EULER-LAGRANGE EQUATION

CONDITION FOR A TRIAL SOLUTION CAUSING A FUNCTIONAL

STATIONARY (EULER-LAGRANGE EQUATION)
As per Principle of Stationary Total Potential (PSTP); for conservative systems (with
no dissipative forces) among all admissible displacement fields, the equilibrium
configuration of the system is that, which makes the total potential of the system
stationary with respect to small admissible variations of displacements.
Let us consider a Functional (i.e., function of functions):
b b
du
I (u )   F (u, , x)dx   F (u, u, x)dx
a
dx a
As per PSTP, let us find u(x) subjected to the end conditions: u(a) = ua and u(b) = ub.
which makes the functional stationary.
I. Variation and differentiation are permutable
Let us consider the possible set of u (x) as trial
solutions for the unknown exact solution u(x) as
shown in the side figure.
It may be clearly observed that all u (x) satisfy
the prescribed end boundary conditions.
If u(x) is slightly changed to u (x)
u ( x)  u ( x)  v( x) Where; ε is a small parameter and v(a) = v(b) = 0.
The difference between u(x) and u (x) is termed as the “variation” in u(x) and is
been denoted as δu(x).   (u ( x))  u ( x)  u ( x)  v( x)
 It is important to understand the difference between the terms “du” and “δu”.
At a given position x, variation (δu) refers to the difference between u(x) and u (x)
while “du” refers to the differential change in u(x) as x changes to x + dx.
 (u( x))  {u ( x)  u( x)}   (u( x))  u ( x)  u( x)
  (u( x))  u( x)  v( x)  u( x)  v( x)
  (u( x))  ( (u ( x)))   (u)  (u )
du d
( )  (u )
dx dx
This indicates that, the process of variation
and differentiation are permutable.

II. Expression for variation of Functional (ΔF, ΔI)

For a given x, as we move from u(x) to u (x) , the given functional also undergoes a
change such that:
F  F{u ( x), u ( x), x}  F{u ( x), u( x), x}  F  F (u  u, u  u, x)  F (u, u, x)
Expanding the first term in Taylor’s series, we have:
F F 1 2F 2 2F 2F 2
F (u  u, u  u, x)  F (u, u, x)  u  u  [ 2 u  2 uu  2 u ]  ....
u u 2 u uu u
F F 1 2F 2 2F 2F 2
F  ( u  u)  [ 2 u  2 uu  2 u ]  ....
u u 2 u uu u
 The first variation of “F” is defined as: F  ( F u  F u)
u u
The second variation of “F” is defined as: 2 1  F 2
2
2F 2F
 F  [ 2 u  2 uu  2 u2 ]
2 u uu u
F F 1 2F 2 2F 2F 2
F  ( u  u)  [ 2 u  2 uu  2 u ]  .... F  F  1  2 F  ....
u u 2 u uu u 2
Considering the change in the Functional as for a given x, we move from u(x) to
b b
u (x) du du
I  I (u , u , x)  I (u, u, x)   F (u , , x)dx   F (u, , x)dx
a
dx a
dx
b b b
1
  {F (u , u , x)  F (u, u, x)}dx   F dx   (F   2 F  ....)dx
a a a b 2
The first variation of “I” is defined as: I   F dx b

