Engineering Structures 203 (2020) 109804

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: [Link]/locate/engstruct

New and extended design moment formulations for slender columns in T

frames with sway
Jostein Hellesland
Mechanics Division, Department of Mathematics, University of Oslo, P.O. Box 1053 – Blindern, NO-0316 Oslo, Norway

A R T I C LE I N FO A B S T R A C T

Keywords: Present first-order based design code formulations for slender columns in frames with sway, employ sway
Buckling magnification factors (for global second-order effects) also as moment magnifiers for the individual columns of
Columns (supports) the frame (storey). This approach ignores differences in magnification of individual column moments caused by
Design local second-order effects in the columns. This difference can be significant. Better understanding of this aspect
Elasticity
will strengthen approximate first-order based elastic methods, for which the important superposition principle is
Frames
Stability
valid. Towards this end, local second-order effects are considered for shears, end moments and maximum mo-
Second-order theory ments, applicable over the full range of axial loads. Specifically, end moment and maximum moment expressions
Sidesway of columns with sway are derived. These represent novel contributions that are suitable in typical design code
Design moment formulations formats, and in practical design work. They will allow more rational column assessments, and will allow more
Structural engineering economical designs than present structural code procedures. Proposals are verified by comparison with results
from second-order in-plane elastic analyses of single restrained columns and columns in frame panels. Also,
extensions to the general case of load combinations that include both gravity and sideways loading are briefly
presented.

1. Introduction major design codes (e.g., ACI [1], AISC [3], Eurocode 2 [2], Eurocode 3
[4], etc.), does not presently reflect the fact that the sidesway actually
Elastic second-order frame analysis methods are important in may affect the column moments differently in the different columns on
practical design for bending and stability of both reinforced concrete the same level of the frame. Normally, the larger end moment increase
and steel columns (beam-columns). For reinforced concrete structures, will be smaller than that reflected by the sway magnification factor.
effects on section stiffnesses of nonlinear material behaviour, cracking, Some, but limited, attention to this matter can be found in the lit-
creep, etc., are generally accounted for through the use of reduced, erature. The AISC Commentary dealt with a reduction in the maximum
“effective” linear (elastic) section stiffnesses (e.g., [1,2]). For steel moment due to sidesway already in 1969, and later in the 1978 edition
structures, nominal section stiffnesses are often used (e.g., [3] App. 8, of the Commentary. The same reduction factor expression was sug-
[4]), or reduced stiffnesses that reflect yielding and other effects (e.g., gested for the sway action of all columns of the storey, and it was al-
[3]). Possible local, cross-section dependent instability effects, ductility, lowed to approximate it by a constant factor (0.85). Both the expression
inelasticity and out-of-plane buckling phenomena, such as covered for and the constant are unconservative in the general case. In later com-
instance in [5–8], are not considered in such methods themselves. mentary revisions, this reduction factor is omitted. Hellesland [9], as
Sidesway due to lateral and vertical loading of reinforced concrete recapitulated in a slightly different form elsewhere [10], derived an
and steel structures can be handled by existing second-order analysis expression for end moments in individual column axes for special cases.
computer programs, and also by approximate methods, that have been, At about the same time, LeMessurier [11] derived a similar expression
and still are, important in parallel with, or as a complement to, more and proposed an extension to a general column of a sway frame.
exact methods. In approximate, elastic based methods for frame struc- However, the extension is incorrect in that it yields the same moment
tures, global (overturning) second-order effects of the vertical loads are magnifier for moments at both ends of all columns of the storey. Hel-
often accounted for by a storey sway magnification factor that are ap- lesland and MacGregor (Mechanics and design of columns in sway
plied to the first-order end moments of the columns of the frame (or frames, 1981, unpublished draft) presented an extension of the previous
storeys of the frame). work by Hellesland. Lui [12] also suggested an approach allowing for
Such approximate procedures, which is in line with practice in different moment magnification factors in different column axes. The

E-mail address: josteinh@[Link].

[Link]
Received 23 May 2019; Received in revised form 11 October 2019; Accepted 14 October 2019
Available online 04 November 2019
0141-0296/ © 2019 The Author. Published by Elsevier Ltd. This is an open access article under the CC BY license
([Link]
J. Hellesland Engineering Structures 203 (2020) 109804

Nomenclature SB lateral stiffness of external bracing(s)

V0, V first-order and total (first + second-order) shear force in a
Bb approximate rising moment branch factor column, respectively
Bm approximate maximum moment magnification factor kj rotational restraint stiffness (spring stiffness) at end j
Bmax exact maximum moment magnification factor moments αb , α s load index of column considered fully braced, and free-to-
Bs system (storey) sway magnification factor sway, respectively
Bv shear multiplication factor αss system (storey) stability index
B1, B2 end moment multiplication factors αE nominal axial load index of a column (= N / NE )
EI , EIb cross-sectional stiffness of columns, and of beams β effective (buckling) length factor
Gj scaled rotational restraint flexibility at member end j βb, βs effective length factor corresponding to Ncb and Ncs , re-
L, Lb length of columns, and of restraining beam(s) spectively
M0j , Mj moment from first-order and second-order analysis, re- Δ0 , Δ first-order and total lateral displacement, respectively
spectively, at end j γn axial load dependent flexibility factor
N axial (normal) force γs, γ0 flexibility factor at free-sway and at zero axial load, re-
Ncb, Ncs critical load of columns considered fully braced, and free- spectively
to-sway, respectively κj relative rotational restraint stiffness at end j (=kj /(EI / L) ).
NE Euler buckling load of pin-ended column (= π 2EI / L2 )

approach does not seem to be well founded, and breaks down in the has its largest absolute value.
general case. Better understanding of this aspect would strengthen the Analytical expressions will be established for all of these B factors.
base for approximate first-order based elastic methods, for which the Prior to this, consider some basic cases with reference to Fig. 1(b).
important superposition principle is valid. Column 1, with no axial force, is affected by the global second-order
The main attention of the present paper is directed towards devel- effects, but has no local second-order effects. Thus, local B factors are
oping approximate shear and moment formulations, suitable in design all equal to unity. The column has a positive shear, implying that it
code formats, for slender elastic columns in frames with sidesway. The contributes to the bracing of the frame. The moment diagram stays
full range of axial loads are considered, thus covering both supporting linear, and the final moments are directly proportional to the sway
and supported columns. More specifically, the main objectives are (1) displacement Δ = Bs Δ0 .
to derive appropriate multiplication factors that account for global and Although its moment diagram varies linearly along the axis, it does
local second-order elastic effects on column end moments, and, not not comply with the definition of a first-order moment (“obtained from
least, (2) to establish improved, more realistic, magnification factors for equilibrium based on the undeformed geometry”), since it includes
the prediction of maximum moments between column ends. Towards global second-order effects through Bs . For the sake of precision and
this end, some basic column mechanics are reviewed and exact second- distinction, Bs M0 is labelled “sway-magnified first-order moment” [30].
order analysis results are presented for evaluation of the proposals. The Column 2 has moderate local second-order effects. All local B fac-
scope is limited to elastic bending in the plane of loading. tors, but for Bmax , are different from unity. It has a positive shear, and
thus it contributes to lateral frame support (bracing).
2. Mechanics of moment formation and definitions Column 3 has significant local second-order effects, and the

