Sie sind auf Seite 1von 12

INELASTIC TORSION OF STEEL I-BEAMS

By Yong Lin PP and N. S. Trahair,2 Member, ASCE

ABSTRACT: A large twist rotation model for the nonlinear inelastic analysis of the non-
uniform torsion of I-section beams is presented in this paper. A finite-element procedure
has been developed based on the model. The nonlinear equilibrium equations of beams
in nonuniform torsion including the effect of the large twist rotation have been derived.
The elastic-plastic behavior of beams in nonuniform torsion is studied using the finite-
element procedure and the results are compared with tests. It is found that I-section
beams have much larger torsional capacities than can be predicted by linear plastic
collapse analysis, and that torsional failure occurs not by the formation of a mechanism
but by the tensile rupture of the flanges. A method is proposed for calculating the full
plastic nonuniform torque for practical design purposes.

INTRODUCTION
Although the torsional analysis of beams has attracted the interest of many researchers since
Saint-Venant (1855) solved the problem of elastic uniform torsion, it appears difficult to predict
the ultimate strengths of steel I-beams in nonuniform torsion. Much research has been carried
out into elastic uniform torsion and expressions have been obtained for the warping displace-
ments of the cross section. Some analytical solutions for the uniform elastic-plastic torsion of
beams with various cross sections have also been obtained.
Nadai (1931) developed the sand-heap analogy for the full plastic torque of solid sections,
Christopherson (1940) obtained an elastic-plastic solution for an I-section, and, later, Nadai
(1954) used the rooftop membrane analogy for the elastic-plastic solution of various cross sec-
tions. Sokolovsky (1946) developed an elastic-plastic solution for an oval section, and Smith
and Sidebottom (1965) derived an elastic-plastic solution for prismatic bars of rectangular sec-
tions. Billinghurst et al. (1991) used the mitre method to obtain elastic-plastic solutions for
various cross sections.
The linear elastic nonuniform torsion of thin-walled open sections was studied by Timoshenko
and Gere (1961), Vlasov (1961), and Wagner (1936) by considering small angles of cross-section
rotation <j> (so that sin <j> = <j>, cos <j> = 1). Because of the complexity of the elastic-plastic analysis
of the nonuniform torsion of beams, no analytical solutions have been obtained so far. Ap-
proximate solutions for the small-rotation elastic-plastic nonuniform torsion of I-section canti-
levers were suggested by Boulton (1962), Dinno and Merchant (1965), and Augusti (1966).
Boulton (1962) presented a lower-bound theory, while Dinno and Merchant (1965) developed
a so-called Merchant upper bound for the full plastic torque-carrying capacities of I-section
cantilevers. Augusti (1966) presented an additional theory that gave results between the upper
and lower bounds and proved that the Merchant upper bound was a true upper bound.
Since thin-walled open sections have low torsional stiffness, the torsional deformations can
be of such magnitudes that it is not adequate to treat the angles of rotation as small. When
finite twist rotation angles are considered, the elastic uniform torsion problem becomes nonlin-
ear. Ashwell (1951) and Gregory (1960) studied both theoretically and experimentally the elastic
nonlinear behavior of twisted cantilevers of different cross sections under uniform torsion con-
ditions. Tso and Ghobarah (1971) presented a study of the nonlinear nonuniform elastic torsion
of thin-walled open sections.
Recently, numerical methods have been used by some researchers to investigate the elastic-
plastic uniform and nonuniform torsional behavior of beams. Yamada et al. (1972), Johnson
(1973), and Itani (1979) studied elastic-plastic uniform torsion. Baba and Kajita (1982) used a
two-node, four-degree-of-freedom beam element for the uniform torsion analysis and a four-
node, 12-degree-of-freedom rectangular section element for the warping analysis of the section.
Their linear and nonlinear predictions of the deformations were quite similar and they concluded
that the solutions for the plastic torque of beams should lie in-between the sand heap analogy
and Merchant upper bound. Bathe and Chaudhary (1982) used warping displacement functions
for beams of rectangular cross section in the formulation of a two-node Hermitian-based beam
and in the formulation of a variable number of nodes isoparametric beam for the linear and

'Post-Doctoral Fellow, School of Civ. and Min. Engrg., Univ. of Sydney, NSW. 2006. Australia.
'Challis Prof. of Civ. Engrg.. School of Civ. and Min. Engrg.. Univ. of Sydney, NSW. 2006. Australia.
Note. Associate Editor: Amde M. Amde. Discussion open until September 1. 1995. To extend the closing
date one month, a written request must be filed with the ASCE Manager of Journals. The manuscript for this
paper was submitted for review and possible publication on September 30. 1993. This paper is part of the Journal
of Structural Engineering. Vol. 121, No.4. April. 1995. ASCE. ISSN 0733-9445/95/0004-0609-0620/$2.00 +
$.25 per page. Paper No. 7073.

