Torsion in Structural Design

Torsion in Structural Design
1. Introduction
1.1. Problems in Torsion
The role of torsion in structural design is subtle, and
complex. Some torsional phenomena include
(a) Twist of beams under loads not passing through the
shear center
(b) Torsion of shafts
(c) Torsional buckling of columns
(d) Lateral torsional buckling of beams
Two main types of situation involve consideration of
torsion in design
(1) Member's main function is the transmission of a
primary torque, or a primary torque combined with
bending or axial load (Cases (a) and (b) above.)
(2) Members in which torsion is a secondary
undesirable side effect tending to cause excessive
deformation or premature failure. (Cases (c) and
(d) above.)
Torsion in Structural Design - Notes
11/30/01
1.2. Development of Torsional analysis- A few key

contributors
1853 - French engineer Adhemar Jean Barre de
Saint-Venant presented the classical torsion theory
to the French Academy of Science
1899 - A. Michell and L. Prandtl presented results
on flexural-torsional buckling
1905 - S. P. Timoshenko presented a paper on the
effects of warping torsion in I beams
1909 - C. Bach noted the existence of warping
stresses not predicted by classical torsion theory
when the shear center and centroid do not coincide.
1929 - H. Wagner began to develop a general
theory of flexural torsional buckling
V. Z. Vlasov (1906-1958) developed the theory of
general bending and twisting of thin walled beams
1944 - von Karman and Christensen developed a
theory for closed sections (approximate theory)
1954 - Benscoter developed a more accurate
theory for closed sections.
Numerous other contributors, these are just a few
highlights.
11/30/01
2. Uniform Torsion of Prismatic Sections

Consider a prismatic shaft under constant twisting moment
along its length.
Classical theory due to St Venant. Assume

Cross-sections do not distort in plane during twisting, so
every point in the section rotates (in plane) through
angle ( x ) about the center of twist.
Out of plane warping is not constrained
Out of plane warping does not vary along the bar
The resulting displacement field is
u=
d
( y , z )
dx
d
dx
d
w = y( x ) = yx
dx
v = z( x ) = zx
11/30/01
In plane displacements v and w are seen from the figure

below
11/30/01
Out of plane distortion (warping) of the section is assumed

to vary with the rate of twist
d( x )
=constant ( x) = x
dx
and to be a function of the position (y,z) on the crosssection only. Several models may be constructed
Warping function model.

Conjugate Harmonic function model
St Venant's stress function model
11/30/01
Warping Function Model

Substituting the displ. fields into the diff. eq. of equilibrium
from elasticity, we obtain
2 2
+ 2 = 2 = 0
2
y
z
(Laplace's equation)
with b.c.
= zany yanz
n
where n is the normal direction to the boundary, and
( any , anz ) are the components of the unit normal vector n on
the boundary.
It can be shown that the St Venant torsional stiffness of
the section is given by
J = y 2 + z 2 + y
A
z dA
y
z
and that the angle of twist is related to the torque by

T=
JG
d
or T = JG
L
dx
a result that reduces to the usual polar moment of inertia

when the section becomes circular, and the warping
function vanishes.
Problem: The eqn. is hard to solve with the b.c given.
11/30/01
St Venant's Stress Function Model

Assume that the non-zero stresses xy , xz are related to a
stress function ( y , z ) by
xy =
xz =
The function ( y , z ) automatically satisfies equilibrium. In

order for the resulting displacements to be compatible (i.e.
satisfy continuity) the d.e.
2 2
d
+
=
2
G
= 2G
y 2 y 2
dx
be satisfied, where G=the shear modulus.

The boundary conditions for this model are
( x, y ) = constant on
where is the boundary of the section. In many cases, it is
convenient to simply take =0 on the boundary.
Given the stress function, it can be shown that
J = ( y , z )dydz
A
over the section.
11/30/01
This d.e. is somewhat easier to solve because of the

simpler b.c. We are particularly interested in rectangular
sections. Consider such a section, of dimension 2 x 2,
as shown.
It can be shown (using a Levy type solution) that, for this

section
32G2 1
cosh(nz / 2) ny
( n 1) / 2
=
(
1
)
1
cos
3
3
n =1, 3, 5,... n
cosh(n / 2) 2
The corresponding stresses are
16G2 1
cosh(nz / 2) ny
( n 1) / 2
xz =
(
1
)
1
sin
2
2
n =1,3,5,... n
cosh(
/
2
)
2
16G2 1
sinh( nz / 2) ny
( n 1) / 2
xy =
(
1
)
cos
2
2
n =1,3,5,... n
cosh(
/
2
)
2
11/30/01
is shown below for two sections
Assume overall dimensions b=2, t=2. Assume that b>t.

In general, it is useful to write
max = ktG
d
dx
J = k1t 3b
Then, since max T , we may write
max =
T
k2t 2b
9
11/30/01
Values of the coefficients are tabulated below for different

aspect ratios
Table 1 Coefficients for Torsion of a Rectangular
Section
b/t
1.0
1.2
1.5
2.0
2.5
3
4
5
10
k
0.675
0.759
0.848
0.930
0.968
0.985
0.997
0.999
1.000
1.000
k1
0.1406
0.166
0.196
0.229
0.249
0.263
0.281
0.291
0.312
0.333
k2
0.208
0.219
0.231
0.246
0.258
0.267
0.282
0.291
0.312
0.333
It is seen that for the thin section, the response away from
the ends is almost independent of y. Hence, a simplified
model takes the form
d 2
= 2G subject to ( ) = 0
dz 2
10
11/30/01
1.2
1
0.8
k
k1
k2
0.6
0.4
0.2
0
1 1.2 1.5 2 2.5 3
10
This leads to the simple solution
z2
( z ) = G 1 2 with resulting shear stress

2
xy = 2G( z / )
Hence, at z=t/2,
max = Gt = Gt
d
dx
Integrating this approximate function over the area,
1
J = bt 3
3
11
11/30/01
consistent with the limit of the 2-D solution as the width to

thickness ratio approaches . The shear stresses in a
narrow rectangular beam are as shown below
Thin-walled open section beams

Some thin walled open sections: angles, channels, W and
S sections, T sections, etc.
Key result 1. The section torsional stiffness J is

approximately equal to the sum of the J's for the
constituent thin walled plates. Thus,
1 n
3
J = bi ti
3 i =1
Key result 2. The maximum shear stress is estimated as

max =
Ttmax i
J
except that larger stresses may occur at the corners .
Key result 3. The shear flow around the section, caused

by St Venant torsion, is as shown below.
12
11/30/01
That is, the stresses vary linearly through the thickness on

any of the constituent plates, achieving a maximum on
each edge.
13
11/30/01
3. Shear Stresses due to bending - thin walled open

sections
Provides important background needed for analysis of

torsion
Three items are needed
shear stresses and shear flow concept
shear center
sectorial moments
These items can be obtained from a generalization of the
analysis studied in undergraduate Mechanics of Materials.
We summarize the results of the analysis below.
Normal Stresses and Resultants
From generalized analysis of beams,
xx =
I yy M z I yz M y
I yy I zz I yz
where
14
I zz M y I yz M z
I yy I zz I yz
I yy = z 2 dA
A
11/30/01
I zz = y 2 dA = the moments of inertia.