Fig. 1(a) shows an unbraced, laterally loaded three bay frame sub-
jected a lateral load (H) and vertical loads (Pi ). The load H, alone, gives
rise to a first-order sway displacement (Δ 0 ), equal in all axes when axial
beam deformations are negligible, and to first-order column shears (V0i )
that are proportional to the relative lateral stiffness of the columns. The
effect of the axial loading is to increase the sway displacement beyond
Δ 0 to Δ = Bs Δ 0 , and to cause a redistribution of the shears from their
first-order values to their final values (Vi ) in the respective column axes
i. Bs is the system (storey) sway displacement magnifier, that reflects
global second-order (overturning “N Δ ”) effects.
The formation of moments in the columns will depend on the axial
loading (local second-order effects). Fig. 1(b) shows possible moment
distributions along the individual column axes. In the general case,
shears (V), end moments (M1 and M2 ) and maximum column moments
Mmax can be given in terms of first-order quantities (identified by sub-
scripts zero) as follows:
V = B v (Bs V0) (1)

M1 = B1 (Bs M01) (2)

M2 = B2 (Bs M0j ) (3)

Mmax = Bmax (Bs M02) (4)

While the same global second-order Bs factor is applied to all columns,

the factors B v , B1, B2 and Bmax , are individual column factors, different
in the different axes. Above, it is chosen to express Mmax in terms of M02 Fig. 1. (a) Laterally loaded, unbraced three-bay frame; (b) Possible moment
at end 2, which is defined as the end at which the first-order moment diagrams in laterally supporting and supported columns.

2
J. Hellesland Engineering Structures 203 (2020) 109804

maximum moment may form away from the column end. All local B respectively, and βs and βb are the corresponding effective length
factors are different from unity. Its shear force is negative, implying (buckling) factors. They are computed for the columns considered in
that it requires lateral support. isolation with representative restraints. Except when the frame consists
The straight, pin-ended Column 4 in the figure, with negative of a single column, these are strictly pseudo-critical loads, that are
shears, requires lateral support. It contributes to the overturning global useful in column characterisation and discussion. NE is the socalled
second-order effects, but has no local second-order effects itself. In Euler load (critical load of a pin-ended column), which is a convenient
practice, though, moments will be introduced through imperfect hinges reference load parameter in several contexts. As defined above, the load
and through imperfections along the axis. indices are interrelated. For instance, αs = αE βs2, αb = αE βb2 ,
αb = αs (βb/ βs )2 .
3. Global second-order effects
4.2. Local second-order analysis
It is appropriate for the sake of perspective and completeness, and
useful for later discussions, with a brief review of a suitable approx- For the verification of the proposals of the present paper, results
imate storey (system) sway magnifier. With reference to Fig. 2, where were obtained using the second-order analyses presented in detail in
beam deformations are assumed to be negligible, Bs can be defined by Hellesland [13]. Columns considered are either single restrained col-
umns, as shown in Fig. 3(b), or columns that are part of a panel frame,
1 such as shown in Fig. 3(c). The columns are initially straight and have
Bs =
1 − αss (5) lengths L and uniform section stiffnesses EI. The deflection shapes in-
where αss is the storey (system) stability index defined by dicated by the solid lines in the figure, are those due to an initial, im-
posed top (joint) displacement Δ = Bs Δ0 . For this Δ to remain constant
∑ (γn N / L) ∑ (γn N / L) for increasing axial loading, the column shears V (lateral loading) will
αss = or αss =
(H /Δ0) ∑ (γs Ncs / L) + SB (6a,b) have to decrease (see Fig. 3(b)) to compensate for the increased over-
turning effect of the vertical loading acting on the relative joint dis-
Here, the ratio H/Δ0 is the first-order lateral storey stiffness (including
placement. No gravity load induced moments (such as from loading on
possible external bracing SB ). In multistorey structures, Δ0 is the in-
beams) are included. The dashed lines are deflection shapes developing
terstorey first-order displacement and H the corresponding storey shear.
as the critical axial loading is approached. For the panel, other de-
The γ factors, reflecting local second-order effects, will be discussed
flection shapes may result depending on the relative stiffness and axial
later (General shear formulation). The summations are over all inter-
loading in the members.
acting columns. Ncs is the pseudo-critical load defined below. Provided
The restrained single column may be the complete structure, or it
Ncs is determined with the same first-order restraints implicit in the
can be considered isolated from the two-column panel, or from a
H/Δ0 calculation, the two Bs expressions above are equivalent.
greater frame. In the latter case, the rotational end restraints should
These forms were presented by Hellesland [10]. With the γn in-
reflect the rotational interaction at the joints with restraining beams
cluded, sway-braced column interaction can be accounted for. Several
(“horizontal interaction”), and possible other columns framing into the
authors and codes give similar sway magnifiers. They can all be derived
considered joint (“vertical interaction”). Procedures for this are given in
from simplifications of the αss in Eq. (6a,b), as discussed in [10].
structural codes, and discussed in the literature (e.g. [10,16]).
Common, and often justified, simplifications are constant column
The rotational end restraints can conveniently be represented by
lengths, neglect or introduction of constant values for γn and γs (typi-
rotational restraint stiffnesses (or spring stiffnesses) labelled k1 and k2
cally about 1.15).
(equal to the moments required to give a unit rotation), or in non-
dimensional form by κj at end j = 1 and j = 2 defined by
4. Local second-order effects and analyses
kj
κj = j = 1, 2
4.1. Second-order effects and load indices (EI / L) (9)
In the present paper, end restraints are due to beams. Then, kj is equal
The local member second-order (Nδ ) effects can be quantified by the to the rotational stiffness kbj of all beams connected to the joint con-
ratios of results obtained from second-order analyses of a column with a sidered. Thus,
specified sidesway Δ = Bs Δ0 and given axial loads N and results from
the same analyses for N = 0 (first-order, determined from an analysis kj = kbj = ( ∑ bEIb/ Lb)j (10)
that is based on the original, geometry).
where b is the bending stiffness coefficient, typically equal to 3, 4, 2 and
The local second-order factors (B v , B1, B2 , and Bmax ), defined pre-
6 for beams (with negligible axial forces) pinned at the far end, fixed at
viously through Eqs. (1)–(4), computed using the analysis method
the far end, bent into symmetrical single curvature bending and anti-
presented briefly below, will be given for verification purposes in a later
symmetrical double curvature bending, repectively.
section.
Alternatively, instead of κj , inverse flexibility factors (1/ κj ) may be
For frames with sway due to lateral and axial loading only, these
coefficients depend on (1) the end restraints, which then uniquely de-
fine the first-order moment gradient, and on (2) the axial load level
defined for instance by the nondimensional load parameters αs, αb or
αE . For an elastic member of length L, uniform axial load and sectional
stiffness EI along the length, they can be defined by
N N N
αs = ; αb = ; αE =
Ncs Ncb NE (7a-c)

where
NE N π 2EI
Ncs = ; Ncb = E2 ; NE =
βs2 βb L2 (8a-c)

