Beruflich Dokumente
Kultur Dokumente
http://hdl.handle.net/2027/mdp.39015000992241
■':■■' -'i"--«iS'SW-■?.S'V
■ :> ^- .-^fe^g 1 1 ■ ^H
§111 -■
.ililiiili
V^xB'aSF ■.-'S-l
WBSS.BBSS"BS■.-
BBffifMi^mffm■MWMIflll
r£gmeFjz&.
+
+ + + +
W M
'*«AB»
+ +
H 1
Vgjp
+ + + +
+ +
W
M
***1U&&
+ + + ♦
**S
H M
Energy Methods
in Applied Mechanics
Energy
Methods
in Applied
Mechanics
Henry II Langhaar
Department of Theoretical
and Applied Mechanics
University of Illinois
New York
London
Engin. Library
77)
Copyright © 1962
L%r] by John Wiley & Sons, Inc.
Preface
BIBLIOGRAPHY 340
INDEX 347
Notations
Qv 62.
' ' '
, Qn components of generalized force [Eqs. (1-6) and
(1-8)]
Px, P2, • • •
. Pn components of generalized external force [Eqs. (1-7)
and (1-9)]
Rv R2,
• • •
, Rn components of nonconservative generalized force
[Eq.(7-ll)]
Q potential energy of external forces that act on a
mechanical system
V total potential energy of a mechanical system, V =
U + Q, where U is the potential energy of internal
forces
N tension in a bar, beam, or strut
E Young's modulus
v Poisson's ratio; also, any positive integer
G shear modulus
L length of a bar, beam, or strut; also, L = T— V,
the Lagrangian function
/ length of a bar, beam, or strut
A cross-sectional area of a bar, beam, or strut; also, A
denotes action
e extension of a bar; also volumetric strain [Eq.
(4-10)]
e' initial or residual value of extension
P magnitude of a force
F magnitude of a force; an integrand; Airy stress
function
h\, h2,---,hn arbitrary of the generalized coordinates
increments
X Lagrange multiplier; shear rigidity of a beam [Eq.
(2-13)] Lame constant for a Hookean material [Eq.
(4-47)]
a tensile stress
e tensile strain of a bar or a fiber; also an arbitrary
infinitesimal
M bending moment
/ moment of inertia of the cross section of a beam
t shearing stress
5 shearing force on the cross section of a beam
k shape factor in formula for strain energy of shear
[Eq. (2-10)]
/S slope of a beam due to shear deformation
6 temperature; an angle or an angular coordinate
GJ torsional stiffness
NOTATIONS xiii
p mass density
a radius of a circle; a length
z ordinate in a cross section of a curved beam, shell, or
plate
q displacement vector of a deformable body
u, v, w components of the displacement vector of a deform
able body in the x, y, z directions
u, v, w values of u, v, w at the middle surface or centroidal
axis of a shell, plate, or curved beam
Z Winkler constant in the theory of curved beams
[Eq. (2-22)]
of
<f>
a
Euler angle; function symbol; in theory of strain
a
= + $e2; in theory of shells of revolution
<f>
<f>
is
e
'
axis of symmetry
K stiffness factor of beam (K = 2EI/L); also Gaus
a
sian curvature.
erj(x) variation of function.
a
ex, ev, ez, yyz, y^., yxy strain components [Eqs. (4-3) and (4-28)]
m, n direction cosines
/,
I2,
13
0\, ff2,
• • •
at stress [Eq. (4-36)] also, ev e2, e3 and av a2, a3
,
[Eq.(5-8)]
XIV NOTATIONS
ex, ey, e^, values of ex, e„, yxy on the middle surface of a plate or
shell
Um strain energy due to stretching of the middle surface
of a plate or shell
Ub strain energy due to bending of a plate or shell
Ue strain energy due to heating of a plate or shell
h thickness of a plate or shell; depth of web of a beam
p a load parameter; lateral load per unit area of a
plate or shell
p„ of load parameter
critical value
D flexural rigidity of a plate or shell [Eq. (5-10)]
Nx, Ny, Nxy, Nyx intensities of normal and shearing tractive forces in a
plate or shell [Eqs. (5-11) and (5-59)]
Qx, Qy intensities of transverse shearing forces in a plate or
shell [Eqs. (5-22) and (5-59)]
Mx, My, M^, Myx intensities of bending moments and twisting mo
ments in a plate or shell [Eqs. (5-22) and (5-59)]
A, B Lame coefficients for the middle surface of a shell or
plate [Eqs. (5-48) and (5-49)]
n unit normal vector of a surface
e,f,g coefficients of the second fundamental form of a
surface [Eq. (5-51)]
rv r2 principal radii of curvature of
a surface [Eq. (5-53)]
P (Px, Py, Pz) intensity of external force acting on a shell (Sec. 5-7)
=
R = (Rx, Ry, Rz) intensity of external couple acting on a shell (Sec.
5-7)
kx, Ky, kxu incremental curvatures and twist due to bending of a
shell [Eqs. (5-69) and (5-71)]
r radius of cross section of of revolution;
a surface
radial coordinate for polar coordinates ; rank of a
matrix
H Hamiltonian function
p\i p2,' ' '
,pn components of generalized momentum
I, rj, £ body-centered rectangular coordinates (Sec. 7-8)
<f>,
co1, co2,
rigid body
C
a
1 General concepts and
principles of mechanics
If any system of bodies or particles, whatsoever, each being acted upon
by arbitrary forces, is in equilibrium, and if one gives to this system a
small arbitrary movement, by virtue of which each particle traverses an
infinitely small space, which will determine its virtual velocity, then the
sum of the forces, each multiplied by the distance through which the
corresponding particle moves in the direction of the same force, will
always equal zero, where we regard as positive the small distances
traveled in the directions of the forces, and as negative the distances
traveled in the opposite directions.
J. L. LAGRANGE
incompressible fluid are such that the volume of any part of the fluid
remains constant. The constraints of an ideal cantilever beam are such
that the displacement vector vanishes at the clamped end.
- =
(a) |X0
(b)
Xol
If X, =£ Xo,
0.
|XX
- Xo|
= |Xo
- XJ > 0.
If Xo, Xx, — —
-
(c) X2 are any three points, |X0 Xx| + |XX X2| >
|Xo X2|.
Condition (c) is known as the "triangle law," since it is analogous to the
Euclidean theorem that the sum of two sides of a triangle is no less than
the third side.
A "neighborhood" of a point X0 in configuration space is defined as the
* A number c is said to be an upper bound of a set S of real numbers if there is no
number in S which is greater than c. If a number c with this property exists, set S is said
to be bounded above. If c is an upper bound of S, so is any number greater than c. It
is shown in real variable theory (36) that the set of all upper bounds of a set S always
contains a minimum number b, called the "least upper bound" of S. The number b is
the maximum number in S if b belongs to S. However, not every set of real numbers
that is bounded above contains a maximum. For example, let S be the set of all real
numbers less than 3. Then, if x is any given number in S, there are infinitely many
numbers in S that are greater than x; hence S contains no maximum. In this case the
least upper bound of 5 is 3, but 3 is not a number belonging to S.
Similarly, if a set S of real numbers is bounded below, it possesses a "greatest lower
bound" which coincides with the minimum number in 5, whenever a minimum exists.
GENERAL CONCEPTS AND PRINCIPLES OF MECHANICS 5
single configura
a
t
tion X. This relationship indicated by the conventional function nota
is
tion, = X(t), a <, <, For example, may denote time. Then the
X
b.
/
t
relation = X(t) represents the motion of the system. The function X(t)
X
then continuous, since the system does not leap finite distance in zero
is
a
time. With the preceding definition of distance in configuration space,
continuity of the function X(/) defined as in real variable theory; that is,
is
the function X(f) continuous at the point X0 = X(/0) to each positive
if
is
d
e
e
for all in the interval — < The function said to be continuous
d.
is
t0\
\t
t
-
= = —
-
at the end point < for all in the interval
if
(or |X
b)
a
X0|
e
/
/
t
b
6
t
If
the coordinates of particle in Euclidean space are continuous
a
a
t,
is
is
a
<,
b
a
t
in configuration space. The points X(a) and X(b) are called the "end
points" of the curve. The curve said to have no double points (in other
is
words, the curve does not cross itself) any pair of distinct values of in
if
t
the interval a < <b (say and correspond to distinct points Xx
t2)
tx
t
and X2. In real variable theory continuous curve of this type called
is
a
called "path."
is
space
a
Nonholonomic Systems. A
system may be constrained to follow certain
paths in configuration space, even though the space connected. In other
is
words, although the system can pass from any point to any other point in
its configuration may be constrained to follow routes that coincide
it
space,
with certain dense network of paths. These paths may be regarded
a
as tracks that guide the system. If such constraints exist, the system
is
ible."
The word "holonomic" was introduced by H. Hertz, (33), who stated:
"A material
system, between whose possible positions all conceivable con
tinuous motions are also possible motions, called holonomic system."
is
a
6 ENERGY METHODS IN APPLIED MECHANICS
z).
Suppose
there mechanism that constrains the particle so that any infinitesimal
is
a
displacement (dx, dy, dz) conforms to the relationship
dx — zdy =
0
(a)
it
lead to finite equation among the coordinates, which would imply that
a
a
interpretation incorrect, since Eq. (a) does not possess an integrating
is
factor. Any point in space to the particle, even though Eq. (a)
accessible
satisfied. In other words, is
(x0, y0, z0) and (xv yv Zj) are any two points,
if
is
the particle can travel from (x0, y0, z0) to (xj, yv z^ by paths that satisfy
Eq. (a).* Since the particle not free to follow every path between these
is
is
to be nonintegrable). A similar two-dimensional example does not exist,
since linear differential expression in two variables always possesses an
a
a
plane. There are two nonintegrable differential equations which specify
that any infinitesimal displacement effected without slipping (see Prob.
is
a
car wheel, whose configuration determined completely by the angle
is
systems
laws are phrased to apply to nonholonomic cases unless the contrary
is
If x0
= Xj and y„ = y,, the particle may follow the path x = x0, = y0. If yn # yu
*
let follow the line x = x„, = y„ from (x„, y0»zo) to (x0, y0, z'), where z' = (x, — x0)/
it
— yol; then let follow straight line from (x„, y0, z') to (x,, yu z'). Finally, let
it
it
(Hi
a
follow the path x = xu = y, from (x,, yu z") to (x,, yu z,). In the special case for
y
which yj = y„ the preceding method may be used to transfer the particle from (x„, j/0, z„)
to (x,, yz, Zj), where (x,, yz, z,) any point such that y, ^ y0. Then,
by
the particle may be transferred from (x,, yif z,) to (x,, yu z,). Thus, in all cases the
particle may be transferred from (x0, y0, z„) to (x,, yu z,) paths that satisfy the equation
by
the bead lie at a point at which the tangent to the curve is horizontal. If f
were not required to be a regular coordinate, we could set t = z, provided
that there is a single value of z for each point on the curve, and vice versa.
However, if
/ = z, the equilibrium points are no longer determined by
dzjdt = 0, since dzjdt is now identically equal to 1 . The trouble is that
is,
dsjdt becomes infinite at the equilibrium points; that the coordinate
= irregular at the points where equilibrium occurs.
is
z
t
n
a
ates (xx, x2, • • •
xn), the number of degrees of freedom defined as
is
,
is
n
it
space,
a
z
A rolling sphere has five-dimensional configuration space, but has
0.
it
a
will not serve this purpose, the system said to have infinitely many
is
degrees of freedom.
a
a
Newton's laws of mechanics are valid* only for reference frames that do
not rotate with respect to the remote stars and that do not accelerate with
respect to the center of mass of the solar system. Such reference frames
are said to be "Newtonian" (or "Galilean" or "inertial"). The earth
is
not exactly Newtonian reference frame, but, for most engineering pur
a
poses, the rotation and the acceleration of the earth are negligible. Unless
the contrary specified, the reference frame under consideration New
is
is
tonian.
Let position vector of particle with respect to
be the Newtonian
a
a
r
is
Newton's laws are exactly valid. However, for phenomena in which the particles are
large compared to the elementary particles of physics and for which the speeds are
small compared to the speed of light, Newton's laws are close approximation to
a
reality.
GENERAL CONCEPTS AND PRINCIPLES OF MECHANICS 9
Accordingly, the work that force F(/) performs on the particle during
the time interval (t0, Q is
-f
W=\ F-vdt (1-1)
where the sum extends over all particles on which the force acts. Let the
displacement Ar, occur in a time interval A^. Then Eq. (b) may be written
as
w=2^-^^h
(<0
10 ENERGY METHODS IN APPLIED MECHANICS
Now ArJA^ is the mean velocity vt of the ith particle during the interval
Af<. Hence Eq. (c) yields
w-2¥riM (d)
If the increments A/< approach zero, the sum approaches a limiting value
which is represented by an integral. Accordingly, Eq. (1-1) remains valid
for a force that is applied in any continuous way to a mechanical system.
When Eq. (1-1) is applied to a mechanical system, y must be interpreted
as the velocity of the particle on which force F acts at time t; it must not be
regarded as the velocity of the point of action of the force. For example,
a rigid ball that rolls on a fixed rigid body receives no work from the
reactive force of the body, since the velocity of the particle of the ball that
is in contact with the body is zero. Likewise, Eq. (1-1) shows that if a
brick slides on a pavement the friction of the pavement performs work
on the brick, although the friction of the brick performs no work on the
pavement. These examples show that Eq. (1-1) is more general than the
equation W = J• F • dr; the latter equation applies only if the force F acts
continuously on the same particle. This fact was emphasized by Osgood
(62).
Newton's third law asserts that if body A exerts a force F on body B
then body B exerts the force — F on body A. This law signifies that forces
may be mated, "action" and "reaction." The reaction of a given force F
is understood to act on the body that causes or exerts force F. a force F If
acts on a mechanical system, its reaction — F acts on another part of the
same system, or it acts on a body outside the system. In the first case, it is
called an "internal force;" in the second case, it is called an "external
force." Accordingly, all the forces that act on a mechanical system may be
classified as internal or external. Hence the work W of all the forces that
•
act on a mechanical system is separated into a sum, W = We + Wt,
where We is the work of the external forces and Wt is the work of the
internal forces.
dW = mv • d\ = — (imv2) dt (e)
dx
Fig. 1-2
rests on the table. The virtual displacements are required to be small, since
otherwise we could admit a virtual displacement for which the brick slides
to the edge of the table and falls off. Then the virtual work of all the forces
(friction and gravity) might be positive.
The work W that all the forces perform when a mechanical system
experiences a virtual displacement from configuration X$ to configuration
X generally depends on the path in configuration space that connects
points Xq and X. Unless the system is provided with an unlimited source
of energy, the set of all values of W corresponding to all paths from X0 to
X is bounded above; hence it possesses a least upper bound W (see
footnote, p. 4). In other words, W <, W for any admissible path from
Xq to X, and there is no number less than W with this property. Since W
depends only on the terminal configurations X0 and X, and not on an
intermediate path, it is designated as W(Xo, X). This means that W is a
function of X0 and X. The function W does not necessarily exist if the
system is infinite. Although all our experience is limited to finite systems,
there are important problems of elasticity, plasticity, and fluid mechanics
concerning media that are conceived to extend to infinity. To analyze
an infinite system, we may consider finite free bodies cut from it.
If W < 0 for every admissible path that leads from X0 to X, W(Xq, X) <,
0, and vice versa. Consequently, Fourier's inequality may be expressed as
follows :
The point Xq in configuration space represents an equilibrium state if
WiXf,, X) < Ofor all points X in a neighborhood o/Xq. i~tA«> ^"
We may specify that virtual displacements are performed adiabatically.
This means, for example, that if a brick on a table receives a horizontal
virtual displacement the heat generated by friction does not pass into the
brick. For an adiabatic virtual displacement, the heat flux Q is zero.
Consequently, by Eq. (1-4), Wt' = —At/, where At/ is the increment of
internal energy. Therefore, since W = W( + WJ, Fourier's inequality
may be expressed by the relation, WJ < At/. Accordingly, equilibrium
exists if there is no small adiabatic virtual displacement for which the virtual
work of the external forces exceeds the increment of internal energy.
\x\
lim W/s = > Nevertheless, by Newton's law, the point x = an
0.
is
F0
0
equilibrium configuration for the particle.
W = dW + - d2W
+
0(s3) (1-5)
quantity less in absolute value than Ks3, where K positive constant and
is
a
If the system has only finite number of degrees of freedom, the resolution
a
d2
According to Eq. (1-5), the equilibrium criterion lim Wjs <, repre
is
0
is
0
d
of expression of the principle of virtual work, but the relation lim Wis <,
0
more general, since W cannot always be resolved into the form of Eq.
is
P
is
(1-5).
a
is
fi
cient of friction, and (dx, dy) are the virtual increments of rectangular
16 ENERGY METHODS IN APPLIED MECHANICS
is,
the limit of the sum of the amounts of work performed on the individual
particles equals the sum of the limits.
The condition lim Wjs = necessary and sufficient for equilibrium of
is
0
x2,
a
a
,
is
represented by
+
dxn)
a
that
;
6W = 0i + • •
+ Qn
+
where • •
are certain functions of xv • •
The
•
Qv
•
customary notation dxt used instead of dxt to denote that the displace
is
+ +
is
Qn dxn
Fy
when the particle on which acts undergoes the displacement (dx, dy, dz),
it
is
but this terminology may cause confusion with the thermodynamic meaning of
reversibility.
GENERAL CONCEPTS AND PRINCIPLES OF MECHANICS 17
+ + • -- + = 0
fa dxx <ff>2dx2 <f>ndxn
We
- dWt +
1 d2We + 0(s3)
in dxv • • •
is,
- + P2 + •
+
Pn
•
where Pv P2,
•
Pn are certain functions of •
In general,
•
•
•
•
• • •
the expression dxx + not an exact differential.
Px
+ + Pn
P2
is
dx2 dxn
By an obvious extension of the terminology introduced previously, the
is
18 ENERGY METHODS IN APPLIED MECHANICS
functions Pv P2,
• • •
, Pn are called "components of generalized external
force." If the variables
xt denote lengths and angles, the variables Pt may
usually be identified as the components of prescribed external forces and
couples that act on the system.
Even though the loads on a structure or mechanism are constants, the
generalized external force components Pt need not be constants. For
example, consider the system shown in Fig. 1-3. Let 6 be the generalized
coordinate, and let the force F be constant. If 0 receives an increment d6,
Fig. 1-3
the first variation of the work of the force F is= FL cos 6 d0 = P d6,
dWe
where P is the generalized external force. Hence P = FL cos 6. Accord-
lingy, P depends on 6, even though Fis constant. On the other hand, if the
vertical displacement x is taken as the generalized coordinate, the virtual
work of the load F is F dx = P dx, whence, F = P. Thus it is seen that
the question of constancy of the generalized force depends not only on
the nature of the loads but also on the choice of the coordinates.
where L is the length of the path. Consequently, for given terminal points,
W attains a maximum when the object describes a straight path; in this
20 ENERGY METHODS IN APPLIED MECHANICS
AK = wx Azx + w2 Az2 + • • •
where w is the weight of the entire system and z is the elevation of the center
of gravity of the system. The last relation follows from the definition of
the center of gravity. Consequently, V = wz. This equation means that
if the acceleration of gravity is constant the part of the potential energy of a
system that results from the gravitational field of the earth is equal to the
weight of the system, multiplied by the elevation of its center of gravity.
Since the zero configuration of potential energy is irrelevant, the datum
plane from which elevations are measured need not be specified.
Conservative External and Internal Forces. The work W that all the
forces perform when a system experiences a virtual displacement from
configuration X0 to configuration X may be separated into a sum, W =
WJ + W{, where WJ is the work performed by the external forces and
W- is the work performed by the internal forces. It may happen that We'
GENERAL CONCEPTS AND PRINCIPLES OF MECHANICS 21
Also, — dV is the work performed by all the forces when the system
receives an infinitesimal displacement (dxv dx2,
• • •
, dxn). Therefore, if
22 ENERGY METHODS IN APPLIED MECHANICS
e"-r ox,
(1-8)
dV = — dx1 + — dx2 + • • •
+ — dxn
ox1 ox2 oxn
U=jNde=j^de
Accordingly,
EAe2
U = (1-10)
2L
GENERAL CONCEPTS AND PRINCIPLES OF MECHANICS 23
dQ = - \p + -C de
ef.
(a
Integration yields
Q(e) =
-pe - C
a — e
AV = dV + -
2!
d2V + 0(s3) (1-12)
in which 6V and d2V are the first and second variations discussed in Art.
1-6.
24 ENERGY METHODS IN APPLIED MECHANICS
If
V exists, the principle of virtual work (lim AV/s = 0) is equivalent
d
P,-^ axt
(.-.3)
U = iiiaiixixj (1-14)
I aaXj = Pt (1-15)
the potential energy of the two particles due to their mutual attraction is
•=/,
dr
F
(1-16)
is
a
bonds in system of attracting particles equals the number of distinct pairs
a
n
— bonds.
«(« l)/2
Let the bonds in system of particles be numbered • • •
m. If we
2,
1,
a
,
give the system an infinitesimal displacement, the increment of potential
energy (i.e. the work that, we perform) dV = ^Ft drt, where the sum
extends from to m. Here, is
the tension in the zth bond and the
is
Ft
is
rt
1
is
a
tial energies of all the bonds. This conclusion signifies that system of
a
attracting particles conservative, irrespective of the law of attraction.
is
a
movable particle. The force that acts on the movable particle depends
F
Fz
z)
dx
=
dy
=
dz „ j_
rx r r
F
F
y
curves, called "lines of force." The lines of force are everywhere tangent
to the vectors of the force field. The theory of integration of equations
of the form of Eq. (1-17) discussed in books on differential equations (35).
is
a
a
generated by fixed particles the sum of the potential energies of the bonds
is
between the movable particle and the fixed particles. An additive constant
that represents the potential energy of the bonds among the fixed particles
26 ENERGY METHODS IN APPLIED MECHANICS
is,
dV= — dx + — dy + — dz (a)
o.r dy 0Z
F,
is
dV = -Fx dx - Fydy-
Fz
dz (b)
Since Eqs. (a) and (b) are valid for any infinitesimal displacement,
ox ay dz
or
= -grad (l-18a)
V
F
A
force field that possesses potential energy function said to be
is
a
a
determined completely by its potential energy function. The lines of
is
V
Equation (1-18) special case of Eq. (1-8).
is
a
is
= k^
F
(1-19)
r2
is
a
mass. The same relationship applies for the force between electric charges
(Coulomb's law).
Let the units be such that = Then, by Eq. (1-16), the potential
1.
k
energy of two particles that attract each other in conformity with the
Newtonian law
is
V=-^ (1-20)
r
mn
,
V=-fi^i
i=i
(1-21)
rt
where the distance between the ith fixed mass and the movable mass.
is
ri
'--f*^ d-22)
Fig. 1-5
r2 = a2 + x2 — 2ax cos 6
2r dr = 2ax sin 6 dd
dv=_Z>rpadr
x
Therefore,
X Jx-tt X
center of the
body, since the solid sphere may be decomposed into homo
geneous spherical shells of infinitesimal thickness, and the preceding rule
applies for each of these shells.
capsize yet
it
a
a
state of stable equilibrium, yet jolt may cause to "snap through" into
it
a
a
badly deformed shape. Such illustrations indicate that the infinitesimal
theory of stability must be used with discretion. Nevertheless, the infinitesi
mal theory has had innumerable practical applications.
In the following the word "stability" used with the classical meaning;
is
a
in state of unstable equilibrium an arbitrarily small amount of energy,
if
is
supplied from external sources, can cause the system to experience large a
displacement. However, this not true. For example, table that rests on
is
cally we can move the table as far as we please without any expenditure of
work. To do this, we lift the table slightly so that the reaction of the floor
to act, and we then displace horizontally. Certainly, external
if
ceases
it
is
;
than the theory for nonconservative systems. In view of the law of con
servation of mechanical energy, = constant, an infinitesimal
T
V
+
is
30 ENERGY METHODS IN APPLIED MECHANICS
of
brium the value the potential
is
if,
if,
energy
a
minimum.
This conclusion it is known "law of minimum potential energy." In
as the
a
qualitative way indicates the meaning of weak stability. The stability of
a
Fig. 1-6
system
is
a
the given equilibrium state with only small increase of potential energy.
a
is
a
a
a
it
a
if,
a
a
bars with hinged ends. The bars assume positions such that the elevation
of the center of gravity of the system relative minimum. The catenary
is
a
(curve
a
case of this type of linkage. Among all curves of given length and given
end points, the catenary has the lowest centroid. With the aid of the cal
culus of variations, the equation of the catenary may be derived from this
principle (Sec. 3-7). A stationary cup of water also illustrates the fore
going principle. The surface of the water plane and horizontal, since
is
GENERAL CONCEPTS AND PRINCIPLES OF MECHANICS 31
(1-25)
stands collectively for all the x's; for example, V^(x) denotes d2V/dxt dxt
at the point (xv x2, • • • , xn). For brevity, Eq. (1-25) is written
AV = dV + i&V + R3 (1-26)
Since the derivatives of V with respect to the x's are continuous in R,
the principle of virtual work signifies that a necessary and sufficient
condition for equilibrium is that d V vanish for all A<; that dV/dxt =
is,
0,
= • • •
n. We consider point x/ that solution of these equa
2,
1,
is
i•
a
,
tions.
The second variation of the potential energy may be written
(1-27)
Accordingly, d2V quadratic form in the variables hv A2, • • •
is
a
hn.
,
Although R3 not exactly cubic form in the A's, differs from cubic
it
is
quently,
a
Fig. 1-7
Example. A
thin hemispherical shell of constant thickness is balanced on
a hemispherical dome of the same radius as the shell. The stability of the
configuration is to be investigated. If the shell rolls slightly, the configura
tion in Fig. 1-7. The center of gravity of the shell lies at the
is as illustrated
midpoint of the radius. Consequently, by Fig. 1-7, the elevation of the
b J is
center of gravity fl , .
y = 2r cos 0 — \r cos 20
^
=
if,
if,
MK-) n4
y
Ay = 6y +
1
±- 62y +
+
•
+
d3y <34y
^
^
we obtain
= = = = -6r04
0,
PROBLEMS
1. Prove the triangle law for displacements in configuration space.
2. A wheel of diameter D rolls on a straight track. Calculate the magnitude
of the displacement in configuration space corresponding to an angular
displacement 8 of the wheel.
3. A gas flows steadily through a tube of circular cross section. The velocity is
distributed parabolically on a cross section, and the velocity at the center is
three times the velocity at the wall. Express the rate of displacement in
configuration space dsjdt in terms of the mean velocity V.
4. A frictionless incompressible fluid flows at constant mass rate through a
conduit of variable cross section. Consider the fluid between two cross
sections, 1 and 2. By applying the law of kinetic energy to this fluid, derive
Bernoulli's equation for liquids.
5. A body that slides on a straight horizontal bar oscillates under the action of
a restoring force that is proportional to the displacement. Assuming that
the coefficient of friction is a constant, show by the law of kinetic energy
that the difference between the amplitudes of successive oscillations is a
constant.
7.
z
is
is
opposite
a
proportional to the speed. Calculate the work that the resisting force per
forms during the interval = to =
3.
1
t
/
that directed toward the origin. Prove that the work the force
is
always
performs on the particle during any displacement the product of the
is
magnitude of the force and the reduction of the distance from the origin.
9. Two mass particles are connected rigid massless rod. Prove
by
by
direct
a
application of Newton's laws that the tension in the rod performs no net
work on the particles when the system moves freely in any way.
A particle in plane moves in the force field Fx = —ky, Fy = kx,
y)
Fig. Pl-11
12. The thrust force F has constant magnitude, and it is always directed along
the axis of the lower bar. The hinges are frictionless. Prove that the system
is nonconservative (Fig. Pl-12).
Fig. Pl-12
GENERAL CONCEPTS AND PRINCIPLES OF MECHANICS 35
U = \El[Lym*dx,
Fig. Pl-19
potential energy of the beam. Derive the relation between the bending
moment and the curvature by the principle of stationary potential energy.
21 A homogeneous solid rests on a plane inclined at angle a to
hemispherical
the horizontal. The body may roll, but it cannot slide. Determine the
equilibrium configuration (angle of roll 0 relative to the plane) by the
principle of stationary potential energy.
22. A spherical soap bubble contains a constant mass of air. The pressure-
volume relation of the air is pv1* = constant. The potential energy of the
air is — J/» dv. The potential energy of the soap film is 2kS, where k is the
surface tension and S is the area of the soap film. Express the potential
energy of the system in terms of the radius r of the bubble. Hence derive a
formula for the radius of the bubble by means of the principle of stationary
potential energy.
23. A soap film spans a circular hole of radius a in a plate. Internal pressure p
causes the film to adopt the form of a segment of a sphere. The strain energy
of the soap film is 2kS, where k is the surface tension = constant) and
(A:
S
the area of the soap film.
By geometry, = na\l + x2) where x = tan
\i
S
is
(Fig. PI -23). Also, geometry, the volume of the bubble above plane
by
O-O =
frcPix3 + 3x). Adopt x as generalized coordinate. Regarding
is
Q
^W^ ^o
I
r?:
Fig. Pl-23
GENERAL CONCEPTS AND PRINCIPLES OF MECHANICS 37
point of a straight line perpendicular to the plane of the disk through its
center and at a distance x from the center is —2np(Va2 + x2 — x), where p
is the mass per unit area. If a -* °o , what is the force on a unit mass at point
xl
26. A wire that is bent into a circle carries a constant density of electric charge.
Derive the potential at any point on the axis of symmetry that is perpendi
cular to the plane of the circle. Hence, compute the force that acts on a unit
charge that lies on the axis.
27. Calculate the Newtonian potential of a thin straight uniform rod of length 2L
at any point outside of the rod. Calculate the attraction of the rod on a point
mass that lies in the plane perpendicular to the rod at its center. Calculate
the limiting value of this attraction as the length of the rod becomes infinite.
Take origin at center.
28. A homogeneous body consists of a cone ofheight h attached to a hemi
spherical base with radius r (Fig. Pl-28). The body rests on a table in an
upright position. By the principle of minimum potential energy, determine
the range of h/r for which this configuration is stable.
29. Prove by the principle of minimum potential energy that among all attitudes
of a floating body for which the weight of displaced liquid equals the weight
38 ENERGY METHODS IN APPLIED MECHANICS
of the body the stable equilibrium configurations are those for whichZr, — Zcb
is a relative minimum, where Ze, is the height of the center of gravity of the
body, and Zcb is the height of the centroid of the submerged part of the body
(called the "center of buoyancy"). Let Z„ > Zcb. Hint. Suppose that the
liquid is contained in a tank. Note that the attitude of the body affects the
potential energy of the liquid.
30. A long homogeneous timber with a rectangular cross section of width b and
depth a floats in water. The specific gravity of the timber is \. Show that if
b\a > V3/2 the timber floats with its wide face horizontal and that if
1 < b\a < V3/2 it does not float with its face horizontal. Determine the
angle of the wide face to the horizontal if bja = V4/3. How does it float if
b\a = 1 ? Hint. Use the theorem stated in Prob. 29.
31. A long homogeneous timber of rectangular cross section floats in water with
its top face horizontal. Prove that a necessary and sufficient condition for the
timber to be in stable equilibrium is s2 — s + A2/6a2 > 0, where s is the
specific gravity of the wood, b is the width of the cross section, and a is the
depth of the cross section. Hint. Use the theorem stated in Prob. 29.
2 Elastic beams and frames
The methods that I expound require neither constructions nor geometrical
nor mechanical reasoning, but only algebraic operations, subject to an
exact and invariable procedure.
J. L. LAGRANGE.
The principle of virtual work finds its most important engineering appli
cations in the statical problems of deformable bodies. Some illustrative
applications to deformable systems with enumerable degrees of freedom
are presented in this chapter.
JA I
dA = 0, = 0, fy dA =
JAVdA
0 (a)
\
J\A
where A denotes the cross-sectional area. The bending moment M, being
represented by a vector that lies in the plane of the cross section, may be
resolved into components Mf and Mn on the f- and ?y-axes. It is assumed
that the fiber stress at the point (f ,
rj)
dA, N =
jff
a,
a-dM + Me +
Z
(2-D
A
/,
I;
39
40 ENERGY METHODS IN APPLIED MECHANICS
Here, N denotesthe net tension, and (/{, Q are the moments of inertia of
the cross section about the |- and r?-axes, respectively.
By Eq. (1-10) and Hooke's law, a = Ee, the strain energy of an element
of the beam of length ds and cross-sectional area dA is a2 dA ds\2E.
Consequently, the strain energy of a beam due to bending and direct
tension is
Fig. 2-1
U = —\ [^L+^L
2E Jo \ J,
+ —)ds
AI
(2-3)
Is
Equation (2-3) shows thatif a beam is bent simultaneously in both of its
planes the contributions to the strain energy from these two
principal
components of bending are additive. Also, Eq. (2-3) shows that the strain
energy due to direct tension may be added to the strain energy of bending.
In view of these conclusions, we consider only pure bending in one
principal plane ; that is, we set N = 0 and = 0. Also, for simplicity,
we write Ms = M and If = I. Then Eq. (2-3) yields
Jo
^2EI
(2-4)
It is to of strain-energy components
be noted that the superposition
implied by Eq. (2-3) is a special circumstance; in general, the effects of
several loads on the strain energy of a structure are not simply additive.
For example, if an elastic bar of length L and cross-sectional area A is
ELASTIC BEAMS AND FRAMES 41
M = — (2-5)
R
Equations (2-4) and (2-5) yield
I'll
Jo
If primes denote derivatives with respect to arc length s, Eq. (2-6) may be
written
EI(y"f dx (2-8)
Jo
Strain Energy Due to Shear. Besides the strain energy of bending, there
is strain energy due to shearing stresses on the cross-sectional planes. The
shearing stress distribution on a cross-sectional plane is approximated by
the elementary formula of beam theory
r = W (2-9)
lb
the volume of the beam, we obtain the following formula for the strain
energy due to shearing forces :
*-n <S2dx
2GA
(2-10)
Q'dV
(2-11)
I2J
For a rectangular cross section, k = 1.20. Except for short beams or
beams with thin^ebs, the energy of shear is usually small compared to the
energy of bending.
If shear is significant, the strain energy of bending is not proportional
to the square of the curvature, since the curvature is augmented by the
shear deformation. Then Eq. (2-8) is inapplicable. The deflection separ
ates into the sum of the displacements caused by shear and by bending.
The slope y', being the derivative of the deflection, likewise separates into
a sum of two terms — slope due to shear and slope due to bending. The
slope due to shear will be denoted by Since the stress-strain relation
is
/?.
L
U8 = PSdx
r
i
Jo
With Eq. (2-10), this yields
fi-£
AG
(2-12)
is
strut may be regarded as beam. Evidently, Eq. (2-12) does not apply for
a
plates or the skin, and very little transmitted to the stringers. Also,
is
slight buckling of the plates of ship or the spar webs of an airplane wing
a
/3
= ty
S
(2-13)
a
a
A
Us = $ )W2 dx (2-14)
Jo
The slope caused by bending is y' — where the total deflection.
ft,
is
y
±.M/EI = —
y"
Consequently, the curvature due to bending ft'. There
is
fore, the strain energy of bending
is
Jo 2EI Jo
With Eq. (2-14), this yields the following formula for the total strain
energy:
= fi[£/(y" - + dx
U
ft')2 (2-15)
i
Jo
In the statical problems of beams the functions and must be deter
ft
mined to minimize the total potential energy among functions that
satisfy the end conditions and the continuity requirements. Since the
slope at clamped end of beam results entirely from shear deformation,
a
y'
the
ft.
y
0
a
a
= = =
y"
is
S
y
are generally
is
applied,
a
/S
— —
y"
ft'
/S
y'"
shear, discontinuous.
is
negligible,
is
ft
Eq. (2-8).
the
L
is
a
the formula
(2-16)
G
is
/
a
for hollow shaft of annular cross section, the polar moment of inertia
/is
a
of the cross section about its center. For any other shape of cross section,
/
less than the polar moment of inertia. Formulas for are developed in
/
is
Fig. 2-2
mrx
y = 2K
l
K=
sin
L
(a)
ELASTIC BEAMS AND FRAMES 45
y' =
Li
2nbncos
—
xj
,
"
y = - -£n 2bn sin —
Xj
(b)
X-i
The strain energy of bending Eq. (2-8). Squaring the series for
is given by
y", we obtain a double series. However, the integrals of the cross products
cancel because of the relations
cl cl (0 if m^n
. rrnrx . nirx . mnx nnx . '
-
I |
I sin sin ax = I cos cos ax =
Jo L L Jo L L ,f m = n
where m and n are any positive integers. Consequently, Eqs. (b) and (2-8)
yield
'
41? £i " V
It is assumed that the beam bends without any change of the length 11
the centroidal axis. Since dx2 + dy2 = d£u dxjdx = \/\ — {rlyjHi)2 Ex
panding the square root by the binomial series and retaining only the first
two terms, we obtain dxjds = 1 — $(dy/ds)2. The distance between the
ends of the bent beam is
[Ld.x ds
Jo ds
QP--Z(L(gYdx (2-19)
2 Jo
where y denotes dyjdx. Substituting Eq. (b) into Eq. (2-19) and utilizing
Eq. (c), we obtain
= -^
4L
I "V n=l
(d)
= ~QiVi - Q^i
46 ENERGY METHODS IN APPLIED MECHANICS
O0 = -CilMin^-
n-l L,
Q2 |bnsin^?
11=1 A,
- --- (e)
Q„ =
- I W dx
Jo
Hence, by Eq. (a),
n<
= -lK \\
n=l Jq
sin
^
L
dx (f)
CL
q sin (nirx/L) dx + Q1 sin (nirc^L) + Q2 sin (nircJL) +
• • •
n~~
(tt4£//2L3)n4
- (ir*P/2L)n2
lg
This is the Euler column formula. Accordingly, Eq. (g) may be written
as follows:
b" =
2J3 J 9 sin (mrx/L) dx + Q1 sin
-
(nircJL) + Q2 sin (nircJL) +
•--
VeI n4 (Pn2/Pe)
(2-21)
Substituting Eq. (2-21) into Eq. (a), we obtain the sine series for the deflec
tion curve.
