Wagner Effects

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Engineering Strucw-es, Vol. 18, No. 2, pp.

125-132, 1996
Copyright 0 1995 Elsevier Science Ltd
Printed in Great Britain. All rights reserved
0141-0296(95)00112-3 01414296/96 $15.00 + 0.00
ELSEVIER

Wagner term in flexural-torsional


buckling of thin-walled open-
profile columns*
W. A. M. Alwis and C. M. Wang
Department of Civil Engineering, National University of Singapore, Kent Ridge Crescent,
Singapore 051 I
(Received November 1994; revised version accepted February 1995)

This paper considers the Wagner term in flexural-torsional buckling


of thin-walled open-profile columns, which has drawn much atten-
tion following its rejection by Ojalvo in 1981. During the ensuing
debate on the subject, further issues of flexural-torsional buckling
were raised highlighting the need for a comprehensive clarification.
This paper collates various questions that have not been convinc-
ingly answered. In tackling these questions, two simple bar models
are first used to clarify important differences between alternative
representations of internal stress resultants, and their relationship
with the Wagner term. With this background, a statically admissible
free-body diagram is described, establishing the Wagner term with-
out resorting to disputed means such as the use of a fictitious lat-
eral load. The Wagner term is shown to be an internal stress result-
ant of the longitudinal prebuckling forces distributed across the
member section; this stress resultant is of the same order of magni-
tude as the St Venant and warping torque. The spatial effect of
stresses contributing to the Wagner term cannot be captured by
the single line representation as proposed by Ojalvo.

Keywords: flexural-torsional buckling, thin-walled open-profile


members, columns, Wagner term

1. Introduction shearing stresses that would cause a torque of -EI,#“,


known as the warping torque. As a result of nonuniform
Cross-sections of a free-ended thin-walled open-profile torsion, the applied torque T, and its gradient mZ per unit
member subjected to a constant torque will warp equally length of the member are, respectively, given by
according to St. Venant’s theory. The only stresses pro-
duced are the shearing stresses in member sections. These
T, = -EI&” + GJ+’ (la)
stresses yield a torque equal to GJ+‘, where GJ is the tor-
sional rigidity, 4 the angle of twist, (.)’ = d(.)/dz and z the mZ = -Tr = EIuv - GJ$’ (lb)
longitudinal co-ordinate. Under this pure torsion condition,
the relative longitudinal displacement is described by w@ This torque expression was derived first for doubly sym-
where w is the double sectorial area co-ordinate. When the metric I-beams by Timoshenko’ and later extended to gen-
torque is not uniform along the member length, 4’ will not eral thin-walled open sections by Wagner’.
remain constant along its length. However, it is still The conventional static-kinematic derivation of the gov-
assumed that the warping displacement remains’ as ~4’. erning flexural-torsional buckling equation for axially
The longitudinal stresses due to this nonuniform longitudi- loaded members (see for example, Timoshenko and Gere3)
nal distribution of warping yield neither a resultant force entails the introduction of a fictitious lateral load to replace
in the axial direction nor moments in the planes passing the effect of axial load P on the deformed member. This
through the member axis. They do, however, produce fictitious distributed load is derived by equating the
moment distribution in a laterally loaded beam to that of a
flexurally deflected member under an axial load. The fic-
* This paper is dedicated to Professor N. S. Trahair on his 60th birthday. titious load produces an applied torque gradient given by

125
126 Wagner term in flexural-torsional buckling: W. A. M. Alwis and C. M. Wang

