pVl - p) (24a)

416

J. Struct. Eng. 1992.118:408-428.


K[, = Kj, = K0(l -. p)(l - p) (246)

Comments on the Use of K = 1


Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

In the preliminary evaluation of the beam-column strength approach by


the authors, elastic effective lengths based on ordinary design procedures
(Liew et al. 1991) were utilized in the calculation of the P„-teim of (22).
The results of these tests indicated that for unbraced frames, the use of the
elastic effective length is ineffective in predicting the inelastic response of
the structure. This is because the member stiffness is degraded based on
the computed value of p before (1) is satisfied. As a result, the sway capacity
of a beam-column member predicted by the beam-column strength ap-
proach, with p defined by (23), is generally less than the capacity predicted
by (1). For unbraced frames and the use of (23) for p, the best results are
obtained when the effective length is assumed essentially equal to one. This
assumption is conservative for braced frames if significant end restraint is
provided to the beam-column members.
An alternative beam-column strength approach not considered in the
present work would be to modify (23) such that p remains equal to zero
until the member strength based on an elastic or inelastic effective length
is achieved. However, this approach may also have limitations since the
/^-factors computed based on ordinary design procedures do not account
for changes in the beam-column end restraint due to aspects such as for-
mation of plastic hinges in the beams.
Of course, the true effective length of the beam columns in a framing
system after the development of distributed yielding in various beam-column
members and the formation of plastic hinges in a number of beams cannot
be calculated by simple methods, and thus, it is largely subject to engineering
judgment. For example, consider a typical column in a rigid-jointed sway
frame. If plastic hinges have formed in all the beams at the connections to
this column, it might be reasoned, based on a local subassemblage buckling
analysis that the effective length of the column is infinite in value. However,
if other columns in the story still provide sidesway resistance, the column
would act essentially as a leaned column, and K might be said to equal one.
Furthermore, if the beam-column member has yielded substantially and the
adjacent beams are elastic (such that the relative flexural stiffness of the
adjacent beams is large), the member could have an effective length essen-
tially equal to 0.5 if the other members in the story provide substantial
sidesway resistance. The only way to determine the true effective length is
to perform a second-order plastic zone analysis. However, K-tactors would
not be needed if such an analysis were performed. Fortunately, for the range
of practical design cases that have been considered by the authors, the
second-order system behavior predicted by the beam-column strength ap-
proach is fairly insensitive to the value of the effective length in the vicinity
of K = 1. This appears to be true for a wide range of values of beam-
column slenderness Llr and column end restraint. The implications of K =
1 are considered further in the discussion of the examples in a later section.

Calculation of M„
The term Mu in (22) is interpreted as the maximum second-order elastic
moment within the beam-column length. For beam-column members sub-
jected only to end loads, and for P less than the Euler load Pe = TT2EI/L2,
this term may be evaluated by multiplying the maximum calculated end
moment by
417

J. Struct. Eng. 1992.118:408-428.


B, = —£*— > 1.0 (25)
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

where C,n may be computed based on the standard formula (Manual 1986)

Cm = 0.6 - 0.4 | ± (26)

or by a more precise expression recently derived by Duan et al. (1988) as

c. . , + o.25g) - o , ( | f ( * + i) (-)
The term MJM2 in both of these expressions is positive when the member
is bent in reverse curvature bending and negative when it is bent in single
curvature. M1 and M2 are the second-order end moments obtained in the
analysis, where Mx is the smaller of the two moments in absolute value. For
end-loaded members, the combination of (25) and (27) gives essentially the
same results as the exact differential equation solution for the second-order
elastic moment within the member length.

