FORUM

Three Hundred Years of Bar Theory

Morris Ojalvo crement becomes the initial configuration for the next increment
Professor Emeritus, The Ohio State Univ., Columbus, Ohio. E-mail: until the body is fully loaded. Although much has been done with
[email protected] finite displacement theory, it remains true that results may gener-
ally not be obtained with a single application of the pertinent
This analysis reviews the ideas, insights and assumptions that are elasticity equations.
critical for a comprehensive formulation of a theory for bars. One Bar theory achieved a certain flexibility as it developed. When
assumption, the Wagner hypothesis, was thought to be necessary the bar’s original configuration is used for expressing equilibrium,
for explaining what appears to be pure torsional buckling when it is called a first-order theory and the equations derived are not
certain bars are compressed. The characterization of the phenom- coupled in the displacement terms. The equations will also be
enon as pure torsional buckling is incorrect. The hypothesis is not linear if the displacements are small and the material the bar is
needed, and without it, the theory returns to a simpler and more formed of has a linear stress-strain behavior. Within bar theory
rational earlier form. there is also a second-order component for which small displace-
Bar theory had its early beginnings in 1705, when James ments are identified and whose analysis proceeds with the under-
Bernoulli proposed that the curvature of a bent bar is proportional standing that products of such displacements and distortions are
to the bending moment and that internal resisting couples result small enough to discard and the equilibrium is expressed with a
from the extensions and contractions of the bar’s longitudinal first-order theory estimate of the final configuration. Additionally,
filaments. Daniel Bernoulli, James’s nephew, communicated the the stress resultants used for the constitutive equations that deter-
idea to Leonhard Euler in 1742, and Euler, acting on it, solved the mine local curvatures, twists, and axial extensions are determined
elastica problem in 1744 共Love 1944兲. Later, Euler defined buck- for planes perpendicular to the deformed centroid line. The equa-
ling and computed buckling loads. His work laid the foundation tions for this version of bar theory are coupled in the displace-
of what is known as stability theory. ment terms but will still be linear if the material’s stress-strain
Euler considered a bar as though it were a line of particles that relationship is linear. Euler’s buckling solutions are from the
curved in accordance with Bernoulli’s principle. In 1776, Charles second-order bar theory and are encountered when the loads are
Augustin de Coulomb 共Love 1944兲 examined the normal planes such that the equations are homogeneous and the boundary con-
of a bent bar. By assuming that the profiles did not warp 共plane ditions needed for their solution are homogeneous as well.
sections remain plane兲 and with known or assumed properties for The second major innovation to bar theory was the inclusion
the bar’s material, he was able to determine the neutral axis for of the St. Venant torsion theory to the analysis of bars. In 1899, A.
bending. From that time, the bar was considered as a three- G. M. Michell 共1899兲 and L. Prandtl 共1899兲 used it for their
dimensional body for purposes of analysis. analyses of flexural-torsional buckling of rectangular cross sec-
It is evident that bar theory rests on “special hypotheses” that tion beams with large depth-to-width ratios. Although St. Venant
make it approximate. Plate and shell theories are also approxi- torsion theory applies only to cases of uniform twist, it seems to
mate for their dependency on special hypotheses. Such theories work well with flexural-torsional buckling problems where the
are useful and occasionally needed, as when the elasticity solution twist is rarely uniform. From this point, the most important con-
is not available. They typically supply the stresses, strains, and tributions to basic bar theory were in the area of thin-walled open-
deformations that most concern the structural engineer and ma- profile bars with cross sections that can be assumed rigid either
chine designer and are simpler to use than the elasticity theory. because transverse stiffeners are used or because the sections are
The drawback to an approximate theory is that it often pro- sufficiently compact to hold their shape.
duces inconsistent and incomplete results. For example, bar The third major innovation came about from a meeting be-
theory is inconsistent when it predicts a transverse shear stress on tween L. Prandtl and S. Timoshenko. In 1904 and 1905, Russia
the normal plane of the bar and neglects the shear strain it should was troubled by political problems and a war with Japan, which
produce. It is incomplete as well because a prismatic bar of ir- broke out in 1904. The Petersburg Polytechnic Institute closed
regular cross section will generally not have a true shear center down, and Timoshenko, who had a position in its mechanics labo-
and shear centerline so that it is not possible to know whether ratory, sought to increase his knowledge by studying under
transverse loads will twist the bar as well as bend it. S. Timosh- Prandtl and other professors at Göttingen University. Prandtl sug-
enko suggested that for some bars, the error might be small if the gested to Timoshenko that he might try to verify Prandtl’s solu-
centroid line is also taken to be the shear centerline. tions experimentally, but Timoshenko chose instead to extend
Elasticity theory is considered to be exact because its results Prandtl’s work to I-beams. Timoshenko reported that his progress
are complete and consistent. However, it includes one glaring was rapid once he recognized the additional torsional strength and
inconsistency that is masked by specifying that it is meant for use stiffness supplied by the flanges when the warping of a cross
when strains and displacements are small. An analysis begins section is restrained. This concept was his major contribution to
with the initial configuration of the body, and that configuration is bar theory and became the basis of his doctoral dissertation. In his
used to express the equilibrium of the deformed body in its final memoir, Timoshenko 共1968兲 expressed the belief that his work on
state. If a truer representation is desired, the loads may be incre- plates was his most important in the field of stability. However, if
mentally applied and the configuration determined after each in- one considers that the onset of buckling in plates is often not the

