CHAPTER - 5

Application Of Finite Element Method To The

Partial Differential Equation Of One As Well As
Two Space Variables And Comparison With
Spline Results

1 5.1 Introduction
5.2 Finite E l e m e n t M e t h o d To Solve Parabolic Linear Partial
Differential Ekjuation Having One Space Variable
5.3 Finite E l e m e n t S o l u t i o n Of The Flow Of Electricity In The
T r a n s m i s s i o n Lines
5.4 Finite E l e m e n t S o l u t i o n Of The Heat C o n d u c t i o n Problem
5.5 Finite E l e m e n t M e t h o d To Solve Parabolic Linear Partial
Differential Ek[uation Having Two Space Variable
5.6 Finite E l e m e n t S o l u t i o n Of The Heat Flow In Thin
Rectangular Plate
5.7 Finite E l e m e n t M e t h o d To Solve Hyperbolic Linear Partial
Differential Equation Having One Space Variable
5.8 Finite E l e m e n t S o l u t i o n Of The Flow Of Electricity In The
T r a n s m i s s i o n Line
5.9 Finite E l e m e n t S o l u t i o n Of Vibrating String Problem
5.10 Finite E l e m e n t M e t h o d To Solve Hyperbolic Linear Partial
Differential Equation Having Two Space Variables
5.11 Finite E l e m e n t S o l u t i o n Of Vibrating Membrane Problem
5.12 Conclusion

255
5.1 INTRODUCTION:

We develop the finite element models of one and two

dimensional time dependant problems and describe time
approximation schemes to convert ordinary differential
equations in time to algebraic equations. The finite element
formulation of time -dependant problems involves two
steps:
1.) Spatial Approximation : where the solution u of the
equation under consideration is approximated by
expressions of the form
u(x,t)«U^(x,t) = u;(t)y;(x) ....(5.1.1)
And the spatial finite element model of the equation is
developed using the procedures of static or steady - state
problems, while carrying all time dependant terms in the
formulation. This step results in a set of ordinary differential
equations in time for the nodal variables Uj(t)of the element.
Equation (5.1.1) represents the spatial approximation of u
for any time t.
2.) Temporal Approximation : where the system of
ordinary differential equations are further approximated in
time. This step allows conversion of the system of ordinary-
differential equations in to a set of algebraic equations

among Uj at time t^^, =(S + 1)A), where At the time

increment and s is is an integer.

All time approximation schemes seek to find u at
time tj^i using the known values of Uj from previous times:

256
C o m p u t e {u},^, u s i n g {uj^Ju},^,,...
T h u s , at the e n d of two-stage approximation, one
h a s a c o n t i n u o u s spatial solution at discrete interval of
time:

u'(x,tj = tu](tjxif]ix) (s = 0, 1,2...)

j=i

5.2 FINITE ELEMENT METHOD TO SOLVE PARABOLIC

LINEAR PARTIAL DIFFERENTIAL EQUATION
HAVING ONE SPACE VARIABLE :

We consider differential equation t h a t c o n t a i n s second

order spatial derivatives a n d first order time derivatives.
Al 2 ^ U
0<x<l ...(5.2.1)

where c^=^
Re
subject to appropriate b o u n d a r y condition
u(0, t) = 0
u ( l , t) = 0 ...(5.2.1a)
a n d initial condition
u(x,0) = f(x) ; 0<x<l ...(5.2.1b)
We divide the region 0 < x < 1 into say, n equal s u b
intervals called elements. Let these be given by
Xo,Xi, X2, ,Xn. The elements are n u m b e r e d a s (1), (2),
(3), , (n) a n d h^ t h e length of the node from node e to
e+1. Let we have t h e solution on h a n d at time S A t at point
^05 X j , X 2 , , Xjj .

Now we begin with semi discrete model.

257
S e m i d i s c r e t e Finite E l e m e n t Model :
The semidiscrete formulation involves approximation of
the spatial variation of the d e p e n d e n t variable, which
follows essentially t h e s a m e steps a s described in section
(2.13). The first step involves t h e construction of t h e weak
form of t h e equation over a typical element.
Following the three - step procedure of constructing
t h e weak form of a differential equation, we c a n develop the
weak form of (5.2.1) over a n element. Integration by p a r t s is
u s e d on t h e first t e r m once a n d on t h e second t e r m twice to
distribute the spatial derivatives equally between the weight
function v a n d t h e d e p e n d e n t variable v :
f du
0 = f V dx
at dx V ax
e+1
Xe+1 nx
au 2 av au dx + au"
= J V + C - c
ax,
Xe
at dx dx
Xe+1
au 2 av au dx
= J V + C Q l V ( X e ) - Q2 v(Xe+i)
at dx dx
...(5.2.2a)
2 au
where Qi - - c —
5x_- I X f

2 au
Q2 = c — ...(5.2.2b)
dx_-•xe+l
Next, we assume that u is interpolated by an
expression of t h e form (5.1.1). Equation (5.1.1) implies t h a t ,
at a n y arbitrarily fixed time t > 0, t h e function u c a n be
a p p r o x i m a t e d by a linear combination of (//J, with Uj (t)

258
being the value of u at time t at the j * node of element

Q^. In other words, t h e time a n d spatial variations of u are

sepairable. We obtain t h e finite element solution in t h e form,

u(x,t3) = ZujCtJ^^JCx)
,..(5.2.3)
= i(upVJ(x) (s = l,2, ,)
j=i

where (up^ is the value of u(x, t) at time 1 = 1^ a n d node j of

t h e element Q®.
Substituting v = 4^i(x) a n d (5.1.1) into (5.2.2a), we
obtain
Xe+l ^
d^j ^ •

0 = J ¥i + c dx - Q i
r , dt
u=i J
dx u=i dt
n 1 1 du.-
=z ij
dt
- R
j=l

...(5.2.4)
In matrix form we have
[ K ] { u } + [ M ] {u} = { F } ..(5.2.4a)
where
Xe+l
Mij = \¥iW] dx

K , = T c ^ ^ ' ^ d x .(5.2.4b)
•J ^ dx dx

Fi = J ^ i f d x + Qi
Xe

This completes the semi discrete finite element

formulation of (5.2.1) over a n element.

259
T i m e Approximation :
Using the 6 family of approximation in section (2.12),
we write the equation (5.2.4a) a s set of linear adgebraic
e q u a t i o n s in matrix form.
[K] {uh.i ={F} ...(5.2.5a)
where
[K] = [M] + 6'At[K]
[F] = [[M]-(l-^)At[K]{u}, +At(^{FUi+(l-^){F}J
...(5.2.5b)
The above equation (5.2.5a) is u s e d to solve equation
(5.2.1).
Here At is time step. Equation (5.2.5) is valid for a
typical element. The assembly, t h e imposition of b o u n d a r y
condition, a n d the solution of the assembled equation are
t h e s a m e a s described before. The calculation of [K] a n d {F}
requires the knowledge of t h e initial conditions {u}o a n d t h e
time variations of {F}.
Now UQ a n d u^ are k n o w n d u e to t h e prescribed
boundary condition. The set of s i m u l t a n e o u s equation
c o n t a i n s only (n-1) u n k n o w n s . These (n-1) e q u a t i o n s in
(n-1) u n k n o w n s c a n be solved by a n y stamdard m e t h o d .
Once the value of u are k n o w n at (s + 1)*^ level, we c a n
proceed to c o m p u t e t h e next level (s+2) by repeating t h e
s a m e process.
The convergence a n d stability of C r a n k - Nicolson is
unconditionally stable for 6-Ml.

260
5.3 FINITE ELEMENT SOLUTION OF THE FLOW OF
ELECTRICITY IN THE TRANSMISSION LINES :
The problem is already d i s c u s s e d in c h a p t e r - 3 section
(3.2).
Let the telegraphic equation is
1
u< .(5.3.1)
RC u
XX

Now we solve t h i s problem with the s a m e b o u n d a r y

a n d initial condition. Let t h e length of t h e t r a n s m i s s i o n line
1
is divide into 20 s u b intervals each of width h^ =
20

Let RC = c^ = At =
100 400

Now u s i n g equation (5.2.4a) a n d (5.2.4b) we get

Qo-Po
4 1 0 0 0 0 0 0 0 0 0 1 -1 0 0 0 0 0 0 0 0 0 '
Qi-Pi
1 4 1 0 0 0 0 0 0 0 0 -1 2 -1 0 0 0 0 0 0 0 0
0 -1 2 -1 0 0 0 0 0 0 0 Q2-P2
0 1 4 1 0 0 0 0 0 0 0
0 0 1 4 1 0 0 0 0 0 0 0 0 -1 2 -1 0 0 0 0 0 0 Q3-P3
0 0 0 1 4 1 0 0 0 0 0 0 0 0 -1 2 -1 0 0 0 0 0 Q4-P4
0 0 0 0 1 4 1 0 0 0 0 1 0 0 0 0 -1 2 -1 0 0 0 0 = Q5-P5
.0 0 0 0 0 1 4 1 0 0 0 Rch
. 0 0 0 0 0 -1 2 -1 0 0 0
Q6-P6
0 0 0 0 0 0 1 4 1 0 0 0 0 0 0 0 0 -1 2 -1 0 0
Q7-P7
0 0 0 0 0 0 0 1 4 1 0 0 0 0 0 0 0 0 -1 2 -1 0
0 0 0 0 0 0 0 0 1 4 1 0 0 0 0 0 0 0 0 -1 2 -1
Qs-Ps
0 0 0 0 0 0 0 0 0 2 4 0 0 0 0 0 0 0 0 0 -2 2 Q9-P9
Qio -Pio

