Vous êtes sur la page 1sur 30

Calculus of Variations

Functional Eulers equation


V-1 Maximum and Minimum of Functions
V-2 Maximum and Minimum of Functionals
V-3 The Variational Notaion
V-4 Constraints and Lagrange Multiplier

Eulers equation Functional
V-5 Approximate method
1. Method of Weighted Residuals
Raleigh-Ritz Method
2. Variational Method
Kantorovich Method
Galerkin method

2
V-1 Maximum and Minimum of Functions
Maximum and Minimum of functions

(a) If f(x) is twice continuously differentiable on [x
0
, x
1
] i.e.

Nec. Condition for a max. (min.) of f(x) at is that

Suff. Condition for a max (min.) of f(x) at are that
also ( )

(b) If f(x) over closed domain D. Then nec. and suff. Condition for a max. (min.)

i = 1,2n and also

that is a negative infinite .

0 1
[ , ] x x x e
( )
0 F x ' =
0 1
[ , ] x x x e
( )
0 F x '' s
( )
0 F x ' =
( )
0 F x '' >
D D e c
0
0
x x
i
f
x
=
c
=
c
of f(x) at x
0
are that
0
2
x x
i j
f
x x
=
c
c c
Part A Functional Eulers equation
3
(c) If f(x) on closed domain D
If we want to extremize f(x) subject to the constraints
i=1,2,k ( k < n )


Ex :Find the extrema of f(x,y) subject to g(x,y) =0

(i) 1st method : by direct diff. of g








1
( , , ) 0
i n
g x x = K
0
x y
dg g dx g dy = + =

x
y
g
dy dx
g
=
0
x y
df f dx f dy = + =

( ) 0
x
x y
y
g
f f dx
g
=
To extremize f
4
We have
and

to find (x
0
,y
0
) which is to extremize f subject to g = 0

0
x y y x
f g f g =
0 g =
(ii) 2nd method : (Lagrange Multiplier)


extrema of v without any constraint
extrema of f subject to g = 0

To extremize v


( , , ) ( , ) ( , ) v x y f x y g x y = +

0
x x
v
f g
x

c
= + =
c
0
y y
v
f g
y

c
= + =
c

0
x y y x
f g f g =
0
v
g

c
= =
c
Let
We obtain the same equations to extrimizing. Where is called
The Lagrange Multiplier.

5
V-2 Maximum and Minimum of Functionals
2.1 What are functionals.


Functional is functions function.
2.2 The simplest problem in calculus of variations.
Determine such that the functional



where over its entire domain , subject to y(x
0
) = y
0
, y(x
1
) = y
1
at the end
points.

( ) ( ) ( ) ( ) ( )
1
0
, ,
x
x
I y x F x y x y x dx =
}
&
2
0 1
( ) [ , ] y x c x x e
2
F c e
as an extrema
1
y
2
y
( ) y x
1
x
2
x
( ) x q
( ) ( ) y x x q +
x
y
6
On integrating by parts of the 2nd term

------- (1)


since and since is arbitrary.

-------- (2) Eulers Equation





Natural B.Cs



or /and

The above requirements are called national b.cs.

0 )] , , ( ) , , ( [ ] ) , , ( [
1
0
1
0
=
'

'

'
}
' '
dx y y x F y y x F
dx
d
y y x F
x
x
y y
x
x y
q q
0 1
( ) ( ) 0 x x q q = = ( ) x q

( , , ) ( , , ) 0
y y
d
F x y y F x y y
dx
'
' ' ( =

0
0
x
F
y
( c
=
(
'
c

1
0
x x
F
y
=
( c
=
(
'
c

7
Variations

Imbed u(x) in a parameter family of function the
variation of u(x) is defined as


The corresponding variation of F , F to the order in is ,

since



and

( , ) ( ) ( ) x u x x | c cq = +
( ) u x o cq =
1
0
( ) ( , , ) ( )
x
x
I u F x u u dx G cq cq cq c
' '
+ = + + =
}
( , ) ( , , ) F F x y y F x y y o cq cq ' ' ' = + + +
F F
y y
y y
o o
c c
' = +
' c c
1
0
( , , )
x
x
I F x y y dx o o ' =
}
1
0
( , , )
x
x
F x y y dx o ' =
}
Then
1
0
x
x
F F
y y dx
y y
o o
| | c c
' = +
|
' c c
\
}
V-3 The Variational Notation
8
Thus a stationary function for a functional is one for which the first variation = 0.

For the more general cases




,

1
1
0
0
x
x
x
x
F d F F
ydx y
y dx y y
o o
(
| | ( c c c
= +
( | (
' ' c c c
\

}
1
0
( , , z ; , z )
x
x
I F x y y dx
' '
=
}
Ex :
Ex : ( , , , , )
x y
R
I F x y u u u dxdy =
}}

Eulers Eq.

