Professional Documents
Culture Documents
Torsion, Bending and Buckling of Steel Beams by Trahair 1997
Torsion, Bending and Buckling of Steel Beams by Trahair 1997
Very few steel design standards give rules for designing against
torsion. It appears that torsion is largely ignored because it is
thought to occur rarely, and when it does, to be unimportant. In
addition, difficulties in predicting torsional effects tend to discour-
age designers from considering torsion at all. However, the grow-
ing use of three-dimensional computer analysis programs has
alerted many designers to the presence of secondary torsion
actions in the structures they design, and has led them to question
their significance and to seek advice on how to design for them.
In addition, there are some circumstances where designers wish
to transfer primary loads by torsion, either alone, or in combination
with bending actions, but are prevented from doing so by difficult-
ies in predicting torsional behaviour and a lack of information on
how to design for torsion. This paper summarizes a series of inves-
tigations into the behaviour, analysis and design of members sub-
jected to combined torsion and bending. © 1997 Elsevier Science
Ltd. All rights reserved.
372
Torsion o f steel beams: N. S. Trahair a n d Y.-L. Pi 373
to be related to the twist per unit length dO/dz, in which G and a mitre model ]~ of the uniform torsion shear strain dis-
is the shear modulus of elasticity, J is the torsion section
tribution. In equation (7), w' is the axial strain, -o90" is the
constant, 0 is the angle of twist rotation and z is the distance
warping strain, and ½(x2+y2)0": is the nonlinear Wagner
along the member. The maximum uniform shear stress ~',,m
strain arising from the change in length of a longitudinal
can be used to define the uniform torsion elastic section
element through the point (x, y) as it becomes helical due
modulus
to the twist rotations 0. The cross-section distributions of
the warping, Wagner and uniform torsion strains are shown
Z,, =M,,/r .... (2)
in Figure 2. These strains are associated with the linear
stress resultants of the bimoment B, the Wagner stress
Cross-section analysis under warping torsion 5,7 9 allows the resultant
warping torsion normal stresses f,. and the bimoment
25
Member analysis allows the twist rotations 0 and the stress
resultants M,, B and M,. to be determined from the torque
distribution M= by solving 20 I o.SL_!_ 0.SL_]
Z L = ! ~ _ _ _.._....
= _ _:_._._._
M_ = GJdO/dz - El,.d30/dz 3 (6) 15
=o ............... -.."..-.-------..--.'..7.-.7--.--.-.:
o ...°, - ...........
subject to the boundary conditions. This can be done manu-
ally 9 or by using computer programs such as TW BEAM ~°.
.[ . First yield
The linear elastic method of torsion analysis is most logi- <
cally used tor serviceability design, for which most if not 5
all of the member remains elastic, so that the linear elastic - -- Plasdc coUapse predic~on
o Test results by Fro'well and G a l a m b o s 12
analysis closely predicts the twist rotations. These and any I I I I I
related deflections can then be assessed by comparing them 0.2 0.4 0.6 0.8 i.O 1.2
Cena'al twist rotation ~(radian)
with what are considered to be limiting values. Linear elas-
tic torsion analysis is less logically used for strength design, Figure 3 Comparison of analysis and test results
374 Torsion o f steel beams: N. S. Trahair and Y.-L. Pi
ing lowers the stiffness initially, but at higher twist respectively, and h is the distance between the flange cen-
rotations the member stiffens, and substantially higher tor- troids. The bimoment plastic section modulus is defined by
ques can be resisted than are predicted by first yield. The
variations of the end stress resultants M, and W are shown s,, = & J L (~2)
in Figure 4. It can be seen that while the uniform torque
M, reaches a maximum value and then decreases, the The flange plastic moment is
Wagner stress resultant W, which is initially quite small,
continues to increase, even after substantial yielding has Mo, = B,,/h : f,.b~t//4 (13)
occurred. It is the Wagner stress resultant which is respon-
sible for the stiffening effect at high twist rotations shown It has not been possible to develop a completely satisfying
in Figure 3. Final failure of the beams tested was reported theory for the plastic collapse of torsion members because
as being by tensile fracture at the flange tips, which is con- the separate plastic collapse modes for uniform and warp-
sistent with the higher Wagner strains which occur at these ing torsion cannot be simply but rigorously combined.
