Académique Documents
Professionnel Documents
Culture Documents
When a beam vibrates in flexure, it may have to be treated as a Timoshenko beam; that is
to say corrections may have to be made to the familiar Bernoulli-Euler theory to allow for
the effects of shear deformation and rotary inertia [l]. Unless the beam is very thin (so the
distance between adjacent nodal points is much greater than the depth of the beam) the
error incurred if the corrections are not made may be substantial-particularly
as regards
natural frequencies. The theory is, in particular, of great importance in the analysis of symmetric vibration of a ship hull.
The purpose of this paper is to show how the correction can also be made in a discussion of
coupled bending and torsion. While, to be sure, this objective is essentially that of filling a
comparatively minor gap in the existing theory of linear vibration, the matter is very far from
trivial. It is well known that the antisymmetric response of ships, to waves and to propeller
excitation, cannot be analysed with much confidence and, in that context, the theory to be
given is very much to the point.
2. EQUATIONS OF ANTISYMMETRIC MOTION
Figure 1 shows a slice of a beam of open section in which C is the centre of mass of the
slice and S is the shear centre. The shear force is represented by V while A4 is the bending
Y(x,<Ax
K(x,t)Ax
Figure 1. A slice of the beam showing the notation and convention employed. The shear centre is S, and C
is the ccntre of mass of the slice. Y represents shearing force and the bending moment is M.
469
470
moment. The axes OXYZ are stationary, the beam occupying the region 0 < x 6 I when it is
in its equilibrium condition with its plane of symmetry coincident with the plane OXZ. For
the sake of definiteness we shall assume that the ends of the beam are both free. Other ideal
end conditions may be used, as can readily be established.
The equation governing motion parallel to 0 Y is
p(x) &(x, t) = V(x, t) + Y(x, t),
where p(x) is the mass per unit length, u,(x, r) is the deflection of C and Y(x, t) is the applied
force per unit length. A prime signifies differentiation
with respect to x and an overdot means
differentiation
with respect to t. Since the deflection of S is
0,
where i(x) is the distance by which S is below C and 4(x, t) is the angle through
slice rotates, this equation may be written
which the
(1)
C is such that
(2)
Here Z=(x) is the moment of inertia per unit length and 0(x, t) is that contribution
to the slope
which is due to bending. Deflection of the beam in the 0 Y direction is due to bending and to
shear deformation.
It is such that
u(x, t) = 0(x, r) + Y(X,I),
(3)
beam.
We have
Mb, t) = Mx)
(4)
where kAG(x) is the shear rigidity and El(x) is the flexural rigidity, while a(x) and j?(x) are
distributions
of shear and bending damping, respectively.
In general, antisymmetric
bending is associated with twisting of the beam. The equations
governing the twisting motion are
Z,(x) 6;(x, t) - p(x)f(x)
where Z,(x) is the moment
(5)
axis through
C. The quantity
K(x, t)
(6)
In this expression, C(x) is the torsional stiffness, C,(x) is the torsional stiffness due to warping
of the cross-section (see reference [2]) and T(x) represents damping of torsional motion.
3. PRINCIPAL
471
OF A BEAM
MODES OF ANTISYMMETRIC
MOTION
0= e + y = e + V/kAc,
Z*B=M+
V,
M = EZe.
(We shall take up the rotation equations separately, later on.) Assume that motion in the
rth principal mode is such that
e(x, t) = e,(x) sin W,t,
V(x, t) = V,(x)sin 0, t.
(7)
u; = e, + V,/kAG,
(8)
v, = -44; - z, of e,,
(9)
Iu, = Eze;.
(10)
- z, 0f e,,
so
-PO:
0, + pa:
4r = -[(Eze;)
+ z, ~0: e,].
and, when the term on the right hand side is evaluated by parts, it gives
The integrated term is nil because the contents of the square brackets are -V, which vanishes
at the boundaries x = 0 and x = 1.The integral may again be evaluated by parts to give
and again the integrated term vanishes since the bracketed term is M, which is zero at the
two extremities. In other words,
-u:j~v,v.dx+~~~~~,v,bi=-~Eze~:~+~~~z~e~~;~.
