Vous êtes sur la page 1sur 17

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245301529

Geometric Nonlinearities in Nonsway Frames

Article  in  Journal of Structural Engineering · December 1983


DOI: 10.1061/(ASCE)0733-9445(1983)109:12(2770)

CITATIONS READS

11 82

3 authors, including:

Jostein Hellesland
University of Oslo
101 PUBLICATIONS   544 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Jostein Hellesland on 01 June 2014.

The user has requested enhancement of the downloaded file.


GEOMETRIC NONLINEARITIES IN NONSWAY FRAMES
By Shu-Ming A. Lai, 1 James G. MacGregor, 2 F. ASCE,
a n d /ostein Hellesland 3

ABSTRACT: The geometric nonlinearities in elastic nonsway frames are exam-


Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

ined and used to review the rationale and accuracy of the ACI slender column
design method. The use of nonsway effective length factors to account for ef-
fects of end restraints, and the assumptions of idealized end restraints in the
conventional alinement chart for effective length factors are investigated. The
discussion is also applicable to the moment magnifier portion of the AISC col-
umn interaction equation.

INTRODUCTION

Two major difficulties in the analysis and design of slender reinforced


concrete frames result from the "material nonlinearities" caused by the
inelastic properties of the materials and changes in section due to crack-
ing, and the "geometric nonlinearities" due to the influence of displace-
ments on the equilibrium of structures. Current design approaches for
slender concrete columns are based on approximate methods that in-
clude these two nonlinearities by modifying the first-order elastic anal-
ysis (12). Starting from a complete nonlinear analysis that includes both
effects, two distinct steps are needed to derive the current design pro-
cedures. In step 1, the material nonlinearities are approximated by the
introduction of equivalent EI values for the individual members, such
that an elastic second-order analysis gives results close to those from a
second-order inelastic analysis at a defined stage in the loading history
close to the ultimate limit state. In step 2, the second-order elastic anal-
ysis is replaced by a first-order analysis modified to give results close to
those from the second-order elastic analysis. Thus, in this procedure the
effects of the nonlinearities are approximated independently of each other.
In step 1, it has been tacitly assumed that inelastic structures behave
in a way that can be represented by the elastic behavior modified with
appropriate effective EI values. This basic assumption was made in the
derivation of the present ACI Code (3) procedure which allows the use
of elastic theory in analysis. The ultimate strength limit state, implicit in
the ACI Code, is defined by the onset of the first plastic hinge or by the
simultaneous occurrence of plastic hinges in several critical sections. It
is generally assumed that elastic analysis is still reasonable at the stage
immediately prior to the occurrence of this limit state.
The modification of the first-order elastic analysis in step 2 is the pri-
mary objective of this and a companion paper (9). Frequently in design,
a frame will be analyzed separately for lateral loads (assuming sway frame
'Struct. Engr., Hong Kong.
2
3
Prof. of Civ. Engrg., Univ. of Alberta, Edmonton, Canada.
Sr. Struct. Engr., Dr. Ing. A. Aas-Jakobsen Ltd., Struct. Consulting Engrs.,
Oslo, Norway.
Note.—Discussion open until May 1,1984. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Technical and Profes-
sional Publications. The manuscript for this paper was submitted for review and
possible publication on November 24, 1982. This paper is part of the Journal of
Structural Engineering, Vol. 109, No. 12, December, 1983. ©ASCE, ISSN 0733-
9445/83/0012-2770/$01.00. Paper No. 18464.
2770

J. Struct. Eng. 1983.109:2770-2785.


action) and for gravity loads on what is essentially a nonsway frame.
The resulting moments and forces are then superimposed. This paper
deals with the nonsway frame while the companion paper (9) considers
the sway frame and the superposition of the force resultants from the
two analyses. According to the principle of superposition (14), the axial
forces in the members of the sway and nonsway frames must be equal
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