ELASTIC BEAMS AND FRAMES 47
Fig. 2-3
is the reaction of the center support. The other Q's are zero. Equation
(2-21) accordingly yields
(i)
yields approaches
(i)
1,
r
= (2/ir)qL. Equation applies also for negative values of P; that is,
(i)
R
R
if
yields
a (i)
r
0.516qL. These results indicate that the axial load has only small effect
on the center reaction.
a
that symmetrical about the plane of its centroidal axis considered
is
is
deformed in its plane. The following notations are used (see Fig. 2-4):
Fig. 2-4
(Fig. 2-4)
the moment of inertia of cross section of the ring about an axis
/
Jf-*L.
a + z
4(1 + 2),
J\l*±--ZA,
a + z
f^* = ZAa
Ja (2-22)
a + z
Z^—a
Aa2
(2-23)
u = u, v=
o -zu' + II
= —zu' (l +-)v
+ -It! (a)
=
u + av'
e (b)
a + z
., u — azu , .
e = t> H (c)
a + z
The net tension in the beam is N= E$e dA. With Eqs. (2-22) and (c),
this yields
FA
N = — [(« + a&) + Z(u + aV)] (2-24)
a
The bending moment M about the centroidal axis of the cross section is
defined as positive if it tends to reduce the radius of curvature. Then,
M = EjzedA. With Eqs. (2-22) and (c), this yields
If
hoop stress predominates, we obtain the strain energy of the beam
by integrating the strain energies of the fibers. A volume element of the
beam is (a + z) dA dd. Consequently, if the material obeys Hooke's law,
the strain energy is
U = dA
lE^ddjia + z)e2 (2-26)
(125).
Equations (2-24), (2-25), and (2-27) yield
u = — +
2
J
n =1
an cos vn6,
oo
v = 2 bn sin vnd (d)
n =1
Therefore,
u + av =
^
2
+
n.l
| - vVK
(l
cos vnd
u + a& = &
2
+ |
n=l
(an +
j (e)
Fig. 2-5
COS
ttEA
u = 'i±sh (i + ZW + ^ 2 [K
ttEA
+ vnin)2 + Z(vv - i)V] (f)
4a 2a n =i
„= vP vP £ -ji
cos vnd
N ~
1 r (0
(j)
applied are obtained from Eqs. and set equal to zero. The
if
(i)
is
6
resulting series may be evaluated by means of the following series for the
cotangent (45):
-
00
2z
v cot itZ = 1
+ - (k)
^
n =1 —
n
Z
- =i--cot-
(1)
f
n .i rV 2v
2
v
1
(j)
yield
- cot^- a Pa\ -
P
ki — ^n]
v
N0 = M„ = cot (m)
,
rJ
+ Z)
2
Ltt(1
2
v
With Eq. (m) the tension and the bending moment at any section may be
obtained by statics.
hinged joints or ball-and-socket joints are then justified. Such joints are
designated as pin joints. Since the members of pin-jointed truss are
a
system
a
is
algebraic equations are also nonlinear, and they are usually difficult to
ELASTIC BEAMS AND FRAMES 53
solve. Even in cases of linear elasticity, the algebraic equations often are
tedious to solve, but they may be handled readily by network or digital
computers. Direct or iterative numerical methods may also be used with
desk calculators (15, 23, 47).
If the displacementsof the joints are so small that the members are not
rotated appreciably, the extension of a member is approximately the
difference of the axial components of the displacements at its ends. If the
displacements of the joints are large, they still determine the extensions of
the members,
deflections
but the geometrical
of trusses rarely occur in practice.
relations
a £T 1,[_
The strain energy of a uniform tensile member witn modulus of elasticity
r /
become nonlinear. Large
FA
N = —(e' + e) (2-29)
U = — (e' + ef (2-30)
2L
Pentagonal Truss. As an example, a regular pentagon-shaped truss with
no residual stresses is considered to be loaded symmetrically by a vertical
force F, as shown by Fig. 2-6. All members have the same cross-sectional
area A. The length of an outer member is L. By geometry, the length of
any cross brace is 1.61804L. The stress-strain relation of the material is
linear. The displacement components of the joints are x, y, z, w (Fig. 2-6).
Resolving the displacements into axial components, we obtain the
following formulas for the extensions of the members :
ex
= O.80902y - 0.58779z + 0.80902w
e2
= -0.95106x + 0.30902^
e3
= -0.58779x + 0.80902y + 0.58779z
et
= 0.309022/
- 0.95106z - 0.30902w,
(a)
e5
= 2w
e6
= 2y
54 ENERGY METHODS IN APPLIED MECHANICS
Fig. 2-6
0.58779e2 + 0.58779e3
= -—
2EA
-0.36327^ + 0.58779e3
- 0.95106e4
= 0
(b)
0.50000^ + 0.19098e2
- 0.30902e4 + e5
= 0
ELASTIC BEAMS AND FRAMES 55
2EA
0.4755x - 2.3906y + 0.1123z - 0.3090w = 0
The symmetry of these equations follows from the fact that the strain
energy U is a quadratic form in the generalized coordinates x, y, z, w.
The solution of Eq. (c) is
FT
— Fl
—
x = 0.7025 , y = 0.1438
EA EA
(d)
z = 0.1769 —
FT
, w = 0.0325 —
FT
EA EA
e1
= 0.0387
FT
— , e2= -0.6581
—
FT
, e3
= -0.1926
FT
—
EA EA EA
FT FT FT
e. = -0.1338 — , e5
= 0.0650 — , e6
= 0.2877 —
EA EA EA
JVi = 0.024F, N2
= -0.407F, N3 = -0.193F
Simple Space Truss with Residual Tensions. The lower ends of the
bars of the four-bar, pin-jointed truss shown in Fig. 2-7 are fixed and the
upper joint is free. When P = 0, the residual tension in member 1 is N1'.
When the load P is applied, the tensions in the bars are Nv N2, N3 = A^4.
Bars 3 and 4 are identical.
Positive senses are assigned to the bars, as indicated by the arrowheads.
56 ENERGY METHODS IN APPLIED MECHANICS
The following table gives the cross-sectional areas, the lengths, and the
direction cosines of the members. The lengths and the direction cosines
have been computed by means of the designated coordinates of the ends of
the bars.
The displacement vector of the upper joint due to the load P is q =
hi + jv + kw. By symmetry, u = 0. The extension et of any member that
0,-20,20
et
= u cos a, + v cos pt + w cos yt
Therefore,
ex
= -L (-3» + 2w), e2
= i(-3v + 4w), (e)
= =
5V6
e3 e4 (—v + w)
The initial extensions of the members due to the residual tensions are
= e4'. By Eq. (2-30), the total potential energy of the truss
e3'
e{, e2',
is
ELASTIC BEAMS AND FRAMES 57
i=i Li ov
(f)
<=i Lf ow E
The derivatives dejdv and dejdw are constants ; they are determined by
Eq. (e).
TABLE 2-1
61 (g)
E
EA1
The residual extensions and = condition
0. e2'
e3'
0.
0
v
e1
Hence, by Eq. (f),
di£i §£• = ^— =
*
|-
(h)
0
dv i=i Li dw
Lt
•=1
Since known and e3 = e4', Eq. (h) determines e2' and e3'. Introducing
is
the derivatives dejdv and dejdw from Eq. (e), and numerical values of At
and Lf from Table 2-1, we obtain
25
= — " = e,- =
648
,
e9 e.i"
e
61
1
'
52 325
0.031876t> - 0.033476w = 0
--
EE
-0.033476U + 0.037742w =
E
The solution of these equations is
= -406- = -387-
EE
(j)
v , w
(j)
- - -
P
E P
= 123 = —65.6 = = 13.3 (k)
e2
e1 e3 e4
,
,
1.22AV
These equations may be checked by the condition that the upper joint
must be in equilibrium under the action of the load and the tensions in
P
the members.
(Fig. 2-8). The deflected beam has the general form indicated by the dashed
curve. The equation of the deflection curve = y(x). Hence
is
y
2/(0) (a)
where (yv y2) are the transverse displacements of the ends, and (01, 02) are
the angular displacements of the ends. Primes denote derivatives with
respect to x. The rotation of straight line through the ends
is
a
= y*~LJb
4>
(2-31)
L
Supposing that single lateral load acts on the beam at the point
F
a
0
ELASTIC BEAMS AND FRAMES 59
Integration yields
2M^x _S^f FH(x - af (c)
KL KL KL
-L-
Si
Mi
C
j n
T3*
i
Fig. 2-8
where K=
2EI/L. The constant K is called the "stiffness factor" of the
beam. Integration of Eq. (c) yields
y-y1 + o1x+
kl 3Kl+ (d)
Equations (a), (c), and (d) yield two linear equations that determine
Ml and Sv Then M2 and S2 may be determined by statics. Thus there
results
-Kiie, + 02
- 3<f>) +
^f-
- -
(2-32)
-Kid, + 262 3<f>)
^
Si
L L3
(2-33)
The lengths a and b that appear in these equations are designated in Fig.
2-8. Equations (2-32) are fundamental in the slope-deflection theory of
60 ENERGY METHODS IN APPLIED MECHANICS
is,
at the ends ends were clamped (that
if
dl
y2 0).
02
y1
For any type of loading, these terms are replaced by the appropriate
fixed-end moments.
of Eqs. (2-32) and (2-33),
by
Eliminating M1 and from Eq. (d) means
S1
we obtain
- - a
2
= + 61x- (20! + + (0, + + constant (2-34)
y
02
y1
02
3<f>)- 2<f>)—
I?
L
The additive constant represents the part of the deflection that does not
depend on yv y2,dvd2; may be called the "fixed-end deflection,"
it
since the deflection that would exist the ends of the beam were
if
is
it
clamped and the lateral load were applied. Insofar as the potential energy
of the external forces concerned, the fixed-end deflection contributes
is
Equations (2-8), (b), (2-32), and (2-33) yield the following formula for
the strain energy of bending:
= K[62 + + - + + + constant
U
is
irrelevant, since our purpose Then the shears
02.
to determine yv y2, 6v
is
and moments at the ends are determined by Eqs. (2-32) and (2-33).
If the beam (Fig. 2-8) part of structure, the variables yv y2, 6v
is
02
a
(2-35),
is
a
i<W,
l if
=
U
(2-36)
= =
l
i
are the equations of equilibrium. For plane frames, these equations are
the same as those provided by the slope-deflection theory (63). After the
x's are determined, the bending moments and the shears may be computed
by the slope-deflection equations [Eqs. (2-32) and (2-33)].
To systematize
the procedure, convenient to set up the matrix
it
is
10-6C/12
= Xj* + x1x2 + x22
—
3xs(x1 + x2) + 3x52
The potential energy of the load Fis Q = — Fya, where ya is the deflec
tion at the point of application of the load. Referring to Eq. (2-34), we
have for the horizontal member y1 = 0, = = 100, x = 25, —
<f>
0,
L
x2,
61
0,
Ps
In forming
the matrix (aw), we must observe the factor in Eq. (2-36).
J
Accordingly, the terms on the principal diagonal of the matrix are twice
the sums of the corresponding coefficients from Eq. (e). However, the
other terms are the sums of the corresponding coefficients from Eq. (e),
since in Eq. (2-36) each nonsquared product of the x's occurs twice; for
example, the terms containing x1 and x2 are \a12x1x2 + ^a^x^.
The accompanying symmetric matrix (Table 2-2) of coefficients
is
coefficients au and the components are reduced by the factor 10-6; this
procedure does not alter the solution of the equations. Also, since the
F,
TABLE 2-2
xi xt x» x* x5 P
», 2 1 0 0 -3 0
1 6 2 0 -3 3
XZ 0 2 6 1 -3 -1
Xi 0 0 1 2 -3 0
-3 -3 -3 -3 12 0
3xs
= 3, etc. The solution of these equations is
= -5/42, x2
= 31/42, x3
= -17/42, x4
= 19/42,
^5
= 1/6
25" -75"
200"
The five equations obtained from the respective rows of the preceding
matrix may be interpreted as follows : first equation — no bending moment
at joint 1. Second equation — net moment applied to joint 2 by the
members is zero. Third equation — net moment applied to joint 3 by the
members is zero. Fourth equation — no bending moment at joint 4. Fifth
equation — sum of shears in the vertical members is zero. These interpre
tations follow immediately from Eqs. (2-32) and (2-33). In the slope-
deflection method the preceding equilibrium conditions are used to set up
the equations that determine the x's . Although the foregoing results apply
only for F=
(16/75) X 106 lb, the values corresponding to any value of F
may be obtained by direct proportioning.
Space Frame. The space frame shown in Fig. 2-10 consists of three
members respectively parallel to the coordinate axes. The ends 0 and 3 are
clamped. At joints 1 and 2 the members are welded together. A tri
angularly distributed load is applied to member 12. The cross sections are
symmetrical, so that EI has a constant value for all axes through the cen-
troid of a cross section. Also, EI has the same value for all members. The
torsional stiffnessof a member is GJ = (3/4)£Y. This relation is typical for
solid or hollow prismatic bars. Setting £//50 = C, we obtain K01 = K12 =
C, Kw = 2C, (GJ/L)01 = (GJ/L)12 = 0.375C, (GJ/L)^ = 0.75C.
The displacement vectors of joints 1 and 2 are denoted by q1 and q2.
The (x, y, z) projections of these vectors are (qlx, qlv, qlz) and (q^, q2v, q2z).
Since the members are practically inextensional, qlz = 0, q2x = 0, and
qlv = q2y. of joints 1 and 2 are small, they may be
Because the rotations
represented by vectors 8! and 82, in accordance with the right-hand-screw
rule. The following generalized coordinates are introduced :
l00xj 100x2
—
Ilv — ?2v, 100x3
= qtz
= olx, = 0lv X6
= 012, = o2x
x8
= 02*. ** = 02,
w = x$ — 0.01
(2x4 + x7
— 3x3)y2 + 0.0001 (xt + x7
- 2x^y* + constant.
J o
pw dy
64 ENERGY METHODS IN APPLIED MECHANICS
1000
= -1500p0) P4 = -500ft,, =
P3 P7 ^p0
The other P's are zero.
The strain energies of the three members are
U0l
= C(x* - 3xjzs + 3^) + C(x42 + 3a;2x4 + 3x22) + 0.1875Cx62
= C[x£ + + x* — + +
V - Ix&s - xsf
U12 xtx7 3x3(a:4 x7) 3x32]
xl xi x3 xt XS x7 xs x9 PJC
12 0 0 0 -3 -3 0 0 -3 0
0 54 0 3 0 0 0 0 -12 0
0 0 54 -3 0 0 -3 12 0 -ISOOpJC
0 3 -3 4 0 0 1 0 0 -SOOpJC
X6 -3 0 0 0 2.375 0 0 -0.375 0 0
xt -3 0 0 0 0 2.375 0 0 1 0
x1 0 0 -3 1 0 0 2.75 0 0 333.33/>0/C
x6 0 0 12 0 -0.375 0 0 4.375 0 0
x9 -3 -12 0 0 0 1 0 0 6 0
— — — — 0
12x! 3x9
-
3x5 3x6
54a;2 + 12xe
= 0
-
3x4
54x3
- 3x4 3x7 + 12x8
= -1500/>0/C, etc.
ELASTIC BEAMS AND FRAMES 65
The solution of these equations may be obtained easily with a desk com
puter, since there are numerous zeros in the matrix. Using an elimination
scheme (23), we obtain
xi = -252.2PolC, xb
= Ul.SpjC, x6 45.49/>0/C
x7
= 105.6/>0/C, xs
= 281.2/>0/C, x9
= 98.96/>0/C
U= ^JaiFiXj
where the x's denote displacements or rotations of the joints. Usually, the
external forces are conservative. Then the components of generalized
external force are Pt = — dCl/dx^ where Q is the potential energy of the
external forces. The principle of stationary potential energy yields
n
2 auxi = pi
PROBLEMS
1. For analytical purposes, a uniform beam of length L and stiffness EI is
approximated by a chain of n equal rigid links with linear springs in the
hinges. The spring constant for a hinge is k. The hinges are frictionless.
Express k in terms of EI, L, and n, so that the strain energies due to constant
bending moments M are equal for the beam and the linkage. How do the
relative angular displacements of the ends compare for the two systems?
2. An airplane wing is regarded as a cantilever beam of length L. The stiffness
of the wing is assumed to be given by EI = a0 — a^, where x is the distance
of a cross section of the wing from the root and (a0, a^) are constants. The
66 ENERGY METHODS IN APPLIED MECHANICS
air load per unit length is approximated by p = be — b^x, where (b0, bj are
constants. The deflection is approximated by y = Ax* + Bx3, where
(A, B) are constants. Derive the expression for the total potential energy of
the wing. Hence derive linear algebraic equations that determine the
coefficients (A, B) by the principle of stationary potential energy.
3. Set a = aS + br\ + c, where (a, b, c) are constants, and derive Eq. (2-1) by
the conditions of statics :
Mn
=
-Jff| dA, M( = fa>j dA, N =
$odA
Generalize Eq. (2-1) for the case in which the origin is at the centroid of the
cross section but Iin ^ 0. Generalize Eq. (2-3) for this case.
4. Calculate the factor k in Eq. (2-11) for an I-beam without fillets, having the
following dimensions: flange width, 5 in., flange thickness, 1 in., over-all
depth, 10 in., web thickness, 0.25 in.
5. A simple beam carries a linearly distributed load that varies from zero at one
end to/>0 at the other end. Calculate the deflection function in the form of a
sine series.
6. A simple beam carries a concentrated load at the center. If the beam is
subjected to a tensile force equal to three times the Euler critical load, by
what percentage is the deflection at the center reduced?
7. Set PjPe = r, and derive a formula for RjFior the beam shown in Fig. P2-7.
Fig. P2-7
equilibrium, derive the coefficients in the series for the displacement com
ponents (u, v). Suppose that the reacting shear q acts at the centroidal axis
of the ring (Fig. P2- 11).
*-F
Fig. P2-11
12. Derive the formula for the section constant Z of a curved beam with an
elliptical cross section. Show that Z is independent of the width of the cross
section.
13. Let a = Ee, and write Eq. (c) of Sec. 2-3 in the form a = (a + /?z)/(a + z).
Express the constants (a, in terms of the bending moment M and the net
/?)
J•cr
M =
\az dA. Thus derive the Winkler formula which expresses a in terms
of M, N, a, A, Z.
z,
14. The tension N in any bar of the semicircular frame N = Ve, where
is
is
k
a
constant and the extension of the bar. The joints are hinged. By the
is
e
is
wires, as shown in Fig. P2-16. For each wire, E = 107 lb/in.2 and = 0.04
A
17. Beam is rigid and uniform. It weighs 3001b. When it is lowered into
position in a horizontal attitude, the wires become taut simultaneously
(Fig. P2-17).
r
r
H
.1
H
600#
TH
J.
-L/2- -L/4-
Fig. P2-17
ELASTIC BEAMS AND FRAMES 69
30',
100#
fF
Fig. P2-18 Fig. P2-19
20. The three wires have the same EA. Other members of the frame are rigid
(Fig. P2-20).
10,000#
Fig. P2-20
22. The four wires have the same EA. If the slab is lowered into place in a
horizontal attitude, the wires become taut simultaneously. The slab weighs
4000 lb, and a 2000-lb concentrated load is applied on the diagonal, 5 ft from
the center, as indicated by Fig. P2-22. Determine the tensions in the wires.
Hint. Let the deflection of the slab be w = ax + by + c and adopt (a, b, c)
as generalized coordinates.
70 ENERGY METHODS IN APPLIED MECHANICS
Fig. P2-22
23. The frame is a regular with cross braces on the center square.
octahedron
All members have the same EA. Compute the tensions in the members by
the principle of stationary potential energy (Fig. P2-23).
24. Solve Prob. 20, assuming that the three wires have the same cross-sectional
of a wire is a = kVe.
area and that the stress-strain relation
25. By means of the principle of stationary potential energy, derive the tensions in
the members of the wall bracket, assuming that the displacement of the joint
is small and that the stress-strain relationvf the material is unknown. Show
that the results agree with those obtained by balancing forces at the right-
hand joint (Fig. P2-25).
ELASTIC BEAMS AND FRAMES 71
Fig. P2-25
(for the cross section of any member). Compute the rotations of joints 1 and
2 in degrees.
33. Each member of the lambda frame has length L, cross-sectional area A,
modulus of elasticity E, and moment of inertia /. As a result of the external
moment M applied at the vertex, the joint experiences displacement com
ponents (m, v) and a rotation d. Taking the extensions of the members into
account, derive formulas for («, v) by the principle of stationary potential
energy (Fig. P2-33).
100# 2 5
"*T
60"
10 10
100"
0 15
Fig. P2-29
ZK 2 2JC
h4 3
K K K
P2-30
Fig. P2-31
72
ELASTIC BEAMS AND FRAMES 73
34. The bent is hinged at its feet. The diagonal tie has no flexural stiffness.
Derive formulas for the rotation of each joint and for the displacement of
the top member by the principle of stationary potential energy (Fig. P2-34).
35. All members of the frame have the same EI. The members are clamped at
the feet. Calculate the horizontal reactions at the supports. (Fig. P2-35).
36. The stiffness factors of the vertical and horizontal members of the two-story
bent are respectively Kv and Kh. The diagonal tie has cross-sectional area A
and modulus of elasticity E; it has no flexural stiffness. Derive the equations
that determine the translations and rotations of the joints by the principle of
stationary potential energy (Fig. P2-36).
74 ENERGY METHODS IN APPLIED MECHANICS
Fig. P2-37
37. The bent is loaded normal to its plane by a force Fat one corner. Represent
the rotations of the joints by vectors with components parallel to the members.
The torsional stiffness of any member is GJ. The stiffness factors of the
vertical and horizontal members are, respectively, Kv and Kh. By the
principle of stationary potential energy, derive the equations that determine
the rotations and translations of the joints (Fig. P2-37).
a
11
MAX BORN
ml**
"*■ l
p A/CTN
rx*
bF
Fig. 3-1
+eEI \y""ri dx
Jo
^EljVrjdx (f)
In view of Eq. (f), the following conditions are necessary* and sufficient
for dV to vanish for all admissible functions rj(x):
y»»(x) = 0 (g)
The equations in (h) are the natural boundary conditions of the problem.
They may be identified as the moment condition and the shear condition
for the free end. Equation (g) is known as the Euler equation of the
problem.
The general solution of Eq. (g) is y = A0 + A1x + A^c2 + A3a?, in
which the A's are constants. The forced boundary conditions yield
A0 = A1
= 0. The natural boundary conditions yield
A* = A3 =
2EI' 6Fl
* The for these conditions is proved in Sec. 3-2.
necessity
78 ENERGY METHODS IN APPLIED MECHANICS
- pJql - *)
Hence
y 0)
6EI
The equilibrium configuration represented by Eq. stable, since PVis
(i)
is
always positive in this example.
*
:
function y(x) that minimizes definite integral,
a
(3-1)
is
a
2
n
a
F
with all its partial derivatives to order n + for all real values of y, y',
1
y", y{n) and for all values of x in the interval x0 <. x < xv The ad
•
•
•
,
is
V
a
n
the interval x0 < x < xv In some cases these limitations are too restrictive.
For example, the admissible values of might be restricted to certain
y
a
then the properties of the function for values outside of that
F
range;
range are irrelevant. Also, the curve = y(x) that minimizes
V
might
y
example,
is
x0.
conditions into account. Although the discussion limited to regular types
is
+ + a$" =
+
a2y' any<n-1)
0
a0 a-M
h
-\
• •
in which are constants.
By requiring the forced boundary
•
a0, av
conditions to be linear, we ensure that an admissible variation
is
r](x)
e
«
calculus of variations that are important for physical applications presented in the
is
treatise
METHODS OF THE CALCULUS OF VARIATIONS 79
Ja
»Vi + wf dx
Without undertaking an analysis of the problem, we can perceive that it
has no solution. Among continuous curves that which consists of seg
ments of the two straight lines y = 0 and x = 1 provides an absolute
minimum to the moment of inertia. However, this curve does not belong
to class r, since it has no continuous tangent. Nevertheless, the value J,
which is the moment of inertia of this curve about the x-axis, is evidently
the greatest lower bound of the values of the moment of inertia for curves
of class I\ There is no curve belonging to class V that attains this lower
bound. The problem has a solution in class T that is expressible by
elliptic functions if other end points are chosen for the curve ; for example,
if the end points are (±1, 5).
Without verifying the existence of the solution of a variational problem,
we may investigate necessary conditions for a minimum. Before con
sidering the absolute minimum B, we must consider relative minima. The
curve C:y = y(x) is said to provide a "relative minimum" to V if there
is a strip of width 2p (bounded by curves with ordinates y + p and
y — p), in which every other curve C of class V conforms to the inequality
V ^ K(Fig. 3-2). In other words, if curve C furnishes a relative minimum
to V among curves of class V, there is a positive number p such that
V ;> V for all curves C of class V in the p-strip, \y — y\ < p. More
specifically, these conditions are said to define an "improper relative
80 ENERGY METHODS IN APPLIED MECHANICS
minimum." IfP>
K (that =
is,
excluded), the minimum
if
F
the case
K
is
said to be "proper."
is Suppose that the curve C:y = y(x) provides relative minimum to
V,
a
of highest order occurring in
y"
and suppose that the derivative the
is
of Eq. (3-1). Let comparison curve with ordinate
C
integrand be
a
+ rj(x), in which r\(x) any admissible variation and an arbitrary
is
is
y
e
y
*0 *1
Fig. 3-2
proper,
is
e. Denoting V(C) by
y"
+ + y, y', y")] dx
+
•'Xo
Here the symbol 0(e3) denotes quantity that less in absolute value than
is
a
is
a
The terms <5Kand d2Vare called the first and second variations of V; they
represent, respectively, the first and second integrals in Eq. (a).
If not zero, controls the sign of AK when sufficiently small.
V
is
is
it
d
a
d
METHODS OF THE CALCULUS OF VARIATIONS 81
(3-2)
(3-3)
This condition follows from the fact that may be chosen so that the
rj
rj'
rj
(x0, a^). Equation (3-3) must be satisfied for all values of the constants
^(x0), rjix^), rj'(x0), rj'(x^) that satisfy the forced boundary conditions. In
general, these conditions impose certain restrictions on the function at the
y
end points. These restrictions are known as the "natural boundary
conditions;" they are necessary conditions in order that shall take
V
a
minimum value.
Equations (3-2) and (3-3) yield
(3-4)
by
T,
rj
may infer that the integrand zero. To verify this conjecture, we forma
is
point
is
interval (x0, xj, containing the point x', in which M > Choose to be
0.
a
rj
function of the type shown in Fig. 3-3 that is, continuous function
is
a
rj
;
82 ENERGY METHODS IN APPLIED MECHANICS
s£0
*0 *1
Fig. 3-3
Sl-£§l + ^Jl = 0
(3.5)
dy dx dy' dx2 dy"
problem
ATfintegral said to be stationary its first variation vanishes. Equa
if
is
tions (3-3) and (3-5) are necessary and sufficient conditions for to be
V
stationary, but they are merely necessary for to have minimum value.
V
are
V
= (3-6)
0
integrand
a
for y, • •
be satisfied.
z,
equations
•
Weierstrass (6) has shown that function y(x) with isolated dis
if
a
dy dy' dyM
-L
(dy
d*F =
V
+ dy' • • •
dyM
+
(3-8)
f
dy dy' dy(n)/
\
AF = dF + - d2F + - d3F + • • •
+ - dkF
+
0(e*+1) (3-9)
2! 3! k\
C
is
b
a
I
defined by
(3-10)
dkV=^dkFdx
Hence by Eq. (3-9)
0(e*+1)
2! 3! k\
84 ENERGY METHODS IN APPLIED MECHANICS
V=\F(x,y,y')dx (3-11)
Case 1: y' absent. If y does not occur in the integrand Fof Eq. (3-11).
the Euler equation (3-6) reduces to
j-
dF
oy
= 0 (3-12)
Since Fis now function of x and y only, Eq. (3-12) ordinarily yields only
a
one extremal, y =f(x). Equation (3-12) means that the value of y corre
sponding to any value of x provides a stationary value to F(x, y). This
condition is merely necessary for a minimum of V; a sufficient condition
is that the value of y corresponding to each value of x provide a minimum
to F(x, y). An extremal with this property does not necessarily exist.
Even if it does exist, it can satisfy the forced boundary conditions only by
accident. Consequently, if y' is absent, the variational problem possesses a
solution only under exceptional circumstances.
-(¥-)-"
dx\oy'l
Consequently,
dF
— = constant (3-13)
ay'
turn, this condition requires that the constant on the right side of Eq. (3-13)
be zero.
(a)
In view of the Euler equation, the right side of Eq. (a) is zero. Conse
quently, Eq. (a) yields
constant (3-14)
Equation (3-14) is a special form of the Euler equation that applies when
x does not occur explicitly in the integrand of Eq. (3-11).
— = — p, — =0, — = Ely
dy dy' dy"
(Ely")" = p (3-15)
is,
load is the inertial force; that
p= —p d2y/dt2, where pis the mass per
unit length. Consequently, the differ
ential equation of freely vibrating
a
uniform beam
is
34v d2y
(3-16)
9
dx* dt2
lever beam that loaded by force F at the centroid of its end section.
is
fa
— C*/2
= -Fflf>/2) + v'f + u")2] dd
+
+
V
Eqs. (3-7) and (3-10), the part that linear in (<5k, dv)
is
is
72 EA
— C"/2
dV= -Fdv\ W)
+
+
+
[(« S'X<5"
Jo
a
du")] d6
+
dV= -Fdv
*/2
-—
F47
(„' + u")
a
du
*/2
x
r*/2
Jo
{[(" + v') + Z(u"" + 2u" + u)] du - (u' + v") dv} d6 =
u = v = u = 0 for 0 = 0
u + u" = 0 for 0 = -
2
-F + — (u + «}')
= 0 for 0 = -
a 2
From the integral terms of Eq. (c), we obtain the Euler equations,
u + v + Z(u" + 2u" + u) = 0
u' + v" = 0
Equation (g) yields
u + v = -KZ
where A• is a constant. Hence Eq. (f) yields
u" + 2u" + u = K
The general solutions of Eqs. (h) and are
(i)
u
- -
0
0
0
0
/42 ^43
= — K6{\
+Z) + A1 cos A2 sin cos
0)
/43(sin
0
0
v
— + sin +
0)
0
>44(cos A5
A2 + K = ^ - + As = + At =
0,
0,
^4 i41
HH ENERGY METHODS IN APPLIED MECHANICS
=
Fa
(sin + cos - cos -
2)
(o)
0
0
u
0
2EAZ
=
Fa
[2(1 + Z)0
+
cos - sin + sin 0-2] (p)
0
2
0
0
_ * ~
£fi
Sin
9>1
(q)
+
„
Z(fl +
/4
J
z)
L
-
The bending moment
is
=
/ ,
cr
(s)
+
/(
^4Za(a
z)
The net tension and the bending moment are given by the equations,
(0
N=jadA, M=jzadA
Hence, Eqs. (q) and (2-22),
by
0)
(u)
we approximate
- a_ Mz
/I F
.
a =
,
H,
(v)
+
<j
/
z
by
+
approximated
z)
is
Equations (t) and (w) yield N = F. This relation disagrees grossly with
statics. Accordingly, it is not permissible to discard z from the denomina
tor in Eq. (v) if the strains are used to calculate the net tension in the beam.
This is true because the strains caused by the net tension and the bending
moment often have quite different orders of magnitude. Relatively small
inaccuracies in one of these strain components may consequently obliterate
the other one entirely. However, the linearizing of the expressions in z
ordinarily causes only small percentages of error in the stresses if the depth
of the cross section is small compared to the radius a. Consequently,
complete linearization with respect to z may be acceptable if the stresses
are desired only for the prediction of failure of the material. The foregoing
considerations apply also to curved shells, and they suggest why the theory
of shells has proved to be very sensitive to approximations.
= (F + XG) dx
'
(b)
length L that has given end points (x0, y0) and (xv y^). Since the differen
tial arc length ds is equal to Vl + (y'Y dx, the potential energy of the
string in the gravitational field of the earth is proportional to the integral
(c)
J*t>
The forced boundary conditions are that the end points of the required
curve shall be the points (x0, y0) and (xv yj. Furthermore, any admissible
curve must conform to the relation
L= Vl + (y')2dx (d)
V(y +
m- K2
A)2
Integration yields
y = K cosh
h+CJ-X (g)
C2. By definition, any solution of Eq. (a) that conforms to class Y also
conforms to class V.
Having determined the general solution of Eq. (a) and having adapted
the constants (Q, Q) to class Y, we may evaluate the integral V by Eq.
(3-1). Thus V becomes a function of Cx and C2. The condition that Vbe
stationary in class V is then satisfied if 6V = 0 for all variations of Cx and
C2 that are consistent with class Y.
It may happen that the forced boundary conditions fix Cx and C2 so
that variations of V are impossible. The foregoing case is consequently
too specialized to have much importance. More significance attaches to
the case of two unknown functions, y(x) and z(x). Then
V=\F(x,y,z,y\z',y",z")d* (b)
Class T now consists of pairs of functions y(x), z(x) that satisfy continuity
requirements and linear forced boundary conditions. Let be the V
subset of class Y that consists of solutions of the differential equation
</>.
<f>
expressed in terms of By giving variations that conform to the forced
boundary conditions, we may derive the natural boundary conditions
<f>.
and the Euler equation for Any solution of the Euler equation for
<f>
that satisfies the forced boundary conditions and the natural boundary
conditions provides stationary value to in class V.
V
a
The Lagrange multiplier provides way of treating auxiliary differential
a
equations in variational problems without the use of the general solutions
of the differential equations at the outset. Referring to Eqs. (b) and (c),
consider the modified integral
(F XG) dx
+
(d)
is
X
designating the function we may give the functions y, arbitrary varia
X,
z
tions that conform to the forced boundary conditions. Thus we derive the
Euler equations and the natural boundary conditions for the integral V.
We may then seek functions y(x), z(x), X(x) that satisfy the Euler equations,
the auxiliary differential equation (c), the forced boundary conditions,
and the natural boundary conditions. These functions provide stationary
a
value to among functions of class V, but, in class V, reduces to
V,
V
V
V
a
z
X,
we may
sometimes dispense with its derivation we can supply proof that
if
it
a
exists. The Lagrange multiplier method does not really evade difficulties
that are inherent in the auxiliary differential equation but sometimes it
leads to relations that have interesting or useful interpretations.
Both of the preceding methods may be extended to problems of multiple
integrals and auxiliary partial differential equations. Care must be used
to include the natural boundary conditions, since one may encounter
puzzling situations he fails to heed these conditions.
if
(x,
cance, these variables may be regarded as rectangular coordinates in
a
represented by
is
space
a
plane.
R
may
a
a
;
METHODS OF THE CALCULUS OF VARIATIONS 93
more generally, it may be any irregular area with holes in it. The region
R will be closed; that
is,
shall contain all its boundary points.
it
A class of admissible functions w(x, defined as follows
y)
r
is
:
(a) Any function of class continuous with its partial derivatives to
V
is
the fourth order in the region R.
(b) Any function of class satisfies given linear forced boundary condi
V
tions of the type aw + dw/dn = c, in which a, are given point
b,
b
c
functions on the boundary of the region and dw/dn denotes the normal
R
derivative of w at the boundary.
derivatives to the third order for all real values of w, wx, wy, wxx, wxy, wyy,
and for all values of (x, in the region R. Subscripts x and denote partial
y)
y
derivatives. Let the function w receive variation erj(x, y), where an
is
a
e
arbitrary constant and an arbitrary function that conforms to class I\
is
rj
is
^v =
[f
(*• V, w +
+
er]j, wx «?». wy
* (b)
+ eVxy, wm + — F(x, y. w, wx, wy, wxx, w^, w„„)] dx dy
wxV eV™)
we obtain
dF dF
+
,
Vx
dwy dwxx
^
*
+ Ixy-z — -— dx dy
+
rjyy
)
e.
is
is
two functions u(x, that are continuous with their first deriva
y)
and v(x,
y)
in the region
R
tives
:
94 ENERGY METHODS IN APPLIED MECHANICS
-^ dx dy = - fo dy
<f>
y- dx dy +
<j>
y;
|J
1
1
R R
(3-18)
!(<j,^dxdy=-[[V^dxdy-i<l>ydx
JJ
R
oy JJR dy Jc
*
R R
=
Hn^dxdy -jy^(f^) dxdy-j^^d.
R R
*
*
&-
i'h
Biit)
£i
m
dx dy dz dy + dy
=/Jr/
R R
R R
♦£'*©*
METHODS OF THE CALCULUS OF VARIATIONS 95
Depending on the order in which the two integrations by parts are per
formed, we may express the integral of the term containing rjxy in either of
the following two forms :
JJ\\vxv Jr~
dwxv
dx dy =
JJ v oxdy\owxy/11
dxdy + iv
Jc dx\owxyl
dx
c dw,y
or
Jr~
dx dy ~
=
(it-
fh
\\vxv owxy dx dy Jc
dy
f
T
oxoyXdw^/
)
dy\owxy/
JJ
JJ
R R
Jc owxv
—— — dx dy
+
^7
(~~~~
)*/J
:
dx dy\dwx
)
dy2\dwvy/J
(3-19)
8F dF
+ dy — e<to
e
r)
— — - dy — —— • dx
(P
(——
e
e
9
Jc Jc
»?x
^
'
)
dyXdw^,' dwxv
l
• •
ds +
<j>
i
<fs
e
•)
•)
Jc Jc
»?n(•
96 ENERGY METHODS IN APPLIED MECHANICS
where s is arc length on the curve C, and r\n denotes the normal derivative
of rj on curve C. The natural boundary conditions are obtained from the
vanishing of the integrands of these two integrals. Rather than to treat
the general case, we consider the case in which the curve C is formed from
segments of coordinate lines. This is not a severe restriction, since, in
physical problems, we usually seek coordinates such that the physical
boundaries coincide with coordinate lines. When such coordinates are
chosen, Eq. (3-19) is in a suitable form for the determination of the natural
boundary conditions. On a part of the boundary represented by x =
constant, dx = 0, and those line integrals in Eq. (3-19) that contain dx
vanish. In this case it is convenient to use the second form for the last two
integrals in Eq. (3-19). Likewise, on a part of the boundary represented by
y — constant, dy = 0, and those line integrals in Eq. (3-19) that contain
dy vanish. In this case it is convenient to use the first form for the last two
integrals in Eq. (3-19). The double integral must vanish independently of
the line integrals, since d V vanishes for all admissible functions rj(x, y).
