2

Weighted Residual Methods

Fundamental equations Consider the problem governed by the dierential equation:

u = f in D.

(83)

The above dierential equation is solved by using the boundary conditions given as follows:
on Su
u=u
t Bu = t on St = S\St .

(84)
(85)

where the boundary S consists of Su and St . The boundary conditions given in eqs.(84) and (85) are called
rigid and natural boundary conditions, or Dirichlet and Neumann boundary conditions, respectively. In
most engineering problems, and B are dierential operators in the forms of = and B = n ,
where is another dierential operator and n is the normal vector on St .
For the Laplace problem, the governing equation is given by
2 u = f in D,

(86)

and the Neumann boundary condition on St is given as t = n u = t. Then the dierential operators
, B and are dened by = 2 , B = n and = , respectively.
Trial functions

of
Suppose that u is approximated by trial function u
(x) =
u

N J (x)
uJ

(87)

J=1

J are coecients and N J (x) are linearly independent functions called trial bases, test functions
where u
or shape functions.
does not satisfy the governing equation and the boundary
Weighted residual Since the trial function u
condition rigorously, the residuals dened in the following are nonzero
RD (x) =
RS (x) =

{
u(x) + f (x)} 6= 0 in D
u
(x) =
(x) t(x) 6= 0 on St .
u
6 0 on Su , or B u

(88)
(89)

to make zero the residuals weighted by the appropriate function w(x) as

The problem is to nd u
follows:

w(x) RD (x)dV +
w(x) RS (x)dS = 0
(90)
D

where w(x) is called the weight function. Eq.(90) is also called a weak form to the original dierentiral
equation (83), called a strong form, because the solution that satises eq.(90) does not always satisfy the
original dierential equation (83) and the boundary condition (85).
Substitution of eq.(87) into eq.(90) yields the following equation:

wI RD dV +
wI RS dS
D
S

=
wI (x) (N J (x)
uJ f (x))dV
D

(x))dS +
+
wI (x) (N J (x)
uJ u
wI (x) (BN J (x)
uJ t(x))dS
Su
St
{
}

J
=
wI (x) N J (x)dV +
wI (x)N J (x)dS +
wI (x) BN J (x)dS u
D
Su
St
|
{z
}
K IJ
{
}

wI (x) f (x)dV +
wI (x)
u(x)dS +
wI (x) t(x)dS
D
Su
St
|
{z
}
fI
=0

(91)
13

or
J = fI.
K IJ u

(92)

where M independent weight functions wI (I = 1, 2, . . . , M ) are chosen. This equation can be solved
J . As shown in the sequel, there are several residual methods depending on the selection of weight
for u
functions.
satisfying the boundary condition u
=u
on Su . In such a
It is not dicult to nd trial functions u
case, no residual exists on Su and the integrals on Su disappear in eq.(91). In the following sections, it
satisfy u
=u
on Su .
is assumed that trial functions u

2.1

Point collocation method

In the point collocation method, we select at least the same number of collocation points as the unknown
J such that the residual is zero at the selected points.
parameters and determine the parameters u
The collocation method is equivalent to the residual method in which the weight function is chosen as
wI = 1(x xI )

(93)

where 1 denotes the unit matrix and (x) is the Dirac delta function satisfying the equations

(x xI )R(x)dV = R(xI ), for xI D

(94)

or

(x xI )R(x)dS = R(xI ),

for xI St

(95)

St

In this case, the matrix K IJ and the vector f I in eq.(92) can be expressed by
K IJ = N J (xI ) or BN J (xI ), f I = f (xI ) or t(xI )

2.2

(96)

Least squares method

In the least squares method, the sum of the squares of the residuals dened by

2
I=
RD dV +
R2S dS
D

(97)

St

is minimized.
The necessary condition to minimize I is
I
I
u
yielding the system of equations:

RD
D

or

= 0, I = 1, 2, . . .

RD
dV +
I
u

RS
St

RS
dS = 0
I
u

J
N I (x) N J (x)dV +
BN I (x) BN J (x)dS u
D
St
|
{z
}
K IJ
[
]

I
I

N (x) f (x)dV +
BN (x) t(x)dS = 0
D
St
{z
}
|
fI

(98)

(99)

(100)

As shown in the above equation, the least squares method is considered as a weighted residual method
with the weight functions N I (x) in the domain D and BN I (x) on the boundary St .

14

2.3

Classical Galerkin method

I are used
If only the domain integral in eq.(91) is taken into account and the shape functions N I (x)w
I
are arbitrary coecients, then we have
as the weight functions wI , where w

{
{
}
}

M
J
I
J
I

N (x)N (x)dV u
N (x)f (x)dV
=0
D
J=1 |
I=1
{z
}
{z
}
| D
K IJ
fI

2.4

(101)

Galerkin method

In the classical Galerkin method, the boundary condition is not considered. Therefore, the classical
Galerkin method is not convenient for practical use. Here the Galerkin method taking account of the
boundary condition is formulated.
Consider the weighted residual as follows:

