Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title no. 94-S36

Stability Problems of Columns and Frames

by J. Dario Aristizabal-Ochoa

In spite of the alignment charts’ limitations to be valid only to fully • Columns and frames are classified into braced and
symmetrical rectangular frames with rigid/hinged connections unbraced cases only (ACI 10.11.2).4
(i.e., frames with identical columns under equal axial loads and • The elastic critical axial load and the effective length k-
boundary conditions), they have been indiscriminately used in all factor for each column in a framed structure is a func-
types of framed structures. Construction codes and the technical tion of its ends flexural conditions only, i.e. ψtop and
literature have promoted their use since their inception more than ψbottom [ ψ = ∑ ( EI ⁄ h ) c ⁄ ∑ ( EI ⁄ L ) g ] , (ACI R10.11.2).4
forty years ago, and only recently some codes have tried to address
• The effective length k-factor for columns with lateral
certain inadequacies. This indiscriminate use has created the
following major misconceptions: 1) columns and frames are clas-
sidesway buckling shall be greater than one (ACI
sified into braced and not braced cases only; 2) the critical load 10.11.2.2).4
and effective length k-factor for each column in a frame is a function • Pin-ended columns (leaning columns) in frames with
of its ends’ flexural conditions only; 3) the k-factor for columns sidesway buckling have no axial load capacity, [ACI
with lateral sidesway shall be greater than one (k > 1); 4) pin- Fig. R.10.11.2(b)} for ψA = ψB = ∞.4
ended columns with sidesway have zero critical axial capacity; • The total critical axial load of a frame with sidesway
5) the total critical axial load of a frame with sidesway buckling is buckling is independent of the axial load distribution
independent of the axial load distribution among the columns; and among the columns (ACI 10.11.5.1 and Eq. 10.8).4
6) the stability behavior of framed structures can be analyzed by Unfortunately, these fallacies, which will be discussed
breaking the structure into plane frames and is independent of the thoroughly in this paper, have become intrinsic parts of most
column layout. Because of its extreme importance this last miscon- construction codes3-5 and of the technical literature on struc-
ception is analyzed in a separate paper.1 The first five misconcep- tural analysis and design.6-7 The writer has investigated
tions are thoroughly discussed and their solutions presented in this these fallacies and has come up with a comprehensive and
paper, and a new stability classification for column systems is practical approach to calculate the effective length k-factor
proposed. The proposed criteria and equations not only overcome
for a single column as well as the total critical axial load for
the misconceptions and paradoxes of the classical alignment
charts, but they are more general and simpler to apply than any
an entire story of plane and space framed structures. The
other known available criteria. Examples are presented that writer’s approach is not only free of fallacies and paradoxes
demonstrate the effectiveness and accuracy of the proposed but simple to apply, and most importantly it is reliable. The
criteria. k-factor for partially braced columns was first introduced by
the writer, as well as the criterion for minimum amount of
Keywords: buildings; column design; computer applications; earthquake
interstory lateral stiffness (or diagonal bracing) required by
design; lateral loads; moment magnification; P-Δ effects; reinforced con- frames to achieve a non-sway buckling mode.1,8-11
crete; slenderness effects; structural analysis; stability. The main objective of this paper is threefold: 1) to present
a definitive classification method for single- and multi-
column systems; 2) to discuss the five fallacies with examples
INTRODUCTION
and their solution; and 3) to expose the narrow limits of
The stability analysis and design of vertical members are applicability of the procedure in the ACI Code 318-89.4 The
critical to the structural behavior of RC buildings and author is proposing an accurate and practical classification
bridges, particularly during earthquakes. The stability anal- method for single- and multi-column systems. The proposed
ysis is generally carried out using the k-factor method in method and equations for multi-column systems are particu-
conjunction with the classical alignment charts-nomograms larly suitable for the calculation of the sway-amplification
known as the Julian & Lawrence charts.2 The alignments factor, and the k-factors for each column in framed structures
charts are limited to fully symmetrical rectangular frames with rigid, semi-rigid, and simple connections. The proposed
with rigid/hinged connections (i.e., frames with identical method and equations also avoid confusing and paradoxical
columns under equal axial loads and boundary conditions),
and since their inception forty years ago, fallacies have been
created by their indiscriminate use in all types of framed ACI Structural Journal, V. 94, No. 4, July-August 1997.
Received February 2, 1995, and reviewed under Institute publication policies.
structures, in spite of efforts made to address certain inade- Copyright © 1997, American Concrete Institute. All rights reserved, including the
quacies by the most recent AISC Code.3 The five fallacies to making of copies unless permission is obtained from the copyright proprietors.
Pertinent discussion will be published in the May-June 1998 ACI Structural Journal
be discussed and analyzed are: if received by January 1, 1998.