Representation of Beam Behavior in Beam-Column Strength


Approach
In rectangular frames, the axial load in beam members is close to zero.
Therefore, for members of compact cross section that are adequately braced
against out-of-plane failure, the member strength is controlled by the cross-
section strength. That is, if strain-hardening effects are neglected, the be-
havior predicted by the plastic-hinge models is essentially equivalent to the
physical behavior. Regions of large curvature are actually developed in the
physical member. If a transverse concentrated load is applied to a beam
with fixed ends, a plastic-hinge approach will predict three plastic hinges in
the beam when the maximum strength is achieved. In the beam-column
strength approach, this beam is considered as failed essentially when just
one plastic hinge is formed in the beam. That is, the beam-column strength
model does not predict the development of concentrated plastic-hinging
action in the limit that the axial force approaches zero. Therefore, in the
present work, the modified plastic-hinge element is still employed for all
beam members in analyses performed with the beam-column strength
approach.

NUMERICAL STUDIES

In this section, a number of benchmark frames are analyzed by the elastic-


plastic-hinge, modified plastic-hinge, and beam-column strength ap-
proaches, and the results of the analyses are compared and contrasted. In
all of the studies, only one element per member has been employed for the
simplified approaches unless noted otherwise.

Inelastic Behavior of Isolated Beam Column


Tests results for 37 full-scale, as-rolled, wide-flange beam columns have
been reported (Van Kuren and Galambos 1964). These tests were conducted
to study the strength and deformation behavior of beam columns in the
418

J. Struct. Eng. 1992.118:408-428.


inelastic range. The effects of axial force, member length, cross-section size,
loading condition, and lateral bracing were each considered by the inves-
tigators. In all the tests, the axial force was applied first and then held
constant while the end moments were applied by hydraulic jacks through
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

lever arms. Bending was about the strong axis of the member. The end
conditions were essentially pin-ended in the plane of the applied moments
and fixed perpendicular to the plane of bending.
Fig. 4 compares the analysis results based on the modified plastic-hinge
and beam-column strength methods to the experimental results for test
A-7. As shown in the figure, this test specimen is a 4WF13 member, 4.877
m long (L/rx = 112), with an applied moment at one end only. A concentric
axial load of 0.158.P,, is applied to the member prior to the introduction of
the moment. The test specimen is braced such that out-of-plane failure is
prevented until well after the maximum strength of the member is reached.
The peak moment obtained in the experiment is 0.884MP while the peak
moments obtained by the modified plastic-hinge and beam-column strength
analyses are 0.91MP and 0.88MP, respectively. It should be pointed out that
elastic-plastic-hinge analysis also predicts a peak moment of 0.91MP for this
member. However, the elastic-plastic-hinge method does not capture the
degradation in stiffness exhibited by the curves in Fig. 4 as the member
strength is approached. The modified plastic-hinge method essentially du-
plicates the moment-rotation behavior of the physical test up to the point
that its peak moment capacity is reached. However, the applied moment
in the modified plastic-hinge analysis continues to increase slightly after this
point is reached, whereas, the test specimen unloads slightly after this point.
Nevertheless, for this member, the peak moments predicted by the elastic-
plastic hinge and modified plastic-hinge methods are only about 3% higher
than the experimental results. The beam-column strength method gives a
better answer for the peak moment capacity of this member, but its overall

Elastic-Plastic Hinge

0.0 0.02 0.04 0.06 0.08 0.10


Rotation Angle 6 0 (Radians)

FIG. 4. Moment-Rotation Curves for Beam-Column Test Reported by Van Kuren


and Galambos (1964)

419

J. Struct. Eng. 1992.118:408-428.


prediction of the moment-rotation behavior is not as good as that of the
modified plastic-hinge approach.
Since only one element is used for the analysis of this member, none of
the three analysis methods discussed are capable of following the unloading
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

portion of the experimental moment-rotation curve shown in the figure.


However, if a large number of elements is employed with the modified
plastic-hinge or beam-column strength methods, or with a refined plastic-
zone approach, the gradual unloading behavior can be captured. For this
member, the lack of ability of these analysis models to predict the unloading
response is due to the approximation of the member inelastic P• — 8 effects
by use of only one element for the entire member length. Finally, it should
be noted that the elastic-plastic-hinge model predicts essentially the same
results for the analysis of this member regardless of the number of elements
employed. This is because the maximum second-order elastic moment occurs
at the end of the member for the level of axial load specified in this problem.

