M.E.

STRUCTURAL ENGINEERING

MA 9103 APPLIED MATHEMATICS L T P C
281103 3 1 0 4

OBJECTIVE:

- To familiarize the students in the field of differential and elliptic equations to solve boundary
value problems associated with engineering applications.

- To expose the students to variational formulation and numerical integration techniques and
their applications to obtain solutions for buckling, dynamic response, heat and flow problems
of one and two dimensional conditions.

UNIT I ONE DIMENSIONAL WAVE AND HEAT EQUATIONS 10+3
Laplace transform methods for one-dimensional wave equation Displacements in a long string
longitudinal vibration of an elastic bar Fourier transform methods for one-dimensional heat
conduction problems in infinite and semi-infinite rods.
UNIT II ELLIPTIC EQUATION 9+3
Laplace equation Properties of harmonic functions Solution of Laplaces equation by means of
Fourier transforms in a half plane, in an infinite strip and in a semi-infinite strip Solution of Poisson
equation by Fourier transform method.
UNIT III CALCULUS OF VARIATIONS 9+3
Concept of variation and its properties Eulers equation Functional dependant on first and higher
order derivatives Functionals dependant on functions of several independent variables
Variational problems with moving boundaries Direct methods Ritz and Kantorovich methods.
UNIT IV EIGEN VALUE PROBLEMS 9+3
Methods of solutions: Faddeev Leverrier Method, Power Method with deflation Approximate
Methods: Rayleigh Ritz Method
UNIT - V NUMERICAL INTEGRATION 8+3
Gaussian Quadrature One and Two Dimensions Gauss Hermite Quadrature Monte Carlo
Method Multiple Integration by using mapping function


TOTAL (L:30+T:15) : 45 PERIODS
REFERENCES:

1. Sankara Rao, K., Introduction to Partial Differential Equations, Prentice Hall of India
Pvt. Ltd., New Delhi, 1997.

2. Rajasekaran.S, Numerical Methods in Science and Engineering A Practical Approach,
A.H.Wheeler and Company Private Limited, 1986.

3. Gupta, A.S., Calculus of Variations with Applications, Prentice Hall of India Pvt. Ltd.,
New Delhi, 1997.

4. Andrews, L.C. and Shivamoggi, B.K., Integral Transforms for Engineers, Prentice
Hall of India Pvt. Ltd., New Delhi, 2003.





Fourier Transform Formulae

Fourier Transform Pair

1. F{ f(x) } =
1
( )
2
isx
f x e dx
t

}
OR F{ f(x) } = ( )
isx
f x e dx

}

= F(s) = F(s)

2. f(x) =
1
( )
2
isx
F s e ds
t

}
OR f(x) =
1
( )
2
isx
F s e ds
t

}



Fourier Coine Transform Pair

3. F
C
{ f(x) } =
0
2
( ) cos f x sx dx
t

}
OR F
C
{ f(x) } =
0
( ) cos f x sx dx

}


= F
C
(s) = F
C
(s)


4. f(x) =
0
2
( ) cos
C
F s sx dx
t

}
f(x) =
0
2
( ) cos
C
F s sx dx
t

}




Fourier Sine Transform Pair

5. F
S
{ f(x) } =
0
2
( ) sin f x sx dx
t

}
OR F
S
{ f(x) } =
0
( ) sin f x sx dx

}


= F
S
(s) = F
S
(s)


6. f(x) =
0
2
( ) cos
c
F s sx dx
t

}
OR f(x) =
0
2
( ) cos
S
F s sx dx
t

}


Fourier Transform of Derivatives

7. ( ) ( )
n
n
n
d f
F is F s
dx

=
`
)
if f, f',.f
n1
0 as x where F(s) = F { f(x) }







Fourier Sine and Cosine Transform of Derivatives


8. F
S
{ f '(x) } = s F
C
{ f(x) } and F
C
{ f '(x) } = s F
S
{ f(x) } f(0)


9. F
S
{ f ''(x) } = s
2
F
S
{ f(x) } + sf(0) and F
C
{ f ''(x) } = s
2
F
C
{ f(x) } f '(0)


10. F { u(x,t) } = U(s,t)


11. ( , ) ( , ) F u x t isU s t
x
c
=
`
c
)