Eulers Eq.
F
y
c
c
( ) 0
F d F
y dx y
c c
=
' c c
( ) 0
F d F
z dx z
c c
=
' c c
(b) Several Independent variables.
(a) Several dependent variables.
( ) ( ) 0
x y
F F F
u x u y u
c c c c c
=
c c c c c
9
Variables Causing more equation.
Order Causing longer equation.
Ex :
1
0
( , , , )
x
x
I F x y y y dx ' '' =
}

Eulers Eq.
2
2
( ) ( ) 0
F d F d F
y dx y dx y
c c c
+ =
' '' c c c
(C) High Order.
10
V-4 Constraints and Lagrange Multiplier
Lagrange multiplier can be used to find the extreme value of a multivariate
function f subjected to the constraints.




Ex :
(a) Find the extreme value of





------------(1)

From -----------(2)
1
0
( , , , , )
x
x x
x
I F x u v u v dx =
}
1 1
( ) u x u =
2 2
( ) u x u =
1 1
( ) v x v =
2 2
( ) v x v =
where
and subject to the constraints
( , , ) 0 G x u v =
2
1
0
x
x
x x
F d F F d F
I u v dx
v dx u v dx v
o o

( (
| | | | c c c c
= + c =
` ( ( | |
c c c c
\ \
)
}
Lagrange multiplier
11
Because of the constraints, we dont get two Eulers equations.

From













The above equations together with (1) are to be solved for u , v.
0
G G
G u v
u v
o o o
c c
= + =
c c

v
u
G
v u
G
o o =

So (2)
1
0
0
x
v
x
u x x
G F d F F d F
I vdx
G u dx u v dx v
o o

( (
| | | |
c c c c
= + =
` ( ( | |
c c c c
\ \

)
}

0
x x
G F d F G F d F
v u dx u u v dx v
( (
| | | |
c c c c c c
=
( ( | |
c c c c c c
\ \

12
(b) Simple Isoparametric Problem

To extremize , subject to the constraint

(1)

(2) y (x
1
) = y
1
, y (x
2
) = y
2


Take the variation of two-parameter family
(where and are some equations which satisfy
)
( )
2
1
, ,
x
x
I F x y y dx
'
=
}
( )
2
1
, , .
x
x
J G x y y dx const ' = =
}
( ) ( )
1 1 2 2
y y y x x o c q c q + = + +
( )
1
x q
( )
2
x q
( ) ( ) ( ) ( )
1 1 2 1 1 2 2 2
0 x x x x q q q q = = = =
Then ,



To base on Lagrange Multiplier Method we can get :
2
1
1 2 1 1 2 2 1 1 2 2
( , ) ( , , )
x
x
I F x y y dx c c c q c q c q c q
' '
'
= + + + +
}
2
1
1 2 1 1 2 2 1 1 2 2
( , ) ( , , )
x
x
J G x y y dx c c c q c q c q c q
' '
'
= + + + +
}
13
( )
1 2
0
1
0 I J
c c

c
= =
c
+ =
c
( )
1 2
0
2
0 I J
c c

c
= =
c
+ =
c

2
1
0
x
i
x
F d F G d G
dx
y dx y y dx y
q

( (
| | | | c c c c
+ =
` ( ( | |
' ' c c c c
\ \

)
}
i = 1,2

So the Euler equation is




when , is arbitrary numbers.

The constraint is trivial, we can ignore .
( ) ( )
0
d
F G F G
y dx y

( c c
+ + =
(
' c c

0
G d G
y dx y
| | c c
=
|
'
c c
\


14
( , , ) ( , )sin( ) f x y t P x y t e o = +
( , )sin( ) u v x y t e o = +

2 2
2 2
2 2
( ) 0
v v
c v p
x y
e
c c
+ + + =
c c
Eulers equation Functional
Examples
we may write the steady state disp u in the form
if the forcing function f is of the form
(1)
2 2 2
2
2 2 2
( ) ( , , )
u u u
c f x y t
t x y
c c c
+ =
c c c

Ex : Force vibration of a membrane.