Columns Subjected to Eccentric Axial Loading


A total of 93 tests of eccentrically loaded columns were conducted by
Johnston and Cheney (1942), of which 89 tests were made with 315.7 mem-
bers and 6 with 6WF20 columns. A number of these tests were considered
in studies performed by Galambos and Ketter (1959). The boundary con-
ditions of the columns were essentially pin-ended for strong-axis bending
and fixed in the weak-axis direction. The results for maximum strength from
the series of tests C-49 through C-54 is compared to the results predicted
by the elastic-plastic-hinge, modified plastic-hinge and beam-column strength
methods in Fig. 5. In these tests, the slenderness ratio Llrx varies from 22.6
to 62.0, but the eccentricity of the load is held constant at 1.01 in. (25.7
mm).
The analysis results are shown for the elastic-plastic and modified plastic-

Member: 315.7 /
1.0
Fy = 291 MPaand
281 MPa

0.8

p
y
0.6

0.4
Elastic-Plastic Hinge or Modified Plastic Hinge (1 element)
Elastic-Plastic Hinge (2 elements)

0.2 Beam-Column Strength (1 element)


Modified Plastic Hinge (2 elements)
Test

""0 10 20 30 40 50 60 70
L/rx

FIG. 5. Axial Load versus Llrx for Eccentrically Loaded Column Tests Reported
by Johnson and Cheney (1942)

420

J. Struct. Eng. 1992.118:408-428.


hinge methods based on the use of both one and two elements for the
member, and for the beam-column strength method based on use of only
one element. Fig. 5 shows that the beam-column strength method is suc-
cessful in predicting the peak strength of the member with only one element,
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

whereas, the modified plastic-hinge method requires the use of two equal-
length elements to capture the peak strength of the member accurately. The
reason for this behavior is that the peak second-order moment does not
occur at the member ends. Rather, the peak moment occurs at the midspan
of these members. The modified plastic-hinge element developed in this
work is capable of forming plastic hinges only at its ends.
It is important to note that, if two elements are used, the elastic-plastic-
hinge approach is capable of predicting the peak load capacity of the column
member to within 2% for the Llrx = 22.6 case. However, this approach
still overpredicts the moment capacity by 6% with two elements when Llrx
= 62.0. The modified plastic-hinge and the beam-column strength methods
give slightly more accurate results for the member strength.

El-Zanaty's Portal Frame


The portal frame shown in Figs. 6-8 has been analyzed using plastic-zone
methods for PIPy = 0.6 by El-Zanaty et al. (1980) and for PIP = 0.2, 0.4,
and 0.6 by White (1985). These plastic-zone solutions were based on as-
sumed peak compressive residual stresses at the flange tips of Frc = 0.333/^,
a constant tensile residual stress in the web of Frt = 0.210/^, and a linear
variation of the residual stresses in the flanges. No initial geometric imper-
fections were specified in these studies. The gravity loads are applied first,
and then the frames were subjected to lateral load up to collapse and through
the postcollapse range of the behavior. In Figs. 6-8, the elastic-plastic-
hinge, modified plastic-hinge, and beam-column strength results are com-
pared to the exact plastic-zone solutions for PIPy = 0.4 and 0.6.
The El-Zanaty frame is considered here because it is one of the most
sensitive problems available for testing the ability of second-order inelastic
analysis programs to capture the effects of distributed plasticity in beam-
column members. This is because the value of Xc [see (3)] based on an elastic
K-iactor of 2.328 is equal to 1.044 for the columns of this frame. Since the
difference between the column design strength curves and the column strength
predicted by an elastic-plastic hinge model is roughly a maximum for Xc =
1.0, as shown in Fig. 9, the limitations of the elastic-plastic hinge approach
for predicting the maximum strength tend to be highlighted for frames that
have values of Xc within this range and in which the columns are subjected
predominantly to axial load effects. As shown by Fig. 9, the elastic-plastic-
hinge approach predicts that the strength of a concentrically loaded column
is equal to Py unless the elastic critical load of the member Pcr is less than
Py (in which case, the strength is equal to Pcr). Since X? is the ratio between
the elastic critical load Pcr and the squash load Py of the columns, the elastic
buckling load for the El-Zanaty frame is P = 0.91/*,,.
Another reason for the sensitivity of the El-Zanaty frame for testing
second-order inelastic analysis capabilities is that there is essentially no
capacity for inelastic force redistribution in the frame. That is, both columns
are subjected to essentially the same loads through the loading (there are
some minor differences due to the overturning effects of the lateral load
H), and there are no other members in the frame to which the lateral load
moments can be redistributed.
The lateral load-displacement curves for the different types of analysis
421