ACI Structural Journal/July-August 1997 389


STRUCTURAL MODEL
J. Dario Aristizabal-Ochoa is a 125-Year Generation Professor, National University,
School of Mines, Colombia. He is also partner of A-O Ingenieros Asesores, A. A. The linear elastic models of a single column and a multi-
75267 Medellin-Colombia. He was manager of Civil and Seismic Applications for column system (or story system) shown in Fig. 1 and 2 have
MTS Corporation, Mn., and research engineer for the Portland Cement Association,
Skokie, Il. Awarded the ASCE Engineering foundation grant in 1982 and the ACI been previously presented and discussed elsewhere.1,8-11
Reese structural award in 1986 for his work on R/C Structural Walls, Aristizabal- However, they will be presented again for quick reference.
Ochoa was an international speaker for ACI in the First Spanish Seminar on Lateral
Loads on R/C Buildings in 1985 and a past member of various ACI-ASCE technical The multi-column system of Fig. 2(b) consists of n pris-
committees. He has been responsible for the design and construction of numerous matic columns, each one with different cross sectional
buildings in Colombia.
properties (Ai , Ii), height (hi), end flexural restraints (κai,
κbi, whose units are in force-distance/radian), under
different axial load (Pi), and all sharing the same lateral
results that the alignment charts give when dealing with spring restraint SΔ and sidesway Δ. A typical column
framed structures that are not fully symmetrical, frames with element AiBi of the structural system is made up of the
leaning columns, and frames with sidesway partially inhib- column itself Ai ′Bi ′ and the lumped flexural connectors
ited as it has been demonstrated by Cheong Siat-Moy12 and Ai Ai ′ and Bi Bi ′ at the top and bottom ends, respectively. It
the author.1,8-11 The proposed method is not only practical is assumed that a typical column AiBi is made of a homoge-
and “friendly” (since it can be carried out with the usage of neous linear elastic material with: 1) a modulus of elasticity
a pocket calculator), but can be trusted to provide reasonable Ei; 2) straight line centroidal axis; and 3) buckling taking
estimates of the stability behavior of framed structures. place about one of the principal axis of the cross section.
Examples are presented that demonstrate the faults of current The lumped flexural connectors Ai Ai ′ and Bi Bi ′ have stiff-
methods based on the alignment charts and the accuracy and nesses κai and κbi, respectively. The ratios Rai = κai /(EiIi /hi)
effectiveness of the proposed equations. and Rbi = κbi /(EiIi/hi) are denoted as the stiffness indices of
the flexural connections of column i. Where Ii = column’s
RESEARCH SIGNIFICANCE moment of inertia about the principal axis in question, and hi
This publication presents and discusses with examples the
misconceptions in the stability analysis of framed structures = column’s height. These indices vary from zero (i.e., Rai =
using current methods accepted by most construction codes. Rbi = 0) for simple connections (i.e., hinged) to infinity (i.e.,
Solutions to these misconceptions are also presented using a Rai = Rbi = ∞) for fully restrained connections (i.e., rigid).
new classification criteria for columns and frames. The For convenience the following two parameters are intro-
proposed column classification criteria is of great research duced by the author:8
significance because it allows the designer to determine the
stability of buildings including the k-factor for each column, 1
the total critical load, and the sway magnification factor of ρ ai = ----------------- (1)
3
each floor in framed structures. But most importantly, the 1 + -------
R ai
proposed classification and stability method can be trusted
since it is free of paradoxes and fallacies and more general
than any other method. and

Fig. 1—Classification of single-column systems: (a) sidesway uninhibited; (b) sidesway partially inhibited; and (c) sidesway
inhibited.

390 ACI Structural Journal/July-August 1997


1
ρ bi = ----------------- (2)
1 + ------ 3-
R bi

where ρai and ρbi are called the fixity factors at the top and
bottom ends of column Ai Bi. For hinged connections, both
the fixity factor ρi and the rigidity index Ri are zero; but for
rigid connections, the fixity factor is 1 and the rigidity index
is infinity. Since in real structures the fixity factors can only
vary from 0 to 1.0 percent (while the rigidity index Ri may
vary from 0 to ∞), they are more convenient to use in the
analysis of structures with semirigid connections.
The relationships between the fixity factors ρai, ρbi and the
alignment charts ratios ψai and ψbi [i.e., ψ =
∑ ( EI ⁄ h )c ⁄ ∑ ( EI ⁄ L )g at the top and bottom ends, respec-
tively] of column i in a symmetrical rigid frame with side-
sway uninhibited or partially inhibited are: ρai = 2/(2 + ψai)
and ρbi = 2/(2 + ψbi) [see Ref. 8, pp. 1276-1277]. For column
i in a symmetrical rigid frame with sidesway totally inhib-
ited, the relationships are: ρai = 2/(2 + 3ψai), and ρbi = 2/(2
+ 3ψbi). For symmetrical frames with semirigid beam-to-
column connections, the fixity factors ρai and ρbi can be
determined using structural principles shown by the author
elsewhere.8 Finally, for unsymmetrical frames with semi-
rigid connections, their buckled configuration and the corre-
sponding fixity factors ρai and ρbi can be estimated by
applying small lateral loads at each floor level of the frame
as shown in Example 2 of this paper.

