Chapter 4

In this chapter, we begin with a solid circular shaft under torsion. Then
we specialize it to a thin-walled tube. Arbitrary cross-section single cell and
multi-cell tubes will be considered after it. Finally, we conclude with opencell cross-sections.

4.1

Solid Circular Shaft

Let us recall the torsion of a solid circular shaft of radius c and length L.
We use the axis of the shaft as the z-axis. The basic assumption concerning
deformation is that a thin slice of the shaft perpendicular to its axis rotates
as a rigid body under the applied torque T . If the relative rotation of the
right end with respect to the left end is , a dashed line on the surface moves
to a solid line, as shown in Fig. 4.1, which is part of a helix. The surface
elements undergo a shear strain of
max = c/L.
This distortion is shown on a slice below the shaft in the gure. Interpolating
this into the interior using the rigid body assumption, the shear strain at a
radius r is
= r/L.
The quantity
/L
is called the rate of twist. Thus
= r,

= Gr,
1

CHAPTER 4. TORSION

111
000
111
000
111
000
111
000

1
0

L
11
00

,
Figure 4.1: Circular shaft under torsion.
where is the shear stress. A relation between the applied torque and the
rate of twist is obtained from
c

T =

Gr 3 dr = GJ,

r dA = 2
A

where
J=

c4
.
2

(4.1)

Tr
.
J

(4.2)

The stress distribution is given by

tau =

4.2

Thin-walled Circular Tube

Consider a tube of outer radius co and inner radius ci . The thickness of the
tube t is given by
t = co ci .
The mean radius c is dened as
c=

co + c + i
= co t/2 = ci + t/2.
2

4.2. THIN-WALLED CIRCULAR TUBE

By thin-walled we mean that t/c is negligible compared to unity. The eective

value of J for this case is

J = [c4 c4 ] = [(c + t/2)4 (c t/2)4 ].

o
i
2
2
Using the thin-wall assumption, we can expand the terms using the binomial
theorem to get
J=

c4
2

1+4

t
2c

43
2

t
2c

t
2c

+ 14

43
2

t
2c

This can be simplied by neglecting higher order terms in t/c to get

J = 2c3 t.

(4.3)

Fig. 4.2 shows the linear stress distribution across the thickness of the tube.
When the thickness is small compared to the radius, the variation of the shear
stress across the thickness is negligible and the mean shear stress corresponds
to the mean radius c. The torque created by the mean shear stress can be

Figure 4.2: Thin-walled circular tube under torsion.

found from

T
.
2c2 t
From this J comes out as shown in Eq. (4.3).
T = 2ct c,

(4.4)

4.3

CHAPTER 4. TORSION

Single cell tube

For a thin-walled closed tube of arbitrary shape the mean surface can be
described by two coordinates: z and s. The parameter s is measured from
a chosen starting point s = 0 and as s is increased it eventually reaches the
starting point when s = smax . It is the convention that the mean surface is
traced in the counter clock-wise sense. What follows is known as the BredtBatho theory of thin-walled tubes. We need an important assumption at
the outset as to the means of maintaining the shape of the cross-section.
This is called the CSRD assumption using initials of Closely Spaced Rigid
Diaphragms. We assume that rigid diaphragms (ribs) are inserted inside
the tube to maintain its cross-sectional shape. For an airplane wing, the
cross-sectional shape represents the selected airfoil and the lift and drag
calculations are based on maintaining its shape. Under the CSRD assumption
when a torque is applied the entire cross-section would rotate as a rigid
body. We may allow variable thickness as long as the thin-wall assumption

q + dq

s
q

dz

q
ds

s=0
Figure 4.3: Coordinate s and the force balance of an element of a thin-walled
arbitrary cross-section tube.
is satised. With variable thickness, a quantity called shear ow q is very
useful. This is dened as
q = t.

