Summary AE3-914 2008-2009

Course Dynamics & Stability

Taught by Prof.dr.ir. Miguel A. Gutirrez
I Introduction .................................................................................................................................................... 2
Laws of Newton ........................................................................................................................................... 2
Relative motion............................................................................................................................................. 2
True motion................................................................................................................................................... 2
Fictious forces ............................................................................................................................................... 2
II. Qualitative Analysis .................................................................................................................................... 3
Kinetic energy ............................................................................................................................................... 3
Inertia Tensor ................................................................................................................................................ 3
Virtual work .................................................................................................................................................. 3
III. Generalised coordinates ............................................................................................................................ 4
Degrees of freedom ...................................................................................................................................... 4
Generalized velocities .................................................................................................................................. 4
Generalized momenta.................................................................................................................................. 5
Generalized work ......................................................................................................................................... 5
IV. Lagrangian Dynamics ............................................................................................................................... 5
V.......................................................................................................................................................................... 6
VI. Ignorables and constraints ........................................................................................................................ 6
Jacobi energy integral .................................................................................................................................. 8
Steady motion ............................................................................................................................................... 8
Raleigh dissipation function ....................................................................................................................... 8
Lagrange multipliers.................................................................................................................................... 9
VII. Dynamic systems ...................................................................................................................................... 9
Linear dynamic systems ............................................................................................................................ 10
Linearization ............................................................................................................................................... 10
VIII. Stability of conservative systems ........................................................................................................ 11
Stability of Lagrangian systems ............................................................................................................... 12
IX. Dynamics of rotating bodies: Gemist .................................................................................................... 12
X. Lagrangian dynamics ................................................................................................................................ 12
Euler angles ................................................................................................................................................. 12
Rotation symmetric bodies ....................................................................................................................... 13
Rotation non-symmetric bodies ............................................................................................................... 13
XI. Functionals ................................................................................................................................................ 14
Euler-Lagrange equation ........................................................................................................................... 14
XII. Variational operator................................................................................................................................ 15
Generalisation ............................................................................................................................................. 15
XIII. Hamiltons principle ............................................................................................................................. 16
XIV. Ritz method ............................................................................................................................................ 17

I Introduction
The course is based on Analytical Mechanics with an introduction to dynamical systems by
J.S.Torok, which can be used for background material and exercises.
Laws of Newton
In the second law of Newton momentum is referred to as motion and equals

p = mv

, such that

d
F = p = dt mv
F = F ji
. The third law yields ij
. All motion is defined with respect to an inertial
Vxyz =
xyz = 0

reference system, so

constant and

Relative motion

r rel = xi + y j + zk
+ y j + zk

v rel = xi

arel =
xi +
y j +
zk
True motion
The position in the xyz-frame with respect to the inertial reference frame can be found using:

r p = r xyz + r rel

dj
dr xyz dr rel
d
di
dk
+ y j + zk
+x +y +z
rp =
+
= v xyz + xi
dt
dt
dt
dt
dt
dt
= v xyz + v rel + xxyz i + yxyz j + zxyz k

vp =

= v xyz + v rel + xyz r rel

From which may be concluded that

d
vp =
dt
= v xyz + r rel + r rel

+ u , such that:
u = ue

ap =

+ v rel

= a xyz + r rel + ( v rel + r rel ) + arel + vrel

= a xyz + r rel + ( r rel ) + 2 v rel + arel
Where

2 vrel is referred to as the Coriolis-acceleration.

Fictious forces

marel = F m a xyz + r rel + ( r rel ) + 2 vrel

Now

F = ma

or

Example: double deck

marel = F F fict

)
.

II. Qualitative Analysis

Work is defined as

such that

r2

t2

r1

t1

W = F dr = F rdt . In a conservative force field a potential can be defined

F = V =

V
V
V
i
j
k and = T + V = constant.
x
y
z

Example:

Fx =

dV
dV
=0
= mx mx +
dx
dx
or

dV
dV
d 1

x mx +
x = mx 2 + V ( x) 1 mx 2 + V ( x)
= mxx +
dx
dx
dt 2

or 2
=constant.