to those of the original frame under the actual state of loading in order
for superposition to apply.
Scope.—The derivation of the ACI (3) design method for nonsway
slender columns was based on pin-ended columns (12). The method was
extended to single restrained columns using nonsway effective length
factors to account for effects of the end restraints. An effective length
factor alinement chart, which assumes idealized end-restraints for a col-
umn, is generally used to compute the effective lengths of restrained
columns in a multistory frame. The rationale of the derivation, and the
accuracy of the ACI design method for braced slender columns, are in-
vestigated. The analysis of the effective length design procedure is also
applicable to the moment magnifier portion of the AISC column inter-
action calculation.
Exact Elastic Analysis.—The method of "exact" analysis used in this
paper is an elastic second-order slope-deflection analysis (7) based on
the following assumptions:
1. Only in-plane behavior is considered.
2. The analysis is based on an elastic small deflection theory.
3. All individual members are prismatic.
4. No transverse or axial loads act between ends of columns.
5. Shearing and axial deformations are neglected.
6. Geometric effects due to axial forces in the beams are neglected.
The approximate design method evaluated in this paper is based on
these plus additional assumptions, as described later.
Notation.—The notation is defined in Appendix II. It is helpful to note
that shears, V, moments, M, and displacements, a, without subscripts
refer to the final second-order values. The subscript "0" as in V0i or a0
refers to the first-order value. The subscript " 2 " as in M2 refers to the
larger end moment in a column.

PIN-ENDED COLUMNS

The deflected shape of an elastic pin-ended column at failure (when


deformations approach infinity corresponding to small deflection anal-
ysis) approaches the symmetrical sine wave buckling or critical mode
regardless of the initial deformations produced by the end moments (8,11).
As a result, when the axial force increases, the inflection point, if one
exists, tends to shift to the nearest end of the column while the point
of maximum moment approaches midheight. In the case of a column
that is initially bent into almost an antisymmetrical double curvature,
the column will unwind rather suddenly into a single curvature as the
critical load or Euler load, ir2EI/L2, is reached.
The maximum moment, M max , in a pin-ended column can be written as
2771

J. Struct. Eng. 1983.109:2770-2785.


Mmax = SMQJ (1)
in which M02 = the numerically larger first-order end moment; and 8 =
the moment magnification factor which is generally separated into two
terms:
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

8 = Cm8r„=i > 1.0 (2)


in which 8ro=i = the value of 8 corresponding to a column with sym-
metrical end moments (r0 = -M 0 i/M 0 2 = 1.0); and Cm = the correction
factor employed to account for r0 ¥= 1.0. The magnifier, 8, can be ap-
proximated by
1 + 0.25a
(3)
l-o
in which a = the load index, N/Ne; N = the compressive axial load in
the column; and Ne = the Euler load. When Eq. 3 is compared to the
exact elastic solution, described earlier, the errors are within +0.2% and
-0.1% for N < 0.5N e , and - 2 % for N <Ne.
An approximation for Cm, neglecting the dependence on N/Ne, was
suggested by Austin (1):
Cm = 0.6 + 0.4r0 > 0.4 (4)
The lower limit of 0.4 was derived for elastic lateral-torsional buckling
and is unnecessary for in-plane bending.
In the ACI Code (3), the maximum moment in a pin-ended column is
determined according to

FIG. 1.—Comparison of Approximate with Theoretical Magnification Factor


2772

J. Struct. Eng. 1983.109:2770-2785.


(5)
1-a
with Cm given by Eq. 4. (The resistance factor, <j>, is neglected here be-
cause this analysis deals with ideal elastic members.) The errors in Eq.
5, compared to exact elastic solutions, are shown in Fig. 1. The approx-
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

imate equation is unconservative in many cases because the term 0.25a


is neglected. It appears reasonable to include this term in the ACI design
equation.
Although the limit of Cm S: 0.4 is unnecessary for in-plane bending, it
seems reasonable to retain this limit because of "the uncertainty of frame
action when values of r are between -0.5 and - 1 . 0 " (12). Tests of rein-
forced concrete columns bent into double curvature of r ~ - 1 (11,13)
have indicated that under high axial loads the column tends to unwind
rather suddenly with the column approaching the instability mode for
a column subjected to pure axial load. Since the derivation of the ACI
design equation assumed that "material failures" would occur rather than
"stability failures," it appears advisable to retain the limit to safeguard
against an instability mode of failure.