A
I yz = yz dA = product of inertia
A
From equilibrium of the slice of the section cut above,
q = t =
I yyQz ( s ) I yzQ y ( s )
I yy I zz I yz
Vy +
where
15
I zzQ y ( s ) I yzQz ( s )
I yy I zz I yz
Vz
Qy =
11/30/01
zdA = ztds ,
A*( s )
Qz =
ydA = ytds
A*( s )
q is the shear flow on the section. If the shear resultants

act through the shear center, there is no twist.
Summing moments about the centroid, it can be shown
that the coordinates of the shear center are
e y = 2
ez = 2
I zz I y I yz I z
I yy I zz I yz
I yy I z I yz I y
I yy I zz I yz
In the above,
S
I y
1
= Q y ( s ) rds = Q y ( s )d( s )
20
0
I z
1
= Qz ( s ) rds = Qz ( s )d( s )
20
0
are the sectorial products of inertia of the section, and

r ( s )ds = 2d( s ) define the sectorial area (the area shaded
in the figure above.)
r(s) is the perpendicular distance from the origin to the

force q ds needed in calculating moments.
16
11/30/01
4. Bending and torsion of open sections

For combined bending and torsion, a number of additional
sectorial area properties are needed for the section.
Let point A undergo
displacements v, w, and rotation .
A different point s , located at
coordinates (y,z) displaces by
v S = v ( z a z )
ws = w + ( y a y )
This can be resolved into a
normal, and a tangential
displacement component. The
tangential component of
displacement is
=v
dy
dz
+ w + r
ds
ds
The shear strain in the tangential plane is given by
xs =
u
+
0
s x
(Wagner's assumption)
From which the out of plane displacement is approximately
u = y(s)
dv
dw
d
z( s)
2( s ) + u0 ( x )
dx
dx
dx
In the above eqn.

17
11/30/01
2( s ) = rds = 2 d( s )
is the sectorial area at point s. The resulting normal stress
is then
d 2w
d 2 du0
u
d 2v
= E ( y 2 z 2 2 2 +
)
xx = E
dx
dx
dx
x
dx
Integrating the normal stress over the area leads to the
axial and bending resultants.
These may be written as
d 2
du0
ES 2
N x = EA
dx
dx
d 2w
d 2
d 2v
M y = EI yz 2 EI yy 2 + 2 EI y 2
dx
dx
dx
d 2v
d 2w
d 2
M z = EI zz 2 EI yz 2 + 2 EI z 2
dx
dx
dx
In the above, the first sectorial moment
S = 2( s )dA
A
has been used. The sectorial products of inertia I y , I z

were defined in connection with the shear center.
Solving for the displacements,
18
11/30/01
du0 N x S d 2
=
+
dx AE A dx 2
M z I yy M y I yz
I z I yy I y I yz d 2
d 2v
=
+2
2
2
dx 2
E ( I yy I zz I yz )
I yy I zz I yz dx 2
M y I zz M z I yz
I y I zz I z I yz d 2
d 2w
=
2 +2
2
dx 2
I yy I zz I yz dx 2
The coefficients multiplying " are the distances from A to
the shear center, so if point A is taken at the shear center,
these terms disappear, and the equations for v and w
reduce to
M z I yy M y I yz
d 2v
=
2
dx 2
M y I zz M z I yz
d 2w
=
2
dx 2
The equation for u0 " can be written in alternative form as
du0 N x
d 2
=
+ 20 2
dx AE
dx
where 0 = the sectorial area between the outside edge of
the section and the sectorial centroid, a point for which S
vanishes.If the s origin were taken at a sectorial centroid,
instead of at an outside edge, then this term would
disappear.
19
11/30/01
Shear Stresses and Resultants

For a section under bending and twisting, the total shear
stress at any point is
xs = ( xs ) SV +
q
t
where
( xs ) SV = St Venant torsion stresses

q
= shear stresses caused by shear flow on the
t
section.
Integrating the twisting moments caused by the shearing
stresses,
T = GJ
d
+ T
dx
where T is the torque caused by constrained warping

induced shear flow. It is sometimes convenient to express
this in terms of applied torque per unit length t ( x ) . Here,
d 2 dT
dT
= GJ 2
t( x) =
dx
dx
dx
If the free edge is taken as the origin of the s coordinate,
the shear flow is given by
s
q( s ) = xx dA
dx
0
20
11/30/01
Hence,
I Q I yzQ y
I zzQ y I yzQz Q
q( s ) = yy z
V
+
Vz V
y
2
2
I I I
C
I yy I zz I yz
yz
yy zz
where
W = 2 xx ( 0 )dA = the bimoment
A
V =
d
d
W = 2 xx ( 0 )dA
dx
dx A
s
Q = 2( 0 )dA =the first sectorial moment

0
C = 4( 0 ) 2 dA =the warping constant

A
Integrating q over the section, and recognizing that the

bending related shear flow vanishes if the load passes
through the shear center (which we may always impose)
we obtain, after some manipulation
d 3
T = ECw 3
dx
from which the equation for twist of a thin-walled section
d 4
d 2
ECw 4 GJ 2 = t ( x )
dx
dx
is obtained.
Example Consider an I beam, simply supported at each
end, free to warp at the ends, loaded by a twisting moment
in the middle.
21
11/30/01
The conditions at the left and right end are similar to those
at the end of a simply supported beam in bending. Hence
(0) = 0
( L) = 0
" (0) = 0
" ( L) = 0
Rather than use the fourth order differential equation which

has discontinuous load at midspan, use the third order
differential equations
EC' ' 'GJ' = T / 2
0 x L/2
EC' ' 'GJ' = T / 2
L/2 < x l
The homogeneous eqn.
EC' ' 'GJ' = 0