Kinetic energy
Kinetic energy is defined as
With v = v0

+r:

1
1
dT = v vdm or T = v vdm .
2
2

1
(v0 + r ) (v0 + r )dm = 1 mv02 + v0 rdm + 1 ( r )( r )dm

2
2
2
1
1
= mv02 + T I
2
2

T=

Inertia Tensor

I xx

The inertia tensor I = I xy

I xz

I xy
I yy
I yz

I xz

I yz can be defined for two specific cases: I0 wrt a fixed point

I zz

and IG wrt the mass center of a body. Now fixed point rotation yields
motion

T=

1
T = T I 0 and general
2

1 2 1 T
mvG + I G .
2
2

Using principal axes such as symmetry axes gives a diagonal matrix.

For a xy-planar body

I zz = I xx + I yy and I xz = I yz = 0 .

For a body with at least three symmetry axes in the xy-plane:

Virtual work
-

I xx = I yy and I xy = 0 .

III. Generalised coordinates

Example: Pendulum

l,m

g
m + m sin = 0
l
g
+ sin = 0
l
which is conservative!

g
d 1 2 g

cos = 0
l
dt 2
l

1 2 g
cos =constant.
2
l

+ sin =

This is however not expressed in the right coordinates and units and therefore can not be energy.
The generalised coordinates

{q , q ,, q } are defined as q
1

= qi ( x1 , x2 ,, xn , t ) or

xi = xi q1 , q 2 , q n , t with i = 1 n and n is the number of degrees of freedom of the system.

Degrees of freedom
The number of degrees of freedom are the number of values that need to be fixed to determine the
state of the system. In general:
2D particle: n=2,
2D body: n=3,
3D particle: n=3,
3D body n= 6.
N particles or bodies have N n degrees of freedom.
Holonomic constraints are defined as

f q1 , q 2 , q n or f q1 , q 2 , q n , t being constant, such that

ndof = ncoord ' s nconstr ' s .

This way a body can be interpreted as a collection of particles with fixed relative distances.

Generalized velocities
When

xi = xi q1 , q 2 , q n , t , xi =

xi
t

xi

q
j =1

Now

or
, such that

Example: Moving pendulum with spring(no gravity)

, with
.

y = l cos , such that x = s + l cos and y = l sin .

2
2
1
1
1
2
2
2 2
2
Now T = m x + y = m s + l cos + l sin = m l + 2ls cos + s which is
2
2
2
defined as T2 .
Determine x = s + l sin and

((

) (

))

Generalized momenta
Furthermore the generalized momentum is defined as

pi =

T
.
qi

Example:

1 2
T
mx ; p =
= mx (linear momentum)
x
2
1 2 2
T
Pendulum: T = ml ; p =
= ml 2 2 (angular momentum)

T=

Generalized work
with

Defining

generalized and

virtual work gives the generalized force Q via

Example: Pendulum

r = l cos i + l sin j ; F = mg i

r = l sin i + l cos j
W = mgl sin , such that Q = mgl sin .
The generalized force can also be determined directly
conservative system

Qj =

V
.
q j

Example: Moving pendulum with spring(with gravity)

In handwritten notes!

IV. Lagrangian Dynamics

Fi =

d pi
dt

Because xi

only if

p i = mi ri

xi n xi
x
x
= xi (q, q ) f (t ) it can be derived that i = i , such that:

t j =1 qi
qk qk

Example: Rotating spring

or for a

Finally, it can be said that

d T T

= Qk with k all degrees of freedom. This gives a set of

dt qk qk

differential equations that describe the behaviour of the system, the equations of motion. This method
is faster than Newtonian mechanics.