mZ = Px,~v~,- PY,~u~- Pi$@ (2) Nevertheless, there remain some important issues raised
by Ojalvo and others, that do not appear to have been clari-
which when substituted into equation (lb) yields the gov- fied during this decade-long debate on flexural-torsional
erning flexural-torsional buckling equation3m8 buckling. Examples are: the dubious use of a fictitious
external load and the absence of a statically admissible free-
EI&Y - GJfl + P<y - Px,p:’ + Pysu; = 0 (3) body diagram that can substantiate the Wagner term. It is
thus the aim of this paper to collate the various unresolved
where x,,y, are the co-ordinates of the shear centre S with issues and provide some answers to them. In tackling these
respect to the centroid C; u, and v, are the displacements issues, we begin by considering two simple bar models
of the shear centre in the x- and y-directions, respectively, where important differences between alternative represen-
(see Figure I ); and r,Tis the radius of gyration of the section tations of internal stress resultant and their relationship with
about its shear centre. Note that equation (3) has also been the Wagner term are pointed out. With this background, a
derived via the work-energy approach. statically admissible free-body diagram for establishing the
The term Pr$#? of equation (3) is commonly referred to Wagner term is described. The derivation of the governing
as the Wagner term. The validity of the Wagner term in torsional buckling equation, does not make use of the fic-
the governing equation has been persistently questioned by titious lateral load. This paper will also highlight the inad-
Ojalvo9-‘4. Using both the static-kinematic and work- equacy of Ojalvo’s single-bar model for flexural-torsional
energy approaches, he argued that the radius of gyration r,, buckling. It is hoped that the present critical examination
in the Wagner term should be replaced by the distance d, of flexural-torsional buckling will contribute to a better
between the shear centre and the centroid of the cross-sec- understanding of the individual terms in the established the-
tion. Accordingly, Ojalvo’s bar theory not only yields a ory, including the Wagner term.
significantly different buckling load solution, but it also
implies that pure torsional buckling is impossible when the
shear centre and the centroid coincide. 2. ‘Unresolved’ issues of flexural-torsional
Ojalvo’s argument against the Wagner term has been buckling theory
broadly rejected’5m20 on the basis of experimental evidence
as well as alternative analyses, including work-energy 2.1. Absence of a free-body diagram
based derivations. There have been suggestions” that the Ojalvo’2,‘3 has noted that hitherto static-kinematic deri-
single line representation of a member for deriving a the- vations in which the Wagner term appears, do not provide
ory, which is viable for bending problems, may not be a free-body diagram with which flexural-torsional equilib-
applicable for torsion of open-profile members. rium is expressed. The writers’ interpretation of Ojalvo’s’
In the most recent rebuttal in the open literature, Goto argument is as follows. Consider a pin-ended open-profile
and Chen*O have claimed that Ojalvo’s energy derivation bar and take a free-body diagram of a segment from one
is fundamentally flawed. They pointed out that Ojalvo has end to a generic section along the deformed bar. The appro-
allowed the axial load to change by a second-order magni- priate axis of the torque at that generic section is the one
tude as the bar buckles, which according to them is not that passes through the shear centre parallel to the cen-
admissible. This latter argument has in turn been disputed troidal axis. This torque T”zdue to the axial force P applied
by Alwis and Wang*‘, who argued that it is not necessary at the member end is given by
to impose a constant load for a valid bifurcation buckling
analysis. They showed that the conventional strain energy
functional, which incidentally was used by Ojalvo” as well 4 + Y,4' v:- x,4’ 1
as Goto and Chen2’, implicitly presumes a constant load, c = det u, + x,, v.s+y., z
thus leading both parties to inexact conclusions. Alwis and 0 0 P
Wang*’ presented a general energy formulation accommo-
dating second-order changes in both the axial load and the
bar length, and concluded that the conventional total poten- = e&4’ - w: +Y.#i> (4)
tial energy functional is correct, despite the constant load
limitation in the conventional strain energy functional. With where & =x5 + y$ By assuming that 71 is the same as Tz
this latest clarification, one may consider that the validity of equation (la), one obtains by substitution in equation (4)
of the governing equation (3) inclusive of the Wagner term,
is established conclusively by the work-energy approach. ,I&“’ - GJ@ + Pdf$’ - Px,>v: + Py,,u: = 0 (5)