Here, Ncs and Ncb are the free-to-sway and the fully braced critical load, Fig. 2. Partially braced frame with sidesway.

3
J. Hellesland Engineering Structures 203 (2020) 109804

frame, it is worth noting that Bs may reach unacceptable values well

before the present “local instability” is reached (at “αb = 1” in the fig-
ures).
Panel columns. The panel columns are rigidly connected to beams
at the top and bottom. EI / L values of the panel members are
EI / L, 1.1EI / L , 0.333EI / L and 1.667EI / L for Column 1 (left hand),
Column 2 (right hand), top beam and bottom beam, respectively. The
bottom beam is considerably stiffer than the upper beam, and will at-
tract the larger first-order end moments. Axial forces are neglected in
the beams. The columns have the same axial force N. Then, because of
the stiffness difference, the load index in Column 1 (left hand) becomes
1.1 times that in Column 2 (αE1 = 1.1αE 2 ). Thus, Column 1 is the more
flexible of the two panel columns, and the one at which system in-
Fig. 3. (a) Sign convention; (b) Single column model; (c) Two-column panel stability will be initiated.
model. Dashed curves show possible deflection modes at member instability The results in Fig. 6 are for the most flexible Column 1 in the panel
(one of several for the panel).
(broken lines) and for the isolated Column 1 (solid lines). Results are
plotted versus αE1 (αE for Column 1). The αs1 (=αE1 βs21) abscissa, added
used, or scaled flexibility factors, such as the well known G factors. In for convenience and information, is computed with the free-sway cri-
their generalized forms [10], they can be defined by tical load of the isolated Column 1 defined above (with effective length
factor βs1 = 1.483). Similarly for the stiffer Column 2, the αs2 (=αE 2 βs22)
(EI / L) ⎛= bo ⎞ abscissa (with βs2 = 1.522 ), is added in Fig. 7.
Gj = bo ⎜ ⎟ j = 1, 2
kj ⎝ κj ⎠ (11) Stationary versus non-stationary restraints. Unlike restraints of
single columns, those of the panel columns will not remain stationary in
where bo is simply a reference (datum) factor by which the relative the general case. As a consequence, the critical loads of the panel col-
restraint flexibilities are scaled. Datum values adopted for instance in umns are lower than those of the isolated columns. This is due to the
ACI [1] and AISC [3], are bo = 6 and bo = 2 for unbraced and braced reduced restraints offered by the panel beams as the axial column
frames, respectively. It should be noted that bo = 6 is used throughout loading increases towards the braced critical load. This is indicated in
this paper, unless otherwise noted. Fig. 3(c), and in the inserts in Figs. 6 and 7, by the unwinding of the
beams from nearly antisymmetrical double curvature (k ≈ 6EIb/ Lb) to
4.3. Typical column response in frames with sway nearly symmetrical single curvature bending (k ≈ 2EIb/ Lb ). Associated
with the reduction in beam restraints, a rather sudden reversal of end
Characteristics study. In a another study [13], response char- moments is seen to take place in the stiffer Column 2 for loads close to
acteristics were presented and discussed for single columns and panels the critical panel load.
with various restraint combinations. Also presented were closed form End moments. The largest first-order end moment (absolute value)
expressions defining a number of key characteristics, or ”landmarks” in is obtained at end 2 with the largest rotational restraint stiffness. End
the moment versus axial load “map”, useful for for enabling a quick moments at the two ends become equal at αE = 1 for single columns,
establishment of moment-axial load relationship of laterally displaced and for common frame (panel) columns [13]. The linear approxima-
columns. Moment results from that study, to be used for the verification tions, B2lin and B1lin , in the figures are discussed later (Section 7.5).
of approximated moment proposals, and points of interest for the pre-
sent paper, will be briefly recaptured and discussed where appropriate.
Single columns. Typical moment and shear responses versus in-
creasing axial load, computed with the second-order analysis, are
shown in Fig. 4 and Fig. 5 for single columns with unequal, rotational
end restraints with different degree of rotational stiffness, as given in
the insert in the upper left hand corner of the figures. In a real case, the
column top displacement Bs Δ0 will be maintained by the action of the
overall frame of which the column may be considered isolated from.
Similar results for two single columns, isolated from the panel by
assuming hinged supports at the first-order inflection points (near
midlengths) of the panel beams, are shown in Figs. 6 and 7. This allows
comparison with the response of the panel columns. The isolated
column is seen to describe the panel end moment response almost ex-
actly up to fairly high axial load levels. More detailed discussion of
response characteristics of this and a panel with a stiffer Column 2, are
given elsewhere [13].
The moments and shear are shown nondimensionally in terms of the
respective B factors, Eqs. (1)–(4). Axial forces are given non-
dimensionally in terms of axial load indices αs and αE (Eq. (7)). All
moment results in the figures are given in terms of Bs M02 . Therefore, B1
and B1lin are multiplied by the ratio M01/ M02 in the figures.
Both moments and shears approach infinity (in either the positive or
negative direction) as αb = N / Ncb = 1 (or αE = 1/ βb2 ) is approached. Ncb
for an elastic column is independent of whether the column is fully
braced at zero or at a non-zero end displacement. The results in the
figures are independent of the sway magnification factor Bs , as stated Fig. 4. Moments and shear versus axial load level in column with unequal,
previously. However, if the considered column was part of a larger relative flexible end restraints (βs = 1.932, βb = 0.785, αE = 0.268αs ).

4
J. Hellesland Engineering Structures 203 (2020) 109804

Fig. 5. Moments and shear versus axial load level in column with unequal, very Fig. 7. Moments and shears for two cases: (a) Right hand Column 2 of the panel
stiff end restraints (βs = 1.065, βb = 0.532, αE = 0.88αs ). (dashed lines); (b) Column 2 considered in isolation (solid lines). B2lin are secant
predictions of B2 .