0
0
0
0
This result must still hold if the sufficesr and s are interchanged, so that
(11)
472
are subtracted,
/H.J,v, dx - 05 /&,v,dx+
(w,z - o:,I
j EZ(B:v: - 0:v:)dx
0
I
--
of
J
0
*I,8,v:dx+w:j
I, 8, v:dx.
0
(13)
If equations
it is found that
v: = 0, - [((EZ&) + Z, w: B,}/kAG].
When this result is differentiated
with respect to x, it gives
I EZe: 8:dx
in this equation
is performed
+ z, 0: 8,
(Eze;) + z,of 8,
(zzze:y
,!%I; v; dx = j EZe: e: dx +
0
EZe: v: dx =
(15)
0
it
are interchanged
is found that
LAG
kAG
?^
0
(16)
IEz(e: v: -
e: f$) dx = w;
(14), multiply
i,
Interchanging
kAG
'(Ezex
kAG
dx,
(17)
kAG
s
0
it throughout
[zzeresdx-of
1
dx _ o2 4 wm)
I
1
O~jzze,v:dx=w:
0
4 wm
j
by 05 Z, 8, and integrate
z,8,dx
w2
OL
s
1: 6 6
I
kAG
with respect
dx.
(18)
r and s gives
1
0:
wze:)
wze;)
dx+ o2 z,e,(Eze:ydx,
EZe; 8: dx +
(Eze;y dx + w2
kAG
(Eze;)
dx,
1
kAG
term vanishes
Subtraction
kAG
1
j-
(Ezey
I[
EZO; 0; dx =
(14)
!zze,v:dx=w: 5zze,e,dx-a:
(19)
0
OF A BEAM
473
(EZ&)
8,Z,
dx
kAG
If equations (17) and (20) are now added together, they give
jEl(e:.:-s:.:)~+o:jz~s.jdx-o:jl,8,v:dr=(~~-~~)~z~B.B.dr.
(21)
(22)
0
We next go back to equations (5) and (6) governing rotation. For undamped free vibration
in the rth principal mode,
&l,
(23)
(24)
0
0:
I(~~~~,-~v*~~)dx=-iT~~,~.
0
When they are evaluated by parts the integrals on the right-hand side of this last equation
become
-ILAlb+j
G&b+
0
I~~&+~&dx.
0
integrated terms vanish because T, and T, are both zero when x = 0 and x = 1. Furthermore
The
~(T,9:-T,~;)br=jfC~:-(C,6:))~~dx-~(C~~-(C~~~)~~~
0
474
This is because c#$,45vanish at a free end of the beam since warping is unrestrained. It follows
that
(W:-w:)kl,$,8,dx_w:ipiv,~,dx+w:jpiv,b,dx=O.
0
(25)
0
bw
Us+
44~4,
+ h&f4
- P%$J,
+ Us
4r)ldx = ~shs,
(26)
where 6,, is the Kronecker delta function. Notice, however, that a special case arises when
o,=o=o
A seconiorthogonality condition can now be found. Let
yr = VJkAG =
U: - 8,.
It can now be shown by substituting back into equation (26) from equations (11) and (15),
and also from equation (24) noting that
that
I
I[EZe; e; + kAGy,
ys + T, &] dx = O:
(27)
Explicit expressions for the generalised mass and generalised stiffness corresponding to
pS can now be written down. If r = sin the orthogonality equations, these quantities are found
to be
and
c,, = co,a,, =
(EZei2 + kAGyf
+ T, c$:)dx,
(29)
respectively.
4. RIGID
BODY
MODES
Although they arise only from the particular choice of end conditions that we adopted,
the rigid body modes are of special interest. If the x co-ordinate of the centre of mass of the
beam is 2, they are
00(x) = 1,
ul(x) = l-G,
e,(x) = 0 =
y,(x) = c$~(x)
(for
r =
l),
42(x) =
(for
r =
2).
u2w=o=e2(4=Y2w,
(for r = 0),
(30)
475
All are scaled to have unit deflection at x = 0. The modal shaper&) does not satisfy equation
(8) but the reason is obvious and trivial--equation (8) effectively excludes deflection in just
such a mode as this and the exclusion can readily be remedied. The modes corresponding to
r = 0 and r = 1 satisfy the orthogonality condition (26) and are therefore mutually orthogonal.