PROPOSED CLASSIFICATION CRITERIA AND


EQUATIONS
Stability criteria
In multi-column systems with sidesway uninhibited or
partially inhibited (as shown in Fig. 2[a] and [b], respec-
tively) every column is defined as having reached its critical
load when sidesway buckling of the frame occurs, with the
load distribution among the columns being as specified by
the designer. The effective length k-factor of each column,
then, is defined as that value that yields the appropriate critical
load when applied to the classic Euler’s formula, π2EI/
(kh)2.8 The k-factor so obtained for each column must be
greater than that calculated for the same column but
assuming the story system with sidesway inhibited (Fig. 2[c]).
Thus, the k-factor would be infinity for unloaded columns,
since the critical load, as defined, is zero for those columns.
Obviously, in this concept, effective length k-factor of each
column is a function of its own properties, the properties of
the entire frame (support and bracing conditions), and the
distribution of load among the columns in the frame.
Figures 1 and 2 show the proposed classification for single
and multi-column plane systems, respectively. The proposed
equations are included at the bottom of each case for easy
reference and convenience. The proposed method and equa-
tions are discussed next.

Single column system


Column with sidesway buckling uninhibited—For
columns in which the lateral sidesway Δ between the ends is
totally uninhibited (Fig. 1[a]), Eq. (3) has been proposed8:

2 2 Fig. 2—Classification of multi-column systems: (a) story


2 40 + 8 ( ρ a + ρ b ) + ρ a ρ b ( ρ a ρ b + 3ρ a ρ b – 34 )
k ≅ ------------------------------------------------------------------------------------------------------------
- > 1.0 (3) with sidesway uninhibited; (b) story with sidesway partially
3 ( 4 – ρa ρb ) ( ρa + ρb + ρa ρb ) inhibited; (c) story with sidesway inhibited

ACI Structural Journal/July-August 1997 391


Eq. (3) is equivalent to the classical alignment chart for βi =
(EiIi)/(EjIj) = ratio of flexural stiffness of column i
“unbraced” columns (i.e., with sidesway totally uninhibited) to that of representative column j
and yields results within 1 percent of the theoretical values. γi = hi/hj = ratio of height of column i to that of repre-
Column with sidesway buckling partially inhibited—For sentative column j
columns in which the lateral sidesway between the ends is kj = effective length k-factor of the representative column j
partially inhibited (Fig. 1[b]), Eq. (4) has been proposed:8 Eq. (7) yields results within 1 percent of the values
obtained from the corresponding theoretical transcendental
⎛ 2 2⎞ equation (or alignment chart for sidesway uninhibited) for
40 + 8 ⎝ ρ a + ρ b ⎠ + ρ a ρ b ( ρ a + ρ b + 3ρ a ρ b – 34 ) ( 1 – 0.392ρ a ) ( 1 – 0.392ρ b )
2
k ≅ ------------------------------------------------------------------------------------------------------------------------------------------------- > ------------------------------------------------
2
(4) fully symmetrical rectangular frames. For the extreme cases
5 ( 4 – ρa ρb )
3 ( 4 – ρ a ρ b ) ( ρ a + ρ b + ρ a ρ b ) + ---------------------------------S
3
( 1 + 0.216 ρ a ) ( 1 + 0.216 ρ b )
of fixed-free columns (ρai = 1 and ρbi = 0) and fixed-fixed
Δ h ⁄ ( EI )

2
columns (ρai = ρbi =1) all under the same loading and iden-
tical properties (flexural stiffness EI and height h), the k-
Eq. (4) also yields results within 1 percent of the theoretical factors are exactly equal to the theoretical values of 2.0 and
values. Eq. (3) is a particular case of Eq. (4) when SΔ = 0. 1.0, respectively.
Column with sidesway-buckling inhibited—For a single Multi-column system with sidesway buckling partially
column system in which the lateral sway between the inhibited—For a story system with lateral sway between the
column’s ends is totally inhibited, Newmark13 proposed the columns ends partially inhibited by a spring SΔ (Fig. 2[b]),
following simple equation in terms of the stiffness indices: Eq. (8) has been proposed:11

n
2 2 α
π
⎛ ----- π ∑ ----γi-i 40 + 8 ⎛⎝ ρai + ρbi⎞⎠ + ρai ρbi ( ρai + ρbi + 3ρai ρbi – 34 )
- + 2⎞ ⎛ ------ + 2⎞
2 2 2
⁄ ( 4 – ρ ai ρ bi )

2
⎝R ⎠ ⎝R ⎠ 2 i=1
k j ≅ ------------------------------------------------------------------------------------------------------------------------------------------------------ (8)
a b
k ≅ -----------------------------------------
- (5) 5 β
n
2 2 h ⁄ ( E j I j ) + 3 ∑ ----3i ( ρ ai + ρ bi + ρ ai ρ bi ) ⁄ ( 4 – ρ ai ρ bi )
3
π - + 4⎞ ⎛ -----
⎛ ----- π - + 4⎞ --------S
2 Δ j
⎝R ⎠ ⎝R ⎠ 2π γ
i=1 i
a b