BEHAVIOR OF SINGLE RESTRAINED COLUMNS

A single restrained elastic column subjected to external moments, Mn


and M t2 , and axial compression, N, is shown in Fig. 2 with the related
symbols and sign convention. The linearly elastic springs represent the
rotational restraints at the ends. A portion of M tl and Ma is resisted by
the restraints, the rest by the column. As the compressive force in-
creases, the deflection mode of such a column will ultimately change to
its critical mode regardless of the first-order condition (8). The critical
mode for a column restrained at both ends is triple curvature, i.e., two
inflection points between the ends.
The process for reaching the final mode is shown in Fig. 3 for a sym-
metrically restrained column subjected to constant external moments but
increasing axial forces. In the figure, the first-order end moment ratio,
r0, is defined by -M 0 i/M 0 2 in which M02 = the numerically larger first-
order end moment. The first-order moment diagrams are shaded, stage
1. The points of maximum moment are shown with solid dots.
For 1 > r0 a 0 shown in Fig. 3(a), M t and M2 both decrease from the
beginning [axial load stage 2, Fig. 3(a)], then reverse direction and in-
crease negatively (stage 3).
For 0 s r0 s -0.9 (the lower limit on r0 depends on the magnitude of
the end restraints but it is close to —1), the smaller end moment, Mx,
increases with increasing axial load while M2 decreases at the outset as
shown in Fig. 3(b). After a change of direction, M2 increases in the op-
posite direction.
For -0.9 £ /"o — ~i, both Mi and M2 decrease gradually initially [stage
2, Fig. 3(c)]. As the critical load is approached, Mj rapidly increases, and
M2 rapidly decreases, changes direction and increases in the negative
direction. This drastic change from double curvature to triple curvature
is referred to as "unwrapping" or "unwinding."
In each of the preceding cases, a restrained elastic column is seen to
2773

J. Struct. Eng. 1983.109:2770-2785.


Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

(c) -0.9*> r0 >-1 'Aproximate. see text.


Restrained column Internal forces (unwrapping)

FIG. 2.—Symbols and Sign Convention FIG. 3.—Bending Momenta and De-
fer Restrained Column fleeted Shapes of Symmetrically Re-
strained Columns under Increasing
Compression

develop tri-curvature with two points of contraflexure as failure is ap-


proached. The moment diagrams and deflected shapes plotted in Fig. 3
were obtained by "exact" analyses based on the assumptions made earlier.
Increasing axial compression will reduce the maximum moment in the
column if the maximum remains at the end of the column as shown in
Fig. 3(c), stage 2. With further increase in the compressive force, the
maximum moment will be forced away from the end and will increase
thereafter. Fig. 3 also shows the change in deflected shapes due to in-
creasing axial forces. As the end rotations increase the moments in the
beams increase, changing the column moment needed to equilibrate the
external moment on the joint. A comparison of the moments and de-
flections show that the end moments change in the opposite direction
to the change of end rotations. In other words, the change of end mo-
ments restrains the column deformations, resulting in a larger critical
load than if the column were unrestrained.
Approximation of Frame Action Using Effective Length Method.—In
the current design approach in the United States (3,5) and Canada for
restrained nonsway columns, the restrained column is replaced by an
equivalent pin-ended column whose length is equal to the effective length
of the real restrained column, knsL [Fig. 4(a)]. This equivalent pin-ended
column is then analyzed for the axial compression plus the first-order
internal end moments of the real column.
This "effective length method" was proposed by Winter (15) based on
2774

J. Struct. Eng. 1983.109:2770-2785.


./'
=fi
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

M
0lyA; *•

J
- „ * f
N (a)

first order
moment \ / |
actual f |
moment \ ^ | |

^M„:
first order V
moment

actual I ^
moment \ \ \

FIG. 4.—Effective Length Method for Single Restrained Columns

the work of Lee (10) and Bijlaard, et al. (2). Lee suggested that the max-
imum moment in a restrained elastic column with equal end restraints
and symmetrical joint moments can be approximated using the effective
length method. Bijlaard, et al., extended Lee's approach, with some
modification, determining the collapse load of an inelastic nonsway steel
column with symmetrical elastic end restraints and symmetrical joint
moments. Although the two studies were limited, the method as used
nowadays became generalized.
Figs. 4(fr)-(c) compare the actual and assumed moment diagrams in
two restrained columns at an advanced stage of loading. The actual mo-
ment curve intersects the first-order moment curve at two points where
the actual moments are identical to the first-order end moments. The
column can then be considered as a pin-ended column subjected to end
moments equal to the first-order end moments of the real restrained col-
umn, with a column length equal to the distance between the two in-
2775

