ME 256

Variational Methods and Structural Optimization

Solution to Mid-term Examination
Open notes but closed books

Marks: 30

Aug.-Dec., 2005

Time: 90 minutes

Question 1 (5 marks)
x2

Consider J = F ( x, y, y , y ," , y n )dx and f ( x) . What possible terms can you add to the integrand
th

x1

of J using f ( x) and y ( x) and their derivatives so that the new functional and J would have the
same Euler-Lagrange extremizing differential equation as the extremizing solution?
Clearly, adding y , y, y,", y n individually or as a product with a constant does not contribute
th

any new terms to the E-L equation of the functional: J = x F ( x, y, y , y, y," , y n ) dx .

x2

th

Likewise, f , f , f ,", f n behave the same way.

th

Furthermore, consider f y + yf . Its terms in the E-L equation will be

( f y + y f )
y

d ( f y + y f )

= f f = 0 . So, it too does not add any new terms to the E-L
dx
y

equation.
Similarly, f n y + y n f would also not add any new terms.
th

th

Question 2 (5 marks)
What functional, when optimized, would lead to the Poissons equation: 2 z ( x, y ) = g ( x, y ) ? Please
also write the boundary condition that arises when such a functional is extremized.
2 z ( x, y ) g ( x, y ) =

2 z 2 z
+
g ( x, y ) = 0
x 2 y 2

Since for a functional of the form J = x

x2
1

F z z


z dx z x dy z y

y2
y1

F ( x, y , z ( x, y ), z x , z y ) dx dy , the E-L equation is:

= 0 , it can be seen that the terms in the Poissons equation arise from

2
1 z 1 z
1
1
2
F = + + g ( x, y ) z = ( z x ) + ( z y ) + g ( x, y ) z .
2 x 2 y
2
2
2

x2

The general form of the boundary condition for J = x

y2
y1

F ( x, y, z ( x, y ), z x , z y ) dx dy is:

F
F
dy
dx z = 0 where = the boundary of the domain . This in this case, gives:

zx
z y

October 14th, 2005

1 of 4

Ananthasuresh, IISc

ME 256

Variational Methods and Structural Optimization

( z dy z dx ) z = 0
x

Aug.-Dec., 2005

where = the boundary of the domain .

Question 3 (3+4+1+2 = 10 marks)

It is desired that an axially loaded fixed-free bars cross-section area, A( x) , be designed with a
given volume of material, V * , such that the overall stiffness and the displacement at a particular
point, x = x , are both maximized.
(a) Justify that the following problem statement meets the above requirements and explain
what the symbols in the problem statement stand for.
L

Minimize J =
A( x )

P2

2 EA dx
0
L

Pq

EA dx
0

Subject to

Adx V * 0

Data: E = Young's modulus, f ( x) = axial load, L = length, V *

(b) Write down the Euler-Lagrange necessary conditions for this problem (You need not
write the boundary conditions).
(c) How do you justify that the volume constraint must be active/inactive?
(d) Determine the expressions for the optimal A( x) and the Lagrange multiplier
corresponding to the constraint.
(a) Recall that the minimizing strain energy is equivalent to maximizing the stiffness of an elastic
L

structure. For an axially loaded bar, the strain energy is given by EAu 2 dx . This can be re2

0
L
L
( EAu )
P2
1

written as EAu 2 dx =
dx =
dx where P( x) is the internal force due to the
2
2 AE
2 AE

0
0
0
applied load f ( x) . Since it is a fixed free-bar, it is statically determinate. In a statically
determinate structure P( x) can be determined without knowing the area of cross-section.
Furthermore, it is independent of A( x) . Thus, the numerator in the functional to be minimized
L

serves the purpose of maximizing stiffness.

L

Next, note that the mutual strain energy,

( EAuv)dx =
0

( EAu )( EAv)
AE

dx =
0

Pq
dx is the
AE

displacement at x = x under f ( x) where q( x) is the internal force and v( x) the deformation

when a unit load is applied at x = x . Since this term is in the denominator, minimizing the
functional implies maximizing the displacement at x = x .