Eq. (7) is a particular case of Eq. (8) when SΔ = 0. The deri-


which in terms of the fixity factors Eq. (5) becomes: vation of Eq. (8) is beyond the scope of this paper, and it is
presented elsewhere1 with further verifications and examples.
2 ( 1 – 0.392ρ a ) ( 1 – 0.392ρ b ) 2 Once the k-factor of the representative j-column has been
k ≅ -------------------------------------------------------------------
- ≥ 0.5 (6) calculated using either Eq. (7) or (8), the critical load of the
( 1 + 0.216ρ a ) ( 1 + 0.216ρ b )
representative j-column and that of the rest of the columns
can be obtained directly from Eq. (9) and (10), respectively.
Eq. (6) yields results with an error less than 2.5 percent of The k-factor of any column can be calculated from Eq. (11).
the values obtained from the corresponding theoretical tran- Then, the total buckling load of the entire floor system can
scendental equation for cases in which Ra +Rb <10 (ρa + ρb be calculated from Eq. (12), and the sway-magnification
<1.5385) and giving maximum errors less than 4 percent in factor from Eq. (13). The use of these equations in the
the range of Ra + Rb from 10 to 20 (1.5385 < ρa + ρb stability of RC framed structures is thoroughly discussed in
<1.7391). The values of k given by these two equations are Ref. 11.
exact at the two extreme cases (ρa= ρb = 0, perfectly pinned;
and ρa = ρb = 1, perfectly fixed). Since the k values from Eq. 2
(5) and (6) are generally less than the true values, the design π Ej Ij
( P cr ) j = ---------------
- (9)
results are slightly unconservative.13 2
( kj hj )
Multi-column system
To carry out the stability analysis of the multi-column (Pcr)i = αi(Pcr)j (10)
plane system shown in Fig. 2(a) through (c), a representative
column is first selected from the n-set of columns. This is 2 2 2
denoted as the j-column. The column under the largest axial k i = [ β i ⁄ ( α i γ i ) ]k j (11)
load stress is generally the one selected.
Multi-column system with sidesway-buckling uninhib- n
ited—For a story system in which the lateral sway between
the columns’ ends is uninhibited (Fig. 2[a]), Eq. (7) has been
( P cr ) Total = ( P cr ) j ∑ αi (12)
proposed11: i=1

n
α 2
1 - ≥ 1.0
δ s = ------------------------------------------------------------------- (13)
∑ ----γi-i 40 + 8 ⎛⎝ ρai + ρbi⎞⎠ + ρai ρbi ( ρai + ρbi + 3ρai ρbi – 34 )
2 2
⁄ ( 4 – ρ ai ρ bi )
⎛ n ⎞
2
k j ≅ i----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
=1
n
βi
(7) 1–⎜
⎝i = 1 ⎠

P iu⎟ ⁄ ( ΦP cr )
total
3 ∑ ----- ( ρ ai + ρ bi + ρ ai ρ bi ) ⁄ ( 4 – ρ ai ρ bi )
3
i = 1 γi n

where ∑
P
is the factored axial load on the entire floor
i=1
iu

where under consideration, and Φ is the understrength factor for


αi = Pi/Pj = ratio of axial load of column i to that of (Pcr)Total. The ACI Code4 specifies Φ = 0.7 and 0.75 for tied
representative column j and spiral columns, respectively.

392 ACI Structural Journal/July-August 1997


It must be noticed that the results calculated from Eq. (9)
and (10) will be of practical interest only when the Pcr of
each individual column is smaller than the critical load
corresponding to the braced case, because otherwise the
story system (Fig. 2[b]) becomes unstable owing to indi-
vidual column buckling (i.e., localized buckling) and not as
a result of the overall sidesway Δ. Therefore, the k-factor of
each column obtained from Eq. (8) and Eq. (11) must be
larger or equal to the braced column case (i.e., story with
sidesway inhibited) given approximately by Eq. (14) or more
accurately by the classical alignment chart for braced
columns.