JOURNAL OF STRUCTURAL ENGINEERING 609

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
nonlinear analysis of torsion. Bathe and Wiener (1983) employed a Hermitian beam element
and a nine-node shell element for the elastic-plastic nonuniform torsion of I-beams. Gellin et
al. (1983) presented a strip finite-element model for the analysis of the nonlinear material
behavior of thin-walled members in nonuniform torsion. May and AI-Shaarbaf (1989) used a
standard three-dimensional 20-node isoparametric quadratic brick element in the elastic-plastic
analysis of uniform and nonuniform torsion of members subjected to pure and warping torsion.
Their results supported Baba and Kajita's conclusions. Chen and Trahair (1991) developed a
finite-element model for the inelastic analysis of nonuniform torsion of I-section beams by using
the mitre model to describe the shear strain distribution over the cross section.
Some experimental studies for the elastic-plastic uniform torsion of I-sections were carried
out by Boulton (1962), Dinno and Gill (1964), and Farwell and Galambos (1969). Boulton's
experiments included two tests with ends free to warp. The experimental torques were much
higher than those predicted by the Merchant upper bound and very large rotation angles were
sustained before failure occurred.
The purposes of this paper are to present an elastic-plastic model for analyzing large torsional
deformations of I-section beams, and to investigate the elastic-plastic behavior of I-section beams
in nonuniform torsion.

RELATIONSHIPS BETWEEN STRAINS AND DISPLACEMENTS


The origin of the member axes ox, oy, oz is defined at the centroid of the cross section. The
ox- and oy-axes coincide with the principal axes of the cross section. The axis oz coincides with
the centroidal axis of the beam.
The longitudinal normal strain 1' at point P(x, y, z) for nonuniform torsion can be obtained
by simplifying the general expression of longitudinal strain given by Pi and Trahair (1994) as
1' = w' - w<l>" + (1I2)(x~ + Y~)<I>'~ (1)

which consists of the axial shortening strain w', the linear warping strain - w<l>", and the nonlinear
Wagner (1936) strain (l/2)(x 2 + y2)<I>'2. The variations of the warping and Wagner strains around
an I-section are shown in Fig. 1.
The shear strains due to warping are very small compared with the shear strains caused by
uniform torsion. The mitre model developed by Billinghurst et al. (1991) for the shear strain
"Y I' due to uniform torsion is

'YI' = 11<1>' (2)


where n normalized uniform torsion shear strain function. For an I-section member
11 = 2f3[IYI - (1I2)(D - t,)]; 11 = 2f3[lxl - (1I2)(B - t,)]; n = 2f3lxl (3i1-c)

for the long and short sides of the flanges, and the web, respectively. In these equations, I I
indicates the absolute values; D = depth of the cross section; It = thickness of the flanges;
B = flange width; and f3 an elastic stiffness correction factor given by (Billinghurst et al.
1991)
J
f3=---...,----...,---...".----,---:-::---::-:-...,------:-::...,---...,-------;------,:-:=-:--:-:- (4)
2Bt}/3 + (D - 2t/)t~/3 + (D - 2t,)(t/ - t,,)I~./4 - (t; - t~./2)t,/6

where I", = thickness of the web. The variations around the cross section of trajectories of these
strains are also shown in Fig. I.
Combining (1) and (2) for the longitudinal strain and shear strain yields

fur 100029
o Oy=250MPa E= 2OO.000MPa
I B
1 y=0.00125 G = 80.000MPa
. =0.01375 E.= 6000MPa

E.

, E
y
D=0.1524m l';=O.ooSOm fur Falwell and Galambos', ~ts
B=O.1509m h=O.I402m Oy=285MPa E = 213,400MPa
y y Ir=O.OI22m y =0.00134 G = SO,OOOMPa
y
, =0.01474 E.= 6000MPa

<a) Warpillll Strains (b) Waper Strains (e) Uniform ThrsloD Shear Strains (a) (b)
-m.' +<x'+f>.-' n. FIG. 2. Cross Section for Farwell and Galambos's Tests and
FIG. 1. Strain Variations around Cross Section Stress-Strain Curve

610 JOURNAL OF STRUCTURAL ENGINEERING

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
(5)
in which

[S] = [~ -ow (6a,b)

RELATIONSHIP BETWEEN STRESSES AND STRAINS


The material is initially elastic. When a certain level of stress or strain is reached, then yielding
starts. The von Mises yield criterion is used to describe the onset of yielding. After the onset
of yielding, the material plastic deformations may be described by a flow rule and a hardening
rule. The incremental relationships between the stress and strain can be given by (Pi and Trahair
1994)
(7)
where {da} and {de} = incremental stresses and strains; and the elastic-plastic stress-strain matrix
[E]ep is given by

(8a,b)

in which a = H'a; + alB + 9'1"~G, where H' = strain hardening parameter; E = Young's
modulus of elasticity; G = shear modulus of elasticity; a p = longitudinal stress at point P;
'1"1' = shear stress at point P; and the effective stress a e is given by

(9)

NONLINEAR EQUILIBRIUM
The nonlinear equilibrium equations of nonuniform torsion for a thin-walled member can be
derived from the principle of virtual work

Iv {d~Y{a} dV - r m d<l> dz - BM d<l> = 0 (10)

where {a} = stresses; and m and M = distributed and concentrated external torques, respec-
tively.
For the elastic case, (10) can be written as

r {d'PF[DH'P} dz - r m d<l> dz - B M d<l> = 0 ( 11)

where
{d'P} = {dw', d<l>", <1>' d<l>', d<l>T (12)
-ES wEA EI"
f
and
[D] =
A
[S)T[E]e[S] dA = -ES
[ EI"
w El w
-EI"w
o o
-EI"w
EI""
o 1] (13)

A = L dA; Sw = L w dA; I" = L (X 2 + y2) dA; Iw = L W2 dA (14a-d)

Ipw = L W(X 2 + y2) dA; 1,,1' = L (X 2 + y2)2 dA; J = L fV dA (l5a-c)