(4.5)

Dimensionally q is between a force and a stress, as it has the dimension of

force per unit length. As shown in Fig. ??, consider the equilibrium of a
small element of length dz and width ds. If q varies with s, one side we have

4.3. SINGLE CELL TUBE

a force of qdz and the other side (q + dq)dz, where
dq =

q
ds.
s

As there is no net force in the z-direction, q/s = 0. Thus, q is a constant

along s, even when the thickness changes. Where the thickness is small there
is higher shear stress and vice versa. This is analogous to a steady state
uid ow in a pipe, where the ow rate is constant. Fluid speed increases
when pipe cross-sectional area decreases. This ow analogy suggests the term
shear ow. Fig. 4.3 shows a small element ds with a shear ow q. The torque

1111111111
0000000000
1111111111
0000000000 ds
dA
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
O 0000000000
1111111111
0000000000
p
,

s=0

Figure 4.4: Evaluation of torque due to the shear ow.

produced by the force qds about an arbitrary point O is given by
dT = pqds,

(4.6)

where p is the perpendicular distance from O to the line of action of the

force. From the hatched wedge, it can be seen that the area of the wedge is
dA = pds/2.

(4.7)

dT = 2qdA.

(4.8)

Using this

CHAPTER 4. TORSION

As q is a constant, we can integrate this to get a relation between the torque

T and the shear ow q,
T
T = 2Aq, q =
.
(4.9)
2A
Remember: A is the area enclosed by the mid-surface curve and not the
material area. Then, shear stress at any point is given by
T
.
(4.10)
=
2At
To complete this theory, we would like to nd the rate of twist = /L in
terms of T . We use the complementary energy to do this. The energy stored
in an element ds along the length L of the tube is
dU =

q 2 Lds
T 2 ds
2 tLds
=
=
.
2G
2tG
8A2 tG

Integrating this, we nd
T 2L
ds
U =
,
8A2
tG
where the integral has to be taken around the tube. As this formula stands,
variable t and G can be allowed. Although it is common to nd variable t,
variable G is rare. Taking G outside of the integral and dierentiating with
T , we have
T
T L
, =
,
(4.11)
=
2G
4A
4A2 G
where
ds
=
.
(4.12)
t

4.3.1 Example: A rectangular tube

Consider a tube of rectangular cross-section with height 2a, width 4a, and
wall thickness t. For this A = 8a2 . Then the shear ow is given by
T
T
T
=
= 0.0625 2 .
q=
2
2A
16a
a
The integral
ds
12a
=
=
.
t
t
The rate of twist is obtained from
3T
T
=
.
=
2G
4A
64Ga3 t

4.4. MULTI-CELL TUBES

4.4

Multi-cell Tubes

Fig. 4.4 shows a sketch of a three-cell tube. The cells are numbered as
1, 2, 3, from left to right. We use the analogy of denoting electrical currents in a circuit to indicate the shear ows q1 , q2 , q3 , . For example, in
cell 2 the bottom and top walls have shear ows q2 , the left wall has (q2 q1 )
and the right wall (q2 q3 ). The net torque produced by all three shear ows
is given by
T = 2A1 q1 + 2A2 q2 + 2A3 q3 ,
(4.13)
where Ai is the enclosed area of the ith cell. This gives one equation involving

q1

q3

q2

Figure 4.5: Shear ows in a three-cell tube.

three unknowns: qi , i = 1, 2, 3. Compatibility requires that all three cells
rotate equally. We may nd the angle of rotation of a particular cell, say the
ith , using the complementary virtual work. Let us apply a virtual torque T
on this cell which produces a virtual shear ow q around the ith cell, with
q =

T
.
2Ai

(4.14)

Equating the complementary virtual work with the change in stored energy
we get
qq
T
q
T = L
tds = L
ds.
(4.15)
2
2Ai i Gt
i Gt
This can be written as
qds
1
.
(4.16)
=
2GAi i t
From this, noting that qi goes around the cell and qi1 opposes it on the left
wall and qi+1 on the right wall, we may expand Eq. (4.16) as
=