Figure 7 Thin-walled open-profile section


Wagner term in flexural-torsional buckling: W. A. M. Alwis and C. M. Wang 127

Equation (5) is the version suggested by Ojalvo and it The area referred to is of the cross-sectional surface which
replaces r,T in the conventional equation (3) by d,. The at all points remains normal to the longitudinal fibres pierc-
issues to be examined are: firstly, the absence of a free- ing the cross-section. The issue is whether it matters in this
body diagram for the c80nventional derivation; and sec- instance, that the deformed section is not contained in a
ondly, which one of equations (3) and (5) is valid? single plane and parts of the sectional surface are orthog-
onal neither to the deformed centroidal axis nor the
2.2. Bar idealization deformed shear centre axis.
Ojalvo has insisted that an open-profile member should be 2.6. Effect of end-traction distribution
represented as a single line. He argued that to be compatible
In the conventional work-energy approach for Euler buck-
with other engineering theories of structural elements, all
ling analysis of a bar, the potential energy of the load is
internal and external actions in the member must be cap-
calculated by taking the product of the load and the second-
tured by parameters defined along that line, which would
order change in distance between the centroidal axis ends.
then represent the three-dimensional member as a bar. Each
This change in distance (or shortening) is evaluated by inte-
point of the line represents a section of the actual member.
grating the square of deflection slope of the centroidal axis,
The lateral movement of a point of the line denotes the
on the basis that this axis does not change in length.
average lateral movement of the section. The bending
0jalvo’3 adopted the same approach to establish the poten-
moments, torsional moment, and axial forces are defined
tial energy for the case of torsional buckling. Silva23
on planes that are perpendicular to the line. Ojalvo chose
showed that if the square of the slope of the centroidal axis
the centroidal line as the reference line for deriving his the-
is replaced by the sectional average of the square of the
ory. The issue is whether Ojalvo’s line idealization is satis-
slope, one would obtain the conventional potential energy
factory and if so, whether it is superior to the conven-
expression. He added that Ojalvo’s approach for determin-
tional model.
ing potential energy is valid if the applied load is con-
sidered as a point load acting at the centroid. He contended
2.3. Filament model
that if the end traction is equally distributed across the end
Lenz and Vielsack** had remarked that in reality the length sectional surface, then the averaged square of slope should
of a closed-profile thin-walled member does not shorten be adopted. Ojalvo I4 disagreed with Silva’s proposal of
under torsion although a model based on longitudinal fibres having two alternatives and insisted that his approach is the
or filaments would predict that it would shorten. Ojalvoi3 only valid one for a bar theory. The important questions
appears to suggest that even open-profile members will not arising are: what is the correct potential energy expression
shorten in contrast to the conventional theory which pre- compatible with the established principles of mechanics;
dicts otherwise. The issue is whether Lenz and Vielsack’s and whether statically equivalent end tractions can lead to
remark on closed-profile members also applies for open- substantially different buckling loads.
profile members.
3. Flexural buckling of axially loaded pin-ended
2.4. Fictitious lateral load
column
Ojalvo objected to the use of a fictitious external load that
Before addressing the issue of flexural-torsional buckling,
leads to the Wagner term through the static-kinematic
the familiar flexural buckling of an axially loaded pin-
approach. He claimed that such an interpretation is equival-
ended column is revisited to highlight important aspects of
ent to a ‘spaghetti-like’ twisting idealization for the
its free-body diagram and the alternative representation of
deformed configuration which does not correspond to a bar
the stress resultants, which will elucidate the ensuing dis-
idealization. The objection here is to the application of such
cussion.
fictitious loads independently on each and every longitudi-
Referring to the straight column and its deformed con-
nal fibre, in order to represent the effect of the axial load
figuration as shown in Figure 2, equilibrium dictates that
on a twisted open-profile column. He suggested that in
reality where such fictitious loads are not present, the Ncos 8+ Vsin B=P (6a)
torque denoted by the Wagner term violates statical equilib-
rium. The issue is whether the governing equations can be Nsin 0- Vcos tI=H=O (6b)
derived by the static-kinematic approach without resorting Pv’-Iv’=0 (6~)
to the use of a fictitious load.
P P
e
2.5. Skew plane of deformed section
1 1 LLI -H
As an axially loaded column bends and twists, the longi-
tudinal fibres which were originally parallel to the column
V
axis would twist and warp around the shear centre axis. In
the conventional theory, the longitudinal stresses in these
inclined fibres are thought to produce a lateral effect on the
sloping surface of the warped member section. The total
torsional effect of the lateral effects about the shear centre
axis is P@’ - Px,v’,’+ Pysuns; that is how some texts on
the subject explain the Wagner term. The P?, factor in the tP tP tP tP
Wagner term is viewed as a reduction effect on the torsional
rigidity GJ of the column l5. Ojalvo’ pointed out that this (al lb) (cl Id)
torsional moment applies to an area that is not a plane and Figure 2 Flexural buckling of pin-ended column: alternative
not normal to the displaced longitudinal axis of the column. representation of stress resultants
128 Wagner term in flexural-torsional buckling: W. A. M. Alwis and C. M. Wang