has the solution
22
11/30/01
h ( x ) = C0 + C1 sinh(x ) + C2 cosh(x )
GJ
.
Since the r.h.s. of the nonEC
homogeneous equations are constant, and the lowest
order derivative of in the o.d.e. is one, assume a
particular solution
where =
p = x
Substituting into the o.d.e.'s,
GJ = T / 2 = T / 2GJ 0 x L / 2
GJ = T / 2 = T / 2GJ L / 2 < x L
Hence, the general solution is given by
= C0 + C1 sinh(x ) + C2 cosh(x ) + Tx / 2GJ
0 x L/2
= C0 + C1 sinh(x ) + C2 cosh(x ) Tx / 2GJ
L/2 < x L
The two solutions must be identical at midspan, and by

symmetry, ' ( L / 2) = 0 .
Consider the left half of the beam first. For this portion of
the beam, the b.c. at the origin are
(0) = 0 C0 + C2 = 0
" (0) = 0 2C2 = 0
23
11/30/01
Hence, C0 = C2 = 0 , and the solution reduces to
= C1 sinh(x ) + Tx / 2GJ
Then, the symmetry condition at midspan
' ( L / 2) = 0 = C1 cosh(L / 2) + T / 2GJ

yields C1 =
1
T
.
2GJ cosh(L / 2)
Hence, for the left half of the beam,
T
sinh(x )
2 JG
cosh(L / 2)
For the right hand side of the beam, we may place the
origin at the right end, running to the left.
Then, it can be shown that the resulting solution (after
converting to the original coordinate system) is
T
sinh( ( L x ))
(
L
x
)
2 JG
cosh(L / 2)
Next, consider the relative magnitudes of the two

components of torsion over the span. We have, for the left
half of the span,
TSV = GJ' =
T
cosh(x )
1
2 cosh(L / 2)
24
TW = EC' ' ' =
11/30/01
T cosh(x )
2 cosh(L / 2)
For a typical W section, the values range from roughly

.04 for a heavy column section to less than .01 for some
very slender beams.
Contributions of the warping torsion are shown below for a
20' long beam.
When combined bending and twist of a beam occurs, it is

necessary to include all three components of shear flow
Shear Flow due to bending
St Venant Torsion
Warping Torsion
to evaluate the total shear stresses.
25
11/30/01
The shear flow from bending is calculated in the usual way,

as are the St Venant torsion shear stresses. The shear
flow from torsion may be calculated as
q =
Q
V
C
However, it is more direct to calculate the shear flow from

the fact that
d 2
d 3
W = EC 2 V = EC 3
dx
dx
d 3
So, q = EQ 3 provides the easiest calculation, given .
dx
How do we calculate Q ? By definition,
s
Q = 2( 0 )dA
0
26
11/30/01
S
, where S = 2dA was defined
A
A
earlier. In these integrals, is a continuous function of
position on the section, starting from 0 at the sectorial
origin.
In the integral, 0 =
In general, calculating these integrals is messy, but we

may often use a simpler approach.
27
11/30/01
Example: Suppose the above problem involves a W

section with properties shown.
Since the beam is doubly symmetrical, the shear center

coincides with the centroid. Taking this point as the origin
of sectorial areas, and as the sectorial center, the sectorial
areas (taken positive when rotated counterclockwise) are
as shown below.
The negative and positive
contributions cancel, so
S = 0 0 = 0 . Integrating the
sectorial areas over the flanges,
(from the outside edges), the Q
functions are as shown.
Within the web, the + and - areas
from the flanges cancel, so the
warping torsion induced shear flow
vanishes.
28
11/30/01
Then,
d 3 TQ cosh(x )
q = EQ 3 =
dx
2C cosh(L / 2)
In the above,
3
C =
h 2b f t f
qmax ( x )
24
3 T cosh(x )
=
4 h b f cosh(L / 2)
The same result can be achieved in a much more intuitive

manner. The sketch below illustrates the procedure
The lateral translation v introduced into the flange by the

rotation causes a bending moment unless ' is constant.
29
11/30/01
M = EI f v ' '
V = EI f v' ' '

h
' ' '
2
3
bf t f h
=E
' ' '
12 2
= EI f
This shear produces a torque

2
3
b
t
h
h

T f(1) = V = E f f ' ' '
2
12 2
A similar torque develops on the bottom flange, so
3
T = Tf
(1)
+ Tf
(2)
bf t f h 2
= E
' ' '
24
But the warping torsion is given as

3
bf t f h 2
T = EC' ' ' C =
24
in agreement with the previous result.
Within the top flange, bending introduces a shear flow

30
11/30/01
VQ f
q =
If
From the above sketch,

tf
Qf =
2
b f 2
2
y
2
bf t f
If =
12
So
b f 3t f
t f b f
h ' ' '
E
24
2 2
q =
3
bf t f
2
y
12
h tf
=E
2 2
b f
2
y ' ' '
31
11/30/01
For the beam analyzed previously,
' ' ' =
T
cosh(x )
2 EC cosh(L / 2)
so
h tf
q =
2 2
h tf
=
2 2
b f 2
T cosh(x )
2
y
2
2C cosh(L / 2)
b f 2
24 T cosh(x )
2
y 3 2
2
b f t f h 2 cosh(L / 2)
b f 2
3T cosh(x )
2
= y 3
2
b f h cosh(L / 2)
The peak value of warping shear flow occurs at the middle
of the flange (y=0) at the midspan of the beam (x=L/2),
where
q =
3 T
4 bf h
The third part of the sketch on p. TO.30 re-establishes

compatibility by rotating the flanges, and the web about an
angle about their respective centroids. This introduces
the St Venant torsion, as well as a secondary warping
torsion caused by the plate bending of the individual flange
and web plates. We usually ignore the secondary warping
except for certain torsionally weak sections.
32
11/30/01
Secondary Warping:
For certain sections, the only resistance to warping is the
secondary warping resistance. Several such sections are
shown below
The secondary warping torsional resistance can often be

computed for such sections using an analogy to the
approach used for the flange above. Consider a plate,
rotated about its base through an angle which varies with
x. At y,
v ' ' ' ( y ) = y' ' '

t 3dy
dVs ( y ) = EI s ( y )' ' ' = E
y' ' '
12
t 3dy 2
dTs ( y ) = dVs ( y ) y = E
y ' ' '
12
Integrating over the height of the plate,

33
hp
11/30/01
t 3h p
t3 y2
' ' '
Ts = E
dy ' ' ' = E
12
36
0
t 3h p
Cs =
36
All of the cases shown above can be constructed by

variations of the above. For example, a plate of height h
rotated about its centroid is equivalent to two plates of h/2
rotated about their respective bases. Hence,
t 3 ( h / 2 ) 3 t 3h 3
=
Cs = 2
36
144
Likewise, the T section has warping stiffness
3
t1 b3 t2 h 3
Cs =
+
144 36
and the cruciform shown has warping stiffness
t 3h 3 t 3h 3
=
Cs = 2
144 72
(You should calculate the corresponding warping stiffness
for an angle with unequal legs.)
If the plates are thin, the warping stiffness may be
relatively small. This can have an important effect, as we'll
see.
34
11/30/01
Design Approximations for Torsion