γn
B v = 1 − αs
γs (12b)

The novelty of Eq. (12) is represented by the distinction between γn and

γs (= γn at αs = 1), and with the γn factor as defined below. Both factors
represent the increased column flexibility caused by the column axial
load acting on the deflection of the column away from the chord
through its ends (Nδ effects).
An approximate γn factor, reflecting the full transition from free-
sway to nearly fully braced, was proposed by Hellesland [10] in terms
of γs and two additional flexibility terms:
γn = γs + Δγ1 + Δγ2 (⩾ γs ) (13a)

8
α − 1⎞
Δγ1 = 0.12 (γs − 1)(αs − 1); Δγ2 = 0.6 αs, b ⎜⎛ s ⎟

⎝ αs , b ⎠ (13b,c)

2
Ncb β
αs , b = = ⎛⎜ s ⎞⎟
Fig. 6. Moments and shears for two cases: (a) Left hand Column 1 of the panel Ncs β
⎝ b⎠ (13d)
(dashed lines); (b) Column 1 considered in isolation (solid lines). B2lin are secant
predictions of B2 . For a cantilever column fixed at the base and with zero axial load, Eq.
(13a)) gives γn = 1.19 (exact 1.20). The limitation on γn (⩾ γs ) may be
adopted when this represents a simplification. For pin-ended columns
Maximum moment. The maximum moment (Mmax , Bmax ) is initially
(Ncs = 0 ), γn = 1. This value may also be taken in the rare case of a
equal to the larger end moment (M2, B2 ). Following an initial decrease
column with axial tensile loading.
along the B2 path, Bmax starts forming away from end 2. Thereafter,
The γn factor may take on large values as the braced critical load is
following a continued small decrease, maximum moments increase and
approached. In other cases, the two first terms of Eq. (13a)) may be
approach infinity for loads approaching the braced critical load
adequate. Often, the first term is sufficient (for frames with reasonably
(αb = 1).
similar columns and loads in the various axes). A linear approximation
The B2m curves, and bullets on the curves, will be discussed later
of γn beyond γs was derived theoretically [14], but the expression was
(Section 8). So will the maximum moment factor approximations,
rather cumbersome, and not very convenient in practical contexts. The
Bm, Bm, t and Bm, c (Section 9).
first to suggest γ values in excess of γs , was possibly Stevens [15].
Shears. For the range of shears shown in the figures, the B v varia-
The γs factor, defined below, varies between 1 and 1.216 (1.22) for
tion is close to linear (B v = 1 − αs , Eq. (12b) for γn = γs ).
columns with positive end restraints, but may become greater in single
curvature regions of multistorey frames.
5. General shear formulation Several diagrams have been established for easy determination of γs
[9–11,16]. Furthermore, general γs expressions are available, notably
In order to establish end moments, it is necessary to know the shear one first given by Rubin [17], and in a modified form by Lai and
force. A general shear expression V = B v Bs V0 (Eq. (1)) can be taken, MacGregor [18]. A detailed discussion and summary of available γs
with B v according to Hellesland [10], as diagrams and expressions are given in [16]. A rather simple approx-
γn N Δ0 imate expression, proposed initially by the author in 1981 during a
Bv = 1 − research stay at Univ. of Alberta, Edmonton, based on observations of
V0 L (12a)
the variation of γs with changing restraints, such as illustrated in Fig. 8,
or, equivalently, is adequate in the present context. It is given by

5
J. Hellesland Engineering Structures 203 (2020) 109804

M1
B1lin = = 1 − (1 − B1s ) αs
Bs M01 (17)
M2
B2lin = = 1 − (1 − B2s ) αs
Bs M02 (18)
Here, B1s and B2s are end moment factor values at αs = 1 as shown in
Fig. 10. B1lin and B2lin predictions with these factors have been found to
provide excellent agreement with exact results [13]. Below, attempts
are made to derive closed form expressions for B1s and B2s .
Results for B2s and B1s coincide in the case with equal end restraints,
and are shown by the dash-dot borderline (labelled G1 = G2 ). Results for
B1s , shown by dashed lines in the figure, and B2s , shown by solid lines,
are located above and below the borderline, respectively. G2 is by de-
finition taken to represent the end with the stiffer restraint (with the
smaller G value). Corresponding B1s and B2s curves terminates therefore
at the dash-dot curve. At G2 = 6/ κ = 0 (fixed end), B2s may have values
Fig. 8. The flexibility factor γs at the free-sway condition versus positive rota- between 0.79 and 0.82, and B1s between about 0.82 and 1.05.
tional end restraints in terms of G factors (reproduced from Hellesland [9]).
7.2. B2s factors for two special cases
1 + [1 − (0.5 Gmax ) p ] 3
γs = 1 + 0.108 For cases with only one unknown end moment (e.g., M2 ) and linear
(1 + 0.5 Gmin ) 2 (14)
shear (γn = γs ), Eq. (16) transforms to
where p = 1 for Gmax ≤ 2 (κ ⩾ 3) and p = −1 for Gmax > 2 (κ < 3). Gmax M2 1
is the larger and Gmin the smaller of the G factors at the two column B2 = = 1 − (1 − ) αs
Bs M02 γs (19)
ends. Eq. (14) breaks down into a pin-ended case considered later (in
conjunction with Eq. (20)). This equation applies to the case of a column pinned at one end
(M1 = 0 ) and to a symmetrically restrained column (M1 = M2 ). Eval-
uated at αs = 1, B2 becomes B2s = 1/ γs . For the two special cases con-
6. General end moment formulation
sidered, for which γs can be found in the literature, B2s can be given as
follows:
From moment equilibrium of a laterally loaded (displaced) column
(Fig. 3(a)), M1 + M2 + N Δ + VL = 0 , where Δ = Bs Δ0 and V is the shear
(1) Column pinned at end 1; γs = γs, pin1:
given by Eq. (12b), the end moment sum may be written as
1 0.216
M1 + M2 γ −1 B2s = with γs, pin1 = 1 +
=1− n · αs γs, pin1 (1 + 0.5 G2)2 (20a,b)
Bs (M01 + M02) γs (15)

From this equation, end moments can be computed directly in cases in (2) Column with equal end restraints; γs = γs, equal :
which there is only one unknown end moment. Two such cases are: (1) 1 0.216
columns pinned at end 1 (M01 = M1 = 0 ), and (2) columns with equal B2s = with γs, equal = 1 +
γs, equal (1 + G2)2 (21a,b)
end restraints (M1 = M2 ). In these cases, Eq. (15) reduces to
The pin-ended γs (Eq. (20b)) was derived by Hellesland [9], and along
M2 γ −1
=1− n · αs different lines and in a different form (CL = 1 − γs ) by LeMessurier [11].
Bs M02 γs (16)

End moment predictions with this equation are shown in Fig. 9 for two
columns pinned at the top and base restraints as shown. The agreement
with second-order theory, labelled “Exact” in the figure (reproduced
from Hellesland [10]), is seen to be good, in particular in the positive B2
range. Maximum moment predictions (Bm ) will be discussed later.
In the general case, with unequal end restraints at the two ends, Eq.
(15) is not directly useful. The distribution of the moment sum to the
two ends must be established before individual end moments can be
calculated. Efforts at accomplishing this have not been successful, and
remains a task for future research.
The simpler task of establishing moment expressions that are valid
for the important class of supporting columns (αs < 1), and over a load
range somewhat beyond this, is pursued below.