where v is the lateral displacement; P the vertical load; N, by a series of parallel thin rigid links, rigidly fixed orthog-
V and M are, respectively, the axial force, shear force and onally to the bar axes and spaced arbitrarily close to each
bending moment at a generic cross-section; H is the hori- other, as shown in Figure 3a. The midpoint of each link is
zontal component of internal forces and 0 the slope of the prevented from lateral displacement (see Figure 3b), but
deformed column at the considered section. Note that, each link is free to rotate and move vertically. The distance
equations (6~) and (6b) can be alternatively written as between the individual links is allowed to change. The cen-
tral reference axis runs vertically through the midpoints of
N=Pcos 0 (7a) the links. Note that the centroidal axis of the system
coincides with the shear centre axis.
V = P sin 19 (76) Consider a vertical conservative load P on each bar.
Apart from axial deformation, the only other possible mode
We examine the following two potentially misleading
of deformation is symmetric twisting of the bars about the
notions pertaining to the relationship between P and the central axis as shown in Figure 3c. Considering the free-
stress-resultants, N, V, M and H.
body diagram of a longitudinal element of the model as
shown in Figure 3d, the pertinent equilibrium equations can
Shear force and lateral force: In this bifurcation analy- be written as follows. For the bar components, by taking
sis, the deformation variables v and 8 are infinitesimally moments about the radial and vertical axes, respectively,
small. Owing to the difference between P and N being of one obtains
second-order of 8, both can be taken to be of the same
magnitude. However, it will be questionable if one equates Pv’+H-M’=O (8)
V to H. The reason is that H is actually zero; whereas V=
P sin f3and is of the same order of magnitude as the terms T’ + H’r = 0 (9)
in the governing equation. If one accounts for lateral forces
and their moments, the use of H would yield completely where T, M and H are, respectively, the torsional moment,
different results from those obtained by the use of V. In the bending moment and the lateral force (acting orthog-
this pin-ended column, the fact that H = 0 highlights that onally to P) on the cross-section of each bar. Note that
the load P cannot produce a torque about any axis parallel v = r4, for infinitesimally small displacements; and the rad-
to the line of action of P. In contrast, if one replaces H ial forces by the mechanical links do not enter the foregoing
with V when determining the torque about such an axis, a equations. Equations (8) and (9) form the basis for the fol-
‘paradoxical’ first-order term will result. In view of equ- lowing discussion on the Wagner term and other related
ation (6b), this term would, however, be cancelled out by issues.
the torque produced by N. This point is crucial in view of Equation (8) is the same equilibrium equation as is nor-
the widely held notion that the vertical load P produces an mally written for bending of the Euler column as shown in
external torque about the vertical axis due to the inclination equation (6c), except for the presence of the lateral force
of the longitudinal fibres under torsional buckling. H. Equation (9) signifies the essential explicit link between
the internal lateral force gradient and the internal torsional
Role of jictitious load: Another issue of conern is the moment gradient. In other words, if a boundary condition
use of a fictitious lateral load Pv” (see Timoshenko and stipulates that the overall torsion at any section must be
Gere3) to represent the horizontal action of the vertical load zero, an internal torsional moment T can still exist together
P. The fictitious load produces H = Pv’. Such a represen- with a nonzero internal lateral force H. If one attempts to
tation is not identical to the correct one, i.e. H = 0 and V= represent the foregoing two-bar example by a single equiv-
Pv’. Furthermore, the fictitious load introduces horizontal alent bar along the centroidal axis, this T and H combi-
reaction forces at the supports while there are none in the nation will nullify each other. Hence, this particular tor-
original configuration. Thus it would become impossible to sional effect cannot be captured in an equivalent single line
have a statically admissible free-body diagram containing or single bar model.
one of the supports, once the axial load has been replaced By eliminating H from equations (8) and (9), one obtains
by the fictitious load. The possible incorrect inferences sug-
gest that one must be careful where such fictitious loads p$q - T’ - M”r = 0 (10)
are used. This is of interest in the present context since in
one form of derivation of the flexural-torsional equilibrium The substitution of the familiar constitutive relations T =
equatiot?, the Wagner term arises directly from a fic- GJ+’ and M = -EIpff where GJ is the torsional rigidity
titious load. and EI, the bending rigidity of one bar, into equation (10)
leads to

4. A simple torsional buckling model PI-=@ - GJV + ,EIx~ = 0 (11)


Recognizing the significant difference between H and V in
flexural buckling, a mechanically connected two-bar model Equation ( 11) takes the form of the familiar torsional buck-
is considered next to demonstrate the crucial role of the ling equation for the case of coincident shear centre and
horizontal stress resultants in torsional deformation. This centroid, if one replaces ?I, by 1,. Note that in the conven-
model simulates an axially prestressed bar-like three- tional theory of torsional buckling of doubly-symmetric I-
dimensional system. The reason for considering such a sim- columns, this replacement is admitted with r denoting the
plified model before addressing the general thin-walled radius of gyration of a flange about the web axis.
open-profile column is to avoid any confusion arising from In the following, the Wagner term will be interpreted in
the complications of section geometry and deformation. the context of the foregoing formulation. First we note that
This model consists of two vertical bars 2r apart, connected P and H acting on the bar sections are internal forces and
Wagner term in flexural-torsional buckling: W. A. M. Alwis and C. M. Wang 129