Analysis using the differential equation for torsion is
complicated, even for simple loads, so various
approximations have been introduced.
Idea: In warping torsion, the flanges act like beams
bending laterally.
A (conservative) approximate analysis might be

1. convert the torque into couples acting on the flanges by
dividing by the beam height.
2. Apply the "equivalent" lateral loads as forces on the
flanges.
3. Analyze the flanges as beams under these loads.
35
11/30/01
Example:
For the beam analyzed before, the torque converts into
two lateral loads
P = T /h
applied at the top and bottom flanges.
The resulting shear is half of this, on each end
V = T / 2h
The shear flow is thus
b f 2
2
3 y T
2
q VQ f / I f =
3
h bf
36
11/30/01
The max. shear flow at y=0 is

qmax =
3T
4h b f
which clearly is an upper bound to the actual shear flow.

This approach is often overly conservative, because
The normal stresses caused by the bimoment are most

important quantitatively, and may be as large as the
bending stresses for a beam in combined bending and
torsion
The flange shear flow predicted by the bending analogy
will over-estimate the actual warping torsion shear flow
over much of the beam. Since, in the same way that
x
dM
V=
M ( x ) = M (0) + V ( )d
dx
0
We also have
x
dW
V =
W ( x ) = W (0) + V ( )d
dx
0
37
11/30/01
For example, by the beam analogy,
The beam analogy yields
M f
T L
T L bf / 2
=
from which xx =
2h 2
2h 2 I f
However, the actual shear is
Vf =
T cosh(x )
2h cosh(L / 2)
Hence, the integrated lateral flange moment is
M f
T
=
2h
L/2
cosh(x )
T L T L
dx =
<
cosh(L / 2)
2h 2 2 h 2
where is a reduction factor, dependent upon

L
The end conditions
The specifics of the applied torque
S & J give values of for some common load cases.
38
11/30/01
Are the stresses caused by torsion large enough to be a

problem? Let's see.
Example: A W18x71 beam spanning 24 ft. is loaded with a
concentrated load of 20 kips at midspan. The load acts 2''
away from the Z axis. The beam is fixed at both ends.
(Variation on S&J example.)

For the beam, the properties are
3
bf t f h 2
C =
24
(7.635" )3 (.810" )(18.47".81" ) 2
=
= 4,685 in 6
24
bt 3 1
J =
= [2(7.635" )(.810" )3 + (18.47 1.62)(.495" )3 ]
3 3
= 3.39 in 4
E / G = 2(1 + ) = 2(1 + 0.3) = 2.6

39
11/30/01
JG
3.39
0.01668
=
=
EC
( 4,685)2.6
in.
For the fixed-fixed case, we can show that
T
[x sinh(x )
2 JG
(1 cosh(L / 2) )
+
(1 cosh(x ))
sinh(L / 2)
so
(1 cosh(L / 2) )
T
' =
1
cosh(
x
)
sinh(
x
)
2 JG
sinh(L / 2)
(1 cosh(L / 2) )
T
" =
sinh(
x
)
cosh(
x
)
2 JG
sinh(L / 2)
(a) Saint-Venant Torsion

SV = Gt f '
Tt
[1 cosh(x )
2J
+ (1 cosh(L / 2)) sinh(x ) / sinh(L / 2)]
Here, sinh(L / 2) = 5.477 , 1 cosh(L / 2) = 4.567 , so
40
SV =
11/30/01
( 40" k )(0.81" )
[1 cosh(.01668 x ) + 0.8338 sinh(.01668 x )]
4
2(3.39 in )
For the fixed-fixed case, the Saint-Venant torsion is zero at

both support and centerline, and is maximum at L / 4 ,
where
SV 1 = 2.141 ksi
(b) Warping Torsion Shear: For the W section, following
the above reasoning,
2
max
b h
= E f ' ' '
16
From the solution,
(1 cosh(L / 2) )
T2
' ' ' =
cosh(
x
)
sinh(
x
)
2 JG
sinh(L / 2)
so
2
bf h T
1 cosh(L / 2)
max =
cosh(
x
)
sinh(
x
)
32C
sinh(L / 2)
= 0.2556[ cosh(.01668 x ) + 0.8338 sinh(.01668 x )]
max reaches its max. amplitude at the end, and at the

centerline, where the warping restraint is greatest. There
(0) = ( L / 2) = 0.2556
41
11/30/01
It is smallest at the quarter points where warping restraint

is least.
(c) Lateral bending stresses introduced by warping torsion

The bending stress caused by warping is given by
xx
d 2
= E 2
dx
From the above analysis, the largest magnitude stress

occurs at the flange tips, where = b f h / 4 . Thus,
bh
xx = f E"
4
b h E T
= f
[ sinh(x ) (1 cosh(L / 2) ) cosh(x ) / sinh(L / 2]
4 G 2J
at the flange tips
Substituting the numerical values leads to

42
11/30/01
(7.635" )(17.66" )
(0.01668 / in.)( 40 " k )
( 2.6)
4
2(3.39 in 4 )
[ sinh(.01668 x ) + 0.8338 cosh(.01668 x )]
= 8.625[ sinh(.01668 x ) + 0.8338 cosh(.01668 x )]
xx =
The maximum values of xx occur at the ends, and at

midspan, where xx = 7.192 ksi (compressive on one side
of the flange, tensile on the other.
The maximum normal stress on one edge of the flange is

shown over a half beam length. Maximum warping
restraint exists at the support, and at midspan.
(d) Bending analogy For this case
43
11/30/01
h = 17.66" , T = 40" k Pf =
40
= 2.265 kips
17.66
For a fixed-fixed beam, max moments are

M=
Pf L ( 2.265 kips)( 288" )

=
= 81.54" k
8
8
2
t f bf
(0.81" )(7.635" ) 2
Then, if S fl =
=
= 7.8696 in 3
6
6
xx equiv . =
81.54" k
= 10.361 ksi
3
7.8696 in
As noted above, this is overly conservative.