Q can be found via

Qj =

V
for a conservative system and else via virtual work.
q j

Example: Rotating spring

In practical cases of conservative systems the potential V is only a function of position, such that

V
q

= 0
j

d T T
V

=
dt q j q j
q j
d T V T V

=0
dt q j q j q j q j

d
(
T V )
(T V ) = 0
q

dt q j
j

Defining the Lagrangian L=T-V this can be written as

d L L
L
=0
dt q j q j
And for a non-conservative system

d L L

= Q nc
j
dt q j q j

V.
-

VI. Ignorables and constraints

When

or

L = L(q ) L(q) the general coordinate q is ignorable and

L
d L
= 0 , which yields
=0
q
dt q j

L
=constant= Cq . The generalised momentum of q is the integral of motion, because
q j

L T V
T
=
=
= p.
q
q
q
Example: Satellite

The Rothian is defined as

R=

c q

i i
i = n m +1

L and is equivalent to the Lagrangian without ignorables,

R
L
R
L
R
R
=
and
=
, for ignorables:
= 0 and
= 0 . Now
q
q
q
q
q
q
d R R
the Lagrangian equation for non-ignorables becomes:

=0.
dt qk qk

because for non-ignorables:

In practice one can find the equations of motion by:

1. Setting up the Lagrangian
2. Determining the ignorable(s)
3.

Find integrals of motion:

4.

Now R

i = n m +1

5.

L
= Cqign qign
qign

Ci qi L .

Set up the Rothian equation:

d R R

=0
dt qk qk

Example: Satellite
1+2:
The same way as in the previous example the Lagrangian

L=

1
km
m r 2 r 2 2 +
, such that is the
2
r

ignorable coordinate.
3:

C
L
= mr 2 = C = 2
mr

4:

C
1
km C 2 1 2 2 C 2 km
R = C L = C 2 m r 2 + r 2 2
=
m r + r 2 4
mr
r
mr 2 2
mr r
2
C2 1
km
= 2 mr 2
2mr
2
r

5:

R
= mr
r
d R
= mr
dt r
2C 2 km
R
= 3 + 2
r
2mr
r
C 2 km
C 2
d R R
k

= mr + 3 2 = 0 or r + 2 3 2 = 0 .
dt r r
mr
r
mr r

Jacobi energy integral

mn

The Jacobi energy integral is defined as

h = R qk
k =1

R
, where the second term represents the
qk

non-ignorable coordinates.
Example: Satellite
With the Rothian as defined in the previous example, the Jacobi energy integral

h = R

C2 1
C2 3
R
km
km
mr 2 = 2 mr 2
r = 2 mr 2
.
r
2mr
2
r
2mr
2
r

Steady motion
A motion is steady (constant in time) when all

qk = 0 and p k =

d L

dt qk

R
= 0.
= 0 , such that
qk

Example: Satellite

r = 0
pr =

(Speed and acceleration both are zero)

d L
= mr = 0
dt r

R
=0
r
C
r = 0 means that r is constant and thus that = 2 =constant, more specific:
mr
2
mr 2
2C 2 km km C 2
k
R
=
+ 2 or 2 = 3 =
= mr 2 or = 3 .
3
3
r
2mr
r
r
mr
mr
r

Raleigh dissipation function

For a non-conservative system dissipative forces like Fx = cx x ,

Fy = c y y and Fz = cz z may play

a role. Those can be taken into account via the Raleigh dissipation function:

D=

1
D
cx x 2 + c y y 2 + cz z 2 ) . Now Q =
, such that via virtual work the Lagrangian equation for
(
2
q

k number of freedom becomes:

d L L D
+
= Qk* .

dt qk qk qk

Lagrange multipliers
For a system with n variables ( {q1 qn } ), m constraints (

f
R = f ( =
x

f
y

f j ( q1 qn ) = 0 ) introduce reaction forces

f
for the 3D-case) perpendicular to f . The Lagrangian multiplier is
z

an extra unknown and can be used to model contact between two flexible bodies.
Via virtual work the Lagrangian equation now becomes:

d L

dt qk

m
f
L

=
j j .

qk j =1 qk

Example: Pendulum (with gravity)

1
m r 2 r 2 2 + mgr cos (no ignorables!)
2
The constraint in this case r = l or f ( r ) = r l = 0 .
From earlier problems it is known that

L=

d L L
f
=

dt r r
r
d L L
f
2.
=

dt

3. f ( r ) = 0
1.

mr mr 2 mg cos =
2
2. 2mrr + mr + mgr sin = 0
3. r l = 0
1.