Integrating (11) by parts and considering Sw = 0 and IpM' = 0 for doubly symmetric I-sections
lead to the differential equilibrium equations
[EAw' + (1I2)EI,,<I>'2]' = 0 (16)
[EIA'']'' - [EI"w' + (1I2)EII'I,<I>'2],<I>' - [GJ<I>,], = m (17)
for the large twist rotation model of nonuniform torsion.
At the same time, the natural boundary conditions at z = 0 and z L can be obtained as
EA w' + (I/2)E(,<I>'2 = 0 (18)
- [EIH.<l>'']' + [Elpw' + (1I2) Elpp <l>'2]<I> , + GJ<I>' = M; Elw<l>" = 0 (19, 20)

JOURNAL OF STRUCTURAL ENGINEERING 611

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
For the small twist rotation model, the nonlinear component (1I2)(x 2 + y2)<\>/2 of longitudinal
strain Ep is ignored and the differential equilibrium equations of nonuniform torsion then reduce
to
[EAwT = 0; [El w <\>")" - [Gl<\>T = m (21, 22)
and the corresponding boundary conditions to
EAw' = 0; Gl<\>' - [Elw<\>'T = M; EI...<\>" = 0 (23-25)
From (21), w/ is constant along the beam and from (23) the constant is equal to zero, i.e.
the axial shortening w/ is equal to zero for the small twist rotation model. Accordingly, (21)
and (23) can be omitted for the small twist rotation model. The remaining (22)-(25) are con-
sistent with Timoshenko (1961) and Vlasov (1961). In these equations, Gl<\>' = Mil is referred
to as the uniform torque; - [Elw<\>'T = M w is referred to as the warping torque; and - El w <\>"
= B is referred to as the bimoment.
For convenience in the finite-element formulation, the nonlinear equilibrium equation can
also be expressed in terms of equivalent nodal forces as

r [N],~[B]T{R} r dz - m{N}; dz - ~ M{N}b = {O} (26)

in which the element external forces are - f if m{N}: dz - ~1.2 M{N} J; the element internal
forces are fir [N]:r B]7{R} dz, where

[B]
[~ ~, ~] (27)

[N)., = a matrix; {N}q and {N}Q = vectors whose elements are the shape functions; and {R} =
stress resultant vector defined by

{R} = L [SV{u} dA (28)

where the stress vector {a} is given by

{u} = {UP}
TI'
(29)

Applying the principle of virtual work to both the incremental equilibrium state and the
previous equilibrium state will lead to the incremental equilibrium equations
(30)

where the increment of the equivalent element external forces {~p} is given by

{~p} = J.L ~m{N}~dz


(I
+ ~Ml{N}bl + ~M2{N}b2 (31)

and the element tangent stiffness matrix [k]r is given by

[k]r = J.L {[Nl.~([BF[D][B]


(I
+ [M).,)[N).,} dz (32)

The matrix [M)., is given by

[M)., = [~ *~]
where the stress resultant W corresponding to the Wagner effect is the third component of the
(33)

stress resultant vector {R} in (28) given by

W = f (X" + y2)UI' dA (34)

In the case of elastic analysis, W = (1/2)11'1'<\>'2.


The tangent modulus matrix [D] is defined by the integration

[D] = f [SV[E]el'[S] dA (35)

612 JOURNAL OF STRUCTURAL ENGINEERING

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
A modified Newton-Raphson method is used for the incremental-iterative solution required
for the nonlinear analysis and the incremental-iterative equilibrium equations
(36)
where {~pt = incremental external forces; and {~P}r = unbalanced forces in the last iteration
given by

{~pt = J m{N}~ dz + ~ M{N}~ - r [Nl.~[BnR} dz (37)

ELASTIC-PLASTIC BEHAVIOR IN NONUNIFORM TORSION


Comparison with Test Results
The finite-element procedure developed in this paper has been used to predict the nonlinear
inelastic behavior of two simply supported I-beams tested by Farwell and Galambos (1969). The
dimensions of the beams and their material properties are shown in Fig. 2. The beams are
assumed to be initially straight, and to have zero residual stresses. One beam is subjected to a
central concentrated torque and the other beam is subjected to two equal concentrated torques
symmetrically placed at O.3L from each end. Analyses using the large twist rotation inelastic
model with end axial restraints at the centroid (case A) are performed for each beam. The
elastic-plastic relationships between the external torque T at the midpoint and the central twist
rotation <I> for the first beam are shown in Fig. 3, and the elastic-plastic relationships between
the external torques Tl2 and the twist rotation <l>O.3L at O.3L are shown in Fig. 4. It can be seen
that at small torques, the behavior is elastic and almost linear for all the analyses. At higher
torques, the twist rotations increase nonlinearly due to yielding, and at even larger torques, the
nonlinear behavior stiffens slightly as a result of the large rotations, and no maximum torque
is reached.
Analyses using the large twist rotation inelastic model without end axial restraints at the
centroid (case B) are also performed for the first beam. The elastic-plastic relationships between
the external torque T at the midpoint and the central twist rotation <I> are also shown in Fig. 3.
It can be seen from Fig. 3 that the effect of axial restraints at the centroid for the large twist
rotation analysis is to decrease the twist rotations significantly at large twist rotations because
significant axial stresses are induced that increase the Wagner effect.
The finite-element predictions using the large twist rotation inelastic model with end axial
restraints at the centroid agree closely with the test results of Farwell and Galambos (1969),
and the predictions using the large twist rotation inelastic model without end axial restraints at
the centroid disagree with the test results in the large twist rotation range. This suggests that
axial restraints may exist in the test. If there had been no axial restraints, then axial shortening
should have been detected in the test as the applied torque increased. However, no axial
shortening strains were detected at the centroid in the tests until the external torque exceeded
13.56 kNm (Farwell and Galambos 1969).