J. Struct. Eng. 1983.109:2770-2785.


tersection points. For a column subjected to uniform first-order mo-
ments [Fig. 4(b)] the design process approaches the actual behavior,
although the assumption that the intersection length is equal to the ef-
fective length is questionable. In the case of a nonuniform first-order
moment [Fig. 4(c)] where the moments at the intersection points are dif-
ferent from the first-order end moments, the method loses its rationality.
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

According to the effective length approach, the maximum moment in


a restrained elastic column is equal to
M„ + V l + r\ - 2r0 cos p
for r0 > cos P;
Ma- sin (3
= 1.0 for r0 < cos p (6)
2 2
in which p = irVN/N„, and N„s = tt EI/(knsL) . Eq. 6 is the exact elastic
solution for a pin-ended column (7), except that L has been replaced by
kmL, and M2 by M02 as used in the effective length approach. The ac-
curacy of Eq. 6 for symmetrically restrained columns with symmetrical
single curvature end moments is shown in Fig. 5(a) for different values

^ ^ X J n s = °- 55
K = K
1 2
-0=1-0

^***«^a65 ^X^

6
^^^^"""-OZ ^"s,"»

l l i ^i™~--2SJB N
1.0 N

(b)

FIG. 5.—Examination of Effective Length Method for Ki = K2 and r0 = 1.0


2776

J. Struct. Eng. 1983.109:2770-2785.


of km as a function of N/Nns. The theoretical Mmax is from the exact elas-
tic analysis described earlier. This figure shows the effective length method
to be unconservative for this case, the errors increasing with increasing
axial compression and with stronger restraints or lower k„s.
A modified effective length factor, km, which gives exact values of the
maximum moment when used in Eq. 6 is plotted in Fig. 5(b) for the same
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

columns as in Fig. 5(a). Because the first-order bending moment is uni-


form along the column, k„L is identical with the distance between the
intersection points (intersection length) in Fig. 4(b), Fig. 5(b) shows that
with increasing compression km approaches the value of kns and is ex-
tremely close to kns at the critical load, Nm. However, in the practical
range of N/N„s, km is larger than k„s.
When Figs. 5(a)-(b) are examined together, they indicate that although
k„sL becomes a more accurate approximation of the intersection length
at higher axial loads, the resultant errors in the maximum moment in-
crease. This is also indicated in the extreme condition of N = N„s, where
the ratio km/k„s is extremely close to 1.0 while the difference in the mo-
ments is still distinct. This is because the maximum moment given by
Eq. 6 increases very rapidly with increasing compression, and thus the

FIG. 6.—Examination of Effective Length Method for Kt = K2 and rB =£ 1.0


2777

J. Struct. Eng. 1983.109:2770-2785.


change in Mmax becomes more sensitive to the change in N/N„s (i.e., kns)
when the axial load is larger.
The preceding observation suggests that a small overestimation of k„s
can eliminate the unconservative errors in the maximum moment for
some particular cases. Thus, if for k„s = 0.75 the value of kns is overes-
timated by about 3.5% [Fig. 5(b)], the maximum moment for N = 0.5N„S
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