This condition yields the Euler differential equation. Hence the Euler
equation is
dF _ dtdF\ _ dtdF\ & I dF \
dw dx\dwj dy\dwj dxAdwJ
rj [dF d(dF\
—I-
dx
—I - —I-
a/3F\l
—
.
dy \owvil/
I
.
=0 onedge
,
y = constant
(3-21)
Lowv
r]v
= 0 on^Sdge y = constant
dwvv
derivatives to the third order for all real values of w, wx, wv, wz, wxx, wvv,
wzi, wVz, wz*, wxv ar»d for all values of (x, y, in region R.
z)
is
a
e
constant and an arbitrary function such that w + erj lies in class T.
is
r\
AV ffl7 dF J. dF -L 3F j. 8F
JJJ ow owx owv ow2
\
R
dF dF dF dF ...
+ r— O)
+,
+
+
z— dx dy
dF
— dF . . .
\
+
Vzx dz
rixv
dwj
z
dwzx
in which the part of that linear in e.
V
is
A
V
is
d
dF _ dldF\ _ dldF\ _ dF
d2
dtdF.\
/
\
+
rU-)=°
+
(3-22)
ox ay \0Wxyi
98 ENERGY METHODS IN APPLIED MECHANICS
If the integrand contains several unknown functions (u, v, w), each of these
functions must satisfy Eq. (3-22). To derive natural boundary conditions,
we must examine the surface integrals that are obtained when Eq. (b) is
transformed by integration by parts.
Since the amplitude of the sine wave is a, the potential energy of the load
P is —Pa. The total potential energy is accordingly V = U — Pa. The
assumption that has been used reduces the beam to a system with one
degree of freedom. The generalized coordinate is a. By the principle of
stationary potential energy, the value of a is determined by dV/da = 0.
Consequently,
2PL3 2PL3 .
a = and y =
774£J 77*E/
The difference between the deflection given by this formula and by the
more accurate formula obtained by integration of the equation, M =
—Ely", is everywhere less than 3 per cent. The agreement between the
METHODS OF THE CALCULUS OF VARIATIONS 99
2/n
= tfogo + aigi + <hgi + • ' -
+ angn (b)
where the a's are undetermined constants and the g's are given functions
that satisfy the forced boundary conditions. To permit the sequence to
converge in the mean-square sense to the exact solution y(x), Ritz required
that the functions gt(x) form a complete set in the interval (a, b). To
explain this statement, we suppose that F(x) is a given function in (a, b) and
that F is approximated by
Mn = \\F-Fnfdx
•la
when F(x)
is
(a, b)
coefficients exist so that
bt
lim M„ =
0
100 ENERGY METHODS IN APPLIED MECHANICS
This is not the same as the condition F(x) = lim Fn, unless the operations
n-*oo
"limit" and "integration" may be interchanged. Therefore, without
supplementary investigations, there is no rigorous assurance that the Ritz
method will converge to the correct result. It is known from the theory
of Fourier series that the functions, 1, cos x, cos 2x, • • •
; sin x, sin 2x, • • •
,
form a complete set in any interval of length 2n and, indeed, that any
piecewise continuous function F(x) can be represented exactly by a con
vergent series of these functions in any interval of length 2-rr. It is apparent
that not every infinite sequence of functions is complete, for, if even one of
the trigonometric functions is omitted, the preceding sequence is not com
plete.
We obtain from Eqs. (a) and (b), Vn = Vn(a0, av a2,
• • •
, an), where Vn
is the approximation of V that results if y is approximated by Eq. (b).
The condition d V = 0 is approximated by dVjdOi = 0; i = 0, 1, 2, • • • , n.
These equations determine the a's; accordingly, they give an approxima
tion to y by means of Eq. (b). Mathematical complications often restrict
the values of n to the first few positive integers. Comparisons between
solutions for n = 1,2,
3, etc. indicate the rate of convergence (38).
PROBLEMS
1. Show by the calculus of variations that a straight-line segment is the shortest
path between two points in a plane.
2. Show that the shortest path between two points on a circular cylinder is
helical, provided that the given points do not lie on the same generator nor
on the same cross section.
3. As surface coordinates on a circular cone, adopt the distance x from the
vertex of the cone and the dihedral angle 0 between a fixed plane and a
variable plane that intersect on the axis of the cone. A path on the cone
that does not coincide with a generator may be defined by an equation of the
form, x = x(6). The path is said to be a "geodesic" if it is an extremal for
the distance on the conical surface between any two points on the path.
Derive the general equation of the system of geodesies (excluding generators)
for a cone whose generators intersect its axis at 30°.
where k and a are constants. The end x = 0 is clamped, and the end x = L
is free. The beam carries a distributed lateral load p(x). State the forced
boundary conditions. Derive the natural boundary conditions and the
differential equation that determines the deflection.
7. Solve the problem of the curved cantilever beam (Sec. 3-6) for the case in
which the load F acts horizontally to the right.
8. Solve the problem of the curved cantilever beam (Sec. 3-6) for the case in
which a bending moment M is applied at the free end. Let positive M be a
closing moment.
9. A uniform semicircular arch of radius a is clamped at its ends. Supposing
that the arch is loaded only by its own weight, derive formulas for the radial
and tangential displacement components («, v). Hence derive formulas for
the bending moment and the net tension at any cross section.
10. Free vibrations of a cantilever beam of length L are defined by the equation
y = f(x) sin nt. By means of Eq. (3-16), derive the function f(x). Prove that
n is determined by the equations k* = prPjEI, cos kL cosh kL = —1. Hint.
The boundary conditions for the free end are yxx = yxxx = 0.
11. Solve Prob. 10 for a simple beam.
12. Calculate the constants in the equation of the catenary for the case in which
the curve passes through the points (0, 0), (1, 1), and ( — 1, 1).
13. A catenary is symmetrical with respect to the y-axis. The arc length,
measured from the vertex, is s. Show that the curve has the following
F(X, l/, U, UX, Uv, Uxx, Uxv, Uyv, UXXX, UXXv, UXyv, Uyyy) <iX dy
R
15. A surface of revolution is generated by connecting two points (x0, y0) and
(xi, yi) by a plane curve C and rotating the curve C about the a;-axis. Show
that the area of the surface is stationary if the curve C is a catenary. (The
surface is then called a "catenoid").
16. The strain energy of an orthotropic plate is given by
Fig. P3-17
V = + wyyf dx dy
i^JJW
where D is constant called the "flexural rigidity" of the plate. By the
a
ROBERT HOOKE
= ds^-ds
ds
is
£e2
e
\VvM + \yvzmn
+ +
evm2 (4-2)
\yzM +
+
e2n2
where =
+ + +
+ + +
+ +
K"v2
v
vy2
v
*, = w. \W v2 + w2)
= = », + +
+ + +
+ +
fryx ey fry
IVzz hzv
is known as the "strain tensor," since it obeys the tensor law of trans
formation when the coordinates are changed.
If two line elements have the initial direction cosines (Iv mv rtj) and
(l2, m2, 7J2) and these line elements are initially perpendicular to each other
(i.e. /j/2 + m1m2 + n-jx2 = 0), the angle 6 between them after the deforma
tion is determined by
Here, e1 and e2 are the strains of the respective line elements; they are
determined by Eq. (4-2). The angle (tt/2) — 6 is known as the "shearing
The quantities (fa, fa, fa) are the three roots of the determinantal equa
tion,*
- fry =
<t>
iyvx (4-5)
0
ey
fr„ \yzv -
i>
I
is
106 ENERGY METHODS IN APPLIED MECHANICS
<£,
e, > 1, > . Expansion
1
- ltf +
<f>3
I2<f>-I3 =
0
(4-6)
where
= +
+
h
ex «v ez
(4-7)
<f>f
a
- + \y*jn + \y.jn =
0
(«« 4>j)l
\yj + - ^> + =
J
\yyj>
o
(«» (4-8)
ly,s» + - id" =
to
(«.
o
+ + =
/2
m2 n2
1
The solution (/„ m„ n<) of these equations the set of direction cosines of
is
=
+
+
fa
fa
fa
h h h
(4-9)
—
fafafa
{x, y, coincide with the principal axes. Since the principal strains are
z)
is,
Tj,
/2,
13
;
I2,
1
3
tensor. Any other invariant of the strain tensor can be expressed in terms
of /2, and I3.
Iv
DEFORMABLE BODIES 107
V* — V
e = (4-10)
V
Eq. (4-12). Also, since = + \ef, the invariants Ixv I2, may be
/3
et
2/2
e
ex
— ux,
ev
= vv» ez
— Wz
(4-14)
yvz = wv + vz, yzx = uz + wx, yxv = vx + uv
If quadratic terms in the strains are neglected, Eqs. (4-2) show that ex, ev, ez
are the strains of line elements that initially lie parallel to the x-, y-, and
z-axes. In addition, Eq. (4-4) shows that yyz, yzx, y^ are the shearing
strains between pairs of line elements that initially lie parallel to the axes
indicated by the subscripts.
It is apparent from Eqs. (4-14) that the linearized strain components
satisfy the differential equations
dy2 dz2 dy dz
d% d\_ d2yz*
dz2 dx2 dz dx
d\ ,
d*ex
= S2yxv
dx2 dy2 dx dy
(4-15)
, d2e
cg
|
d2v
fvz _ d2v
Yzx
j
"d2v/xv
dy dz dx2 dx dy dzdx
, 3% d2yzx
= d2yxy d2yyz
d2e z d2v
w r xv d2vYvz
^ d2v
w Yzx
2 | j
dx dy dz2 dz dx dy dz
(ex, ev, ez, yyz, yzx, yxv) are possible strain components and only
if,
they
satisfy Eqs. (4-15). The corresponding relations in the large-displacement
theory for which quadratic terms in Eqs. (4-3) are retained are also known,
but they are very complicated (56).
DEFORMABLE BODIES 109
TVX Oy Tyz
, Tzx Tzy az ,
For example, the first row of this matrix represents the x-, y-, and z-com-
ponents of the vector ax. Evidently ax, <x„, az are normal stresses on the
• • •
corresponding coordinate surfaces, and rxy, rxz, , rzy are components
of the shearing stresses on the coordinate surfaces.
By balancing moments on an arbitrary free body cut from the material,
tyz = rzy,
is,
m, and
ri)
S
(4-16)
p
110 ENERGY METHODS IN APPLIED MECHANICS
Denoting the components of vector p on the coordinate axes by (px, py, p2),
we may write Eq. (4-16) as follows:
Jy = ax + ay + az
ay
The quantities Jv
J2, J3 are stress invariants. They are related to the
principal stresses by the equations
Jl ~ a1 + a2 + a3
ri)
direction cosines of the outward normal to S. The stress vector on
is
S
given by Eqs. (4-17). The material may be subjected to spatially distrib
uted forces such as weight, centrifugal force, or magnetic effects. These
forces are known as body forces. If the vector represents the intensity of
F
jFv + kF2) dx dy
+
dz
j\jp(iFx
R
mTvx ™Oy
+JJ[('ff* nrzx)•1
s
+ dS =
+
(1txz
0
mtvz naz)k~]
dx dy dz = • dS
Jjjdiv (4-23)
q
R
S
112 ENERGY METHODS IN APPLIED MECHANICS
\ ox oy dz I
+ kfe
\ dx
+
dy
+ d-Ei
dz
+ pFY] dxdydz =
IJ
0
Since this integral vanishes for any region R within the body, the integrand
is zero. Consequently,
dx dz
^ + ^ + ^+^» =
dx dy dz
0 (4-24)
dx dy dz
Equations (4-24) remain valid for a vibrating body if the body force F
includes inertial force.
(y,
a
Similarly, there are two other families of coordinate surfaces with the
respective parameters and z. The intersection of surface = constant
y
y
a
DEFORMABLE BODIES 113
If
dy
0, z).
dz.
+
is
rv
rz
dy = dz = the vector dr tangent to an x-coordinate line. Therefore,
is
rv
rz
tangent respectively to y- and 3-coordinate lines. Accordingly, the co
if,
•
rv
0
r2
rv
rx rx
ds2
rz
rv
=
p2
+
y2
a2 dx2
+
(4-27)
system.
In view of Eq. (4-27), the magnitudes of the vectors rx, rv, are, respec
rz
tively, a, y. Consequently, the unit vectors tangent to the x-, y-, and
j8,
coordinate lines, dxjds = //a, dy/ds = mfP, dzjds = njy. The strain
components are defined as for rectangular coordinates by Eq. (4-2). The
derivation of the expressions for ex, ev, ez, yyz, yzx, yxv routine problem
is
a
114 ENERGY METHODS IN APPLIED MECHANICS
of tensor calculus. The following results are also derived without the use
of tensors in some treatises on elasticity (61):
For many applications, the quadratic terms in Eq. (4-28) are neglected.
When ex, ev, ez, yyz, yzx, yxy are determined by Eq. (4-28), Eqs. (4-2), (4-4),
(4-6), (4-7), (4-8), (4-9), (4-11), (4-12), and (4-13) remain valid for any
orthogonal coordinates. Equations (4-28) are easily specialized for
particular coordinates. For cylindrical coordinates, x = r, y = 6, z = z;
then, x = I, P = r, y = \. For spherical coordinates, x = r, y = 6 =
= = longitude; then a = =
<f>
colatitude, z — sin
r,
8.
1,
r
/?
DEFORMABLE BODIES 115
With reference to any orthogonal coordinates (x, y, z), the stress nota
tions of Sec. 4-2 are retained. For example, ax denotes the tensile stress on
a plane element that is normal to an x-coordinate line ; t„ and rxz denote
the y- and z-components of the shearing stress on the same plane element.
For equilibrium of moments on an infinitesimal element, the symmetry
relations, tvz = tzb, t2X = rx„ txv = tvx, must be retained for all ortho
gonal coordinates. Furthermore, if
denote the direction cosines
(/,
m,
ri)
of the normal to plane element relative to the local coordinate lines,
a
Eqs. (4-17), (4-18), (4-19), (4-20), (4-21), and (4-22) remain valid for any
orthogonal coordinates. The differential equations of equilibrium take
the following general form:
d
ox ay d
oz
yPxVy
o
^ ^ ^
d
Tz d
+ ya<T" + +
Tx
dy (4-29)
yPxTvx
o
— y*rzv + —
d
These equations are purely statical; they must be satisfied even though
creep, plastic behavior, or thermal stresses occur. If the body-force terms
are generalized to include inertial force, Eq. (4-29) applies for vibrating
bodies.
4-4.
DEFORMATION PROCESS. The displacement vector field that
carries a deformable body from configuration X0 to neighboring con
a
figuration X denoted by (du, dv, <5r,). The functions (du, dv, dw) are to
is
be (u, v,
total differentials of («, In other words, du = ef
dv = erj, dw = e£,
v,
w).
,
z)
plasticity only the incremental displacements (du, dv, dw) have significance,
since the zero stateof deformation has no absolute meaning. Deforma
tions in the large may be regarded as integrated effects of infinitesimal
deformations.
116 ENERGY METHODS IN APPLIED MECHANICS
(du), dyxy =
^-(dv) + f(du),--- (a)
n)
of the outward normal to S. As in Sec. 4-2, the intensity of body force
is
denoted by (Fx, Fv, Fz). In view of Eqs. (4-17),
dS
+
+
+
(/'
+ p(Fx du Fy dv Fz dw) dx dy dz
+
R
DEFORMABLE BODIES 117
Jjj
-dWi = dWe = [a. dex + ay fc» + az dez
R
(4-30)
+ t„ dyyz + tzx t„ dyxy] dx dy dz
+
dyzx
R
is
U
=jjju0dxdydz (b)
^0 = + fyyz + fyzx +
+
if
a
deformation (du, dv, dw) not performed adiabatically.
is
is
it
express the components of stress in the following form
:
a* = ax' + = aj + s, = +
s,
az a2'
s
<tv
'
= \(PX
+
ay az)
s
ax 'Xy
a;
/
ryz'
r,x' <-
,
Tzv
is
av'
When the body undergoes the infinitesimal strains (dex, dev, dez), the
volumetric strain [Eq. (4-13)] de — dex dez. The increment of
+
+
is
dev
internal energy density due to the volumetric change dU0v = de =
is
dU0 = dU0v + dU0d. The term dU0" called the increment of internal
is
= + + + rj + tzx' tJ
+
(4-33)
Tzx
118 ENERGY METHODS IN APPLIED MECHANICS
a
vanish. This condition unavoidable there are residual stresses due to
if
is
a
state of deformation as the zero configuration for potential energy. For
example, in buckling problems, the zero configuration may be considered
as the unbuckled state immediately before buckling.
Usually elastic deformations of solids do not cause significant changes of
temperature. However, the strain energy density may depend on the
prevailing temperature Problems of thermal stress involving transient
6.
temperatures may often be analyzed with the assumption that the time rate
of change of temperature so slow that associated inertial effects may be
is
is
though the temperature distribution at that instant were maintained.
Processes for which this type of approximation admissible are said to be
is
"quasi-stationary."
In view of the preceding remarks, the strain-energy-density function U0
generally depends on the strain components, the coordinates, and the
temperature; that is, U0 = f(ex, ev, e„ yyz, yzx, yxv, x, y,
If
the dis
z,
6).
placement vector (h, v, w) receives variations (du, dv, dw), the strain tensor
takes variations • •
and the corresponding variation of
•
is
dex, dyyz, U0
,
...
°U0 =
dU0— ,dU0.
— ,dU0.
— de2 +,dU0Adyyz
*
+
dex dey
de* de* 3e, dVv,
(4.24)
dU0 dU0 .
r— fyzx + t— fyXy
+,
°yzx oyxv
Since Eqs. (4-31) and (4-34) are valid for arbitrary variations (du, dv, dw),
Although Eqs. (4-35) have been derived for rectangular coordinates, they
remain valid for any orthogonal coordinates (x, y, z). Furthermore, since
rigid-body displacements have no effects on the stresses and strains,
Eqs. (4-35) remain valid irrespective of the magnitudes of the translations
and rotations of the particles, provided that the strains are small and the
stresses are referred to the deformed coordinate system. For example, if a
thin rod with axial coordinate x is bent into a loop, ax is to be interpreted
as the normal stress on a cross-sectional element of the deformed rod.
If the rotations are large, the derivatives of the displacement vector are no
longer small, and the quadratic terms must be retained in the equations
that define the strain tensor [Eqs. (4-3)]. However, this circumstance does
not invalidate Eqs. (4-35) if the strains are small. If the strains are large,
a generalization of Eqs. (4-35) is required (56, 61).
= ff4, Tzx
= a5»
= 0-6
Oi
- 5 i = •
1, 2, • • , 6 (4-37)
After this solution of Eq. (4-37) has been determined, U0 may be expressed
as a function of (av a2, • • • , a^).
*
Through the courtesy of the Franklin Institute, the material in this article is extracted
from the author's paper, "The Principle of Complementary Energy in Nonlinear
Elasticity Theory," Jour. Franklin Inst., 256, 3, September 1953.
120 ENERGY METHODS IN APPLIED MECHANICS
defined by
dT0
da,
% dat dat da J
Also, the chain rule of partial differentiation yields
W0 =
da,
|
i=i
dUo djk
dek da.
Consequently,
Off,-
= 0 if k^i and ^=
OCTj
1 if A: = i
= = • • •
2,
1,
(4-39)
6
*i ,
/
,
;
is
it
that there function T0 which satisfies Eq. (4-39). This function may be
is
a
and T0.
For structures built of beams, struts, or torsional members, there may
be only one significant component of stress, say av Then, in view of
Eq. (4-35), the function U0 depends on the single strain component ev
Consequently, by Eqs. (4-35), a1 depends on et only. Figure 4-1 illustrates
typical form of the (av ei) graph, called "stress-strain curve." By
a
Fig. 4-1). This area may be measured from an arbitrary abscissa; for
simplicity, the fixed abscissa is taken to be e1 = 0. Since the product
is represented by the area of a rectangle in the plane, Eq. (4-38)
(av
shows that the area above the stress-strain curve (shaded in Fig.
4-1)
represents the function T0. How
ever, this graphical interpretation
of T0 is applicable only if there
is but one nonzero component of
stress.
It is assumed in some theories
of plasticity that the increase of
internal energy during an adiabatic
deformation depends only on the
final state of stress, and not on the
path by which that state is attained,
provided that the stresses increase
monotonically during the loading
(60, 76). These theories do not dis
tinguish between inelasticity and
nonlinear elasticity unless unload
ing occurs. Consequently, the concept of complementary energy finds
applications in analyses of plastically deformed bodies.
U0
= hl iw,-2(c«0 +
i .H.l
+ constant (4-40)
» =1
ai = 1 b<i*t -W + gd ; i - 1, 2, • •
, 6 (4-41)
These equations are valid only if the strains are small, since Eq. (4-37)
is subject to this restriction. For a nonhomogeneous material, the quanti
ties bH, Ci, gi are functions of x, y, z.
The state of zero strain may be designated arbitrarily; if this origin is
122 ENERGY METHODS IN APPLIED MECHANICS
Substitution of Eq. (4-42) into Eq. (4-40) yields, with Eq. (4-43),
6 6
The additive constant in Eq. (4-44) contains the temperature, but it does
not contain the stress components ; hence it is irrelevant.
Equations (4-39) and (4-42) yield
T0 = *2
i =1 3 =1
Iw+2W i =1
(4-45)
If thermalstrains are absent (6 = 0), Eqs (4-44) and (4-45) yield T0 = U0.
Consequently, if there are no thermal strains, Eqs. (4-37) and (4-39) show
that homogeneous Hookean materials obey the conjugate relations,
ai = — !? , e<
= — 0
; i = 1, 2, • • • , 6 (4-46)
is,
no symmetry at
all) constitute the triclinic system. Crystals of this class have 21 elastic
constants bu and thermal constants ct.
6
The coefficients bu in Eq. (4-40) depend on the orientation of the co
ordinate axes (x, y, with respect to the lattice of the crystal, since U0
z)
invariant under coordinate transformations. The transformation of the
is
a
tensor theory. If
by
is
a
an K-fold axis of symmetry, the coefficients bu are evidently unchanged.
This condition determines the number of elastic constants for crystal of
a
given class.
For example, crystal that possesses only single 2-fold axis of sym
a
a
metry belongs to the monoclinic system. A rotation of the crystal through
the angle it about this axis leaves the elastic coefficients unchanged. If
the axis of symmetry the z-axis, the coordinate transformation corre
is
y'
z'
is
el —
_ du' _— _— dv' _— —
,
~. el, e2 e2, e3 e3
,
ox ay
+ +
,
«4=^ &<
£5
a? ^ (a)
dv' du'
,
ox dy
is
et
3
1
<.14.1 t .1
Since permutations of the quantities Cj in this equation are irrelevant
because of the isotropy,
*ii = b2i = b^ = X + 2G
*23
= b31
= ba = X (4-47)
Ci = ct = c3
= c
The constants and G were introduced by Lame. In view of Eqs. (c) and
A
Tx»
= Gy^ e = ex + ev + e,
= (/ + G)g, -Xa,
, e, = e,=
G(3/. + 2G) 2G(3/ + 2G)
Yn = y« = y„ = 0
DEFORMABLE BODIES 125
= "f . «v
= = -»**. y„ = yzx = y** = 0 (4-52)
£
where £ is Young's modulus and v is Poisson's ratio. Hence
£y G(3A + 2G) A
/ + G 2(A + G)
(4-53)
2Gv „ £
A =
1 - 2v 2(1 + v)
ex
- IE - [ax v(ay + a,)] + kd
=
l [<r»
- K<r, + <r.)] + k6 (4-54)
«
e2
=
I£ - [<rz v(ff, + »»)] + fc0
yvz ~
^vZ
, yzx
^~Z$)
' yxu.
~
^
xv
where
k = (4-55)
£
The constant k is known as the coefficient of thermal expansion. Equa
tions (4-45) superpose strains due to stress on strains due to temperature.
Substituting Eqs. (4-54) into Eq. (4-50), we obtain the strain energy
density of an isotropic Hookean body in terms of the stresses:
U0 =
— [a* + ay2 + a2 - 2v(aya2 + azax + axay)
2E
+ 2(1 + v)(rj + rj + rj)-] (4-56)
Since Eq. (4-56) does not contain 6, the strain energy density is determined
by the stresses alone. Of course, the temperature distribution may affect
the stresses. Stress-free thermal expansion of a body contributes nothing
to the strain energy.
Equations (4-39) and (4-54) yield
Y0 =
— \a2 + a2 + a* - 2r(ff»ffz + a,ax + axay)
2E
+ 2(1 + v)(rj + rj + rj)} + kd(ax + ay + az) (4-57)
(/,
body may have cavities or holes. On the surface of cavity, the normal
a
(a)
0
dT. (b)
DEFORM ABLE BODIES 127
r=jjjrodxdydz (4-58)
R
For Hookean body in the absence of residual stresses and thermal strains,
a
T isidentical to the strain energy, for then T0 = U0. With Eq. (b), Eqs.
(4-39) and (4-58) yield
= + + +
<5Y
JjJ(«, &ax ev &ay ez &az yyz &Tyz
R
+ 7zx fazx + 7xv foxy) dx dy dz (c)
R
+ ("z + w x) &Tzx + (vx + "») dtxv1 dx dy dz
dT = + +
^
drxy w dtX2)
(u
dax
v
r
—
d
+ + — dx dy dz
(u
vday w dryz)
+
+
(u
w daz)
+
ay dz
J
+ «[|-(^) + |-(H)
+
Lox dy |(«5tm)
dz
+ w (drzx) + + dx dy dz (d)
f
(drzv)
I
Lox ay dz
(daz)\
JI )
In view of Eq. (a), the second integral in Eq. (d) vanishes. Hence by the
divergence theorem [Eq. (4-23)], Eq. (d) may be expressed as follows:
dT =
Jj[(" + fay* + vday
+
+
+
w w dryz)m
("
&ax
v
&Txv
s
+ + + w daz)n] dS
(u
Let the stress vector on the surface S be p = \px + \py + kpz. Then, by
Eqs. (4-17), Eq. (e) may be expressed as follows:
(4-59)
s
• • •
where <5Y0 is a linear form in (dax, , dt^), <52T0
is a quadratic form in
(dax,
• • •
, drxv), etc. If T0 is a quadratic function of the stress components,
DEFORMABLE BODIES 129
6 6
<32T0
= 2. i2=1au dat dOj
< 1
By virtue of Eq. (4-44), SHa,, dat da t is positive definite (see Appendix A-l)
Consequently, <52Y0 is positive definite. Hence <52T is positive definite,
since
d2Y = !!d2r0dxdydz
Among all states of stress of a Hookean body which satisfy the differ
ential equations of equilibrium and the boundary conditions on Sx, that
which represents the actual equilibrium state provides an absolute
minimum to XV.
This theorem remains true even if there are thermal strains, since the linear
terms in the complementary-energy density do not enter into the argument.
For the same reason, the theorem does not require that the strains be
measured from the unstressed state ; any origin of strain is admissible.
The theorem incidentally establishes the uniqueness of the equilibrium
state of stress for a Hookean body. It is to be recalled, however, that
small-displacement approximations have been used in the derivation of
Castigliano's theorem.
Castigliano's principle of least work is useful for deriving compatibility
equations for the stress components, particularly if the material has
anisotropic or nonlinear stress-strain properties. If the boundary condi
tions are entirely of the stress type, the region S2 vanishes, and, by Eq.
(4-60), W = Y. Then Castigliano's theorem is expressed by the varia
tional equation, oT = 0. In any case, the equations of compatibility for
stress components are determined by the condition <5T = 0, since these
equations are independent of the boundary conditions. Variations of the
stresses are restricted to the class of functions that satisfy the equilibrium
conditions.The equilibrium equations (4-24) may be treated by the
Lagrange-multiplier method (Sec. 3-8), and the Euler equations for the
modified complementary function T may be identified as the compatibility
equations for the stress components. This method has been used as an
alternative derivation of Castigliano's principle (120). It is also possible
to represent the general solution of the equilibrium equations by means of
the stress functions of Maxwell or Morera or by a combination of the two.
130 ENERGY METHODS IN APPLIED MECHANICS
T = u=
JE JT [FJ + Fy* - 2vF*xF™ + 2(1 + v)F*y2] dx dy (g)
where h is the thickness of the plate. The Euler equation for this integral
is V2V2/r = 0. This is the known compatibility equation of plane-stress
theory for isotropic Hookean bodies. The same method could be used to
obtain the compatibility equation of plane-stress theory for an aniso
tropic Hookean body, but a generalization of the strain-energy formula
(4-56) would be required.
Castigliano's principle of least work may be used to obtain approximate
solutions of elasticity problems by the Ritz method. This application of
the principle has received relatively few applications, but it has important
potentialities for stress-type problems.
m,
(/,
the
ri)
is
stress tensor
is
S
are/>x
pz = Pz, where (px, py, pz) are given functions on Sv The forced boundary
DEFORMABLE BODIES 131
it,
w) are given func
v,
v,
v
tions on 52. The complementary energy density T0 regarded as func
is
a
tion of the stress components. The body force, per unit volume,
is
(pFx, pFv, pFz), where the mass density. The following functional
is
p
introduced
is
+
J
- KuFx + -
it
("/>* vpv
ff
[(«
- u)Px + (v- v)pv + (w- w)pz] dS (4-61)
of
elasticity, the equilibrium state
of
•
•
0
of
ensures the satisfaction
of
the differential
the stress-displacement relations, and the
of
equations equilibrium,
boundary conditions.
The fact that the equilibrium equations and the stress-displacement
relations are satisfied follows immediately from the observation that these
equations are simply the Euler equations for the triple integral in the
formula for J. The double integrals in the equation for are introduced
J
av,
,
J
drxy.
,
is
- [px du + py dv + pz dw + - + - dpy
«)
(u
dpx
(v
v)
JJ
+ (w - vv) dpz] dS •
(a)
132 ENERGY METHODS IN APPLIED MECHANICS
jjjux dx dy dz
=jjl" dS
R S
w
=![(«.-
R £)«*+•--+(*+ *-£)*•-
" *
I<>F»)
'F*)
(lT
~
IT +
IT *" +
+
+
+
ox oy oz ox 7T
ay az
1
\
+ +
+
+
K'ff* WT** nrzx (Wxy mcr,,
Si
+ (frxi + + — pz) dS
mryz naz <5w]
dpv (w
s,
Since vanishes for arbitrary variations <5h, or, dw, dax, dav, • •
•
<57 drxv,
,
«. = = -r-2 =
•
+ r~ (b)
-
-
»»
•T"2 x «»
,
,
.
ox oy oz
ox oy oz
^=
T-'
+
+
T2
°
Or Tar
OJ/ 02:
Equations (b) are the stress-displacement relations; Eqs. (c) are the
DEFORMABLE BODIES 133
hx + mryx + nr„ - px
'
U = U, v = v, w = H' on S2 (e)
of the constraints. The part S2 may consist of a few small spots, commonly
called "point supports." Since the displacement vector vanishes on S2,
the region of integration in Eq. (4-59) reduces to Sv The variation of
boundary stresses on is not restricted by equilibrium conditions, since
equilibrium may be maintained by compensating reactions in S2.
Fig. 4-2
dT = w dp,) dS
+
So
Equation (a) may be simplified by the theorem of the mean for integrals
If are continuous real functions in of
y) y)
y)
<[>(x,
jjfi*, dx dy = K JJ#x, dx dy
y)
V)
V)<K*, (4-63)
A A
where K the value of f(x,y) at some point in A. Accordingly,
if
is
(dpx, dpv, dpz) all retain constant signs in S0, Eq. (a) yields
<5Y = ds + ds + ds
»oJJ<5P, vojjdpv woJJ^2
So So So
<5T = q0 dS (b)
-JJ^p
So
DEFORMABLE BODIES 135
"If.pdS
So
Consequently, Eq. (b) yields
<3T = q0 • dF (c)
Since dF is arbitrary, we may let collinear with F. Then dF =N dF,
it be
where N is a unit vector in the direction of F. Hence, by Eq. (c),
«5T = q0 • N dF = qF dF (d)
where qF is the component of the displacement vector q0 in the direction
of force F. Also,
dT = —dF
dF
Therefore, by Eq. (d),
q< =
|J
dFi
(4-64)
dxt
where U is the potential energy of internal forces and Pt are the components
of generalized external force. In this case the complementary energy is
defined by
T- -17 + J PA (4"65>
*=1
136 ENERGY METHODS IN APPLIED MECHANICS
f£
- *, (4-66)
Fig. 4-3
and that is subjected to external forces F1, F2, • • • (Fig. 4-3). The body
may also be subjected to distributed loads and temperature gradients. If
two parallel concentrated forces, Fj and Fa, with opposite senses, act
perpendicular to the ends of a straight-line segment of length a (Fig. 4-3),
Eq. (4-64) shows that the rotation of the line segment due to the deforma
tion is
a dFl a dF2
* If the body is supported by springs, the supports may be regarded as a part of the
body.
DEFORMABLE BODIES 137
Fig. 4-4
ar =
ar ar
dF dF1 dF2
Consequently,
a 3F
Since the equal and opposite forces Fj and F2 constitute a couple of
magnitude M = Fa, this equation may be written
6 = — (4-67)
dU
= (a)
*1
^
138 ENERGY METHODS IN APPLIED MECHANICS
dU F2 = 0
q* = (b)
W2
For Hookean body, the deflections are proportional to the loads, and
a
the strain energy is accordingly a quadratic form in the loads. Therefore,
in the present case
U = aFf + bFxF2 + cF2* (c)
where a, b, c are constants. Equations (a), (b), and (c) yield qx = bF2,
q2 = bFv Consequently, if Fx
= F2, qx = q2. This conclusion may be
expressed as follows :
For a Hookean body, the deflection component at point 1 in the direction
Dx due to a load F
at point 2 in the direction D2 equals the deflection
component at point 2 in the direction D2 due to a load at point 1 in F
the direction Dv
This is Maxwell's law of reciprocity. It may be illustrated by the solutions
of elementary problems of mechanics of materials. For example, in Sec.
2-2, the deflection of a simple beam of length L due to a concentrated
lateral load Q at the point x = c is shown to be
20L3
-f—
£
2, "
_4 .
sin —
nnc .
sin —
rnrx
TT EI n = l L L
Since this equation remains unchanged if x and c are interchanged, the
deflection of a simple beam at point x due to a lateral load Q at point c
equals the deflection at point c due to a lateral load Q at point x. This
statement may be regarded as a special case of Maxwell's theorem.
Incidentally, Sec. 2-2 shows that Maxwell's law of reciprocity applies also
for beam columns, since the only change in the above formula is the
replacement of the term n4 by n* — rn2, where r = PjPe.
It is apparent from Eq. (4-67) that there is also a reciprocal theorem for
couples and rotations.
9
BF
Z BF
Since T, depends only on the tension Nt in the ith member, this yields
* dNt dF
By Eq. (4-68) dTJdNt is the extension et of the ith member due to the
actual loads and thermal strains. Accordingly, if dNjdF is denoted by n„
q = 2>A (4-69)
is
a
is
a
N<
:
/th member = and all other loads are removed from the structure.
if
F
1
tween two joints of truss — say joints J' and J" equals the displacement
—
a
= 2>Xn/ + O (a)
q
applied loads, «/ (or n") the tension in the ith member due to unit
is
if
100,000# 100,000#
Fig. 4-5
parts with identical deflections, say s parts. Then loads of magnitude 1/s
should be applied to the parts for the computation of the tensions nt.
Compression Tension
^12
= ^24
= ^34
= 1-5 in.2
^23
= ^13
= ^35
= 1 in-2
The stress-strain curve for any member is represented by Fig. 4-6. Since
there is no unloading, the distinction between inelasticity and nonlinear
elasticity is irrelevant.
The reduction of the distance between joints 2 and 5 is required. The
tensions in members 23, 24, and 35, as determined by the equations of
equilibrium of the joints, are N23 = 115,470 lb, #24= —115,470 lb,
N35 = 1 1 5,470 lb. Consequently, the stresses are a23 = 11 5,470 lb/in.2,
cr24
= -76,980 lb/in.2, a35 = 115,470 lb/in.2. Hence, by Fig. 4-6, the
strains are e23 = 0.0065, «24 = —0.0041, e35 = 0.0065. Accordingly, since
the members are 100 in. long, their extensions are = 0.65 in., eM =
—0.41 in., £35 = 0.65 in.
Since the truss is statically determinate, superposition of tensions is
DEFORMABLE BODIES
143
legitimate. Since the truss consists of two identical parts with symmetrical
loadings, we apply tensions of magnitude £ lb between joints 2 and 5 and
between the corresponding joints 3 and 6 (Fig.
4-7). The tensions in the
members due to these loads are
= nl3 = A*
n12 «M = 0, «23 = «21 = 6 4
-0.2887 lb, = -0.5773 lb.
By Eq. (4-69) the reduction d of
the distance between joints 2 and
5 is
d = 2/123*23 + +
2n2ieU «35e3S
d = 2(-0.2887)(0.65) +
2(-0.2887)(-0.41) + (-0.5773X0.65)
Therefore, d = —0.51 in. The negative sign signifies that the distance
between joints 2 and 5 is increased 0.51 in.
6000 #
Fig. 4-8a
eAB
= -192,000 - 38.4F, eBC
= -162,000 - 10.8F
By statics the tensions in the members due to a unit tension acting between
joints A and C (Fig. 4-8c) are
»cd = -0.8
6000#
NAB =
- 1906 lb, NBC = -7430 lb, NBD = 2383 lb,
NAD = 1570 lb, NCD = 2094 lb
Jo R
f(M) = — (a)
Jo K
Then, for a beam that is subjected to uniform bending, Y = Lf(M).
Consequently, for nonuniform bending, energy in an
the complementary
infinitesimal length ds is f(M) ds. Therefore, if any beam is subjected to a
bending moment M(s), where s is an axial coordinate,
Y =
I /(M ) ds (4-70)
Jo
This formula does not take account of shear deformation or axial forces.