Equation 3 yields r=l, r =0 and

r =0, such that:

2
1. ml mg cos =

2.

g
ml 2 + mgl sin = 0 or + sin = 0 .
l

VII. Dynamic systems

A first order differential equation can describe a dynamical system like:

x = F ( x, t ) with x ( 0 ) = x0 .

x* is an equilibrium point when F ( x* , t ) = 0 . The stability of x* can be judged on by a phase

diagram.
By definition

x* is stable if and only if > 0 ( ) > 0 | x0 x* < x(t ) x* < or the

solution x(t) is always inside the area of the initial condition x0 around an equilibrium point
This represents the effect of a small deviation in initial condition.

x* .

Example: pendulum (with gravity)

g
sin = 0 is a second order differential equation, which can be converted to a first order d.e. via = .
l
g
T
Then = sin , such that x = [ ] . The system is in equilibrium when x = 0 or when = 0 and
l
= 0 , which is the fact for = k or when the pendulum is vertical.

Linear dynamic systems

When the function F is strictly linear the system can be described by

x = Ax with A constant. The

equilibrium point(s) can once more be found because of x = 0 .

This yields a solution

x ( t ) = ki ci ei t with Ac = c . The solution is stable when:

i R and i < 0 ,

i C and Re ( i ) 0 ,

or combinations of those.

Example: Trolley with horizontal spring

k
k
x = 0 , which can be converted to a 1st order d.e. with x = y and y = x .
m
m
0 1
x
x
y = k 0 y .


m

x
In an equilibrium point =0, which yields x*=0.???
y

x+

i
x
k
k
k
and = c1e
= + = 0 or = = i
m
m
m

y
2

k
m

+ c2 e

k
m

, this solution is stable

because both Eigen values are complex with a zero real part.
Linearization
Non-linear dynamic systems can be linearized about equilibrium points by a Taylor expansion:

F ( x ) = F ( x 0 ) + F ( x 0 )( x x 0 )

Example: pendulum (with gravity)

=
g
l

= sin
This systems equilibrium points are 0,0 (I) and

,0 (II).

Linearization about point I gives:

F ( x0 ) = 0

F1
x
1
F ( x0 ) =
F2
x
1
0

= g

l

F1
0

x2

= g
cos
F2
l
x2 x = x

1
0
0 0

1
0

= g

0
x = x0 l

= 2 +

g
g
, such that is complex with Re ( ) = 0 , thus equilibrium point
= 0 = i
l
l

I is stable.
Linearization about equilibrium point II gives:

= g
cos
l

g
l

= 2

0
x = x0

0

0 = g

1

0

g
g
, such that is real with one of >0, thus the equilibrium point II is
=0 =
l
l

unstable.
In general the stability of any system can be checked by:
1. Setting up the 2nd order differential equation (equation of motion) by one of the methods from
previous chapters (in general via the Lagrangian equation)
2. Deriving the 1st order system of differential equations
3. Finding the equilibrium points
4. Linearizing about those points
5. Analysing the Eigen values.
Example: vertical rod with springand force acting on it
Example: satellite
Elaborations of this example can be found in handwritten notes.

VIII. Stability of conservative systems

Sometimes it is hard to set up the Lagrangian equation (kinetic and/or potential energy). Because in a
conservative system the applied forces

V
q

Q=

V
= 0 ,
and in an equilibrium position q = 0 and q
q

=0.
q = q*

Around a maximum of V(q), Q is opposite to the disturbance in q and around a minimum Q is in the
same direction. This means that:

2V
When
q 2

q = q*

2V
When
q 2

q = q*

> 0 , V(q*) is a minimum and therefore stable.