Effects of Large Deformations and Strain Hardening


The analyses for the first beam by a finite-element procedure based on the small twist rotation
inelastic model are also performed for cases A and B. The elastic-plastic relationships between

25,-----....----,.---.,....----....----,.----.".,

a.';X (
20 e-. X (
mr.-I!;."
_._-
15
-'- -- - _.- _.- _._._.- -'- - --
10

. . ~~ .. SJIII11lwisuowlon iuelasW: IlIOdclfor.bolh.cues A IDd B


- - 1..IrIe twist IOWloa ille1ulic rqodel for case A _ Large twill rowIoo iIIe1ulic IlIOdcI
- Large twist roWioD iIIe1ulic IlIOdcI for case B
TeSt IeI1I1II by Firwe111Dd Ga1ambos (1969} Teil results by F~eU IDd Ooiambos(l969)
0.2 0.4 0.6 0.8 1.2 0.2 0.4 0.6 0.8 1.2
CeDlra1 Twist RowioII (rldim) Twisl RotalioD ~ II C (radian)
FIG. 3. Comparison of Analysis and Test Results FIG. 4. Comparison of Analysis and Test Results

JOURNAL OF STRUCTURAL ENGINEERING 613

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
::' 5
z
e _._. End Wagner stress resullaDt .
4.5 _ End Wliform torque
~ Full plastic Wliform torque 7 . 0 Flrsr yield' ...

"~~ 4 -,~,":"':'
-~- Formation of eeottal plastic warping binge
First strain hardening

iii 35
] x , x
c 3
(j
5 I: "'~, =-.IL I
e
z
~
!I
E"
{:.
~
'S
:>
~
00
0.2 0.4 0.6 0.8 1.2 1.4 2
Centtal Twist RotatiOIl (radian) End Twist.' (radiao/m)

FIG. 5. Effect of Large Twist Rotations on Blmoment and Uni- FIG. 6. Effect of Large Twist Rotations on Wagner Stress Re-
form Torque sultant and Uniform Torque

25
T
~ .. :z: ( X
T
20 . e-a X ( x
o.s~.. ,t:.sL
e
z free fO warp T warping preveoted
~
f-
IS ~*:"'J+-_....:"1::f (a) Thrsion Member

~ I,l!, I. La .1
{:.
";l
e 10
!l
Ul (b) Warping Thrsion Collapse Mecbanism

.. ResullS. for case A wilbnut sa:aln bardeoiog.


- - Re.sults for case A wilh sa:aln bardeoiog
v."1--~
a .
v.-~
"2- L-a
.._._. Results for case B without sa:aln bardeoiog
- - Results for case 11 wilh sttaiD hardening
0.2 0.4 0.6 0.8 1.2 (c) 1bp Flange Moment Distribution
Centtal Twist Rotation. (radian)

FIG. 7. Effects of Strain Hardening FIG. 8. Warping Torsion Plastic Collapse

the external torque T at the midpoint and the central twist rotation <I> are also shown in Fig. 3.
The small twist rotation inelastic model ignores the nonlinear strain term (1/2)<1>'2 and any Wagner
stress resultants W. The predictions using the small twist rotation inelastic model are in agreement
with test results only in the small twist rotation range, but differ significantly from the test results
in the large twist rotation range. This indicates that the Wagner effect and the effect of large
twist rotations are significant and that the small twist rotation inelastic model is not realistic.
The relationships between the end uniform torque M" and the central twist rotation <1>, and
the relationships between the central bimoment B and the central twist rotation <I> are shown
in Fig. 5 for both the small and the large twist rotation inelastic models. For both models, first
yield occurs at midspan, and subsequently the bimoment becomes fully plastic at midspan where
a warping hinge is formed. The bimoment continues to increase after strain hardening starts.
For the small twist rotation inelastic model, the end uniform torque continues to increase after
the formation of the warping hinge and approaches the full plastic uniform torque. Finally, two
uniform torsion hinges form at the ends and the beam becomes a mechanism and collapses.
However, for the large twist rotation inelastic model, which includes the Wagner effect, the
end uniform torque increases only slightly after the formation of the warping hinge but does
not reach the full plastic uniform torque. Hence, plastic uniform torsion hinges do not form at
the ends and the beam does not collapse by forming a mechanism.
To illustrate the Wagner effect and the effect of large twist rotations further, the relationships
between the end uniform torque and the end twist <1>' , and the relationships between the Wagner
stress resultant and the end twist <1>' are shown for the large twist rotation inelastic model in
Fig. 6. The Wagner stress resultant is small at first, but increases after the warping hinge is
formed at midspan. As the end twist increases, the Wagner stress resultant continues to increase
while the end uniform torque does not but decreases after reaching a maximum value, which
is lower than the full plastic uniform torque. Finally, the Wagner stress resultant becomes very

614 JOURNAL OF STRUCTURAL ENGINEERING

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
large and the tension stresses induced at the flange tips may lead to tension rupture as observed
by Farwell and Galambos (1969).
The effect of strain hardening is investigated for both case A and case B by the large twist
rotation inelastic model for the first beam as shown in Fig. 3. The results without strain hardening
are compared with the results with strain hardening in Fig. 7. It can be seen that the effect of
strain hardening is small even in the large twist rotation range compared with the effect of large
twist rotations and the Wagner effect.