From table 8.6.2, the interpolated value is
= (0.8)(0.68) + (0.2)(0.76) = 0.696

Then,
xx 0.696(10.361 ksi) = 7.211 ksi

which is pretty close to the maximum value obtained from
the closed form solution.
Observations:
44
11/30/01
The Saint-Venant torsion shear stresses are somewhat

larger than the warping torsion shear stresses, but both
are relatively small.
The warping normal stresses may be significant.
(e) To see how significant, let's compute the bending
stresses
M=
PL ( 20 kips)( 288" )
=
= 720" k
8
8
S x = 127 in 3
xx bending = 720" k / 127 in 3 = 5.67 ksi
In this case, the warping normal stresses are larger than
the bending stresses!
45
11/30/01
5. Other Approaches:
Calculating may be simplified by tabulated solutions for
certain idealized boundary conditions, or a matrix stiffness
equation
T1
a2
a1
W
a3
1 JG a2
=
T
2 a1 a2
a
W
a4
2
2
a1
a2
a1
a2
a2 1
a4 1 '

a2 2
a3 2 '
where
= 2 L sinh(L) 2 cosh(L) + 2
a1 = 2 sinh(L)
a 2 = [cosh(L) 1]
a 3 = L cosh(L) sinh(L)
may be useful.
a 4 = sinh(L) L
This stiffness matrix may be used to replace the 2x2

torsional stiffness submatrix
T1 JG 1 1 1
=
1 1
T
L
2
2
commonly used in the beam (or frame) equations, and
incorporates the end conditions commonly encountered.
46
11/30/01
This approach has the advantage that it can produce

reasonable torsional stiffnesses for applications involving
torsional moments caused by eccentrically connected
beams (See S&J, section 8.7)
The associated solution
( x ) = f1 ( x )1 + f 2 ( x )1 '+ f 3 ( x )2 + f 4 ( x )2 '
can then be used to obtain all derivatives of needed to
calculate the shear flow, where
f1 = [a1(L x) a2 (1 + cosh(x)) + a1 sinh(x) / ]/
f2 = [ a2 x + a3(1 cosh(x)) + (a1L a2 )sinh(x) / ]/

f3 = [a1( x sinh(x) / ) a2 (1 cosh(x))] /
f4 = [a4 (1 cosh(x) a2 ( x sinh(x) / ]/
are the "shape functions" for the problem. Distributed

torques can also be accomodated.
47
11/30/01
S&J suggest that the best approach for determining the

applied torque is one that establishes compatibility
between the twisting rotation of the beam supporting a
torque producing member, and the end bending rotation
of the attached element. They note that
Applying the shear at the face of the web
underestimates the torsion
Applying the shear at the centerline of the connection
overestimates the torsion
Applying the shear so as to achieve rotational continuity
with the end of the beam framing in tends to be about
right.
48
11/30/01
6. LRFD for Torsional Moments

Assumption: Torsion stresses add to bending stresses, but
lateral torsional instability is prevented.
i.e. lateral restraint exists, but laterally restraining
members may induce self-limiting torsion before restraint
becomes effective.
Other problems exist as well, for which torsional rotation
may not be self limited.
AISC Philosophy: (consistent with ASD)
Assume the limit state occurs when yielding is reached.
Elastic analysis is applicable.
Compute factored stresses using wD , wL , etc, e.g.
wu = 1.2 w D + 1.6 w L ,
Once the factored design moment and factored design
lateral flange moments are known, the stresses are
superpositioned with the limit state. Formally,
M bz
Sz
M
b Fy
S
or, in terms of the equivalent lateral flange moment
M bz
Sz
2M f
b Fy
Sy
In the above, b = 0.9 =resistance factor.
49
11/30/01
Since is a section property that influences the M

term, through the factor, or through the d.e. solution,
iterative solution is necessary.
Guess a value of in the
right approximate range.
Use the corresponding value of L , to determine the
value of M to be used with the design bending
moment.
Use the bounds to estimate a section size.
Use the section size to update the estimated value of
., etc.
Example: An A36 beam with torsionally simply supported
ends must carry two 20 kip loads acting at an eccentricity
of 6" (5 Kips DL, 15 Kips LL) (Variation on S&J example)
Pu = 1.2(5 kips) + 1.6(15 kips) = 30 kips

Guess wbeam = 150 plf wu = 1.2(150) = 180 plf
Bending moments:
50
11/30/01
Torsion:
Tu = (6" )(30 kips) = 180" k.
Using beam analogy
Pf = Tu / h = 180" k / h
M f = Pf L / 3 = (180" k )(144" ) / h
= 25,920 in 2 k / h
The reduction factor depends upon L. Guess = 0.015 ,

(for a majority of W sections, .01 < < .02 ) Thus,
L = .015( 432" ) = 6.48 0.36
51
11/30/01
The value for the torques in the middle of the beam

aren't exactly known, but the estimates tend to be
somewhat conservative.
Guessing a 14" beam, the reduced flange moment is
M f = 0.36( 25,920 in 2 - k ) / 14" = 667" k
For a midsized W14,

S y / S z 2.6 , taking both flanges
into account, so
4670" k 667" k 2
+
0.9(36 ksi)
Sy
S y / 2.6
S y 251 in 3
Try W14x159. For this section, S y / S z = 2.64 , so the guess

is o.k. for now. Also, for this section
1
=
EC
= 68.2 (p. 1-148 of AISC-LRFD)
GJ
so L = 432" / 68.2 = 6.33

This implies a revised = 0.37
Now,
52
11/30/01
M f = 0.37( 25,920) / 13.79" = 695" k

So
Sy
4,670 " k (695" k )( 2)

+
= 257 in 3
32.4 ksi 32.4 / 2.64
The W14x159 is not quite big enough, since S y = 254 in 3 .

(actually, it's probably o.k., in view of the approximations in
calculating , but without a more precise calculation, we
can't be sure.)
Try a W14x176 Revise the moment upward to
4,720"k to account for the extra dead load.
From the tables,
= 1 / 62.9 L = 432 / 62.9 = 6.86
Sy =
S y / S z = 2.63
Revised = 0.35
h = 13.91"
M f = .35( 25,920) / 13.91 = 652" k
4720 652 2
+
= 29 ksi < 32.4 ksi
281 281 / 2.63
o.k. A W14x176 is more than adequate here
53
11/30/01
7. Presence of Axial Loads (Torsional Instability)