I 2P

r
r
%

@I (b)

CC) cd,
Figure 3 Two-bar torsional buckling model

that H can be expressed in terms of the normal and shear Further, note that the shear force V and the normal force
forces in the form N in the column are given by

H= VCOS e-Nsin 8 (12) V=HcostI+Psine (13)

The V cos 8 part of H is the contribution from the internal N=Pcose-Hsine (14)
moment gradient and leads to the EZ&“’ term. The N sin
8 part of H is the contribution from the internal axial force The Wagner term should not be interpreted as the torsional
and produces the Wagner term PP#‘. Thus the Wagner moment due to P sin 0 which is a part of V as shown in
term is a torsional effect of the internal normal stress equation (13). Although the magnitude of the result of
components in a deformed section. using P sin 0 is correct, it produces a torque acting in the
In Ojalvo’s single bar derivation for torsional buckling, opposite sense of the correct internal moment, leading to
the inclusion of GJ@’ and EI,@“’ terms in the torsional the misconception that the Wagner term is an external
equation amounts to admitting the contributions of T and effect. Such an interpretation of the Wagner torque as an
V cos 8, respectively. However, the contribution of N sin external effect is fraught with difficulties when considering
0 in its spatial form had been replaced erroneously by the the statical equilibrium of a free-body segment. The intro-
contribution of the statical resultant axial force passing duction of a fictitious lateral load, to represent the effect of
through the centroid. In the present two-bar model, the spa- axial load on the deformed configuration, further reinforces
tial form of N sin 6 produces a nonzero torque while Ojal- the inaccurate notion that the Wagner term is an external
vo’s resultant force representation leads to a null moment. effect. Moreover, in such a representation, the Wagner term
As a result, Ojalvo’s theory predicts that torsional buckling arises from the shear stresses generated by the fictitious
is not possible for the #considered model. This pinpoints load and this has led to the linking of the Wagner term to
the inconsistency of Ojalvo’s static-kinematic derivation, shear stresses on the sectional surface. This is incorrect,
where the spatial effects of some internal forces are admit- and as pointed out earlier, the Wagner term arises from the
ted and others are compromised. internal normal stresses in the axially loaded member.
130 Wagner term in flexural-torsional buckling: W. A. M. Alwis and C. M. Wang

5. Flexural-torsional buckling of open-profile face forms a valid boundary along the member axis. The
columd incremental stresses due to flexural-torsional deformation
are as follows: 7: is the warping shear stress, T; the shear
In the preceding section, the two-bar model has illustrated stress due to bending, r;1, the St Venant shear stress, a:
the basis of the Wagner term and clarified its misleading the normal stress due to bimoment and ai the normal stress
interpretations. It remains now to derive the governing due to bending. Note that in the conventional theory, it is
equations with the Wagner term for a general thin-walled assumed that the transverse shear stress across the wall sec-
open-profile column using the static-kinematic approach in tion in the direction orthogonal to the middle surface is null.
a similar manner. The torsional moment Tz of these stresses about an axis
The following points must be noted before attempting to parallel to the original member axis is derived next. As
derive the governing equation from a free-body diagram. discussed in Section 1 of this paper, the terms -EZ&“’ and
GJ4’ are due to r: and r:v. The stresses ri, a; and a:
In the conventional linear theory of structures, the consti- do not produce first-order moments. What remains to be
tutive relations between generalized stresses and strains determined is the contribution from a,. The components of
are expressed subject to the assumption that geometrical oO normal to the z-axis are -a,~’ in the x-direction and
changes due to deformation do not affect the basic -a,~’ in the y-direction, where the displacements u and v
relations. The linear theory guarantees a unique solution at (x,y) are given by
for a given structure and loading. The alternative solution
provided by the Euler concept of buckling, is due to a
u = 4 - (Y - YJ4 (15)
modified application of the linear theory. In bifurcation
buckling analysis, linear relationships are expressed link- v = v, + (X -X,)4 (lo)
ing strains of an assumed deformed configuration, and
stresses that are in equilibrium with the loading in the in which U, and v, are the displacements of the shear centre
same deformed configuration. In other words, in the con- located at (x,,y,). Since the resultant force of all sectional
ventional analysis, equilibrium is expressed for the undis- stresses has to be parallel to the z-axis, the moment of
turbed form of the structure whereas in buckling analysis, stresses about any axis parallel to the z-axis will be the
equilibrium is expressed for a deformed state of the struc- same. It follows that, by taking moments about an axis par-
ture. allel to the z-axis through the displaced shear centre of the
The introduction of such a variation does not pose any section, the contribution to T, from a, is given by
difficulty in flexural buckling analysis of columns since
the bending moments in an initially imperfect column
can be readily related to the change in curvature from -a<> [U’(Y - Y,) - V’(X - -41 dA
an initial straight form to an assumed imperfect shape. i A
However, in flexural-torsional buckling analysis of thin-
walled open profile columns, the application of the Euler which on substitution from equations ( 15) and ( 16) yields
concept of buckling requires careful consideration of the
assumed deformed configuration. The main difficulty lies
in the fact that when an initially straight member is
twisted subject to the conditions of no cross-sectional where A denotes the area of the section and P = u,,4. Hence,
distortion and no shear deformation in the middle sur- the torsional moment of the internal stresses acting on the
face, the initially parallel longitudinal lines of the mem- deformed section, about the original member axis is
ber take spiral forms which are no longer parallel to given by
each other.
TI = -EZw+“’ + GJ+’ - P(+#J’ + u,& - v,;x,~) (17a)
Consider the stress components acting on a warped open-
profile section as shown in Figure 4 which is subjected to The moment gradient about the z-axis is thus given by
a uniform axial prestress o<,. The sectional surface is a plane
that is orthogonal to the axis of the undeformed member, ml = EIwv - GJ$” + P(r&V + u;yy, - v’$~) (17b)
but after torsional deformation the sectional surface would
not be a plane in general. Nevertheless, this sectional sur- For flexural-torsional buckling of the open-profile column
under axial load