< 0 , V(q*) is a maximum and therefore unstable.

Example: Pendulum without gravity

V
2V
= mgl sin = 0 = k ; 2 = mgl cos , which is
q
q
> 0 and thus stable for = , 3 , 5 ,
< 0 and thus unstable for = 0, 2 , 4 ,

V = mgl cos ;

Example: Bar in triangle

Stability of Lagrangian systems

When all applied forces can be expressed in terms of the generalised potential V,the same procedure
can be used as for conservative systems.
Because the Jacobian energy integral becomes h = T2 T0 + V =constant and for equilibrium all q = 0

T2 =0) T0 + V = h =constant.
The effective potential is defined as Veff = V T0 and the equilibrium condition becomes:
(such that

Veff
q

= 0 . Again for a minimum of Veff the system is stable and for a minimum of Veff it is unstable.

Examples

IX. Dynamics of rotating bodies: Gemist

X. Lagrangian dynamics
The equations of motion of a solid body are based on its angular velocity, therefore rotational
transformations are needed. Because this transformation of coordinate systems depends on sequence,
the rotation angles about coordinate axes are no suitable general coordinates.
Euler angles

about the Z-axis,

about the X axis and about the z axis. There angular velocities are respectively = K ,

The transformation from X,Y,Z to x,y,z can be performed via the Euler angles

= ex '

and

= k , with =precession, =nutation and =spin. These vectors do obviously

not form an orthogonal basis:

= + + .

The actual transformation can be done via matrices or by projection. The latter case yields:

x = cos + sin cos

y = sin + sin sin
z = + cos
1
Now T = ( x
2

Ix
z ) 0
0

0
Iy
0

0 x

0 y
I z z

Rotation symmetric bodies

For a body that is rotation symmetric about the z-axis Iz=Is (axial moment of inertia) and Ix=Iy=I
(transverse moment of inertia), such that
that and

T=

((

1
2
I 2 sin 2 + 2 + I S ( cos + ) , such
2

are ignorable coordinates.

d T T
T

= 0 for non ignorable coordinates and

= Cq for ignorable coordinates, this
dt q q
q
results in three expressions. For steady motion = = 0 , which yields sin =0 or
I S
=
, the relation between precession and spin for steady motion of an rotation
( I S I ) cos
Because

symmetric body.
When

I S < I , I I S > 0 , then: > 0 > 0 evv. Rotation about the Z-axis (spin) causes motion of

the Z-axis itself (precession) in the same direction, this effect is called direct precession.
When I S > I , I I S < 0 , then: > 0 < 0 evv. Rotation about the Z-axis (spin) causes motion of
the Z-axis itself (precession) in the opposite direction, this effect is called retrograde precession.

Rotation non-symmetric bodies

M x = I x x ( I y I z ) y z
In general for an arbitrary rotating body:

M y = I y y ( I z I x ) z x .
M z = I y y ( I x I y ) x y

x =
Without applying any forces ( M x

= M y = M z = 0 ), this yields

I y Iz
Ix

y z

I I
y = z x z x
Iy

I y Iz

0
z

Ix
x
x , with its
This can be rewritten as the linear dynamic system: =
y
y I z I x
0

z
I y

Iy Iz

z
Ix
( I y I z ) ( I z I x ) 2 . For stability should be
eigenvalues via
= 2
z
Iz Ix
IxI y
z

Iy
purely imaginary, such that

(I

I z ) ( I z I x ) < 0 : I z > I x , I y or I z < I x , I y , the rotation axis should

have the largest or smallest moment of inertia.

XI. Functionals
A functional makes a real value from a function, in contrary to a function that makes a real value of a
real value. Finding the function (extremal) that maximizes or minimizes a functional is called calculus
of variations.