October 14th, 2005

1 of 4

Ananthasuresh, IISc

ME 256

Variational Methods and Structural Optimization

Aug.-Dec., 2005

The volume constraint is there as per the statement of the problem. Note that this being a
statically determinate structure, governing differential equation is absent in the constraint.
L

(b) First, we write the Lagrangian for the problem as L =

P2

2 EA dx
0
L

Pq
0 EA dx

Adx V *

} . Note that the

ratio of two integrals is dealt with by using the rules of ordinary calculus while taking the
variations.
By taking the variation with respect to A( x) , we get:
L P 2 Pq
P2
dx

2
EA EA2
0

2
EA
L =0 L

2
Pq
L Pq
dx

dx

0 EA
0 EA

L Pq P 2
L P 2 Pq
L Pq
dx
dx
dx
+ = 0
+
=

2
2

0 EA 2 EA 0 EA EA
0 EA

(c) If = 0 , from the last equation it implies that

L Pq P
L P 2 q
+
dx
dx 2 = 0 .

2
0 EA 2 EA 0 EA EA

L = 0 P

Since P cannot be zero in the entire span of the

bar for any given load f ( x) , we note that

L Pq P
L P2 q
L Pq P L P 2
dx
dx 2 = 0
dx =
dx q

+
2
0 EA 2 EA 0 EA EA
0 EA 2 0 EA
L P2
dx

EA P
0

L
=
Pq 2q
dx

0 EA
P
, is determined once f ( x) and x are given. So, this
The last result in the above equation, i.e.,
2q

means that the functional to be minimized is already determined! Clearly, that value is not the
mimimum. So, we argue that 0 . So, the constraint must be active.
(d) From the E-L equation in (b), the expression for
L Pq P 2 Pq
dx
+

0 EA 2 E E
A( x) =
L Pq

dx
0 EA

A( x)

can be obtained as

The expression for the Lagrange multiplier can be determined by substituting the above
expression for A( x) into the active volume constraint.

October 14th, 2005

1 of 4

Ananthasuresh, IISc

ME 256

Variational Methods and Structural Optimization

Aug.-Dec., 2005

The expressions for A( x) and need to be simultaneously solved here by using a numerical
technique.

Question 4 (10 marks)

Write down the extremizing differential equation and the boundary conditions for the following
functional if the end points of y ( x) lie on curves 1 ( x) and 2 ( x) .
x2

J = F ( x, y, y )dx + ( x1 , y1 , x2 , y2 )
x1

Note that y1 = yx = x and y2 = yx = x .

1

By taking variation with respect to y ( x) , we get:

x
2

x1 +
y1 +
x2 +
y2 = 0
y1
x2
y2
x1

J = 0 ( Fx x + Fy y + Fy y )dx +
x1

The first term in the integrand comes out of the integral to join the other (boundary) terms and
integration by parts is needed to get rid of y . This gives:
x2

x2

x1 +
y1 +
x2 +
y2 = 0
Fy ( Fy ) y dx + ( Fy y ) x + F x2 F x1 +
1

y2
x
y
x

1
1
2
x1

Note that y x = y1 y x x1 and y x = y2 y x x2 . Since the end points lie on give curves
1

and y2 = 2 ( x) , we can write y1 = 1 x1 and y2 = 2 x2 . By substituting these

relationships into the long equation preceding this paragraph, we get:
y1 = 1 ( x)

+
1 x1 + F + Fy (2 y ) +
2 x2 = 0
Fy ( Fy ) y dx F + Fy (1 y )
x1 y1 x
x2 y2 x

x1

1
2

Since y , x1 and x2 are arbitrary, by fundamental lemma of calculus of variations, we write

Fy ( Fy ) = 0

1 = 0
F + Fy (1 y )
x1 y1 x

+
2 = 0
F + Fy (2 y ) +
x2 y2 x

The first in the above set is the E-L differential equation and the other two are boundary
conditions.
October 14th, 2005

1 of 4

Ananthasuresh, IISc