By Eq. (4-70) the complementary energy of a structure that consists of
flexural members (e.g., a frame) is
T = I J/(M<)
ds (b)
where A/,- is the bending moment at any point in the ith member, and the
integral extends over the length of the ith member. The sum extends over
all members. If P is a force applied to the structure, the component of
deflection q at the point of application of force P and in the direction of
force P is, by Eq. (4-64), q = 3Y/3P. Hence, by Eq. (b)
4 =
l[^S (4-71)
If the bending moments vary linearly with the external forces, m,- is inde
pendent of the actual loads on the structure. We may then interpret nti
to be the bending moment in the ith member if P = 1, and all other loads
are removed from the structure.
146 ENERGY METHODS IN APPLIED MECHANICS
0 = 2f— (4-72)
M = 2bK \eHr)dr)
Jo
Since e = rj/R, this yields
„ = 4bKhH
M T7- or —
1
=
25M2
5R}4 R l6b2K2hs
fl =
CLmds = 25P2
—— CLt .
6, q \ s m as (c)
Jo R 16 b2K2h5 Jo
since M = Ps. Applying a unit vertical load at the free end, we obtain
m = s, and Eq. (c) yields
9
_
~
25P2L4
64b2K2h5
the deflections be small, and let the deflection component of the point of
application of load P along the line of action of P (or the angular displace
ment of the arm of the couple P) be denoted by q. Since the material is
Hookean, U = T. Consequently, by Castigliano's principle [Eq. (4-64)
or (4-67)], q = dU/dP.
If the structure consists of beams, columns, ties, and torque bars, its
strain energy is determined by the formulas in Sec. 2-1. Consequently, if
the tension, bending moment, shear, and torque in a member are denoted
by N, M, S, and T, respectively, and if x is an axial coordinate along a
member,
Fig. 4-9
The tensions, moments, and shears in the various members, due to load
F, as determined by statics, are
Ni = 2F M1 = 0 Si = 0
N2 = -iV2F M2 = 0 = 0
N3 = 2F M3 = 0 = 0
Ni = -2F M4 = Fx St = F
Ns = 0 Mb = Fx s6
= -F
Since the angular deflection of member 2 is required, we place a unit
we impose equal and opposite forces of
is,
0
*1
20
= V2/40 = =
0
«2
sa
m2
= = =
0
«8 m3 *3
= mi = =
l
*4
0
»4
2
0
= = =
0
*i
«5 »»5
DEFORMABLE BODIES 149
Since the /w's and j's are zero, Eq. (4-73) reduces to
= »i^i^i + n2N2L2
+
W4N4L4
M = Rx- iwx2
To apply the unit-dummy-load method, we consider the moment due to a
1-lb load applied at the free end (Fig. 4-10c). This moment is m = x.
Neglecting the effect of shear, we then have, by Eq. (4-73),
q - \L ^E dx = i- - iw*3) dx
EI
Jo EI Jo\\R*»
Hence , .
RI? wL*
q =
3EI SEI
Since R is the reaction of a fixed support, q = 0. Therefore,
150 ENERGY METHODS IN APPLIED MECHANICS
w*/~
MtiilllMnitlMTll
^
UUIIUUUUIU
(6)
'
Fig. 4-10
1#
Figure 4- 12a shows that the bending moment due to the applied loads is
q = (
J EI
MUL ds = —
EI
\M(l - cos 6)dd = 0
Jo
or
-
I M(l cos 6) dd = 0 0>)
= = L = o
J[M?Lds EI Jo\*Mde
ai
EI
or
M dd = 0 (c)
1
Equations (b) and (c) yield
M cos 6 dd = 0 (d)
152 ENERGY METHODS IN APPLIED MECHANICS
f - - $PR -
)\l
dd + NoR cos dd '"sin dd dd =
0)
M0 ±PR
f
0
0
Jo Jo Jo Jm
cos ddd + NoR - cos cos dd
6)
M0
(1
6
Jo Jo
0
d
d
I
J* Jrtt
\
These equations yield
No =
— Mo =
—
.
Lu
4
PROBLEMS
Determine the principal strains, the direction cosines of the principal axes of
1.
5,
6,
ex tz yyz
= -2.
Y„
2. Determine the principal strains, the principal axes of strain in the undeformed
medium, and the volumetric strain corresponding to the displacement vector,
u = x — 2y, = 3x + 2y, w = 5z.
v
angle between the and x-axes be For plane stress, express an, t.
I-
d.
1,
az 1,
5. az
,
f,
P = 30,000 #
Fig. P4-10
10. The bars in the truss shown in Fig. P4-10 have the following cross-sectional
\
areas: A1 = 1 in.2, A2 = At = 2 in.2. A3 = in.*, As = A6 = in.*. The |
elastic behavior of a bar is represented by the stress-strain relation
a = 200,000«^ lb/in.2. Compute the vertical deflection of the right-hand
joint.
11. The numbers adjacent to the members of the symmetric truss (Fig. P4-11)
denote the cross-sectional areas (in.2). Calculate the horizontal displace
ments of the end supports and the vertical displacement of the center joint,
using the stress-strain diagram of Fig. 4-6.
154 ENERGY METHODS IN APPLIED MECHANICS
12. Solve the problem of the pentagonal truss (Sec. 2-4) by the Maxwell-Mohr
method.
Fig. 4-14
14. The numbers adjacent to the members of the truss (Fig. P4-14) denote
relative values of EA . Compute the tensions in the members by the Maxwell-
Mohr method.
15. By Castigliano's theorem, derive a formula for the deflection at the center of
a simple Hookean beam of rectangular cross section with a single load at the
center, taking shear deformation into account.
16. A simple beam with rectangular cross section of depth h and width b carries a
concentrated load P at its center. The stress-strain relation of the material is
a = Ke1^. Neglecting shear deformation, derive a formula for the deflection
at the center by Castigliano's theorem.
20. A steel piston ring with 3.3-in. mean diameter and 0.125-in. square cross
section has 0.375-in. gap at the ends. What tangential force at the ends will
close the gap?
Fig. P4-21
22. Assuming that the shear q is proportional to |sin 0| and that equilibrium
exists (Fig. P4-22), derive formulas for the bending moment and the tension
at the point at which the load is applied. Neglect shear deformation and
suppose that the material is Hookean. Also, assume that the shear q acts at
the centroidal axis.
23. A uniform semicircular arch that is hinged at the ends carries a vertical
concentrated load P at the top. Calculate the horizontal reactions at the
ends.
24. The uniform semicircular member (Fig. P4-24) is clamped at the left end and
is loaded by a force P perpendicular to the plane of the semicircle at the
mid-point. The right end is free. The torsional stiffness is GJ and the
flexural stiffness is EI. Derive a formula for the deflection at point A in
the direction of force P.
ENERGY METHODS IN APPLIED MECHANICS
Fig. P4-25
•
tension N0 at the top. Consider only strain energy of bending. Express the
integrals in terms of L, h, x, y, J^, and Ixx, where (x, y) are the coordinates
of the centroid of the right half of the curve, and
26. The uniform semicircular Hookean bar is clamped at one end, and it is
subjected to torque T0 at the free end. (The vector T0 in Fig. P4-26 represents
27. By the unit-dummy-load method, determine the end moments Af0, M1 and
the end reactions R0, R1 for the clamped beam (Fig. P4-27). Neglect shear
deformation and suppose that the material follows Hooke's law.
28. Using the strain-energy formula for a ring [Eq. (2-28)], derive the formula
for curved beams corresponding to Eq. (4-73). Use the formula to determine
the end deflection of the curved cantilever beam of Sec. 3-6 by the unit-
dummy-load method.
158 , ENERGY METHODS IN APPLIED MECHANICS
29. ax
= ay = az = 0, t„, _ ra —
r„ _ t > 0
Determine the principal stresses and the maximum shearing stress. Deter
mine the direction cosines of the principal axis for which the corresponding
principal stress is tensile.
30. Let (x, y, z) be curvilinear coordinates such that the coordinate lines coincide
with the lines of principal stress. Obtain the differential equations of equi
librium for this case.
5 Theory of plates and shells
This function (strain energy) being known, we can immediately apply the
general method given in the Mecanique analytique, and which appears to
be more especially applicable to problems that relate to the motions of
systems composed of an immense number of particles mutually acting
upon each other. One of the advantages of this method, of great
importance, is that we are necessarily led by the mere process of cal
culation, and with little care on our part, to all the equations and
conditions which are requisite and sufficient for the complete solution of
any problem to which it may be applied.
GEORGE GREEN
[Eqs. (4-3)]. The same argument applies for v2 and w2 in the formulas
for ty and «2 and for vyvz, wvwz, uzux, wzwx, uxuv, vxvy in the formulas for
yv», yzx' yxv The other quadratic terms in u and v are dropped because
they have about the same magnitude as the squares of the strain compon
ents and the terms that have already been discarded. On the other hand,
wx and wv are the slopes of cross sections of the deformed middle plane;
apparently, they may be quite large compared to the strain components.
It will be convenient to denote the displacement of an arbitrary point
(x, y, z) in the plate by (u, v, w) and to denote the displacement of the
corresponding point (x, y, 0) of the middle plane by («, v, w). Likewise,
(*«, *»♦'"• yxv) wi'l denote the strain components at any point (x, y, z)
and (ex, ev, • • • , e^) will denote the corresponding strain components
at the point (x, y, 0). After discarding all quadratic terms from Eqs. (4-3),
except w2, w2, wxwv, we obtain
+ £W*2, «»
= vy + K2' ez
= Wz
(a)
Vy, = ^y + v„ yzx = + Wx, yxy = vx + Uy + WxWy
«, *» ««
- «»
= ey
- zwvv, yxy =- exy
- 2zwxy (5-2)
where
The shearing stresses txi and tv, have already been neglected. In the
stress-strain relations az is also neglected, since it has about the same magni
tude as the external pressures applied to the faces of the plate. In other
THEORY OF PLATES AND SHELLS 161
«. - ~
h,
fax
- vav) + kd
(5-4)
«»
= - K ~ ™x) +
1
y„ =
T
E G
E , , ,
;(«* + vtv) - Ekd
(5-5)
„ E , , . Ekd
1 — v* 1 — v
Ut « ± [a*
2E
+ av*
- 2vaxay + 2(1 + v)tw*] (5-6)
= |V + 2i*A Kl - - + vX«.
e»*
U0 —
2(1 «»)*»]
v
1
(5-7)
The strain energy of the entire plate
is
Integration
is
A
f/i/2 fh/2
= kd dz, = zkd dz (5-8)
0!
00
J-h/2 J-h'2
When Eqs. (5-2) and (5-7) are substituted into the preceding formula for
U, the strain energy separates into sum, U = Um + Ub Ue, where the
+
a
is
h
The term U9 represents the part of the strain energy that results from
162 ENERGY METHODS IN APPLIED MECHANICS
=
j]W + + J(l - v)ej] dx dy
ev'
+
Um 2vexev
+
I™** 2vw**w*y 2(1
If
2(1 37)
1
dy
v,
If lateral distributed load p(x,
y)
applied to the plate, the potential
is
a
= — wp dx dy
Q
+
V
Ub
Euler equations for the integral are the equilibrium equations of the is
V
— - ~ = -7—
+ 6exy vdd0
^
Kl
1/1
v)^
+
1
\
t
+,
(ex vey)
ox dy dx
h
... .dexy
+
. vdd0
d
1
,
(5-10)
+
v)
32(1
V2v2w + (ex2w + v2e,) =
12
+ + -
+
Ti
(1
where ,
12(1
- v2)
The constant D called the "flexural rigidity" of the plate. The last of
is
Eqs. (5-10) has been simplified with the aid of the first two equations.
A rectangular element of the plate with dimensions dx and dy subjected
is
to the forces Nx dy, Ny dx, Nxy dy, and Nvx dx (Fig. 5-1). The tractions
Nx, Nv, Nxv, Nvx are expressed in terms of the stresses by means of the
statical relations
(5-11)
THEORY OF PLATES AND SHELLS 163
Eh
= -
1 — yr2(ex
+ vev)
1 -
Ed0
v
Eh
N„ = •
1 — vi
(ev + vex)
1 - v
(5-12)
N„ = N„ = GheT
dy
dx
Fig. 5-1
'
dx dy (5-13)
dx dy
DV2V2h' +
1 — v V% - p + Nxwxx + Nvwvv + 2Nxvwxv (5-14)
DW*W + __ v201 = p +
FvvWxx + FxxWyy _ 2FxvWxv (5-16)
164 ENERGY METHODS IN APPLIED MECHANICS
he„ =
N
ZZ* = - F-2i G
G
,
—
d2ex- ,
H
d2evt d2e,w
— = w.z — w,,w„„
oyt oxr ox oy
Equations (5-16) and (5-18) are the fundamental relations in the von
Karman theory of plates. The temperature term in Eq. (5-18) was intro
duced by Gossard, Seide, and Roberts (114).
Strain Energy of Bending. The formula for the strain energy of bending
[Eqs. (5-9)] may be written as follows:
Ut = + wvv)2
- 2(1
- vX"M>v„
- O] dx dv (5-19)
The expression K =
w„wvv
—
w^2 is an approximation for the Gaussian
curvature of the deflected middle surface (Sec. 5-6). It satisfies Euler's
equation of the calculus of variations automatically [Eq. (3-20)]; hence
it contributes nothing to the differential equations of the plate [Eqs. (5-10)].
This circumstance reflects the fact that K
may be transformed into a line
integral on the boundary (121). However, K must not be discarded from
Eq. (5-19) indiscriminately, since it may contribute to the natural boundary
conditions. If the edge of the plate is clamped, there are no natural
boundary conditions, and then K may be dropped. Furthermore, the
Gauss-Bonnet theorem of differential geometry shows that K contributes
nothing to the natural boundary conditions if the edges of the plate are
geodesics on the deformed middle surface (121). In particular, if the plate
is polygonal and the edges remain straight when the plate is deformed, K
is irrelevant. This case was noted by Nadai (59).
THEORY OF PLATES AND SHELLS 165
1),
are unchanged. Also, the first two equations in (5-10) are unchanged.
The third equation of (5-10) becomes
+
V\ -
v)
12(1
£
V2V2w
+
(5-20)
ff
D
Equation (5-18) becomes
V2V2F £V200 =
+
(5-21)
0
Accordingly, the coupling between the lateral deflection w and the in-plane
displacements («, eliminated. The stresses due to the deflection w and
is
v)
the stresses due to the displacements (m, are simply additive. The use
t;)
of complex variables for solving boundary-value problems for differential
equations of thetype in (5-20) and (5-21) has been developed extensively
Muskhelishvili and his school (57).
by
In addition to the tractions AT,., Nv, Nxv, defined by Eq. (5-11), the
following integrals of the stresses are employed in plate theory
:
Qx = r„ dz, Qy = tvi dz
-h/2 -hlt
J
J
M,
fxv
= Mvx = ZTxy &
J-KU
The quantities (Qx, Qy) represent intensities of transverse shearing forces
[dimension (F//L)]; the quantities (Mx, My) represent intensities of bending
moments [dimension (FL/L) = (F)]. The quantity M„ = Myx represents
intensity of twisting moment [dimension (F)]. The intensities of forces
and moments on an infinitesimal element of the plate are represented by
Figs. 5-2 and 5-3. Equations (5-11) and (5-22) are retained in the large-
deflection theory. However, then the deformation of the element (Figs.
5-2 and 5-3) must be taken into account. In the large-deflection theory
the obliquity of the forces Nx dy, Ny dx, etc., with respect to the undeformed
middle surface
is
important.
In the small-deflection theory the deformation of the plate considered
is
the equilibrium equations are derived with the supposition that the middle
surface of the deformed plate plane. These equations are obtained most
is
166 ENERGY METHODS IN APPLIED MECHANICS
1
-*~N*
h Ny,
Nzy
Fig. 5-2
f
OxJ-h/S
axdz + l\
Oy J-h/2
rxvdz +
J-h/i
\
*±,dz = 0
oz
(a)
Fig. 5-3
dNT . dN,
= 0
dx dy
Similar results are obtained from the second and third of the equations in
to. |
dN„ = Q
3Nyx
|
3Ny
= Q
dx dy dx dy
&+ + p = 0 (5-23)
dx dy
dMyx
+
dMy _ q
dx dy dx dy
Equations (5-23) are purley statical; they are equally valid for elastic
plates and plastic plates. The boundary conditions, t„ = rn = 0 at
z = ±A/2 have been used for deriving Eqs. (5-23), but, by a slight genera
03r Bx By By*
Ox = _E
— v*\
(ex +
, .
vev)
- Ekd Ez
(w„ + vwvv)
1 1 — v 1 — r
" =
E
_ y2i
(ev' + vex)
- Ekd
-
Ez
- 2 (wvv
+ vwxx) (5-25)
1 1 V 1
Mv =
- D(wvv + vwxx)
- Fft
— v
(5-26)
- D(l -
1
Substitution of Eqs. (5-26) into Eqs. (5-24) again yields Eq. (5-20).
If forces in the plane of the plate and 6 = 0,
there are no boundary
Eq. (5-21) yields F = 0. Then, by Eq. (5-15), Nx = Ny = Nxy = 0, and,
by Eqs. (5-12), ex = ev = exv = 0. By Eqs. (5-25) the stresses are then
determined if w is known. With the boundary conditions, Eq. (5-20)
determines w.
168 ENERGY METHODS IN APPLIED MECHANICS
y = b
1 x =
wz = 0<
*
m = 0
Fig. 5-4
K= I | [i D(wxx + wv¥)2
- D(l - v)(wxxwvv
- wj) - pw] dx dy
Jo Jo
The last two integrals represent the strain energy of the flange [see Eqs.
(2-8) and (2-17)]. Note that wxv(x, b) is the twist of the flange per unit
length.
The terms in Eq. (a) that represent the strain energy of the flange do not
enter into the part of the first variation of V that is expressed by a double
integral. Consequently, the Euler equation of the problem is not affected
by the presence of the flange. Accordingly, Eq. (5-20) determines the
deflection of the plate (with 0! = 0). The natural boundary conditions
are determined by Eqs. (3-21) and Eq. (a). Let us suppose that the forced
boundary conditions are
In view of these conditions, the variation r\, with its derivatives r\x and
r]v, vanish on the edges x = 0 and x = a. Also because of the forced
boundary conditions, rj and rjx vanish on the edge y = 0. Hence, by
Eqs. (3-21), the natural boundary condition for the simply supported edge,
y = 0, is dF/dwvv = 0 or wvv + vwxx = 0. As might be expected, Eqs.
wvv
= 0.
To obtain the natural boundary conditions for the flanged edge, we must
augment Eqs. (3-21) by terms that come from the line integrals in Eq. (a).
By Eq. (a), the first variation of the strain energy of the flange is
+
r]y[-GJwxxy
where F denotes the integrand of the double integral in Eq. (a). Since rj
and rjv are arbitrary, these equations yield the natural boundary conditions
for the flanged edge:
wvvv + (2
- v)wxxv
-—EI
wxxxx
= 0 for y = b
wvv + vwxx
- ^ wx„ = 0 for y = b
(5-27)
wvvv + (2 - v)wxxv
= 0 for y = b
wvv + vwxx
= 0 for y = b
Qx
= -D(wxxx + wxvv), Qy = -D(wxxy + wvvv) (5-29)
VX
= QX + ^,
oy
Vy = Qy + ^
ox
(5-30)
w = 2 2
2. .
sin — sin niry
rnirx
—r
.
(5"32)
m=i n=i a b
Ub = tfabDZ f i^ +
^Ya^ (5-33)
Q = — dx dy
jjpw
JJab
where p(x, y) is the lateral distributed load. Hence, by Eq. (5-32),
In the small-deflection theory forces in the plane of the plate have no effect
on bending. Consequently, if there is no bending caused by heating, the
deflection w may be computed with the understanding that the total
potential energy is V =
Ub + O. The coefficients are determined by the
principle of stationary potential energy, 3 V/da^ = 0. Differentiating, and
subsequently replacing by (m, ri), we obtain
w = poa
— —b'
41.4
„ « sin — sin —
(5-36)
tt4Z)(a2 + ft2)2 a b
Fig. 5-5
"
_ -p0a2b
[a2 + (2
- v)b2] sin
itX
—
a
at y = b
77(a2 + 62)2
-!ws£')'
(5-38)
Body forces will be neglected. Then, by substituting Eqs. (5-38) into the
equilibrium equations (4-24) and simplifying the results by Eqs. (5-23) and
the boundary conditions, we obtain
(5-39)
T =
2£^ JI[N*2
+ ^ ~ 2vNxN" + 2(1 + V)Nx"2
+
jfz jj[Mx2 + M/ - 2vMxMv + 2(1 + v)Mj + \(\ + v)h*
x (Q*2 + Qy2)
- (ivph* - 2Efl1XA#« + M»)] dx dy (5-40)
174 ENERGY METHODS IN APPLIED MECHANICS
Additive terms that do not contain Nx, Nv, Nxv, Qx, Qv, Mx, Mv, or Af„
are not included in Eqs. (5-40), since they are irrelevant. The terms 00 and
0j are denned by Eqs. (5-8).
According to the generalized form of Castigliano's theorem of least
work (Sec. 4-8), the quantity Y
that is to be rendered stationary is obtained
by appending to the complementary energy T a certain line integral on the
boundary of the plate. However, this line integral does not affect the Euler
equations for the double integrals. Consequently, it will be disregarded.*
The first and second integrals in Eq. (5-40) may be treated separately.
When Nx, Nv, Nxy are expressed in terms of an Airy function F(x, y)
by Eq. (5-15), the first integral in Eq. (5-40) becomes
IF J + Fyy2
- 2vFxxFvv + 2(1 + v)Fj
+ (vph + 2Ed0)(Fxx + dx dy
Fvv)~]
The Airy function ensures that the equilibrium equations for Nx, Nv, Nn
are satisfied. The Euler equation [Eq. (3-20)] for the preceding integral is
2
+ My2 - 2vMxMy + 2(1 + v)Mj - (ivph2 - 2E6J
X (Mx + Mv) + 4(1 + v)h
A dx dy / \ dx dy / -
•')}■
\ oar ox oy oy
Writing the Euler equations [Eq. (3-20)] for this integral and simplifying
Reissner included the line integral in order to obtain natural boundary conditions.
*
He also used the line integral to interpret Lagrange multiplier. However, certain
a
approximations (136, 137, 110) concerning the displacement vector at the edge of the
plate are required in this procedure. Reissner did not include Nx, Nv, and N„, in his
theory. Also, he did not include temperature terms.
THEORY OF PLATES AND SHELLS 175
Mx - vMv + £0,
- hvph2 - i(l + v)h2
ox
+ kh* = 0
M»
- vMx + E6,
- hvph2 - Kl + v)h2
d-& +
ay
\Xvv
= 0 (5-43)
- Ml +
y-')
(1 + v)Mxv + + \K =
0
v)/»2(^v
Equations (5-23) and (5-43) yield
-U„
- ft'gg,= + ^»)
- -
£fli vph*
M*
■
3x
v)
-qvv
- -
+
-
£0! h2 dQv
M ?£ft? = vkxx)
I
|
"
v)
xv
10\ dx dy +
v)
2(1
J/
With the boundary conditions, the six differential equations (5-23) and
(5-44) determine the functions Mx, Mv, Mxv, Qx, Qv,
X.
vN vp\ 3vp\
Fc _{NX (12MX 12vMv
+ Ekd
+
2vp
f' h3
F _ (Nv vNx
+
vp\
+
/12M, _ \2vMx _ 3vp\
z
(5-45)
+ 2vp- + Ek6
h3
176 ENERGY METHODS IN APPLIED MECHANICS
dx dy dz
_ "12(1 + v) dQ. ,
6vplz2 _ Ek
86
. h3 dx h3J dz
^(-f)(f+t)-T^
Defining the flexural rigidity D as in Eq. (5-10), and noting Eqs. (5-23), we
may write these equations as follows:
= - D(wxx + vwvv)
= - D(wvv + vwxx)
Reissner (136, 137) identified the quantity 6XjEh3 as the mean value of
the deflection w. This relationship reconciles Eqs. (5-44) and (5-46),
except for the temperature term, provided that the mean value of 1 — 4z2/h2
is taken to be However, according to the usual definition of the mean,
|.
this value should be § instead of f . The temperature terms in Eqs. (5-44)
THEORY OF PLATES AND SHELLS 177
Mx + D(l + v)W +
8(1
—
v)
-TIT = -
4 ox
+ D^w" vvv-)
Mv + D(l + v)k6' + —
~
v¥"V=
4
- D(w» + «J
ay
(5-47)
8(1 v)
"--f(t^)~«'-'>-
where 6' denotes the value of 30/3z at z = 0.
Equations (5-47) were derived by Goodier (110). They are consistent
with the initial assumptions, and they may be used instead of Eqs. (5-44).
With the boundary conditions, Eqs. (5-23) and (5-47) determine the
functions Mx, Mv, Mxv, Qx, Qy, w. Goodier remarked that if z = ±hj2
Eqs. (5-46) reduce to the classical equations (5-26), except for the tempera
ture term, which he did not discuss.
and only
0.
•
rx
rv
The distance ds between points with the surface coordinates (x, y) and
{x + dx, y + dy) is determined by ds2 = dr • dr. Hence, for orthogonal
surface coordinates,
ds2 = A2dx2 + B2 dy2 (5-48)
where
A* = Tx.rx = X*+ Y* + Z*
-
K '
B2=ry.ry = X2+ Y2 + Z2
The area of any part of the surface is evidently determined by
area = dx dy
JJ^.B
For orthogonal surface coordinates, the magnitudes of the vectors
r,.
and are and B, respectively. Therefore, the unit vector normal to the
A
rv
surface
is
r^
n = (5-50)
AB
dx2 + dx dy + = — dr • dn
2/
dy2 (5-51)
g
e
e,
g
by
Y y
Y
yv xy
xxx ZXy
= xx
Yx
Yx
Zx
e
AB AB
Xy
Z»
Yv
(5-52)
y
Y
'yv Z»»
= x.
Yx
Z*
g
AB
Xy
Z»
Yy
r2
if,
with the coordinate lines
if,
coincide and only the coordinates are
orthogonal and = In this case
0.
/
(5-53)
These equations affix signs to the principal radii of curvature, and r2.
r1
They signify that or the corresponding center of curvature
positive
if
is
r1
r2
lies on the negative side of the tangent plane of the surface, the positive
side of the tangent plane being the side toward which the surface normal
directed. For example, for sphere of radius a, = =
if
is
is
a
n
n
r1
r2
directed outward, but = = —a directed inward.
if
is
r2
n
r1
If the lines
of principal curvature are coordinate lines (that =
is,
if
•
rx
rv
= 0), a theorem of Rodrigues expressed as follows
is
/
:
dn _ dr 8n _ dr
1
(5-54)
'
dx dx dy dy
r1
r2
(5-55)
A
is
and do not change the surface bent without straining, since ds does
if
B
is
not change [see Eq. (5-48)]. Therefore, by Eq. (5-55), K does not change
the surface bent without straining. In particular, K = for any
if
is
plane.
a
R = r + nz (a)
Hence, by differentiation,
Rx = r, + a,*, R„
= r„ + n^ (b)
Since the coordinate lines are lines of principal curvature, the Rodrigues
formulas apply [Eq. (5-54)]. By Eq. (b),
B. - (l +
y)rx,
R, =
(l
+
yjry,
R, = n (c)
THEORY OF PLATES AND SHELLS 181
(5-58)
The factors a, are the Lame coefficients [see Eq. (4-27)]. Equations
/?,
y
(5-56), (5-57), and (5-58) are basic geometric relations in the theory of
shells.
Equation (c) shows that the vectors R« and Rv are parallel to the vectors
ix and rv, respectively. Accordingly, all x-coordinate lines (or all y-
coordinate lines) that intersect given straight normal N to the middle
a
a
ential element of shell, cut out by surfaces x = constant and =
y
a
ingly, the coordinate lines on the middle surface are lines of principal
curvature. By Eqs. (5-58) and (5-57), the elements of area of the cross
sections (Fig. 5-7a) are
A^l
dAx = a dx dz = dx dz
+
—J
-
b(i
= $dydz = dy dz
+
dAv
j
where r1 and r2 are the principal radii of curvature of the middle surface.
Let Nx be the tensile force on cross section per unit length of y-
a
a
coordinate line (Fig. 5-76). Then the total tensile force on the differential
element in the x-direction NxB dy. Hence
is
N, = M paxdz =
aJl+-)dz
\
Similarly, the tension Nv, the shears Nxv and Nvx, the transverse shears Qx
and Qv, the bending moments Mx and Mv, and the twisting moments Mxv
182 ENERGY METHODS IN APPLIED MECHANICS
Fig. 5-7c
THEORY OF PLATES AND SHELLS 183
and Mvx are defined (see Figs. 5-lb and 5-7c). The complete set of relations
is
^
Paxdz =
\"12
Nx=l- ax(l + dz
rj
f
-h/t J-h/2
B
J
\
Nv=- = aJl
-)
*av dz + dz
J-h/2 J-h/2
A
\
\
T11
i
»Xv
= B i
V
r2/
= -
r*/a
-z\ r»/2
/
=
i
t„\
*rxv dz dz
+
Nvx
1
J-h/2 J-h/2
A
\
r1/
\
-\J-h/2
r*/2 r*/2
-z\
/
Qx = Prxz dz =
i
dz
\
J-h/2
Txz[l+ (5-59)
B
V
r2/
= - p"2
= r*/2 z\
A i
I/
-J
Q» aT« T« + dz
J-h/2 J-h/2 \ iy
z\
-\J-h/2 fraxdz=\J-h/2 z[\+-\axdz
C"/2 Ch/2
M* =
I
l
B
r2'
\
My =
r*/2
=
ch/2
—z\
i
—
axaydz
1
1
z
crv
I
J-»/2
/4
-h/2
^
T]/
\
P/2 [m -z\It^
\ ZlI
1
Mxv
= ~
fcTxv dz
= + </z
1
J-h/2 J-h/2
B
r2/
- f"/2 f*/2
zll
z\
/
1
r-J
/I
J-/1/2 -»/2
J
\
The positive senses of forces and moments are shown in Fig. 5-7.
Equations (5-59) are also valid for flat plates, with \/r1 = l/r2 =
0.
Equation (5-59) shows that Mxv not exactly equal to Myx, unless
is
Mxv_Myx =
N^_ N^ (5-60)
The element of the shell shown in Fig. 5-7a may be subjected to external
forces caused by gravity and by external pressures and shears applied to the
outer and inner surfaces of the shell. Since the area of the element dx dy
of the middle surface AB dx dy, the resultant external force on the ele
is
ment of the shell will be denoted by PAB dx dy. The vector the result
P
is
ant external force per unit area of the middle surface. function of
It
is
a
184 ENERGY METHODS IN APPLIED MECHANICS
We obtain the equilibrium equations for Nx, Nv, Nxv, Nyx, Qx, Qy by
integrating the differential equations of equilibrium through the thickness.
Consider, for example, the first of the equations in (4-29). The first term
is d/dx(Pax). Integrating this expression with respect to z between the
limits — /j/2 and h/2, and utilizing Eqs. (5-59), we obtain
r
/•ft/2 g
Q fft/2
(bnx)
J -ft/2 5VI
X
OX
X/
OX J -ft/2 ox
r- K») dz = — dz
J-h/ioy dy J -ft/2 <x.txv
dy
ft/2 g ft/2
m 3«
Tm, dZ
J
h/2 dy
J-h/idyL \ r/J
THEORY OF PLATES AND SHELLS 185
Hence
rw 3a rn/2
/dA xdA\
J-hnoy
— Txydz = \
J-h/2\oy
— +-riayl
— .
)rxvdz
oy J-h/2\ r2/
Hence by Eqs. (5-59)
Ch<2 dx , „ dA
T-Txydz=Nxy —
J-hnoy ay
Applying Eqs. (5-58) and (5-59) to the fifth term in Eqs. (4-29), we obtain
r
J-h/2
fl
OZ
rxz dz = r
J-h/2
p±rxzdz =
r1
*
Tj
r
J-h/2
(h„ dz
Finally, by Eqs. (5-56), (5-58), and (5-59), the integral of the sixth term in
Eqs. (4-29) is
r?**-rr[>(,+i)]**
J-M20X r2/J J-h/2 OXL \
»n,
oxJ-hi2\
+
i)v*_w»
r1I ox
With the foregoing results, the first of the equations in (4-29) yields
i- (BNX)
ox
+
I (ANvX) + Nm
oy
~
T
ox
+ — +
r1
= 0
Similarly, the second and third equations in (4-29) yield the equilibrium
equations for forces in the y- and z-directions.
186 ENERGY METHODS IN APPLIED MECHANICS
f
a a a
fh/2 fh/2
z (J}ax) dz=f\ zpax dz=~ (BMX)
J-h/2 OX OX J-h/2 OX
K)
rh/2 a a a
p/2
z r- dz = — z*rxv dz =
— (AMvX)
J-h/2 oy oy J-h/2 oy
fh/2
J-h/2
z -
OZ
a
(a/Sr,.2) dz =
-\J-h/2
fh/2
*pTxz dz + ABRy
'
—
J-h/2
z -T (*frxz) dz + z/S t„ dz
OZ J-h/2 OZ
/ n/2
=
f*/2
-a + za\
— )pTxz dz + ABRv = -A]J-h/2 frxzdz + ABRy
J-h/2\ r1 I
= -ABQX + ABRy
Also,
z— I-
/^i
rh/2 rh/2
aa rdA
— d
ZT-Txydz = \ z + \\rxydz
J-h/2 oy J-h/2 Ldy oi/V/yJ
= f*/2 z
IdA,
r- + -z dA\
— \rxv dz .
=
dA
— f»/s
I.
z[l
, z\
+ -\Txvdz = Mxv
. . . dA
—
J-h/2 \oy r2oy I oy J-h/2 \ r2/ ay
and
d\
T•
x
W f
— , '- x < ' +
oy
\ y j- ' +
(^MvJ • - xy —
• Mra
^
^
oy
- Mv
M —25 - ^BQ, +
V -S
ox
ABRv = 0
THEORY OF PLATES AND SHELLS 187
I
ox
(BNX) + j-
oy
(ANyx) + Nxy^-Ny^
oy ox
+
r1
+ ABPX = 0 (5-61)
I
ox
(BNXv) + I
ay
(AN,) + nJ^-Nx^+ABQx+
ox oy r2
ABPv - 0 (5-62)
Y
ox
(BQX) + f
dy
(AQy) -—
r1
Nx
-—
r2
Ny + ABP, = 0 (5-63)
f
ox
(BMX) + |-
oy
(AMyx) + Mxy^-My^-ABQX
ay ox
+ ABRy = 0 (5-64)
~ (BMxy)
ox
+ f
oy
(AM,) + mJ-£-mJ±-
ox oy
ABQy - ABRX = 0 (5-65)
y»
=^
P
+
».-^
P
(5-66)
yzx = uz + — -—
a a
y:XV
^ + E* - i*l -<h¥ + V»
yS a aj8 a^S AB
188 ENERGY METHODS IN APPLIED MECHANICS
dz\J
+
«2
'
dz V + f?
With Eqs. (5-58), these equations yield
u - —
wx |*
dz v - —wy
|*
dz
Hence by integration
A B
The additive functions, f(x, y) and g(x, y), are determined by the condi
tions u = u and v = v if z = 0. Accordingly, with Eq. (5-58), there results
w = A-^<xu
— zwx), v = B~\Pv — zw„), w = w (5-67)
1 d
(§£
- zwy\ px(<xu - zwx\ pzv> wy2
ex
= ex + ZKx, ev
= ev + ZKv, yxv = «xv + ZKxv (5-69)
Here, (ex, ev, e^) are the values of (ex, ev, y^) on the middle surface of the
shell. The factors (kx, k^) are closely related to the changes of curva
ture of the middle surface caused by bending. Setting z = 0 in Eqs. (5-68),
we obtain
(5-70)
B AB r2 2B2
e,
Equations (5-70) is a generalization of Eq. (5-3) which was derived for flat
plates.
The factors (kx, kv, are obtained from Eqs. (5-68) and (5-69) if the
z-terms are expanded to first powers by the binomial series. The first-order
effects of u and v on kx, kv, kxv represent changes of curvature due to the
circumstance that the elements of a surface are usually bent if the surface
is deformed into itself. For example, if a surface element of a cone is
displaced along a generator, it must be bent to remain in the original conical
surface. Also, if a surface element of a cylinder is rotated about its normal,
it must be bent to remain in the original cylindrical surface. These effects
are quite small compared to the bending that results from the normal
displacement w. Consequently, u and v will be discarded from the formulas
for kx, kv, Also, the bending effect of w results mainly through the
derivatives wx, wv, w^, w^, wvv. There is a slight bending that depends
explicitly on w; it remains if w is constant. For example, a spherical shell
experiences changes of curvature if w is a nonzero constant, since the radius
of the sphere is changed. However, this effect is quite small. Consequently,
h> will be discarded from the formulas for kx, kv, k^, although the deriva
AyWy
Adx\Al
(5-71)
190 ENERGY METHODS IN APPLIED MECHANICS
The stress-strain relations for a flat plate [Eqs. (5-5)] remain valid for a
curved shell. Accordingly, if the terms z/rt and z/r2 in Eqs. (5-59) are
disregarded, Eqs. (5-12) expresses the quantities Nx, Nv, Nxy in terms of
ex, ev, exv. For a curved shell, ex, ev, exv are determined by Eqs. (5-70).
Likewise, when zjr1 and z/r2 are discarded from Eqs. (5-59), we obtain
with Eqs. (5-5), (5-8), and (5-69)
Mx = D(kx + vkv)
- E61
1
My = D(Ky + vkx)
- -^i- — v
(5-72)
1
The quantities kx, kv, /cct are determined by Eqs. (5-71). Equations (5-72)
is a generalization of Eqs. (5-26) which was derived for flat plates. After
Mx, Mv, Mxy are determined by Eqs. (5-72), the shears Qx, Qy may be
obtained from the equilibrium equations (5-64) and (5-65).