Example: The length of an arbitrary path between position a and b

sab =

xb

1 + y '( x) 2 dx , which is a

xa

functional. Minimalizing this one finds the shortest path y(x).

The travel time between position a and b of a particle influenced by gravity can be described by dt

v = 2 gh = 2 gy ( x) , such that tab =

xb

xa

For a function f(x) the extremum

ds
,
v

1 + y '( x)2
dx .
2 gy ( x)

x* can be found by f '( x* ) = 0 . Then for a minimum

f ( x* + dx) > f ( x) , for every dx.

y* ( x) can be determined by variation of y(x). y ( x) = y* ( x) + ( x) , with the
arbitrary function ( x ) . For now ( xa ) = ( xb ) = 0 .
This way the extremal

Defining the functional

Now

I ( y ) , y* is a minimum of I when I ( y* + ) I ( y* ) for any ( x) .

I ( y * + ) = I ( ) for a given ( x) . The extremal can be found for

d
I ( ) = 0 .
d

Euler-Lagrange equation
Because x is integrated, y is the anwer and
*

( x)

xb

is fixed.

I ( y ) = F ( x, y, y ' )dx can be written as

xa

xb

I ( y * + ) = F ( x, y* + , y '* + ' )dx = I ( ) .

xa

dI
d

chain rule

b
F y F y '
F
F
x y + y ' dx = x y + y ' ' dx
a
a

xb

xb

b
F
F
d F
dx
+

dx
y
x y

dx

y
'

x
x
a
a
a

partial integration xb

b
F d F
F
F
=

( xb ) ( xa )
dx +

y dx y '
y xb
y xa
xa

b
F d F
=

dx = 0 for all .

y
dx

y
'

xa
F d F
The fundamental lemma now yields

= 0 , which is called the Euler-Lagrange equation.

y dx y '

Example: sab, tab, min resistance

XII. Variational operator

y ( x) = y* ( x) + ( x) = y* ( x) + y* ( x) , where is the variational operator.

This yields I ( y + y ) = I ( y ) + I , such that y is an extremal when I = 0 .
Now

I is the difference in shape of function y and thus the variation of a function, comparable to the
differential dx, which is the variation of a x-value.
xb

chain rule

I = F ( x, y, y ')dx =
xa

xb

y y + y ' y ' dx

xa

xb

b
F
F
d F

ydx
+

ydx
y
x y

y
dx

y
'
x
x
a
a
a

partial integration xb

b
=0
=0
F d F
F
F
=

y ( xa ) = 0
y ( xb )
ydx +
y ' xa
y dx y '
y ' xb
xa
F d F
for every y , such that

=0.
y dx y '
y ( xa ) = 0 means that y ( xa ) has a fixed value ya , which is called an essential boundary condition.

Non-fixed boundaries lead to natural boundary conditions:

y ( xa ) = ya
y ( xb ) = ?

y ( xa ) = 0
y ( xb ) 0

=0
F d F
F
F

y ( xb )
y ( xa ) = 0 for every y ( xb ) yields the
ydx +
y dx y '
y ' xb
y ' xa
xa
xb

I = ... =

natural boundary condition

F
= 0.
y ' xb

Example: Shortest distance

sab =

xb

1 + y '2 dx

xa

F
y'
=
y '
1 + y '2
y '( xb )
F
=
y '( xb ) = 0
y ' xb
1 + y '( xb ) 2
In part XI it is shown that Euler-Lagrange equation yields a straight line and the natural boundary condition
now yields the slope being 0.
Generalisation
It can be shown that in general for

