See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/266139178

Solution of Weighted Residual Problems by using Galerkin's Method

Article  in  Indian Journal of Science and Technology · March 2014

DOI: 10.17485/ijst/2014/v7i3S/48657

CITATIONS READS
6 4,729

1 author:

Prof.(Dr).Mrs.J.EVANGELINE CICELIA
Bharath University
20 PUBLICATIONS   8 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Prof.(Dr).Mrs.J.EVANGELINE CICELIA on 27 September 2014.

The user has requested enhancement of the downloaded file.

 ISSN (Print): 0974-6846
Indian Journal of Science and Technology, Vol 7(3S), 52–54, March 2014 ISSN (Online) : 0974-5645

Solution of Weighted Residual Problems

by using Galerkin’s Method
J. Evangeline Cicelia
Department of Mathematics, Bharath Institute of Science & Technology,
Bharath University, Selaiyur, Chennai - 600073, Tamil Nadu, India evangelinecicelia@gmail.com

Abstract
For simple one-dimensional problems, direct stiffness method is limited. PMPE (Principle of Minimum Potential Energy)
is limited to potential problems and FEM (Finite Element Method) can be applied to many engineering problems that are
governed by a differential equation. Hence, there are systematic approaches, needed to generate FE equations, such as
Weighted Residual method, Energy method and ordinary differential equation (second-order or fourth-order). Weighted
residual method can be solved, in particular using Galerkin’s method.

Keywords: Residual, Weighted, Stiffness, FEM, FE, Galerkin’s Method, Differential Equations

1.  Introduction the trial solution is as close to the exact solution as

The theory behind finite difference solutions is intuitive
and these solutions are easy to understand and straight-
forward to implement. However, the finite difference
technique is most appropriate for problems where the
computational domain is regularly shaped and the nodes
are positioned in a regular pattern. The finite element
technique is an alternative numerical method. The finite
element technique is much less intuitive than the finite
difference technique to most engineers, and the underly-
ing theory is more involved.
However, once the theory has been developed, the
finite element technique can be applied as easily to irregu-
lar geometries with an arbitrary arrangement of nodes as
it can be to a regular geometry with a regular pattern of
nodes. The simple approach of Galerkin’s method is dis-
feasible.
3. Consider the boundary value problem y'' = f(y, y', x) =
f(x, y, y')  (i)
with boundary conditions y(a) = A and y(b) = B
4. Write the differential equation R = y–'' – f(x, y–, y–') (ii)
where, R is the residual of the equation when R=0 for
the exact solution y(x) only
which will satisfy the boundary conditions.
5. Consider the trial solutiony– (x) = C  1Ø1(x) +
C2Ø2(x) (iii)
when y– (a) = A and – y (b) = B
6. Differentiate trial solution (iii) twice and substitute in (ii)
7. To find C1, C2, ... ... weight the residual by trial func-
tions Ø1(x), Ø2(x), ... ... ... and set the integrals to zero.
∴ ∫ Rf1 (x )dx = 0, ∫ Rf2 (x )dx = 0,  
c c

cussed as follows: 8. Solve simultaneous equations to find the unknowns,

1.1  Procedure
1. Guess a solution to the differential equation, which
satisfies the boundary conditions.
2. The trial solution will contain certain parameters,
which can be adjusted to minimize the errors so that
*Author for correspondence
substitute these unknowns in the appropriate solution,
y–(x) is obtained.
9. By using this procedure carefully the following steps,
many weighted residual problems can be solved. Here,
we discussed a sample solved problem, which gives the
approximate solution of the problem and comparison
with exact solution.
 J. Evangeline Cicelia

2.  Problem Step 7: WKT ∫ R.f(x )dx = 0

1
Use Galerkin’s methods to solve the boundary value i.e.   ∫ ( 4Cx 2 + (1 − 4C ) x − 8C ) ( x (1 − x )) dx = 0
problem y'' – y + x = 0, 0 ≤ x ≤1, y(0) = 0 and y(1) = 0. 0
1
Compare the approximate solution with exact solution. ∫ ( 4Cx + 4Cx + x − 8c ) ( x − x ) dx = 0
2 2