-----(1)

Helmholtz Equation
15
2 2
2 2
2 2
( ) 0
R
v v
c v p vdxdy
x y
e o
(
c c
+ + + =
(
c c

}
-----(2)
2
xx
R
c v vdxdy o
}
2
[( ) ]
x x x x
R
c v v v v dxdy o o =
}
Consider
2
yy
R
c v vdxdy o
}
2
y y y y
[( ) ]
R
c v v v v dxdy o o =
}
( )
x x xx x x
v v v v v v o o o = + ( )
y y yy y y
v v v v v v o o o = +
cos sin n i j u u = +
%
x y
V V i V j = +
x x
V v v o =
y y
V v v o =
y
x
( v)
( v)
=
y
x
V
V
V
x y x y
v o
v o
c c
c c
V = + +
c c c c
g
Note that
16
2 2
xx yy
R R
c v vdxdy c v vdxdy o o +
} }
2 2 2
1
sin ( )
2
y y
R
c v v ds c v dxdy

o u o +
} }
(2)

2 2 2 2
1
( cos sin ) [( ) ( ) ]
2
x y x y
R
c v v vds c v v dxdy

u u o o + +
} }
2
2
1
( ) 0
2
R R
v dxdy P vdxdy e o o + + =
} }

2 2 2 2 2
1 1
[ ( ) ] 0
2 2
R
v
c vds c v v Pv dxdy
n

o o e
c
V =
c
} }
( ) V da V nds
9
V =
} }
g g

( cos sin )ds


x y
v v v v o u o u = +
}
2 2 2
1
cos ( )
2
x x
R
c v v ds c v dxdy

o u o =
} }
17
Hence :
(i) if is given on
i.e. on
then the variational problem

-----(3)


(ii) if is given on
the variation problem is same as (3)

(iii) if is given on




( , ) v f x y =
0 v o =

2
2 2 2
1
[ ( ) ] 0
2 2
R
c
v v pv dxdy o e V =
}
0
v
n
c
=
c
( )
v
s
n

c
=
c

2 2 2 2 2
1 1
[ ( ) ] 0
2 2
R
c v v pv dxdy c vdx
t
o e

V =
`
)
} }

18
Ex : Steady state Heat condition

in D
( ) ( , ) k T f x T V V =
B.Cs :
on B
1

on B
2
on B
3

multiply the equation by , and integrate over the domain D. After integrating
by parts, we find the variational problem as follow.





with T = T
1
on B
1

1
T T =
2
kn T q V =
3
( ) kn T h T T V =
T o
0
2
1
[ ( ) ( , )
2
T
D T
k T f x T dT d o t

' ' V +
`
)
} }
2 3
2
2 3
1
( ) ] 0
2
B B
q Td h T T d o o + + =
} }
Diffusion Equation
19
2
2 V =
0 = on
in R
Ex : Torsion of a primatri Bar
where is the Prandtl stress function and

,

The variation problem becomes



with on

z
G
y
t o

o
c
=
c
zy
x
o

o o
c
=
c
{ }
2
[( ) 4 ] 0
R
D dxdy o =
}
0 =
Poison Equation

20
V-5-1 Approximate Methods

(I) Method of Weighted Residuals (MWR)

in D
+homo. b.cs in B
Assume approx. solution



where each trial function satisfies the b.cs
The residual


In this
method (MWR), C
i
are chosen such that Rn is forced to be zero in an average
sense.

i.e. < w
j
, R
n
> = 0, j = 1,2,,n
where w
j
are the weighting functions..

[ ] 0 L u =
1
n
n i i
i
u u C|
=
= =

i
|
[ ]
n n
R L u =
21
(II) Galerkin Method
w
j
are chosen to be the trial functions hence the trial functions is chosen
as members of a complete set of functions.

Galerkin method force the residual to be zero w.r.t. an orthogonal complete set.

j
|
Ex: Torsion of a square shaft




2
2 V =
0 , on x a y a = = =
(i) one term approximation
2 2 2 2
1 1
( )( ) c x a y a =
2 2 2
1 1
2 2 [( ) ( ) ] 2
i
R c x a y a = V + = + +
2 2 2 2
1
( )( ) x a y a | =
22
From
1 1
0
a a
a a
R dxdy |

=
} }

1
2
5 1
8
c
a
=
therefore
2 2 2 2
1
2
5
( )( )
8
x a y a
a
=
the torsional rigidity
4
1
2 0.1388 (2 )
R
D G dxdy G a = =
}
the exact value of D is
4
0.1406 (2 )
a
D G a =
the error is only -1.2%
23
2 2 2 2 2 2
2 1 2
( )( )[ ( )] x a y a c c x y = + +
By symmetry even functions

2 2
2 R = V +
2 2 2 2
1
( )( ) x a y a | =
2 2 2 2 2 2
2
( )( )( ) x a y a x y | = +
From 2 1
0
R
R dxdy | =
}
and 2 2
0
R
R dxdy | =
}
we obtain
1 2
2 2
1295 1 525 1
,
2216 4432
c c
a a
= =
therefore
4
2 2
2 0.1404 (2 )
R
D G dxdy G a = =
}
the error is only -0.14%
(ii) two term approximation
24
(I) Kantorovich Method
1
( )
n
i n i
i
u C x U
=
=
Assuming the approximate solution as :
where U
i
is a known function decided by b.c. condition.
Ex : The torsional problem with a functional I.
2 2
( ) [( ) ( ) 4 ]
a b
a b
u u
I u u dxdy
x y

c c
= +
c c
} }
V-5-2 Variational Methods
C
i
is a unknown function decided by minimal I.