can be accurately determined using Eq. 6. On the other hand, if k„s is


underestimated, the maximum moment will be underestimated to a much
larger degree.
The preceding results for columns with equal end restraints are also
applicable to the unequal end restraints, provided the first-order mo-
ment is also uniform. Lai (8) has shown that the maximum moments
are essentially independent of the ratio of end restraints for a given value
of k„s and uniform first-order moment.
The results of a similar study for columns with equal end restraints
corresponding to k„s = 0.75 but nonuniform moments are shown in Fig.
6. The conservatism of Eq. 6 depends on r0, becoming more conserva-
tive as r0 decreases. For restraint magnitudes other than kns = 0.75, the
results are similar except that magnitude of errors will be larger for stif-
fer restraints and smaller for more flexible restraints. This is because the
actual value of Mmm/Mo2 for r0 < 1.0 is initially decreased by the axial
load; whereas the value of M^/Mcc determined by Eq. 6 is always greater
than or equal to 1.0.
For the case of unequal end restraints, it was found (8) that for a given
kns, a column with a flexible restraint, K2, at the end where the larger
end moment, M 2 , acts, and a stiff restraint, Kx, at the other end (sym-
bols defined in Fig. 2) has the greatest maximum moment at any axial
load for a given k„s and r0 value. The extreme case of K2 = 0.0 (a hinge),

F!G. 7.—Examination of Effective Length Method for KJK2 = <* and r0 s 1.0
2778

J. Struct. Eng. 1983.109:2770-2785.


shown in Fig. 7, indicates that for r0 < 1.0 the maximum moment ap-
proximated by Eq. 6 is accurate up to a certain load level because the
maximum occurs at the end (Mmax = M2 = Mm). When the maximum
moment moves away from the end of the column, the approximate value
becomes unconservative. Unlike the case shown in Fig. 6, the approxi-
mate values tend to be much more unconservative for smaller values of
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

r0, This is also reflected by the values of km in Fig. 7(b).


In conclusion, the theoretical maximum moment in an elastic column
subjected to uniform first-order moment is always greater than that ob-
tained using the effective length method. For columns subjected to a
first-order moment gradient, the results range from conservative to un-
conservative. In a design situation where the maximum moments need
to be estimated, the preceding results suggest that a restrained elastic
column analyzed by the effective length method may have a safety factor
less than that implied for a pin-ended column.
ACI Design Method for Single Restrained Columns.—According to
the current ACI Code (3), the maximum moment in a single restrained
column can be taken as
Mmax = 8„sMo: (7)
in which 8„s = (C,„/l - a„s) s 1.0; Cm is given by Eq. 4 with r replaced
by r 0 ; and a„s = N/Nm. The limit of Cm s 0.4 is also required for re-
strained columns since unwrapping can occur near failure (11), similar
to the problem of the unwinding of a pin-ended column analyzed earlier.
The ACI moment magnification (Eq. 7) is compared to the upper and
lower bounds of Mmax/M02 from theoretical elastic solutions for a given
N/N„3 in Figs. 8-10 for r0 = 1.0, 0.0, and -0.5, respectively. For a given
N/N„s, the maximum moment in an elastic column is a function of kns

0,8 1.0

FIG. 8.—Comparison of Approximate with Theoretical Magnification Factor for


r„ = 1.0
2779

J. Struct. Eng. 1983.109:2770-2785.


Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

FIG. 9.—Comparison of Approximate with Theoretical Magnification Factor for


r0 = 0.0

and the relative magnitudes of the two end restraints (8). The corre-
sponding upper and lower bound values represent the extremes from
all possible cases.
In the case of r0 = 1.0, the ACI equation falls below the lower bound
curve for all values of N/Nm for two reasons. First, the term 0.25a„s has
been neglected in Eq. 7. If it were included, Eq. 7 would be almost the

N/N„,

FIG. 10.—Comparison ©f Approximate with Theoretical Magnification Factor for


r0 = -0.5

2780

J. Struct. Eng. 1983.109:2770-2785.


same as the lower bound curve. Second, the use of effective length itself
makes the solutions unconservative for this case as reviewed earlier. For
smaller values of r0, Eq. 7 approaches the lower bound or moves into
the bounded region. As r0 decreases, Eq. 7, which is derived from the
approximate equation for a pin-ended column (Eq. 5), approaches closer
to the theoretical equation for a pin-ended column as shown in Fig. 1.
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