Solution 0
1
∫ (4Cx − 4Cx − 4Cx + 4Cx + x − x − 8Cx + 8Cx ) dx = 0
3 4 2 3 2 3 2
Step 1: Th
 e given differential equation is y'' – y + x = 0, 0
0 ≤ x ≤1.
Integrating we get,
Step 2: Consider the residual R, 4Cx 4 4Cx 5 4Cx 3 4Cx 4 x 3 x 4 8Cx 2 8Cx 3
− − + + − − + =0
R = y ′′ − y + x  (i) 4 5 3 4 3 4 2 3
4Cx 5 4Cx 3 x 3 x 3
Step 3: To find the trial solution, which satisfies the 2Cx 4 − + + − − 4Cx 2 = 0
5 3 3 4
boundary conditions, derive lagrangian polynomial
which passes throw the points, Substitute the limits 0 to 1, we get
1 4C 4C 1 1
x :0 1 1 2C − + + − − 4C = 0
x :0 2 1 5 3 3 4
2
y :0 C 0 4 4 1 1
y :0 C 0 C  2 − + − 4 +  −  = 0
 5 3   3 4
4 4 1 1
Step 4: The resulting polynomial is C  − + − 2 +  −  = 0
y(x ) = 4Cx(1 − x )  (ii)  5 3   3 4
−12 + 20 − 30   4 − 3 
= C1 .x(1 − x )    Where C1 = 4C C + =0
 15   12 
Since the trial solution is y(x ) = C1f1 (x ) + C2f2 (x )    −22  1
C = − 
Where f(x ) = x(1 − x )  15   12 
15
C=
Step 5: Differentiate twice (ii) w.r.t x, we get 12(22)
d y(x ) d 5
= [ 4Cx(1 − x )] C=
dx dx 88

= 4C [(1)(1 − x ) + x(−1)] Step 8: Substitute this C value in equation (ii), the

approximate solution is
= 4C [1 − x − x ]
5
y ′(x ) = 4C(1 − 2 x ) y ( x ) = 4 x(1 − x )
88
Again differentiating this, we get 5
y ( x ) = x(1 − x )
22
y ′′ (x ) = 4C(−2)
Step 9: To find exact solution
y ′ ′ (x ) = −8C  (iii)
Let the given differential equation be,
Step 6: substitute the values of y ′′ (x ) and y– (x) in equa- y ′′ − y + x = 0
tion (i), we get
y ′′ − y = − x
R = −8C − 4Cx(1 − x ) + x
i.e. (D 2 − 1) y = − x
2
= −8C − 4Cx + 4Cx + x Auxiliary equation is,
2
R = 4Cx + (1 − 4C )x − 8c m2 − 1 = 0

Vol 7 (3S) | March 2014 | www.indjst.org Indian Journal of Science and Technology 53
 Solution of Weighted Residual Problems by using Galerkin’s Method

m2 = 1 m = ±1, m = 1, m = −1 1
C2 =
The roots are real and different
(e − e −1 )
∴ The complementary function is,
Step 10: The exact solution:
C. F = Ae m1x + Be m2 x
Substitute the values of C1 and C2 in equation (iv),
i.e. y = C1e x + C2 e − x we get,
Particular integral −1 1
y= ex + e−x + x
P .I = 2
1 ( e − e −1
) ( e − e −1
)
(D − 1)(− x )
 (e x − e − x ) 
1 y=x− −1 
= 2
−(1 − D )
(− x ) = (1 − D 2 )−1 (x )  (e − e ) 
This is the required exact solution.
(
= 1 + D 2 + D 4 +  (x ) )
P .I = x
3.  Conclusion
General solution: C.F + P.I
y = C1e x + C2 e − x + x  (iv) This procedure of solving weighted residual method is
found to be much simpler than the other methods. Also,
Since y(0) = 0 and y(1) = 0 it gives an approximate solution and can be compared
0 = C1e 0 + C2 e 0 + 0 ⇒ C1 + C2 = 0 ⇒ C2 = −C1 with exact solution.
0 = C1e1 + C2 e −1 + 1 ⇒ C1e + (−C1 )e −1 + 1 = 0
4. Bibliography
C
⇒ 0 = C1e − 1 + 1
e 1. Boyd JP. Chebyshev and Fourier spectral methods.
Dover;1999.
⇒ −1 = C1 (e − e −1 )
2. Franklin J. Advanced mechanics and general relativity.
−1 Cambridge University Press; 2010.
⇒ C1 =
(e − e −1 ) 3. Zienkiewicz OC, Taylor RL. Finite element method. 2000;
 −1  1, The Basis.
1
C2 = −C1 = −  −1 
= 4. Cicelia J. Mathematicals method, thermal engineering. S&S
 (e − e )  (e − e )
−1
Publication; 2013.

54 Vol 7 (3S) | March 2014 | www.indjst.org Indian Journal of Science and Technology

View publication stats