7. Linearized end moment formulations

7.1. Formulations

End moments will generally be discussed in terms of the end mo-

ment factors B1 and B2 . Linear moment relations for the individual end Fig. 9. Moments versus axial load level for two columns pinned at the top: (a)
moments, taken as secant approximations (between αs = 0 to 1.0) can Fixed at the base (βs = 2, βb = 0.7 ); (b) Partially fixed at the base
be expressed by (βs = 2.635, βb = 0.843); αs = αE βs2 .

6
J. Hellesland Engineering Structures 203 (2020) 109804

bending. This ratio in terms of G2 factors can easily be established [13].

Sufficiently accurate γs values above can be obtained from Eq. (14). An
overestimation of B2s will lead to an underestimation of B1s , and vice
verse. Prediction accuracies are found to be within about ± 2 to 2.5%.
Considering the lesser importance of obtaining very accurate B1s
predictions, a simpler approach may be justified. Based on Fig. 10, it is,
alternatively, proposed to calculate B1lin (Eq. (17)) with
B1s = 1 for G1 > 1.25 (25a)

0.22
B1s = 1 − for G1 ⩽ 1.25
0.22 + (1 + G1)2 (25b)
Eq. (25b) is given by the curve labelled G1 = G2 in Fig. 10. The pre-
diction accuracy of Eq. (25) is close to that of Eq. (23), and is con-
sidered sufficiently accurate. At one stage [19], G1 greater or smaller
than 1.0 was considered in Eq. (25).

Fig. 10. End moment factors B1s and B2s at αs = 1 versus end restraints in terms
7.5. B2lin and B1lin compared with exact results
of G factors (from Hellesland [13]).
Linear end moment predictions at end 2 by B2lin (Eq. (18)), with B2s
given by Eq. (22), are shown in Figs. 4 and 5 for single restrained
Also, Eq. (14) breaks down into this expression in the pin-ended case.
columns, and in Figs. 6 and 7 for the two columns of a panel. Predic-
The equal restraint case (Eq. (21b)) can be obtained from Eq. (20b) by
tions at end 1, B1lin (Eq. (17)), computed with the simplified B1s (Eq.
replacing L by L/2 . The factor 0.216 may clearly be rounded off to 0.22,
(25)) are shown only in Figs. 4 and 5.
or even 0.2, in practical applications.
It can be seen that the linearized end moment approximations
B2s = 1/ γs for the two cases above can be identified in Fig. 10 by the
generally provide good B2 and B1 predictions, not only for supporting
solid line marked “∞” and by the dash-dot line marked “G1 = G2 ”, re-
sway columns (αs < 1) , but also well beyond this range in the cases
spectively. The equal-ended case can be seen to provide an upper bound
considered.
on the B2s results, and the pin-ended case a practical lower bound,
except at low G2 values (less than about 1.5; κ > 4 ) where slightly, but
8. Lower bound on maximum moments
insignificantly, smaller B2s values result.
In the figures (Figs. 4–7) it is seen that maximum moment factors,
7.3. Approximate B2s factor – general case Bmax , are less than 1.0 up to fairly high load indices, and that they are
well below 1.0 in cases with relatively stiff end restraints.
The equal-ended case (Eq. (21)) provides an upper bound on the B2s An approximate lower bound on the maximum moment can be
values, and might be chosen to provide conservative estimates of the given by a B2m factor defined by a bilinear curve that follows the des-
maximum end moment in the general case. However, since the B2s re- cending linear approximation B2lin (Eqs. (18), (22))) up to αs = 1.25, and
sults lie within a reasonably narrow band (shaded in Fig. 10), it is then stays constant:
proposed to adopt an “average” B2s value, located approximately in the
middle of the shaded band, given by B2m = B2lin ⩾ B2in (αs = 1.25) (26a)

1 0.24 which transforms to

B2s = with γs, aver = 1 +
γs, aver (1 + 0.75G2)2 (22a) B2m = 1 − g2 ·αs ⩾ 1 − g2 ·1.25 (26b)
or, when rewriting, by 0.24
g2 = 1 − B2s =
0.24 0.24 + (1 + 0.75G2)2 (26c)
B2s = 1 −
0.24 + (1 + 0.75G2)2 (22b)
Predictions by this B2m are included in Figs. 4–7. It can be seen to
This expression is independent of G1, reflecting the relative insensitivity give good estimates of the maximum moment for αs less than about
to G1, and gives B2s values within ± 2.5% of correct values. This ac- αs = 3. A more detailed discussion is given in Hellesland [20], where
curacy is generally quite acceptable. This B2s is adopted in this paper for also other cases are presented.
linearized B2lin predictions by Eq. (18). Bullet points are included on the B2m curve at load indices of
αb, t = 0.7 , computed with single curvature beam bending (horizontal
7.4. Approximate B1s factor tangent at beam midlength). As a potential application limit indicator,
αb, t of 0.7–0.8 was considered. However, on the overall, αs = 3, com-
When B2s is known, the moment factor B1s at column end 1 can be puted with “sidesway restraints”, represents a more consistent limit
solved for from Eq. (15) for αs = 1 (at which γn = γs ). It becomes indicator.
In an initial phase, αs = 1 was considered instead of 1.25 in Eq. (26).
1 − μ0 ⎞ 1 A value of 1.5 has also been considered, and found to be acceptable in
B1s = ⎜⎛B2s − ⎟

⎝ γs ⎠ μ0 (23) all cases except for those with very stiff restraints, such as in Fig. 5.

where
9. “Rising” moment branch
M G +3
μ0 = − 01 = − 2
M02 G1 + 3 (24) 9.1. Approximate formulation

With the adopted sign convention for end moments (Fig. 3(a)), this
The exact maximum moment (Eq. (4)), is here approximated by
ratio between the first-order end moments becomes positive for single
first-order curvature bending and negative for double curvature Mmax = Bm Bs M02 (27)