( 1 – 0.392ρ ai ) ( 1 – 0.392ρ bi )
Fig. 3—Example 1: Asymmetrical bent frame.
2
k i ≅ ----------------------------------------------------------------------
- (14)
( 1 + 0.216ρ ai ) ( 1 + 0.216ρ bi )
makes this situation even worse is that by underdesigning
Story system with sidesway buckling totally inhibited—For column 2 by 1.96 (i.e. EI2 = EI1/1.96), the frame becomes a
a story system in which the lateral sway between the columns’ frame with sidesway-buckling uninhibited with k1 ≅ 0.83 (or
ends is totally inhibited (Fig. 2[c]), Eq. (14) presented previ- Pcr1 = Pcr2 = π2EI/(0.83h)2 = 1.44π2EI/h2) resulting in a net
ously for a single column system is proposed by the author. underdesign of 4/1.44 = 2.77 for column 2. These results
indicate that the classification of frames as braced or
FIVE FALLACIES: SOLUTIONS AND EXAMPLES unbraced, based purely on the frame’s geometry, is not
First fallacy correct. In addition, the calculation of the k-factors for the
Columns and frames are classified into braced and columns based on the alignment charts or “judgmental” k-
unbraced cases only. factors should be avoided because it might result in deficient
Solution—Single- and multiple-column systems should be designs.
classified according to the buckling mode and the magnitude
of the interstory lateral bracing into one of the following Second fallacy
three categories: 1) systems with sidesway-buckling totally The elastic critical axial load and the k-factor for a column
uninhibited (Fig. 1[a]/2[a]); 2) systems with sidesway-buck- which is part of a framed structure is a function of its ends’
ling partially inhibited (Fig. 1[b]/2[b]); and 3) systems with flexural conditions only (i.e., ψtop and ψbottom).
sidesway-buckling inhibited (Fig. 1[c]/2[c]). Solution—This is only true for totally symmetrical multi-
Example 1—Consider the simple asymmetrical frame column systems (i.e., systems with identical columns, under
shown in Fig. 3. To simplify calculations and for conve- the same axial loads and boundary conditions). Otherwise,
nience, assume that the girder’s flexural stiffness EI/L is infi- the columns in a multi-column system are coupled together.
nitely larger than that of the columns. This frame appears to It is wrong to assume that each column has its own critical
be a frame with sidesway-buckling uninhibited. However, load and that the total critical load of the system is the sum
this classification needs to be checked according to Eq. (7) of the critical load of each individual column. To find the
as follows: correct critical load of the system, the k-factor for the column
Selecting column 1 as the representative column, then: carrying the largest axial load (j-column) should be evalu-
• For column AB: α1 = 1.0; β1 = 1.0; γ1 = 1.0; ρa1 = 1.0; ated first utilizing Eq. (7) or (8). The magnitude of kj is a
ρb1 = 0 function of the loading pattern (αi), stiffness ratios (βi),
• For column CD: α2 = 1.0; β2 = 1.0; γ2 = 0.5; ρa2 = 1.0; height ratios (γi), fixity factors (ρai, ρbi), and the relative total
ρb2 = 1.0 lateral bracing provided at that inter-story level, SΔ/(EjIj /hj3).
3 + 6 - = 4 ⁄ 11
Furthermore, kj should be less than that calculated assuming
• then from Eq. (7): k 21 ≅ -------------------------- sidesway inhibited according to Eq. (6) or (14), otherwise
( 3 ⁄ 4 ) + 24
Therefore k1 ≅ 0.60; however, this value is less than the the multi-column system (or story system) becomes a system
value corresponding to a column with sidesway inhibited (k with sidesway inhibited.
= 0.70). Therefore, for the given loading conditions, geom- This fallacy is perhaps the most troublesome of them all.
etry, flexural stiffnesses, and boundary conditions, this is a Example 2—Consider the rigidly-connected one-story
frame with sidesway-buckling inhibited, and as a conse- unsymmetrical frame shown in Fig. 4(a). Assume L = 40 ft,
quence Pcr1 = Pcr2 = π2EI/(0.7h)2 = 2.04π2EI/h2. Had the I = 10,000 in.4 and E = 3000 ksi, and neglect axial elonga-
height of column 2 been 0.6h instead of 0.5h, the frame of tions of all three members. This frame would be considered
Fig. 3 would have been classified as a frame with sidesway- a frame with sidesway-buckling uninhibited by most
buckling uninhibited (because in such a case k1 ≅ 0.74 > designers with the k-factors obtained from the alignment
0.70). Notice that k1 < 1.0 in both cases, in spite of the frame, chart for “unbraced” frames as follows:
appears to be “unbraced.” • For column AB: ψtop = 0.4; and ψbottom = 0: then kAB =
1.10
It is interesting to note that according to the alignment
chart for sidesway uninhibited k1 = 2.0 (or Pcr1 = π2EI/4h2) • For column CD: ψtop = 0.4; and ψbottom = ∞: then kCD
and k2 = 1.0 (or Pcr2 = 4π2EI/h2). This would result in an = 2.10
extremely conservative design for column 1 (overdesigned According to Eq. (7), however, the fixities factors must be
by a factor 4 x 2.04 = 8.16), but very unconservative for established first in order to determine the k-factors. These
column 2 (underdesigned by a factor of 4/2.04 = 1.96). What factors depend on the buckling mode or configuration of the
frame which can only be determined “exactly” from the

ACI Structural Journal/July-August 1997 393


Fig. 4—Example 2: One-story frame: (a) frame model; (b)
variation of k of both columns with α.