The formula for the strain-energy density U0 that was used for flat plates
[Eq. (5-7)] remains valid for curved shells. The volume element of the
shell is a/S dx dy dz. In view of Eq. (5-58), this will be approximated by
ABdxdydz. Then,
rr pi/a
U = dx
JJ\\AB dy\J-h/2 U0dz
When Eqs. (5-7) and (5-69) are substituted into the preceding equation,
the strain energy separates into a sum of three terms, U = Um + Ub + Ug,
where the membrane energy Um is linear in h and the bending energy Ub
is cubic in h. The term Ue represents the part of the strain energy that
results from heating. Again, the notation in Eq. (5-8) is used. Then, if G
and v are considered to be constants, integration with respect to z yields
= —-
JJW +
- v)eJ]hAB dx
ey2
+ dy
+
Um 2vexey £(1
+ tfl - vyJ^AB dx
Uh =
- JJl>*2
+ «v + 2vkxkv dy
■
v)
12(1
= + (kx k^AB dx dy
+
Ue
jjl(ex ey)60 (5-73)
expressed
as functional of «, w.
v,
a
stationary potential energy requires that the strain energy U satisfy the
Euler equations of the calculus of variations for u and v. With Eqs. (5-12),
these conditions yield the equilibrium equations for tangential tractions.
The results agree with Eqs. (5-61) and (5-62), except for the terms con
taining Qx and Qy, which have been lost because of the approximations.
Various refinements of the theory of shells have been developed (102,
124, 129), and many of them possess particular merits, but the foregoing
A = A(x), B = r (5-75)
The positive normal n of the surface is directed outward.
One principal curvature \frl is the curvature of a meridian. The center
of curvature corresponding to r2 lies on the Z-axis. Consequently, with
the notation of Fig. 5-8,
= r
<f>
r2 csc (5-76)
a
r1
d<f>,
= x; then
<f>
Sometimes, coordinates may be chosen conveniently so that
= rv However, for generality, we shall not fix the coordinate x. From
A
a
apparent that
it
is
— =
<f>
cos
A
(5-77)
dx
since the
r,
o
ox dd r1
(5-78)
^8
AN,, + ArPg =
<f>
+
A
0
j>
ox dd
<f>,
and
0,
is
is
is
along
a
are no tangential external forces and Qx and Qe are neglected, Eqs. (5-78)
reduce to
THEORY OF PLATES AND SHELLS 193
a 2a
(5-81)
exe
= vx H r
By Eqs. (5-71)
". = t#
- - ~T a2
. *.» -- ~
a
»>* (5-82)
+ 2H». + J».*)(^
\
+
2aV a
+ Kl - \
»(<>.
+ * + —J
a a /J
+
^r^k2
r)L —
+ w4
a*
+
a'
+ 2d J pw) de
J
12(1
a
(5-83)
If the integrand in Eq. (5-83)
J,
ow
a
owx
a
_ N„we _
dJ
Gh3 WgA
+ dJ_ +
I
=
a2' dwM 6(l-v)a2\a2 "7
3
dWxe
+ — — — = + Nxwxx
+
+
D
~ N«y»« + N,w„ D =
+
5-84)
\
;
cr —
12(1 vl)
a
194 ENERGY METHODS IN APPLIED MECHANICS
= = ^
fc2
weeee + ^wee 0,
D
The general solution of this equation is
= sin kd + cos kd + Cd + Q
B
w
A
is
k
k
0
1
corresponds to rigid-body displacement. Consequently, the buckling
a
pressure
is
4D
Actually, this result too high. The correct buckling pressure, determined
is
qcT
open
the pressure at ordinate x
is
top,
Fig. 5-9 = y(L —
x).
p
THEORY OF PLATES AND SHELLS 195
Ne = (a)
a
w— + w = (b)
75 d
Equation (b) may be integrated readily when the preceding linear expres
sion for p is introduced. The solutions of this equation are discussed in
detail in the book by Timoshenko and Woinowsky-Krieger (86). The four
arbitrary constants in the general solution may be adapted to the two
forced boundary conditions (w = wx = 0 at x = 0) and the two natural
boundary conditions (w^ = = 0 at x = L). The natural boundary
conditions signify that the bending moment and the shear are zero at the
top edge. They may be derived by applying the variational procedure to
the potential energy integral which results from Eqs. (5-83) when simplifi
cations due to axial symmetry are introduced.
0.
By
Eqs. (5-59), = and = Mer.
The equilibrium equations for
tractions in the plane of the plate are obtained immediately from Eqs.
(5-78):
dNr 3Nre N, - Ne _
1
dr d6
r
or od
r
Pe
v
£vvr2,
r
By Eqs. (4-71)
196 ENERGY METHODS IN APPLIED MECHANICS
= + K2)2 + + 2r«(«r + dr
Um j[r(ur ~r i*,1)]
V-^
Mr= -D(w„ +
(5-87)
a
supported circular plate that carries an axially symmetric distributed
load p(r) and concentrated load at the center. If small-deflection
F
a
is
approximations Um 0.
= By Eqs. (5-86) the total potential energy
0.
Ug
is
V= tt
\{D[r(w"f + r~Xw'f
- 2pwr} dr -
+
2vw'w"~\ Fw0
Jo
-
"
= -Fw0 + + tt
\{D\r(w"f + r-\w'f] 2pwr} dr
V
ttvD(w')2 (a)
Jo
o
is
a
a
remains finite that is, the integral proper. The variation rj(r) restricted
is
is
;
+ 2tt
I
Jo
{D(rw"" + 2w"' — r~ V + r~V) - pr]»? dr - 0
?w»'_ivv" Ivv' = £
+
r +
r
w»» (b)
r D
(5-8,)
r dr
If the plate is simplysupported at the edge, the forced boundary
condition for the = 0 at r — a. The condition »v'(0) = 0 may
edge is w
also be regarded as a forced boundary condition. The preceding expression
for dV yields the following natural boundary conditions:
vW + rw" = 0 at r = a (c)
= 2B + 3C + 2C log r -— '
r
w"
. .. 2C . 2C
198 ENERGY METHODS IN APPLIED MECHANICS
--
Hence
rw'" + w" w' = 4C
r
+
Fa2(3
v)
16ttD(1 +
v)
Consequently, by Eq. (e)
(5-90)
It lim = — oo of curvature
is,
is apparent that w" ; that the radius at
a
difference between plate theory and beam theory. In plate theory the
second derivative of the deflection function discontinuous at point of
is
a
concentrated load. In beam theory the second derivative remains con
tinuous, but the third derivative discontinuous at point where
is
a
concentrated load applied.
is
PROBLEMS
A elastic isotropic plate of constant thickness heated. The
1.
is
rectangular
6(x, y). By the principle of stationary potential energy,
is
temperature
derive the differential equations for the displacement components (a, v).
The edge x = free; derive the natural boundary conditions for that
is
a
The edges = ±b are free. The load isp — p0 sin x. Set w = f(y) sin x and
y
obtain the solution of Eq. (5-20) that satisfies the boundary conditions.
Let = 0.
6
3. Suppose that the flange (Fig. 5-4) carries distributed lateral line load q(x).
a
pressure on square region of width 2c that has sides parallel to the edges
p
of the plate and center at the center of the plate. Outside this square, there
no load. Determine the deflection w(x, in the form of double sine
y)
is
13.
a
colatitude.
A, B, a, ljrv for an elliptic cylindrical shell with middle
/S,
0,
0,
0.
b
Ny = A^F^ - A-^AyFy + KF
+
A~3AXFX
NXy = -A^B^y + A-*B~1AyFX + A-^B^y
where F(x, an arbitrary function (generalized Airy function).
y)
stress
is
200 ENERGY METHODS IN APPLIED MECHANICS
17. Specialize the equations given in Prob. 16 for a spherical shell of radius a
<f>,
referred to spherical coordinates (x = = d).
y
18. Let (X, Y) be rectangular coordinates in the middle plane of flat plate. Set
a
X = cosh x cos y, = sinh x sin y. Prove that the coordinates (x, are
y)
Y
c
c
orthogonal and that the x- and y-coordinate lines are hyperbolas and ellipses,
respectively, with foci at the points ±c, 0). Determine and and write
B
A
(
the general solution of the equilibrium equations for Nx, Ny, JVCT by means
of the equations given in Prob. 16.
<f>
Specialize Eq. (5-68) for spherical shells. Let x = = where
6,
is
19.
y
colatitude.
20. Specialize Eqs. (5-68) for conical shells. Let x be distance along generator
a
from the vertex and let =0 be longitude.
y
21. Specialize the strain-energy expression for cylindrical shells [Eqs. (5-83)] for
the case of axially symmetric loading and small deflections (quadratic terms
neglected). Hence derive the differential equations and the boundary
conditions for semi-infinite cylindrical shell that subjected to constant
is
a
a
uniform lateral pressure p. Determine the deflection w(r). Compare with
the approximate solution obtained in Prob. 20, Chap.
3.
23. Solve Prob. 22 for plate with clamped edge.
a
a
Theory of buckling
If the system is influenced only by internal forces, or if the applied forces
follow the law of doing always the same amount of work upon the system
passing from one configuration to another by all possible paths, the
whole potential energy must be constant, in all positions, for neutral
equilibrium; must be a minimum for positions of thoroughly stable
equilibrium; must be either an absolute maximum, or a maximum for
some displacements and a minimum for others when there is unstable
equilibrium.
tive system, the theory of stability developed in Sec. 1-11 may be used to
determine the buckling load.
In practice, the mere existence
of stability provides no assurance of
safe design. What important in engineering the degree of stability.
is
is
Von Karman and Tsien (118) noted that some structures — particularly,
THEORY OF BUCKLING 203
Fig. 6-2
dV = - V' +
fc2
d' d6 sin 66 ds
6)
1 I (h)
Jo
/o
Since the variations of at the end points are arbitrary, Eq. (h) yields the
6
0'(0) = ^=0
(i)
g„ _~ du _du dd _~ "
du
ds~ dd ds dd
(j)
is
s
Hence
^
~ ~ ~ — cos (k)
k\/2(cos
a)
6
is
obviously
is
a)
0
= a when =
by
0.
is
C1
6
Jo Vcos — cos a
Vsin2a/2
•/<>\/sin2a
- sin2 6/2
0)
8
<f>.
obtain
fr,2
kL=2 d*
(m)
Jo — a2 sin2
<f>
V
1
kL =
<f>
• •
<f>
fa4 sin4
(1
2
•)
d<f>
Jo
Hence
A:L = tt[1 + (i)2 a2 + (f)2 a4 + (A)2 a6 + • •
(n)
•]
206 ENERGY METHODS IN APPLIED MECHANICS
If the functions au(p) are analytic in tt, if stable equilibrium exists when
p < A (where X is some constant in tt), if the equation D(p) = 0
* The terms
"positive definite," "negative definite," etc., are defined in the Appendix
(Sec. A-l).
THEORY OF BUCKLING 207
possesses a minimum rootp0 in n, and ifD changes its sign from positive
to negative as p passes through p0, thenp0 is the buckling loadpa.
Fig. 6-3a
Fig. 6-36
constants for the hinges are kx and k2, as indicated in Fig. 6-3a. The
hinges are frictionless and they are free to rotate either way.
The straight form of the linkage is obviously an equilibrium con
figuration. However, if the compression load Fis large, this configuration
is unstable. To determine the buckling load, we consider the system in a
deflected form, as indicated by Fig. 6-3b. The angles a and fl are adopted
as generalized coordinates. The strain energy of the hinges is
sin a = —
m
, sin p = - , sin y
n — m
a 1.5a
where m and n are the deflections of the joints (Fig. 6-36). Consequently,
sin y = \ (sin /3
— sin
a)
«)
(b)
03
7
f
208 ENERGY METHODS IN APPLIED MECHANICS
a = -F(ia - L)
With Eqs. (b) and (d), this yields
V= V0 + dV+$d*V+--- (g)
where K0 is a constant, <5Kis a linear form in (a, /S), <52Kis a quadratic form
in (a, /?), etc. Since Eq. (f) represents V as a quadratic form in (a, /S),
K0
= 0 and dV = 0. Actually, K contains cubic and higher degree terms
in (a, but these terms have not been derived, since Eqs. (c) and (e) are
/S),
A2)
analytic; see Eq. (f)], we conclude from the preceding theorem that the
buckling criterion that the determinant of coefficients in Eq. (f) vanish;
is
W - (ito - W2
is,
that
= (to ff*2 ~ =
D
(h)
0
aF =
A*2
(j) (i)
+
k1
aF = $k1 + lk2
THEORY OF BUCKLING 209
The smaller value of F obtained from these two equations is the buckling
(j)
load Hence Eq. 3^ < lk2 and Eq. 3^ > lk2.
if
if
(i)
FCII. applies applies
To verify that Eqs. (i) and
(j)
provide the buckling load, we must show
that D > < FCI and D <
B,
FCI < < where some con
if
F
if
B
F
is
0
0
stant. This condition easily seen to be true, since the graph of D versus
is
aF the parabola shown in Fig.
is
(10/7)^.
it
if
=
(j)
if
is
/S.
x
if
/3.
3k1 < lk2, and buckles antisymmetrically 3k1 > lk2 (Fig. 6-5).
if
it
Symmetric buckling
Antisymmetric buckling
Fig. 6-5
ample,
a
is
a
ft
= sin ^ (k)
bn
2
V
n=l L
210 ENERGY METHODS IN APPLIED MECHANICS
(1)
Jo 4 n =l
The strain energyof bending and the potential energy of the axial load are
given by Eq. (2-18) and Eq. (d) of Sec. 2-2. The sum of these energy
expressions the increment of potential energy due to buckling; does
it
is
not include the strain energy due to direct compression before buckling.
Also, because of quadratic approximations, the formula for correct
A
V
is
only to second-degree terms. Accordingly,
(m)
is
configuration, although need not be stable.
it
Evidently, d2Vis positive definite all coefficients in Eq. (m) are positive;
if
is
load Per the smallest load for which coefficient in Eq. (m) zero.
is
is
a
Therefore,
(n)
I.
may be written
=
Per 2JEIP
Sec. 1-11; for system with infinitely many degrees of freedom, d2V
is
a
in the sufficiency theory of the calculus of variations are adequate for only
the simplest buckling problems. Consequently, buckling problems are
often treated by means of approximations that reduce continuous elastic
structures to finite degrees of freedom. Although this method is effective,
there is usually no feasible way to estimate the accuracy of the results.
At the buckling load, d2V ordinarily changes its character from positive
definiteness to negative definiteness, negative semidefiniteness, or in-
definiteness. Consequently, we may anticipate that d2V is positive semi-
definite when p = pa. With the hypothesis that the coefficients au in
<52Kare analytic functions of p, it has been proved (Appendix, Sec. A-5)
that d2V is necessarily positive semidefinite at the buckling load. Con- f\
sequently, if p = />„, there exist nonzexo virtual displdceineiits-fox which
^2K_^_Q_b-Ut there is no- vuluaLdispjacement for which d2V < 0. This
conclusion may be written in a slightly different form by means of the
relation V = U + Q., where U and Q are the potential energies of internal
and external forces, respectively. Also, Q = — We, where W, is the work
performed on the system by the conservative external forces. Con
sequently, if/7 — pa, there exist nonzero virtual displacements such that
62U = d2We, but there is no virtual displacement such that d2U < d2We.
If p < pcT, d2U > d2W( for all nonzero virtual -displacements Kelvin
ana Tail (42) expressed the buckling criterion in the form At/ = AWe,
where AU and AWe denote the increments of U and We corresponding to
an infinitesimal virtual displacement. Timoshenko (84) employed the
criterion AU = AWe to solve many practical buckling problems. How
ever, since d V = 0 for the prebuckling configuration and second-degree
approximations were used to compute AU and AWe, the condition
d2U = d2We was actually applied. In 1891 Bryan (98) applied the criterion
62U = 62We to investigate buckling of plates.
It is to be emphasized that if p ^^Jhejelation d2U =d2W& is not
true for all virtuardisplacements ; it is valid onlylor_s£eciaj_yirtual dis
placements that coincide with the actual buckling_pattern. Consequently,
the criterion (Ptr^ SPW, has ufUiirbeen applied in an approximate way
by means of estimations of deflection functions that represent the buckled
form. The relation has also been applied in an exact manner (84) by means
of the condition that the load parameter p shall have the minimum value
that is consistent with a nonzero solution of the equation d2U = d2We.
reduces to the condition that d2U be positive definite. Therefore, when the
buckling load is reached, d2U becomes positive semi-definite; that is, the
minimum value of d2U is zero, and this value is attained for some nonzero
virtual displacements ht. Trefftz set d2U = Q, and he observed that the
condition that Q possess a nontrivial minimum signifies that the variational
equation dQ = 0 be satisfied identically for certain nonzero values of the
functions Variations of are effected by adding arbitrary variations
ht.
Q
to ht. The equation dQ = yields linear homogeneous differential
0
equations (the Euler equations of the calculus of variations). Also, the
resulting natural boundary conditions are homogeneous in other words,
;
certain functions satisfy the natural boundary conditions (or the
if
Q
d
0
has occasionally been used to determine the tangential components of
displacement after an equation containing few undetermined parameters
has been assumed for the normal displacement. a
if
applies generally
Q.
Bounds of the Buckling Load. The set of unbuckled configurations
is
denoted by T. If < pa, the value of relative minimum at the
is
V
a
p
0
x
X,,
hand, pc<$ p, there exist points
if
which lies in S'. This the usual type of approximation the assump
is
T
tions restrict the deformations that accompany buckling, but they do not
restrict the unbuckled form.
Let N' be the points of neighborhood N that he in S'. Let pcI' be the
buckling load, computed with the assumption that the system constrained
is
pa, but the configuration space taken to be S'. If pcI > pcT' and
A is
lies in the range pcI' < < pcI, > for all points in N. Hence
X
V
X p
points
0
deleted neighborhood.
a
THEORY OF BUCKLING 213
conclusion conflicts with the definition of pa', for, when pcI' < p, the
point X0 does not provide a relative minimum to V in S". Consequently,
the inequality pcr > pcI' is impossible.
(2-15) and (2-19), the second variation of the total potential energy is
EI(y'" - n + W = 0 (b)
b
a
/S'
y" + —^ 0, k2 = — (e)
I -P/?. EI
Equation (e) yields
(g)
15 \ 11 EI
Solving Eq. (g) for P, we obtain
n2El
P= Per = p, = (h)
1 + PJX 1}
Ny
i. NxZ
^ ^ ^ ^ ^ ^ ^ ^ ^ y
Ny
Fig. 6-6
edges coinciding with the x- and y-axes (Fig. 6-6). All edges are hinged.
The plate is subjected to uniform compressions Nx and Ny. For con
venience, positive Nx and Nv are considered compressive. This is opposite
to the sign convention in Chap. 5.
THEORY OF BUCKLING 215
v =
JJ K2)
{(«.
+ hwx2f + K2)2 +
+
(vv 2H««
-^-v
+ Ki - +
+
vX«v «» wxwv)2
+ + + 2vwxxwvv + 2(1
- v)^2]} dx dy (a)
Y2
y)
(0,
is
- = -
y)
is
Jf
Similarly, the potential energy of Ny determined. Accordingly, the
is
If
(u, v, w) receive variations f(x, y), rj(x, y), t,(x, y), and receive
Q
U
increments A£/ and AQ. As in Chap. we write
3,
•
•
<5Q £<52Q •
0.
Since
is
vx
«i - +
+
+ iU CJ + 2*JL, + 2(i - dx dy
+
(c>
f2
v>^2]}
«* = (N*
- - tt - (d)
p*
(Ar»
Equilibrium
if
is
is
f
r?
216 ENERGY METHODS IN APPLIED MECHANICS
condition from the outset. Eliminating u and v from Eq. (c) by means of
Eq. (d), and dropping £ and rj, we obtain
.u =
JJ{DR. 2vlJn+ + + 2(1
- V)C„2]
-NXQ-Nytf}dxdy (e)
g - -
'gf
*U ±11
^
= + m*
[MD „']«mi,2
(f)
if,
if,
Apparently d2U positive definite
and only
is
positive. Consequently, buckling occurs when any term in Eq. (f) equals
zero. Supposing that NJNx = = constant, we may express the critical
r
(Nx)CI = (g)
kt^
Letting a\b = c, we obtain from Eq. (f)
(m2 + «2c2)2
c\m2 + n2c2r)
Equation (h) means that the positive integers m and are to be chosen to
is n
0
r
integer to c.
displacements of the middle plane are denoted by u(r) and w(r), where
r
the radial coordinate (Fig. 6-7). The radius of the plate a. By Eqs.
is
is
(5-86),
*
[K«,
= + h»r2?
U
+
+
2vu(ur
"Jo"{^72 K2)]
- + dr
+
D(rw„*
+
wr2 (6-1)
2vwrwjj
THEORY OF BUCKLING 217
+ D(r^ + Ur2 +
2vUn)]dr
(a)
Fig. 6-7
where £(r) is the variation of w.
Before buckling, Nr = Ne = N and er = ee
= ur. Hence by Eqs
(5-17)
Q = d*U =
2nD^(-o?rtf
+ t& + Ur2 +
2vUr)
dr (c)
_ddF d?dF
drdtr dr*dtrr
dF d dF
(d)
(c)
218 ENERGY METHODS IN APPLIED MECHANICS
Since £r vanishes at the center of the plate, only one natural boundary
condition is obtained at the center. Letting £ take a variation <5£ and
integrating by parts as in Sec. 3-2, we obtain the natural boundary con
dition,
r£rrr]
tyr
lim =0
+
£rr
[(«2r
is
(e)
stitution = we obtain <f>,
£r
+ - 1)* =
+
(f)
0
r2<£rr r<ff>r (a2r2
is is
a
surface. Equation (f) Bessel's differential equation of the first order.
The solution, which regular at the center,
is
is
(7)
= AJ^r) (g)
^
is
A
J1
is
a
=
<f>
gration. Equation (g) satisfies the center condition
0.
If the plate clamped at the edge, the forced boundary conditions at
is
= are = and = If
<f>
it
is
0
a
r
it +
is
r<f>r v<f>
dition = irrelevant, since determines the constant of inte
is
merely
0
£
0.
0
whence JVCI
= 4.20Z)/a2.
symmetrical cross
is
a
by
* An approxi
are hinged the hinge lines lie in the plane of symmetry.
;
mation that reduces the beam to a system with two degrees of freedom is
employed. y
M
~T
4E
cc
h
3E^
Fig. 6-8
length
Poisson's ratio.
v,
ac
moments of inertia of the cross section about principal axes
Iv
I2,
through the centroid.
torsion constant for the entire cross section.
J,
TTX
w = by) sin
+
(a
(a)
hinged ends. Also, conforms to the observation that the cross sections
it
admitted (84).
220 ENERGY METHODS IN APPLIED MECHANICS
are rotated and translated without distortion. With Eq. (a), Eqs. (5-9)
yield the following formula for the strain energy of the web:
i --(»-)
The curvature of the compression flange due to sidewise bending is
v =h
^2 tJ.X
= — 0 + bh) sin —
L2 L
Hence the strain energy of the compression flange due to bending is
W (L± dx =
'Jo
Rt2 4L3
wxv = —
Ttb
cos
tTX
,
y
Hence, by Eq. (2-17), the strain energy of the flanges due to twisting
is
D 4L
Jo
is
about the «/-axis, since otherwise the strain energy of buckling would be
by
augmented
Collecting the preceding results, we obtain the following formula for the
second variation of the strain energy
:
(c)
41? 4L3 4L
More general results for beam with an off-center web were derived by the author
*
(123). Goodier has investigated the stability of any cylindrical or prismatic elastic bar
subjected to pure bending (112).
THEORY OF BUCKLING 221
= _ (d)
jea„dA
V-*r(»."
Jo
+ w.Vx (e)
— bz sin —
A2e =
4L
|W + (a + by)*] (f)
= -
^J|>V + (a + byf](y - ct) dA
4L Jo
222 ENERGY METHODS IN APPUED MECHANICS
Here, the notations tf and cf on the integrals denote "tension flange" and
"compression flange." Evaluating the integrals, we obtain
=
-^{[y(»- 2c<) + ceAc
-
ctA^a*
+ +
[!j(2h-3ct) 2hceA^ab
- -
[jy
h\Ac ctIt + ceIc
bj
+
+
(3ft 4c4) (g)
A
= ^ (h)
<52Q,
(i)
2a12ab a22b*
:
11
-
hi
Mt - ccAc)
+
^
T -
Tl
-
" Ac +
Jc
J<
«22 (,2r2
•1"
o/l „\J,4 •t• U'
1-3
2(l +
,
ft/ ft5V"
ft3
12ft
\
!^ ^ _ W2 _ _ K±
-
£.A
= +
fl
12
ft/
ft3
The buckling criterion that the determinant of the matrix (ai3) shall
is
is,
vanish that
;
(j)
and yields
approximately
c\6ft ft2/
+
v)fc4
12
2/i2L2
2ft2 '\6ft W
Hence, by Eq. (k),
'
*
ft2L2U2L2 2(1 + v)/j4J 4A4L* \6h
For a beam with flanges, the moment of inertia of the cross section of the
web about its vertical center line ordinarily negligible. Consequently,
is
\Iv = \Iv Moreover, the moment of inertia of the cross section of the
/„
Ach2 h3t
= —
,
1
2
12
2
e<
= •
277 4L2 2(1
4-i^Hi 'Wf
+
a
MCI =
+
(6-2)
v)J
2JL2
L
L2(l
L
This result agrees with the classical formula for the critical bending moment
of doubly symmetric I-beam (84).
a
and (k).
For example, beam with the cross section shown in Fig. 6-9 considered.
is
a
is
lations yield
I2
cc
h
= = = 0.030
0,
0.0000180,
/, Jt
Jc
At =
-
0.0600, = 0.0200, = 1.000
h
a22
an = -0.0090ec
224 ENERGY METHODS IN APPLIED MECHANICS
By Eq. (k),
219(10.4
- 3000ec) = 81 X 106ec2
The positive root corresponds to bending for which the flange is placed in
tension; its value is ec = 0.00262. Consequently, MCI = 0.0000236£.
This agrees very closely with the result obtained by Goodier's theory (123).
Equation (6-2), which applies only
for beams with equal flanges, gives
about twice the correct result in
so that the
z-axis the centroidal axis. The ends of the bar lie at = and = L.
is
0
z
Modes of buckling are considered for which the transverse stresses ax,
Consequently, by
Eq. (4-50),
+ AU = iGjJJW yj y„a] dx dy dz
+
+
U
v)e? (a)
z)
terms in e, we obtain
+ G + + (£, U2 U2] dx dy dz
+
[2(1 v)?22 (V. (b)
0.
£2
is
represented by the function To represent translation and rotation of the
£.
= a + yy, = - yx (c)
fi
£
r\
£
r\
d2U. The second integral in Eq. (b) denoted by A. The first variation of
is
6A = v)t, Q di. dQ dx dy dz
Q
+ +
+
d1, (rjz
.(f
2
J]J[2(1
Transformation of this integral by integration-by-parts [Eq. (4-62)]
yields
dA = v)U dx dy dz
dl
-zJJJk.. +
+
lvv 2(1
U
jj
m(t), +
+
2(1
2
surface of the bar. The surface integral extends over the lateral surface
and the ends of the bar. On the end faces = m = and on the lateral
0,
/
=
0.
surface
n
The volume integral in Eq. (d) must vanish for all variations d£. Con
solution of the differential equation,
is
sequently,
a
£
in + + v)t„ = (e)
+
C« 2(1
0
226 ENE8.GY METHODS IN AJKJED MECHANICS
The latfraJ surface of the bar is free from constraint. By Eq. <d i. the naiural
boundary condition for the lateral surface is
an as
where dL,dn is the directional derivative along the outward normal to the
lateral surface, and d]ds denotes the derivative with respect to arc length on
the periphery of the cross section.
If the
column is clamped at the ends, the end conditions are z = 8 =
7 = 4 = 0. If it is loaded by point forces at the centroids of the end
sections and rotation of the ends is prevented (pin-ended case), the
quantities (a, B, again vanish at the ends and Eq. (d) yields the natural
-/)
it
is
a = a sin — = sin — = sin — (g)
y
B
c
,
,
L
L
where (a, are constants. These functions satisfy the end conditions
6,
c)
= = 0. A solution of Eqs. (e) and (f) that satisfies the natural end
Tj
f
condition =
is
£z
= (h)
F(*,j,)cos^r
S
F« + F„-2(l + =
(i)
v)^F
0
^=
an as
+ +
L
(i)
b,
c).
a
Equation (b) now yields the following expression for d2U an irrelevant
if
d*U = -e\.A(a2 + -
1
+ F* dx dy + -
+
/pc2]
JJ
!\
f>2)
+
v)
2(1
THEORY OF BUCKLING 227
is,
d2U = ana2 + + +
(1)
+
+
a22b2 a^c2 2a23bc 2a31ca 2a12ab
if
the last term in Eq. This appears to be reasonable
(i)
is is
neglected.
a
approximation,* since ordinarily large compared to the dimensions of
cross section. L
convenient to introduce the substitution,
It
is
a
by + (m)
(i)
neglected,
is
-
+
V** (n)
0
and the boundary condition
is
—=
+
y*) (o)
an as
2
In view of Eqs. (n) and (o), the function Y identified as the warping
is
function of the Saint Venant torsion theory (51). The solution of Eqs.
(n) and (o) contains an arbitrary additive constant. This loss of uniqueness
results from the discarding of the last term in Eq. (i). The additive constant
must be chosen to minimize d2U. This condition satisfied the mean
if
is
of T
is,
= (p)
jjwdxdy
0
b2)A c2/P]
I x)2]}
~ dx dy
T^— Vf + 0Fv + (q)
+
[Q¥x
2(1 +
v)
- yf x)2] dx dy
=
+
(r)
J
JJ[0^
An example treated in (122) shows that the last term in Eq. (i) introduces only
a
*
dx dy
r=jjv*dxdy,
(s)
R,=jjxVdxdy
Consequently, the coefficients in Eq. are
(1)
AI*e AL*e
aU = '« =
,
°22
,
=
JI3 _ IpL*e
+
r
°33 (t)
2(1 + 77*
where /„, /w /„
are the moments and product of inertia with respect to
the x- and y-axes. to be noted that au = the section
if
It
is
is
symmetrical
about the x-axis and = 0
symmetrical about the y-axis.
if
is
it
0
11 "12 "13
= (u)
0
l«31 fl32
a
by
With the coefficients given Eq. (t), Eq. (u) identical to the result of
is
Kappus. To apply Eqs. (t) and (u), we must have practical way to
a
The conditions an = and a22 = yield the Euler formulas for flexural
0
fCI = +
f
(v)
is
is
/
/
a
length
b
THEORY OF BUCKLING 229
PROBLEMS
1. The frictionless pin bears on the rigid semicircular body which is supported
by perfect rollers (Fig. P6-1). If the body moves horizontally a distance s,
the restoring force of the two springs is ks, where A: is a constant. Compute
the load Pcr at which the body slips from under the pin.
1
mw 1588M
i
B.P6-1
Solve Prob. 1 for the case in which the profile of the body is the curve
x* + y* = a* and the springs provide the nonlinear restoring force ks3.
$3. The bar is rigid and the hinge is frictionless (Fig. P6-3). The hinge contains a
spring such that the moment resisting an angular deflection 6 is kd. Compute
the buckling load PcT, supposing that the force P remains vertical when the
bar rotates. Compute Pcr, supposing that the force P rotates only one third
as much as the bar.
mmmmt.
Fig. P6-3 Fig. P6-4
230 ENERGY METHODS IN APPLIED MECHANICS
Fig. P6-5
[EI(y')2 -P(y')2]dx
11. The potential energy of a system with two generalized coordinates (x, y) is
V = (2 — 6p)x2 — 4(1 — p)xy + (9 — 2p)y2, where p is a load parameter.
Determine the value of/> for which the configuration x = y = 0 ceases to be
stable.
12. A uniform upright elastic column of length L is clamped at the bottom and
free at the top. A vertical compression load F
acts at the top. Assuming
that the deflection y of the buckled column is given by y = ax2 + bx3, derive
a potential energy expression correct to second powers of (a, b). Hence, by
investigating d2V, calculate the buckling load FCI. Compare with Euler's
solution.
13. A flat uniform rectangular elastic plate is subjected to axial compression of
intensity in the x-direction. The edges x = 0 and x = a are simply
supported. The edges y = ±b are clamped. Set w = f(y) sin mrxja and
derive the characteristic equation that determines NQI by means of von
Karman's large-deflection equation (5-14). Hint. N/D > rPiPja2.
14. Study postbuckling behavior of a simple column by means of the assumption
that the axis of the buckled column is a circular arc. Compute F/FCI for a
30° rotation of either end and compare with the result obtained in Sec. 6-2.
15. A flanged rectangular plate is simply supported on the edges x = 0, x = a,
and y = 0. The flanged edge (y = b) is free. The flange is subjected to an
232 ENERGY METHODS IN APPLIED MECHANICS
W. R. HAMILTON
X, = fy(xl, x2,
• • •
, xn), ^»
= • ' •
, xn),
(a)
• • •
Zv — hjx}, x2, , £n)
Such equations exist for each particle; they are mathematical expressions
of the constraints.
233
234 ENERGY METHODS IN APPLIED MECHANICS
I
V
T = |2mv(^ + f,2T^) (c)
l
v=
T = hl ibifrx, (7-1)
of
freedom
is
a
a
quadratic form in the components
of
of
that the generalized coordinates specify the configuration the system
with respect to the Newtonian reference frame for which the kinetic
is
energy computed.
velocity
figuration of the system relative to reference frame that itself moves in
a
a
prescribed way with respect to the Newtonian reference frame for which the
kinetic energy computed. For example, block of mass m slides on
if
is
the elevator x, and the speed of the elevator relative to the earth
if
is
is
v,
x2).
Although this expression in the generalized velocity x,
is
quadratic
it
a
It
is
is
is
a
always computed with respect to some Newtonian reference frame, but its
value depends on the choice of that reference frame. Likewise, work
is
a
a
Hamilton's principle and the equations of lagrange and Hamilton 235
/a.
motion to be determined. Instead of concentrating on this
is
•subsequent
problem, which poorly formulated for variational treatment, Hamilton
is
" The
and txv are known? significance of this problem clarified we
if
is
t0
solve Hamilton's problem for simple special case — particle that moves
a
a
without the of Let particle lie at
in
a
a
point (x0, y0) at given instant t0, and let lie at another given point
it
a
b,
c,
d
x
= = = =
d;
+ +
b,
b,
+
Hamilton adduced the principle of D' Alembert, which asserts that any
law of statics becomes transformed into law of kinetics the driving forces
if
a
with the true course of the motion. To avoid confusing the virtual dis
placement with the actual motion, we suppose that the performance of the
virtual displacement consumes no time. In other words, the time variable
/ is conceived to remain constant while the virtual displacement is executed.
An equivalent point of view is that the real motion of the system is stopped
while the virtual displacement is performed. However, we must suppose
that the inertial forces corresponding to the real motion persist during the
virtual displacement, since we wish to calculate the virtual work of these
forces.
When the system experiences the virtual displacement, the rectangular
coordinates(x, y, z) of any given particle receive infinitesimal increments
(f , r), To first-degree terms in (f, rj, £), the work that inertial forces
£).
is
a z)
a
+
is
is is
+
(x
£)
y
it
(f
rj, £),
,
variation of the virtual work of the inertial force of the particle. Con
sequently, the first variation of the virtual work of all the inertial forces
of the system
is
—
w(*£ Pi + &0
+
Z
where the sum extends over all particles. As in Sec. 1-6, the first variation
of the virtual work of the noninertial forces denoted by W. Extending
is
d
the principle of virtual work to kinetics by augmenting 6W with the virtual
work of the inertial forces, we obtain
dW-^m(xi + yrj zO =
+
(a)
matical transformation of Eq. (a) which renders the equation more useful
for analytical purposes. He supposed that at the time the system has
a
/0
(where given
a
tx
/x
single configuration
F
corresponds to each value of in the range (t0, fj). The set of con
X
F(g = F*(/0) = X0
Hti) = F*(f,) = X1
£)
r],
on the virtual displacements. Since (f, r\, are functions of time, the
£)
+
2
m(x£ (c)
0
at
system
T= £l>(i2
y2
+ + z2)
Consequently, as the system describes the varied path, its kinetic energy
point X*
is
at
T+ Ar= If + tf)
(y
+ + +
+
(i
T))2
(£,
is
t)
T. Hence
238 ENERGY METHODS IN APPLIED MECHANICS
Hamilton disposed of the last term in Eq. (d) by integrating the equation
with respect to t between the limits f0 and tv Since the last term in Eq.
(d) is a time derivative, it may be integrated explicitly. Since S vanishes
at the end points of the path, the functions (f , r\, £) vanish at the times t0
and tv Consequently, integration of Eq. (d) yields \ ' yj? «A
I to
(dT + dW) dt = 0
Q {
A
•
(7-2)
where 6 is the angle through which the wheel has turned and lis the moment
of inertia of the wheel about its axis. The Euler equation for this integral is
0 = 0. Consequently, 6 = constant; that is, the wheel rotates at constant
angular velocity.
I (F - mi) -drdt = 0
J
A=\ Ldt,/ L\=T-
,
dA = 0, V (7-3)
Among all motions that will carry a conservative system from a given
initial configuration X0 to a given final configuration Xx in a given time
interval (t0, tj), that which actually occurs provides a stationary value
to the integral A.
(7-4)
<f>)
x = /(sin 0 + sin
<f>)
= /(cos + cos (a)
6
Hence
<f>)
x =
10
cos cos
+
0
<f>)
= -10
<f>
sin + sin (b)
y
0
The kinetic energy of the lower bob
is
=
+
(<f>
<j?
200 cos
+
+
0)]
system
-
(<f>
6<f,
=
|02 + cos
T
<f>)
— mgy - + cos
—mgl(cos
6
<f>)
(d)
0
-
(<f>
<f>)
ml2
L
\<j?
+
to]
0)
(2
(e)
0
.
I
^
Hamilton's principle and the equations of lagrange and Hamilton 241
- ' -
d<j>
dt 30 30 dt d<ff>
^
<f>2
(<f>
+ cos (4>-6)- sin sin =
+
20'
0)
0
0
$
(<ff>
- -
(<f>
=
<j>
<f>
62
^
+ cos + sin + sin
0)
0)
6
0
Equations (f) are complicated nonlinear differential relationships. In
the classical theory of vibrations certain linearizing approximations are
introduced. In the present case these approximations are equivalent to
-
<f>,
(<f>
<ff>
(<f>
cos — = sin = sin = and sin — =
02
0,
1,
0)
setting
0)
0
<j>2
(<f>
a
cal system to constraints that with time in
is
a
way. For example, certain parts of vibrating system may be subjected
a
to given oscillations
scribed movements of these parts are time-dependent constraints.