Plastic Collapse
The results of many experiments (Boulton 1962; Dinno and Gill 1964; Dinno and Merchant
1965; Farwell and Galambos 1969) and the predictions of large twist rotation inelastic model
(Figs. 3 and 4) indicate that torsional failure does not occur until substantial rotations have
taken place and that torsional failure is not suitable to be used as practical limit for the design
of nonuniform torsion of beams. Because of this, it is more appropriate to base a "strength"
model of torsion on the commencement of large rotations. This is similar to the use of a general
yield load <T y A 8 for the strength of tension members as an alternative to the local fracture load
<T,.An (Trahair and Bradford 1991).
Such a large rotation strength model for torsion may be developed as a parallel to the rigid-
plastic collapse theory (Neal 1977; Horne 1979) used for flexural structures, according to which
a structure collapses when there are a sufficient number of hinges (frictionless and plastic) to
transform the structure into a mechanism. For this flexural rigid-plastic theory, the material is
modeled as being rigid-plastic so that the deflections are small, and strain-hardening is ignored.
This flexural rigid-plastic theory can easily be transformed for members that resist torsion
actions solely through the warping torque M w , such as very thin-walled I-sections for which the
uniform torque M" is negligible. In this case, plastic collapse occurs when there are a sufficient
number of warping hinges (frictionless and plastic) in the flanges to transform them into mech-
anisms, as indicated in Fig. 8(b). The warping torsion plastic collapse load factor Awl' can then
be determined from a consideration of the warping plastic torques
(38)
in which h distance between the flange centroids; and Vfl' plastic collapse flange shears
dMj
Vji' = d; (39)

in which dMr/dz = gradients of the flange moment Mf at plastic collapse. For the torsion member
of Fig. 8(a), the flange gradients are Mfl'/a and 2Mfi,/(L - a), in which

(40)

is the full plastic moment of the flange. Thus

A "'"
= Mfi,h
T
(~ +
a
_2_)
L-a
(41)

The warping torsion plastic collapse load factors Awl' for some beams with commonly used
support and applied torque conditions are shown in Fig. 9.
A somewhat different parallel of the flexural rigid-plastic theory is required for members that
resist torsion actions solely through the uniform torque M", such as narrow rectangular cross
sections, for which the warping torque M w is negligible. In this case, plastic collapse occurs
when the fully plastic uniform torque M"I' is reached at all supports, as indicated in Fig. lOeb).
The full plastic torque M"p for the I-section is derived from Nadai's (1931) sand heap analogy
as (Billinghurst et al. 1992)

M"I' = Ty [Bt J (1 - ;~) + (D - 2t,) ~ + t] (42)

For the torsion member of Fig. lO(a), the uniform torsion plastic collapse load factor is

2M""
A""=r (43)

More generally, however, both uniform and warping torsion are important, and the total
(nonuniform) torque M z at any cross section is the sum of the uniform and warping torques, so
that

JOURNAL OF STRUCTURAL ENGINEERING 615

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
Beam Awp Beam Awp

,3 I ~ 0
T
'"' I
~Mt~h
L x
T
1 x: 'X
T
'"'
0 x: 1. Mtph
~

'.U2.I.U2'pl
! Li2 ' LI2
T
.LI2 , ~

LI2
I ,.U2 1 U2 i
L T

X
T
'"'
0 I X 0
T
.....
I
. Mtph
L T X"'"
v
T
X
.1
xJ
T
X
4.5 ;\t/ tph
L T
:.. LI2.I. U2 ,i ILI2.'.U2' I~I 2UJ Wl' 2lJ3 ,I

TiL ~Mtph 1. Mtph


)() . . . '"' '"'s :K)JJ':S, I
T

,. """ .' L I. L
I'
L T
L
~,
,
L T

X'"'
T/2

"
IMW~:141
TR
oX XJ ':X
1141l.QI
TI2

LIl'
. Mtph
L T X
L
;J
T
..... 0
I 1
I
FIG. 9. Warping Torsion Plastic Load Factors

T
c: wbco
VCbtiOn of warping ineffective

::2E~:::J-t---'-i1
10101. La .1
M""O

<a) 1brsion Member

(b) Unlfonn 1brsion Collapse Mechanism

D =0.2613 m =
b 0.2490 m q.= 0.50, q;= 0.3S<J,
B = O.I~IO m t,= 0.0123 m
t,,=0.0077 m
<c) Unlform'lbrque Distribution
<a) Cross Section (b) Residual Stresses
FIG. 11. Cross-Section and Residual Stresses
FIG. 10. Uniform Torsion Plastic Collapse of 10UB29

It is not easy to develop a rigorous model for plastic collapse under nonuniform torsion, because
(1) different stresses (shear stress T and normal stress (J) are associated with uniform and warping
torsion collapse; (2) the different stresses T and (J are distributed differently across a section;
and (3) the uniform torque Mil and warping torque M w are distributed differently along the
member length.
As a result, a number of approximate theories have been proposed. These include Boulton's
(1962) lower-bound theory for cantilevers and Merchant (Dinno and Merchant 1965) upper
bound; and Augusti (1966) developed an analysis that predicted results falling between these
lower and upper bounds.
The simplest theory is the Merchant upper-bound approximation (Dinno and Merchant 1965)
for cantilevers. This can be generalized to a wide range of loading and support situations by
assuming that
(45)

approximates the nonuniform torsion plastic collapse load factor A.p This is equivalent to as-
suming that the separate uniform and warping torsion collapse mechanisms are completely
independent. For this to be so, there must either be no interaction between the shear and normal
stresses T and (J at full plasticity, which violates the von Mises yield criterion