Euler Equation of C
i
25
Assuming the one-term approximate solution as :
2 2
( , ) ( ) ( ) u x y b y C x =
Then,
2 2 2 2 2 2 2 2
(C) {( ) [C( )] 4 C ( ) 4( )C( )}
a b
a b
I b y x y x b y x dxdy

'
= +
} }
Integrate by y
5 2 3 2 3
16 8 16
(C) [ C C C]
15 3 3
a
a
I b b b dx

'
= +
}
Eulers equation is
2 2
5 5
C ( ) C( )
2 2
x x
b b
''
=
where b.c. condition is
C( ) 0 a =
General solution is
1 2
5 5
C( ) A cosh( ) A sinh( ) 1
2 2
x x
x
b b
= + +
26
1 2
1
A , A 0
5
cosh( )
2
a
b
= =
where
and
5
cosh( )
2
C( ) 1
5
cosh( )
2
x
b
x
a
b



=
`


)
So, the one-term approximate solution is
2 2
5
cosh( )
2
u 1 ( )
5
cosh( )
2
x
b
b y
a
b



=
`


)
27
(II) Raleigh-Ritz Method

This is used when the exact solution is impossible or difficult to obtain.

First, we assume the approximate solution as :

Where, U
i
are some approximate function which satisfy the b.cs.
Then, we can calculate extreme I .

choose c
1
~ c
n
i.e.

1
n
i i
i
u CU
=
=

1
( , , )
n
I I c c = K
1
0
n
I I
c c
c c
= = =
c c
K
y xy x
''
+ =
(0) (1) 0 y y = =
( )
1
0
0 y xy x ydx o
''
+ + =
}
( )
1
2
2
0
1 1
2 2
I y xy xy dx
(
' =
(

}


Ex: ,
Sol :
From
28
Assuming that



(1)One-term approx










(2)Two-term approx
( )
( )
2
1 2 3
1 y x x c c x c x = + + K
( ) ( )
2
1 1
1 y c x x c x x = =
( )
1
1 2 y c x
'
=
( ) ( ) ( ) ( )
1
2 2 2 2 3 4 2
1 1 1 1
0
1
1 4 4 2
2 2
x
I c c x x c x x x c x x x dx
(
= + +
(

}
2 2
1 1
1
4 1 2 1 1 1
1 2
2 3 2 4 5 6 3 4
c c
c
| | | | | |
= + +
| | |
\ \ \
2
1
1
19
120 12
c
c =

1
0
I
c
c
=
c
1
19 1
0
60 12
c =

1 2
(1 )( ) y x x c c x = +
2 2 3
1 2
( ) ( ) c x x c x x = +

(1) 0.263 (1 ) y x x =
1
0.263 c =
Then,
29
Then
2
1 2
(1 2 ) (2 3 ) y c x c x x
'
= +
( )
{
( )
1
2 2 2 3
1 2 1 1 2
0
1
( , ) [ 1 4 4 2 2 7 6
2
I c c c x x c c x x x = + + +
}
}
{ ( ) ( )
2 2 3 4 2 3 4 5 4 5 6
3 1 1 2
1
(4 12 9 ) 2 2 2
2
c x x x c x x x c c x x x + + + + +
}
( ) ( )
{ }
2 5 6 7 2 3 3 4
2 1 2
( 2 ) ] c x x x c x x c x x dx + + +
2
1
1 2
4 1 2 1 7 3 1 1 1
1 2 1
2 3 4 5 6 3 2 5 3 7
c
c c
| | | |
= + + + + +
| |
\ \
2
1 1 2
4 9 1 2 9
3
2 3 5 6 7 8 12 20
c c c
| |
+ +
|
\
2 2
1 2
1 1 2 2
19 11 107
120 70 1680 12 20
c c
c c c c = + +
30
0.317 c
1
+ 0.127 c
2
= 0.05 c
1
= 0.177 , c
2
= 0.173
1
0
I
c
c
=
c

1 2
19 11 1
60 70 12
c c + =

2
(2) (0.177 0.173 )(1 ) y x x x =
( It is noted that the deviation between the successive approxs. y(1) and y(2) is
found to be smaller in magnitude than 0.009 over (0,1) )






2
0
I
c
c
=
c

1 2
11 109 1
70 840 20
c c + =