...(5.3.2)
Using boundary and initial condition and from
e q u a t i o n (5.2.5a) a n d (5.2.5b), we get

261
h (9At h
fiiAt
0 0 0 0 0 0 0 0
3 Rch 6 Rch
h (9At 2h 26111 h
0 0 0 0 0 0 0
6 Rch 3 Rch 6 Rch
h 2h lOM h OM.
0 0 0 0 0 0 0
6 Rch 3 Rch 6 Rch
h 2h 2(9At h (9At
0 0 0 0 0 0 0
6 Rch 3 Rch 6 Rch
h 2h 2<9At h OM
0 0 0 0 0 0 0
6 Rch 3 Rch 6 Rch
h (9At 2h 2<9At h 6'At
0 0 0 0 0 0 0
6 Rch 3 Rch 6 Rch
h 9AX 2h 29At h 6'At
0 0 0 0 0 0 0
6 Rch 3 Rch 6 Rch
h 6'At 2h 26'At h 6At
0 0 0 0 0 0 0
6 Rch 3 Rch 6 Rch
h 6'At 2h 26'At h 6'At
0 0 0 0 0 0 0
6 Rch 3 Rch 6 Rch
h 6'At 2h 26'At
0 0 0 0 0 0 0 0
6 Rch 3 Rch

262
 h -f < +
h . ( -^)At
+ - Rch
1
u;+AtPo"

l-^)At 2h
—+ Rch u"
j"0 v3 !p,h JJ •
Rch IV6 fi Rch j ^ '
6
^2h -^)At^ ' h I(l-<9)At
I - f7l/\T
n+l —+ uf + u5 + u?+AtP,"
Ur

.n+1 u Rch V

2h
Rch
-^)At^
V 6 Rch j

—+ u- u U ^h (l-g)At < + AtP"

V' Rch Rch 6 Rch j
„n+l l-6')At^ 2h l-^)At^
• + uV + uS +
n+l V' Rch v3 Rch v6 Rch /
U
.n+l h l-^)At\n ^2h l-^)At^
U, —+ + u" +
6 Rch ) 3 Rch J 6 Rch j ^ ^
n+l
U,
2h
n+l + + u^ + + •
U
6 Rch J 3 Rch J \6
U Rch yj ^ ^
.n+l
l-^)At^ r2h 1 - g)At
—+ u. + U7 + — - + +^^ ^— Uo+AtP."
.n+l

n+l
u Rch j
l-^)At^
v3
f2h
Rch j '
l-^)At^
"
' h (l-^)At
Rch j ^ '
MO —+ u? u" + u^ + A t Pg"
Rch j .3 Rch j , ,6 Rch ,j

—+
l-^)At^
Uc
+.3 ii^Vs+f^+^i^^luro+Atp,"
u Rch J
l-^)At^ ^2h
Rch J ^ I6 Rch /i '" ^

Rch j + v 33
• + U9"
_V 6 Rch J'^'^^U Rch J^'i^^'^'«

for n = 0

take h = Re 0 = 0.5 At -
20 100 400
we get
2.516667 -2.49166 0 0 0 0 0 0 0 0 0
2.491666 5.033334 -2.49166 0 0 0 0 0 0 0 0
0 - 2.491666 5.033334 -2.491666 0 0 0 0 0 0 0
0 0 2.491666 5.033334- 2.49166^) 0 0 0 0 0 0
0 0 0 2.491666 5.033334 -2.491666 0 0 0 0 0
0 0 0 0 -2.491666 5.033334 -2.491666 0 0 0 0
0 0 0 0 0 -2.491666 5.033334-2.491666 0 0 0
0 0 0 0 0 0 2.491666 5.033334- 2.491666 0 0
0 0 0 0 0 0 0 - 2.4916665.033334 -2.491666 0
0 0 0 0 0 0 0 0 -2.491666 5.033334 -2.491666
0 0 0 0 0 0 0 0 0 - 2.491666 5.033334

263
[ur ~
2.483333) u;)+(2.508334) ul+AtPo
u| 2.508334) uj)+(4.966666) u|+(2.508334) u^2+^tP/
"^ 2.508334) u]+(4.966666) u^+(2.508334) u^+At P]
u^ 2.508334) u^2 + (4.966666)u^+(2.508334) U4+At P]
< 2.508334) u^ + (4.966666) u^ + (2.508334) u^ + At PJ
. u' > = 2.508334) u\ + (4.966666) u^ + (2.508334) u^ + At P5'
ui 2.508334) u^5 + (4.966666) u^ + (2.508334) u\ + At P^
ui 2.508334) u^6 + (4.966666) u\ + (2.508334) u^ + At P^
u^ 2.508334) u\ + (4.966666) u^ + (2.508334) u^ + At Pg^
u> 2.508334) u^ + (4.966666) u^ + (2.508334) UJQ + At P^

u'o 2.508334) u^ + (4.966666) u[o + (2.508334) u|i + At P/o

Using initial a n d b o u n d a r y condition

ui =0 0 for i = 1,2,...., 10 , t h e n we get

2.516667-2.49166 0 0 0 0 0 0 0 0 0
•2.491666 5.033334-2.49166 0 0 0 0 0 0 0 0
0 -2.491666 5.033334-2.491666 0 0 0 0 0 0 0
0 0 2.491666 5.033334-2.491666 0 0 0 0 0 0
0 0 0 -2.491666 5.033334 -2.491666 0 0 0 0 0
0 0 -2.491666 5.033334-2.491666 0 0 0 0
0 0 2.491666 5.033334-2.491666 0 0 0
0 0 0 -2.491666 5.033334-2.491666 0 0
0 0 0 0 -2.491666 5.033334-2.491666 0
0 0 0 0 0 -2.491666 5.033334 -2.491666
0 0 0 0 0 0 -2.491666 5.033334

264
 u.
-0.001837
u- -0.003634
U:
-0.005343
U4 -0.006916
-0.008318
-0.009513
u,
-0.010481
ui
-0.011188
Us
-0.011614
U(
-0.011764
U 10

Hence we get 10 algebraic e q u a t i o n s in 10 u n k n o w n s

with tri-diagonal matrix. This c a n be solved by any
s t a n d a r d m e t h o d . Once the values of u are k n o w n at (n + 1)*
level, we c a n proceed to c o m p u t e the next level (n + 2) by
s a m e t e c h n i q u e s a s above. Due to t h e symmetry of the
solutions, results are given for 0 < x < 0.5 at
1 3 1 1 , , , . , ,,
t = , , ,— etc. a n d they are shown m the table
200 400 100 80 ^
(5.3.1) ajid plotted in figure (5.3.1).

265
 Table (5.3.1)
Current distribution in the cable through Finite element
method

Current in the cable u —•

X t = 0.0 t= 1/400 t =1/200 t =3/400 t= 1/100 t = 1/80

0.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000

0.05 0.156434 0.156395 0.156356 0.156317 0.156278 0.156239

0.10 0.309017 0.308941 0.308865 0.308789 0.308713 0.308634

0.15 0.453991 0.453879 0.453767 0.453655 0.453543 0.453429

0.20 0.587785 0.587640 0.587495 0.587350 0.587208 0.587063

0.25 0.707107 0.706932 0.706757 0.706582 0.706407 0.706232

0.30 0.809017 0.808817 0.808617 0.808417 0.808217 0.808017

0.35 0.891007 0.890787 0.890567 0.890347 0.890127 0.889907

0.40 0.951057 0.950822 0.950387 0.950352 0.950117 0.949882

0.45 0.987688 0.987444 0.987200 0.986956 0.986712 0.986468

0.50 1.000000 0.999753 0.999506 0.999259 0.999012 0.998764

266
1.200000

4 - t = Q000
• - t = 1/400
t = 1/200;

t = 1/100
t = 1/80

0.000000
0.0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

Figure (5.3.1)
Current distribution in the cable through Finite element method

267
5.3 (A) DISCUSSION OF RESULTS :
Here table (5.3.2a) gives the comparison of finite
element solution of flow of electricity cable of transmission
line problem with spline solutions as well as exact solution
at t = l / 8 0 . Table (5.3.2b) and figure (5.3.2) represents the
error analysis. And from the figure (5.3.2), it is concluded
that spline solutions are closer than the finite element
solution in national conference on Recent
Developments in Fluid Dynamics at Jaipur [2010].
Hence spline solutions are more reliable and accurate.
Table (5.3.2(a))
Comparison of solution (at t = l / 8 0 )

X UFEM UEXT USE USI

0.0 0.000000 0.000000 0.000000 0.000000

0.05 0.156239 0.156242 0.156241 0.156241

0.10 0.308634 0.308636 0.308635 0.308636

0.15 0.453429 0.453431 0.453429 0.453430

0.20 0.587063 0.587060 0.587059 0.587060

0.25 0.706232 0.706235 0.706233 0.706233

0.30 0.808017 0.808019 0.808017 0.808018

0.35 0.889907 0.889908 0.889906 0.889906

0.40 0.949882 0.949884 0.949882 0.949882

0.45 0.986468 0.986471 0.986468 0.986468

0.50 0.998764 0.998767 0.998764 0.998764

[UFEM denotes Finite element method solutions]