locations, as shown in Figure 2b. Instead, a simple approximate theory has been developed 2,
according to which the plastic collapse load factor A,:, can
2.3. Plastic collapse analysis be obtained from
Although methods of analysing plastic collapse in flexural
structures are well established 9,13:4, it is only recently that A,p = A,e + A.p (14)
a simple but sufficiently accurate method has been
developed ',2 for the analysis of the plastic collapse of mem- in which A,p and A,.p are the independent plastic load fac-
bers subject to torsion. This method also consists of two tors for uniform torsion collapse and warping torsion col-
parts: cross-section analyses to determine the fully plastic lapse. While this simple extension is not rigorous, compari-
uniform torque Mupand the fully plastic bimoment BI,; and sons with numerical and experimental data (References 1
independent member analyses of the collapse mechanisms and 12 and Figure 3) for members under a wide range of
for uniform and warping torsion. support and loading conditions have shown that it is con-
The sand heap analogy 9 can be used to analyse a thin- servative, principally because it neglects the strengthening
walled open section to determine the uniform torsion plas- effects of the Wagner stresses that develop at large
tic torque rotations ~, and also because of strain-hardening.
In uniform torsion, plastic collapse occurs when the uni-
form plastic torque Mup is reached at the supports
Mup "~ %~bt2/2 (9) (Figure 5). In warping torsion, plastic collapse occurs when
Prevention of warping
in which %, (=f)~/3) is the shear yield stress,f~, is the tensile ineffective in uniform
yield stress, and b and t are the width and thickness of each
rectangular element of the cross-section, respectively. The
uniform torsion plastic section modulus is defined by
B,, = fyb~th/4 (l l)
Uniform torsion
in which b: and t: are the flange width and thickness, plastic shear hinge
8 Bottom f l a n g c ~
- - . - End Wagner stress resultant W
7 End uniform torque Me
--4[
~ o First yield ~pfl~
7~ ' ~ 6 • Formation of cenmfl plastic warping hinge. ~ - "
~ • First strain hardening ~"
0 ~'¢ .G~'°1 I I I I I I I
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
End twist ~' (radlan/m)
(c) Uniform torque distribution
Figure 4 Effect of large twist rotations on Wagner stress result-
ant and uniform torque Figure 5 Uniform torsion plastic collapse
Torsion o f steel beams: N. S. Trahair and Y.-L. Pi 375
each flange becomes a collapse mechanism as a result of lapse analysis does not predict the twist rotations. Because
the formation of plastic bending hinges and the existence of this, serviceability design should be carried out using
of any frictionless bending hinges at points where warping elastic analysis.
is unrestrained (Figure 6). The warping torsion plastic col-
lapse load factor can be determined from a consideration
of the warping plastic torques 3. Strength design for torsion
in which the uniform torsion section capacity is taken as or secondary, depending on whether the torsion action is
required to transfer load (primary torsion), or whether it
M,, = M,/, = ~-,S,, (21) arises as a secondary action. Secondary torques may arise
as a result of twist rotations compatible with the joint
and the bimoment section capacity is taken as rotations of primary frames 5, and are predicted by three-
dimensional analysis programs. They are not unlike the sec-
B, = B~, = f, Sh (22) ondary bending moments which occur in rigid-jointed
trusses, but which are usually ignored (a procedure justified
by many years of satisfactory experience based on the long-
3.4. First yield design standing practice of analysing a rigid-jointed truss as if pin-
In first yield design, the strength of the member is assumed jointed). Secondary torques are usually small when there
to be reached when the most heavily strained cross-section are alternative load paths of high stiffness, and may often
first yields (ignoring any residual stresses and local stress be ignored.
concentrations at re-entrant corners). First yield design is Primary torsion actions may be classified as being
appropriate for members whose width-thickness ratios are restrained, free, or destabilizing. For restrained torsion, the
at the limit between slender and noncompact sections. First member applying the torsion action also applies a
yield design can be carded out by making linear elastic restraining action to the member resisting the torsion. In
analyses of the member under the strength load combi- this case, the structure is redundant, and compatibility
nation and determining the maximum values of the design between the members must be satisfied in the analysis if
uniform torque M*, and the design bimoment B*. The the magnitudes of the torques and other actions are to be
design is then satisfactory if equations (19) and (20) are determined correctly. Free torsion occurs when the member
satisfied with the section capacities now taken as applying the torsion action does not restrain the twisting of
the torsion member, but does prevent its lateral deflection.