1
qi
2GAi

ds
qi1
t

Li

ds
qi+1
t

Ri

ds
,
t

(4.17)

CHAPTER 4. TORSION

where i has to be evaluated around the ith cell, Li and Ri are evaluated
over the left and right walls of the cell. We may use the notations
i =
i

ds
,
t

Li =
Li

ds
,
t

Ri =
Ri

ds
,
t

(4.18)

to write the equations for the rate of twist for the three cells,
1
[i qi Li qi1 Ri qi+1 ] = ,
2GAi

i = 1, 2, 3.

(4.19)

In the case of variable G, it has to be kept inside the integrals. Eqs. (4.13)
and (4.19) form four equations for fours unknowns qi and . Also, note that
the left-most cell has no left wall and the right-most cell has no right wall.
4.4.1 Example: Three-cell tube
Consider the three-cell tube shown in Fig. ??. All walls have thickness t and
modulus of rigidity G.

2a

2a

Figure 4.6: A three-cell tube with uniform thickness t.

We assume a torque T acts on it. Find shear ows in the walls and the
rate of twist.
The cell areas are
A1 = 2a2 ,

A2 = 2a2 ,

A3 = 4a2 .

With shear ows q1 , q2 , and q3 , the torque-shear ow relation is

2A1 q1 + 2A2 q2 + 2A3 q3 = T,

i.e. q1 + q2 + 2q3 =

T
.
4a2

(4.20)

4.5. OPEN THIN-WALLED TUBES

The rate of twist is given by
i qi Li qi1 Ri qi+1 = 2GAi ,

i = 1, 2, 3.

For this case

1 = 6a/t,
L1 = 0,

2 = 6a/t,

R1 = L2 = 2a/t,

3 = 8a/t,

R2 = L3 = 2a/t,

R3 = 0.

Thus
3q1 q2 = 2Gat,
q1 + 3q2 q3 = 2Gat,
q2 + 4q3 = 4Gat.
Solving the three equations we get
q1 = 1.1724Gat,

q2 = 1.5117Gat,

q3 = 1.3793Gat.

The rate twist is found from the torque relation, Eq. (4.20), as
3.3427Gat =

T
,
4a2

or = 0.0459

T
.
Ga3 t

Then

T
.
a2
We note that the walls separating the cells have much less amount of shear
ows compared to the top and bottom walls. The shear ows in the top and
bottom walls are comparable to the value we obtained for a single cell tube
of the same outer dimensions in the previous example.
(q1 , q2 , q3 ) = (0.0538, 0.0696, 0.0630)

4.5

Open thin-walled tubes

Open tubes are really not tubes, but thin sheets as shown in Fig. 4.5. When a
torque T is applied on this cross-section, shear stresses, , develop, creating
an equivalent torque. The material in the positive -direction has shear
stresses pointing in the negative s-direction and vice versa. The simplest
assumption we can make is that is linearly distributed as shown in the
sketch on the left, varying from 0 to a maximum of 0 . There is no shear
stress on the mid-surface, = 0. Eectively, we may think of this sheet as

10

CHAPTER 4. TORSION

a single-cell tube with a hole of zero area at the mid-surface. The sketch on
the right in Fig. 4.5 shows a shear ow q at a distance going around this

tube. Equating the force per unit length due to the shear stress to q, we get
1 t
(4.21)
q = 0 , or 0 = 4q/t.
2 2
Now, we have to decide the location, eta of the line of action of q. For this

we equate the moment produced by the linearly distributed shear stress to

that due to q. That is
2 t 0 t
1
t
= tq, or = .

32 4
3
3
The area enclosed by our ctitious tube is
q =

(4.22)

A = 2at/3.

1
0 s=a
1
0
1
0
0

(4.23)

11
00
11
00

s=a
q

11
00
11
00
11
00 s = 0

s=0

Figure 4.7: Open cross-section with assumed shear stress distribution.