J. Struct. Eng. 1992.118:408-428.


Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

0.01 0.02 0.03 0.04 0.05


A/L

FIG. 6. Load-Deflection Curves for El-Zanaty's Portal Frame with P/Py = 0.4

H,A

L/r-40
Fy - 250 MPa
0.20 E -200,000 MPa
F,= - 0.333 F»
PI P, - 0.6

0.16
Elastic-Plastic Hinge ''

2M„
0.12
y
Modified /
Plastic Hinge /'

0.08

0.04

0.00
0.0 0.004 0.008 0.012 0.016 0.020
A/L

FIG. 7. Load-Deflection Curves for El-Zanaty's Portal Frame with PIP, = 0.6

are given for this problem in Figs. 6 and 7. The difference between the
collapse loads obtained from the elastic-plastic-hinge and the plastic-zone
analyses is large. The elastic-plastic hinge results for the maximum value of
H ate 30% and 72% greater than the plastic-zone results for PIPy = 0.4
and 0.6, respectively. These results support the general conclusion made by
El-Zanaty that "the elastic-plastic [hinge] method of analysis is found to
overestimate the ultimate capacity of frames under the action of lateral
422

J. Struct. Eng. 1992.118:408-428.


P P
H, A | W8x31

" L / r - 40
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

Fy-250MPa
L E - 200,000 MPa W
F l c -0.333 F,


oQ>, ' Mh

n Plastic Zone
* Beam-Column Strength
A Modified Plastic Hinge
o Elastic-Plastic Hinge

0.0 0.2 0.6 O.E


HL/2M„

FIG. 8. Comparison of Maximum Strengths Obtained by Different Analysis Meth-


ods to Various Beam-Column Strength Curves for El-Zanaty's Frame

Column strength predicted by


an elastic-plastic hinge approach

CRC Column Strength \,


(Eq. 2) \

0.0 0.25 0.5 0.75 1.0 1.25 1.5 1.75 2.0 2.25

• KL/rTt / F„ / E

FIG. 9. Column Strengths Obtained by CRC Column Strength Equations versus


Those Predicted by Elastic-Plastic Hinge Analysis

loads, together with large axial loads and column slenderness ratios," (El-
Zanaty et al. 1980). The figures illustrate that, for this type of test case, the
solutions obtained with the modified plastic-hinge and the beam-column
strength methods are significantly more accurate than the elastic-plastic-
hinge results.
Fig. 8 further illustrates the differences between the various methods in
their prediction of the strength of the El-Zanaty frame, and compares the
423

J. Struct. Eng. 1992.118:408-428.


analysis solutions for strength to several beam-column interaction curves.
It is significant to note that the two points representing the plastic-zone
results for P/Py = 0.4 and 0.6 essentially fall on the AISC-LRFD interaction
curve based on the CRC column strength (2) and a K-factor of 2.25 deter-
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

mined from an inelastic buckling analysis of the frame. Of course, the elastic-
plastic-hinge results fall on the interaction curve defined by the cross-section
strength expression, (6). The results from the modified plastic-hinge and
beam-column strength analyses fall in between these two interaction curves.
Also, it is interesting to note that the beam-column strength based on the
Duan and Chen interaction equation (1) with K = 1 is essentially the same
as the interaction curve based on cross-section strength for this particular
frame. This curve is the member strength that is employed by the beam-
column strength method. Since in the beam-column strength approach, the
member stiffness is assumed to degrade as the member strength curve is
approached, the actual maximum strengths obtained by this method are, in
general, less than the strengths represented by the beam-column strength
equation.