12.
2
2
2
( , ) ( ) ( , ) F u x t is U s t
x
c
=
`
c
)


13. ( , ) ( , )
t
F u x t U s t
t
c
=
`
c
)


14. | |
2
2
2
( , ) ( , )
S S
F u x t s F u x t
x
c
=
`
c
)
if u(0,t) = 0

15. | |
2
2
2
( , ) ( , )
C C
F u x t s F u x t
x
c
=
`
c
)
if u
x
(0,t) = 0

Note: If u at x = 0 is given, take Fourier sine transform and if
u
x
c
c
at x = 0 is given take Fourier
cosine transform





















One dimensional heat equation

Example1: Solve the diffusion equation
2
2
,
u u
K
t x
c c
=
c c
< x <, t > 0 with the conditions
u(x,0) = f(x) and
u
x
c
c
, u tend to zero as x tend to .
Here u = u(x,t), < x <, t > 0
We know that the Fourier transform of u(x,t) is
1
( , ) ( , )
2
isx
u x t u x t e dx
t

=
}

Taking Fourier transform of the given differential equation, we get

K F{ u
xx
} = F{ u
t
}

K { s
2
( , ) u s t } =
1
2
isx
u
e dx
t
t

c
c
}


=
{ } ( )
d
F u
dt


=
du
dt


du
dt
+ K s
2
u = 0 ------(1) where u is the Fourier transform of u(x,t)
Solving (1), we get u (s,t) = A
2
s Kt
e

----------(2)

Since u(x,0) = f(x), u (s,0) = F(s)

Using this in (2), we have A = F(s)

u (s,t) = F(s)
2
s Kt
e

Taking inverse Fourier transform, we have
u(x,t) = F
1
{ F(s)
2
s Kt
e

}

= f(x) * F
1
{
2
s Kt
e

}

= f(x) *
2
4
2
x
Kt
e
Kt

=
2
( )
4
1
( )
4
x
Kt
f e d
K t
u
u u
t

}


putting
2
x
Kt
u
|

= , we get 2 x Kt u | = and du = 2 d Kt |

u(x,t) =
2 1
( 2 ) f x Kt e d
|
| |
t

}


Example2: Solve
2
2
,
u u
k
t x
c c
=
c c
for t > 0 with
u(x,0) =
1,
0,
x a
x a
<
>
and bounded (x,t)

Taking Fourier transform w.r.t. x on both sides of the heat equation, we get


2
2
isx isx
u u
e dx k e dx
t x



c c
=
c c
} }



2
( ) ( ) ( ) F u k is F u
t
c
=
c




2
( ) ( ) 0 F u ks F u
t
c
+ =
c
-----(1)


Solving, we get F(u) = A
2
ks t
e

-----(2)

Taking Fourier transform for the initial condition, we get

F{ u(x,0) } = ( )
isx
f x e dx

}


=
a
isx
a
e dx

}


=
2sin sx
s


putting t = 0, (2) gives A =
2sin sx
s


F(u) =
2sin sx
s

2
ks t
e

using inverse Fourier transform, we get

u(x,t) =
2 1 2sin
2
ks t isx
sa
e e ds
s t

}


=
2 1 sin
cos
ks t
sa
e sx ds
s t

}


=
2
0
2 sin
cos
ks t
sa
e sx ds
s t

}



Example3: Solve
2
2
,
u u
k
t x
c c
=
c c
for x > 0 subject to the conditions
u(x,0) =e
x
, u(0,t) = 0,

Taking Fourier sine transform for the given equation, we get


2
2
0 0
2 2
sin sin
u u
sx dx k sx dx
t x t t

c c
=
c c
} }



2
( ) ( ) ( )
s S
d
F s k is F s
dt
=


2
( ) ( ) ( )
s S
d
F s k is F s
dt
=


2
( ) ( ) 0
s S
d
F s ks F s
dt
+ =

The solution of the equation is F
S
(s) = A
2
ks t
e

-----(1)