268
 Table (5.3.2(b))
Error analysis (at t = l / 8 0 )

X UEXT-UFEM UEXT-USE UEXT-USI

0.0 0.000000 0.000000 0.000000

0.05 0.000003 0.000001 0.000001

0.10 0.000002 0.000001 0.000000

0.15 0.000002 0.000002 0.000001

0.20 0.000003 0.000001 0.000000

0.25 0.000002 0.000002 0.000002

0.30 0.000002 0.000002 0.000001

0.35 0.000001 0.000002 0.000002

0.40 0.000002 0.000002 0.000002

0.45 0.000003 0.000003 0.000003

0.50 0.000003 0.000003 0.000003

269
 0.000004O

0.0000035 1\
0.0000030
- • - UEXT-UFB/I
0.0000025
O • • - UEXT-USE
0.0000020
HI UEXT-USI
0.000001S

0.0000010

0.000000
0.0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45

Figure (5.3.2)
Error analysis (at t = l / 8 0 )

270
5.4 FINITE ELEMENT SOLUTION O F T H E HEAT
CONDUCTION PROBLEM :
We solve this problem with s a m e p a r a m e t e r a n d initial
and b o u n d a r y condition a s we d i s c u s s e d in chapter-3
section (3.4).
Let the equation of h e a t flow is
5^
= a 0<x<l (5.4.1)
5X'
a n d b o u n d a r y a n d initial condition is
u(x,0) = x(l-x) 0<x < 1 ,.(5.4.la)
and
u(0,t) = 0
,(5.4. lb)
u(l,t) = 0
Let t h e length of h o m o g e n e o u s rod of length one meter
is divided in ten s u b equal element.
1 . 1
.-. h = At = a =1
10 500
Now u s i n g equation (5.2.5a) a n d (5.2.5b), we get

"QO=PO'
4 1 0 0 0 0 0 0 0 0
2 -1 0 0 0 0 0 0 o" Qi=Pi
141 0 0 0 0 0 0 0
-1 2 -1 0 0 0 0 0 0
014 1000000 Q2=P2
0 -1 2 -1 0 0 0 0 0
Q3=P3
0014 100000
0 0 -1 2 -1 0 0 0 0
00014 10000 Q4=P4
0 0 0 -1 2 -1 0 0 0
00001 41000 Q5=P5
0 0 0 0 -1 2 -1 0 0
00000 14100
0 0 0 0 0 -1 2 -1 0 Q6=P6
0000001 410
0 0 0 0 0 0 -1 2 -1 Q7=P7
0 0 0 0 0 0 0 141
0 0 0 0 0 0 0 -1 2
0 0 0 0 0 0 0 014 Q8=P8
.Q9=P9.
.(5.4.2)

271
 Now using equation (5.2.6a) and (5.2.6b) and (5.4.2)
can be written as a set of linear algebraic equation.
Using initial and boundary condition and n = 0, we get
u
"0.087 0.0067 0 0 0 0 0 0 0
0.0067 0.087 0.0067 0 0 0 0 0 0 u
0 0.0067 0.087 0.0067 0 0 0 0 0 u
0 0 0.0067 0.087 0.0067 0 0 0 0 u
0 0 0 0.0067 0.087 0.0067 0 0 0 u
0 0 0 0 0.0061 0.087 0.0067 0 0 u
0 0 0 0 0 0.0067 0.087 0.0067 0 u
0 0 0 0 0 0 0.0067 0.87 0.0067
0 0 0 0 0 0 0 0.0067 0.087 u
u
' 0.008522
0.015530
0.02055
0.023562
0.024566
0.023562
0.02055
0.015530
0.008502

Hence we get 9 algebraic equations in 9 unknowns

with tri-diagonal matrix form. This can be solved by any
standard method. Once the values of u are known at first
level, we can proceed to compute the next level and so on.
The result obtained by finite element method is shown in
table (5.4.1) and plotted in figure (5.4.1).

272
 Table (5.4.1)
Displacement of temperature in a rod through Finite
element method

Temperature in rod u(x, t) — •

X t = 0.0 t= 0.002 t= 0.004 t= 0.006 t= 0.008 t = 0.01

0.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000

0.1 0.09 0.085707 0.082490 0.079787 0.077409 0.075260

0.2 0.16 0.156039 0.151912 0.147996 0.144307 0.140824

0.3 0.21 0.206014 0.202047 0.198022 0.194016 0.190071

0.4 0.24 0.236016 0.232029 0.228052 0.224060 0.220063

0.5 0.25 0.246015 0.242029 0.238043 0.234064 0.230079

0.6 0.24 0.236016 0.230029 0.228052 0.224060 0.220063

0.7 0.21 0.206014 0.202047 0.198022 0.194016 0.190071

0.8 0.16 0.156039 0.151912 0.147996 0.144307 0.140824

0.9 0.09 0.085707 0.082490 0.079787 0.077409 0.075260

1.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000

273
 0.300000 |—

0.250000 -•-t=0.0
0.200000 ^ ^ ^
-«-t=0.0Q2
t=0.004
3 0.150000
-^ t=0.006
O100000 -)K-t=0.008
-*-t=0.01
0.050000

0000000 1 ! 1
1 m

00 01 02 03 0.4 0.5 06 07 0.8 0.9 1.0

Figure (5.4.1)
Displacement of temperature in a rod through finite element method

274
5.4 (A) DISCUSSION OF RESULTS :
The result presented in table (5.4.2(a)) gives the
compression of finite element solutions with spline explicit
and implicit as well as exact solution at t = 0.006. Table
(5.4.2(b)) and figure (5.4.2) indicate the error analysis and
from the figure it is clear that solution of all methods
converges to exact solution. But spline results are more
accurate than finite element solution. This justifies the wide
applicability of spline collocation method. It is also
concluded that spline solutions are more truthful than
the finite element solution in national conference of
Gujarat Science Congress at Surat [2009].
Table (5.4.2(a))
Comparison of solution (at t=0.006)

X UFEM UEXT USE USI

0.0 0.000000 0.000000 0.000000 0.000000

0.1 0.079787 0.079683 0.079756 0.079149
0.2 0.147996 0.148269 0.14795 0.147771
0.3 0.198022 0.198467 0.197975 0.198686
0.4 0.228052 0.227979 0.228006 0.228024
0.5 0.238043 0.237569 0.237998 0.237713
0.6 0.228052 0.227979 0.228006 0.228024
0.7 0.198022 0.198467 0.197975 0.198686
0.8 0.147996 0.148269 0.14795 0.147771
0.9 0.079787 0.079683 0.079756 0.079149
1.0 0.000000 0.000000 0.000000 0.000000

275
 Table (5.4.2(b))
Error analysis (at t=0.006)

X UFEM-UEXT USE-UEXT USI-UEXT

0.0 0.000000 0.000000 0.000000

0.1 0.000104 0.000397 0.000534

0.2 0.000273 0.000109 0.000498

0.3 0.000445 0.000467 0.000218

0.4 0.000073 0.000021 0.000045

0.5 0.000474 0.000431 0.000144

0.6 0.000073 0.000021 0.000045

0.7 0.000445 0.000467 0.000218

0.8 0.000273 0.000109 0.000498

0.9 0.000104 0.000397 0.000534

1.0 0.000000 0.000000 0.000000

276
 0.000600

0.000500

0.000400
••-UFEM-UEXT
O 0.000300 -*—USE-UEXT
m USI-UEXT
0.000200

0.000100

0.000000
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

X - "

Figure (5.4.2)
Error analysis (at t=0.006)

277
5.5 FINITE ELEMENT METHOD TO SOLVE PARABOLIC
LINEAR PARTIAL DIFFERENTIAL EQUATION
HAVING TWO SPACE VARIABLE :
Consider the partial differential equation having two
space variables.

du 2 5 u o u
0<x<a, 0<y<b, t>0
— = c
at
...(5.5.1)
with the b o u n d a r y conditions
u(x,0,t) = 0 0<x,y<l
.(5.5.2a)
u(0, y, t) = 0
a n d the initial condition are of t h e form
u(x,y,0) = Uo(x,y) 0<x,y<l ...(5.5.2b)
We s h o u l d s u b divide t h e regions 0 < x, y < a into say N
intervals, each of widthNAx. Let these elements are
n u m b e r e d a s (1), (2), ...., (n) a n d let u s have t h e solution on
h a n d at time SAt at point XQ,XI,....,X^.

Now the semidiscrete variational formulation of

equation (5.5.1), (5.5.2a) & (5.5.2b) over a n element is given
by
Xe+1
^U 2 r dv du dy du
0 1 V + C
dx dx ay ay
dxdy ...(5.5.3)
X
e u at
S u b s t i t u t i n g finite element interpolation of the form

u(x,y,t) - ^ U j ( t ) ^^j(x,y) ...(5.5.4)

and v = ^i into equation (5.4.3), we obtain

278
 —I -\
{f. e + l A du^
0 ' 7 ' ^2 [^¥i ^^j , ^¥i ^j Adxdy
Z J ^i^j dx dy
dt + dx dx 5y 5y
u
V ^e

...(5.5.5a)
In m a t r i x form
[M^"^] {u} + [K^^>]{u} = {F^")} ...(5.5.5b)
where a s u p e r p o s e d dot on u d e n o t e s the time derivatives
, . du
(u- —) .

and
xe+l
Mi j J v^i^j dx dy ,(5.5.5c)

Xe+1
Kf, = J c^ dxdy
dx dx dy dy ^

Xe+1
i J ^^i f dx dy

This completes the semidiscretization step.