Vogel's Six-Story Calibration Frame


The six-story frame shown in Fig. 10 was proposed (Vogel 1985) as one
of three frames for verifying the reliability and accuracy of second-order
inelastic analysis programs. Initial residual stresses as specified in "Ulti-
mate" (1984) are assumed in Vogel's plastic-zone analysis along with initial
sway imperfections as specified in Fig. 10. The same sway imperfections are

Elastic-Plastic Hinge
1.2
Modified Plastic Hinge
Beam-Column Strength
Plastic Zone
1.0 10-23 kN j, j , j . j, j , j , ,L j . j , ,L ,L j , j .
IPE240
HEB160 HEB200
n
H., = 20.44 k N , 1
IPE300
0.8 HEB160 HEB200 11
1
H .,. J. L
. i i r . i .i. .i. TT:
IPE300
HEB220 HEB240
CC q1
-o 0.6 H ) j n n
IPE330 @
o HEB220 HEB240
Hi jc J. .1 i J. i ~ J. i .i. i J.
IPE360
0.4 HEB220 HEB260
H 1 1 .1. J. 1
1 —:
Frame Imperfection, W0 1/450 IPE400
(each story) HEB220 HEB260
0.2 q , = 49.1 kN/m
= 31.7kN/m
rmTm nrftm mrrm
2 @ 6.0m

0.0
0.0 6.0 12.0 18.0 24.0 30.0
Lateral Deflection at Top Right-hand Joint (cm)
FIG. 10. Load-Deflection Curves for Vogel's Six-Story Frame

424

J. Struct. Eng. 1992.118:408-428.


considered in each of the simplified analyses presented here. The loads are
applied proportionally in all the studies.
All four of the analysis methods that have been previously discussed give
essentially the same results for the limit load of this frame. The maximum
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

capacity is reached at a load parameter of 1.11 in Vogel's plastic-zone study,


at 1.10 for both the modified plastic-hinge and beam-column strength so-
lutions of the present work, and at 1.13 for the elastic-plastic-hinge solution
obtained in the present work. The differences in the computed deflection
at the top of the frame by the four methods is somewhat more significant,
as illustrated in Fig. 10, but these results are also remarkably close in value.
These results support the conclusion that has been drawn by El-Zanaty and
others that "when the overall behavior of the frame is considered, local
losses in frame stiffness due to inelastic action are often compensated for
by the behavior of other members and by the effects of strain-hardening,"
(El-Zanaty et al. 1980). Also, whenever the nonlinear behavior is dominated
by inelastic action in the beams, which is the case for the six-story Vogel
frame, the elastic-plastic-hinge type of model gives an accurate represen-
tation of the behavior.
Generally, in local regions of a large frame where columns are subjected
to high axial loads, the prediction of the distribution of loads and the mag-
nitude of deformations by the elastic-plastic-hinge approach will be less
accurate than the prediction of the overall strength of the system and the
overall load-deflection results. It may be argued that, as long as inaccuracies
in the prediction of local effects do not influence the prediction of overall
strength, the estimation of the local distributions of loads and deformations
is of minor importance. However, it is possible that the precipitation of an
overall collapse by a local failure may be missed by an elastic-plastic-hinge
analysis. This problem justifies the consideration of alternative methods
such as those discussed in this paper for analysis of any structural system.

SUMMARY AND CONCLUSIONS

Several important conclusions can be drawn from the aforementioned


numerical examples and from additional numerical studies that have been
reported (King 1990) with regard to simplified methods for second-order
inelastic frame analysis.