In presence of axial loads, or under certain lateral load
conditions, instability can occur. In this case, formulation
in the displaced state is necessary.
The relevant displacements are v, w and . So, the first
equation (for u0 ) is not changed.
Formulation is complicated, (as is the result) so, we'll just
present the resulting equations for v, w and .
The complete equations are nonlinear. A linear version
which omits the non-linear terms is presented here.
d 4v
d 2v
d 2w
E ( I yy I zz I yz ) 4 I yy N x 2 + I yz N x 2
dx
dx
dx
d 2
+ [ I yy M ly + I yz M lz N x ( ez I yy + e y I yz )] 2
dx
d
+ ( I yyVz + I yzV y ) = q y I y qz I yz
dx
2
54
11/30/01
d 2v
d 2w
d 4w
E ( I yy I zz I yz ) 4 I zz N x 2 + I yz N x 2
dx
dx
dx
d 2
+ [ I zz M lz I yz M ly + N x ( e y I zz + ez I yz )] 2
dx
d
( I zzV y I yzVz ) = qz I zz q y I yz
dx
d 4
IE
H d 2
EC 4 GJ + N x + C z M lz + C y M ly +
W 2
dx
A
C
dx
2
d 2v
d 2w
dv
dw
+ ( M ly ez N x ) 2 ( M lz e y N x ) 2 + Vz V y
dx
dx
dx
dx
d
H
C zV y + C yVz + V = t ( x )
C dx
In these equations, a number of additional quantities are to

be defined.
M ly , M lz = Moments caused by transverse loads only
I E = I yy + I zz + ( e y + ez ) A = polar moment of inertia about

shear center
2
Three new geometrical quantities are
55
11/30/01
H y = z ( y 2 + z 2 )dA
A
H z = y ( y 2 + z 2 )dA
A
H = 2( 0 )( y 2 + z 2 )dA
A
From these,
Cy =
Cz =
I zz H y I yz H z
I yy I zz I yz
2 ez
I yy H z I yz H y
2e y
2
I yy I zz I yz
The equations are three simultaneous equations in three
unknowns. Although the geometrical properties are
constant, the load coefficients M ly , M lz ,V y ,Vz ,W ,V all vary
with x.
The d.e. coefficients are variable.
Often, the d.e.s can be simplified, by taking advantage of
special properties of different classes of problems.
We will want to do this in practice, since the original set of
equations is unwieldy.
56
11/30/01
First, assume that the y-z coordinates are principal

coordinate axes. Then I yz = 0 , and the eqns reduce to
d
d 2
d 2v
d 4v
EI zz 4 N x 2 + ( M ly N x ez ) 2 + Vz y
= qy
dx
dx
dx
dx
d 4w
d 2w
d 2
d
EI yy 4 N x 2 + ( M lz + N x e y ) 2 V y
= qz
dx
dx
dx
dx
d 4
IE
H d 2
EC 4 GJ + N x + C z M lz + C y M ly +
W 2
dx
A
C
dx
dw
dv
d 2w
d 2v
+ ( M ly ez N x ) 2 ( M lz e y N x ) 2 + Vz
Vy
dx
dx
dx
dx
d
H
C zV y + C yVz + V = t ( x )
C dx
If, in addition, the transverse loads q y , qz are zero, and the

transverse moments are constant, then the shears V y ,Vz
are also zero, and the additional simplification
d 4v
d 2v
d 2
EI zz 4 N x 2 + ( M ly N x ez ) 2 = 0
dx
dx
dx
d 2
d 2w
d 4w
EI yy 4 N x 2 + ( M lz + N x e y ) 2 = 0
dx
dx
dx
57
11/30/01
d 4
IE
d 2 H d d
EC 4 GJ + N x + C z M lz + C y M ly 2
W
dx
A
dx
C
dx
dx
d 2v
d 2w
+ ( M ly ez N x ) 2 ( M lz e y N x ) 2 = t ( x )
dx
dx
is possible.
Now, assuming that the bar is subjected only to an axial
compressive load through the centroid, the moments due
to transverse loads are zero, the applied torque vanishes,
and if the loading is applied so that the bimoment is zero,
the equations become
d 4v
d 2v
d 2
EI zz 4 + N x 2 + N x ez 2 = 0
dx
dx
dx
d 4w
d 2w
d 2
EI yy 4 + N x 2 N x e y 2 = 0
dx
dx
dx
2
d 4
IE
d 2v
d 2w
d
EC 4 GJ N x 2 + ez N x 2 e y N x 2 = 0
dx
A
dx
dx
dx
These equations represent a generalization of the elastic

column buckling problem. (Note sign change in N x
because of compression.)
Assume a pinned end column for which the appropriate
boundary conditions are
v=0
w=0
=0
v" = 0 w" = 0 " = 0
58
11/30/01
Then a solution of the form
v = v0 sin( x / L)
w = w0 sin( x / L)
= 0 sin( x / L)
where v0 , w0 , 0 are the values of v, w, at midspan, will
satisfy the resulting equations provided
2
( EI zz 2 N x )v0 N x ez 0 = 0
L
2
( EI yy 2 N x ) w0 + N x e y 0 = 0
L
2
I
ez N x v0 + e y N x w0 + EC 2 + GJ E N x 0 = 0
L
A
or, in matrix form
2
0
EI
N
N
e
x
x z
zz L2
v0 0
w0 = 0
0
EI yy 2 N x
N xey
L
IE
0 0
N xey
EC 2 + GJ N x
N x ez
L
A
Non-trivial solutions exist iff the determinant of coefficients

is zero. Rather than solve for the general case, let's
consider two special cases.
59
11/30/01
(a) The section is doubly symmetric. Then e y = ez = 0 , and

the equations reduce to
EI zz
2
Nx
L2
2
EI yy 2 N x
L
v 0 0
w0 = 0
0

2
IE
0 0
EC 2 + GJ N x
L
A
The corresponding determinantal equation
2
2
2
I
EI zz 2 N x EI yy 2 N x EC 2 + GJ E N x = 0
A
L
L
L
is already factored, and has the three solutions

2
N x = EI zz 2 = PE zz (buckling about z axis)
L
2
N x = EI yy 2 = PE yy (buckling about y axis)
L
2
A
N x = EC 2 + GJ = PT (Torsional, or twist buckling)
IE
L
rd
The 3 solution corresponds to a new mode of buckling,
shown below, for a cruciform section.
60
11/30/01
Dividing by A and noting that I E = I p for a doubly

symmetrical section yields the critical stress
xx cr
2
1
= EC 2 + GJ
Ip
L
If this critical stress is less than the minor axis Euler

stress, then torsional buckling will preceed Euler buckling
of the section, and must be accounted for in design.
61
11/30/01
Example: Consider an equal leg cruciform section with

h=6",t=1/4".
Section properties:
I yy
th( h 2 + t 2 ) (6" )(.25" )[( 6" ) 2 + (.25" ) 2 ]