In addition, the effective length approach leads to conservative results


up to certain load levels for some cases as shown previously.
The preceding studies for single restrained elastic columns indicate the
shortcomings of the effective length approach. On the other hand, the
approximate method has the advantage of simplicity. If simplicity out-
weighs accuracy, it appears advisable to have conservative approximate
methods. In fact, two such approaches currently exist. One is the graph-
ical method suggested by Wood (16), also plotted in Figs. 8-10. These
curves tend to be close to the upper bounds, providing conservative re-
sults in most cases. The second approach is to overestimate the effective
length factors in the effective length method. The ACI Commentary (4)
suggests that the effective length factor for braced columns can be taken
as the lesser of
km = 0.7 + 0.05(Gj and G2) < 1.0; Ks = 0.85 + 0.05GS < 1.0 (8)
in which Gx and G2 = the ratios of column to beam stiffness at the two
ends; and Gs = the smaller of Gi and G 2 . Eqs. 8 are a conservative em-
pirical approximation to kns for a single restrained column derived by
Cranston (6). The use of kns obtained from Cranston's equation, plus the
inclusion of the 0.25a„s team in the ACI equation (Eq. 7), will provide
conservative results for single restrained columns.

RESTRAINED COLUMNS IN MULTISTORY FRAMES

The following analysis assumes that the first-order moments in elastic


nonsway multistory frames can be represented by the unbalanced end
moments at the joints due to the gravity loads. The column moments
resulting from this type of loading are the same as those from the actual
gravity loads.
The previous study of single restrained columns assumed that the
stiffness of an end restraint remains constant as the axial load increases.
When a column is a component of a frame, the end restraints of any
column may be affected by the geometric action in two ways: the "hor-
izontal interaction" between adjacent columns, and the "vertical inter-
action" of columns continuous through several stories.
Horizontal interaction results from the change of the deflected shapes
of the beams due to the second-order deflections of the columns con-
nected to the ends of the beams. This is not significant except when the
columns carry axial loads close to those which would initiate elastic fail-
ure of the frame (8).
The vertical interaction of columns is a function of the relative stiff-
nesses of columns in successive stories, and the stiffnesses of the beams
between the stories. The change in the relative stiffnesses of the columns
due to axial loads affects the distribution of moments between columns.
If the beams are stiff relative to the columns, any changes in the end
2781

J. Struct. Eng. 1983.109:2770-2785.


moments of a column will be absorbed primarily by the beams and,
therefore, will not significantly affect the columns above or below. On
the other hand, if the beams are very flexible, any changes in one col-
umn will affect directly the other stories through the continuous col-
umns. In effect, the strong columns will assist the weak columns through
the mechanics of vertical interaction when the beams are flexible (8). The
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

strong columns lose end restraints while the weak columns gain the
restraints.
Effective Length Method for Restrained Columns in Multistory
Frames.—The effective length method considered earlier is based on sin-
gle restrained columns whose end restraints are not affected by the axial
forces in the columns. In order to extend the method to restrained col-
umns in multistory frames, it is necessary to make the same assumption.
As a result, any column can be isolated from the frame in such a way
that the end restraints for the isolated column are equal to the end re-
straints derived from the first-order condition (first-order end restraints).
The restraint offered by a beam is a function of the signs of the mo-
ments at its ends (Eq. 10), and this restraint is distributed to the relative
values of the column end moments at the corresponding joint. Thus the
first-order end restraints can be determined from the first-order mo-
ments. The restraint stiffness, K2, at joint 2 of the isolated column ob-
tained in this way is equal to
M02 / EIB\
K2==
7^I7V\lmT) (9)
(ZM0)coi \ LB/beam
3
in which m= ., w (10)
IMOF
2M 0 /
and (2M0)Coi denotes the sum of first-order column end moments at joint
2; and M0J and M0r = the first-order moments at the near and far ends
of a beam connected to joint 2. The terms EIB and LB are the flexural
stiffness and length of the beam, and (2)beam denotes summation for all
beams rigidly connected to joint 2. Note that the first-order beam mo-
ments are those determined from the unbalanced fixed-end moments at
the joints due to the gravity loads. The restraint stiffness, Kt, at the
other joint of the isolated column is similar to the preceding equations.
The assumption of constant end restraints, i.e., neglecting the hori-
zontal and vertical interaction of columns due to geometric effects as
reviewed previously, should be reasonable for large regular multistory
frames where the member stiffnesses and axial loads of neighboring col-
umns do not vary appreciably. In cases where the vertical interaction is
significant due to large differences in a from story to story, this as-
sumption is conservative for weak columns but unconservative for strong
columns (8). For horizontal interaction, the further assumption intro-
duced by the following ACI method will make the solution tend to be
conservative.
ACI Method for Restrained Columns in Multistory Frames.—In the
derivation of the alinement charts given in the ACI Commentary (4), a
further simplification was made. The end restraints used to calculate the
2782