(x) t(x))dS = 0.
w(x)(
u(x) f (x))dV +
w(x)(B u
(102)
D

St

satises the boundary

Here we assume that w is a weighted function satisfying w = 0 on Su and u
=u
on Su .
condition u
Since = , the rst integral in eq.(102) is written as

w(x)
u(x)dV =
w(x)
u(x)dV
D
D

=
w(x)n
u(x)dS
w(x)
u(x)dV
St
D

(x)dS
=
w(x)B u
w(x)
u(x)dV
(103)
St

where the divergence theorem is used. Note that in the above equation, the integral on Su vanishes due
to w = 0 on Su . Then eq.(102) becomes

w(x)
u(x)dV
w(x)f (x)dV
w(x)t(x)dS = 0.
(104)
D

St

This equation has the weak form which involves the Neumann boundary condition on St , which can be
=u
and the constraint condition w = 0 on Su .
solved under the Dirichlet boundary condition u
is expressed by
Now it is assumed that the trial function u
(x) =
u

N J (x)
uJ + N 0 (x)
u

(105)

J=1

where N J (x) and N 0 (x) are the shape functions equal to zero and one on the boundary Su , respectively.
Also the weight function w is selected as
I
w(x) = N I (x)w

(106)

and w = 0 on Su
where N I (x) is the same shape function used in eq.(105). Then the conditions u = u
are satised.
Substitution of eqs.(105) and (106) into eq.(104) yields

J
I
N I (x) N J (x)dV u
w

D
I=1
J=1 |
{z
}

KIJ

(
)

I
I
0
I

N (x)f (x)dV
N (x) N (x)dV u
N (x)t(x)dS = 0.

D
D
St
|
{z
}
fI

15

(107)

Example 2.1 Consider the solution u satisfying the equation

d2 u
+ f0 = 0 0 x l
dx2

(108)

subjected to the boundary conditions

u = 0 at x = 0, du/dx = f0 l at x = l.

(109)

The exact solution for the above problem is u(x) = f0 x(2l x/2) and the value at x = l/2, u(l/2),
is obtained as u(l/2) = f0 l/2(2l x/4) = f0 7l2 /8.
Now the boundary value problem mentioned above is solved by means of both the classical Galerkin
method and the Galerkin method assuming that the solution is approximated by u
(x) = N 1 (x)
u1 ,
1
where N (x) = x(2l x). Note that the approximated function is slightly dierent from the exact
solution.
Classical Galerkin method For N 1 (x) = x(2l x) and N 1 (x) = d2 N 1 (x)/dx2 = 2, eq.(101)
becomes
l
l
x(2l x)(2)dx
u1 +
x(2l x)f0 dx = 0
(110)
0

From eq.(110), we obtained u

= f0 /2. The solution u
(x) is obtained by u
(x) = f0 x(2l x)/2.
2

At x = l/2, u
(l/2) = 3/8f0 l , which is much dierent from the exact solution u(l/2) = 7/8f0 l2 .
Galerkin method From eq.(107), the following equation based on the Galerkin method is formulated.
(
)
l
l
dN 1 dN 1
1
1
1

dx
u
N (x)f0 dx + N (1)f0 l = 0
(111)
0
0 dx dx
1

Substitution of N 1 (x) = x(2l x) into the above equation yields

(
)
l
l
2(l x)2(l x)dx
u1
(2xl x2 )f0 dx + l(2l l)f0 l = 0,
0

(112)

from which u
1 = 5f0 /4. Hence, u
(x) = 5f0 x(2l x)/4. Also at x = l/2, u
(l/2) = 15/16f0 l2 .
Thus we can conclude that the Galerkin method with the boundary condition gives a value
closer to the exact solution than the classical Galerkin method, in which no boundary condition
is considered.
Example 2.2 Solve the governing equation
d2 u
+1=0 0x1
dx2

(113)

2
uJ , where N J (x) = xJ .
by using the Galerkin method with the trial function u(x) = J=1 N J (x)
The boundary conditions are given by u(0) = 0 and du/dx(1) = 20.
The Galerkin method yields the following equation for the above mentioned problem:

M 1

dN I dN J

dx u
J
w
I
dx
dx

I=1
J=1 | 0
{z
}

KIJ

N I (x)fdx

dN I dN 0

dx
u + N I (1)t = 0
dx dx

{z
}

(114)

fI

where f = 1, u
= 0, t = 20, M = 2, N (x) = x and N 2 (x) = x2 . The calculation of the matrix
KIJ and the vector fI leads to the equations
[
]{ 1 } {
}
1
1
u

20 21
=
(115)
1 4/3
u
2
20 13
1

which can be solved for u

J (J = 1, 2).
16

Problem 2.1

Consider the governing equation

d2 u
u=0 0x1
dx2

(116)

subjected to the boundary condition u(0) = 0 and du/dx(1) = 20. Solve the problem
2
J
by using the Galerkin method with the trial function u(x) =
uJ , where
J=1 N (x)
J
J
N (x) = x .
Problem 2.2 Derive the weak form based on the Galerkin method for the multi-dimensional
Laplace problem, for which the governing equation is given by
u = f in D

(117)

and the boundary conditions are as follows.

u=u
on Su
t n u = t on St = S\Su
Problem 2.3

Solve the same problem as Problem 1.9.1 using the weak form.

17

(118)
(119)