= I zz =
=
= 4.508 in 4
12
12
I p = I yy + I zz = 9.016 in 4
A = 2th t 2 = 2(6" )(.25" ) (.25" ) 2 = 2.938 in 2 ry = rz = 1.239"
t 3h 3 (.25" ) 3 (6" )3
=
= 0.04688 in 6
C =
72
72
( h / 2)t 3 2ht 3 2(6" )(.25" ) 3
J =4
=
=
= 0.0625 in 4
3
3
3
For Euler buckling, both axes have the same I, so
( Euler )
xx cr
2 E
2 (30,000)
454,530
2
=
=
=
(
1
.
239
"
)
( L / r)2
L2
L2
For torsional buckling,
62
( Torsion )
xx cr
11/30/01
1
2
= ( EC 2 + GJ )
Ip
L
1
2
=
[(30,000)(0.04688) 2 + (11,540)(0.0625)]
9.016
L
1,540
= 2 + 80
L
(The GJ term dominates the torsional buckling equation

unless the column is very short.)
)
( Euler )
Torsion controls if (xxTorsion
<
, so
xx cr
cr
1540
454,530
+
80
<
L < 75.2"
L2
L2
In a symmetrical section, torsional buckling tends to

control only for short columns.
The thinner the walls of the column, relative to the plate
widths (Thin plates) the more likely it is that torsional
buckling will control. (Increasing the wall thickness to
3/8" in the above calculations decreases the length
column for which torsion controls to 50".)
It may be feasible to ignore the warping torsion term in
the above in the practical range of interest.
For the torsion buckling, it is appropriate to replace E
with E /(1 2 ) consistent with dominant plate action.
(b) The section is singly symmetrical. If the z axis is an
axis of symmetry, then e y = 0 , and
63
2
EI zz L2 N x
N x ez
0
2
EI yy 2 N x
L
0
11/30/01
v 0 0
w0 = 0
0

2
IE
0 0
EC 2 + GJ N x
L
A
N x ez
From which, the determinantal equation is
2
2
2
I
2 2
EI yy 2 N x EI zz 2 N x EC 2 + GJ E N x N x ez = 0
L
L
L
A
Before solving this problem, it is useful to substitute the

relations
2
2
2
IE
PE zz = EI zz 2
PE yy = EI yy 2
PT = EC 2 + GJ
L
L
A
L
Then, the determinantal eqn. takes the form
(P
I
2 2
N x E (PE zz N x )(PT N x ) N x ez = 0
A
This is an important case, as T beams, one section

commonly used for compression elements (e.g. truss
members) is singly symmetrical.
E yy
The leading factor is just Euler column buckling in the

plane of symmetry.
The other factor leads to a coupled flexural-torsional
buckling mode. The formal solution is
64
( PE zz
Nx =
11/30/01
Aez
PE zz PT
+ PT ) ( PE zz + PT ) 41
IE
Aez
21
I
E
Typically, this critical load will be smaller than either the

twist buckling load, or the flexural buckling mode.
65
11/30/01
Dividing the equation by the area permits the relationship

to be written in terms of the critical stresses for the
individual buckling modes.
(
(E)
xx cr
(xxFTcr ) =
(T )
xx cr
Aez ( E ) (T )
xx cr xx cr
) ( + ) 41
I
E
Aez
21
IE
(E)
xx cr
(T ) 2
xx cr
where
E
xx cr
2
=E
( L / ry ) 2
(T )
xx cr
E 2
J
= C 2 +
I E L 2(1 + )
One can use the latter equation to define an equivalent

radius of gyration for torsional buckling. Equating
2 E
E 2
J
=
+
C
( L / rE ) 2 I E L2 2(1 + )
yields the resulting equivalent r
rE =
1
JL2
C + 2
2 (1 + )
IE
66
11/30/01
Example: Consider a WT12x27.5:

A = 8.10 in 2
I yy = 117 in 4
I zz = 14.5 in 4
rz = 3.8"
ry = 1.34"
ez = 3.248"
J = 0.588 in 4
C = 2.764 in 6
I E = 117 in 4 + 14.5 in 4 + (8.1 in 2 )(3.248" )2 = 217 in 4
2
2
531,700
E
2
xx cr = E
=
=
30
,
000
(
1
.
34
"
)
( L / ry ) 2
L2
L2
2 0.588
E 2
J 30,000
= C 2 +
=
2.764 2 +
I E L 2(1 + )
L
217
2. 6
(T )
xx cr
3,771
+ 31.26
2
L
Substituting into the critical stress eqn yields
535,500
535,500
1,289,300 3,771
2
31
.
3
)
(
31
.
3
)
31
.
3
+
2
L2
L2
L2
L
=
1.213
(
(xxFTcr )
This result is shown below in two plots.

First plot shows the interaction of the critical stresses
over a large theoretical range
Second plot shows the interaction of the critical stresses
over a practical range
67
11/30/01
Flexural-Torsional interaction significantly reduces the

buckling load for this column, extending well into the
elastic range!
68
11/30/01
7. References
Bleich, F. (1952), Buckling Strength of Metal Structures,
McGraw Hill Book Co., New York, N.Y.
Boresi, A. P., Schmidt, R. J., and Sidebottom, O. M.,
(1993) Advanced Mechanics of Materials, 5th Edition,
John Wiley and Sons, Ch. 6 and 8.
Chang, F. K., and Johnston, B. G., (1953), "Torsion of
Plate Girders," Transactions, ASCE, 118, 337-396.
Chen, M.-T., and Jolissaint, D. E. Jr., (1983), "Pure and
Warping Torsion Analysis of Rigid Frames," Journal of
Structural Engineering, ASCE, 109, 8, 1999-2003.
Chu, K-H., and Johnson, R. B., (1974), "Torsion in Beams
with Open Sections," Journal of the Structural Division,
ASCE, 100, ST7, 1397-1419.
Chu, K-H., and Longinow, A., (1967), "Torsion in Sections
with Open and Closed Parts," Journal of the Structural
Division, ASCE, 93, ST6, 213-227.
Dean, D. L., (1994), "Torsion of Regular Multicellular
Members," Journal of Structural Engineering, 120, 12,
3675-3678
Driver, R. G., and Kennedy, D. J. L., (1989), "Combined
Flexure and Torsion of I-shaped Steel Beams," Canadian
Journal of Civil Engineering, 16, 124-139.
69
11/30/01
El Darwish, I. A., and Johnston, B. G., (1965), "Torsion of