1. The elastic-plastic-hinge analysis model performs quite well in many cases.


Specifically, for realistic multibay and/or multistory frames, often most of the
beam-column members are not subjected to severe distributed yielding along
their lengths and are not loaded close to their buckling load as individual columns
as the maximum strength of the structural system is approached. Furthermore,
much of the inelastic action in plastically designed frames is usually in the beams.
Also, multibay and/or multistory frames often have a significant capacity for
inelastic redistribution of loads prior to failure of the system. For these types
of structures, the elastic-plastic-hinge approach tends to provide predictions of
the overall frame inelastic stiffness and strength that are comparable to those
obtained from an exact plastic-zone analysis.
2. For isolated individual members and for frames in which a significant
portion of the structure experiences significant distributed yielding along mem-
ber lengths at failure, refinements to the elastic-plastic-hinge model are nec-
essary to capture accurately the actual second-order inelastic behavior of the
425

J. Struct. Eng. 1992.118:408-428.


structure. This is particularly true when there is little capacity for inelastic
redistribution of forces within the structural system.
3. Simple modifications to the basic elastic-plastic-hinge model result in sig-
nificant improvements in the prediction of the inelastic load-deflection behavior
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

and strength for cases in which the elastic-plastic-hinge model performs poorly.

Two simple approaches for second-order inelastic analysis, the modified


plastic-hinge method and the beam-column strength method, have been
presented and the characteristics of their solutions discussed. In summary,
it can be stated that both methods provide improvements over the predic-
tions of the elastic-plastic-hinge approach for the cases considered, but still,
each of the methods has certain limitations. The modified plastic-hinge
approach still predicts failure of an axially loaded member either at the
elastic critical load Pcr or at the member squash load Py, as does the elastic-
plastic hinge approach (see Fig. 9). In other words, for a column member
that fails inelastically, the modified plastic-hinge model does riot reduce to
the behavior of the inelastic column in the limit that the member is loaded
by axial force alone. The beam-column strength method alleviates this prob-
lem, but it tends to predict column failure at a load somewhat less than the
strength associated with K = 1. Also, the beam-column strength method
does not reduce to the behavior of a beam member in the limit that the
member is subjected only to bending moment. More recent work by the
authors (White et al. 1991) demonstrates that a plastic-hinge type model
may be formulated that reduces to the appropriate behavior in the limit of
pure column action if the concepts discussed in Section 3 are combined with
a column tangent modulus model to represent distributed plasticity effects
associated with pure axial loading.

APPENDIX I. REFERENCES

Alvarez, R. J., and Birnstiel, C. (1969). "Inelastic analysis of multistory multibay


frames."/. Struct. Div., ASCE, 95(11), 2477-2503.
Chen, W. F., andLui, E. M. (1987). Structural stability: Theory and implementation.
Elsevier, Inc., New York, N.Y.
Chu, K. H., and Pabarcius, A. (1964). "Elastic and inelastic buckling of portal
frames." J. Engrg. Mech. Div., ASCE, 90(5), 221-249.
Duan, L., and Chen, W. F. (1989a). "Design interaction equation for steel beam-
columns."/. Struct. Engrg., ASCE, 115(5), 1225-1243.
Duan, L., and Chen, W. F. (1989b). "A yield surface equation for doubly symmetrical
sections." Struct. Engrg. Report CE-STR-89-19, Purdue Univ., W. Lafayette, Ind.
Duan, L., Sohal, I. S., and Chen, W. F. (1988). "On beam-column moment am-
plification factor." Struct. Engrg. Report Number CE-STR-88-11, Purdue Univ.,
W. Lafayette, Ind.
El-Zanaty, M. H., Murray, D. M., and Bjorhovde, R. (1980). "Inelastic behavior
of multistory steel frames." Dept. of Struct. Engrg. Report Number 83, Univ. of
Alberta, Edmonton, Alberta, Canada.
Galambos, T. V., and Ketter, R. L. (1959). "Columns; under combined bending
and thrust." /. Engrg. Mech. Div., ASCE, 85(2), 1-30.
Goto, Y., and Chen, W. F. (1987). "Second-order elastic analysis for frame design."
/. Struct. Engrg., ASCE, 113(7), 1501-1519.
Guide to stability design criteria for metal structures. (1988). 4th Ed., T. V. Galambos,
ed., John Wiley and Sons, New York, N.Y.
Johnston, B. G., and Cheney, L. (1942). "Steel columns of rolled wide flange sec-
tion." Progress Report Number 2, American Inst, of Steel Constr. (AISC), Chi-
cago, 111.
King, W. S. (1990). "Simplified second-order inelastic analysis for frame design,"
426