mZ = 0 (18)

The substitution of equation ( 17b) into equation ( 18) leads


to equation (3).
It is worth noting that the foregoing derivation is based
on a uniformly distributed end traction a,. This form of end
traction does not produce a bimoment. If other statically
equivalent end tractions are applied, such as a concentrated
load at the centroid, in general there will be a nontrivial
bimoment distribution along the member. The presence of a
bimoment results in a nonuniform normal stress distribution
across the section. Hence equation (17) may not apply in
Figure 4 Free-body diagram such a case. The St Venant principle for end tractions will
Wagner term in flexural-torsional buckling: W. A. M. Alwis and C. M. Wang 131

not be valid in general for longitudinal end forces applied expression (23) for an axially prestressed open-profile thin-
on thin-walled open-proflle members5. Statically null sys- walled member can be written as
tems of end tractions can give rise to primary longitudinal
stresses which may affect the buckling load. u=w-P(6,+6,.)+U, (25)

6. Work-energy approach Hence the total incremental energy functional is given by

Goto and ChenzO claimed that Ojalvo’s energy derivationI


II = u- w= -P(S, + S,,) + u, (26)
is invalid because it violates the constant load condition.
In a recent paper, the writers2’ clarified that there is no need
Note that II is independent of any second-order changes
to insist on the constancy of load for a valid bifurcation
assumed for end-to-end distance 8, or the load AP in the
analysis through the work-energy approach. As Ojelvo’s
buckled configuration considered.
error does not lie in imposing a constant load, it is then
The deficiency of the bar model proposed by Ojalvo can
imperative to determine where the mistake lies in his
also be readily identified in the foregoing work-energy deri-
energy formulation.
vation. A bar model will not contain the 6, term which is
In the work-energy approach, up to second-order changes
the axial shortening due to twisting of a three-dimensional
in incremental strains and stresses need to be accounted for
member. It is clear that the corresponding energy term PS,
when considering a deformed configuration. All terms of
is precisely the difference between Ojalvo’s energy func-
order higher than the second can be neglected. For compati-
tional and the conventional one. Ojalvo had assumed that
bility, it is necessary thal:
the bar length does not change (i.e. S = 0) and has
implemented 8, + 8, = 0. This implementation is correct
; AP = 6, + 6, + a,, = S (19) only when the member is represented by its centroidal axis
ignoring the longitudinal effect of twisting in three-dimen-
sions subject to the Euler-Bernoulli hypothesis.
where L is the member length, EA the axial rigidity, AP The Wagner term in the governing equation (3) results
the second-order increment in the compressive load, 6, the from the -P(& + S,,) term, This term is an exclusive
second-order increment of the average end-to-end distance component of internal strain energy U as can be observed
in the deformed configuration relative to the prebuckling from equation (25), and it is not a part of the external work
configuration, S, the change in end-to-end distance due to W given by equation (22). This result is in agreement with
bowing of the member, a,, is the contraction of the member the observation made in the preceding static-kinematic deri-
due to twisting while maintaining the Euler-Bernoulli vation that the Wagner term denotes an internal stress
‘plane section’ condition, and 6 the total change in member resultant. Many textbook derivations consider the special
length. The second-order changes in length 6, and 8, are, case of constant load, AP = 0 which is tantamount to stating
respectively, given by that 6 = 0. Although this additional constraint is unnecess-
ary and irrelevant to the final result, its introduction makes
the expression for external work W appear as P( 8, + S,),
8, = ; L (us’ + v;‘) dz (20) giving the false impression that the consequent Wagner
I 0
term is an external load effect.
Lenz and Vielsack22 have commented that according to
6, = ; < L @2 dz (21)
I 0
experimental evidence, twisting of a closed-profile member
does not shorten it. If the deformation of the member is
in which r, is the polar radius of gyration of the member taken to be due to changes in lengths of longitudinal lines
section with respect to the centroid. that deform to spiral shapes as the member twists, a short-
The incremental work done by the load and the ening is predicted. It should be noted that when there is
incremental strain energy in the member can be expressed, twisting of such a closed-profile member, the wall middle
respectively, by2’ surface is subjected to shear deformation, or else twisting
cannot occur. In the linear theory, pure shear on a square
w=-P62 (22) element deforms it to a rhombus, yet leaves the direct
strains along edges of the square unaffected. However, the
lJ=-P6+ u, (23)
thin-walled open-profile member theory is based on the
assumption that the middle surface of the walls undergoes
in which U, is the conventional internal strain energy
no shear deformation. This assumption implies that any
component6 given by
change in length of longitudinal lines on the middle surface
(which wrap around the torsion axis in response to the
U, = ; L (E&u;= + EI,v:~ + El,f= + GJc#i2) dz torque) is accompanied by an appropriate change in longi-
I 0 tudinal stresses. The geometry of a twist that satisfies the
(24) condition of no shear deformation in the middle surface
requires the longitudinal lines to tilt with respect to others
It is clear from equations ( 19) and (23) that the conven- depending on their spatial distances across the cross-sec-
tional strain energy functional implicitly presumes that the tion. 0jalvo’s’3 suggestion that Lenz and Vielsack’s com-
member length and the load does not change (i.e. 6 = 0 ment on closed-profile members also applies to open profile
and AP=O). members is unacceptable because it ignores the fundamen-
By the substitution for 8, from equation (19) and then tal difference in middle plane shear deformation in the the-
introducing equation (22), the general form of strain energy ories of closed- and open-profile thin-walled members.
132 Wagner term in flexural-torsional buckling: W. A. M. Alwis and C. M. Wang