1686 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / DECEMBER 2007

Downloaded 17 Aug 2010 to 130.54.110.32. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
limit of their load-bearing capacity as it is for bars, it may be that load systems, the effects are found to be far-reaching even though
his work with I-beams at least rivals it in importance. only one of the equations is altered because the equations are
By the early years of the 20th century, ample evidence indi- coupled 共interrelated兲 in the four dependent variables. The four
cated that thin-walled open-profile bars would twist when the dependent variables are the bar’s rotation about the shear center
action lines of transverse forces do not precisely intersect a cer- and two transverse displacement functions for the shear centerline
tain line running parallel to the bar’s length. We know that line to and the longitudinal displacement function for the points of the
be the shear centerline as defined by the shear center in each cross centroid line.
section. In 1922 and 1924, R. Maillart 共Vlasov 1961兲 showed how To understand how displacements are determined throughout
the shear center could be analytically located. A moment gradient the bar, one must consider the functions of the centroid and shear
is assumed; the shear flow for the entire profile is determined; and centerlines. They are idealizations that may not even occupy a
from it, the line of action of the shear flow resultant is deter- physical part of the bar, but they are also crucial metaphors for
mined. When repeated for a moment gradient causing bending describing the displacements. The longitudinal displacement of a
about another axis of the cross section, the two shear flow result- point on the centroid line corresponds to a longitudinal displace-
ant lines of action locate the shear center of the profile. Timosh- ment of all points on the corresponding cross section by the same
enko remarked that this development came late to the subject of amount. Two transverse movements of a point on the shear cen-
strength of materials. We believe Maillart’s insight is deserving of terline determine the translation of the corresponding profile in its
recognition as the fourth major innovation to bar theory. plane. A rotation angle at a point on the shear centerline corre-
The fifth and sixth major innovations are from a paper by sponds to the rotation of a profile about the shear centerline. The
Herbert Wagner 共1929, 1936兲. The fifth expanded on Timoshen- points in a profile continue to remain in a plane as the cross
ko’s discovery of enhanced torsional properties when warping is section tilts about a centroidal axis for the area. The tilt angle is
suppressed in I-section bars by generalizing it to thin-walled determined by the slope of the deflected shear center 共not cen-
open-profile bars of any cross section. His approach continues troid兲 line. A final longitudinal movement of points in a cross
with the assumption that shear flow forces do not produce corre- section is determined by the rate of twist of the shear centerline in
sponding shear strains. accordance with St. Venant torsion theory. The movement pro-
Wagner’s paper, translated into English for the National Advi- duces a warping of the profile in a pattern for which the average
sory Committee for Aeronautics as Technical Memorandum 807 longitudinal displacement remains the same. Thus, the average
共Wagner 1936兲, is difficult to follow, but the essentials are all longitudinal movement from all causes of points in a profile con-
there. For a fuller appreciation, one may refer to V. Z. Vlasov’s tinues to be that of the corresponding longitudinal movement on
1959 treatise Thin-Walled Elastic Beams, available in English the centroid line. The computed final displacement of a particular
translation as Vlasov 共1961兲 or to the writer’s Thin-Walled Bars point within the bar is the sum of its individual movements from
with Open Profile 共Ojalvo 1990兲. Vlasov used the term law of each of the profile displacements described previously.
sectorial areas for what Wagner called unit increase in warping. Second-order bar theory equations derived with the Wagner
Wagner’s second long-lasting contribution has been problem- hypothesis predict the possibility of pure torsional buckling, and
atic from the start. It is known as the Wagner hypothesis, and its visual observations appear to confirm this as the failure mecha-
original purpose was to explain a curious type of compression nism. But the hypothesis also leads to several insurmountable
buckling behavior that occurs only when the shear centerline of problems that preclude its adoption.
an open-profile bar coincides with the centroid line. If such a The first problem is that the hypothesis proposes a discontinu-
column with flat and parallel ends is placed vertically between ous model when, for the real bar, displacements must be continu-
parallel platens of a testing machine and loaded, it will at first ous throughout. J. Lenz and P. Vielsack 共1980兲 noted that pure
shorten. Then, at a critical value of the load, the onset of one of torsional buckling requires each filament to be inextensible as it
two new types of distortion signals the beginning of buckling. If winds as a helix on a cylinder of constant radius between two
the column is long, the new distortion will be of the flexural type, cross sections. Because of their inextensibility, the ends of the
there is an abrupt reduction of longitudinal stiffness, and further filaments move toward each other in an axial direction. If the
shortening is accommodated by a bowing of the column. If, how- theorem of stationary potential energy is used for determining the
ever, the column is short, the new type of distortion is one where buckling load for axial compression, the potential of the end
all cross sections appear to rotate about the common shear center- forces is determined from the shortening of each filament and the
centroid line while the common line remains perfectly straight. distribution of the compressive stresses over the end faces. Dif-
There is also an abrupt loss of vertical stiffness, and continued ferent distributions produce different potentials for the same end
shortening occurs as for long columns. It is this short column load so that the buckling load changes with the distribution of end
behavior that the hypothesis attempts to explain and that has been stresses. In 1952, H. Bleich noticed the anomaly, although he
called pure torsional buckling. misdiagnosed the reason for it. The real reason is that the station-
In 1950, J. N. Goodier succinctly described the hypothesis as ary potential energy theorem depends on the virtual work theo-
follows: rem, and the latter only works when the displacements are those
of a continuum.
In the 共bar’s兲 untwisted form the 共longitudinal兲 “fibers”
The hypothesis also does not explain how a bar can twist in
have compressive stresses in them. In the twisted form the
torsional buckling while plane sections normal to the axis of twist
fibers run through a cross section in a twisted pattern. The
remain free of a torsional couple about any longitudinal axis. Nor
fibers are supposed to carry their compressive stresses
does it explain the abrupt loss of longitudinal stiffness at buckling
with them. These stresses, acting on a cross section, being
or the continued shortening that follows. Lenz and Vielsack
no longer parallel to the axis, form a couple about the axis
共1980兲 stated that there is experimental and theoretic evidence
as well as a force resultant 共the thrust兲.
that a twisted bar does not shorten and may actually elongate.
When, as eventually happened, the hypothesis is brought to the This writer’s response to the problem 共Ojalvo 1981兲 was to
derivation of the four bar equations for other profiles and other return to bar theory as it was before the Wagner hypothesis. It