Transient Analysis :
Note t h a t t h e form of (5.4.5) is the s a m e a s the
parabolic equation d i s c u s s e d in section (5.2). The spatial
d i m e n s i o n does not a p p e a r in (5.4.5a) a n d (5.4.5b) b e c a u s e
it is t a k e n into a c c o u n t in the spatial approximation. T h u s ,
irrespective of t h e spatial dimension, the finite element
model of all problems with a first time derivative is the
s a m e . Therefore, t h e time approximation s c h e m e s d i s c u s s e d
in section (2.12) for parabolic equation c a n be readily
applied.

279
 Using t h e a family approximation, we c a n transform
t h e ordinary differential equation (5.5.5) into set of algebraic
e q u a t i o n s at time t^+i.

[K:]S+I {U}S+I = {F}s, s+i ...(5.5.6a)

where [K]^i - [M] + a J K ]

[ F ] = ( [ M ] - a3[K]{u}s+ a4 {F}s+ a3 {F}s,i)

and a4=aAt aj = ( l - a ) A t ...(5.5.6b)
E q u a t i o n (5.5.6a), after a s s e m b l y a n d imposition of
b o u n d a r y conditions is solved at each time step to t h e nodal
vedues Uj at timetg = (S + l ) A t . At time t = 0 (i.e. S = 0), t h e

right h a n d side is c o m p u t e d u s i n g initial values { u }, t h e

vector { F }, which is the s u m of source vector { f } is always
k n o w n , for b o t h times t^ and t^+i, at all n o d e s at which t h e
solution is u n k n o w n .
The equation (5.5.6a) a n d (5.5.6b) is u s e d to solve
equation (5.5.1).
We get (N - 1) X (N - 1) s i m u l t a n e o u s equation in
(N - 1) X (N - 1) u n k n o w n s . These e q u a t i o n s with s q u a r e
m a t r i x c a n be solved by a n y s t a n d a r d m e t h o d .
Once t h e values of u are k n o w n at (S + 1)* level, we
c a n proceed to c o m p u t e the next level (S + 2) by s a m e
t e c h n i q u e s a s above.

5.6 FINITE ELEMENT SOLUTION OF THE HEAT FLOW

IN THIN RECTANGULAR PLATE :
The problem is already d i s c u s s e d in c h a p t e r - 3 in
section (3.7).

280
 Here the problem is solved by finite element method
with same parameter and compare the results with spline
explicit - implicit and exact solution.
au a u
Let —T + — ^ 0<x,y<l .(5.6.1)
at ^ax^ dy J

is t h e equation of h e a t flow in thin r e c t a n g u l a r plate.

Initial a n d b o u n d a r y condition is
u(x,0,t) = 0
0 < X, y < 1 (5.6.1a)
u(0,y,t) = 0
and
u(x, y,0) = sin;wsin;^ t>0 ...(5.6.1b)
Due to symmetry of problem we can compute
forO<x, y <0.5. Let t h e r e c t a n g u l a r plate is divide in 2 0 0 x
2 0 0 uniform m e s h e s of 4 0 0 linear r e c t a n g u l a r elements to
discretize a q u a d r a n t of a domain.
Since all elements in t h e m e s h are identical, we will
c o m p u t e t h e element m a t r i c e s for only one element.
We have
4^1 = ( l - 2 x ) ( l - 2 y )
4^2=2x(l-2y)
4^3 = 4xy
4^4 = (l-2x)2y
From equation (5.5.5c), we get
'4 -1 -2 -1 4 2 1 2
1-1 4-1 -2 2 4 2 1
[K^] = [M^] =
-2 -1 4 -1 36 1 2 4 2
-1 -2 -1 4 2 1 2 4

281
 '0" Ql"
0
{F(^>} = + '
0
0
QfJ
Now by assembly of element matrices, we can write
matrix form of equation {5,5.5b) as algebraic equation using
equation (5.5.5c).
0 . 0 0 1 1 1 1 U i + 0 . 0 0 0 2 7 7 7 8 U 2 + 0 . 0 0 0 2 7 7 7 8 u u + 0 . 0 0 0 0 6 9 4 4 u 12 +

0 . 0 0 2 6 6 6 7 Ui+ 0 . 0 0 0 3 3 3 U2 - 0 . 0 0 0 3 3 3 Uu - 0 . 0 0 0 3 3 3 U12 =
r(l)^i7(2) (12)
F4^'^+F3^"^+F^^^^+Fi

0.00027778Ui+ 0.001111 li 2 + 0.00027778 li 3 + 0.00006944 Ujj +

0.002778 Uj2 + 0.0000694 li 13 - 0.000333 Uj + 0.0026667 U2 -

0.000333 U3 - 0.000333 U u - 0.000333 U12 - 0.000333 U13

:(2) :(3) + F2(12)_^P(13)

F^^^ + F3^^^ + F,

0.00027778U2 +O.OOIIIIU3 + 0 . 0 0 0 2 7 7 7 8 u 4 + 0 . 0 0 0 0 6 9 4 4 Ui2 +

0 . 0 0 2 7 7 8 Ui3 + 0 . 0 0 0 0 6 9 4 li 14 - 0 . 0 0 0 3 3 3 U2 + 0 . 0 0 2 6 6 6 7 U3 -

0 . 0 0 0 3 3 3 U4 - 0 . 0 0 0 3 3 3 U12 - 0 . 0 0 0 3 3 3 U13 - 0 . 0 0 0 3 3 3 U14

:(3) r(4) (13)^17(14)

Fr+F3^'^^+F2 + F,

O.OOO27778U3+ 0 . 0 0 1 1 1 l i i 4 + O.OOO27778U5 + 0 . 0 0 0 0 6 9 4 4 11^3 +

0 . 0 0 2 7 7 8 U14+ 0 . 0 0 0 0 6 9 4 U15 - 0 . 0 0 0 3 3 3 U3 + 0 . 0 0 2 6 6 6 7 U4 -

0 . 0 0 0 3 3 3 U5 - 0 . 0 0 0 3 3 3 U13 - 0 . 0 0 0 3 3 3 U14 - 0 . 0 0 0 3 3 3 U15 =

282
0.00027778ii4 +O.OOIIIIU5 + 0.00027778 u 5 + 0.00006944 u ,4 +

0.002778 Ui5 + 0.0000694 u ig - 0.000333 U4 + 0.0026667 U5 -

0.000333 Ug - 0.000333 U14 - 0.000333 U15 - 0.000333 u^^ =

Ff)+Ff) ^p(.5)^p(16)

Hence we get 100 x 100 simultaneous equations in

100 X 100 unknowns, for j= 1,2,...., 100. Now using the
Crank Nicolson scheme, equation (5.6.2) can be written in
the algebraic equation of the form (5.5.6a) for S = 0.
0 . 0 0 1 1 1 4 uj + 0 . 0 0 0 2 7 7 3 U2 + 0 . 0 0 0 2 7 7 uji + 0 . 0 0 0 0 6 9 UJ2 =

0.0000606

0 . 0 0 0 2 7 7 uJ + 0 . 0 0 1 1 1 4 U2 + 0 . 0 0 0 2 7 7 U3 + 0 . 0 0 0 0 6 9 u | i +

0 . 0 0 0 2 7 7 UJ2 + 0 . 0 0 0 0 6 9 ujg = 0.0001198

0 . 0 0 0 2 7 7 U2 + 0 . 0 0 1 1 1 4 U3 + 0 . 0 0 0 2 7 7 U4 + 0 . 0 0 0 0 6 9 u{2 +

0 . 0 0 0 2 7 7 u}3 + 0 . 0 0 0 0 6 9 u!4 = 0.0001760

0 . 0 0 0 2 7 7 U3 + 0.001114 U4 + 0 . 0 0 0 2 7 7 U5 + 0 . 0 0 0 0 6 9 UJ3 +

0 . 0 0 0 2 7 7 u!4 + 0 . 0 0 0 0 6 9 ujg = 0.00022798

Hence we get 100 x 100 simultaneous linear equation

in 100 X 100 unknowns. This system can be solved by any
standard method. Once the values of u are known at first
time level of time, process can be repeated for second time
level and so on. Due to symmetry of the problem, results are
given in the table (5.6.1(a)) and (5.6.1(b)) and (5.6.1(c)) and
plotted in figure (5.6.1) for y=0.25.