7
J. Hellesland Engineering Structures 203 (2020) 109804

where the approximate member magnification factor is denoted Bm to 10. Maximum column moment proposals
distinguish it from the exact Bmax . The “rising” moment branch, above a
lower limit “limBm ”, which will discussed later, is commonly computed 10.1. Maximum moment proposal 1
using variations of Bm below:
It has proven difficult to develop reasonably accurate approxima-
Bm = Bb ⩾ limBm (28) tions for maximum moments in the general case. An attempt is made
below to provide some improvements, with different degrees of con-
where servativeness, to common regular design work procedures at present.
1 + A αb Alternative (1a) – Low to moderately high axial load levels:
Bb = · Cm For most practical load levels, it is acceptable for design purposes to
1 − αb (29a)
compute the maximum moment according to Mmax = Bm Bs M02 , Eq.
(27), with the simplification
Cm = 0.6 + 0.4μ0 (29b)
Bm = 1.0 (30)

Bs M01 M It has been found [13], from elastic second-order analyses of single
μ0 = − = − 01 columns with practical (and invariant) end restraints, that this Bm ap-
Bs M02 M02 (29c)
proximation is conservative for load indices given in terms of the free-
Here, αb = N / Ncb is the critical load index (Eq. (7b)) of the column sway and the braced critical load indices, respectively, by
considered braced, Cm is a moment gradient factor that accounts for αs < 3.5 (3.0) or αb < 0.5 (0.8) (31)
non- uniform first-order bending [21], and μ0 is the first-order end
moment ratio, to be taken as positive when the member is bent in single If some 5–10% underestimation of moments were accepted, the limits
first-order curvature, and negative otherwise. A is a factor, typically above could be increased somewhat. For columns in sway frames, most
about 0.25 for pin-ended columns with uniform first-order bending columns will have load indices well below the values indicated above.
(Cm = 1), but commonly neglected. Also in present computations, A = 0 The numbers in parentheses were found for columns in panels
is chosen. (frames), in which, as discussed before, the critical loads are lower than
With the lower limit taken as limBm = 1, and A = 0 , the approx- those of isolated, single columns due to the reduced restraints offered
imate Bm formulation above can be found in most structural design by the panel beams as the axial column loading increases towards the
codes (such as ACI 318 and AISC 360 (limited to “braced moments”), braced critical load. As a consequence, the rising, maximum moment
Eurocodes 2 and 3). The approximation and its faults are discussed in branch will be pressed upwards, towards larger values (eventually in-
considerable detail by Lai et al. [22]. In the literature, the Cm factor is finity) at smaller load indices than in the single column case.
extensively discussed for both elastic and inelastic applications (e.g. The limit in terms of the free-sway load index, αs , is affected much
[23–29]). However, Cm in Eq. (29b) is the most widely adopted factor less than the corresponding limit in terms αb by the changing of the end
for elastic applications. restraints described above. Consequently, αs is the better suited para-
For regular frames, it is common design practice to assume that meter of the two, to indicate range of applicability of simplified max-
beams bend into symmetrical, single curvature (with rotational bending imum moment expressions.
stiffness 2EIb/ Lb ) at braced frame instability if more correct values are Alternative (1b) – Present practice, for any load level:
not established, and to use such restraints in the calculation of αb . This Axial load indices are not likely to exceed those in Eq. (31), but may
is considered a prudent approach, and is in accordance with most codes possibly do so in partially braced frames. To complement Alternative
of practice. This assumption will also be adopted in the comparisons (1a) above, it is therefore prudent, at very high load levels, to include a
below. rising moment branch. Common practice is, as mentioned before, to
calculate Bm according to the conservative Eq. (28) defined with
limBm = 1 and A = 0 . Then, adopting the same,
9.2. Comparisons with exact results Bm = Bb ⩾ 1.0 (32)

Single column results. Approximate predictions Bm for single re- This case is not specifically identified in Figs. 4–7, but it is clear that it is
strained columns according to Eq. (28) are shown in Figs. 4, 5. equal to the case below, exept that the lower bound on the rising branch
The predictions labelled Bm are computed with the restraints given (Bm ) is taken as 1.0 rather than a value less than 1.0.
in the inserts of the figures. The curves labelled Bm, t are included for
comparative reasons. They are computed assuming that the columns 10.2. Maximum moment proposal 2
have been isolated from a greater frame with rotational beam stiffnesses
of kb = 2EIb/ Lb (single, symmetrical curvature bending). These are 1/ To simplify presentation and discussion, proposal 2 alternatives are
3rd of the restraints given in the figures. illustrated for a specific restraint case in Fig. 11.
Panel column results. Two sets of predictions are shown for the Alternative (2a) – Low to moderately high load levels:
panel columns in Figs. 6 and 7. The curves labelled Bm, t are computed as A more economical design than that in Alternative 1(a) can be
described above (with beam stiffnesses kb = 2EIb/ Lb ). The Bm, c curves achieved for αs < 3 by taking Bm according to
are computed with αb taken as the exact critical load indices of the two
Bm = B2m (33)
panel columns (in terms of αE equal to 1.639 and 1.490, respectively).
The Bm, t curves are seen to be very close to the Bm, c curves. This where B2m is “maximum moment modified B2lin ” given by Eq. (26), and
implies that the assumed single, symmetrical curvature beam bending defined also in Fig. 11. The rotational end restraints to be used in cal-
in the Bm, t computations is close to the exact one for this panel. The culating B2m are those for the column considered free-to-sway. Typi-
difference will be more marked in cases with greater difference between cally, k = kb = 6EIb Lb , and G2 = 6(EI / L)/ k , corresponding to anti-
the columns of the panel. symmetrical beam bending.
As seen, the Bm predictions (incl. Bm, t and Bm, c ) by Eq. (28) are ra- For unbraced frames (SB = 0 in Eq. (6b), individual αs values will
ther conservative. This is primarily due to the Cm approximation, that normally be at most about 1 to 2, and never likely to ever exceed 3.0
tends to become very conservative (too large) for columns with sig- (Hellesland [20]). This estimate is based on storey sway magnifiers, Bs
nificant double curvature bending, as in the present cases. (Eq. (5)), in the range 1.5 to 2.0.

8
J. Hellesland Engineering Structures 203 (2020) 109804

to lateral loads are all considered feasible in regular design work.

From a practical design point of view, the proposal (1) alternatives
(a and b) are believed to be the most suitable in early design phases,
when details of column end restraints are yet to be determined.
The proposal (2) alternatives (a, b and c) will allow more eco-
nomical designs than the proposal (1) alternatives, but are believed to
be most suitable in final design phases, following preliminary design,
and in possible design check situations.
The alternative (2c) formulation has the advantage over alternative
(2b) that it is less conservative. Also, and conceptually more important,
it can be applied in a more rational manner in cases with load combi-
nations that also include moments from gravity loading.