616 JOURNAL OF STRUCTURAL ENGINEERING

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
(46)

or else there must be no spatial coincidence of nonzero values of both (J' and T at any critical
cross section, which is most unlikely. For these reasons, the Merchant approximation and its
extension (45) provide an upper bound to Ap , at least within the small rotation rigid-plastic
assumptions of the plastic collapse theories.
In view of the fact that the attainment of the rigid-plastic collapse load does not lead to failure
because of large rotation and strain-hardening effects, but rather signals the onset of large
rotations (Figs. 3 and 4), there does not appear to be any special merit in seeking a more accurate
approximation than that of (45). In view of its simplicity and ease of calculation, it is therefore
proposed that the extension provided by (45) of the Merchant approximation be used as the
"strength" referetIce capacity of a member in nonuniform torsion.
To investigate the torsional strength further, an initially straight I-section lOUB29 beam with
L = 4.39 m has been used for the analyses by both the small and the large twist rotation inelastic
models for both cases A and B of beams and cantilevers with various support and applied torque
conditions. The dimensions and residual stress distributions of the cross section used in the
analyses are shown in Fig. 11 and the stress-strain curve is shown in Fig. 2. Fig. 12 shows the
variations of the central twist rotation <t> for a beam with fixed ends (<t> = <t>' = 0) subjected
to a central concentrated torque T. Fig. 13 shows the variations of the central twist rotation <t>
for a beam with one end fixed and the other end torsionally simply supported (<t> = <t>" = 0)
subjected to a central concentrated torque T. Fig. 14 shows the variations of the central twist
rotation <t> for a beam with both ends torsionally simply supported subjected to a central con-
centrated torque T. Fig. 15 shows the variations of the central twist rotation <t> for a beam with
both ends torsionally simply supported subjected to uniformly distributed torque TIL. Fig. 16
shows the variations of the twist rotation <t> at C for a beam with both ends torsionally simply

2S
2S e-.~
,-.~ e-.~
20
,-.~ ~
20
~

E
E - -_.- -- - - _._- ;z:
d
l~

....~
l~
.."
....
I:
~ 10

I 10

- Larje twisl iowioo iDelastic model for case A


-
-
LarKe rwisl roratiOll inelastic model for case A
Large rwisl roralioo inelastic model for case B:
- Larje twisl rOwioo iDelastic model for case B: .. ;:.:... smlli iWiSi rowro.i ibelUiiC model fOrbOi1icUes A iDd II
... ,_smillIWislDlwioo melastic illlldeHor bod..,..... A uid B o FlISi yield . . . . .
o Fin! yield FustbiDge :
. Fin! biDBe
~ U U U U M ~ ~ M
0.3 0.4 o.~ 0.6 0.7 0.8 0.9 Cemral Twisl RoIaIioo (radian)
CeDlrll Twisl RoIaIioo (radian)
FIG. 13. Beam with One End Free to Warp and Other End Pre-
FIG. 12. Beam with Both Ends Fixed vented from Warping

2S
e-A~
c..~ e-a~
20 c..~ 20 ~
~ M.E~,,!!I ~~r ~'!!Jd_. _
IS
E
;z: l~

....d
II 10
I:
~ 10

5
~

0.7 0.8 0.9


0.5 0.6
CeDlrll Twisl RoIaIioo (radian)
CeDIrII Twist RoIaIioo (radian)
FIG. 15. Beam Subject to Uniformly Distributed Torque with
FIG. 14. Beam with Both Ends Free to Warp Both Ends Free to Warp

JOURNAL OF STRUCTURAL ENGINEERING 617

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
25.---...--...--.,..--.,..--.,..--...--...--,..--.---...
7r---.,--...--...,....--r---.--....,...-.....,...-~----.-----.

20 6

l~ --- ------
--_._---- 4

.
/.",..;.~; -.:-.-
,.'
-
. - Large tw1st rou.tiOa inelastic model for <:lIR A
- Large tw1st rotatiOa inelastic model for case B
;..;..'SDiill iwlii riiWio.. lrielaStiC modeHcir both cASes AiDa II
o FUll yieid . ..
2
_:-Lara_ twistl'Otation inelastic model
:...._Smi1l twist iotation inelastic model
FUIlbinS_ . 0 F.nt yield:
:. Fu1t binge :
0.3 0.4 O.~ 0.6 0.7 0.8 0.9
Twist Rotation. at C (radian) 0.4 0.6 0.8 1.2 1.4 2

Free End Twist Rotation. (radian)