283
 Table (5.6.1(a))
Temperature distribution in thin rectangular plate through finite element method

Temperature in thin rectangular plate u ^ at t = 1/400

\ y
0.0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

0.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0,000000 0.000000 0.000000 0.000000 0.000000

0.05 0.000000 0.024469 0.048338 0.071015 0.091944 0.110609 0.126551 0.139377 0.148770 0.154500 0,156426

0.10 0.000000 0.048338 0.095486 0.140283 0.181626 0.218497 0,249987 0.275322 0.293877 0.305196 0.309001

0.15 0.000000 0.071015 0.140283 0.206096 0.266834 0.321003 0.307267 0.404488 0.431748 0.448378 0.453967

0.20 0.000000 0.091944 0.181626 0.266834 0.345473 0.415605 0,475504 0.523694 0.558988 0.580518 0.587756

0.25 0.000000 0.110609 0.218497 0.321003 0.415605 0.499975 0.572032 0.630005 0,672464 0.698368 0.707069

0.30 0.000000 0.126551 0.249987 0.367267 0.475504 0.572032 0.654475 0.720803 0.769382 0.799015 0.808976

0.35 0.000000 0.139377 0.275322 0.404488 0.523694 0.630005 0.720803 0.793852 0,847355 0.879992 0.890961

0.40 0.000000 0.148770 0.293877 0.431748 0.558988 0.672464 0.769382 0.847355 0.904463 0.939300 0.951008

0.45 0.000000 0.154500 0.305196 0.448378 0.580518 0.698368 0.799015 0.879992 0.939300 0.975479 0.987639

0.50 0.000000 0.156426 0.309001 0.453967 0.587756 0.707069 0.808976 0.890961 0.951008 0.987639 0.999950

284
 Table (5.6.1(b))
Temperature distribution in thin rectangular plate through finite element method

Temperature in thin rectangular plate u ^ at t = 3/400

\ y
0.0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

0.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000

0.05 0.000000 0.024465 0.048332 0.071007 0.091934 0.110597 0.126537 0.139363 0.148754 0.154484 0.156410

0.10 0.000000 0.048332 0.095476 0.140269 0.181608 0.218475 0.249961 0.275294 0.293874 0.305194 0.308969

0,15 0.000000 0.071007 0.140269 0.206074 0.266806 0.320971 0.367229 0.404446 0.431704 0.448332 0.453921

0.20 0.000000 0.091934 0.181608 0.266806 0.345437 0.415563 0.475456 0.523642 0.558932 0.580458 0.587698

0.25 0.000000 0.110597 0.218475 0.3209711 0.415563 0.499925 0.571974 0.629941 0.672396 0.698302 0.706995

0.30 0.000000 0.126537 0.249981 0.367229 0.475456 0.571974 0.654409 0.720731 0.769304 0.798933 0.808896

0.35 0.000000 0.139363 0.275294 0.404446 0.523642 0.629941 0.720731 0.793772 0.847275 0.879904 0.890871

0.40 0.000000 0.148754 0.293874 0.431704 0.558932 0.672396 0.769304 0.847275 0.904355 0.939206 0.950913

0.45 0.000000 0.154484 0.305194 0.448332 0.580458 0.698302 0.798933 0.879904 0.939206 0.975381 0.987541

0.50 0.000000 0.156410 0.308969 0.453921 0.587698 0.706995 0.808896 0.890871 0.950913 0.987541 0.999814

285
 Table (5.6.1(c))
Temperature distribution in thin rectangular plate through finite element method

Temperature in thin rectangular plate u -> at t = 5/400

0.0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
0.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000

0.05 0.000000 0.024461 0.048326 0.070999 0.091924 0.110585 0.126523 0.139349 0.148738 0.154468 0.156394

0.10 0.000000 0.048326 0.095466 0.140255 0.181590 0.218453 0.249935 0.275266 0.293844 0.305132 0.308937

0.15 0.000000 0.070999 0.140255 0.206052 0.266778 0.320939 0.367191 0.404404 0.431660 0.448286 0.453875

0.20 0.000000 0.091924 0.181590 0.266778 0.345401 0.415521 0.475408 0.523590 0.558876 0.580398 0.587638

0.25 0.000000 0.110585 0.218453 0.320939 0.415521 0.499875 0.571916 0.629877 0.672328 0.698236 0.706925

0.30 0.000000 0.126523 0.249935 0.367191 0.475408 0.571916 0.654343 0.720659 0.769227 0.798852 0.808806

0.35 0.000000 0.139349 0.275266 0.404404 0.523590 0.629877 0.720659 0.793692 0.847191 0.879814 0.890815

0.40 0.000000 0.148738 0.293844 0.431660 0.558876 0.672328 0.769227 0.847191 0.904252 0.939153 0.956714

0.45 0.000000 0.154468 0.305132 0.448286 0.580398 0.698236 0.798852 0.879814 0.939153 0.975256 0.987477

0.50 0.000000 0.156394 0.308937 0.453875 0.587638 0.706925 0.808806 0.890815 0.956714 0.987477 0.999634

286
 0.8
0.7
06
n
t 05 n -^t=1/400
3 04 n -•-t=3/400
03 F» . t=1/80
02
01
Or^ 1 r
00 005 010 Q15 020 025 030 035 0.40 046 O50
X—^

Figure(5.6.1)
Temperature distribution in thin rectangular plate through finite element method
(at y=0.25)

287
5.6 (A) DISCUSSION OF RESULTS :
The table (5.6.2(a)) represents comparison of solution
of the Heat flow in thin rectangular plate by finite element
method with spline explicit-implicit as well as exact
solution. Table (5.6.2(b)) and figure (5.6.2) shows error
analysis. The figure (5.6.2) eilso indicates that spline
solution gives better result than finite element method. So
the spline solutions are quite accurate and reliable. From
the results it is concluded in 12^*' International
Conference of International Academy of Physical
Sciences at Jaipur [2010] that spline solutions are more
truthful than the finite element solution.
Table (5.6.2(a))
Comparison of solution (at y = 0.25 & t = 1/80)

X UFEM UEXT USE USI

0.0 0.000000 0.000000 0.000000 0.000000

0.05 0.110585 0.110589 0.110588 0.110588
0.10 0.218453 0.218454 0.218454 0.218454
0.15 0.320939 0.320941 0.320942 0.320941
0.20 0.415521 0.415524 0.415524 0.415524
0.25 0.499875 0.499877 0.499876 0.499877
0.30 0.571916 0.571920 0.571920 0.571920
0.35 0.629877 0.629881 0.629881 0.629881
0.40 0.672328 0.672333 0.672332 0.672332
0.45 0.698236 0.698228 0.698229 0.698229
0.50 0.706925 0.706932 0.706932 0.706932

288
 Table (5.6.2(b))
Error analysis (at y = 0.25 & t = 1/80)

X UEXT-UFEM UEXT-USE UEXT-USI

0.0 0.000000 0.000000 0.000000

0.05 0.000004 0.000001 0.000001

0.10 0.000001 0.000000 0.000000

0.15 0.000002 0.000001 0.000000

0.20 0.000003 0.000000 0.000000

0.25 0.000002 0.000001 0.000000

0.30 0.000004 0.000000 0.000000

0.35 0.000004 0.000000 0.000000

0.40 0.000005 0.000001 0.000001

0.45 0.000008 0.000001 0.000001

0.50 0.000007 0.000000 0.000000

289
 0.000009
0.000008
0.000007
0.000006
UFEM-UEXT
o 0.000005
USE-UEXT
0.000004
USI-UEXT
0.000003
0.000002
0.000001
0.000000
0.0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

Figure (5.6.2)
Error analysis (at y = 0.25 & t = 1/80)

290
5.7 FINITE ELEMET METHOD TO SOLVE HYPERBOLIC
LINEAR PARTIAL DIFFERENTIAL EQUATION
HAVING ONE SPACE VARIABLE :
We consider differential equation t h a t c o n t a i n s second
order spatial derivatives a n d second order time derivatives.

—;r = c —r- 0<x<L, t>0 ...(5.7.1)

.2
where c = Lc
with Dirichilet b o u n d a r y condition
u(0, t) = 0
...(5.7.2)
u(L,t) = 0
a n d two initial condition at t = 0 (Cauchy condition)
u(x, 0) = f(x)
Ut(x,0) = g(x)

We divide t h e region 0 < x < L into say n equal s u b

interval each of width hg. The points of division are
Xo,Xj,X2, , x^. The elements are n u m b e r e d a s (1), (2), (3),

,(n).
We begin with semidiscrete model a n d the weak form
of equation (5.7.2) is given by

f 5 u
Xe+l 2 5v 5u
0 = J dx - Qiv(xJ - Q2v(Xe+i)
xe V
di^ dx dx

.(5.7.3a)
where
du du
-a dx ...(5.7.3b)
dx '^e+l

291
 Next we assume that u is interpolated by an
expression of the form (5.7.1). E q u a t i o n (5.7.1) implies t h a t ,
at a n y arbitrarily fixed time t > 0. The function u c a n be

approximated by a linear combination of the i//^ with u^^^ (t)

being t h e value of u at time t at t h e j * node of t h e

element Q^. We obtain the finite element solution in t h e form

j=l
,(5.7.4)
= i (uj)Vj(x) (s = l,2, )
j=i

where (ujJT is t h e value of u(x, t) at time t = ts a n d node j of

t h e element Q^.
Substituting v = ^^j(x) a n d (5.7.1) into (5.7.3a), we
obtain
xe+l ^n d^U, r n U/j ^
d 2...
0 = J ^i ^ j +c dx - Q i
j = i dt' dx j=i dx

1 9 dU:
0 = Z - F. ,.(5.7.5)

In matrix form, we have

[K]{u} + [M2]{U} = {F} ,(5.7.6a)
where
xe+l
Mi j J ^i ^ j dx

K,i j 7' e^df^^dx

dx dx

292
 This completes the semidiscrete finite element
formulation of (5.7.1) over a n element.

T i m e Approximation :
Using the New Mark Integration method, we write
equation (5.7.6a) a s a set of a linear algebraic e q u a t i o n in
matrix form
[K2]{U},,I ={F2} ...(5.7.7a)
where

{F^}s.s+i = {F}s+i+[M^]s + i(ai{u}3 +ai{u}3 +a2{ii}s)

...(5.7.7b)

^0 = ^ ^ a, = aoAt a2 = — - 1 ...(5.7.7c)

equation (5.7.7) is valid over typical element a n d it is u s e d

to solve equation (5.7.1).
Note that the calculation of [ K ] and [F] requires
knowledge of initial conditions {u}o,{u}o and {ii }o • In
practice, one does n o t know{ ii }o • As a n approximiation, it
c a n be calculated from (5.7.6a).
{ii}o = [ M ^ r ^ d F l o - [ K ] { u } o ) ...(5.7.8)
Now UQ and u'^ are known due to the prescribed
b o u n d a r y condition a n d the set of s i m u l t a n e o u s equation
c o n t a i n s only (n - 1) u n k n o w n s . These (n - 1) equation in
(n - 1) u n k n o w n c a n be solved by a n y s t a n d a r d m e t h o d .