Fig. 5—Example 2 (modified): one-story asymmetrical


frame: (a) frame model; and (b) variation of k of both
eigen-value solution, but can be approximated by applying a
columns with α.
unit horizontal load at node A (or D) and finding the
moments and rotations at joints A and D of the frame with
sidesway uninhibited. This was accomplished using a Therefore, the stiffness of the flexural connector at A is κa
conventional first-order analysis computer program for ≅ 0.3312/(3.312 × 10–6) = 10,000 kip-in./radian; and the
plane frames with the following results: stiffness index at A is Ra = κa/(EI/h)AB = 10,000/(3000 ×
1.) For column AB: rotation at A = 6.24 × 10–4 radians; and 4000/480) = 4; then using Eq. (1), the fixity factor at A
Moment at A = 170.84 kip-in. becomes ρa ≅ 1/(1 + 3/4) = 4/7. The fixity factor at B: ρd ≅
Therefore, the stiffness of the flexural connector at A is κa 1/(1 + 3/∞) = 1.0.
2
≅ 170.84/(6.24–4) = 273,777 kip-in./radian; and the stiffness Therefore, using Eq. (14): k AB with sway totally inhibited
index at A is Ra = κa /(EI/h)AB = 273,777/(3000 × 4000/480) ( 1 – 0.392 × 4 ⁄ 7 ) ( 1 – 0.392 × 1 )
≅ --------------------------------------------------------------------------------
- = 0.345 ; or (kAB) with sway totally
( 1 + 0.216 × 4 ⁄ 7 ) ( 1 + 0.216 × 1 )
= 10.953; then using Eq. (1), the fixity factor at A becomes inhibited ≅ 0.59.
ρa ≅ 1/(1 + 3/10.953) = 0.785. The fixity factor at B: ρb ≅
1/(1 + 3/∞) = 1.0. 2) For column CD: rotation at D = 1.273 × 10–5 radians;
2.) For column CD: Rotation at D =1.194–4 radians; and and moment at D = 0.0955 kip-in.
Moment at D =107.70 kip-in. Therefore, the stiffness of the flexural connector at D is κd
Therefore, the stiffness of the flexural connector at D is κd ≅ 0.0955/(1.273 × 10–5) = 75,000 kip-in./radian; and the
≅ 107.70/(1.19 × 10–4) = 905,405 kip-in./radian; and the stiffness index at D is Rd = κd/(EI/h)CD = 75,000/(3000 ×
stiffness index at D is Rd = κd/(EI/h)CD = 905,045/(3000 × 4000/480) = 3; then using Eq. (1), the fixity factor at D
4000/480) = 36.232; then using Eq. (1), the fixity factor at D becomes: ρd ≅ 1/(1 + 3/3) = 0.5. The fixity factor at C: ρc ≅
becomes ρd ≅ 1/(1 + 3/36.232) = 0.924. The fixity factor at 1/(1 + 3/0) = 0.
2
C: ρc ≅ 1/(1 + 3/0) = 0. Therefore, using Eq. (14): k CD with sway totally inhib-
( 1 – 0.392 × 0.5 ) ( 1 – 0.392 × 0 )
Once the fixity factors at the top and bottom of each column ited≅ ------------------------------------------------------------------------------- = 0.726
( 1 + 0.216 × 0.5 ) ( 1 + 0.216 × 0 )
are determined and the j-column is selected (column AB in this kCD with sway totally inhibited≅0.85
case), the k-factor can be calculated using Eq. (7) as follows: The values of kAB and kCD and their approximate lower
2
limits are plotted in Fig. 4(b). It can be concluded that: 1. this
k AB ≅ [-------------------------------------------------------------------------------------------------------------------------------------------
40 + 8 ( 1 + 0.785 ) + 0.785 ( 0.785 + 1 + 3 × 0.785 – 34 ) ]
2
particular frame will buckle with sidesway under any
3 ( 1 + 0.785 + 0.785 ) ⁄ ( 4 – 0.785 )
compressive axial load combination (i.e., any α-value); 2. if
2 2 2
the designer uses the alignment charts, columns AB and CD
⁄ ( 4 – 0.785 ) + α [ 40 + 8 × 0.924 ] ⁄ 4
-------------------------------------------------------------------------------------------------- would be underdesigned for load combinations corre-
+ 3 ( 1 + 0.924 ) ⁄ 4
sponding to α > 0.3 and α < 0.25, respectively. Now, by
reducing the height of column CD to 0.5h as shown in Fig. 5(a)
therefore k AB ≅ ( 0.923 + 0.947α ) ≤ ( k AB ) braced , and and maintaining everything equal, the k-factor of the short
column CD is increased substantially, whereas that of the
k CD ≅ 0.923 ⁄ ( α + 0.947 ) ≤ ( k CD ) braced
long column AB is reduced significantly as shown in Fig. 5(b).
Each of these two k values has its own lower limit as indicated Notice the kAB < 1 for values of α < 0.69. These results,
above. These lower limits are obtained from Eq. (6) assuming again, indicate that it is incorrect to design unsymmetrical
that the frame is “braced” (i.e., with sidesway totally inhibited frames using the ACI Code 318-894 resulting in the short
at the top). To calculate these two limits, the fixity factors for columns being under-designed and the long ones over-
each column must be determined again assuming that the frame designed. It is wrong to assume that each column has its own
is “braced” (i.e., zero sidesway at the top joints A and D). An independent critical load. In reality, all columns are coupled
additional structural analysis was carried out with a unit together.
moment at D and the frame restrained along the horizontal
direction AD at the top, yielding the following results: Third fallacy
1) For column AB: rotation at A = 3.312 × 10–6 radians; The k-factor for columns with lateral sidesway shall be
and Moment at A = 0.3312 kip-in. greater than 1.

394 ACI Structural Journal/July-August 1997


Solution—The effective length k-factor for columns with
lateral sidesway partially and totally inhibited shall be
greater than 1/2 and columns with lateral sidesway uninhib-
ited shall be greater than 1.
Example 3—Consider the simple rigid frame shown in
Fig. 6(a). Most designers using current criteria would
consider this frame as a frame with sidesway-buckling unin-
hibited and k = 1 for both columns. This is correct as long as
both columns are identical to each other and subject to iden-
tical axial loads. However, once the symmetry is broken, the
k-factors become different for each column, and for signifi-
cant differences between the two columns either in axial
loading, moments of inertia or height, the buckling configu-
ration changes dramatically. For instance, when one of the
columns has a larger moment of inertia, the k-factor for the
strong column increases (k > 1.0) and that for the weak
column decreases (k < 1) as shown in Fig. 6(b). The buckling
modes change when the ratio β = (EI)2/(EI)1 is less than 1/7
or greater than 7 as shown in Fig. 6(b). The variations of the
k-factors for each column and the buckling modes with the
ratios α = P2/P1 and γ = h2/h1 are shown in Fig. 6(c) and (d).
Using Eq. (7) the k-factors for columns 1 and 2 are as
1+α⁄γ 1+γ⁄α
follows: k 1 ≅ ---------------------- ≥ 0.5 and k 2 ≅ ---------------------- ≥ 0.5 .
3 3
1+β⁄γ 1+γ ⁄β