M,, = M,, = r,Z,, (23) Destabilizing torsion may occur when the member applying
the torsion action does not restrain either the twisting or
B, = By =f,Z,, (24) the lateral deflection of the torsion member. In this case,
lateral buckling actions ~6 caused by the in-plane loading of
When significant uniform torques and bimoments occur at the torsion member amplify the torsion and out-of-plane
the same location, then the design will be satisfactory when bending actions.
The inelastic nonlinear bending and torsion of fully
braced, centrally braced, and unbraced l-beams with central
/14;; + B, -- (25) concentrated loads has been analysed 4, and interaction
equations developed for predicting their strengths. It was
For noncompact sections whose width-thickness ratios lie found that while circular interaction equations are appropri-
between the yield and plastic limits, linear interpolation ate for short length braced beams, these provide unsafe pre-
may be used between first yield design and first hinge dictions for beams subject to destabilizing torsion, where
design. lateral buckling effects became important. On the other
hand, destabilizing interactions between lateral buckling
3.5. Local buckling design and torsion tend to be masked by the favourable effects of
In local buckling design, the reduction in the cross-section the secondary axial stresses that develop at large rotations
strength below the yield strength by local buckling is due to the Wagner effect, and linear interaction equations
accounted for approximately. Local buckling design should based on plastic analyses provide satisfactory strength pre-
be used for members with slender sections. Local buckling dictions, as shown in Figure 7. Proposals based on these
design can be carried out by making linear elastic analyses findings are made below for the analysis and design of
of the member under the strength load combinations and members subject to combined torsion and bending.
determining the maximum values of the uniform torque
M*, and the bimoment B*. The design of an open section
member can be considered to be satisfactory when
1.4
Modlfi©d slenderness Accurate ~'L approximation
B* d~2 (26)
"~
~, 1.2
1.0
~'L = ~ ' ~ - ~ / M y x )
! 0.5
1.0
1.41
....
.......
--" --
which allows for postlocal buckling resistance. Usually, "- ,,, o
M*, and B* will not be maximum at the same location, o~ 0.8
and so only one term of the left-hand side of this equation
0.6
will be significant. ~.~...,~ ~-.. ~. o
4.2 Design First 'hinge' design is based on full plasticity of the most
Member cross-sections should be classified as slender, non- heavily stressed cross-section under either uniform torsion
compact or compact. Plastic or elastic cross-section analy- or bimoment.
ses may be used to determine the fully plastic section For plastic design, two analyses of the plastic collapse
capacities M,,, M,~,, B~, and M#, or the first yield section of the member are made. In the first, the torsion loads are
capacities M,,., M,> and B>. considered to be resisted by uniform torsion alone, and the
The inelastic nonlinear method of member analysis used plastic collapse toad factor is determined by using the sand
in Reference 4 is unsuitable for design. However, separate heap analogy. In the second, only warping torsion is con-
plastic collapse analyses may be used to determine the plas- sidered, and the plastic collapse mechanisms of the flanges
tic collapse load factors A0, for in-plane bending and A,~,for are analysed as in the plastic analysis of beam bending. The
torsion. Alternatively, linear elastic member analyses approximation for the fully plastic collapse load factor is
should be used to determine the in-plane distributions of determined by adding the independent collapse factors for
moment M*, and shear V*,., and the torsional distributions uniform and warping torsion. The use of plastic design for
of torque M* and bimoment B*. There is no need to compact members will greatly simplify both analysis and
amplify any of these actions or to decrease any of the load design, and will lead to more economic structures.
factors because of the finding4 that linear interaction equa- Torsion often occurs in conjunction with bending
tions are satisfactory, even for unbraced beams, as shown actions. Combined bending and torsion may be designed
in Figure 7. against using simple interaction equations.