J. Struct. Eng. 1983.109:2770-2785.


effective length factor were assumed constant to their lowest value based
on the following assumptions:
M 0/ = -M0F (11)
EI
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

M02 L
(12)
(SMo),
2f) col

in which (£EI/L)col denotes summation of columns rigidly connected to


the same corresponding joint. When Eq. 11 is substituted into Eq. 10,
m becomes 2.0. Substituting m = 2 and Eq. 12 into Eq. 9, Eq. 13 is obtained:
2EI
K
^^L <13)
in which G2 = (2EI/L)col/(2EIB/LB)beam . Similarly, Kj = a function of Gt.
Consequently, the effective length factor for any column in the frame
can be computed after obtaining the values of Gi and G 2 . Cranston's
equation (Eq. 8) is also based on the same assumptions.
The assumption in Eq. 11 that the beam effective stiffness is equal to
2EIB/LB tends to be conservative. The other assumption (Eq. 12) that the
ratio of column end moments at the joint is equal to the ratio of their
stiffness parameters, EI/L, is reasonable if the far ends of the upper and
the lower column are restrained in a similar manner. It can be seen that
these two assumptions appear reasonable for a regular multistory frame
with a regular loading pattern.
If the far end of the beam that is framed into the column under con-
sideration is hinged (M0P = 0.0), m becomes equal to 3. In order to use
the alinement chart (or Cranston's equation) that is based on m = 2, the
beam length should be multiplied by 2/3 when calculating the value of
G. Similarly, for a beam with far end fixed against rotation, the beam
length should be multiplied by 0.5. This correction is mentioned in the
AISC Commentary (5) but not in the ACI Commentary (4).

ANALYSIS

In all the previous analyses, the use of effective length factors in the
ACI approach assumed that the beams maintain their assumed stiffness
throughout the loading up to the instant of collapse of the restrained
column. In order that the beam can be expected to behave this way, a
designer needs to know the first-order column end moments as well as
the column ultimate end moments at ultimate load. The ultimate end
moments, however, are unknown to the designer in the effective length
approach. Consequently, whether the beam can remain as stiff as as-
sumed becomes uncertain. This situation is most severe when the re-
straining beam is designed for moments which equilibrate the column
moments, as may be the case for beams restraining the exterior columns
in a frame or the columns in a single-story frame. If in such a case the
beam is designed to equilibrate the first-order end moments of the col-
umns, a beam mechanism will form as ultimate loads are approached
2783

J. Struct. Eng. 1983.109:2770-2785.


and the column e n d m o m e n t s remain equal to the plastic m o m e n t ca-
pacity of the beams which, in turn, is equal to the first-order m o m e n t s .
In this case, the use of kns < 1.0 is unconservative a n d the use of k„s =
1.0 is more reasonable.
For interior columns in a multibay, multistory frame, the column is
designed assuming the worst loading pattern. The beam, however, is
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

designed on the basis of a m o m e n t envelope as required in the ACI


Code (3). Consequently, the beam so designed can adapt to a consid-
erable change in the column e n d m o m e n t s before the beam w o u l d form
a mechanism, provided the e n d m o m e n t s do not change sign. The re-
serve of strength i n the b e a m m a y justify the use of km < 1.0.

CONCLUSION

This paper has examined the rationale, assumptions, and problems


behind the existing ACI design m e t h o d for restrained columns in n o n -
sway multistory frames. These studies make it clear that it is not nec-
essary to use "exact" effective length factors derived from rigorous elas-
tic stability analysis for the whole frame, since rough yet conservative
estimates are more reasonable. The use of Cranston's equation for the
effective length (Eq. 8) is suggested. In short, the design equation is
suggested as

1 + 0.25a„ s
8„s = C m — - (14)
1 - a„ s
in which Cm is given by Eq. 4, including the lower limit of 0.4, a n d k„s
is given by Eq. 8.