Fourth fallacy
Pin-ended columns (or leaning columns) in frames with
sidesway have zero axial load capacity.
Solution—The effective length k-factor for leaning
columns in frames with sidesway buckling can be obtained
directly from Eq. (7) or (8) and it should always be greater
than 1. Fig. 6—Example 3: Asymmetrical bent frame: (a) frame
Example 4—Consider the simple rigid frame shown in model; (b) variation of k of both columns with β = (EI)2/
Fig. 7. Most designers using current criteria would consider (EI)1; (c) variation of k of both columns with α = P2/P1;
this frame as a frame with sidesway-buckling uninhibited and (d) variation of k of both columns with γ = h2/h1.
and k1= ∞ and k2 = 1.0. Using Eq. (7) the k-factor for the
leaning column can be obtained directly as follows:
for each case can be obtained directly from Eq. (7) and Eq.
1 + 1.2α ⁄ γ
k 1 ≅ --------------------------- ≥ 1.0 (9) through (11) as follows:
3
1.2β ⁄ γ
n

The k-factor for leaning columns in framed structures is ∑ αi ⁄ γi


fully discussed by the author elsewhere.9 Case 1) from Eq. (7): k 2j ≅ --------------------
i=1
n
- ≥ 0.5
2

∑ βi ⁄ γ
3

i=1
Fifth fallacy
The total critical axial load of a frame with sidesway buckling n n
is independent of the axial load distribution among the ∑ βi ⁄ γ ∑ αi
3

columns. from Eq. (11): ( P cr ) Total ≅ ----------------------------------π


i=1 i=1 2 2
Ej Ij ⁄ hj (15)
n
Solution—This is only true for multi-column systems with
identical end fixities and heights in all columns and with ∑ αi ⁄ γi
i=1
sidesway uninhibited. Otherwise, the total axial capacity of
a multi-column system with sidesway partially inhibited or n
uninhibited depends on the load distribution αi, stiffness βi, 4 ∑ αi ⁄ γi
and height ratios γi. The total critical axial load of a multi- Case 2) from Eq. (7): k 2j ≅ -----------------------
i=1 - ≥ 0.5
2
n
column system with sidesway buckling can be obtained
∑ βi ⁄ γ
3

directly using Eq. (7) or (8) and Eq. (9) through (11) as i=1
explained previously.
Example 5—Consider the model of the multi-column n n

∑ βi ⁄ γ ∑ αi
3
system in Fig. 2(a) with sidesway uninhibited. Assume two
2 2
cases: 1.) all columns having clamped conditions at both from (11): ( P cr ) Total ≅ ----------------------------------π
i=1
n
i=1 Ej Ij ⁄ hj (16)
ends (Fig. 8[a]); and 2.) all columns having one end clamped 4 ∑ αi ⁄ γi
and the other hinged (Fig. 8[b]). The total critical axial load i=1

ACI Structural Journal/July-August 1997 395


CONCLUSIONS
A powerful new method has been developed to classify
column systems and frames. This criteria is based on the type
of sidesway-buckling mode and the presence of interstory
bracing. The proposed method and equations for multi-
column systems (story systems) are particularly suitable for
the calculation of the sway-amplification factor, and
columns’ k-factors for framed structures with rigid, semi-
rigid and simple connections as it has been shown in
previous publications.1,8-11 The proposed classification
method and equations also avoid confusing and erroneous
results that the alignment charts give for framed structures
that are not fully symmetrical, frames with leaning columns,
Fig. 7—Example 4: Asymmetrical bent frame with a leaning and frames with sidesway partially inhibited. The proposed
column. method is not only practical and “friendly” (since it can be
carried out with the usage of a pocket calculator), but it can
be trusted to provide reasonable estimates within 1 percent
of the true stability behavior of framed structures. Numerous
examples are presented that expose the narrow limits of the
applicability of the procedure in the current ACI Code4 and
the accuracy and effectiveness of the proposed criteria and
equations.
Analytical studies indicate that the use of the alignment
charts in the design of framed structures with columns of
different heights gives faulty designs. This is vital in the
design of highway bridges (at interchanges) and buildings on
slopes (or with different column heights). While the alignment
chart for unbraced cases indicates that the long columns have
the lowest critical load and the short columns have the largest,
the proposed model indicates opposite behavior (see Exam-
ples 1, 2, 3, and 5). For systems with sidesway uninhibited
(Fig. 2[a]) and partially inhibited (Fig. 2[b]), all columns in the
system are coupled together by the lateral sway Δ. It is wrong
to assume that each column has its own independent critical
load and that the total critical load of the system is the sum of
these critical loads. In unsymmetrical frames this assumption
gives faulty designs. Single- and multiple-column systems
should be classified according to the buckling mode and the
provided interstory lateral bracing as proposed by the author
and not according to frame “looks” or engineering “judgment”
as is being promoted by most codes and the technical litera-
ture.
According to the proposed method the k-factors for
columns under light axial loads are very large, becoming
k= ∞ when the applied axial load is zero. Therefore, the
limit imposed by the ACI Code 318-89 to the slenderness
ratio kh/r to all vertical members is not realistic (see ACI
10.11.4).4 In a multi-column system the ratio kh/r for the
columns under large axial load stress shall not exceed a
specified limit, and those columns under relatively low
axial stress should have a limited ratio of (h/r) instead.