293
 Once the value of u are k n o w n at (s + 1)* level, we c a n
proceed to c o m p u t e the next level (s + 2) by repeating t h e
s a m e process.

The a c c u r a c y a n d stability for a = — and>5 = - is k n o w n

to give a n unconditionally stable (in linear problems)

s c h e m e , which corresponds to t h e constant - average
acceleration m e t h o d .

5.8 FINITE ELEMENT SOLUTION OF THE FLOW OF

ELECTRICITY IN THE TRANSMISSION LINE :
The problem is already d i s c u s s e d in c h a p t e r - 3 in
section (3.10).
Here the problem is solved with FEM a n d c o m p a r e t h e
r e s u l t s with spline a s well a s exact solution. Let the
equation

g = (LC)fH ...,5.8.1)

is k n o w n a s hyperbolic partial differential equation with one

space variable x a n d time t a n d t h e c u r r e n t distribution is
obtained a s follows :
u(0, t) = 0
u ( l , t) = 0 ...(5.8.2)
and
u(x,0) = sin;zx 0<x<l
u(x,0) = 0
Let the length of t r a n s m i s s i o n line divide the region
into 10 elements, i.e. h = 0.1 LC = 0.01 a n d At = 0.01.

294
 Now u s i n g equation (5.7.6a) a n d (5.7.6b), we get

2 10 0 0 0 Uf 1-1 0 0 0 0 >/

14 10 0 0 -1 2 - 1 0 0 0 Pl^+P^
u,
h 0 14 10 0 U' LC 0 - 1 2 - 1 0 0 Pl+Pl^
6 0 0 14 10 U'
+ h 0 0-1 2 - 1 0
Pl+Pl'
0 0 0 14 1 u. 0 0 0 - 1 2-1
P'+Pl'
0 0 0 0 2 4 u< 0 0 0 0 -12 p5

...(5.8.3)
Using the New m a r k direct integration m e t h o d and
(5.7.7a), (5.7.7b), (5.7.7c) we write e q u a t i o n (5.8.3) a s set of
linear algebraic equation in m a t r i x form by equation (5.6.7a)
[K^]s,i - [K]s,i +ao[M^]s^i
1 _ 4 _
where 40,000.
° ~ B(At)^ ~ (0.001)^
for S = 0
[K]i = [K]i -K 40,000 [M]i
2666.84 666.54 0 0 0
666.54 2666.84 666.54 0 0
0 666.54 2666.84 666.54 0
0 0 666.54 2666.84 666.54
0 0 0 1333.08 2666.84

from equation (5.7.8) we c a n c o m p u t e { ii }o •

{ii}o = - 0 . 7 5 {u}o
~- 0.231762
- 0.440838
- 0.606762
- 0.713292
- 0.75

295
 Now u s i n g equation (5.7.7b) a n d (5.7.7c), we get

0.309017 - 0.231762
0.587785 - 0.440838
{F}o.i {F}i+[M],- (40,000) 0.809017 + 80(0) + l - 0.606762
0.951057 - 0.713292

.^
1 - 0.75

ere EQ = 40,000 a, = 80 32 = 1

'1215.855287"
2312.693814
{F}o,i - 3183.151038 u
3742.019567
3934.591502

Now u s i n g equation (5.7.7a), we get

1~
2666.84 666.54 0 0 0 1215.855287
1
666.54 2666.84 666.54 0 0 U2 2312.693814
0 666.54 2666.84 666.54 0 u^ — 3183.151038
0 0 666.54 2666.84 666.54 A 3742.019567
0 0 0 1333.08 2666.84 1 3934.591502
Us

Hence we get 5 algebraic e q u a t i o n s in 5 u n k n o w n s

with tri-diagonal m a t r i x form. This c a n be solved by a n y
s t a n d a r d m e t h o d . Once the values of u are k n o w n at first
level, we c a n proceed for the next level a n d so on. Due to the
symmetry of the problem, r e s u l t s are given for 0 < x < 0.5 at
t = 0 . 0 1 , 0.02, 0 . 0 3 , 0.04 a n d 0.05 a n d they are shown in
t h e table (5.8.1) a n d plotted in figure (5.8.1).

296
 Table (5.8.1)
Current distribution in the cable through finite element
method

Current Distribution in the cable u(x, t) ->

X t = 0.0 t= 0.01 t= 0.02 t= 0.03 t= 0.04 t= 0.05

0.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000

0.1 0.309017 0.309010 0.309003 0.308996 0.308989 0.308982

0.2 0.587785 0.587772 0.587759 0.587746 0.587733 0.587733

0.3 0.809017 0.808999 0.808981 0.808963 0.808945 0.808927

0.4 0.951057 0.951036 0.951015 0.950994 0.950973 0.950952

0.5 1.000000 0.999978 0.999956 0.999934 0.999912 0.999890

297
 1.200000

1.000000
0.0
0.800000 • • — t 0.01
t t 0.02
=) 0.600000
-K t 0.03
0.400000 •^K-t 0.04
0.05
0.200000

0.000000
0.0 0.1 0.2 0.3 0.4 0.5
X — •

Figure (5.8.1)
Current distribution in the cable through finite element method

298
5.8 (A) DISCUSSION OF RESULTS :
Table (5.8.2(a)) gives the comparison of finite element
solution with spline explicit and im.plicit as well as exact
solution at t = 0.04. Table (5.8.2(b)) and figure (5.8.2) shows
an error analysis and from the figure, it is clear that the
solution of the methods converges to exact solution but
spline solution gives better approximation than any other
method. This justifies the wide applicability of spline
collocation. It is also concluded that spline solutions are
more truthful than the finite element solution in
national conference of Gujarat Science Congress at
Ahmedabad [2010].
Table (5.8.2(a))
Comparison of solution (at t = 0.04)
X UFEM UEXT USE USI
0.0 0.000000 0.000000 0.000000 0.000000
0.1 0.308989 0.308990 0.308990 0.308992
0.2 0.587733 0.587740 0.587740 0.587738
0.3 0.808945 0.808950 0.808950 0.808953
0.4 0.950973 0.950980 0.950980 0.950981
0.5 0.999912 0.999920 0.999920 0.999920

Table (5.8.2(b))
Error analysis (at t = 0.04)
X UEXT-UFEM UEXT-USE UEXT-USI
0.0 0.000000 0.000000 0.000000
0.1 0.000003 0.000000 0.000000
0.2 0.000006 0.000001 0.000001
0.3 0.000008 0.000000 0.000000
0.4 0.000008 0.000001 0.000000
0.5 0.000009 0.000001 0.000001

299
 0.000010
0.000009
0.000008
0.000007
t 0.000006 -•—UFEM-UEXT
o 0.000005 ••-USE-UEXT

m 0.000004 USI-UEXT
0.000003
0.000002
0.000001
0.000000
0.0 0.1 0.2 0.3 0.4 0.5

Figure (5.8.2)
Error analysis (at t = 0.04)

300
5.9 FINITE ELEMENT SOLUTION OF VIBRATING
STRING PROBLEM :
The problem is already d i s c u s s e d in c h a p t e r - 3 section
(3.12).
Here t h e problem is solved u s i n g FEM a n d c o m p a r e
with the solution of spline explicit a n d implicit a s well a s
exact solution.
Let the e q u a t i o n of vibrating string is
d\ ^, d\
= C^ 0<x<2 ,..(5.9.1)
di' ax-

where C^= T / m
T : Tension in string
M : Mass per unit
length
a n d initial a n d b o u n d a r y condition is defined a s
(7DC\
u(x,0) = sin
\ ^J \ 0<x<2 ...(5.9.1a)
Ut(x,0) = 0

and
u(0,t) = 01
t>0 .(5.9.1b)
u(2, t) = OJ
Let t h e length of string is divide in 10 s u b intervals, i.e.
h = 1/5 At = 1/20 = 0.05 C = 1. We have t a k e n t h e s a m e
value a s in section (3.14) a n d (4.13).
Now u s i n g e q u a t i o n (5.7.6a) a n d (5.6.7b), we get

301
 Pl^+pl
4 1 0 0 0 0 0 0 0 "1 2-1 0 0 0 0 0 0 0 Ul P2 + Pi^
1 4 10 0 0 0 0 0 U2 -1 2 - 1 0 0 0 0 0 0 U2
0 1 4 10 0 0 0 0 U3 0-1 2 - 1 0 0 0 0 0 U3