7. Conclusions Ojalvo in his energy formulation lies in having incompat-


ible strain and potential energy expressions.
A static-kinematic based explanation of the Wagner term in
It is hoped that the present clarification will be useful in
the flexural-torsional buckling of thin-walled open-profile
reaching a common ground on the Wagner term in flexural-
columns has been presented. The Wagner term is an
torsional buckling of thin-walled open-profile columns.
internal stress resultant arising from the normal stresses act-
ing on the member section. This is a torsional moment of
the same order as other torsional effects (such as St Venant References
torque and warping torque) when a member is prestressed
1 Wagner, H. ‘Verdrehung und Knickung von offenen Profilen,’ 25fh
prior to initiation of an infinitesimally small torsional defor- Anniversary Publication, Technische Hochschule, Danzig, 1929,
mation. However, when the prestressing force is of the 329-343 (NACA translation, TM 807)
same order as the applied torsional moment, the torque Timoshenko, S. P. Bull. Polytech. Inst., St. Petersburg, 1905
denoted by the Wagner term would become a higher-order Timoshenko, S. P. and Gere, J. M. Theory of elastic stability,
McGraw Hill, New York, 1961
term when compared with the other two internal stress
Goodier, J. N. ‘The buckling of compressed bars by torsion and flex-
resultants. Some issues raised by Ojalvo and others during ure,’ Bull. 27, Engineering Experiment Station, Cornell University,
various discussions on the subject have also been addressed Ithaca, New York, 1941
herein in the process of clarifying the Wagner term. The 5 Vlasov, V. Z. Thin walled elastic beams, (Translated by Israel Pro-
gram for Scientific Translation for National Science Foundation,
following is a brief comment on the main issues.
Washington DC, 1961) (Russian original published 1941)
The reservation raised about the Wagner term by Ojalvo 6 Bleich, F. Buckling strength of metal structures, McGraw-Hill, New
stems from the fact that there has been no statically admiss- York, 1952
ible free-body diagram to validate the Wagner term. This I Galambos, T. V. Structural members and frames, Prentice Hall,
issue has been clouded by the use of a fictitious lateral load Englewood Cliffs, NJ, 1968
8 Trahair, N. S. and Bradford, M. A. The behaviour and design of steel
to simulate the axial prestressing effect and the belief that structures, (2nd edn) Chapman and Hall, New York, 1988
the Wagner term is an external action. The presentation 9 Ojalvo, M. ‘Wagner hypothesis in beam and column theory,’ /.
herein of a statically admissible free-body diagram Engng. Mech. Div., ASCE 1981, 107 (4) 669-677
accompanied by clarifications of various relevant issues 10 Ojalvo, M. ‘Discussion on “Lateral-torsional buckling of tapered I-
beams”‘, J. Struct. Div., AXE, 1982, 108 (2), 503-504
using two simple models, establishes the validity of the
II Ojalvo, M. ‘Closure of “Wagner hypothesis in beam and column