J. Struct. Eng. 1992.118:408-428.


thesis presented to Purdue Univ., at W. Lafayette, Ind., in partial fulfillment of
the requirements for the degree of Doctor of Philosophy.
Liew, J. Y. R., White, D. W., and Chen, W. F. (1991). "Beam-column design in
steel frameworks—Insights on current methods and trends." /. Constr. Steel Res.,
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

accepted for publication.


Manual of steel construction, load and resistance factor design. (1986). 1st Ed., Amer-
ican Inst, of Steel Constr. (AISC), Chicago, 111.
"Ultimate limit state calculation of sway frames with rigid joints." (1984). ECCS-
CECM-EKS Publication No. 33, 1st Ed., European Convention for Construction
Steelwork, Rotterdam, the Netherlands.
Van Kuren, R. C , and Galambos, T. V. (1964). "Beam-column experiments." /.
Struct. Div., ASCE, 90(2), 223-255.
Vogel, U. (1985). "Calibrating frames." Stahlbau, 10, 1-7.
White, D. W. (1985). "Material and geometric nonlinear analysis of local planar
behavior in steel frames using interactive computer graphics," thesis, presented
to Cornell Univ., at Ithaca, N.Y., in partial fulfillment of the requirements for
the degree of Master of Science.
White, D. W., Liew, J. Y. R., and Chen, W. F. (1991). "Second-order inelastic
analysis for frame design: A report to SSRC Task Group 29 on recent research
and the perceived state-of-the-art." Struct. Engrg. Report CE-STR-91-12, Purdue
Univ., W. Lafayette, Ind.
Ziemian, R. D. (1990). "Advanced methods of inelastic analysis in the limit states
design of steel structures," thesis presented to Cornell Univ., at Ithaca, N.Y., in
partial fulfillment of the requirements for the degree of Doctor of Philosophy.

APPENDIX II. NOTATION

The following symbols are used in this paper:

Bx = P - 8 moment amplification factor;


Cm = equivalent moment factor;
E = Young's modulus of elasticity;
Et = column tangent modulus;
Frc = maximum compressive residual stress located at flange tips of wide-
flange sections;
Fy = yield stress;
/ = cross-section shape factor;
Ix = I = strong-axis moment of inertia;
K = effective length factor of columns in plane of bending;
L = span length of members;
M = second-order moment at cross section under consideration;
Mi = moment at node /;
Mp = cross-section plastic moment capacity = FyZx;
Mpc = cross-section plastic moment capacity as influenced by level of axial
force in member [see (6)];
Mu = maximum factored second-order elastic moment within member
length;
Myc = yield moment of cross section as influenced by level of axial force
in member [see (8)];
Pe = TI2EllL2 = Euler buckling load of member;
P„ = strength of member as column;
Pu = P = factored axial force in beam column;
Py = member squash load = FyA;
rx = radius of gyration about strong axis;
427

J. Struct. Eng. 1992.118:408-428.


plastic section modulus for strong-axis bending;
1.3 + 0.002 (KLIr)x;
member strength ratio [see (22)];
normalized slenderness ratio;
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 08/22/13. Copyright ASCE. For personal use only; all rights reserved.

parameter that characterizes extent of plastification for overall


member in beam-column strength approach;
parameter that characterizes extent of plastification at node / of
element in modified plastic-hinge method; and
rotation at node i.

J. Struct. Eng. 1992.118:408-428.

Potrebbero piacerti anche