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / DECEMBER 2007 / 1687

Downloaded 17 Aug 2010 to 130.54.110.32. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
Fig. 1. 共a兲 Profile, cruciform bar; 共b兲 profile, cruciform bar as four
T-Bars
Fig. 3. 共a兲 Profile, bar which may exhibit pseudo pure torsional
buckling; 共b兲 profile, bar which cannot exhibit pseudo pure torsional
includes a return to the use of the planes perpendicular to the buckling
deformed line of centroids for determining internal stress result-
ants. These stress resultants are axial force, flexural moments, the
bimoment, and the moment of the shear stresses about the shear
center of the profile. His equations for determining buckling for When columns buckle with an enforced axis of rotation, dis-
centrally compressed thin-walled open-profile bars were pub- tributed forces and couples are generally needed along the axis. In
lished in 1981. They show that flexural and flexural-torsional the usual case, they are supplied through an external agency. But,
buckling can occur but that pure torsional buckling cannot. When as can be seen from Fig. 1共b兲, these forces are self-generated
the column buckles in a flexural or flexural-torsional mode, the when all four T-columns buckle in unison.
abrupt loss in longitudinal stiffness, as well as the continued The author’s general second-order theory equations 共Ojalvo’s
shortening, is easily attributable to the bowing of the column 1990兲 were used to determine the flexural-torsional buckling of an
caused by the flexural distortion. Also, when torsion is a part of angle section column constrained so as to buckle with a longitu-
the distortions, the bar is subject to couples in the normal planes. dinal axis of enforced rotation running through the tip of one leg
There remains the problem of explaining the apparent occurrence 共Ojalvo’s 2002兲. Two identical angle sections aligned as shown in
of pure torsional buckling when the shear center and centroid of a Fig. 2共a兲 and rigidly connected form a Z-section. When the
cross section coincide. At the time, the writer had no explana- pseudo pure torsional buckling load is computed for the
tion beyond suggesting that it was a manifestation of a local Z-column, it is twice the flexural-torsional buckling load of one
instability. angle column. The required self-generated restraining forces for a
In Ojalvo 共1989兲, the same equations were arrived at, this time pair are as shown in Fig. 2共b兲. For this study, the postbuckling
by using the theorem of stationary potential energy. They were shortening of the angle column was expressed in terms of the
used to determine the several types of buckling possible for the amplitude chosen for the rotation function defining the angle’s
flanged cruciform whose profile is shown in Fig. 1共a兲. The col- rotation about the enforcing axis of rotation.
umn was considered as four T-section members joined rigidly The profile shapes of columns that can exhibit pseudo pure
where the tips of their stems meet at the shear center-centroid line torsional buckling are limitless. They all require coinciding shear
of the cruciform column. The four T columns can buckle simul- center and centroid lines. But some caution must be exercised.
taneously, each in a flexural-torsional mode with the common line The column with a profile as indicated by Fig. 3共a兲 is a member
of the cruciform column acting as an enforced axis of rotation for of a class of columns that may exhibit pseudo pure torsional
the T-members. At any cross section, all T’s rotate the same buckling whereas a column with a profile indicated by Fig. 3共b兲 is
amount and in the same direction. This behavior mimics a pure not. For the latter, a careful analysis of longitudinal displacements
torsional buckling of the cruciform column. But it only seems so along the shear center-centroid line reveals that, while total short-
to the casual eye, which cannot discern the much smaller longi- ening of the subparts along that line are the same, local disparities
tudinal displacements of a profile. The combined buckling loads exist that preclude pseudo pure torsional buckling.
for the four T-sections exceeded the “pure” torsional buckling Clearly, tests should be conducted to determine if the Wagner
load as computed for the cruciform under the Wagner hypothesis. hypothesis is wrong and, if wrong, also not needed. These tests
This should be anticipated in view of the relaxation of internal should include columns with coinciding shear center-centroid
constraints implied by the hypothesis. lines as outlined in the preceeding. It will be interesting to find
out whether Occam’s razor applies for bars. That ancient philo-
sophic principle holds that for competing theories, the less com-
plex one is preferred and that the explanation of unknown phe-
nomena should be first sought in terms of known quantities.