F ( x, y, y ', , y ( n ) ) the Euler-Lagrange equation

i
F n
F
i d
becomes
+ ( 1) i (i ) = 0 .
y i =1
dx y

The function y can also depend on more variables. For example y

that I ( y ) =

tb xb

F ( x, t , y, y , y )dxdt , with y
x

ta xa

= y ( x, t ) , such

y
y
and yt =
.
x
t

Now

F
F
y+
yx +
y x dxdt
y
yx
y x

ta xa
tb xb

I =

xb

b
b b
F
F
F
+

ydxdt
y
dt

t x y
t yx
t x x yx ydxdt
x
a a
a
a a

partial integration tb xb

t x

tb

F
F
y dx
ydxdt
yt
t yt
xa
ta xa
t
xb

tb xb

xb

tb

b
b
F F F
F
F
=

y dt +
y dx = 0
ydxdt +
y x yx t yt
y x
yt
ta xa
ta
xa
xa
ta
F F F

= 0 and boundary conditions.

for every y , yields
y x y x t yt

tb xb

Example:

F ( x, t , ut , u x , u xx ) can be found in sample problems on Blackboard.

XIII. Hamiltons principle

r = r (t ) , the force working along this path is F ( r ) = mr

and W = F r = mr r . (Note: vector notation is dropped)
d
1
1
Use
( mr r ) = mr r + mr r = mr r + m ( r r ) = W + m ( v 2 ) = W + T
dt
2
2
For a path in space

or mr r t =
tb

tb

W + T .

ta

Defining the essential boundary conditions

For conservative forces Q =
tb

(r ) = 0
a
(rb ) = 0
r (tb ) = rb

r (ta ) = ra

tb

yields

W + T = 0 .

ta

V
V
and W = Q q =
q = V , such that
q
q

tb

tb

ta

ta

( T V ) dt = (T V ) dt = Ldt = 0

ta

Now for I ( r ) =

tb

Ldt , which is defined as action, I = 0 gives the minimum for the motion of a

ta

Lagrangian system. This is called Hamiltons principle.

Example: beam with distributed properties

tb

For dissipative forces

I = L + Qinp qi dt = 0 , where Qinp qi = W np is virtual work.

ta

In statics the system is steady in time and T = 0 , such that

stationary potential energy.

V = 0 , which is called the principle of

Example: Aero-elasticity/wing

XIV. Ritz method

Earlier solutions of the Euler-Lagrange equations and natural boundary conditions were mainly
differential equations; therefore an approximation can be used, called Ritz-method.
Assume an approximate solution as a linear combination of a finite number of linearly independent
n

functions:

y ( x) = ai hi ( x) . The coefficients ai are the degrees of freedom and the functions hi ( x) the
i =1

coordinate- or shape functions. Translate the boundary conditions to

a h (x ) = y
i =1

i i

and

a h (x ) = y
i =1

i i

Example:
1

Try to find the solution of

I ( y ) = y 2 + y '2 dx , with y (0) = y (1) = 1

0

h1 ( x) = 1; h2 ( x) = x; h3 ( x) = x

y = a1h1 + a2 h2 + a3 h3 = + x + x 2
1

I ( y ) = y 2 + y '2 dx = ( + x + x 2 ) + ( + 2 x ) dx = ( , , )
0

The functional now became a function and has a minimum when all partial derivatives are zero:

= 0;
= 0;
= 0.

The boundary conditions are

y (0) = + 0 + 0 = 1 = 1 and y (1) = + + = 1 = .

Now

2
1
11
2
I ( y ) = (1 + x x 2 ) + ( 2 x ) dx = 1 + + 2 = ( )
3
30
0

1 22
5
= + =0 =
, such that the approximate solution becomes
3 30
11
5
5
y = + x + x2 = 1 x + x2
11
11
sinh( x) + sinh(1 x)
This is only 0.05% of w.r.t. the real solution y ( x) =
.
sinh(1)

Ritz method is commonly used in computerized computational mechanics like finite element method,
using non-continuous shape functions are chosen.
The choice of shape functions is rather arbitrarily, as long as they are linear independent:

{ v1 , , vn } is linear independent a1v1 + + an vn = 0 a1 = = an = 0

For computational reasons polynomials should be chosen.
Example: Dynamics of a beam