This gives

T
3T
3T
=
, 0 = 2 .
2A
4at
at
To nd the rate of twist we use the energy method.
q=

U =

L
2G

(4.24)

t/2 2 dds.
0

t/2

(4.25)

4.5. OPEN THIN-WALLED TUBES

11

Here, from the linear stress distribution,

= 20 /t.

(4.26)

Then
2
4La 0 t/2 2
d,
G t2 0
4La 9T 2 t3
,
=
G a2 t4 24
3LT 2
=
.
2Gat3
Using dU /dT = and = /L, we get
T
,
(4.27)
=
GJ
where
at3
.
(4.28)
J=
3
As we may expect, the stiness of the open tube is many orders of magnitude smaller than that of a closed tube.

U =

4.5.1 Example: Circular tube

Consider a circular single-cell tube of radius a and thickness t. For an applied
torque T , the shear ow and shear stress are given by
T
T
, =
.
q=
2
2a
2a2 t
In case this tube develops a crack along its length, we have an open tube.
The maximum shear stress under the same applied torque is
3T
open =
,
2at2
which is (a/t) times larger than the value for the closed tube. The torsional
stiness for the uncracked tube is
GJ = 4GA2 / = 4G 2 a4 /(2a/t) = 2Ga3 t.
For the open tube
G
2at3 = 2Gat3 /3,
3
which is of the order of (t/a)2 smaller.
GJopen =

12

CHAPTER 4. TORSION

4.5.2 Warping of the cross-section

In the case of non-circular tubes, plane cross-sections before deformation do
not remain plane after. This phenomenon is called warping. We assume
we are dealing with a particular cell (say, ith ) of a multi-cell tube. In order
to describe warping, we introduce displacement components w along z and
u tangential to the mid-surface at s. We complete this set by adding a
coordinate and a displacement v perpendicular to the tangent as shown in
Fig. 4.5.2. The displacements are exaggerated in this gure for clarity. Our
shear stress , going around the cell, can be expressed as
= zs .
The corresponding shear strain has the form
zs =

w u
+
= /G.
s
z

(4.29)

The relative rigid body rotations of the tube take place around a xed point

v
111
000
111
000
111
000
rz

1111
0000
1111
0000
111
000
s
111
r 000
111
000
111
000
111
000
1
0
111
000
1
0
111
000
1
0

0000
1111
CT

s=0
Figure 4.8: Tangential and normal displacements around the tube.
called the center of twist, CT . At this stage of the development of our theory
we dont know how to nd this point. Consider a cross-section at z which
has undergone a rigid body rotation of z relative to the section at z = 0.

13

4.5. OPEN THIN-WALLED TUBES

Let a point at s be at a distance r from the center of twist CT . This point

will move perpendicular to the radius r by zr. This displacement can be
resolved to have the tangential component
u = zr cos = zp,

(4.30)

where p is the perpendicular distance from the center of twist to the tangent
at s as shown in Fig. 4.5.2. This gives
u
= p.
z

(4.31)

From the equation for the shear strain, this gives

w
q
=
p.
s
tG

(4.32)

Assuming G is constant, we integrate this expression form s = 0 to s, keeping

in mind, in the case of a multi-cell tube, q is a variable (piece-wise constant)
depending on s, to get
w(s) = w0 +

1
G

qds

pds,

(4.33)

where w0 is the axial displacement at our origin s = 0.