Taking Fourier sine transform of u(x,0) = e
x
, we have F
S
(s) =
2
1
s
s +


When t = 0 equation (1) gives A = F
S
(s) =
2
1
s
s +


Now (1) becomes F
S
(s) =
2
1
s
s +

2
ks t
e

Using inversion formula, we get

u(x,t) =
2
2
0
2
sin
1
ks t
s
e sx ds
s t

+
}



Example4: Solve
2
2
,
u u
t x
c c
=
c c
for x>0, t>0 given that (i) u(0,t) = 0 for t>0
(ii) u(x,0) =
1, 0 1
0, 1
x
x
< <
>
and (iii) u(x,t) is bounded


Taking Fourier sine transform, we have F
S
{ u
t
} = F
s
{ u
xx
}


2
2
0 0
2 2
sin sin
u u
sx dx sx dx
t x t t

c c
=
c c
} }



du
dt
= s
2
u +
2
(0, ) s u t
t
where u stands for Fourier sine transform of u.

using boundary condition (i) we get
du
dt
+ s
2
u = 0

Solving u (s,t) = A
2
s t
e

-----(1)

Taking Fourier sine transform of the initial condition (ii), we get

u (s,0) =
1
0
2
sin sx ds
t
}


=
2 1 cos s
s t
| |
|
\ .
-----(2)

using (2) in (1), we get A =
2 1 cos s
s t
| |
|
\ .


(1) becomes u (s,t) =
2 1 cos s
s t
| |
|
\ .

2
s t
e

By inversion theorem, we have u(x,t) =
2
0
2 1 cos
sin
s t
s
e sx ds
s t

| |
|
\ .
}



Example5: Solve
2
2
,
u u
k
t x
c c
=
c c
0 < x < , t>0 given the conditions (i) u(x,0) = 0, x>0
(ii)
u
x
c
c
(0,t) = a (constant) (iii) u(x,t) is bounded

Taking Fourier cosine transform,

F
C
{ u
t
} = k F
C
{ u
xx
}


du
dt
= k
2
2
(0, ) s u u t
x t

c

`
c

)



= k
2
2
s u a
t



`

)
( using the condition (ii) )

du
dt
+ ks
2
2
u ka
t
=
-----(1)



This is linear in u , solving


2 2 2
ks t ks t
ue ka e dt
t
=
}


=
2
2
2
ks t
e
ka c
ks t t
+

u (s,t) =
2
2 a
s t
+
2
ks t
c e

-----(2)

From condition (i) u (s,0) = 0. Using this in (2)

u (s,0) = c +
2
2 a
s t
= 0


c =
2
2 a
s t


(2) becomes u (s,t) =
2
2 a
s t

( )
2
1
ks t
e

By inversion theorem,

u(x,t) =
2
2
0
2 1
cos
ks t
a e
sx ds
s t

}


Example6: Find the temperature distribution in semi-infinite bar with its end point and the
lateral surface insulated and with initial temperature distribution in the bar is
prescribed by f(x). Deduce the solution when f(x) = e
ax
.

This problem is represented by one dimensional heat equation
2
2
2
u u
c
t x
c c
=
c c

for 0 < x < , t>0 with boundary condition u
x
(0,t)=0 (insulated end) and with the initial
condition u(x,0) = f(x) (given) for 0 < x <


Taking Fourier cosine transform to the equation


2
2
2
0 0
cos cos
u u
sx dx c sx dx
t x

c c
=
c c
} }


( )
C
d
F u
dt
= c
2
s
2
F
C
(u)



d
dt
U + c
2
s
2
U = 0 where U = F
C
(u)
Solving, we get

U(s,t) = A
2 2
c s t
e

-----(1)

When t = 0, U(s,0) = A

Taking Fourier transform for the initial condition u(x,0) = f(x), we get

U(s,0) =
0
( , 0) cos u x sx dx

}


A =
0
( ) cos f x sx dx

}


Taking inverse Fourier cosine transform of (1), we get

u(x,t) =
2 2
0
2
cos
c s t
A e sx ds
t

}


Special Case: When f(x) = e
ax
then


A =
0
cos
ax
e sx dx

}
=
2 2
a
s a +



u(x,t) =
2 2
2 2
0
2
cos
c s t
a
e sx ds
s a t

+
}


Example7: Solve
2
2
,
u u
k
t x
c c
=
c c
0 < x < , t>0 under the conditions u=u
0
at x =0, t>0
with the initial condition u(x,0) = 0 , x>0