The second variation of “I” is defined as: a  2 I    2 F dx

1 2 a
There fore: I  I   I  ....
2
III. Limiting value of variation of the functional (ΔI)
If “I” represents the total potential of a structure and we look for a stable
equilibrium configuration then we wish to find u(x) that minimizes I.
Since, u(x) minimizes “I”; for a given x as u(x) changes to u (x), the F(u,u՛,x)
changes to F (u , u , x) and accordingly I(u,u՛,x) changes to I (u , u , x)
 I  0
 For a given x as u(x) changes to u (x):
  (u ( x))  u ( x)  u( x)  v( x)
As “ε” is reduced, δu→0; ΔF (a function of δu) approaches zero; finally ΔI (a function
of δF) also approaches zero.
For a given x as u (x) approaches to u(x) :  u ( x)  u ( x,) I attains a minimum
value and ΔI = 0. 1
 I  I   2 I  ....  0
2
For sufficiently small values of “ε”, δ2I (a function of “ε2” ) and higher order
vibrational terms become negligibly small as compared to δI (a function of “ε”),
hence can be neglected in the expression of ΔI.
Thus the condition for “I” to be stationary:  I  I  0
We propose to show that the vibrational form given by: I  I  0 reduces to the
Governing Differential Equation of the problem.
F F
b b


I  F dx  0  I   ( u  u)dx  0
a a
u u
du d d
We have already proved that:  ( )  (u )  u  (u )  (u )
dx dx dx
F F
b b

Considering the second term: a u u dx  a u d (u) dx

F
b
F d F
b b
Considering integration by parts:
a u d (u ) dx  [ u] 
 (u )  ( )  dx
u
u  a
a
dx
Putting the above equation in the expression of “δI”, we have:
F F
b
 I   ( u  u)dx  0
a
u u

b
b
F F b
d F
 I    u  dx  [ u ]   (u )  ( )  dx  0
dx u
a u u a
a


b
b
F d F F
 I   [  ( )]  u  dx  [
dx u
u ] 0
a u u a

Since all the trial functions: u(x) satisfy the end conditions at x = a and x = b; we
have: u (a)  u (b)  0
b
F d F
 I   [  ( )]  u  dx  0
a u dx u
For arbitrary δu, we therefore have the equation:
F d F
 ( )0 Euler_lagrange Equation
u dx u
EULER-LAGRANGE EQUATION q(x) = q0
Potential energy of an elemental
portion of the bar is given by :
1 du 2 L
d p  AE ( )  (q)(u ) dx
2 dx x=0

σx+(dσx/dx)dx
So the potential energy of this q(x) = q0
elemental bar portion is actually a
functional, as it is a function of σx
functions:
1 du 2

I  F (u, u , x)  { AE ( )  (q)(u )}
2 dx dx
Let u(x) be the displacement function for which the structure will achieve minimum
potential energy, i.e., the functional will be minimum.
If u(x) be the displacement function causing minimum functional; then as per, Euler-
Lagrange equation: F d F
  ( )0
u dx u
AEu2  qu  F  q, F  AEu
In the present example F 
1
2 u u
F d F d 2u d 2u
  ( )  0  q  AE 2  0  q  AE 2  0 GDE
u dx u dx dx
We already have:

b
F d F
b
F
 I   [  ( )]  u  dx  [
dx u
u ] 0
a u u a

1 du 2
Substituting: F (u, u, x)  { AE ( )  (q)(u )}
2 dx
L
L 2
du du
 I   [q  AE
dx 2
]  u  dx  [ AE u ] 0
0 dx 0
With prescribed boundary conditions at x = 0, and L: δu (0) = δu (L) = 0
L
d 2u
  [q  AE 2 ]  u  dx  0
0 dx
Recapitulating the weighted residual statement for this problem we have:
L
d 2u
  [q  AE 2 ]  W  dx  0
0 dx
Comparing the form of the equations obtained using the weak form and present
discussion; it can be observed that the weighing function W(x) used in case of
Weighted residual statement has the connotation of “variation of u(x)” i.e., δu(x)
here.
Since, all the variations of u(x) being considered satisfy the essential boundary
conditions; δu(x) at x = a, x = b is zero.
In the weak form; it may be recalled that we require the weighting function to be
zero at these point.
Further, for structural mechanics problems, if u(x) denotes the equilibrium
configuration, δu(x) (and hence W(x)) has the connotation of a virtual displacement
consistent with boundary conditions.
So δI = 0, in essence is a statement of the familiar virtual work principle.