11. Overview of possible application steps

1. First-order load effects. Establish end moments (M01, M02 ) and

shears (V0 ) from a conventional first-order analysis based on as-
sumed sectional stiffnesses.
2. Column end restraints. For calculation of local second-order ef-
Fig. 11. Maximum moment proposal alternatives (2a), (2b) and (2c), as illu- fects for columns considered in isolation, establish end restraint
strated for a specific case (G1 = 3 (κ = 2 ), G2 = 0.6 (κ = 10 )). Alternatives (1a) stiffnesses k1 and k2 (and corresponding nondimensional parameters
and the “present practice” alternative (1b) are not shown in the figure, but are as discussed in Section 4.2 (and in Section 4.3).
equal to (2a) and (2b), respectively, when B2m is taken as 1.0.
3. Global sway magnification factor. Establish Bs (Eq. (5)) for the
sidesway action, at the level (floor) of the frame considered, using a
For partially braced frames with sway (SB > 0 ), the column indices storey stability index (αss ) defined by Eq. (6a) or (6b), or by sim-
can be greater than in the unbraced case above without inflicting un- plifications of these.
acceptable Bs indices. However, there are many good reasons in design 4. Maximum design moment. Establish Mmax = Bm Bs M02 (Eq. (27)).
to limit axial loads to levels well below the critical braced column loads – Preliminary and early design phases: the maximum moment factor
[20]. Column loads, in terms of critical loads for braced cases, or better Bm may simplified be approximated by Bm = 1 (Eq. (30)) in most
in terms of the system critical loads, should probably be limited in practical cases, or conservatively by the “present practice” factor
regular design work to values below 0.7αb , or so, in order to avoid Bm = Bb ⩾ 1.0 (Eq. (32)) in very high axial load level cases.
column behaviour sensitive to uncertainties in restraint assessments, – Final design checks: a more rational and economical design is
including possible effects of beam yielding, and unwinding type phe- obtained with Bm = B2m (Eq. (33)) in most practical cases, or more
nomena. conservatively with Bm = Bb ⩾ B2m (Eq. (34)) in very high axial
Alternative (2b) – Any load level: load level cases, or alternatively with Bm = Bb·B2m ⩾ B2m (Eq.
Again, to cover very high load levels, a rising moment branch ap- (35)); see earlier discussion.
proximation may, in lieu of better formulations, be taken according to 5. End moments. If end moments, given by Mj = Bj Bs M0j ( j = 1, 2 ),
Bm = Bb ⩾ B2m (34) are of interest for the design of a foundation or adjacent beams, they
can be estimated with B1 and B2 approximated by B1lin and B2lin ,
Alternative (2c) – Any load level: respectively, as given in Section 7.
A less conservative magnifier than that defined above by Eqs. (32) 6. Shear transfer. Column shear at a foundation attached to a column
and (34) is illustrated in Fig. 11 by the curve marked (2c). It is given by may sometimes be required. The reduced shear V = B v Bs V0 , can be
a product defined by computed with B v given by Eq. (12a) or (11b), or simplified by the
linear version B v = 1 − αs .
Bm = Bb·B2m ⩾ B2m (35)

In this formulation, the rising branch of the proposal Alternative (2b) is 12. Maximum moment in load combinations
lowered by the multiplication with B2m (Eq. (26)).
Predictions by Eq. (35) are not included in the figures (Figs. 4–7), A brief mention on how to deal with load combinations seems ap-
but the effect is that the conservativeness of the rising moment branch propriate. For framed columns with first-order end moments from both
predictions is reduced, yet still quite conservative in most cases in- lateral loading (M0s ) and gravity loading (M0b ), the sway modified first-
vestigated. order moment sum (not including local second-order effects) at the two
An exception is found for the case of a column pinned at the top ends, can in line with the principle of superposition be defined by
(G1 = ∞) and fully fixed (clamped) at the base (G2 = 0 ), for which Bm ∗ ∗
M01 = M01b + Bs M01s and M02 = M02b + Bs M02s (36)
(Eq. (35)) gives somewhat unconservative predictions. However, two
∗
aspects of the fixed-pinned column case deserving attention are the i) Here, M02 is taken, per definition, to be the moment at the end with the
∗
theoretically full fixity at one end and ii) the theoretically perfect hinge larger moment sum (absolute value), and M01 the moment at the end
connection to the adjoining structure. Both are difficult to achieve in with the smaller end moment sum.
practice. It is found (Hellesland [20]) that only a slight fixity reduction Proposal 2, alternative (2c) above, offers the most rational max-
from G2 = 0 to G2 = 0.25 (κ2 = 24 ) is sufficient to provide good predic- imum column sway moment formulation. Adopting this alternative, it is
tion accuracy. Also, it is found that a “practical pinned connection” proposed to calculate maximum column moments from
with a small restraint of about G1 = 10 (κ1 = 0.6) or so, is sufficient to Mmax = Bb (M0b + B2m Bs M0s )2 (37)
compensate for the mentioned unconservativeness. It is felt that these
considerations separately justify the use Eq. (35) also for practical pin- where
ended columns with very stiff base restraints.
Summary and conclusions on maximum moment proposals: (i) B2m is to be taken according to Eq. (26), when the larger moment
The maximum design moment proposals for columns with sway due sum occurs at the end with the larger sway moment M0s (i.e., at the