Taking Fourier sine transform, we have F
S
{ u
t
} = k F
s
{ u
xx
}


du
dt
= k
2
2
(0, ) s u s u t
t


+
`

)



du
dt
+ ks
2
0
2
u ksu
t
=
-----(1)

This is linear in u , solving


2 2
0
2
ks t ks t
ue ku se dt
t
=
}


=
2
0
2
ks t
u
e c
s t
+ -----(2)

since u(x,0) = 0 , u (s,0) = 0. Using this in (2)

0 =
0
2 u
s t
+ c
c =
0
2 u
s t



2
ks t
ue =
0
2 u
s t
( )
2
1
ks t
e

u =
0
2 u
s t
( )
2
1
ks t
e

By inversion theorem

u(x,t) =
2
0
0
2 1
sin
ks t
u e
sx ds
s t

| |

|
|
\ .
}


Example8: Solve
2
2
,
u u
k
t x
c c
=
c c
0 < x < , t>0 under the conditions u(x,0) = e
ax
, a>0
u
x
(0,t) = 0, u
x
(x,t) = 0, t>0

In this problem, then ends of the bar have been insulated and kept at zero temperature.

Taking Fourier cosine transform


2
2
0 0
cos cos
u u
sx dx k sx dx
t x

c c
=
c c
} }


( )
C
d
F u
dt
= k s
2
F
C
(u)


d
dt
U + k s
2
U = 0 where U = F
C
(u)

Solving, we get

U(s,t) = A
2
ks t
e

-----(1)

Taking FCT of the initial condition u(x,0) = e
ax
, we get U(s,0) =
2 2
a
a s +


when t = 0, (1) becomes U(s,0) = A =
2 2
a
a s +


U(s,t) =
2 2
a
a s +

2
ks t
e

By inversion formula

u(x,t) =
2
2 2
2 1
cos
ks t
a
e sx ds
a s t

+
}


Example9: Determine the temperature distribution in semi -infinite medium x>0 when the
end x=0 is maintained at zero temperature and the initial temperature
distribution is f(x).

The given problem is described as solve
2
2
,
u u
k
t x
c c
=
c c
0 < x < , t>0 under the conditions
u(x,0) = f(x) 0 < x < , u(0,t) = 0, t>0 and u, u
x
both tend to zero as x .

Taking Fourier sine transform for the given equation, we get

2
2
0 0
2 2
sin sin
u u
sx dx k sx dx
t x t t

c c
=
c c
} }



2
( )
s S
d
U k is U
dt
= where U
s
= F
S
{ u(x,t) } as x



2
0
s S
d
U ks U
dt
+ =

The solution of the equation is U
S
= A
2
ks t
e

-----(1)

A = U
S

2
ks t
e

Taking Fourier sine transform of u(x,0) = f(x), we have U
S
= F
S
(s) at t = 0.



When t = 0 equation (1) gives A = U
S
= F
S
(s)

Now (1) becomes U
S
= F
S
(s)
2
ks t
e

Using inversion formula, we get

u(x,t) =
2
0
2
( ) sin
ks t
S
F s e sx ds
t

}












ELLIPTIC EQUATION

Solved Problems

1. Solve the following boundary value problem in the half plane y>0, described by the
PDE u
xx
+ u
yy
= 0, < x < , y > 0 with the boundary conditions
u(x,0) = f(x), < x < , u is bounded as y, u and u
x
both vanishes as |x|.