If we evaluate w(s) at the end point s = smax , we should get w0 . This
leads to
qds
1
=
,
(4.34)
2GAi i t
which is in agreement with the formula for the rate of twist, , derived earlier
in Eq. (4.16) using virtual work.
We may specialize the formula for the warp distribution in Eq. (4.33) for
a single-cell tube by noting q is a constant given by
q=

T
.
2A

(4.35)

Introducing the notations

s

s =
0

ds
,
t

As =
0

pds
,
2

(4.36)

14

CHAPTER 4. TORSION

we have the expression for warping displacement,

w(s) w0 =

T
[As As ] .
2A2 G

(4.37)

Clearly, the right hand side of this equation vanishes when s reaches smax .
The warping of the section represented by w(s) is called the primary warping
as in contrast to the warping perpendicular to the mid-surface, in the direction. With our assumption of thin-walled tubes the secondary warping
in the -direction is negligible.
s
Once again, it has to be kept in mind that the area, As = 0 pds/ uses
the perpendicular distance p from a special point, namely, the center of twist.
When a tube cross-section has two perpendicular axes of symmetry (bi-lateral
symmetry), we can argue that the center of twist lies at the intersection of
these axes. Further, the warping displacement has to be zero at points of
symmetry on the walls.
4.5.3 Example: Warping in a single-cell tube
Consider a single-cell symmetric tube of height 2a and width 2b with thicknesses ta for the vertical walls and tb for the horizontal walls, shown in Fig.
4.5.3. The planes AC and BD form symmetry planes and their intersection
P
A
111
000
111
000
111
s000
111
000
111
000

D
S
O
C

B
R

Figure 4.9: Warping of the walls in a rectangular tube.

O is the center of twist. The points A, B, C, and D are symmetry points
where the warp is zero. The area of the cell is 4ab. If we measure s from the

15

4.5. OPEN THIN-WALLED TUBES

symmetry point A, then, with wA = 0,
T
[As As ] ,
2A2 G
T
s
sa 4a 4b
=
4ab
+
2 b2 G
32a
tb
2 ta
tb
T
b
a
=
as.

2 b2 G t
16a
ta
b

w(s) =

Thus, the magnitude of the warp increases linearly from A to Q. At Q, we

have
b
a
T
.

wQ =
16abG tb ta
To compute the warp distribution from Q to B, we may use a new coordinate
s starting at Q. Then
T
s
sb 4a 4b
4ab
+
32a2 b2 G
ta
2 ta
tb
b
a
T
b(a s).

=
2 b2 G
16a
tb ta

w(s) = wQ +

At the point B, we get wB = 0. If we extend s to R, we nd

wR = wQ .
Thus, the warp distribution is anti-symmetric about the symmetry points
and the over all warp distribution is as shown by the dotted line in Fig.
4.5.3. If
a
b
= 0,
ta tb
we nd there is no warp along the walls. Tubes with zero warp are called
Neuber tubes.
From
T
[As As ] ,
w(s) w0 =
2A2 G
to have zero warp

s = As
A

16

CHAPTER 4. TORSION

A
B

111
000

Figure 4.10: Warping in a split circular tube.

This can be expressed as
s
0

ds

=
t
2A

pds.
0

Dierentiating, we nd

1
=
p,
t
2A

or pt = 2A/,

as a general condition for zero warp. Of course, in the case of variable G, it

has to be kept inside the integrals.
4.5.4 Warping in open tubes
The major dierence between open and closed tubes is that the shear stress
along the mid-surface is zero in the former and non-zero in the latter. In our
formula for warping displacement, Eq. (4.33), the rst term is due to the
shear stress. In the open sections we drop this term and what remains is
s

w(s) = w(0)

pds = w(0) As ,

(4.38)

4.5.5 Example: Split circular tube

Consider an open tube of radius a obtained by introducing a small cut between A and B, as shown in Fig. 4.5.5. This section has one plane of
symmetry and the center of twist lies on this. In fact, it is out side the tube

4.5. OPEN THIN-WALLED TUBES

17

at point O as we will see in the next chapter. If we are only interested in the
relative warp between A and B, we use the relation
2a

pds = 2A = 2a2 .

wB wA =
0

Our treatment of the tube is based on freely allowing warping of the walls.
If warping is prevented (say, by xing one end of the tube to a wall) normal
stress zz would develop from this strain. We will study this in detail later.

Exercises