FIG. 16. Beam SUbject to Torque at Third Point with Both Ends
Free to Warp FIG. 17. Cantilever with Warping Restraint

supported subjected to a concentrated torque T at a third point of the beam. Fig. 17 shows the
variations of the end twist rotation (j> for the cantilever with a fixed end subjected to an end
concentrated torque T. Also shown in Figs. 12-17 are the proposed plastic collapse torques Tp
It can be observed from Figs. 12-17 that for the large twist rotation inelastic model, as the
external torques increase, first yield occurs, and then the first warping hinge (or hinges) is
formed. After the formation of the first hinge, the twist rotation increases faster than before
(and probably until tension rupture occurs in the flange). There are no collapse mechanisms
formed. For the small twist rotation inelastic model, first yield occurs, and then the first hinge
(or hinges) is formed. However, after the formation of the first hinge, the twist rotation increases
much faster than before, until uniform torsion hinges are formed and a collapse mechanism is
formed. For the small twist rotation model, when the uniform torsion hinges are formed, the
corresponding twist rotations are very large and are not shown in the Figs. 12-17. It can also
be observed that the proposed extensions of Merchant's approximation Tp are generally a little
higher than the asymptotes of the small twist rotation inelastic model, but not always higher
than the torques at the commencement of large rotations of the large twist rotation inelastic
model, as defined by the knee of the torque-twist rotation curve.
It can be seen that the large twist rotation model can predict the elastic-plastic nonuniform
torsion behavior of the beams and cantilevers, and the small twist rotation model cannot predict
the elastic-plastic behavior. The large twist rotation model provides an upper-bound solution
for the nonuniform torsion collapse of beams and cantilevers while the small twist rotation model
provides a lower-bound solution. The finite-element results indicate that the proposed extension
of Merchant's approximation provides a reasonable reference for the strength capacity of a
member in nonuniform torsion and confirm the suggestion that the extension may be used to
approximate the torsional strengths of steel I-beams in nonuniform torsion.

CONCLUSIONS
A finite-element procedure based a large twist rotation inelastic model has been developed
for nonuniform torsional analysis. The model includes higher-order terms in the nonlinear strains,
a mitre model of the shear strain distribution, and Wagner stress resultant. Comparison of the
finite-element results with the available test results indicates that the model can accurately predict
the nonlinear elastic-plastic behavior of steel I-beams subjected to nonuniform torsion. The
finite-element model is also used to investigate the elastic-plastic behavior of steel I-beams in
nonuniform torsion. It is found that the torsional failure of beams occurs at very large twist
rotations by fracture of the flanges rather than by the formation of a plastic collapse mechanism
because of high tension stresses induced by the Wagner effect and the effect of large twist
rotations. Strain hardening has a small effect on the nonlinear inelastic behavior of beams.
The nonuniform torsional strength of steel I-beams can be based on the commencement of
large twist rotations. An extension of the Merchant upper-bound approximation is developed
using rigid-plastic theory, and formulas for the plastic collapse of several commonly used beams
and cantilevers are obtained. The finite-element results from both the small and the large twist
rotation models indicate that this extension of the Merchant approximation provides a reasonable
reference for the strength capacity of a member in nonuniform torsion, and confirm the sug-
gestion that the extension may be used to approximate the torsional strengths of steel I-beams
in nonuniform torsion.

618 JOURNAL OF STRUCTURAL ENGINEERING

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
ACKNOWLEDGMENTS
This work has been supported by a postdoctoral fellowship made available by the Australian Research Com-
mittee through the Australian Research Fellowship Scheme and by the Centre for Advanced Structural Engineering
of the University of Sydney to Yong Lin Pi.