0 0 14 10 0 0 0 U4 0 0-1 2 - 1 0 0 0 0 U4 P2+Pf
h
0 0 0 14 1 0 0 0 Us + 0 00-1 2-1000 Us P|+Pl'
6
0 0 0 0 14 1 0 0 U6 0 00 0-12-100 "6 P6+P/
0 0 0 0 0 1 4 1 0 iiy 0 00 0 0-12-10 U7
P2+P1'
0 0 0 0 0 0 1 4 1 ug 0 00 0 00-12-1
0 0 0 0 0 0 0 1 4 _U9 0 00 0 000-12 U9 Pi-Pi'
p9 plO
^2 - M
...(5.9.2)
Now b y e q u a t i o n (5.6.7a)

1
where ao = = 1600.
B(At)' (0.05^

for S = 0

[K^ ]l = [ K ] i + 1600 [ M ^ l i
.3328 48.3328 0 0 0 0 0 0 0
3328 223.3328 48.3328 0 0 0 0 0 0
0 48.3328 223.3328 48.3328 0 0 0 0 0
0 0 48.3328 223.3328 48.3328 0 0 0 0
0 0 0 48.3328 223.3328 48.3328 0 0 0
0 0 0 0 48.3328 223.3328 48.3328 0 0
0 0 0 0 0 48.3328 223.3328 48.3328 0
0 0 0 0 0 0 0 223.3328 48.3328

from e q u a t i o n (5.7.8) we c a n c o m p u t e { ii }o

302
 (iilo = 75.000187 {u}o
- 23.176257
- 44.083984
- 60.676351
- 71.329377
- 75.000187
- 71.329377
-60.676351
- 44.683984
-23.176257

Now using equation (5.7.7b) and (5.7.7c), we get

0.309016 - 23.176257
0.587785 - 44.083984
0.809016 - 60.676351
0.951056 - 71.329377
}i+[M]i- 1600 1 + 80(0) + l - 75.000187
0.951056 - 71.329377
0.809016 -60.676351
0.587785 -44.683984
0.309016 _-23.176257
where ao= 1600 ai= 80 a2= 1

'471.249268"
896.372016
1233.749174
1450.360148
{F}o,i = 1524.999813
1450.360148
1233.749174
896.372016
471.249268

303
 Now using the New mark direct integration method
equation (5.9.2) can be written as set of algebraic equation
in matrix form, we get

'"!'
«\
223.3328 48.3328 0 0 0 0 0 0 0 471.24921
48.3328 223.332J! 48.3328 0 0 0 0 0 0 896.37201
'\
0 48.3328 223.3328 48.3328 0 0 0 0 0 1233.7491
0 0 48.3328 223.3328 48.3328 0 0 0 0 < 1450.3601
0
0
0
0
0
0
48.3328
0
223.3328 48.3328
48.3328
0
223.3328 48.3328
0
0
0
0
"l = 1524.9998
1450.3601
0 0 0 0 0 48.3328 223.3328 48.3328 0 4 1233.7491
0 0 0 0 0 0 48.3328 223.3328 48 896.37201
"^
0 0 0 0 0 0 0 48.3328 22 471.24926
'i
4
Hence we get 9 algebraic equations in 9 unknowns
with tri-diagonal matrix form. This can be solved by any
standard method. Once the values of u are known at first
level, we can proceed for the next level and so on. Due to the
symmetry of the problem, results are given for 0 < x < 1 at
t = 0.05, 0.10, 0.15, 0.20, 0.25 and they are shown in the
table (5.9.1) and plotted in figure (5.9.1).

304
 Table (5.9.1)
Displacement of vibrating string in the rod through
finite element method

Displacement of vibrating string u(x, t)

J 0.0 0.05 0.10 0.15 0.20 0.25

i
0.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000

0.2 0.306016 0.294075 0.279856 0.266324 0.253447 0.241192

0.4 0.587785 0.551352 0.532295 0.506547 0.482045 0.458728

0.6 0.809016 0.769884 0.732645 0.697207 0.663483 0.631390

0.8 0.951056 0.905053 0.8612076 0.819616 0.779971 0.742244

1.0 1.000000 0.951630 0.905608 0.861796 0.820111 0.780442

1.2 0.951056 0.905053 0.8612076 0.819616 0.779971 0.742244

1.4 0.809016 0.769884 0.732645 0.697207 0.663483 0.631390

1.6 0.587785 0.551352 0.532295 0.506547 0.482045 0.458728

1.8 0.306016 0.294075 0.279856 0.266324 0.253447 0.241192

2.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000

305
 0.200000

0.000000
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
X

Figure (5.9.1)
Displacement of vibrating string in the rod through finite element method

306
(A) DISCUSSION OF RESULTS :
The table (5.9.2(a)) represents the comparison of finite
element solution of vibrating string problem with spline
explicit and implicit as well as exact solution at t = 0.15.
Table (5.9.2(b)) shows an error analysis. From the figure
(5.9.2), it is clear that spline solutions give better
approximation than finite element method. So spline
solutions are more accurate and reliable than finite element
solution.
Table (5.9.2(a))
Comparison of solution (at t = 0.15)

X UFEM UEXT USE USI

0.0 0.000000 0.000000 0.000000 0.000000

0.2 0.266324 0.300477 0.300404 0.300430

0.4 0.506547 0.571538 0.571402 0.571453

0.6 0.697207 0.786659 0.786468 0.786838

0.8 0.819616 0.904506 0.924548 0.924632

1.0 0.861796 0.951056 0.972129 0.972214

1.2 0.819616 0.904506 0.924548 0.924632

1.4 0.697207 0.786659 0.786468 0.786838

1.6 0.506547 0.571538 0.571402 0.571453

1.8 0.266324 0.300477 0.300404 0.300430

2.0 0.000000 0.000000 0.000000 0.000000

307
 Table (5.9.2(b))
Error analysis (at t = 0.15)

X UEXT-UFEM UEXT-USE UEXT-USI

0.0 0.000000 0.000000 0.000000

0.2 0.034153 0.000073 0.000047

0.4 0.064991 0.000136 0.000085

0.6 0.089452 0.000191 0.000179

0.8 0.084890 0.020042 0.020126

1.0 0.089260 0.021073 0.021158

1.2 0.084890 0.020042 0.020126

1.4 0.089452 0.000191 0.000179

1.6 0.064991 0.000136 0.000085

1.8 0.034153 0.000073 0.000047

2.0 0.000000 0.000000 0.000000

308
 0.100000
0.090000
0.080000
I 0.070000
DC
0.060000 ••-UFBVI-UEXT
o
0.050000 * - USE-UEXT
UJ
0.040000 USI-UEXT
0.030000
0.020000
0.010000
0.000000
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0

Figure (5.9.2)
Error analysis (at t = 0.15)

309
5 . 1 0 FINITE ELEMENT METHOD TO SOLVE
HYPERBOLIC LINEAR PARTIAL DIFFERENTIAL
EQUATION HAVING TWO SPACE VARIABLES :
Consider t h e hyperbolic partial differential equation
with two s p a c e variables is

0<x,y<l ...(5.10.1)

with b o u n d a r y condition
u(0,y,t) = 0 0<x,y<l
.(5.10.2a)
u(x,0,t) = 0
a n d initial condition at t = 0
u(x,y,0) = Uo(x,y)
du ...(5.10.2b)
(x,y,0) = g(x,y)
a
We divide the region by 0 < x, y < a into say N
subinterval each of width NAx = NAyand let the e l e m e n t s
are n u m b e r e d a s (1), (2), (3), , (n). Let u s have t h e
solution on h a n d at time SAt at point XQ, , x^j.
Now the semidiscrete variational formulation of
equation (5.10.1) u s i n g (5.10.1a) a n d (5.10.1b) over an
element is given by
xe+l o u dv du dv du^
0 = J +c dxdy
dt- dx dx + 5y 5y

...(5.10.3)
where v(x, y) is weight function.
The semidiscrete finite element model of equation
(5.10.1) is obtained by s u b s t i t u t i n g the finite element
approximation (5.4.4) for u a n d v = i//i into (5.10.3).

310
 \
n d^u^
0 I Mi j T + ^h ^] ...(5.10.4a)
j=l V dt
J
in matrix form
,.(5.10.4b)
The coefficients
xe+l
Mi j I 11/^ xj/-^ dx dy

Xe+1 a^i a^^j a^i a^j

K, = 1 c^ dxdy
dx dx + 5y 5y
...(5.10.4c)
xe+l
Fj® = J ^i f dx dy

TRANSIENT ANALYSIS :
The hyperbolic equation (5.10.4b) c a n be reduced to a
system of algebraic equation by approximating the second
derivative. As d i s c u s s e d in section (2.12), the new m a r k
time integration scheme is the most commonly used
m e t h o d , therefore it is u s e d here. Since the m a t h e m a t i c a l
form of (5.10.4b) is exactly the s a m e a s t h a t in (5.7.6a), t h e
r e s u l t s in (5.7.7) are immediately applicable to the former.

[K^]s.i{uns.i = {F'}s,s.i ...(5.10.5a)

where

{F}s,s+i = {F}s+i+ [ M ] (ao{u}s+ai{u}s+ a2{u}s)

...(5.10.5b)

311
 1 A 1 ,
and an = z- a, = an At a, = 1
' P{Mf ^ IP
Equation (5.10.5) can be solved fu {u} at time
t = ts = (S +1) At by u s i n g k n o w n {u} at t = t^. At time t = 0, t h e
initiad conditions of the problem are u s e d to initiate the time
m a r c h i n g s c h e m e . For the second order problem {ii} at
t = 0 c a n be c o m p u t e d from equation (5.10.4b).
{u}o = [M]-^ ({F}o - [ K ] { U } o ) ...(5.10.5c)
The equation (5.10.5) is used to solve equation
(5.10.1).
We get (N-1) X (N - 1) s i m u l t a n e o u s equation in (N - 1)
X (N - 1) u n k n o w n s . These e q u a t i o n s with s q u a r e m a t r i x
c a n be solved by a n y s t a n d a r d m e t h o d .
Once the value of u are k n o w n at (S + 1) level, we c a n
proceed to c o m p u t e the next level (S + 2) by s a m e technique
a s above.