Acknowledgments

The writer is indebted to Cindy Sopher, Joseph Snyders, and Dr.

Irving U. Ojalvo who gave generously of their time and talents
toward the task of shepherding this document through the publi-
cation process. The advice, criticism and encouragement supplied
by Dr. I. U. Ojalvo when the need to settle the Wagner hypothesis
Fig. 2. 共a兲 Profile, Z-bar; 共b兲 profile, Z-bar as two angle bars question was not a high-priority item is also greatly appreciated.

1688 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / DECEMBER 2007

Downloaded 17 Aug 2010 to 130.54.110.32. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
References bars.” J. Appl. Mech., 633–638.
Ojalvo, M. 共1990兲. Thin-walled bars with open profiles, The Olive Press,
Bleich, H. H. 共1953兲. “Refinement of the theory of torsional buckling of Estes Park, Colo.
thin-walled columns.” Proc., 1st Midwestern Conf., Solid Mechanics, Ojalvo, M. 共2002兲. “Towards resolving issues concerning torsional and
Urbana, Ill. lateral-torsional buckling of thin-walled open-profile bars.” Proc.,
Goodier, J. N. 共1950兲. “Elastic torsion in the presence of initial axial 15th ASCE Engineering Mechanics Conf., ASCE, New York.
stress.” J. Appl. Mech., 383–387. Prandtl, L. 共1899兲. “Kipperscheinungen.” dissertation, Nuremberg.
Lenz, J., and Vielsack, P. 共1980兲. “Eine kritische Bemerkung zur Theorie Timoshenko, S. 共1968兲. As I remember, R. Addis, translater, Van
des Drillknickens.” A critical remark on the theory of torsional buck-
Nostrand.
ling, Der Stahlbau. Vlasov, V. Z. 共1961兲. Thin-walled elastic beams, 2nd Ed., 464, 465.
Love, A. E. H. 共1944兲. A treatise on the mathematical theory of elasticity,
Wagner, H. 共1929兲. “Verdrehung und Knickung con offenen Profilen.”
Dover, New York.
Michell, A. G. M. 共1899兲. “Elastic stability of long beams under trans- 25th Anniversary Number of the Technische Hochschule, Danzig, Vol.
verse forces.” Philos. Mag., 48, 298. 1904–1929, 329–343.
Ojalvo, M. 共1981兲. “The Wagner hypothesis in beam and column theory.” Wagner, H. 共1936兲. “Torsion and buckling of open sections.”
J. Engrg. Mech. Div., 107共4兲, 669–677. Technical Memorandum No. 807, National Advisory Committee for
Ojalvo, M. 共1989兲. “The buckling strength of thin-walled open-profile Aeronautics.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / DECEMBER 2007 / 1689

Downloaded 17 Aug 2010 to 130.54.110.32. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org