Wagner term, without resorting to the use of a fictitious theory” ’, .I. Engng Mech. Div., ASCE 1983, 109 (3), 924-934
load or other means questioned by Ojalvo. 12 Ojalvo, M. ‘Discussion on “Buckling of monosymmetric I-beams
The deformed member section not being in a single plane under moment gradient”‘, J. Struct. Engng., ASCE 1987, 113 (6),
does not affect the derivation since all moments are taken 1387-1391
13 Ojalvo, M. ‘The buckling of thin-walled open-profile bars,’ J. Appl.
about a single axis, accounting for the direction of different Mech., ASME 1989, 56, 633-638
stress components acting on the section. It is shown that 14 Ojalvo, ‘M. Closure of “The buckling of thin-walled open-profile
torsional deformation in an open-profile member involves bars” ‘, J. Appl. Mech., ASME, 1990, 57, 479-480
three-dimensional effects which cannot be simply modelled 15 Kitipomchai, S. and Dux, P. F. ‘Discussion on “Wagner hypothesis
in beam and column theory”‘, J. Engng. Mech. Div., ASCE 1982,
by a single line representation unless due care is exercised.
108 (3), 570-572
In the case of Ojalvo’s bar theory, two spatial effects (St 16 Studnieka, J. and K&tec, V. ‘Discussion on “Wagner hypothesis in
Venant torsion and warping torsion) have been correctly beam and column theory”‘, J. Engng. Mech. Div., AXE 1982, 108
included, but the spatial effect of the normal stresses in (3). 573-574
the deformed configuration has been oversimplified. This 17 Trahair, N. S. ‘Discussion on “Wagner hypothesis in beam and col-
umn theory”‘, J. Engng. Mech. Div., ASCE 1982, 108 (3). 575-578
oversimplification has led to Ojalvo’s erroneous conclusion 18 Haaijer, G. ‘Discussion on “Wagner hypothesis in beam and column
that torsional buckling cannot occur in members whose theory” ‘, J. Engng. Mech. Div., ASCE 1983, 109 (3), 923-924
shear centre and centroid coincide. 19 Kitipomchai, S., Wang, C. M. and Trahair, N. S. ‘Closure of “Buck-
The conventional theory for flexural-torsional buckling ling of monosymmetric I-beams under moment gradient”‘, J. Strucr.
Engng., ASCE 1987, 113 (6), 1391-1395
of axially loaded columns applies to cases where the bimo-
20 Goto, Y. and Chen, W. F. ‘On the validity of Wagner hypothesis,’
ments at the ends are zero. The case of the uniformly dis- ht. J. Solids Srrucr. 1989, 25 (6). 621-634
tributed end-tractions is one such instance. If the end bimo- 21 Alwis, W. A. M. and Wang, C. M. ‘Should load remain constant
ments are nonzero, the primary longitudinal stresses will when a thin-walled open profile column buckles?’ Int. J. Solids
not be uniformly distributed, leading to a possibly different Strucf. 1994, 31 (21) 2945-2950
22 Lenz, J, and Vielsack, P. ‘Eine kritsche Bemerkung zur Theorie des
buckling solution. Drill-knickens,’ Der Stuhlbau 1980, 49 (8), 245
The work-energy approach presented herein comp- 23 Silva, R. R. ‘Discussion on “The buckling of thin-walled open-profile
lements the static-kinematic derivation. The error made by bars” ’, J. Appl. Mech., ASME 1990, 57, 479

You might also like