Since x has an infinite range of values, we take Fourier transform of the PDE in the variable x



1
2
isx
xx
u e dx
t

}
+
1
2
isx
yy
u e dx
t

}
= 0


( )
isx isx
x x
e u is u e dx

`
)
}
+
isx
yy
u e dx

}
= 0

( )
0
isx isx
is e u is ue dx

(


` (
)
}
+
isx
yy
u e dx

}
= 0

Since u and u
x
both vanishes as |x|, we get

s
2
U(s,y) +
2
2
( , )
d
U s y
ds
= 0

where U = F{u(x,y)}

The solution of the equation is U(s,y) = A(s) e
sy
+ B(s) e
sy

Since u must be bounded as y, its Fourier transform U also should be bounded

A(s) = 0 for s<0 and B(s) = 0 for s > 0

U(s,y) = K e
| s | y
-----(1)

From the boundary condition u(x,0) = f(x), we get U(s,0) = F(s) -----(2)

When y = 0, (1) & (2) becomes U(s,0) = K = F(s)

(1) becomes U(s,y) = F(s) e
| s | y


=
| |
1
( )
2
s y isx
f x e e dx
t

(
}


Taking inverse Fourier transform

u(x,y) = ( )
| |
1 1
2 2
s y i s isx
f e e ds e ds

t t



(
(

} }


=
( ) | |
1
( )
2
is x s y
f d e ds

t



} }

But
( ) | | is x s y
e ds

}
=
0
( ) is x sy
e ds
+

}
+
( )
0
is x sy
e ds

}


=
| |
0
( ) [ ( )]
( ) ( )
is x sy s y i x
e e
i x y y i x


+

(
(
+
(
(
+




=
2 2
2
( )
y
y x +


u(x,y) =
2 2
( )
( )
y f
d
y x

+
}



2. Solve u
xx
+ u
yy
= 0, y>0 subject to the conditions u and u
x
0 as x
2
+y
2


u(x,0) =
1, | | 1
0, | | 1
for x
for x
<
>


Taking Fourier transform for the given pde, we get

s
2
U(s,y) +
2
2
( , )
d
U s y
ds
= 0

where U = F{u(x,y)}

since u0 for large value of y, F{u} = U 0 for large value of y, the solution is


U(s,y) = A e
|s|y
-----(1) ( Refer previous example)

From the initial condition F{ u(x,0) } = U(s,0) =
2sin s
s

When y = 0, (1) gives U(s,0) = A =
2sin s
s

the solution becomes U(s,y) =
2sin s
s
e
|s|y


To find u(x,y), we use convolution theorem. Let F
1
(s) =
2sin s
s
, F
2
(s) = e
|s|y



1
1
2sin s
F
s

`
)
= u(x,0)


{ }
1 | | | |
2
1
2
s y s y isx
F e e e ds
t

=
}


=
0
1
cos
sy
e sx ds
t

}
{ since the integrand is even}

=
2 2
1 y
x y t +


Hence convolution theorem gives,

u(x,y) =
{ }
1 1 1
1 2
( ). ( ) F F s F s



=
2 2
1
( , 0)
( )
y
x dt
x t y
|
t

+
}


=
2 2
1
( )
y
dt
x t y t

+
}


=
1 1
1 1
tan tan
y x x
x y y

( | | | | +
+
( | |
\ . \ .


3. Solve u
xx
+ u
yy
= 0 subject to the conditions u(x,0) = f(x), u
y
= 0 at y = 0

Taking Fourier transform and solving the differential equation, we get the solution as

U(s,y) = A e
sy
+ B e
sy
----(1)

{ Refer example (1) }

Since u(x,0) = f(x), U(s,0) = F(s) where U(s,0) = F{ u(x,y) }

When y = 0, (1) becomes U(s,0) = A + B

A + B = F(s)

Differentiating (1) w.r.t. y, we get U '(s,y) = A s e
sy
s B e
sy


But u
y
= 0 when y = 0. Therefore A s B s = 0

i.e. A = B

i.e. A = B = F(s)

(1) becomes
1
( , ) ( )
2
sy sy
U s y F s e e

( = +



Taking inverse transform,

u(x,y) =
{ }
1 1
( )
2 2
sy sy isx
F s e e e ds
t

+
}


=
( )
1 1
( )
2 2
i x iy s
F s e ds
t

}
+
( )
1 1
( )
2 2
i x iy s
F s e ds
t

+

}


= | |
1
( ) ( )
2
f x iy f x iy + +

4. Solve the Laplaces equation in the semi infinite strip shown in the Figure:



The Laplaces equation is
2 2
2 2
0
u u
x y
c c
+ =
c c
. Here the boundary conditions are
u(0,y) = 0 for 0 < y < b u(x,b) = 0 for 0 < x <

u(x,0) = e
ax
with a > 0.