APPENDIX I. REFERENCES
Ashwell, D. G. (1951). "The axis of distortion of a twist prism." Phil. Mag., Vol. 42, 820-832.
Augusti, G. (1966). "Full plastic torque of I-beams." Int. J. Mech. Sci., Vol. 8, 641-649.
Baba, S., and Kajita, T. (1982). "Plastic analysis of torsion of a prismatic beam." Int. J. Numerical Methods in
Engrg., 18(6),927-944.
Bathe, K.-J., and Chaudhary, A. (1982). "On the displacement formulation of torsion of shafts with rectangular
cross-sections." Int. 1. Numerical Methods in Engrg., 18(10), 1565-1580.
Bathe, K.-J., and Wiener, P. M. (1983). "On elastic-plastic analysis of I-beams in bending and torsion." Computers
and Struct., 17(5-6), 711-718.
Billinghurst, A., Williams, J. R. L., Chen, G., and Trahair, N. S. (1992). "Inelastic uniform torsion of steel
members." Computers and Struct., 42(6), 887-894.
Boulton, N. S. (1962). "Plastic twisting and bending of an I-section in which the warp is restricted." Int. 1. Mech.
Sci., Vol. 4,491-502.
Chen, G., and Trahair, N. S. (1992). "Inelastic nonuniform torsion of steel I-beams." 1. Constr. Steel Res., Vol.
23, 189-207.
Christopherson, D. G. (1940). "A theoretical investigation of plastic torsion in an I-beam." J. Appl. Mech., Vol.
7,1-4.
Dinno, K. S., and Gill, S. S. (1964). "The plastic torsion of I-sections with warping restraint." Int. 1. Mech. Sci.,
Vol. 6, 27-43.
Dinno, K. S., and Merchant, W. (1965). "A procedure for calculating the plastic collapse of I-section under
bending and torsion." The Struct. Engr., 43(7), 219-221.
Farwell, C. R. Jr., and Galambos, T. V. (1969). "Nonuniform torsion of steel beams in inelastic range." 1. Struct.
Engrg., ASCE, 95(12), 2813-2829.
Gellin, S., Lee, G. c., and Chern, J. H. (1983). "A finite element model for thin-walled members." Int. J.
Numerical Methods in Engrg., 19(1),59-71.
Gregory, M. (1960). "The bending and shortening effect of pure torque." Australian J. Appl. Sci., Australia,
Vol. 11, 209-216.
Horne, M. R. (1979). Plastic theory of structures, 2nd Ed., Pergamon Press, Oxford, England.
Itani, R. Y. (1979). "Elasto-plastic torsion of axisymmetric bars." 1. Engrg. Mech. Div., ASCE, 105( I), 1-12.
Johnson, A. F. (1973). "Bending and torsion of anisotropic bars." Int. J. Solds. Struct., Vol. 9,527-551.
May, I. M., and AI-Shaarbaf, A. S. (1989). "Elasto-plastic analysis of torsion using three-dimensional finite
element model." Computers and Struct., 33(3), 667-678.
Nadai, A. (1931). Plasticity. McGraw-Hili, New York, N.Y.
Nadai, A. (1954). Theory of flow and fracture of solids. McGraw-Hili, New York, N.Y.
Neal, B. G. (1977). The plastic methods of structural analysis, 3rd Ed., Chapman & Hall, London, England.
Pi, Y. L., and Trahair, N. S. (1994). "Nonlinear inelastic analysis of steel beam-columns. Theory." 1. Struct.
Engrg., ASCE, 120(7),2041-2061.
Saint-Venant. (1855). "Memoire sur la torsion des prismes." Memoires des Savants Etrangers, France, XIV, 233-
560 (in French).
Smith, J. 0., and Sidebottom, O. M. (1965). Inelastic behavior of load carrying members. John Wiley & Sons,
Inc., New York, N.Y.
Sokolovsky. W. W. (1946). Theory of plasticity. Moscow, Russia (in Russian).
Timoshenko, S., and Gere, J. M. (1961). Theory of elastic stability. McGraw-HilI. New York. N.Y.
Trahair, N. S., and Bradford, M. A. (1991). The behaviour and design of steel structures, revised 2nd Ed.,
Chapman & Hall, London, England.
Tso, W. K., and Ghobarah, A. A. (1971). "Non-linear non-uniform torsion of thin-walled beams." Int. J. Mech.
Sci., Vol. 13,1039-1047.
Yamada, Y., Katagiri, S., and Takatruka, K. (1972). "Elastic-plastic analysis of Saint-Venant torsion problem
by a hybrid stress model." Int. J. Numerical Methods in Engrg., 5(2), 193-207.
Vlasov, V. Z. (1961). Thin-walled elastic beams. Israel Program for Scientific Translations, Jerusalem.
Wagner, H. (1936). "Verdrehung und Knickung von offenen Profilen (torsion and buckling of open sections)."
NACA Tech. Memo. No. 807, NACA, Washington, D.C.

APPENDIX II. NOTATION


The following symbols are used in this paper:

A cross-sectional area;
Ax gross cross-sectional area;
An net cross-sectional area;
a distance to torque;
B concentrated bimoment;
[B] matrices for strain-displacement relationships;
C cos <1>;
D overall depth;
[D] tangent modulus constitutive matrix;
{dE} incremental strains;
fda} incremental stresses;

JOURNAL OF STRUCTURAL ENGINEERING 619

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org
E Young's modulus of elasticity;
E, tangent modulus;
[E]e elastic stress-strain matrix;
[E]ep elastic-plastic stress-strain matrix;
G shear modulus of elasticity;
H' strain hardening parameter;
h distance between the flange centroids;
II" 11'1' section properties;
I,,"" section property;
Iw warping section constant;
J torsion section constant;
[kh element tangent stiffness matrix;
L length of element or length of member;
M applied concentrated torque;
Mf flange moment;
M fp flange moment at plastic collapse;
Mu uniform torque;
M 11P
full plastic uniform torque;
Mw warping torque;
M wp full plastic warping torque;
Mz torque about the OZ-axis;
m applied distributed torque per unit length;
[M)., stress resultant geometric matrix;
[NJo, [N]" shape function matrices;
[N)., shape function matrix;
{R} stress resultant vector;
S sin <1>;
Sw section property;
T applied concentrated torque;
Tp value of T at plastic collapse;
t, flange thickness;
tw web thickness;
U, 1.', W displacements of centroid in the OX, OY, OZ-directions;
Vfp flange shear at plastic collapse;
x,y, Z coordinates in the OX-, OY-, OZ-directions;
W Wagner stress resultant;
IV 1 + w';
C\: = coefficient in (9);
f3 elastic stiffness correction factor;
'(I' shear strain at P;
{~p} increment of external force vector;
{~pt increment of equivalent element external force vector;
{~pt unbalanced force vector at each iteration;
{M} incremental nodal displacement vector;
E normal strain;
{E} strain vector;
normal axial strain at P;
strain at which strain hardening starts;
yield strain;
nonuniform torsion plastic collapse load factor;
uniform torsion plastic collapse load factor;
warping torsion plastic collapse load factor;
normal stress;
stress vector;
effective stress;
normal stress at P;
arc compressive residual stress in flange;
arCK' compressive residual stress in web;
a" tensile residual stress;
au ultimate tensile strength;
normal yield stress;
shear stress;
shear stress at P;
shear yield stress;
twist rotation of cross section about centroidal axis OZ;
displacement vector;
uniform shear strain function; and
w section warping function.

620 JOURNAL OF STRUCTURAL ENGINEERING

Downloaded 03 Feb 2011 to 150.217.251.47. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org

Das könnte Ihnen auch gefallen