APPENDIX I.—REFERENCES

1. Austin, W. J., "Strength and Design of Metal Beam-Columns," Proceedings,


ASCE, Vol. 87, ST4, Apr., 1961, p. 1.
2. Bijlaard, P. P., Fisher, G. P., and Winter, G., "Strength of Columns Elasti-
cally Restrained and Eccentrically Loaded," Proceedings, ASCE, Separate No.
292, Oct., 1953.
3. "Building Code Requirements for Reinforced Concrete (ACI-318-77)," Amer-
ican Concrete Institute, Detroit, Mich., 1977.
4. "Commentary on Building Code Requirements for Reinforced Concrete (ACI-
318-77)," American Concrete Institute, Detroit, Mich., 1977.
5. "Commentary on the Specification for the Design, Fabrication and Erection
of Structural Steel for Buildings," American Institute of Steel Construction,
New York, N.Y., 1978.
6. Cranston, W. B., "Analysis and Design of Reinforced Concrete Columns,"
Research Report No. 20, Cement and Concrete Association, 1972.
7. Galambos, T. V., "Structural Members and Frames," Prentice-Hall, Engle-
wood Cliffs, N.J., 1968.
8. Lai, S.-M. A., "Geometric Non-Linearity in Multi-Storey Frames," thesis pre-
sented to the University of Alberta, at Edmonton, Canada, in 1982, in partial
fulfillment of the requirements for the degree of Doctor of Philosophy.
9. Lai, S.-M. A., and MacGregor, J. G., "Geometric Non-Linearities in Un-
braced Multi-Story Frames," Journal of the Structural Division, ASCE, Vol. 109,
No. ST11, Nov., 1983.
10. Lee, A. Y.-W., "A Study on Column Analysis," thesis presented to Cornell
2784

J. Struct. Eng. 1983.109:2770-2785.


University, at Ithaca, N.Y., in 1949, in partial fulfillment of the requirements
for the degree of Doctor of Philosophy.
11. MacGregor, J. G., and Barter, S. L., "Long Eccentrically Loaded Concrete
Columns Bent in Double Curvature," Symposium on Reinforced Concrete
Columns, Publication SP-13, American Concrete Institute, Detroit, Mich., 1966.
12. MacGregor, J. G., Breen, J. E., and Pfrang, E. O., "Design of Slender Con-
crete Columns," ACl Journal, Vol. 67, No. 1, Jan., 1970.
Downloaded from ascelibrary.org by UNIVERSITY OF OSLO on 10/21/13. Copyright ASCE. For personal use only; all rights reserved.

13 Martin, I., and Olivieri, E., "Tests of Slender Reinforced Concrete Columns
Bent in Double Curvature," Symposium on Reinforced Concrete Columns,
Publication SP-13, American Concrete Institute, Detroit, Mich., 1966.
14 Timoshenko, S. P., and Gere, J. M., Theory of Elastic Stability, McGraw-Hill
Book Co., New York, N.Y., 1961.
15 Winter, G., "Compression Members in Trusses and Frames," The Philosophy
of Column Design, Proceedings of the Fourth Technical Session, Column Re-
search Council, Lehigh University, Bethlehem, Pa., 1954.
16 Wood, R. H., "Effective Length of Columns in Multistorey Buildings," The
Structural Engineer, Vol. 52, No. 7, July, 1974.

APPINDIX II.—NOTATION

The following symbols are used in this paper:

Cm = correction factor to account for u n e q u a l e n d moment;


K = rotational stiffness of end restraint;
k = effective length factor;
L = length of column;
M = moment;
Mmex = maximum second-order m o m e n t along length of column;
N = axial compressive load;
Ne = Euler buckling load;
r = ratio of column end moments;
a = N/Ne;
(3 = IT V N / N „ S ; a n d
8 = m o m e n t magnification factor.

Subscripts
0 = first order;
1 = denotes m o m e n t or stiffness at e n d of column having smaller
end moment;
2 = denotes m o m e n t or stiffness at e n d of column having larger
end moment; a n d
ns = nonsway.

2785

View publication stats J. Struct. Eng. 1983.109:2770-2785.

Vous aimerez peut-être aussi