Hamilton's principle remains valid for systems with time-dependent
constraints, provided that the terminal configurations (X0, Xj) and the
variations are restricted in accordance with these constraints, since such
S
ever,
weperform against noninertial forces in giving the system any displacement
may conceivably be independent of the path. Then 6W = —6V, and
V
is
potential energy
it
time. Also, the initial configuration X0 that serves as the zero configuration
for potential energy function of time. Consequently, the potential
is
a
energy
irrelevant, insofar as Hamilton's principle concerned, since does not
is
it
= constant.
T
V
+
242 ENERGY METHODS IN APPLIED MECHANICS
at time / is cot (Fig. 7-3). A body of mass m slides without friction on the
bar. The body is subjected to the force of gravity and to the action of a
spring which attracts the body toward a point P on the bar with a force kx,
where x is the distance of the body from point P and k is a constant. The
point P lies at distance a from the center of rotation O. The motion of the
body is determined when x is expressed as a function of t, since the
orientation of the bar is a known function of /. Accordingly, with respect
to the rotating reference frame, the body has a single degree of freedom.
The requirement that the body shall rotate with the bar is a time-
dependent constraint. The potential energy of the body is calculated for a
fixed configuration of the bar. If the bar is fixed in its position at time t,
the potential energy of the body is
The origin might be chosen equally well at the center of rotation. Then
V = —mg(x + a) sin cot + %kx2
However, the additive term —mga sin cot contributes nothing to the
equation of motion that is derived by Lagrange's equation.
The kinetic energy of the body with respect to the fixed reference frame
is
T= \m[(a + zfco* + x2] (b)
Hamilton's principle and the equations of lagrange and Hamilton 243
(c)
(7-7)
Differentiation yields
dH
—
at
=Lpixi + l p&i
„ dL ~ 3L dL
-Z-^-xi-ZT7xi--^
oxi oxi
.
ot
..
dH dL
244 ENERGY METHODS IN APPLIED MECHANICS
"
dL dT .
P<
= T7 = = 2 Kxi
oxi oxi 3-=i
t
system (Sec. 1-9). The condition H = constant then expresses the law of
conservation of mechanical energy.
to be emphasized, however, that
It
is
H not generally the total mechanical energy the coordinates xt specify
if
is
the configuration of the system with respect to reference frame that itself
a
has prescribed motion with respect to fixed reference frame for which
a
a
and are computed.
T
,
;
be function of m variables and set = (dF//du^w. The notation
+
n
a
vt
regarded as
is
variables (ut, w,) and that the w's are held constant when the partial
differentiation performed. Inversion of the foregoing equations yields
is
ut
and w's. function of the d's and
F
w's. Set
= - F(v,
w)
w)
2
»)
regarded as
is
Since (dF/du^ = vk, substitution of the last relation into the preceding
one yields
(c)
t).
identify the variables in the preceding Legendre transformation in the
following way: ut = xt; wt = xt\ = L(x, x,
t).
Then =
i\ dL/dxi = pi = generalized momentum,
— — = = Hamiltonian
and H(x,
G
L
2/Vc< p,
t)
function. Consequently, by Eqs. (b) and (c),
3H 3H dL
(d)
dp, dxt dxt
dt dpi dt dxt
Pi(t), p2(t\ •
pn(t). The first
•
set of Hamilton equations (dxjdt =
•
is
that expresses physical law, since this set that derived from
it
is
is
a
Lagrange's equations.
Hamilton's equations are important in statistical mechanics and in some
other branches of physics. Jacobi (48, 90) developed an extensive theory
for their integration. Because of their simplicity and symmetry, they are
better suited for general theory of integration than the Lagrange equations.
a
/.
a
The dihedral angle between the plane AOB and fixed plane containing
a
<f>.
<f>.
with Eq. (7-9), set xl = = The kinetic energy of the pendulum
is
0, x2
<f>2
= \ml202 sin2
0)
(e)
The potential energy
is
= -mg/cos (f)
6
Since = -
V,
Eq. (7-5) yields
T
L
0
ml2<j>
is
is
H = of Eq.
cj>
and are eliminated by means
if
V.
+
Consequently,
T
6
(g),
2
H = 2 mgl cos
+
(h)
0
6
2ml2 2ml2 sin2
0
Accordingly, Hamilton's equations (7-9) yield
-£l, —
Pi
=
l
Y
0
(0
2
p2COSd d n
pl = 2_2c;„3p-mg/Sin =
•
^
6,
i
0
:
m/2 sin3
0
These are four differential equations for the four unknowns, 0(0, <£(0.
piO), MO-
The equation p2 = yields />2 = constant. This result easily inter
is
0
preted, since Eq. (g) shows that p2 the moment of momentum of the
is
pendulum about the vertical axis OA. By Eq. (g), the equation p2 =
constant yields
= constant
(j)
sin2
0
$
yields
cos
g
x p22 . n
0
K
'
m2/4 sin3
6
~ w' ~ ~ ~ W
~dt
dd dt dd
Hence, since p2 constant, integration of Eq. (k) yields
is
=
6*
+ cos constant
+
^
(1)
™P*\
0
mlr smr
a
0
is
solution or
is
0
0
0
Hamilton's principle and the equations of lagrange and Hamilton 247
<£2
Elimination p2 from Eq. (k) by means
6.
(g/l) sec
=
<j>
Another solution obtained Then the pendulum oscillates
0.
if
is
fixed vertical plane. In this p2 = Consequently, Eq.
0.
in
case
(1)
a
is
60),
60
6
7-7. LAGRANGE'S EQUATIONS FOR NONCONSERVATIVE SYS
TEMS. If the coordinates of
mechanical system receive infinitesimal
a
virtual increments &r<, the virtual work, of all the forces that act on the
represented by linear differential expression Eq. (1-6)]:
is
system
a
[see
l
i
The components Qt of generalized force may be functions of (xv x2, • • •
,
xn; t). If there are no nonholonomic constraints, the variations dxt
are arbitrary functions of
/.
' • •
ixn
•
0
xn
,
'i
,
to
If the system holonomic, this relation true for all variations <5xt
is
is
is,
f[£i_!z: = (mo
e<
dt dxi dxt
are conservative and others are not. For example, vibrating system may
a
contain springs and weights. The forces resulting from these parts are
derivable from potential energy function. However, frictional forces in
a
Consequently,
it
oxt
248 ENERGY METHODS IN APPLIED MECHANICS
where V is the potential energy function for some of the forces and
/?,
the cpmponent_of the remaining generalized force. Then, =
is
if
L
—
V,
Eq. (7-10) becomes
T
dL dL _
R
dt dxt dxt
Utilizing Eq. (7-12) instead of Eq. (7-4), we may also augment Hamilton's
equations (7-9) by the term Thus we obtain
d±=_dH
d^M
+
Ri (?13)
dt dpi dt oxi
d
force W may be calculated with the understanding
F
dy. Consequently,
is
that the potential energy of the force — Fy. Although potential energy
is
a
a
is
is
is
located
where (Fx, Fv, Fz) are the components of the force that acts on the particle.
Also, = lm(x2 + + Eq. (7-10) yields Newton's
y2
T
z2). Consequently,
equations, Fx = mx, = my, Fz = mz. Accordingly, Newton's law
Fv
x = — sin = cos
6,
(a)
y
6
z
Consequently,
x = - cos = -Id sin
16
6,
(b)
y
6
z
y2).
is
^m(x2
= \m(z2 - im cos +
T
l262) (c)
6
Suppose that the hinge contains spring that exerts restoring moment
a
a
Hamilton's principle and the equations of lagrange and Hamilton 249
Fig. 7-5
kd and a damper that exerts a resisting moment c6. The potential energy
due to the spring and the weight is
d_dL-dL=~ - g
dt 36 36
Setting L = T— V, we obtain
If the pivot moves with constant velocity (z = 0), Eq. (e) is the same as
though the pivot were stationary. This conclusion exemplifies the fact that
a uniform velocity of translation does not alter the behavior of a mechanical
£)
and let them also have origin O. The direction cosines of the
(f
rj,
£)
,
axes with respect to the (x, y, axes are represented by the rows in the
z)
following table
:
TABLE 7-1
X
z
ll m, »i
f
n*
I*
n
I,
m3 «a
£
Each entry in this table the cosine of the angle between the axes desig
is
nated at the top of its column and the left end of its row.
Let unit vectors along the (x, y,
(i,
be
(f
rj,
c)
j,
z)
Q
,
axes, respectively.Since the projections of unit vector on rectangular
a
i/1
+
a
knj, = a/i + + c/3, etc. Since = • = etc., these equations
1,
1,
•
a
a
bl2
i
i
i
V = =
/j2
/22
/32
yield
If
1,
1,
of
the squares
of
sum
is
1.
Also, = = =
0;
since + +
0,
• etc.,
0,
+
•
a
ntn2
b
l2m2 +1 3W3
of
ing in
is
b
a
a
c
= — m3n2, = — n3l2,
nl = l2m3
—
- - «i4, -
m2n3 m1 n2l3 l3m2
'i
min3. m2 «2
l2
= m1n2
—
m2nv m3
= n-J2
—
n2lv n3
= ljn2 —
l2m1
l3
.
b
b
a
a
c
c
1
1
1. ,
is
by
(8, y>), known as "Euler's angles." These angles are illustrated by Fig.
The angle between
by
axes
z
£
Hamilton's principle and the equations of lagrange and Hamilton 251
<f>.
angle between the planes xOz and zOt, is denoted by The dihedral angle
between the planes zOt, and 0£ denoted by y. In Fig. 7-6 these angles
is
£
are represented with the aid of arcs on sphere with center at the origin.
a
Arc A is the intersection of plane f0£ with the sphere.
are ordinary spherical coordinates of the £-axis,
<f>
Since 6,
<ff>,
<f>,
= sin cos = sin sin = cos (a)
6
l3
0
m3 n3
Fig. 7-6
Since the dihedral angle between the planes z0£ and £0£ the same as
is
angle between the normals to these planes, the angle between the
y>
the
is
77-axisand the normal to plane zOt,. The direction cosines of the normal-
<f>,
<f>,
/2
Also,
y3 + m?Ph = W
+
«2*3
and
+ + V=
/22
(d)
1
m22
The quantities (/3, m3, n^) may be eliminated from Eq. (c) by means of
Eq. (a). Then Eqs. (b), (c), and (d) may be solved for (/2, m2, n2). Thus the
<f>,
direction cosines (/2, m2, n2) are expressed in terms of (0, y>). Then
are expressed in terms of (6, by Eq. (7-14). The
y>)
results
j>,
(d, im15
are assembled in Table 7-2.
252 ENERGY METHODS IN APPLIED MECHANICS
TABLE 7-2
X <J r
6 <f>
0 <f> 0 <f>
sin sin — cos sin
<j> v
sin cos
v
1 — cos — cos
cos cos sin cos
y
y
- cos
^
sin cos cos
y y
y
y y
— sin sin
y
0
n
y
0
sin cos sin sin cos9
y
6
0
^
I
may be verified that the sum of the squares of the terms in any row or
It
any column of Table 7-2 and that the sum of the products of corre
is
1
sponding terms in any two rows or any two columns 0. Also, the vector
is
product of any two rows yields the other row, and the determinant of the
matrix
is
.
1
TABLE 7-3
X
2
- cos (y + y) - sin (y + y)
-
0
(y
sin + y) cos + y)
0
V
1
i
=
<f>
Table 7-3 shows that the coordinates and are not determined
if
y>
0
6
uniquely by the direction cosines given in Table 7-1. Also, Fig. 7-6 shows
=
<f>
tj,
if
angle are
(f
is
0
0
,
scribed in rigid body that rotates about the origin, the Euler angles are
a
computational
a
6
definition given in Sec. 1-2, the Euler angles cease to be regular coordinates
when = since there no longer one-to-one correspondence between
0,
is
a
0
<f>,
motion of rigid body with one fixed point vectorial angular velocity
is
a
a
<o
axes.
It
represented
0
is
<f>,
(0,
desirable to express (wj, co2, w3) in terms of the time derivatives ip).
Let (a, be unit vectors along the follows from the
It
axes.
b,
(f
c)
£)
r?,
,
= X = =
b,
a, u) to
X
fa>
b
c
a
c
HAMILTON'S PRINCIPLE AND THE EQUATIONS OF LAGRANGE AND HAMILTON 253
a = bco3
— cft)2, b = coj1 — aco3, c = aco2
— bcji (e)
Since the (x, y, z) components of the vectors (a, b, c) are the respective
rows in Table 7-1, we obtain from Eq. (e)
k = =
— w2/3,
— (Oa/j, = —
w3/2 /2 ft)1/3 l3 U)2l1 0)^2
Wj = co3m2 — oj2m3, m2
= o)1m3
—
co^m^ m3
= oy2m1
—
co1m2 (f)
'1
= ®ni cos ^
—
mi0 + '"i = sin + /i<£ +
= = — —
<f>
n1 6 cos 0 cos y> + «2y>, /2 $/i2 cos m2<j> Ijy> (g)
= — = —6 cos —
<f>
y>
y>
y;
0
6
rh2 6n2 l2<j> n2
<f>,
=
<j>
0
l3
m3
0
= —6 sin
0
n3
Table 7-2. Thus, nine equations are obtained that relate cov co2, co3 to
<j>,
y>.
= —6 sin
<j>
+ sin cos
y>
y>
6
col
= —6 cos — sin sin
y>
y>
<p
(7-15)
6
co2
=
<f>
cos +
y>
6
co3
By substituting Eqs. (7-15) into Eqs. (f) and expressing the direction
cosines by Table 7-2, we may verify that Eqs. (g) are all satisfied. A
different derivation of Eqs. (7-15), based on elementary geometrical
also available (62). Equations (7-15) are known as "Euler's
is
arguments,
kinematical equations."
7-9.
£)
r?,
,
the principal axes of inertia of rigid body that has fixed point at the
a
couple
£)
y\,
,
the Euler angles are given infinitesimal virtual increments (d6, d<f>, dy>),
If
The coefficients (Qv Q2, Q3) are the components of generalized force
(Sec. 1-7).
Any displacement of a rigid body with one fixed point can be effected
by a rotation about an axis (90). Furthermore (67), an infinitesimal
angular displacement is a vector <5p. The components of this vector on the
axes (f, rj, Q are denoted by d^, dP2, <5/?3. If the angular displacement
occurs in a time interval dt, <5p = to<5/, or <5/S1 = co1 dt, <5/32 = co2dt, etc.,
where to is the angular velocity. Consequently, by Eqs. (7-15),
<5/S!
= —sin rp d6 + sin 6 cos y> d<t>
<5/32
= —cos rp d6 — sin 6 sin y> d<f> (b)
<5/?3
= cos 6 d<f> + dy>
Since the virtual work of the couple M may be expressed in the form,
dW = M •
<5p
= M1 aft + M2 <5/S2 + M3 dp3, we obtain from Eqs. (a)
and (b)
Q1 = —M1 sin rp — M2 cos y>
The principal moments of inertia of the body are denoted by (A, B, C).
These are the respective moments of inertia about the (f , % £) axes. By
the theory of dynamics of rigid bodies (49), the kinetic energy of the body is
Equations (7-15) and (7-16) yield T/dy> = Cco3 and dT/df = Acnlco2
—
(A — .fi)oj1w2 = M3
—
Ccb3 (e)
Bco2 ^)w3w1
= M2 (7-17)
Cci»3
—
(A —
5)w1w2
= M3
/
moments of inertia of the top about the
(f, rj, £) axes are (A, B, C), respectively.
Because of symmetry, A = B. The
components of the angular velocity to
on the (f , rj, £) axes are (cov co2, wj).
Since A = B, Eq. (7-16) yields
Fig. 7-7
<f>.
angle of rotation of the top about the £-axis (Fig. 7-6). Eliminating
y>
is
=
<t>
+
6)
= mgh cos
V
(c)
6
the distance from the origin to the center of mass (Fig. 7-7).
O
is
where
h
V
<j>,
y>), L//dd
Eq. (7-5), the components of generalized momentum are pl = dT/d&,
p2 = dT/d<p, p3 = dT/df. Hence
pl = A&
= (A sin2 cos2 cos
+
+
C
p2 (d)
6
6)<j> Cy>
= C(V> + c°s
6)
p3
4
256 ENERGY METHODS IN APPLIED MECHANICS
<j>,
H = T+ if
y>)
are eliminated from Eq. (b)
by
V. Consequently, (0,
means of Eqs. (d),
~
pi
El
H = (ft ft cos 6?
*
2A 2/4 sin2 2C
0
Hamilton's equations (7-9) are
dH dH dH
•
L .
. _ dH . _ dH . _ dH
pl-~Te' pl~~Tf f3~~dy>
The first three of these equations give relations that are equivalent to
Eqs. (d). The last three yield
. ~(P2
~ p3 cos 6)(p3 - pi cos
g)
+ mgh s.n (f)
p
0
sin3
A
= =
°,
A
(g)
0
/>3
+ cos
y>
- (h)
A C
sin2 + bK cos = M, =
</.
dv dv dd dv
=
«
= — = = —
0 /.
v,
6
dt dd dt d6
Introducing these relations into Eq. (f) and noting that p2 = AM and
p3 = CK, we obtain by integration
z2 = — — cos + constant
(i)
62
t
t
a
6
:
sin2
6
Eq.
- - au) - (M - bKu)2
(i)
=
(j)
ii2 u2)(N
(1
- + bKu = M
(1)
^(1 m2)
Hamilton's principle and the equations of lagrange and Hamilton 257
Equations (j), (k), and are three first-order differential equations that
(1)
<f>,
determine and as functions of Equation (k) merely asserts that
6,
y>
/.
component of angular velocity on the axis of the top constant,
is
the
since,by Eqs. (7-15) and (k), w3 = K.
(j)
Equation separable, but the integral cannot be evaluated in terms
is
of elementary functions. However, the general nature of the solution
is
known (62). may be shown that periodic function of hence
It
t;
is is
u
a
periodic function. After «(0 determined, the function
is
is
a
d(t) <£(0
determined by Eq. (1).
M (b) (c)
Fig. 7-8
The motion may be visualized by means of the curve that the axis of the
top traces on fixed spherical surface with center at the tip of the top.
a
The relations
0
There are three possible types of curves, depending on the initial conditions
(Fig. 7-8).
The motion may be studied with the aid of certain approximations
if
the fluctuations of are small. then convenient to set = + a,
It
is
00
0
is
= bK(cos -
<j>
0)
6
- — — cos
00
= (cos fr2K2(cos
cos
6*
(n)
6)
00
sin20
J
00,
6
if
is
is
Consequently, co3
258 ENERGY METHODS IN APPLIED MECHANICS
, 2mAgh sin 80
\j ZirAmgh
Co C2Jf2
Fig. 7-9
. , CKt
a =
2mAgh sin
2——0n2 sin2 (P)
X2C2 2/1
0 = — — sin2,CKt
2mgh .
(q)
KC 2A
Integration of Eq. (q) yields
. mghl
= —— t
A . CKt\
sin (0
4>
KC \ CK A I
In the 0) plane, Eqs. (p) and (r) are parametric equations of the curve
shown in Fig. 7-9. This curve agrees with Fig. l-Sb; it is a modified
cycloid.
of precession, and
</>
<f>
7-11. THE GYROSCOPE. A rigid body with one fixed point is called
a "gyro" (18). The top discussed in Sec. 7-10 is an example of a gyro.
In devices such as the gyrocompass and gyrostabilizer the gyro is usually
mounted on a shaftwith bearings on the innermost of two hinged rings,
called "gimbals" (Fig. 7-10). Thus the center point of the shaft of the
gyro is fixed. The ring that supports the shaft of the gyro is called the
"inner gimbal," and the other ring is called the "outer gimbal." The
Fig. 7-10
260 ENERGY METHODS IN APPLIED MECHANICS
<f>.
7-8), in fact, <xl
= y, (% = 6, and a3 = These relations are apparent
from Fig. 7-10. The angle has been defined as the angle between the
6
and z-axes; hence identical to the dihedral angle between the planes
is
it
£-
of the inner and outer gimbals. Accordingly, = a2, provided that the
6
zero reference line of
chosen properly.
is
a2
The plane of the and z-axes will be called plane P. The angle
<f>
has
£-
been defined as the dihedral angle between plane and fixed plane that
a
contains the z-axis. Since plane to the plane of the outer
P
is
perpendicular
<ff>
gimbal, accordingly equal to the angle a3, through which the outer
is
P
rotating
xp
a
meridional plane of the gyro. Consequently, likewise the dihedral
is
xp
angle between rotating meridional plane of the gyro and the plane of the
a
that
xp
;
gimbal rotates the gyro at constant angular velocity relative to the inner
= constant.
is,
y;
different from the component of the angular velocity on the £-axis.
is
co3
ip.
+
d
eo3
(a)
where the moment of inertia of the outer gimbal about the axis of its
is
/
fixed hinges.
If the f-axis the hinge line of the inner gimbal and the £-axis the
is
is
axis of symmetry of the gyro, the Euler angles of the inner gimbal are
These variables were used Poritsky, who studied effects of torques applied to
by
*
the gyro and the gimbals and effects of inertia of the gimbals (132).
Hamilton's principle and the equations of lagrange and Hamilton 261
<f>,
6, tt/2 (see Figs. 7-6 and 7-10). Consequently, by Eqs. (7-15), the
components of angular velocity of the inner gimbal on the (f, r], axes
Q
of that gimbal are co, = — =
<f>
sin cos Ordinarily,
6,
6.
cot= —if>
6,
co3
the principal axes of inertia of the inner gimbal are the hinge line of the
inner gimbal (f-axis), the axis of symmetry of the gyro (£-axis), and line
a
perpendicular to the plane of the inner gimbal (»y-axis). Letting the
moments of inertia of the inner gimbal with respect to these (f, rj, axes
£)
(A', B', respectively, we have the following formula for the kinetic
C
•),
be
: CO
= \(A'co* + B'co* +
Tt
With the preceding expressions for wj, w2, w3, this yields
=
T,
0)
+
(b)
6
The gimbals are ordinarily balanced so that their centers of mass lie at
the fixed point on the axis of the gyro. Then the potential energy of the
zero. Consequently, the center of mass of the gyro lies at
if
gimbals
is
distance from the fixed point on the axis of the gyro and the mass of the
h
gyro
Sravityis
-mgh co*
V
(c)
9
is
= $A02 \C(y
cj>
sin2 cos
+
$fy2
+
ft
-
L
6)
0)2
6
0
B'<f>2 C'<f>2
work that they perform when the coordinates (6, receive variations
Consequently, (/?„, R+) are the components of
Re dd +
is
7-10).
PROBLEMS
The lowest corner of the homogeneous cubical block moves along the
1.
homogeneous quadratic
a
of generalized velocity.
262 ENERGY METHODS IN APPLIED MECHANICS
Fig. P7-1
equations.
7. A particle of mass m that moves in space
located by spherical coordinates
is
<f>
attracted
is
is
8,
r
<j>
0
t
Show that line drawn from the origin to the particle sweeps out equal
a
is
(see
/
nn,I
Fig. P7-5
Fig. P7-8
frame. Show that this equation yields the conclusion that the total
mechanical energy is constant.
11. A simple pendulum of length / and mass m is suspended from the rim of a
wheel of radius r which rotates with constant angular velocity co in a vertical
plane. Adopting the angle 0 that the pendulum forms with the vertical as
the generalized coordinate, derive the differential equation for d(t) by
means of Lagrange's equation.
12. A homogeneous solid ball of radius r on a horizontal platform
oscillates
that executes horizontal simple harmonic motion defined by the equation
s = s0 sin cot. Let x be the displacement of the center of the ball relative to
the platform. Derive the function x(t) by means of Lagrange's equation.
13. A small body of mass m falls to earth from a great height while remaining at
constant latitude a. The attraction of gravity is F = km/r2, where r is the
distance from the center of the earth. Let the distance r and the longitude 4>
relative to the earth be generalized coordinates. Taking into account the
angular velocity co of the earth, write the differential equations of motion by
<j>.
means of Lagrange's equations. Eliminate Transform the resulting
<f>,
differential equation by regarding as function of and integrate the
a
r
7,
8,
M(t)
Fig. P7-15
15. Figure P7-15 represents two flywheels that are keyed to uniform elastic
a
shaft. The moments of inertia of the flywheels are and I2. A time-
Ix
dependent torque M(t) applied at the end of the shaft. The angular
is
displacements at sections
2
1
,
stiffness of the shaft GJ. Regard 00, 0t, as generalized coordinates and
is
02
16.
a
y)
with components (Fx, Fy). The coordinate system (x, rotates in its plane
with angular velocity co relative to Newtonian reference frame. Adopting
a
(x,
HAMILTON'S PRINCIPLE AND THE EQUATIONS OF LAGRANGE AND HAMILTON 265
Fig. P7-17
17. The mass m slides on the rigid rod which rotates in a horizontal plane with
constant angular acceleration a (Fig. P7-17). There is a constant frictional
force F. By means of Lagrange's equation, derive the differential equation
for r, letting 6 = 0 for t = 0.
18. Let (a, b, c) be unit vectors along the (f, V, £) axes (see Table 7-1, Sec. 7-8).
Prove that (a — i) • (b — j) x (c — k) = 0. Hence prove that the direction
cosines in Table 7-1 satisfy the relation,
/i-l
m2
- 1 = 0
I, Ho — 1
19, Show that all two-rowed minors of the determinant in Prob. 18 cannot be
zero unless all nine terms in the determinant are zero. Hence, by using
the result of Prob. 18, show that there is a unique line L determined by
x = f, y = V z = £. Show that a suitable rotation of the body about line
L carries the axes (f, V, £) into the respective axes (x, y, z).
20. A rigid body has one particle fixed at the origin. Initially, the Euler angles
are d = 30° $ = 0, y> = 0. After a displacement of the body, 6 = 90",
= 60°, = 120°. Using the result of Prob. 19, determine the fixed axis
ij>
L
yi
1 z
£
1/V5 -2/V5
0
Fig. P7-25
29. The center of mass of gyro lies at the fixed point on the axis of the shaft.
a
Suppose that the bearings are frictionless but that there are linear springs
with constant in the hinges of the gimbals that tend to restore them to the
k
configuration for which =tr/2and^ =0. Taking the inertia of the gimbals
0
<j>
by supposing that does not differ much from and does not differ much
d
from tt/2.
30. A sphere of radius a rolls on plane. Let (x, y) be rectangular coordinates
<j>,
y>)
of the point of contact between the sphere and the plane and let (6, be
Euler angles of the sphere. Derive the differential relations that express the
condition that any infinitesimal displacement (dc, dy, d0,
is
d<f>,dy>) per
formed without slipping.
31. If the lower joint of the double pendulum discussed in Sec. 7-3 locked,
is
=
4>.
Equations (f) are not consistent with this condition. Explain why.
d
8 Theory of vibrations
Pythagoras discovered the importance of dealing with abstractions; and
in particular directed attention to number as characterizing the perio
dicities of notes of music.
A. N. WHITEHEAD
The length r (Fig. 8-1) locates the center of mass of the rigid body. The
velocity of the center of mass of the body is x1 + rx2. Consequently, the
kinetic energy of translation of the body is \m(x + ri^2. The kinetic
Fig. 8-1
*i + r*2 + Ax1 = 0
(d)
rxj + (r2 + p2)*2 + Bx2 = 0
where A = ajm and B = b/m.
Equations (d) are homogeneous linear differential equations with
constant coefficients. Their general solution may be obtained readily by
well-known methods. In the theory of vibrations especial importance
attaches to particular solutions of the form
xi = zi sin (wr — y),
x2
= z2 sin (pit — y) (e)
270 ENERGY METHODS IN APPLIED MECHANICS
in which zj, z2, co, and y are constants. Such a solution is known as a
natural mode of vibration (also called an "eigenvibration"). When the
system vibrates in a natural mode, all particles perform simple harmonic
oscillations that are in phase with each other and that have the same period.
Accordingly, all particles attain their maximum velocities simultaneously
as they pass throughtheir neutral positions, and they all reach their
extreme displacements simultaneously. The constants z1 and z2 are known
as the amplitudes of the motion; the constant co is called the angular
frequency (or "circular frequency") of the motion, and the variable cot — y
is called the phase of the motion. Consequently, the initial phase is —y.
The number of oscillations per second (called the "frequency") is coj2ir.
Accordingly, the time interval in which an oscillation is performed
(called the "period") is 2n\co.
To determine the constants zv we substitute Eq. (e) into Eqs. (d).
z2, co,
Thus the following equations are obtained :
(A - - = 0
-
rcoh2
+ + = 0
U
rco2z1 [(r2 p2)co2 B]z2
if,
z2 and only the constant
root of the determinantal equation,
is
co
a
- co2 -rco2
-
[
=
0
+ p>2
B
rco2 (r2
or
- [A(r2 + + AB =
+
Setting r/p = x and B/(Ap2) = y2, we may express the solution of Eq. (g)
in the following dimensionless form
:
^= (y-
(y
x2
(h)
1
A
co2;
"frequency equation."
If co determined by Eq. (h), one of the equations in (f) redundant.
is
is
(f) yidds
A
rz2
©
1
2
iz
is
y
be any
real constants. Let be determined by Eq. (i), with = and co = co'.
z/
z2
z1
THEORY OF VIBRATIONS 271
Let be determined by Eq. (i), with zx = zx" and co = co". Then, since
z2"
the sum of any two solutions of Eq. (d) is again a solution, we obtain
xx
—
2i sin (co'' - /) + zi sin («»"/ — y")
(J)
= sin (pit —
/) z{ sin
— y")
z2'
+
(j) x2 (co"t
y',
Since Eqs. contain the four arbitrary constants, zx', Zi", y", they are
the general solution of Eqs. (d). In other words, the general solution of
linear combination of two natural modes.
is
Eqs. (d)
a
is
The hinge of the simple pendulum may move
, , ,
mm
up and down between the frictionless guides. 2^
c
The coordinate x measured from the position
is
\mx2.
To compute the kinetic energy of the bob, we
observe that the coordinates of the bob
= sin = x cos The kinetic
6.
6,
are
+
rj
f
/
lml2&2
0
6
= \mH2
T
mx2
+
= —mgx — mgrj
+ \k(x +
V
/i)2
where the extension of the spring for which the system can remain
is
h
— — mgl cos
+ constant
V
\kx2
d
272 ENERGY METHODS IN APPLIED MECHANICS
x + A
— x = o, 0' + £ 0 = 0 (k)
In
2m I
co„
= Vk/2m, coe
= Vgfl
For a natural mode, cox
=
ca. coe
=
Evidently this condition is impossible,
unless kj2m = g/l. The special feature of this system that excludes natural
is,
modes is that the coordinates are decoupled ; that the oscillation of x
does not affect the oscillation of
6.
Ifthe exact expressions for and are retained in this problem, the
T
- - -x=
k
0
6
6
(1)
=
ig
sin
10
x)
0
0
to g, and all nonlinear terms are discarded from Eq. (1). Equation (1) shows
that the system not completely decoupled. The decoupling results from
is
systems
vibrations of systems with finite degrees of freedom was developed in
a
general form by Lagrange. This theory provides one of the most important
engineering applications of Lagrange's equations.
Vibrations of system are said to be free no energy enters or leaves
if if
a
V=il <=i
2<W,
j=i
(8-1)
«=i i=i
in which z}, co, and y are constants. Equations (8-3) and (8-4) yield
i(fl„
j=i
- <A,)2, = 0 (8-5)
if,
if,
the z's and only the
determinant of the equations
is,
zero; that
is
det (au - co%j) =
0
(8-6)
is
nth degree in co2; consequently possesses roots. The symmetry con
it
n
ditions, = aH and bu = bH, combined with the conditions that the
quadratic forms and are positive definite, lead to the conclusion that
V
;
n
con.
These values are called the "natural frequencies" or the "eigenfrequencies"
of the system. The natural frequencies are not necessarily distinct, but the
case in which two roots cot are equal exceptional. If this condition
is
(8-3^
Rt =
0.
less
is
co2bu)
than for one or more of the natural frequencies cok (see Appendix,
—
n
of the z's may be assigned arbitrarily, and Eq. (8-5) may be solved for the
remaining z's (5). If < — there certain amount of decoupling;
1,
is
a
n
r
that some
ally, all coordinates are decoupled. This case illustrated by the second
is
is,
B^B = where the unit matrix. Accordingly, the roots cot are
/,
is
/
identical to the roots of the equation det (B~xA — co2/) = In this
0.
determinant the unknowns co, occur only on the principal diagonal.
Practical numerical methods for finding the roots cot of determinantal
a
equation of this type are discussed in the book by Faddeeva (23).
is
important in physics. Identifying xt as yt, we may employ the dual-
transformation theory of two quadratic forms (Appendix, Sec. A-4). Let
(/„) be the matrix that reduces the two quadratic forms and to the
T
V
canonical form. Then
- n
=
n
det (tu)
#
U,"t,
I
xi '««/.
2
0
*<
where ut and are the new variables. Hence = ut. The variables w<
v,
vt
are called "normal coordinates" for the system. According to the theorem
in Appendix A-4, the kinetic and potential energies assume the following
forms when normal coordinates are used
:
iKV
(8-7)
= co2V +
• • •
+ coBV)
v
wt + co,\ = Rt (8-8)
If/?, known function oft, the variables are separated in Eq. (8-8).
is
w,
a
Integration yields
Hj = /L.sinKf-y<)+./X0 (8-9)
= sin il,t
Rt
Ct
(8-10)
276 ENERGY METHODS IN APPLIED MECHANICS
If Qt does not coincide with one of the natural frequencies of the system, a
cof
- O/
If Qi happens to lie near the natural frequency cu„ Ht is very large; in fact,
Hi approaches infinity if
Q< approaches w<. This condition is known as
"resonance." In practice, resonance is manifested by violent shaking.
Damping prevents the infinite amplitude that is predicted by the preceding
theory.
since the mean values of sin2 — y,) and cosa — yt) are
(ci^r \.
(cott
For a conservative system with one degree of freedom, the equation
T = V or T0 = V1 determines the natural frequency. A system with more
than one degree of freedom may be reduced to one degree of freedom by
means of an assumption about the mode form; that is, an assumption
about the ratios of amplitudes of the various parts. Then the equation
T = V provides an approximation to the frequency of the mode under
consideration. For example, it might be assumed that the vibration of a
uniform cantilever beam is represented by y = A(\ — cos nxj2L) sin cot.
Then, by Eq. (2-8), V = ir*EIA2imL3. Also, T = ip^2w2Z.(f - 4/tt),
where p is the mass per unit length. Accordingly, the equation T= V
yields cu = 3.66V Elj pL*. This is 4.2 per cent higher than the correct value.
THEORY OF VIBRATIONS 277
dW—jtR/dx, (b)
i=l
where R/ is the part of the generalized force that represents viscous friction.
278 ENERGY METHODS IN APPLIED MECHANICS
Since dxt is arbitrary for a holonomic system, Eqs. (a) and (b) yield
«/"-!>„*, (8-12)
Equation (8-12) represents the generalized force that arises from viscous
friction. Force components due to other causes may be superimposed to
provide the total generalized force Rt that appears in the Lagrange
equations (7-12). Conservative forces are accounted for by the potential
energy function V.
By a slight extension of the theory, it may be shown that Eq. (8-12)
remains valid if the frictional forces are proportional to the relative
velocitiesof the particles rather than to their absolute velocities, provided
that the coefficients ctj are modified suitably (71). For free vibrations of a
system with viscous damping, Rt' = /?,. Then, if the coefficients ctj are
constants, Eqs. (8-3) are generalized as follows by means of Lagrange's
equations (7-12):
(c)
in which zi and r are constants. Substitution of Eq. (c) into Eq. (8-13)
yields
i=l
2(rX + ^ + ««)^ = 0 (8-14)
solution
if,
Since the left side of Eq. (8-1 polynomial of degree 2n in r.the equation
is
5)
possesses 2n roots, although they are not necessarily all distinct. Again,
we consider only the case in which the roots are distinct, since this con
dition may always minor modification of the physical
be ensured by
a
gained
is
a
THEORY OF VIBRATIONS 279
the complex roots areof primary interest in vibration theory. It has been
shown in Sec. (8-2) that the particles of an undamped system execute
periodic motions. A damped system approaches this condition if the
damping factors approach zero. Consequently, if damping is small, a
decaying oscillation occurs. In this case, Eq. (c) shows that the roots r, are
complex constants with negative real parts. Furthermore, since the coeffi
cients of Eq. (8-15) are real, the roots of this equation occur in pairs
(/•,., rt), where rt is the complex conjugate of rt.
/3,
possible
given by (d) are equivalent. Consequently, attention will be confined to
the first of these equations.
Let us set a = Aerxv, = Pje'e', — —c + ico, where = V— and
z}
1
i
real. Also, let Then Eqs. (d) yield
A,
co co
0
c
+
6,
(cot (e)
i
Because of the arbitrary constant y, the real and imaginary parts of Eq. (e)
represent the same motion. Consequently, the general real solution
obtained from Eq. (e)
is
xi = APje-* sin
-
y)
+
(8-16)
dj
(cot
no damping, ct, = and Eq. (8-15) yields pure imaginary roots = ico),
0,
(r
6t
z,
damping. It indicates that the various parts of the vibrating system are not
phase with each other.
in
real solution of Eq. (8-13) of the type in Eq. (8-16). Each of these solutions
contains two arbitrary constants, and y. Consequently, the solutions
A
•,
(8-13).
280 ENERGY METHODS IN APPLIED MECHANICS
A = i('1 \L(Py?-ElyJ)dxdt
where p denotes the mass per unit length. If El is a constant, the Euler
equation for this double integral is
Vmmx + £
EI
Vtt
= 0 (8-17)
-f + -E-i-
EI g
= 0
/ EI g
This equation shows that the oscillation of any particle is a simple harmonic
motion.
The first of the equations in (8-18) yields
f(x) = 0 at x = 0 or x = L, f'(x) = 0 at x = 0 or x = L
THEORY OF VIBRATIONS 281
fix) = ^(sinh px
- sin px) + 5(cosh px - cos Px) (8-20)
PL -
^(sinh
(8-21)
^(cosh cos PL) + 5(sinh PL + sin PL) = 0
if,
if,
These equations possess a nonzero solution and only their deter
minant zero. This condition reduces to
is
1
known as the "frequency an
It
Equation (8-22) equation."
is
possesses
infinite number of roots, denoted by P0L, p1L, P2L, etc. These roots are
= = 4.7300408, =
0,
PnL
5,
(n
n
/S,
is
B=l, _A = = ^hPnL-cosPnL
0Ln
sinh PnL — sin PnL
a1
= 0.9825022158, = 1.000777311, a3
= 0.999664501
a4
= 1.000001450, a5
= 0.9999999373
> =
with an accuracy of at least seven decimal places.