The region is 0 < x < , 0< y < b.

Taking Fourier sine transform to the given equation, we get

s
2
U + s u(0,y) +
2
2
d
U
ds
= 0 where U = F
S
{ u(x,y) }


2
2
d
U
ds
+ s
2
U = 0 -----(1)

The solution is U(s,y) = A cosh sy + B sinh sy -----(2)

When y = b, U(s,b) = F
S
{ u(x,b) } = F
S
{ 0 } = 0

From (2) U(s,b) = A cosh sb + B sinh sb = 0 -----(3)

When y = 0, U(s,0) =
0
( , 0) sin u x sx dx

}

But u(x,0) = e
ax
and U(s,0) =
2 2
s
s a +


From (2) U(s,0) = A =
2 2
s
s a +


Using this in (3), we get B =
2 2
s
s a +
coth sb

the solution becomes U(s,y) =
2 2
s
s a +
cosh sy
2 2
s
s a +
coth sb sinh sy

Y
X
O(0,0) f(x) = e
ax
Y=b
0
0
V
2
u = 0

u(x,y) =
1
{ ( , ) }
S
F U s y

=
2 2
0
2
(cosh coth .sinh )sin
s
sy sb sy sx ds
s a t

+
}



5. Solve the following Neumann problem described by u
xx
+ u
yy
= 0, <x<, y>0
subject to the conditions u
y
(x,0) = f(x) and u, u
x
both vanishes as |x|


Let us define |(x,y) = u
y
(x,y). Then |
xx
+ |
yy
=
y
c
c
( u
xx
+ u
yy
) = 0 -----(i)
And |(x,0) = u
y
(x,0) = f(x) -----(ii)

Thus the function | is the solution of the above problem(1). the solution of (i)
subject to (ii) is of the form

|(x,y) =
2 2
( )
( )
y f
d
y x

+
}


u(x,y) = ( , ) x y dy |
}


= ( )
2 2
1
( )
y
f dy d
x y

t

+
} }


= ( ) ( )
2
2
1
log
2
f x y d
t

(
+

}
+ C is the required solution.

6. Using the method of integral transform, solve the following potential problem in the semi
infinite strip described by u
xx
+ u
yy
= 0, 0 < x < , 0 < y < b subject to the boundary
conditions u(x,0)=f(x), u(x,b)=0, u(x,y)=0 and u
x
0 as x .

Taking Fourier sine transform to the given equation, we get

s
2
U + s u(0,y) +
2
2
d
U
ds
= 0 where U = F
S
{ u(x,y) }


2
2
d
U
ds
s
2
U = 0 -----(1)

The solution is U(s,y) = A cosh sy + B sinh sy -----(2)


The sine transform of the boundary conditions becomes U(s,0) =
0
2
( ) sin f x sx dx
t

}

When y = 0, U(s,0) =
0
( , 0) sin u x sx dx

}
and U(s,b) = 0
Using these two relations in (2), we have A =
0
2
( ) sin f x sx dx
t

}


0 = A cosh sb + B sinh sb

B =
cosh
sinh
sb
sb

0
2
( ) sin f x sx dx
t

}


the solution becomes

U(s,y) =
0 0
2 cosh sinh
cosh ( ) sin ( )sin
sinh
sa sy
sy f x sx dx f x sx dx
sa t

(

(

} }


=
0
sinh( ) 2
( ) sin
sinh
a y s
f x sx dx
sa t

(

(

}



0
sinh( ) 2
( ) sin
sinh
a y s
f s d
sa

t

(

(

}



u(x,y) =
1
{ ( , ) }
S
F U s y

=
0 0
sinh( ) 2
( ) sin sin
sinh
a y s
f d s sx ds
sa

t

(

(

} }