But
I
[pzdx=ll20#0
rro2=-
d
I
5( 1
al2 = -
fl
l-:
&=a21#0
(31)
zi
In other words the Kronecker delta function in equation (26) must be ignored for these
combinations of the modal indices.
(32)
wherep,(r) is the rth principal co-ordinate. Equation (1) now becomes
If this equation is multiplied throughout by u, and then integrated over the length of the beam
it is found that
%) 0, dx -
,zo
j [kAG(y,p, + ay,d,)lusdr = j
0
YU,dx.
and this vanishes because the quantity in square brackets is the shearing force which vanishes
at the ends of the beam. That is to say
We now transfer our attention to equation (2) and treat it in a similar manner, this time
multiplying by 6,. Thus
476
By the same reasoning as before the integrated term can be rejected when the second integral
is evaluated by parts-the
bending moment is nil at the boundaries-and
this leaves
-$of4/
crkAGy,O,dx = 0.
(34)
- MaJ&dx
-%
- (C, 4:>pr +
r4:rirlAdx = -
I
-
/]C&P,
to give
Y dx +
2&
s
0
+S Kdx.
J
If the second integral is evaluated by parts, the integrated term vanishes since the term
[C&p, - (C,&)p, + f&J]
is equal to zero at both boundaries.
We are thus left with the
equation
- jZ&
Ydx+
Adding
equations
%b.j]
(35)
+,Kdx.
+ I,& 4s + I,& 6 - PM
now gives
a, + d,
v,>ldx +
$
k0
d,
bkAGy,(
us - e,) + pzze: e: + rg
$:I dx +
I[ Y(vs- 34 + @sl dx
0
. . That is to say
for s = 0, 1, 2, . . ., where
I
a,, =
akAGy, ysdx,
+ uZojj2 = Ydx
s
this gives
(for s = 0),
(for s = l),
(37)
477
C!OUF'LELlBENDINGANDTWISTINGOFABEAM
u&o
-t a&
+ a&,
= j (K - Z Y) dx
(for s = 2),
0
1
4d%+
r-3
c&s + B,,+rJI4+
c**p*= I 1 Y(v* - W
~44 dx
(38)
6. CONCLUSIONS
The simple Bernoulli-Euler theory of beam flexure may be adapted to the theory of coupled
bending and twisting vibrations. We have shown that, in this more complex motion, the
bending distortions may be those admitted in the theory of a Timoshenko beam. In practice,
the effect of this change is often fairly minor so far as the shapes of those principal modes of
lowest order are concerned. But the effect on the corresponding natural frequencies is often
substantial. (Notice that the Timoshenko beam necessarily has lower natural frequencies
than the corresponding Bernoulli-Euler beam since its bending stiffness is reduced by the
allowance for shear deflection and its inertia is increased by the rotary inertia.)
If the cross-section of the beam is box-like, the foregoing theory takes a degenerate form.
The points C and Sin Figure 1 then coincide, and so Z = 0. But if i = 0, equations (1) and (5)
are independent of each other. Variations of v&t), e&t) and r(x,t) are independent of
those of 4(x,t). Thus the last of equations (32) should now be replaced by some such
expression as
4(x, 0 = z 41(r) 4(x),
I-0
since equations (36) and (38) separate out into two distinct sets of equations.
At first sight it may seem surprising that, despite the added complications of shearing
deflection and rotary inertia, the theory produces exactly the same equations as those arrived
at by using the Bernoulli-Euler approach. Thus generalised masses, damping coefficients,
stiffnesses, displacements and forces at the principal co-ordinates all perform the same
functions as before (though of course the expressions for them are different). In fact, however,
it would be surprising if this were not so; for both theories conform to the general pattern of
linear vibration theory which may be developed by the use of Lagranges equation [3].
REFERENCES
1. S. P. TIMOSHENKO, D. H. YOUNG and W. WEAVER1974 Vibration Problems in Engineering. New
York: Wiley, fourth edition.
2. S. P. TIMOSHENKO1945 Journal of the Frunklin Institute 239,201-219,249-268,343-361.
Theory
of bending, torsion and buckling of thin-walled members of open cross-section.
3. R. E. D. BISHOPand D. C. JOHNSON1960 27re Mechanics of Vibration. Cambridge University
Press.