NOTATION
E = Young’s modulus of the material
hi = height of column i
Fig. 8—Example 5: Multi-column systems: (a) columns with
clamped-clamped ends; and (b) columns with hinged- Ig = girder moment of inertia
clamped ends. Ii or Ic = column’s moment of the inertia
κai and κbi = the flexural stiffness of the end connections at Ai and
Bi, respectively
Notice that the only way that in both cases (Pcr)Total would kj = effective length factor of the representative column j
be independent of the loading ratios αi is when all columns Lg = girder span
have the same height (i.e., γi = 1). It is interesting to note that n = total number of columns in the story system
the total critical load of frame of Fig. 8(a) is four times that (Pcr)j = buckling load of representative column j [= π2EjIj/(kjhj)2]
of Fig. 8(b). (Pcr)I = buckling load of column i [=αi(Pcr)j]

396 ACI Structural Journal/July-August 1997


n 3. Manual of Steel Construction, Load & Resistance Factor Design,
(Pcr)Total = critical load of the entire story [= (P )
cr j ∑ α
i
] second edition, American Institute of Steel Construction (AISC), 1994.
i=1
4. ACI Committee 318, “Building Code Requirements for Reinforced
Pu = total factor axial load at the story system
Concrete and Commentary (ACI 318-89/31R-89 Revised 1992),” Amer-
SΔ = interstory lateral stiffness or bracing provided at the top ican Concrete Institute, Detroit, 1992, 347 pp.
of the story system
5. Manual of Steel Construction, Allowable Stress Design, ninth edition,
Rai = stiffness index of the flexural connection at Ai [=κai/(EI/hi)]
American Institute of Steel Construction (AISC), 1990, Part 3.
Rbi = stiffness index of the flexural connection at Bi [= κbi/
6. Salmon, C. G., and Johnson J. E., Steel Structures: Design and
(EIi/hi)]
Behavior, third edition, Harper & Row, New York, 1990.
αi = ratio of axial load of column i to that of representative
7. Wang, C.-K., and Salmon, C. G., Reinforced Concrete Design,
column j [= Pi /Pj]
fourth edition, Harper and Collins, 1992.
βi = ratio of flexural stiffness of column i to that of represen-
tative column j [= (EiIi)/(EjIj)] 8. Aristizabal-Ochoa, J. Dario, “K-Factor for Columns in any Type of
Construction: Nonparadoxical Approach,” Journal of Structural Engi-
γi = ratio of height of column i to that of representative col-
neering, ASCE, V. 120, No. 4, 1994, pp. 1272-1290.
umn j[= hi/hj]
9. Aristizabal-Ochoa, J. Dario, “Slenderness K-factor for Leaning
δs = sidesway amplification factor at top of story system {=
Columns,” Journal of Structural Engineering, ASCE, V. 120, No. 10,
1/[1-Pu/(φPcr)Total]}
1994, pp. 2977-2991.
ρai and ρbi = fixity factors at Ai and Bi, respectively
10. Aristizabal-Ochoa, J. D., “Stability of Columns under Uniform
Φ = critical load understrength factor specified by the ACI
Load with Semirigid Connections,” Journal of Structural Engineering,
Code (<1)
ASCE, V. 120, No. 11, 1994, pp. 3212-3222.
ψai and ψai = ∑ ( EI ⁄ h )-c at ends Ai and Bi, respectively
-------------------------- 11. Aristizabal-Ochoa, J. D., “Story Stability and Minimum Bracing in
∑ ( EI ⁄ L )g RC Framed Structures: General Approach,” ACI Structural Journal, V. 92,
No. 6, Nov.-Dec. 1995, pp. 735-744.
REFERENCES 12. Cheong-Siat-Moy, F., “K-Factor Paradox,” Journal of Structural
1. Aristizabal-Ochoa, J. D., “Amplification Factor for Three-Dimen- Engineering, ASCE, V. 112, No. 8, 1986, pp. 1747-1760.
sional RC Framed Structures: A Nonparadoxical Approach,” ACI Struc- 13. Newmark, N. M., “A Simple Approximate Formula for Effective
tural Journal, V. 94, No. 5, Sept.-Oct. 1997, pp. 538-548. Fixity of Columns,” Journal of Aeronautical Sciences, V. 16, No. 2, Feb.,
2. Kavanagh, T. C., “Effective Length of Framed Columns,” ASCE pp. 116. (or ASCE Civil Engineering Classics: Nathan Newmark, 1976,
Transactions, Paper No. 3366, V. 127, Part II, 1962, pp. 81-101. p. 439), 1949.

ACI Structural Journal/July-August 1997 397

You might also like