Members analysed plastically should satisfy the circular
interaction equation Acknowledgment
This work has been supported by research grants made
I/A~, + I/A~, <
-- Oz (27)
available by the Australian Research Committee and the
Centre for Advanced Structural Engineering.
Members analysed elastically should satisfy the linear inter-
action equation
References
M*~ M* 1 Pi, Y. L. and Trahair, N. S. 'Inelastic torsion of steel l-section beams',
J. Struct. Engng, ASCE 1995, 121 (4), 609-620
+M,,, 2 Pi, Y. L. and Trahair, N. S. 'Plastic collapse analysis o1" torsion', ./.
Struct. Engng, ASCE 1995, 121 (10), 1389-1395
in which M*, and M*. are the maximum bending moment 3 Pi, Y. L. and Trahair, N. S. 'Steel member design for combined tor-
and torsion actions in the member, ~bMbx is the design sion and bending', Aust. Civ. Engng Trans. 1994, CE36 (4), 325-33(I
4 Pi, Y. L. and Trahair, N. S. 'Inelastic bending and torsion of steel l-
moment capacity which includes an appropriate allowance beams', J. Struct. Engng, ASCE 1994, 120 (12), 3397-3417
for beam lateral buckling, and &M, is the design torsion 5 Timoshenko, S. P. "Theory of bending, torsion, and buckling of thin-
capacity for the member when loaded in torsion alone. This walled members of open cross-section', Collected papers ~]'Stephen
design torsion capacity is equal to the value of M*. at which P Timoshenko McGraw-Hill, New York, 1953
6 Timoshenko, S. P. and Goodier, J. N. Theoo' (~f elasticity (3rd edn)
the appropriate design criterion of Section 3 is just satisfied.
McGraw-Hill, New York, 1970
7 Vlasov, V. Z. Thin-walled beams (2nd edn) Israel Program for Scien-
tific Translations, Jerusalem, 1961
5. Conclusions 8 Zbirohowski-Koscia, K. Thin-walled beams Crosby Lockwood, Lon-
don, 1967
Very few standards give rules for designing steel structures 9 Trahair, N. S. and Bradford, M. A. The behaviour and design of steel
against torsion, and those that do only supply limited structures (revised 2nd edn) Chapman & Hall, London, 1991
advice, usually concerning elastic design. This paper sum- 10 Baigent, A. H. and Hancock, G. J. 'Structural analysis of assemblages
marizes a series of investigations into the behaviour, analy- of thin-walled members', Engng Struct. 1982, 4 (3), 207-216
11 Billinghurst, A., Williams, J. R. L., Chen, G. and Trahair, N. S.
sis, and design of members subjected to combined torsion "Inelastic uniform torsion of steel members', Comput. Struct. 1992,
and bending. Steel torsion members may be designed for 42 (6), 887-894
local buckling, first yield, first 'hinge', or plastic collapse. 12 Farwell, Jr., C. R. and Galambos, T. V. 'Nonuniform torsion of steel
Local buckling, first yield and first 'hinge' design require beams in inelastic range', J. Struct. Div. ASCE 1969, 95 (STI2),
2813-2829
the use of an elastic analysis of the member for nonuniform 13 Home, M. R. Plastic theory ~['structures (2nd edn) Pergamon Press,
torsion in which the total torque at a cross-section has uni- Oxford, 1979
form and warping torque components. 14 Neal, B. G. The plastic methods ~[ structural analysis (3rd edn)
Local buckling design is based on the local buckling or Chapman & Hall, London, 1977
postbuckling strength of the most heavily stressed section. 15 Trahair, N. S. and Bridge, R. Q. 'Members subject to combined
actions', Commentary on the AS4100-I990. SAA Steel Structures
First yield design is based on local yielding in the most Code, Standards Australia, Sydney, 1990, Chap. 8
heavily stressed section under the combination of uniform 16 Trahair, N. S. Flexural-torsional buckling of structures E & FN Spon,
torsion shear stresses and warping torsion normal stresses. London, 1993