Structural Shapes," Journal of the Structural Division,
ASCE, 91, ST1, 203-227, Errata: 92, ST1, 471.
Evick, D. R., and Heins, C. P. Jr., (1972), "Torsion of
Nonprismatic Beams of Open Section," Journal of the
Structural Division, ASCE, 98, ST12, 2769-2784.
Felton, L. P., and Dobbs, M. W., (1967), "Optimum Design
of Tubes for Bending and Torsion," Journal of the
Structural Division, ASCE, 93, ST4, 185-200.
Goldberg, J. E., (1953), "Torsion of I-Type and H-Type
Beams," Transactions, ASCE, 118, 771-793.
Goodier, J. N., and Barton, M. V., (1944), "The Effects of
Web Deformation on the Torsion of I Beams," Journal of
Applied Mechanics, ASME, p A-35.
Heins, C. P. Jr., (1975) Bending and Torsional Design in
Structural Members, Lexington Books, Lexington, MA.
Heins, C. P., Jr., and Kuo, J. T. C., (1972), "Torsional
Properties of Composite-Girders," Engineering Journal,
AISC, 9, 2, 79-85.
Heins, C. P. Jr., and Potocko, R. A., (1979), "Torsional
Stiffening of I-Girder Webs," Journal of the Structural
Division, ASCE, 105, ST8, 1689-1698.
70
11/30/01
Hotchkiss, J. G., (1966), "Torsion of Rolled Steel Sections

in Building Structures," Engineering Journal , AISC, 3, 1,
19-45.
Johnston, B. G., (1982), "Design of W Shapes for
Combined Bending and Torsion," Engineering Journal,
AISC, 19, 2, 65-85.
Kubo, G. G., Johnston, B. G., and Eney, W. J., (1956),
"Nonuniform Torsion of Plate Girders," Transactions,
ASCE, 121, 759-785.
Lin, P. H., (1977), "Simplified Design for Torsional Loading
of Rolled Steel Members," Engineering Journal, AISC, 14,
3, 97-107.
Lin, P. H., Discussion of "Design of W-Shapes for
Combined Bending and Torsion," by B. G. Johnston,
Engineering Journal, AISC, 20, 2, 82-87.
Lue, T. and Ellifritt, D. S., (1993), "The Warping Constant
for the W-Section with a Channel Cap," Engineering
Journal, AISC, 30, 1, 31-33.
McGuire, W., (1968) Steel Structures, Prentice Hall,
Englewood Cliffs N.J., pp. 346-400.
Murray, N. W., (1984), Introduction to the Theory of Thinwalled Structures, Oxford Univ. Press, Oxford, Ch. 2 and
3.
71
11/30/01
Oden, J. T., (1967), Mechanics of Elastic Structures,

McGraw Hill, N.Y., Ch. 3, 5 and 7.
Pi, Y. L., and Trahair, N. S., (1995a) "Inelastic Torsion of
Steel I-Beams, Journal of Structural Engineering, ASCE,
121, 4, 609-620.
Pi, Y. L., and Trahair, N. S., (1995b), "Plastic Collapse
Analysis of Torsion," Journal of Structural Engineering,
ASCE, 121, 10, 1389-1395.
Pi, Y. L., and Trahair, N. S., (1994a), "Inelastic Bending
and Torsion of Steel I-Beams," Journal of Structural
Engineering, ASCE, 120, 12, 3397-3417.
Pi, Y. L., and Trahair N. S., (1994b), "Nonlinear Inelastic
Analysis of Steel Beam-Columns I: Theory," Journal of
Structural Engineering, 120, 7, 2041-2061.
Pi, Y. L., and Trahair N. S., (1994c), "Nonlinear Inelastic
Analysis of Steel Beam-Columns II: Applications," Journal
of Structural Engineering, 120, 7, 2062-2085.
Salmon, C. G., and Johnson, J. E., (1996) Steel
Structures, Design and Behavior, Harper and Row, New
York.
Shermer, C. L., (1980), "Torsional Strength and Stiffness
of Steel Structures," Engineering Journal, AISC, 17, 2, 3337.
72
11/30/01
Siev, A. (1966), "Torsion in Closed Sections," Engineering

Journal, AISC, 3, 1, 46-54.
Timoshenko, S., (1941), Strength of Materials, Part II, 2d
ed., New York, Van Nostrand Co. Inc., Ch. 6.
Vacharajittiphan, P., and Trahair, N. S., (1974), "Warping
and Distortion at I-Section Joints," Journal of the Structural
Division, ASCE, 100, ST3, 547-564.
Vlasov, V. Z., (1961) Thin Walled Elastic Beams, Israel
Program for Scientific Translations, Jerusalem.
73

Torsion in Structural Design

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Torsion in Structural Design

Hochgeladen von

Copyright:

Verfügbare Formate

Torsion in Structural Design

Torsion in Structural Design - Notes

1.2. Development of Torsional analysis- A few key

Torsion in Structural Design - Notes

2. Uniform Torsion of Prismatic Sections

Classical theory due to St Venant. Assume

Torsion in Structural Design - Notes

In plane displacements v and w are seen from the figure

Torsion in Structural Design - Notes

Out of plane distortion (warping) of the section is assumed

Warping function model.

Torsion in Structural Design - Notes

Warping Function Model

and that the angle of twist is related to the torque by

a result that reduces to the usual polar moment of inertia

Torsion in Structural Design - Notes

St Venant's Stress Function Model

The function ( y , z ) automatically satisfies equilibrium. In

be satisfied, where G=the shear modulus.

over the section.

Torsion in Structural Design - Notes

This d.e. is somewhat easier to solve because of the

It can be shown (using a Levy type solution) that, for this

Torsion in Structural Design - Notes

is shown below for two sections

Assume overall dimensions b=2, t=2. Assume that b>t.

Torsion in Structural Design - Notes

Values of the coefficients are tabulated below for different

Torsion in Structural Design - Notes

This leads to the simple solution

Integrating this approximate function over the area,

Torsion in Structural Design - Notes

consistent with the limit of the 2-D solution as the width to

Thin-walled open section beams

Key result 1. The section torsional stiffness J is

Key result 2. The maximum shear stress is estimated as

except that larger stresses may occur at the corners .

Key result 3. The shear flow around the section, caused

Torsion in Structural Design - Notes

That is, the stresses vary linearly through the thickness on

Torsion in Structural Design - Notes

3. Shear Stresses due to bending - thin walled open

Provides important background needed for analysis of

From generalized analysis of beams,

Torsion in Structural Design - Notes

I zz = y 2 dA = the moments of inertia.

From equilibrium of the slice of the section cut above,

Torsion in Structural Design - Notes

q is the shear flow on the section. If the shear resultants

are the sectorial products of inertia of the section, and

r(s) is the perpendicular distance from the origin to the

Torsion in Structural Design - Notes

4. Bending and torsion of open sections

The shear strain in the tangential plane is given by

From which the out of plane displacement is approximately

In the above eqn.

Torsion in Structural Design - Notes

has been used. The sectorial products of inertia I y , I z

Solving for the displacements,

Torsion in Structural Design - Notes

Torsion in Structural Design - Notes

Shear Stresses and Resultants

( xs ) SV = St Venant torsion stresses