5.11 FINITE ELEMENT SOLUTION OF VIBRATING

MEMBRANE PROBLEM :
The problem is already d i s c u s s in c h a p t e r 3, section
(3.15). Here we solve the problem with s a m e vadue a n d
c o m p a r e the r e s u l t s with spline explicit - implicit as well a s
exact solution.
Let t h e equation of vibrating m e m b r a n e is given by
5 u 2 I^ u . 5 u
0<x,y<l ...(5.11.1)
ax^ + 5y J
= c 2
dC

where c =—
m

312
 We solve t h e equation (5.11.1) u s i n g t h e s a m e initial
a n d b o u n d a r y condition
u(0,y,t) = 0 0<y<t
.(5.11.1a)
u(x,0,t) = 0 0<x<t
and

u(x, y, 0) = f(x, y) = sin;zx sin;^ 0 < x, y < 1

Ut(x,y,0) = 0

...(5.11.1b)
Now we solve the problem u s i n g the finite element
model (5.10). Let the region 0<x, y < l is divided into 10
sub-interval i.e h = 0.1 At = 0.01 c = 0.1
Now u s i n g equation (5.10.4a) a n d (5.10.4c), we get
algebraic equation.
0.004444 ii'i +0.001111 ii' 2 +0.001111 tig +0.00027788 iiy +
0.026667 Ui- 0.003333 ii 2 - 0.003333 ii^ - 0.003333 117 =

0.001111 ii'i +0.004444 i i ^ +0.001111 113 +0.00027778 ii g +

0.0011111 ii 7 +0.00027778 tig - 0.003333 Uj +0.026067 U2
- 0.003333 U3 - 0.003333 Ug - 0.003333 U7 + 0.003333 Ug =

Ff+F(3)+F(^^+Ff

0.001111 ii'2 +0.004444 113 +0.001111 114 +0.00027778 ii j +

0.0011111 ii 8 +0.00027778 119 + 0.003333 U2 +0.026667 U3

- 0.003333 U4 + 0.003333 U7 - 0.003333 Ug - 0.003333 U9 =

Ff) + F3(^) + Ff) + Ff >

313
0 . 0 0 1 1 1 1 u 3 + 0 . 0 0 4 4 4 4 ii 4 + 0 . 0 0 1 1 1 1 U 5 + 0 . 0 0 0 2 7 7 7 8 iig +

0 . 0 0 1 1 1 1 1 U9 + 0 . 0 0 0 2 7 7 7 8 ii JQ + 0 . 0 0 3 3 3 3 U3 + 0 . 0 2 6 6 6 7

U4 - 0 . 0 0 3 3 3 3 U5 - 0 . 0 0 3 3 3 3 Ug - 0 . 0 0 3 3 3 3 U9 - 0 . 0 0 3 3 3 3 UJQ

...(5.11.2)
From equation (5.10.5b) and (5.10.5c), we can
calculate {F}ss_j.i and {vijs with t h e help of initial amd

boundary condition for S=0.

314
 36.954755 -0.573010
70.292242 -1.089932
96.749172 -1.500168
113.735568 -1.763356
59.794306 -1.854241
70.292242 -1.089932
133.703928 -2.073179
184.02710 -2.853488
116.337729 -3.3354479
113.735457 -3.526974
96.749172 -1.500168
184.027810 -2.853488
{F} 0,1 253.292496 {U}( -3.927489
297.761357 -4.617039
156.543346 -4.854460
113.735568 -1.763556
216.337729 -3.354479
297.763157 -4.617039
350.041658 -5.427658
184.027810 -5.706763
59.794306 -1.854241
113.735457 -3.526974
156.543346 - 4.854460
184.027811 -5.706763
96.749172 - 6.000900
...(5.11.3)
Now using the new mark time integration scheme,
equation (5.11.2) can be written in algebraic equation of the
form (5.10.5a) and using equation (5.11.3), equation
(5.10.5a) can be written as for S = 0.

315
177.8044 ui+ 44.4411 U2 + 44.4411 Ug + 11.1077 U7 =
36.954755
44.4411 uj + 177.8044 U2 + 44.4411 U3 + 11.1077 Ug +

44.4411 u + 11.10777 Ug = 70.292242

44.4411 U2 + 177.8044 U3 + 44.4411 U4 + 11.10777 U7 +

4 4 . 4 4 1 1 Ug + 11.10777 U9 = 96.749172

4 4 . 4 4 1 1 U3 + 177.8044 U4 + 4 4 . 4 4 1 1 U5 + 11.10777 Ug +
1
4 4 . 4 4 1 1 U9 + 11.10777 u 10 = 113.735568

44.4411 u i + 8 8 . 9 0 2 2 U5 + 11.1077 U9 + 2 2 . 2 2 0 5 5 UJQ

59.794306

Hence we get 25 x 25 simultaneous linear equation in

25 X 25 unknowns. This system can be solved by any
standard method. Once the values of u are known at first
time level of time, process can be repeated for second time
level and so on. Due to symmetry of the problem results are
given for the region 0<x, y<0.5 and it is given in the table
(5.11.1a) and (5.11.1b) for y = 0.1 and plotted in figure
(5.11.1).

316
 Table (5.11.1(a))
Velocity distribution in rectangular vibrating membrane
through finite element method u ^^ at t = 0.01

y 0.0 0.10 0.20 0.30 0.40 0.50

X
0.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000

0.1 0.000000 0.095476 0.181606 0.249961 0.293847 0.308969

0.2 0.000000 0.181606 0.345437 0.475454 0.558930 0.587693

0.3 0.000000 0.249961 0.475454 0.654406 0.769300 0.808890

0.4 0.000000 0.293847 0.558930 0.769300 0.904367 0.950908

0.5 0.000000 0.308969 0.587693 0.808890 0.950908 0.901845

Table (5.11.1(b))
Velocity distribution in rectangular vibrating membrane
through finite element method u -^ at t = 0.03

y 0.0 0.10 0.20 0.30 0.40 0.50

X
0.0 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000

0.1 0.000000 0.095446 0.181550 0.249883 0.293756 0.308873

0.2 0.000000 0.181550 0.345330 0.475306 0.558757 0.587511

0.3 0.000000 0.249883 0.475306 0.654203 0.769062 0.808639

0.4 0.000000 0.293756 0.558757 0.769062 0.904087 0.950613

0.5 0.000000 0.308873 0.587511 0.808639 0.950613 0.999535

317
 0.350000
0300000
0250000
0200000
0150000
0.100000
0.050000
0.000000
00 0.1 0.2 03 0.4 0.5

Figure (5.11.1(a))
Velocity distribution in rectangular vibrating membrane through finite element method
(at y = 0.1 &t=0.03)

318
5.11 (A) DISCUSSION OF RESULTS :
The table (5.11.2(a)) represents the comparison
of finite element solution of vibrating membrane problem
with spline explicit -implicit as well as exact solution at
y = 0.1 and t = 0.03. The table (5.11.2(b)) and figure (5.11.2)
shows error analysis. From figure (5.11.2), it is clear that
spline explicit and implicit results are quite accurate than
the finite element results. Clearly, the spline results are
more truthful and reliable than finite element solutions.

Table (5.11.2(a))
Comparison of solution (at y = 0.1& t = 0.03)

X UFEM UEXT USE USI

0.0 0.000000 0.000000 0.000000 0.000000

0.1 0.095446 0.095483 0.095483 0.095483
0.2 0.181550 0.181619 0.181619 0.181619
0.3 0.249883 0.249978 0.249978 0.249978
0.4 0.293756 0.293866 0.293867 0.293866
0.5 0.308873 0.308989 0.308989 0.308989

Table (5.11.2(b))
Error analysis (at y = 0.1 & t = 0.03)

X UEXT-UFEM UEXT-USE UEXT-USI

0.0 0.000000 0.000000 0.000000
0.1 0.000037 0.000000 0.000000
0.2 0.000069 0.000000 0.000000
0.3 0.000095 0.000000 0.000000
0.4 0.000110 0.000001 0.000000
0.5 0.000116 0.000000 0.000000

319
 0.000140

0.000120

0.000100

K 0.000080 -•— UEXT-UFEM

o « — UEXT-USE
g 0.000060 UEXT-USI
0.000040

0.000020

0.000000
0.5

Figure (5.11.2)
Error analysis (at y = 0.1 & t = 0.03)

320
5.12 CONCLUSION :
Chapter 5 deals with finite element method.
The method is applied to linear partiail differential equation
of parabolic and hyperbolic type one and two space
variables. We have solved the same problems which are
discussed in chapter 3 with same parameter and compared
the solution with spline explicit, spline implicit as well as
exact solution. In all the problems, spline results give better
approximation than finite element solutions. Throughout
the work in this chapter, it is also clear that spline
collocation method gives truthful solution with compact
calculation than finite element method. When the time step
is increase, the calculation in finite element method
becomes quite lengthy and difficult. So it is seen that spline
solution are quite accurate and reliable.

321