9
J. Hellesland Engineering Structures 203 (2020) 109804

end with the stiffer end restraint), and References

(ii) B2m = 1 when the larger moment sum occurs at the end with the
smaller sway moment M0s (i.e., at the most flexible end restraint). [1] ACI 318–14. Building code requirements for structural concrete, and Commentary.
Farmington Hills (MI): American Concrete Institute (ACI); 2014.
[2] Eurocode 2: Design of concrete structures – Part 1-1 (NV 1992-1-1). Brussels:
Furthermore, Bb , Eq. (29a), is now subject to the restriction Bb ⩾ 1, and European Committee for Standardization (CEN); 2004.
Cm , Eq. (29b), is now defined with μ0 in Eq. (29c) replaced by [3] ANSI/AISC 360-16. Specification for structural steel buildings, and Commentary.
∗
Chicago (IL): American Institute of Steel Construction (AISC); 2016.
M01 (M0b + Bs M0s )1 [4] Eurocode 3: Design of steel structures – Part 1-1 (EN 1993-1-1). Brussels: European
μ0 = − ∗
=−
M02 (M0b + Bs M0s )2 (38) standard, European Committee for Standardization (CEN); 2005.
[5] Kemp AR. Inelastic local and lateral buckling in design codes. J Struct Eng
This moment ratio is taken positive when the member has single first- 1996;122(4):374–82.
[6] Peng BHH, MacRae GA, Walpole WR, Moss P, Dhakal RP, Clifton C, et al. Location of
order curvature bending, and negative otherwise. Eq. (37) breaks down plastic hinges in axially loaded steel members. J Construct Steel Res
to Mmax = Bm Bs M02 (Eq. (27)) with Bm given by Eq. (35) when M0b = 0 , 2008;64(3):344–51.
and into the conventional, common practice, formulation for gravity [7] Elkady A, Ligno DG. Full-scale testing of deep wide-flange steel columns under
multi-axis cyclic loading: Loading sequence, boundary effects and out-of-plane
load moments when M0s = 0 [30].
brace force demands. ASCE. J Struct Eng 2018;144(2).
It is a drawback of the proposal that it is necessary to check if the [8] da Silva LS, Simões R, Gervasio H. Design of steel structures. ECCS Eurocode Design
larger end moment sum and the larger sway moment occur at the same Manuals, Ernst & Sohn; 2010.
or different column ends. The advantage is a more economical final [9] Hellesland J. Approximate second order analysis of unbraced frames. Technical Rep.
Oslo (Norway): Dr. Ing. A. Aas-Jakobsen Ltd.; 1976.
design. A more extensive discussion, and comparisons with present [10] Hellesland J. Extended second order approximate analysis of frames with sway-
code formulations, are beyond the scope of the present paper. braced column interaction. J Constr Steel Res 2009;65(5):1075–86. [Link]
org/10.1016/[Link].2008.08.008.
[11] LeMessurier WJ. A practical method of second order analysis, Part 2 - Rigid frames.
13. Conclusions Eng J AISC 1977;14(2):49–67.
[12] Lui EM. Geometrical imperfections on inelastic frame behavior. J Struct Eng ASCE
Development of shears, end moments and maximum moments be- 1992;118(5):1408–15. [Link]
118:5(1408).
tween ends of framed columns with sidesway have been studied using [13] Hellesland J. Mechanics of columns in sway frames – Derivation of key character-
elastic second-order theory. The full range of axial loads are covered, istics. Res. Rep. in Mechanics, No.1-Febr. Oslo (Norway): Dept. of Math., Univ. of
and therefore both “supporting sway columns” (αs < 1) and “supported Oslo; 2019. [Link]
[14] Xu L, Liu Y. Story stability of semi-braced steel frames. J Constr Steel Res
(braced) sway columns” (αs > 1). 2002;58(4):150–9.
End moment and maximum moment expressions for columns with [15] Stevens LK. Elastic stability of practical multistory frames. Proc Inst Civ Eng
sway are derived. These represent novel contributions that are suitable 1967;36.
[16] Hellesland J. Second order approximate analysis of unbraced multistorey frames
in typical design code formats, and in practical design work. They will
with single curvature regions. Eng Struct 2009;31(8):1734–44. [Link]
allow more rational column assessments, and will reduce the con- 1016/[Link].2009.02.015.
servativeness of current procedures, thus allowing more economical [17] Rubin H. Das Q_Δ -Verfahren zur vereinfachten Berechnung verschieblicher
designs than presently available structural code expressions. Rahmensysteme nach dem Traglastverfahren der Theorie II. Ordnung. Der
Bauingenieur 1973;48(8):275–85.
A possible extension of the maximum moment formulation to the [18] Lai SMA, MacGregor JG. Geometric nonlinearities in unbraced frames. J Struct Eng
general case of columns with moments from both gravity and lateral ASCE 1983;109(11):2528–45. [Link]
loading, is briefly presented for the sake of completeness, but not dis- 109:11(2528).
[19] Hellesland J, MacGregor JG. Mechanics and design of columns in sway frames.
cussed in any depth. Preliminary draft (unpublished). Dept. of Civ. Engineering, Univ. of Alberta,
Edmonton, Canada; Sept. 1981.
Declaration of Competing Interest [20] Hellesland J. Approximate end and maximum moment formulations for slender
columns in frames with sway. Res. Rep. in Mechanics, No.1-March. Oslo (Norway):
Dept. of Math., Univ. of Oslo; 2019. [Link]
None. 70699.
[21] Austin WJ. Strength and design of metal columns. J the Struct Div ASCE
1961;87(ST4):1–32.
Acknowledgments [22] Lai SMA, MacGregor JG, Hellesland J. Geometric nonlinearities in nonsway frames.
J Struct Eng ASCE 1983;109(12):2770–85. [Link]
During a research stay by the author at the Univ. of Alberta (UA), 9445(1983) 109:12(2770).
[23] Sarja A. Two methods for analysing the stability of slender concrete columns in
Edmonton, Canada, in 1981, a preliminary draft entitled “Mechanics
frames. Nordisk Betong (J Nordic Concrete Federation) 1974;5:21–9.
and design of columns in sway frames”, co-authored with the now de- [24] Trahair NS. Design strength of steel beam-columns. Struct. Eng. Rep., No. 132,
ceased Professor J.G. MacGregor, was neither published nor completed. Edmonton (Alberta): Dept. of Civ. Eng., Univ. of Alberta; 1985.
The topic is still of considerable interest, though. The present paper is [25] Duan S, Chen WF. On beam-column moment amplification factor. Eng J AISC
1989;4(26):130–5.
significantly revised and extended. Early input by Prof. MacGregor is [26] Chen WF, Lui EM. Stability design of steel frames. Boca Raton (FL): CRC Press;
greatly appreciated. So is the running of the initial computer analyses of 1991.
the panels by S.M.A. Lai, then a PhD student at UA. In the early phase of [27] Villette M. Consideration sur le flambement, proposition de révision de l’Europe 3.
Constr Métal 1997;3:15–38. [in French].
this study, support was provided to the author by the Research Council [28] Sheng-Jin C, Wu-Chyan W. Moment amplification factor for P-δ effects of steel
of Norway and a research associateship at UA. beam-column. Tech Note ASCE 1999;125(2):219–23.
[29] Leite LC, Bonet JL, Pallarés L, Miguel PF. C_m-factor for RC slender columns under
unequal eccentricities and skew angle loads at the ends. Eng Struct 2014;71:73–87.
Appendix A. Supplementary material [30] Hellesland J. Mechanics and slenderness limits of sway-restricted reinforced con-
crete columns. J Struct Eng ASCE 2008;134(8):1300–9. [Link]
Supplementary data associated with this article can be found, in the (ASCE)0733-9445(2008) 134:8(1300).
online version, at [Link]

10