If
5,
1,
an
n
The functions fn(x) are known as the "mode forms" of the beam. These
by
functions, and their derivatives, have been tabulated Young and Felgar
(151). Mode forms for other end conditions are also tabulated.
The frequency equation (also called "characteristic equation") (8-22)
analogous to the frequency equation for system with finite degrees of
is
freedom [Eq. (8-6)]. The roots of this equation are known as the "eigen
values" of the boundary-value problem. The mode forms, represented by
Eq. (8-23), are called "eigenfunctions." All problems of free vibrations of
undamped linear systems are eigenvalue problems.
282 ENERGY METHODS IN APPLIED MECHANICS
frLfm(x)U*)dx =
Jo
0, m^n (8-24)
This condition is expressed by the statement that the functions fn(x) are
orthogonal in the range (0, L).
To verify Eq. (8-24), we observe that Eq. (8-18) yields
Integration yields
( fmJf"
\J n
_ JfnJf"
m ff
+' JnJm
" —
f 'f "))
JmJn L+(Pmi-Pni)\Lfmfndx
o Jo
= 0
L
fn2dx =
2
L (8-25)
Not only are the functions fn(x) orthogonal in the range (0, L) but their
second derivatives are orthogonal in this range. This may be shown as
follows :
Equation (a) yields
fmf": - PSfmfn = 0, m * H
Lfmf:'dx = o
f.0
THEORY OF VIBRATIONS 283
Since /m(0) = fm(L) = 0, the terms outside the integral signs vanish.
Integrating by parts again and noting that fm'(0) = fm'(L) = 0, we obtain
= 0, m^n
\fmJn"dx
Jo
(8-26)
\L<Jnfdx
= p*L (8-27)
Jo
The following formulas are also useful :
Ll+(-D
ofJmdX~ JT^l 1 m ^ n
\LfJn dx = «n/Sn(2
- xJnL) (8-28)
Jo
These and many other formulas for the mode forms of beams with various
end conditions are given in a bulletin by Felgar (104).
Influence Function for a Clamped Beam. The mode forms for beams
and various other simple vibrating systems are excellent functions for use
with the Ritz procedure (Sec. 3-11), even though the problem under con
sideration is not concerned with vibrations. As a simple example, we
consider the deflection of a clamped beam due to an off-center concentrated
load (Fig. 8-3). The deflection may be represented by
V=laJn(x)
l n=
(b)
where /n(x) are the mode forms of the beam. The preceding series satisfies
all the boundary conditions, irrespective of the values of the coefficients an.
This fact and the orthogonality property are the main reasons why the
functions fn(x) are advantageous in this problem. Equation (2-8) yields the
following formula for the strain energy of the beam :
Jo
284 ENERGY METHODS IN APPLIED MECHANICS
Since the functions fn" are orthogonal in the range (0, L), this equation
yields
-'o
Hence, by Eq. (8-27),
n= l
Fig. 8-3
X
Vc
= 1l ajn(c)
n=
"= -i>lX/n(c)
«=i
V=iEILZpnian*-pfanfn(c)
n=l n=l
dV
EILpJam - Pfm(c) = 0
PL3 Zfn(x)fn(c)
(8-29)
EI n= l /Sn4L4
With the tabulated values of the functions fn(x), Eq. (8-29) is easily
From the tables of Young and Felgar (151),/i(Z./2) = 1.58815,
applied.
/2(L/2) = 0,/3(L/2) = -
1.40600, /4(L/2) = 0. Hence, using the values of
/SnL from Sec. 8-4, and letting x =
= L/2, we get y = 0.005 UPtf/EI.
c
This is an approximation for the deflection at the center due to a load at
the center if only the first three terms of the series are used. The exact
answer is known to be
PI? PI}
y = = 0.005208
192EI El
In his celebrated paper on variational methods of approximation (142),
Ritz used products of the functions fn(x) for a clamped beam to represent
deflections of a clamped plate.
Fig. 8-4
•-♦/'
Jo
(Py? + IJ?)dx (a)
where the subscript x denotes the partial derivative with respect to x and I
denotes the moment of inertia of the cross section of the structural parts
about the principal axis of inertia through point P'. Also, since the twist
per unit length is 6x, the strain energy due to twisting is
GJ6* dx (C)
Joo
THEORY OF VIBRATIONS 287
t).
a
is
a
wing depend on the elastic response of the wing to these loads. In the
present example this complication need not be considered.
The point at which the force dx considered to act on lamina
is
is
Q
a
arbitrary, since this point may be changed by the introduction of com
a
pensating change in the moment function M. For convenience, we let the
force dx act at point (Fig. 8-4).
P
Q
is
-\\Qy + Md)dx
Jo
= - (Qy M dx
U
(d)
+
V
0)
Jo
where U the strain energy of the wing.
is
•z
IJt2)-hGJdx*
+
Jh jo
- iEI(yxx + 2a A + adxxf (Qy + M0)] dx dt
+
axx6 (e)
pytt + (Elt)^ =
- (GJ6X - 2(EIaJ,)x
Q
where
=
<f>
Vxt + (a0)*x
However, since they are linear, they may be solved by numerical methods
with the aid of modern computing equipment.
288 ENERGY METHODS IN APPLIED MECHANICS
is
/?.
Therefore, the rotary kinetic energy, per unit length, ipl(yxt — j8t)2, where
is
the mass per unit volume and the moment of inertia of the cross
is
is
/
p
section about its centroidal axis. This augmented by the kinetic energy
is
(8-30)
i
Jo
By Eqs. (2-15) and (8-30), the action integral
is
- - - -
h
Jo Jt0
Because of Hamilton's principle, the Euler equations for the action integral
are the differential equations of free vibrations. These equations are
[EI(yxx [PI(yxtt
f
(8-31)
OXi ox
-
[.El(yxx - PI(yxtt - =
+
ptt) Xp
0
|-
(8-32)
ox
Differentiating Eq. (8-32) with respect to x and comparing the result with
Eq. (8-31), we obtain
fM)=:pAytt (8-33)
THEORY OF VIBRATIONS 289
Equations (8-31) and (8-33) are the differential equations of free vibrations.
They are applicable even though A and vary gradually with x. /
If shear deformation and rotary inertia are both neglected, Eq. (8-31)
reduces to
PAytt = 0 (8-34)
^(EIyxx)+
0
= 0 in Eq. (8-31).
If X and / are constants, may be eliminated from Eq. (8-31) by means
/S
of Eq. (8-33). In this way, we obtain
— o2AI
(EA\J
1+ Vxxu +
tLj~ Vtm + pAytt = 0 (8-35)
- wj* - - - vO - ^w,2]
JJJ
the mass per unit area of the plate. If and D are constants, the
is
where
p
V2V2w + = (8-36)
0
wtt
wrote: "Inadopting, like the author, 1/r + 1/r' for the measure of the curvature of the
surface, which the elasticity tends to diminish, and to which one supposes to be
it
°
2
The natural modes are solutions of the type, w = fix, y)g(t). Sub
stituting this equation into Eq. (8-36), we obtain
m.?, ££-,_^ w
/ D i
where /3 is a constant. The second of these equations yields
g^CsmipVDfpt-y) (b)
V2V2/-,Sy=0 (8-37)
This differential equation determines the mode forms of the plate. For a
rectangular plate with simply supported edges on the lines x = 0 and
x = a, the boundary conditions on these edges are w — = 0 (Sec. 5-3).
Consider mode forms of the following type :
Set
With
pn = + («M*2).
= An sinh pny +
<f>
Cn Z>n (f)
For example, let the edge = be simply supported and let the edge
y
= be free. By Sec. 5-3 and Eq. (5-28), the boundary conditions for the
y
/=/w = at =
0
0
2/
= -v)/**» = at =
+
+
(2
fvw
°.
/»v
0
v/k»
Although these boundary conditions were derived for statical loading, we may
*
show, taking the first variation of the action integral, that the same boundary
by
= = =
<£"
at
</,
y
0
0
-^r
(g)
f"-(2-
<f>"
= = =
<f>
r
'0,
at
y
0
b
,2
7
v
„2
The constants in Eq. (f) are to be determined by Eqs. (g). The boundary
conditions at the edge = yield Cn = Dn = The boundary condi
0.
0
tions at the edge = y yield
y
- -
{qn2
(pn2
+
An Bn (h)
0
v
v
(i)
Bnqn
0
v
v
(pn2
j
J
Equation (i) has been simplified by means of the following relations which
result from Eqs. (e):
,2_2
=
nV
pn2-(2-v)-Li-
a1
qn2 +
a
(j)
- —= - —
(2
+
v)
9n
V
/>n2
a2 a2
Equations (h) and determine values of An and Bn that are not both zero
(i)
if,
is
~ n2772\2
—
n2tt2\2
/
tanh = + tan
^n2
v
J
For each positive integer n, Eq. (k) yields an infinite sequence of roots
jSmn; m = • •
These are the eigenvalues of the vibration problem.
2,
3,
1,
With Eq. (b), they determine the natural frequencies. Let the values of n
p
Eq.
corresponding to (}mn, be Amn and Bmn. Then, since Cn = Dn =
0,
Eqs. (f)
and (h) yield
= =
<f>
(1)
+
-
mn
sinh
v
pmnf>
(pmn2
292 ENERGY METHODS IN APPLIED MECHANICS
Accordingly, by Eqs. (b) and (c), the natural mode, corresponding to the
positive integers (m, n), is
wmn
= Bmn(sin qm^y + Kmn sinh pmny) sin
a
sin (fi^VWp t - ymn)
(m)
This solution contains two arbitrary constants, Bmn and ymn. Equation (m)
yields a double infinity of natural modes. Because of the linearity and
homogeneity of Eq. (8-36), the sum of any number of natural modes is a
possible free vibration.
Instead of Eq. (c), we could write the more general relationship,
Such a series can represent any mode form. Nothing new results from this
approach, however. Therefore, all natural modes are represented by
Eq. (m). The foregoing theory is quite parallel to the corresponding theory
for buckling of a plate with the preceding boundary conditions and a
uniform axial load (84, Sec. 65).
The mathematical nature of linear vibrations of undamped systems,
with emphasis on the relationships to the theory of eigenvalues and
eigenfunctions, is treated in the books by Courant and Hilbert (13) and
Collatz(1l).
2- = ^ (a)
Py Poy
v = - (b)
THEORY OF VIBRATIONS 293
V=-\pdv (c)
v=^n-Ply-l
y - 1 Po
(d)
is
t)
a
that Lagrange applied to problems of fluid motion. For studies of oscilla
tions, Lagrange's representation sometimes preferable to the Eulerian
is
Consider the fluid that lies initially in the shell bounded by the spherical
surfaces with radii and dr. As the pressure wave passes, this shell
+
r
of fluid alternately expands and contracts. At the time the inner and
t,
+
is (1
u
r
Consequently, the volume of the expanded shell 4tt(t + «)2(1 + ur) dr.
Since the mass of fluid in the shell the same as the initial mass, ^np^2 dr,
is
es£
P
u)2(l + ur)
+
(r
Therefore, by Eq. (d), the potential energy of the gas in the expanded shell
is
l-y
v — L\ rl
y
is
Consequently,
a
1-7
dr
r2
(e)
v — Jo L\
Wo rl
y
0
r
r
294 ENERGY METHODS IN APPLIED MECHANICS
The mass of gas in the expanded shell is 4ttp0r2 dr, and its radial velocity
is ut. Consequently, the kinetic energy of the gas is
T=2iTpJauydr (f)
Jo
Equations (e) and (f ) yield the following formula for the action :
A =
y - 1
r* r°[2Lzj
Ji, Jo L 2
£0
p0
rV _ a,\ + h)2(i-w(i +
r/
u,hJ dr A
The Euler equation for this double integral is
(g)
(l-2y)H ...
(l
+ «-p=l + +
Expanding the other terms in Eq. (g) similarly and discarding nonlinear
terms that arise from the products, we obtain
u„ +
2
-ur- -
2
u
1
= —utt (h)
r
c02
r2
where c0
= Vypjp0.
The general solution of Eq. (h)
is
= ~ - -
*
(i)
c00]
«
that travel outward with speed c0. The attenuation factor for the first wave
1/r and that for the second wave At
is
1/r2.
a
THEORY OF VIBRATIONS 295
from the source, the wave with attenuation factor 1/r2 has almost dis
appeared. The remaining wave motion would be obtained from Eq. (h) if
the term w/r2 were discarded from that equation. Then Eq. (h) would be
the well-known wave equation for the spherically symmetric case.
t),
considered to be
0.
is
v
is
v)
a
expressed by the differential equation,
+ - =
+
(8-38)
0
vv
ux uxvv uvvx
equation." signifies that the volume of any part of the fluid remains
It
pressure.
is
entirely from the weight of the liquid. The potential energy of particle of
a
is
g
v
is
0
z
(a)
where the double integral extends over the region initially occupied by the
Likewise, the kinetic energy of the liquid
is
liquid.
(b)
the action
is
Consequently,
(c)
296 ENERGY METHODS IN APPLIED MECHANICS
Suppose that the fluid is contained in a tank with end planes x = 0 and
x = a and with bottom plane y = —b. The free surface of the undisturbed
liquid is represented by the plane y = 0. It is convenient to imagine a space
with rectangular coordinates (x, y, The region of integration in Eq. (d)
t).
rectangular parallelepiped in this space. The limits on the time integral
is
a
A
are the dynamical equations
motion. However, we must perform the variational process on the integral
to obtain the natural boundary conditions for the free surface. Let the
A
tj)
v)
0
0
u
a
and = for = —b. Also, in accordance with Hamilton's viewpoint,
y
0
v
tv
v)
t0
The first variation of the integral
A
is
M=
/-JJJ
+ vm) - sv
+ A(f „. + vjx — uyT)x — vjy)] dx dy dt
+
by
multiple [Eq. (4-62)]. Transforming the terms in Eq. (e)
integrals
integration by parts and observing the forced boundary conditions for
f
and r], we obtain
- W»J dxdt dy dt
+
+
rj)
boundary conditions, the integrands of the triple integrals and the double
integrals vanish separately. Thus the triple integrals provide the Euler
equations,
- K»* "tt =
+
4(1 (8-39)
0
».)
-KM, + + =
+
(8-40)
g
vtt
0
ux)
:
A(l + = and = for = (8-41)
0
0
ux)
y
Xvx
y
Xvx
(8-42)
Equations (8-38), (8-39), and (8-40) are the Lagrangian equations of plane
flow of frictionless liquid that moves under the action of gravity alone.
a
Equations (8-41) are the natural boundary conditions for the free surface,
and are the natural boundary conditions for the walls of
Eqs. (8-42)
a
rectangular tank that contains the liquid. Equations (8-38) to (8-41) apply
also for an infinite expanse of liquid, but the boundary conditions per
taining to the walls of the tank have no significance in this case.
There only one known exact solution of the preceding equations that
is
for which the orbits of the particles are circular, he was led to try the
relations,
- -
2tt(-
2tt(-
= sin = cos
+
(8-43)
v
u
r
e
,
where and are functions of alone, and t and are constants. These
y
e
is
/
;
which the waveform advances one wavelength (called the "wave period"),
the radius of the orbit of particle, and the height of the center of
is
is
a
e
r
an orbit above the initial level of the particle that describes that orbit.
necessary to adapt the parameters t, to the equations of
It
r,
is
e,
motion and the natural boundary conditions. Substituting Eq. (8-43) into
Eq. (8-38) and setting = 2t7(//t — x/l), we obtain
6
298 ENERGY METHODS IN APPLIED MECHANICS
e'-^ = /
0 (f)
= r' (g)
?f(l+e')
Equation (f) yields
e = — (h)
is,
from trough to Equations (h) and implicitly express and
(i)
crest. as
e
functions of y.
Substituting Eq. (8-43) into the dynamical equations, (8-39) and (8-40),
we obtain
6)
Xv
0
6
T2
/
00
g
0
6
Equations and
Xx
Xy.
of Eqs. (f) and (g), the determinant of these equations Hence
is
1.
/2tt/ . gl. __
2ttV ...
\
=
^__^cos0_JLlogr +/(0
A,
,
+
THEORY OF VIBRATIONS 299
A = r - gj cos 6 -£ log T- -
^ (a2
- r2) + f(t) (1)
- (~ - ? (r2 - - gy + /(*)
a2)]
A cos 6 + (m)
g) [r
+
(1 = for =
the free surface. Consequently, Eqs. (8-41) require that
0.
y
0
A
Therefore, Eq. (m) yields f(t) = and = 2ttI/t2. Since the wavespeed
g
0
= Vgl/2n (8-44)
c
tappets.
adequate.
Let an isotropic elastic body be referred to rectangular coordinates
z). If the particles of the body perform small oscillations, the kinetic
(x,
y,
is
energy
= dx dy
t;t2
+ dz
+
T
p(uf2 w<2)
where («, v, w) the displacement vector and the mass density. It will
is
is
when body forces and temperature effects are absent, the action
is
- -
t;/
=
iJJJJ - wj2 + +
t)t2
+ w,2)
+
+
+
A
[p(Mt2 2G(ux2
vv
wt2) X{ux
where e = ux + vv + wz. The dots indicate that there are similar equations
for v and w. In Gibbs's vector notation, Eq. (8-45) is written as follows :
= grad + curl H
<f>
q (8-47)
y,
z,
<f>(x,
t)
= (8-48)
e
V2cf>
If = = =
<f>
0,
Consequently,
0,
j>
e
ment relations are used). The wave motion then said to be "equivolu-
is
It
0,
is
q
particle is
0
a
grad
G
(A
Ht<
p
<f>u V2<£
^
-
9
-
G
y>,
P
y,
z,
t)
y>(x,
THEORY OF VIBRATIONS 301
H„ - /S2V2H
= grad x,
4>u
- a2VV = 0, H„ - /?2V2H
= 0
, = A + 2G o2 = -G (8-49)
/S2
a2 ,
P
P
Equations (8-49) are known as the "wave equation." There an extensive
is
literature on in treatises and articles on partial differential equations,
it
/J.
a
vibrating medium, and let be the region contained within S.By Eq.
R
(4-16), the stress exerted by the material outside upon the contiguous
S
m,
+
may
is
is
S
n)
p
W=
-4))V•qtdS
8
is
unit normal and the material on the positive side of dS (the side toward
if
n
302 ENERGY METHODS IN APPLIED MECHANICS
expressed in terms of the derivatives q,, q„, q, by means of Eqs. (4-17) and
(4-35).
in which the angular frequency and the initial phase — y are constants
co
m,
n)
is
is
homogeneous;
satisfies the boundary condition, so does ca, where
is,
that vector
if
is
a
c
any constant. Equation (8-51), in conjunction with the boundary con
dition, possesses solution other than = and only if,
if
ca belongs to
a
0
a
one of an infinite sequence of values cov co2, • • \ These are the natural
frequencies of the body. To any natural frequency eo<, there corresponds
a
nonzero vector field that satisfies Eq. (8-51) and the boundary conditions.
a,
is,
This vector field contains an arbitrary constant factor; that ca, also
is
oscillations of the particles are not determinate, but the ratio of the ampli
tudes for any two particles fixed. Each particle executes simple
is
harmonic oscillation on the straight line in which the vector a< lies.
equations and the boundary conditions of the
Since the differential
vibration problem are linear and homogeneous, very general solution
is
a
=
00
cos - y,)
2
c,a< (8-52)
q
(cott
,=i
complete
a
t).
is
q
that perpendicular to the x-axis. Consequently, the planes
is
any plane
= constant are called "wavefronts." Lines perpendicular to the wave-
x
fronts are called "rays." In the present case the rays are all straight lines
parallel to the x-axis.
= q(«, = <f>(x, and H = H(x, t). In this case the general
<f>
t),
Since
q
t)
solution of Eqs. (8-49)
is
(14)
4>=f(x-xt)+g(x
-
+
*t) "
H = F(x
}
1
G(x +
+
/SO /SO
G
where are arbitrary
F
g
/
= -Fz'(x - -
-
Gz\x + (8-54)
v
/SO /SO
w = Fy'(x
+
F,', Gz are all arbitrary. For the motion represented by Eqs. 54), the
(8
= f" ~ Gz",
V„ = = Fv"
+
except
is
versus
t1
t
with speed a in the positive x-direction. Waves of this type are said
to be "progressive." Similarly, the solution « = g'(x + *t), = w =
0
v
304 ENERGY METHODS IN APPLIED MECHANICS
PROBLEMS
1. A mass m is suspended from a linear spring with constant k. The mass
oscillateson a vertical line under the action of a time-dependent force /••(/).
The displacement of the mass from the statical position is x. Derive the
differential equation of motion by means of Lagrange's equation. Supposing
that F = F0 sin cot, show that x becomes infinite if a>2 = kjm.
2. A uniform simple beam of length / and stiffness El carries a mass m at the
center. Neglecting gravity and the mass of the beam, derive the differential
equation for free vibrations of mass m by means of Lagrange's equation.
3. A heavy mass that is suspended by a slender elastic wire executes torsional
oscillations. One principal axis of inertia of the body coincides with the
wire. Supposing that the frictional torque is c6, where 0 is the angular
THEORY OF VIBRATIONS 305
lb-sec2/in.
wheel has radius of gyration in. and mass lb-sec2/in. The axle of the
4
wheel contains spring such that the restoring moment 8000 lb-in., where
is
a
</>
is
0
6. The three equal masses oscillate freely on vertical line (Fig. P8-6). The
a
constant for each spring k. Let the generalized coordinates (xv x2, x3)
is
represented by
p
mrx
—-
^
= 2«n sin
where means
is
t.
an
a
10.
a
of attraction between any two particles equal to kd, where the distance
is
d
306 ENERGY METHODS IN APPLIED MECHANICS
m 2
Fig. P8-6
wy2) dx dy
+
(wx2
Write the expression for the strain energy. Hence write the expression for
the action integral and obtain the differential equation of free vibrations
by
/).
differential equation for the deflection w(r, by means of Hamilton's
t)
principle.
15. Supposing that the strain energy of circular membrane of radius
is
a
a
proportional to the increment of area due to deflection and using small-
deflection approximations, derive the differential equation for free axially
symmetric vibrations by means of Hamilton's principle. Show that the
natural modes are determined by Bessel's differential equation of order zero.
Determine the first few natural frequencies with the aid of table of zeros of
a
16. Derive the frequency equation for uniform elastic beam that hinged at
is
the end x = and clamped at the end x = a
0
I.
17. Derive the frequency equation for uniform elastic cantilever beam.
a
a
elastic rectangular plate that on edges x =
0,
simply supported the
is
x = = and =
0,
a,
b.
y
y
19. By Hamilton's principle, derive the differential equation for free torsional
vibrations of uniform elastic bar. Determine the natural frequencies and
a
20. By Hamilton's principle, derive the differential equation for plane longi
tudinal waves in slightly tapered elastic bar.
a
21. Write the action integral for uniform elastic circular ring that vibrates
a
freely in its plane. Obtain the differential equations for the mode forms.
Neglect shear effects and rotary inertia.
22. By Hamilton's principle, derive the differential equation for plane com
pression waves in gas, supposing the amplitude to be large. Show that the
a
23. For ocean waves, the length 300 and the vertical distance from trough
ft
is
to crest 20 ft. Supposing the Gerstner theory applies, determine the wave
is
speed. Determine the depth at which the amplitude only per cent of
is
6=i 2aaXixt
>=1 1=x
(A-l)
ofx^
coefficient into the sum aw + a3,is arbitrary; there is consequently
no loss of generality in the specification atj = au.
If the coefficients au are real and the x's are restricted to real values, Q
is real. Evidently, Q = 0 if there's are all zero. Let us consider the range of
Q when the x's independently take all real values except the simultaneous
values 0, 0, • • • , 0. The following conditions may occur: (a) Q maybe
confined to the range Q > 0. Then the quadratic form is said to be positive
definite, (b) Q may be confined to the range Q > 0, there being infinitely
many values of the x's for which Q = 0. Then the quadratic form is said
to be positive semidefinite. (c) Q may be confined to the range Q < 0.
Then the quadratic form is said to be negative definite, (d) Q may be
confined to the range Q < 0, there being infinitely many values of the x's
for which Q = 0. Then the quadratic form is said to be negative semi-
definite, (e) Q may cover the range — oo < Q < oo. Then the quadratic
form is said to be indefinite. Stability of a conservative mechanical system is
usually characterized by positive definiteness of the second variation of the
potential energy. At the critical load this quantity becomes positive semi-
definite. At higher loads it is negative definite, negative semidefinite, or
indefinite.
A quadratic form "definite" if it is either positive definite or
is said to be
negative definite. It is said to be "semidefinite" if it is either positive semi-
definite or negative semidefinite. If a quadratic form vanishes at the point
, xn), it also vanishes at the point (kxv kx2,
• • • • • •
(&x, x2, , kxn), where k
Negative definite, — —
x12 x22
Negative semidefinite, — —
2x^x2
—
x22
= —
(a^ + x^2
Indefinite, XjZ^
(aw) is not zero ; then the rank of the matrix is said to be n. If the deter
minant of matrix vanishes, we consider the determinants of submatrices,
obtained by crossing out rows and columns of matrix (aw). The statement
that the rank of matrix (au) is r means that (auj) contains at least one
r-rowed square submatrix with a nonzero determinant, whereas the
determinants of all submatrices of (a„) with more than r rows are zero.
According to a theorem on matrices (5, Sec. 20, Theorem 3), there is at
principal minor of matrix (au) that is not zero [that
is,
least one r-rowed
a
minor whose principal diagonal coincides with the principal diagonal of
matrix (au)]. This conclusion follows from the fact that au = aH. There
are several other theorems that facilitate the calculation of the rank of
a
matrix (5). proved in algebra (5, Sec. 47) that the
It
is
symmetrical
form can factored into product of two linear forms
if,
quadratic be
a
Q
Q
r
said to be "singular."
is
Suppose that the x's are expressed in terms of other variables (yv y2, • • • ,
(A-2)
2
x<
UiVt
=
\
i
tu
det (rM). Then, conversely, the y's are determined uniquely by the x's.
said to define nonsingular linear transformation of the
is
Equation (A-2)
a
form in the y's. shown in algebra (5) that the coefficients can be
It
is
ti}
y?
= + y22 + • • •
+ vi* (A-3)
Q
2/12
where the rank of matrix (auj). Equation (A-3) called the canonical
is
is
r
form of the quadratic form Q. The number of positive terms in Eq. (A-3)
called the index of the quadratic form. determined
It
(denoted by
/)
is
is
The type of a quadratic form is determined by the rank r and the index
7, as follows :
if,
if,
(a) Q is positive definite and only n.
I
= <
if,
if,
positive semidefinite and only n.
is
(b)
Q
/
r
= =
if,
if,
(c) is negative definite and only and n.
0
/
Q
r
= <
if,
if,
negative semidefinite and only and n.
is
(d)
/
0
Q
r
< <
if,
if,
indefinite and only r.
is
(e)
I
0
Q
if
place, obviously negative
Q
if
are positive. Conversely, Q
is
Q
or semidefinite. If Eq. (A-3) contains both positive and negative terms,
Q
takes both positive and negative values, and therefore indefinite.
is
Q
Accordingly, statement (e) verified. The cases that remain to be con
is
sidered are those in which all terms in Eq. (A-3) have the same sign. If all
terms in Eq. (A-3) are positive (or negative) and = n, zero only all
if
is
Q
the y's are zero. Then, since det (tu) all the x's are zero. Conse r
0,
^
if
Eq. (A-3) are
is
Q
r
= • • • =
yr = However, we may assign arbitrary values to yr+v
0.
y2
• •
without affecting the condition = since yr+v y„ 2, • • • ,yn
0,
•
2/r+2. Vn
»
do not occur in Eq. (A-3). Then there are nonzero values of the x's for
which = Hence positive semidefinite (or negative semidefinite)
0.
is
Q
Q
may be derived from properties of the determinant D of matrix (aw) without
explicit consideration of transformation that reduces to sums and
a
Mr) where the rank of matrix (au), and the quantities Mt are any
is
r
is
a
= (b) M
one row and the same column from Mi+1 •• • — .
0.
2,
1
1
/
,
;
1.
r
diagonal of matrix (au) are zero, set M1 = and let M2 be any nonzero
0,
APPENDIX ON QUADRATIC FORMS 311
two-rowed principal minor of (a^). Such a minor exists, for, if all one-
rowed and two-rowed principal minors of (a,,) are zero, all principal
minors of (au) are zero. This conclusion follows from the following
theorem (5, Sec. 20, Theorem 2) :
0,
of matrix (a„) to form M2. If possible, choose the augmenting row and
column so that M2 Then augment M2 by numbers from one row and
0.
^
0,
if
possible.
If > and M2 = certainly possible to construct M3 ^ 0. This
0,
it
is
2
r
conclusion follows from the following theorem (5, Sec. 20, Theorem 1):
an r-rowed principal minor a symmetric matrix (au) of not zero,
If
is
by
and all the principal minors obtained adding one row and the same
if
column, and also all those obtained by adding two rows and the same
two columns, to Mr are zero, the rank of r.
is
(au)
Continuing with the preceding method, we obtain
nested principal
minors (1, Mv M2, •
M^), such that no two consecutive minors are
•
•
^
0.
Also, Mt1 ^ we can select l)-rowed minors of since all
if
Mr
0,
^
+
(r
(flw) are zero. Hence, there were no nonzero principal minor Mr con
if
1
r
0,
is
Lemma 1):
His real nonzero + )-rowed symmetric determinant, H" the
(k
If
is
a
if if
1
Hi de
of
in
H obtained
by
deleting the
of H, then
HH n = Hk+i,k+iHk,k — Hk+lh
2
(a)
To apply this theorem, let Mk = < r. Then and Mk+1
0.
0,
k
=fi
necessary,
permuted in the same way so that the elements forming Mk1 will lie in
the upper-left corner of the transposed determinant. This permutation
does not alter the values of the principal minors. Denote the permuted
Mk+1 and H and H". Since Mk = =
0,
determinants by Hk
0. 0
lc
0,
Two consecutive terms in the sequence (1, Mv M2, • • • , Mr) are said to
have a "permanence of sign" if they have the same sign. The number of
permanences of sign in the sequence is not affected by signs attributed to
the zeros, since a zero always lies between adjacent terms with opposite
signs. The following theorem is proved in algebra (37, Theorem 4) :
The index I of the quadratic form Q equals the number of permanences
• • •
of sign in any indicia! sequence (1, Mv M2, , Afr), wherein any zero
it
is
is
permuted as explained before, so that the leading principal minors are
Mr) in any preassigned
• • •
numerically equal to the terms (M1, M2,
,
indicial sequence more specifically, the leading /-rowed principal minor
;
a
the variables does not change the range of values of Q, the foregoing
theorem remains valid for any indicial sequence.
The following theorem occasionally useful
is
the determinant D
matrix (atj) contains any principal minor that
of
If
is
negative definite, negative semidefinite, or indefinite.
is
negative,
Q
by in
R
is
represented
(1, Mv M2, • M). Since M < the preceding theorem shows that the
0,
•
•
,
index of the quadratic form less than the number of rows in deter
R
is
if,
if,
n
r
5y2
Q
2s2
APPENDIX ON QUADRATIC FORMS 313
2 13-2
15 4 3
2 1 3
= 0, M3 = 1 5 4 = -35
3 4 2 7
-2 3 7 -10
3 4 2
2 1
M2 = = 9, Mi = 2
1 5
= anx2
+
a12xy dtfcz
Q
+ + (A-4)
+
is,
Let (I, rj, he another rectangular coordinate system with the same
0
origin as the system (x, y, z). Let the two systems be both right-handed or
314 ENERGY METHODS IN APPLIED MECHANICS
both left-handed. The cosine of the angle between any two axes of the two
systems may be represented by the following matrix:
TABLE A-l
X y z
S h mx «x
n h m<i n<£
I /. m3 nz
There are numerous identities among the direction cosines in this table
(see Sec. 7-8). The equations of transformation of the coordinates of any
point are
x = I J
+ l2V + /8£
y = m^ + + m3t
m2r) (A- 5)
z = nj + n2rj + n3t,
If Eqs. (A-5) are substituted into Eq. (A-4), the quadratic form adopts the
following form :
Q = bnP + bx2S-r, + baft
+ *2i»yf + b22V2 + biant (A-6)
+ butt + b^r, + b33?
btj = bH. It
is,
that the direction cosines in the preceding matrix may be chosen so that
Eq. (A-6) reduces to the form,
= bn? + bvtf + b^? (A-7)
Q
This theorem proved most easily by indirect means. Let denote the
S
is
+ attain
y2
+
1.
5
is Q
statement does not preclude the possibility that there are other points on
at which equals its value at Pv In fact, since the values of are equal
S
on two
S,
is is
OPx
a
by
is
S
P2.
OPv OP2, OP3 are mutually perpendicular. The sense of the vector OP3
is
•
APPENDIX ON QUADRATIC FORMS 315
defined as such that the three vectors OP^ OP2, OP3 form a right-handed
or left-handed system, according as the coordinate axes x, y, z are right-
handed or left-handed.
Let the f-, tj , £-axes coincide respectively with the unit vectors OPv OP2,
0P3. In other words, the vectors OPv OP2, OP3 are the unit vectors a, b,
c of the f, r\, £ coordinate system. With reference to the coordinates
of Eq. (A-6). =
(f,
It
V, £),
is
Q
/>31
= = to Eq. (A-7). The proof requires
0,
a
b12
a
= = =
by
0,
The point P1 has coordinates
0.
Consequently,
1,
f
r\
£
of of
by
Eq. (A-6), bn the value at point Pv Therefore, the definition
is
Q
point Pl3 bn the absolute maximum of on S. Accordingly, —
is
bn
0
<;
Q
the region S.
in
V2
(a)
Q
C2)
is
0
Q
proportional to the square of the distance from the origin, since the
R
G
rj,
0
f
£
,
at point
G
region Pv
G
is
u
rj
f
bn)) (b)
= = Then, by Eq. (b), G(l, = Hence,
If
622, set u
0)
bn b12. m, 2bn2.
never positive, =
0.
G
is
since b12
If b1l 622, set = —
bu\. Then, by Eq. (b)
u
b12/\b22
G(l,u,0)=bfb"±ub")
—
(c)
\°22 °n\
G never positive, Eq. (c) shows again that b12 = Accordingly,
0.
Since
is
0.
by
h = - MP + i2)
+
v2
<d)
Q
= 0. Also,
- Ml
f
G(l, 0, «) = + (633
- b^u] (e)
If 633
= bn, set u = bu. Then, since G is never positive, Eq. (e) shows
that bja — 0. If b^ # bn, set u = b^\bm = bn\. Then, again, Eq. (e)
yields b13
= 0.
Thus the theorem
is,
is proved; that reduces to the canonical form
Q
[Eq. (A-7)] when expressed in terms of the coordinates (f, rj, £). The
it
is
are called the "principal axes" of the quadratic form.
f,
axes rj,
£
a
principal axes of quadratic form, but more symmetric method may be
a
a
used. Equation (A-7) shows that the principal axes coincide with the
vector field grad Q. also shows that no two of the coefficients bn, b^,
It
if
633 are equal the principal axes are the only straight lines through the
origin that coincide with grad Q. If two of the coefficients bn, b22, £33 are
equal, Eq. (A-7) shows that the function rotationally symmetric with
is
Q
respect to one of the principal axes. The other two principal axes may be
chosen arbitrarily, provided only that the three axes are mutually perpen
dicular and concurrent at the origin. If all three coefficients bn, b22, b^ are
equal, any three mutually perpendicular lines through the origin are
principal axes.
The problem thus reduces to the determination of the straight lines
through the origin that coincide with grad Q. Any such line principal
is
a
axis and vice versa. Consequently, the principal axes are characterized
by
= 2XR, where = ix yy + kz, and scalar.
R
is
X
a
Q
a
A
is
z2)
to grad = + z2). This, in turn, equivalent to
y2
+
grad
is
0, Q
(x2
grad H = where
H=Q-
y2
+
In other words, for any principal axis, there exists constant such that
A,
a
these
equations yield
- + =
+
(«ii a12y
0
alsz
a^x + (a^
- X)y + = (A-9)
0
a^z
+ - =
+
X)Z
0
A
= (A-10)
0
Equation (A- 10) is a cubic equation in A. Consequently, it has three
roots Aj, A2, A3, although they are not necessarily distinct. By the theory of
linear equations (5), Eqs. (A-9) possess a solution x = a, y = b, z = c,
such that a2 + b2 + c2 = 1, provided that A is a root of Eq. (A- 10).
Supposing that A is a root of Eq. (A- 10), let us substitute the solution
(a, b, c) into Eqs. (A-9). Then, if we multiply the first, second, and third
of these equations by a, b, and c, respectively, and add the resulting
equations, we obtain
Q(a, - + + =
b,
b2
c)
A(a2 c2)
0
= this equation yields Q(a, b, c) = A; that
is,
Since a2 + b2 + c2
is
1,
A
the value of at point on the unit sphere. This conclusion shows that
Q
A1? A2,
the matrix (a„), since a„ may be any real constants, such that au = aH.
may easily be shown by specific numerical example that some of the
It
is
any root of Eq. (A-10) substituted into Eqs. (A-9), then the latter
If
is
satisfy the condition grad = 2AR. Since Aj, Aj, are the values of at
Q
Q
A3
the intercepts of the principal axes with the unit sphere, they do not
depend on the choice of in other words,
the coordinate system (x, y,
z)
;
they are invariants. Consequently, we may study the nature of the solution
letting (x, y, be principal axes. Thena^ = a31 = an =
by
Accordingly,
0.
z)
we may easily verify the following conditions, which remain valid when
are arbitrary orthogonal axes: Eq. (A-10) has no multiple roots,
if
y,
x,
then the principal axes of the quadratic form are unique and are deter
Q
double root, then the function axially symmetrical. The root that
is
Q
The other two principal axes may be any two perpendicular lines that are
also perpendicular to the axis of symmetry at the origin. If the three roots
of Eq. (A-10) are all equal, then the function spherically symmetrical.
is
Q
Since (Xv A3) are the values of at the intercepts of the principal
A2,
Q
axes with the unit sphere and since assumes the form of Eq. (A- with
7)
Q
respect to these axes, we obtain
+
Q
Equation (A- 10) may be expressed in the form
- J? + J2X-J3 = (A-12)
J
A3
0
where
= an + + = det
J3
Ji
a22 c33, (aw)
+
(A-13)
•*32 "33 a31 a33 ••21
0.
2,
is J3
is
is
0.
is
J3
J2
1,
0
A
A
root of Eq. (A-12). In general, the rank of matrix (au) zero root
if
is
is
r,
a
of Eq. (A-12) with multiplicity — r. Hence only nonzero terms appear
3
r
in Eq. (A-ll).
Extension to n Dimensions. By means of the idea of an n-dimensional
euclidean space, the principal-axis theory may be extended to quadratic
forms in variables (13). The principal axes are mutually perpendicular
n
system
(xv x2, xn).
,
coordinates that coincide with the principal axes, then the quadratic form
reduces to the canonical form,
= W W --• + Kyr2
+
(A-14)
Q
where the rank of matrix (aw). None of the coefficients in Eq. (A-14)
is
is
r
an
- Am
X
a12 a13
—
^
a.
a
n2 anZ
Only terms appear in Eq. (A-14) because — of the roots of Eq. (A-15)
n
r
are zero. Some of the A's in Eq. (A-14) may be positive and the others
negative. The number of positive terms in Eq. (A-14) the index of the
is
quadratic form Q.
APPENDIX ON QUADRATIC FORMS 319
jt)
the direction given by Eq. (4-2). convenient to consider the
(/,
m,
It
ri)
is
is
= + £e2. Equation (4-2) shows that when the point (x, y,
<f>
quantity
z)
e
<ff>
is
/,
a
/,
n
cartesian coordinates in space. Only points on the units sphere,
+ = have physical significance, but insofar as the theory of
+
m2
/2
n2
1,
ri)
is
<f>
point function in
(/,
quantity m,
ri)
is
space.
a
-
<f>
fax \Vv,
-
«»
<f>
\y,x hv,v
I
- - =
<£3
(A-16)
/3
/2<£
<f>3
I2,
/3
<f>3
are real. The strains corresponding to <f>v <f>2, are called the "principal
strains." According to the principal-axis theory, one of these roots the
is
absolute maximum value of for line elements through the point (x, y,
<f>
z)
;
stationary value of
<f>
a<f>
value. The stationary character seen from the fact that grad normal
is
is
of
zero at each
these points.
is
sphere
The principal axes of the quadratic form represented by Eq. (4-2) are
called the "principal axes of strain" corresponding to point (x, y, z).
by
Y/J + («, -
0
b>*Sn
+ fa* = (A-
- frn
0
<f>)m 17)
fas" =
J
+ + («,
b'
0
320 ENERGY METHODS IN APPLIED MECHANICS
Hence, if n^) and (/2, m2, n2) are principal directions, Eqs. (A-17)
yield
i(l + «i)(l + «2) cos 6 = 4>iikl2 + mim2 + nin2)
Since the vectors (lv mv and (/2, m2, n2) are perpendicular, this yields
cos 0 = 0. This conclusion means that if a body (e.g., a piece of putty) is
subjected to any continuous and differentiable deformation there exist three
mutually perpendicular line elements through each point that remain
perpendicular under the deformation. These line elements coincide with
the principal axes of strain. We may speak accordingly of the principal
axes of strain for the unstrained body or the strained body. For the
unstrained body, these axes have the directions (/j, mv n{), (l2, m2, n2),
<f>3
(l3, m3, n3) determined by Eqs. (A-17) from the three roots <f>v <f>2, of
Eq. (A-16). Line elements in these directions pass into line elements in the
principal directions m^, n^*), (/2*, m2*, n2*), (l3*, m3*, n3*) for the
strained body. According to the preceding conclusion, the three starred
unit vectors are mutually perpendicular, as are the unstarred ones. An
infinitesimal rectangular parallelepiped with edges parallel to the principal
directions consequently remains rectangular parallelepiped under the
a
deformation.
is
£
APPENDIX ON QUADRATIC FORMS 321
is,
nomial in A; that
F(X) = A0- - V3 + • • •
+ (- 1)MnA"
+
A1X A2P
may be shown that A0 = det (a^) and An = det (bu). More generally,
It
by
replacing columns
is
k
Ak
of det (au) by the corresponding columns of matrix (6w).
The equation = called the "characteristic equation" of the two
T F
is
0
(X)
quadratic forms and V. roots
It
An), provided
•
•
•
possesses A2,
,
that det (bu) although the roots need not all be distinct. shown
0,
It
is
(5, Sec. 59) that the roots are all real, provided that the
in
algebra
A<
quadratic form not indefinite nor singular.
T
is
^
2
(a)
Vi
0
1=1 i=i
the quadratic forms are transformed to
shown in algebra (5, Sec. 57) that the roots of the equation
is
It
det (a</ — A6,/) = are again L,, • An). In other words, the quan
•
0
•,
tities are invariant when the variables xt and yt are subjected to any
Xt
is
If
if
form,
of
the
equations in (a) to the form:,
of
type
= KAl«l2 + V
V
Xnu*)
J
(A- 18)
F(X).
Insofar as this theorem need not all be distinct.
concerned, the roots
is
Xt
The theorem shows that Kis positive definite, the roots are all positive.
if
?H
number.
In vibration theory desirable to find the matrix (tu) that effects the
is
it
a
322 ENERGY METHODS IN APPLIED MECHANICS
Zfo, (b)
i=l
If X is anyone of the zeros Xk of F(X), Eq. (b) yields a solution (tv t2, • • •
, fn),
such that all the r's are not zero. If no two A's are equal, the rank of the
matrix (au — is n — 1; this is apparent from Eqs. (A-18). Con
tn) corresponding to any root
•
sequently, the solution (tv
t2,
•
•
Xk
is
,
unique, except for an arbitrary constant factor. This factor determined
is
by the supplementary condition
= 3=1
l
i
if
is
•
•
•
(tv r2,
a
tn)
a
.
solution of Eqs. (b) and (c) so — • • •
—fn). However, the sign
t2,
is
(— tv ,
immaterial. The solution of Eqs. (b) and (c) for = may be written
Xk
is
X
the fcth column of matrix shown in the theory of quadratic
It
is
as (tuj).
a
forms (5) that the matrix (/0), constructed in this way, effects the reduction
of Tand Kto the forms of Eqs. (A-18).
There geometric interpretation of the preceding theory based on the
is
a
If of the determinantal
by
(48).
equation for the A's presents formidable computational problem.
a
Numerical schemes for treating this problem are presented in the book by
Faddeeva (23). (See also the last paragraph in Sec. 8-2).
The quadratic form 2ris positive definite. The matrices of the coefficients
in the two quadratic forms are
APPENDIX ON QUADRATIC FORMS 323
-X +
+ A 1-3A
This feduces to 2X2 + X — 1 = 0. The roots are X1
= \, X^
= — 1. Hence
-k -i + ^i r-i -f
-l + X1 1-3AJ L-i -iJ
Therefore, by Eq. (b), the r's corresponding to A1 satisfy the equation,
i*2 = 0. Also, by Eq. (c), =
— —
~ih 2t1t2 + 3r22 1. Consequently,
tl = 1/V6, = — These values form the first column of matrix
r2 l/VZ.
('«)•
Likewise,
"
1 -2
-2 4 .
— = — =
ri2
Consequently,
1.
= 2/V1, = l/VI. These values are the second column in matrix (tu).
r2
t1
Therefore,
= 1/76 2/V3•
(tu)
-1/V6 1/V3J
3.
form accordingly
is
»>1 2v2
,
U-,
- Ha »1
+
** y*
h
h
V6 V3
V,
2V = JV - = W
+
X2u22
2T = vj2 + v22
accomplished.
is
324 ENERGY METHODS IN APPLIED MECHANICS
significant range of p : that the functions atj(p) shall admit power series
expansions with nonzero intervals of convergence about any point in n.
Analytic functions have the following properties (a) They are continuous
:
(b)
(c) An analytic function that not identically zero vanishes at only
is
discontinuities in any closed subinterval of tt. The proof follows from the
Indian
Society of Theoretical and Applied Mechanics, Delhi, December, 1960.
APPENDIX ON QUADRATIC FORMS 325
Lemma 1. If
the functions atj(p) are analytic in tt, r(p) = K = constant
in any closed subinterval of n, except for a finite set of points on which
r<K.
A conclusion that follows from Lemma 1 is the following:
For proof, we observe that Lemma signifies that an open interval tt',
1
If the functions atjj(p) are analytic in it, if stable equilibrium exists when
p < (where
X A is some constant in tt), if the equation D(p) = 0
possesses a least root p0 in tt, and if D < 0 in an interval p0 <p < B,
then p0 is the buckling loadpcI.
for, if D
were negative in this range, equilibrium would be unstable in the
range p < X. Since Q is positive definite in the range p < A, r = = n I
in that range. Also, since D > 0, r = n in the range p < p0. Therefore,
by Lemma 2, r = = n in I
the range/> < p0. Consequently, Q is positive
definite in the range p < p0.
On the other hand, Q is indefinite or negative definite throughout the
interval p0 <p < B, since D is negative in this interval. Consequently, p0
is the greatest lower bound of the values of p for which equilibrium is not
stable. In other words, the critical value of p is p0 = pcI.
The preceding theorem implies that the determinant D itself usually
vanishes before any of its principal minors. This theorem consequently
often eliminates the need for an investigation of a sequence of minors of D.
By a different type of reasoning, based on the bifurcation concept, Poincare
(131) concluded that the equation D = 0 is a buckling criterion, but he did
not derive conditions under which the equation D = 0 ensures buckling.
An illustrative application of the theorem is given in Sec. 6-3.
In applications of the energy method to problems of buckling, the
following condition is usually assumed, but the literature on buckling
seemingly has not provided an algebraic proof of it:
/
p
/
r
r
r;
p p is is
/
pcT
tinuous. Therefore, the case that remains to be considered that in which
and are both discontinuous Let = = K for
at pcI. < pei, where
/
/
r
AT
b
I
a
a
Q
b
a
= =
0;
for/>
b
definiteness.
APPENDIX ON QUADRATIC FORMS 327
PROBLEMS
1. Determine to which of the five classes the following quadratic forms belong:
— y2 + 4z2 + Gyz + 8zx + 2xy
(a) 3x2
— Sxy + 2xz — Ayz
(b) 5x2 + 5y2 + z2
+ ly2 + - xy - -
- cd
(c) 10a:2 5z2 + 12w2 2zz 3zw
(d) + *2 + \<? + id2
- 4ad -
a2
3. Q = (2 - 6x)A2 - 4(1
For what range of x
is Q positive definite? Positive semidefinite? Negative
definite? Negative semidefinite? Indefinite?
4. Show that there are infinitely many different linear transformations of the
variables that will reduce the quadratic form anx^ + 2a12a;1a:2 + a22a;22
to the canonical form if none of the a's is zero.
Q in terms of
£).
6. quadratic
Q
a
~
~3
0
2
2-104
(an) =
10
6
5
-2_
4
0
7. Q = x12+ 4x1x2 — 2x22, P = y-f + y22. Determine the matric (fw) and
reduce the two quadratic forms to the canonical form.
8. Q = 3xj2 — 2x1x2 + 4x22, P = y? + 2y1y2 + 2y2. Determine the matrix
(tu) and reduce the two quadratic forms to the canonical form.
9. Q = 3x^ — — + 2x1x2 — 2xjX3 +
-
2x2 x32
Reduce the two quadratic forms to the canonical forms without determining
the matrix (%).
10. Develop the principal-axis theory of stress with the aid of the principal-axis
theory of quadratic forms.
Answers to problems
CHAPTER 1
6. = r/2A.
7. W = -7084A:/15.
11. W/F = 2R/(R -r).
13. Fn = \Lbn.
14. U = Wx2 + y2 + V(L - x)2 + 2/2]F.
15. (dr\dif = 4a/(mr) - 46/(9wre) + constant, r8 = 6/a.
18. co = (nV/L2)V£7/p.
25. F — -2irP.
26. K = 2nkpa/Vx2 + a2, F = 2nkpax/(x2 + a2)*.
X+L+V(x+L)2+y2
27. K = —kp log
x -L + V(x -L)2 +y2
CHAPTER 2
* = (n -
1)£//L,
- 1£a1Ls)Bi
equal.
- - Sa-^AB
1.
3.
4. 4.785.
«
2L4/>0 ( *j
-!)"+! ,W*c
5.
•nsEl
2v
-1= «
- sin
1
6. 74.26%
- r) - r) - r)
1 1 1 1 1
-r - r) - r)
1 1 1
V2 + ; +
1 9(9 25(25
8. an = 0 if « is a multiple of 4. Otherwise, an = FL3/(2if*£/«*).
•
M =Pa r
3 1
10. AT = P/V3, F
l.2ir(l +Z) V3.
Fa Fa cos n6
11. + a1 cos 6 + J,
277£/4(l +Z) TrEAZnt2(n* -l)2
Fa \ Fa "
( sin «0
sin 6 —
nEAZn%n(n2 - l)8
where is an arbitrary constant.
- - -
a1
12. Z = + 2R2/b2
1 2(R/b)VR*/b2 1, where 2b is the width of the cross
section and R is the radius of the centroidal axis.
"
N MV z
13. <*=—+— 1 + , — —7= •
/I (a + z)Z.
Aa[.
14. h = 0.09078F2/A:2, ,/V0
= 0.301 3F, JVj = 0.2804F,
ATg = 0.2130F, N3 = 0.
15. x = vertical displacement of joint y = horizontal displacement of joint 1,
2,
17. Left wire 237 lb, center wire 126 lb, right wire 537 lb.
- -P
23. NA13 NA0 NAC =
2(1 + V2) 2V2
24. AT:
= 14,1501b, Nt = 16,4201b, N3 = 17,6501b.
ANSWERS TO PROBLEMS 331
- = 0.
-
29. 4O0J + 402 + 1004 + 0.02xj 0.20x2
0.20xa = 0.
-
40j + 1802 + 503 + 0.20^
0.20r2 = 0.
-
502 + 1803 + 404 + 0.20x!
0.20r2 = 0.
100! +
-
403 + 6004 +
1006 + 0.02x!
0.18x! = 0.
1004 + 3208
- - -
-
0.020J + 0.2002 + 0.2003 + 0.0204
- - - + 0.0241 3x1 0.01333^,
-
0.1 806 0.
-0.200j 0.2002 0.2003 0.2004 0.01 333^ + 0.01 333x, 100.
30. 0j = 03 = 4Fa/129#, 02
= Fa/129/:.
x = 26Fa2/387tf.
31. x =Pd3/l2EI.
32. 0.003737°, -0.0006228°.
3V3M .
33- v~O' r = Vl'A-
U=2EA(l+9r*/Lr
U
34- B
6°
= 0
°3
=
(1 + 3r)Fa/A.
. h = 02
= -Fa\Kv- .
6r + x+ 2rX 6r k
EAaW
X ~
(1 + 2r)Fa2/Ktt
' * ~ ' A " '
6r + X + 2rX Yv tfvC3
-9b2F
35 "_ •
32a(a + 2b)
36. Let r =
(3ja)xt = (£)Fb/Kv. -
2(2 + r)0j + 02 + r63
p^j30Kv. - -
0! + 2(1 + r)62 + r04 + (3/a)xi (3/a)x2
(3/a)x, = -(A)Ftytf,,. -
r0! + 2(2 + r)03 + 04
(3/a)x, = -p^j20Kv. -
rd2 + 03 + 2(1 + r)04 + (3/a)a^
- - F,//Te.
-(3/a)0j -
(3/a)02 + (3/a)04 + (24/a2 +
(3/a)02
- (3/a)03
-
EAb^fK^x^
(3/a)04
- (12/a2)xi
V2ja2)x2
+ (12/a4)^ - FjKv.
37. Xj and x2 denote displacements of joints
and 2 normal to plane of frame; 1
x3 and xt denote rotations of joints 1 and 2 about axis 12; xs and xt denote
rotations of joints 1 and 2 about axes 01 and 02. Let Kh/Kv — r.
- - - - -
2(l/a2 + r/bt)x1 2(r/b*)x2 (l/a)x3 (r/b)x6 (r/b^
- F/3Kv.
-
-2(r/^)xi + 2(l/a2 + r\b*)x2
GJ/bK^
(l/a)z4 + (r/6)x6 +
- (GJ/bK^
(rjb)xt 0.
= 0.
-(3/a)xi
-Q/aK - (GJ/bKv)x3
+ (2 +
+ (2 + GJjbK^ -
-
0.
-(3r/6)z, + (3r/6)x2 + (2r + + nr, 0.
-Or/b^ + (3r/6)x2 + rxs + (2r + GJ\aK^ - 0.
CHAPTER 3
3. x = asec\(e b). -
A. y = pxiL -
zYL1 + Lx - xt)f2AEI.
5. y = p&HL - zX2L2 - - xi)/l20EIL.• Lsc
y" + jT/fl* + -
6.
At = L,
ar - =0 y/a4
y~ y/a*
/>/2Jfc.
and y0 + //a + y/a* - 0.
332 ENERGY METHODS IN APPLIED MECHANICS
7. u =
Fa „
——
2EAZ
9
.
sin 9,
.
v = —Fa—
2EAZ
(6 cos 6
- sin 6), a = —
—Fz cos
—— 9
.
AZ(a + z)
-M M
-COS0), -sin9],
^EAz^1 +Z)6
8. u
=£^d »
m r / i\z"
+
IT7)L1
,4(a (,+z)« •
9 .M = • - ~ « • + •
^[(7r2/4)1(1++2?)-2(C05 J) 5
^ =
P „ „ • „ (1 + 2Z) cos 9 I
^-cos9-9s,n9-(g2/4)(1+z)_2j,
where w is weight per unit length.
10. f(x) = A [(sin kL — sinh *Z,)(sin kx — sinh
*x)
+ (cos kL + cosh A:L)(cos kx — cosh fcc)].
11. y =f(x) sin «/, /(x) = j4 sin (vnxjL), v = integer,
« = (vV/L2)V£^•.
12. * = 0.617, A = 0.617, C = 0.
dF d dF d 3F d2 3F 32 3F 32 3F
3m dx dUX dy duv dx2 duxx dx dy duxy dy2 duvv
d3 dF S3 dF 33 dF d3 dF
dx3 diixxx dx2 dy du^y dx dy2 du^y dy3 duyyl
2nZ - -
w0
u =
- -
22. -(Pa/SEAZ)[2 sin 9 + (3w + 4 4/77) cos 9
+ (4/jI)9 sin 9
- -
29 cos 9 4].
- -
v = (Pa/8£,4Z)[-4(1 + Z)9 4 cos 9 + (3tt + 2ttZ 4) sin 9
(4/w)9 cos 9 29 sin 9 + 4], 9 > 0.
tepatb* p
ii. W0
=
.j. £2)2'
wxxxx ■ "mv ■ ^vvvv =
£j
•
^2)^2
CHAPTER 4
f3=-l+Vl7, e= = J,
i
1. 2, 35.11. /3
2. X y z ei
5. X y z
= 10,000.
Tmax
20. 5.4061b.
21. PbL3/3E + PaLx\YlE.
<5 =
= 3P/4n.
22. M0 = -3Pa/4w, N0
23. Pfr.
Pa3 Pa3
+
2*-8=2EI 2GJ^-X)-
25-No-2(y>L-Ixx)-
PWxx - hLxy + Ixy(h - y)]
2(?L -U
1).
1),
=
Kr0(l
+
26. « = +
s
4*r.(l
27. M0 = Pa62/L2, A/! - -Pa2b\L2.
= Pa2(36
+
= Pb2(3a + b)/L3, /?! a)/L3.
R0
ui*.M ^1
+
+
= aAT)
+
[(«
28.
?
n
=
29. ffj = 2t,
I
ct2 ff3 Tmax
CHAPTER
5
JO - =
+ +
a„
-
»)«v*
+
1.
=
Id i(i
+
+ +
+
v)cxx vvv
and = 0.
+
= + v)kd vx «v
On edge x = a, iix
(1
vvv
+ By sinh p0/D.
+
=
^4
-
cosh
2/
-
y
2.
/G/)
b]
sinh cosh
- -
b
v)
vP0[(l + v)/(l •
~
cosh v)b]
(1
sinh
6
b
+
v)
D[(3
vPn sinh
- -
b
•
cosh v)b]
(1
sinh
6
+
6
-
v)
-
£>[(3
+ q/D = 0.
-
v)wxxy (EI/D)wxxxx
(2
3-
+ +
H,yyv
= 0.
Wvv VH,xa; (GJ/D)WXXv
16/>
sin (mn\T) sin (/jtt/2) sin (rnncja) sin (nirc/b)
•
°mn = ^Dmn(m2ja2
+
n2/b2)2
«rr
= )t(l
-
+
At = a, v)0.
+
-
uT (>'/a)«
r
= 0.
(S/r2) +
+
[(1 v)/2r]H'r2
6. «rr + (l/r)fir
+ rile,,
%rwr2w„ + rHrrH'r
+
=
0.
+ [(1 v*)/Eh]pr
ANSWERS TO PROBLEMS
U = nD sin
7.
aj jaw^2
+ (l/x)wx2 + 2vwxwxx
Qmn =0 if m or « is even.
16*/S(1 + v)
mn if m and « are odd.
JmnHr#\a2 + «2/62)
r2
= x tan a, a = 1, /S
= x sin a + z cos a.
<£,
r1
n
a = a + =
<£.
+ sin
(a
z,
z)
/S
+
/4
9,
14. 62 cos2
1,
P
<f>
=
a2Ne Fu + F.
= cot
-F+o csc + Fe esc
^.
^
cPN+e
^
cV cosh2 x
= = — cos2
A
18.
B
•4•
AXFX AyFy.
A3NXy — AFXy + AyFX + A%F y.
21. ux+ (vwja) = + (Ehja2D)w =
0,
0.
At x = = -M0/Z>, hw
0,
>vra <20/Z>.
/>(a2 -r2)i
22. w
64 D
CHAPTER
6
Per =
1.
Per =
2.
Ara3.
Per
Pcr = 2A:c3/a2&2, F„ = 2A:c/A2, = a2 + b*.
c2
4.
Per = 42P//L2.
7.
is
r
336 ENERGY METHODS IN APPLIED MECHANICS
9. = A sin kx + B cos kx + Cx + D
r)
kL
= tan kL. Hence PCI = 2Q\^EI\12.
^- + V(AT/i))(«V/a2) , 0 =
- + ^(Ar/DX«V/a2) .
. a J [_ a
14. Fw
- 12£7 '
F
- •
L2 Fcr 3(sin 6 0 cos 6)
i5-pcr~—
Tr2^/ t^AD 2Z>(1
- v)
a2 +^~T+ b
16. L = ir^AnEIjy^ -
4) or L3 = 8.3£//y.
17. £V*f + + Nytvv = 0.
—
6(1 v')L
CHAPTER 7
- F
1. J> = 0, r r&2 + — = 0, r26 = constant.
+ 2/tf sin = 0.
+
g
x = —
12. -fs0 sin cof.
-
x0
13. r(j>
f
-
= constant.
- C/r3 -
+
r2(<f> co)
(<£
e
r
ANSWERS TO PROBLEMS 33
- flj) = M/Iv -
-
15. S1 (G//a/1)(02
02 + (Gy/a/2)(02 Sj) = 0.
Hence + /jo", = A/.
x — co2^ —
16. <by — 2coy = Fxjm.
y —
+ 2cox = Fv//w.
r -
oA/ + cox
C-
^
(-A C
(/I
-
29. + /4')fl + + + B')42 sin 0 cos 0 + sin 0
C<f>y>
+ k(8 Trj2) = 0.
$[(A + £') sin2 0 + (C + C") cos2 6 + /] + Cip cos
+ 6^2(A + B') sin 6 cos 0 2(C + C) sin 6 cos 6] - -
6
<*
+
0
17/2
^
+ 5' + = = constant.
+
0,
(y4
y
/)<£ A:^
<j>^
c/0
a
dy>
0
^
CHAPTER
8
mx + kx = F(t).
1.
mx + (48F///3)x =
2.
0.
+ k6 =
c6
+
3.
16
0.
- V5)/2]*/m.
V
= =
[(3+ V5)/2]*/m,
4.
V[(3
- l)/2.
«>!
co2
= (^5 D/2, -(V5
+
co2 0.116.
a1=0.198, a3
If
a = av 22/2!
= 1.802, zjz1 = 2.247.
If a = a2, zj)^ = 0.445, zjt, = -0.802.
If a = a3, z^Zj = —1.247, z3/zj = 0.555.
con (MU*)VEi[>.
otxj 2A:Xj — fcc2 — kx3 =
26^ -
■
ky2 - ky3 =
0, 0,
10.
+ +
wj/j ••
-.
co = V3A:/w.
g" + o>2g
= 0.
- 1/r) Dfr + v Df„]
- pm2rf = 0.
(d*/dr*)(r Dfrr +
- (l/r>rr
v
Dfr) (<//*>•)[(
= Pco2/£/.
/J4
16. tan = tanh pV,
/S/
/9*
p7
1
p7
sinh sin
+
0/
—— fa — cos px)
T
.
= (sinh fix
/I
J
[.
p
18. comn
f(x, y) = Amn sin (rmxjd) sin (mryjb).
19. flf< = ,c20ra,
= G7R =f(x)g(t).
c2
0
= sin (mrxjl).
A
o)n
= /(a?)
= (djdx)(EAux)\ — mass per unit length.
20. putt
p
= 0.
+ \\Z)u + ve/Z + (a2pjEZ)uH
+ 2uee +
-
(1
21. ugooe
=
+ v
u
= 0.
0" + 2f 1/Z a2pco2\EZ)<\> + tp'jZ
(1
+
= 0.
<£'
V* + + (a2co2p/£)y
2Z
•
C"
W+1 Po
(1
G(u„ - w„)
ft.
- (l/r>
23. 39.2 ft/sec, 219
puu. =
- - - -
+
+ 2C)(«rr + (l/r)«j.
+
wZT)
(A (A
24.
+ (l/r)«, + (G/rX«, G(«„
h>„)
w„) = wr) Pwtt.
+ 2G)(uzr
APPENDIX
(c) Positive
definite, (d) Positive
(a) Indefinite, (b) Positive semidefinite.
1
.
semidefinite.
= (7a
(e) Indefinite.
2c 4d)(a + 3b - - - 5c + d).
2. (a) Not factorable,
+
b
Q
(b)
iy)(x — iy). (d) Not factorable.
-
=
+
(x
Q Q
(c)
= (2x + 2/)(3x 5y 7z).
+
(e)
Set a = 4(25
- = 4(25 + V50).
VJTT),
b
3.
x = a, indefinite
if
if
positive definite
is
a Q
x >
if
b.
x
y
(5)
\
= +
|2
Q
-|^2
ANSWERS TO PROBLEMS
—
6. Q -7.582^ + 9.15l2/22 + 3.43l2/3s
"2/V5 I/V5I
7. (t„)
.1/V5 -2/V5J•
Q = 2^ - «,2, i> = %2 + >;22.
"I/V5
8. (/„) = -3/V5]
.1/V5 2/V5J•
Q
9. Q = -0.664^ + 0.095f22 + 7.9261a2
P
Bibliography
The following incomplete bibliography lists some books and articles
that are commendable for their presentations of special topics or for
general developments in mechanics. The well-known treatises by Whit-
taker and Appell are classics in the mathematical aspects of the sub
ject. The works of Routh, Osgood, Goldstein, and many others are
also noteworthy. The introductory book by Planck provides an excellent
physical insight into the foundations of mechanics. Lanczos' work on
variational principles of mechanics combines an advanced treatment with
a lucid and penetrating philosophy. The treatise by Courant and Hilbert
provides an excellent mathematical background for the variational theory
of mechanics. Variational methods of approximation are treated in the
works of Kantorovich and Krylov, Courant and Hilbert, Collate, Biezeno,
and Grammel, and Temple and Bickley. The recent historical work by
Dugas gives a comprehensive survey of the development of ideas in
mechanics.
Among the engineering works on mechanics, the treatise by Biezeno and
Grammel is noteworthy. It contains a good presentation of variational
principles in elasticity theory. This topic is also treated admirably in the
book by Sokolnikoff. General principles and engineering applications in
the mechanics of rigid and elastic bodies are developed precisely and
concisely in the recent work by Parkus. Applications of energy principles
in analyses of engineering structures are emphasized in the works of
Argyris, Hoff, Phillips, and Charlton. Variational principles play a sig
nificant part in modern plasticity theory; the work of Prager and Hodge
treats this topic.
BOOKS
1. Appell, P., Traiti de mecanique rationnelle, vols. 1 (1941), 2 (1953), 3 (1952), 3rd
ed., Gauthier Villars, Paris.
2. Argyris, J. H., and S. Kelsey, Energy Theorems and Structural Analysis,
Butterworth., London, 1960.
3. Biezeno C, and R. Grammel, Engineering Dynamics, vols. 1 and 2, Blackie and
Son, London, 1956.
340
BIBLIOGRAPHY 341
73. Schleusner, A., Strenge Theorie der Knickung und Biegung,Teubner, Leipzig, 1937.
74. Seely F. B., and J. O. Smith, Advanced Mechanics of Materials, 2nd ed., Wiley,
New York, 1952.
75. Sokolnikoff, I., Mathematical Theory of Elasticity, 2nd ed., McGraw-Hill,
New York, 1956.
76. Sokolovsky, V. V., Theorie der Plastizitdt, VEB Verlag Technik, Berlin, 1955.
77. Sommerfeld, A., Mechanics of Deformable Bodies, vol. 2 of Lectures on Theoretical
Physics, Academic, New York, 1950.
78. Sternberg, W., and T. Smith, The Theory of Potential and Spherical Harmonics,
University of Toronto Press, Toronto, 1946.
79. Struik, D. J., Differential Geometry, Addison- Wesley, Cambridge, Mass., 1950.
80. Synge, J. L., The Hypercircle in Mathematical Physics, Cambridge University
Press, 1957.
81. Synge, J, L., and B. A. Griffith, Principles of Mechanics, 3rd ed, McGraw-Hill,
New York, 1959.
82. Temple, G, and W. Bickley, Rayleigh's Principle and Its Application to Engineering,
Dover, New York, 1956.
83. Timoshenko, S., Vibration Problems in Engineering, 3rd ed., Van Nostrand
Princeton, N.J. 1955.
84. Timoshenko, S. and J. M. Gere, Theory of Elastic Stability, 2nd ed, McGraw-
Hill, New York, 1961.
85. Timoshenko, S., and J. N. Goodier, Theory of Elasticity, 2nd ed., McGraw-Hill,
New York, 1951.
86. Timoshenko, S., and S. Woinowsky-Krieger, Theory of Plates and Shells, 2nd ed.,
McGraw-Hill, New York, 1959.
87. Tong, K. N., Theory of Mechanical Vibrations, Wiley, New York, 1960.
88. Van den Broek, J. A., The Elastic Energy Theory, Wiley, New York, 1931.
89. Weinstock, R., Calculus of Variations, McGraw-Hill, New York, 1952.
90. Whittaker, E. T., Analytical Dynamics, Dover, New York, 1944.
ARTICLES
91. Benjamin, R. J., Analysis of Oval Rings by Fourier Series, /. Aeronaut. Sci., 19, 9,
September 1952.
92. Bleich, F., and H. Bleich, Bending, Torsion, and Buckling of Bars Composed of
Thin Walls, Prelim. Pub., 2nd Congr. Internat. Ass'n Bridge & Structural Engrs.,
English ed., 871, Berlin, 1936.
93. Boresi, A. P., A Refinement of the Theory of Buckling of Rings under Uniform
Pressure, /. Appl. Mechanics, 22, 1, March 1955.
94. Brown, C. L., The Treatment of Discontinuities in Beam Deflection Problems,
Quart, appl. Math., 1, 4, January 1944. -jy, ^ y <ir~
95. Brown, E. H., The Energy Theorems of Structural Analysis, Engineering: 1-
Definitions and Fundamentals, 305-308, March 11, 1955; 2-Derivation and
Discussion, 339-342, March 18, 1955; 3- Worked Examples, 400-403, April 1,
1955.
96. Brush, D. O., Strain-Energy Expression in Nonlinear Shell Analysis, J. Aeronaut.
Space Sciences, 27, 7, 555-556,July 1960.
97. Bryan, G. H., On the Stability of Elastic Systems, Proc. Cambridge Phil. Soc,
VI, 199-210, 1886-1889.
98. Bryan, G. H., On the Stability of a Plane Plate under Thrusts in Its Own Plane,
with Applications to the "Buckling" of the Sides of a Ship, Proc. London Math.
Soc, XXH, 54-67, 1891.
344 ENERGY METHODS IN APPLIED MECHANICS
99. Castigliano, A., Theorie de l'equilibre des systemes elastiques et ses applications,
Turin, 1879.
100. Charlton, T. M., Some Notes on the Analysis of Redundant Systems by Means of
the Conception of Conservation of Energy, /. Franklin Inst., 250, 6, December
1950.
101. Charlton T. M. The Analysis of Statically Indeterminate Structures by the
Complementary Energy Method, Engineering, 174, 4521, September 19, 1952.
102. Chien, W. Z., The Intrinsic Theory of Thin Shells and Plates, Quart. Appl. Math.,
1, 4, January 1944.
103. Donnell, L. H., Stability of Thin-Walled Tubes under Torsion, NACA TR 479,
1933.
104. Felgar, R. P., Jr., Formulas for Integrals Containing Characteristic Functions of a
Vibrating Beam, Circular No. 14, Bureau of Engineering Research, University of
Texas, Austin, 1950.
105. Fourier, J., Memoire sur la statique contenant la demonstration du principe des
vitesses virtuelles et la theorie des moments, Oeuvres de Fourier, 2, Gauthier
Villars, Paris, 1890.
106. Friedrichs, K. O., On the Minimum Buckling Load for Spherical Shells, Theodore
von Karman Anniversary Volume, Calif. Inst, of Technology, 1941 .
107. Friedrichs, K. O., and J. J. Stoker, Buckling of Circular Plates Beyond the
Critical Thrust, J. Appl. Mechanics, 9, A-7, 1942.
108. Goldberg, J. E., On the Application of Trigonometric Series to the Twisting of
I-Type Beams, Proc. First U.S. Nat. Congress Appl. Mech., ASME., New York,
1952.
109. Goodier, J. N.,Some Observations on Elastic Stability, Proc. First U.S. Nat.
Congress Appl. Mechanics, ASME, New York, 1952.
1 10. Goodier, J. N., Discussion of paper by E. Reissner, J. Appl. Mechanics, 13, 3,
A-251, 1946.
111. Goodier, J. N., Torsional and Flexural Buckling of Bars of Thin-Walled Open
Section under Compressive and Bending Loads, Trans. ASME, 64, A-103,
1942.
112. Goodier, J. N., Flexural-Torsional Buckling of Bars of Open Sections Under
Bending, Eccentric Thrust, or Torsional Loads, Cornell Univ., Eng. Exper. Sta.
Bulletin 28, 1942.
113. Goodier, J. N., The Buckling of Compressed Bars by Torsion and Flexure,
Cornell Univ. Eng. Exper. Sta. Bulletin No. 27, Dec, 1941.
114. Gossard, M., P. Seide, and W. Roberts, Thermal Buckling of Plates, NACA TN
2771, 1952.
115. Hamilton, W. R., The Mathematical Papers of Sir W. R. Hamilton, II, Dynamics,
Cambridge University Press, 1940.
116. Hoff, N. J., A Strain Energy Derivation of the Torsional-Flexural Buckling Loads
of Straight Columns of Thin-Walled Open Sections, Quart, appl. Math., 1, 4,
341-345, January 1944.
117. Kappus, R., Drillknicken zentrisch gedriickter Stabe mit offenem Profil im
elastischen Bereich, Luftfahrtforschung, 14, 44, 1937. Also: Twisting Failure of
Centrally Loaded Open-Section Columns in the Elastic Range, NACA TM 851,
1938.
118. Von Karman, T., and H. S. Tsien, The Buckling of Spherical Shells by External
Pressure, J. Aeronaut. Sci., 7, 2, December 1939.
119. Langhaar, H. L., General Theory of Buckling, Appl. Mechanics, Revs, 11, 11,
November 1958.
BIBLIOGRAPHY 345
145. Temple, G., The General Theory of Relaxation Methods Applied to Linear
Systems, Proc. Roy. Soc. (London), A, 169, December-March 1938-39.
146. Trefftz, E., Zur Theorie der Stabilitat des elastischen Gleichgewichts, Z. angew.
Math. Mech., 13, 160-165, 1933.
147. Trefftz, E., Konvergenz und Fehlerschatzung beim Ritzschen Verfahren, Math.
Ann., 100, 503-521, 1928.
148. Vlasov, V. S., Basic Differential Equations in the General Theory of Elastic Shells,
NACA TM 1241, February 1951.
149. Wagner, H., Verdrehung und Knickung von ofTenen Profilen, 25th Anniversary
Volume of the Technische Hochschule, Danzig, 1929. Also: Torsion and Buckling
of Open Sections, NACA TM 807, 1936.
150. Westergaard, H. M., On the Method of Complementary Energy, Trans. ASCE,
107, 765-803, 1942.
151. Young, D., and R. P. Felgar, Jr., Tables of Characteristic Functions Representing
Normal Modes of Vibration of a Beam, Engineering Research Series No. 44,
University of Texas, Austin, 1949.
152. Ziegler, H., On the Concept of Elastic Stability, Advances in Applied Mechanics,
IV, Academic, New York, 1956. TA 1 f' ' ,
153. Ziegler, H., Die Stabilitatskriterien der Elastomechanik, Ingen. Arch., 20; 49-56,
1952.
Index
Action, 239 Buckling, of rectangular plate, 214-216
Adiabatic process, 13, 14 relation to quadratic forms, 324-327
Airy stress function, 130, 163, 193 Trefftz's criterion of, 21 1
Atwood machine, 262 Buoyancy, center of, 38
70 2
6296
r CONTROL.MARK