6 APPROXIMATE METHODS
This
ε^
is made equal to zero at i according to equation (1.11), which
gave the same result as central finite differences, for this particular
case.
METHOD OF MOMENTS
Equation (1.9) can be used to generate a wide variety of weighted
residual techniques such as the method of moments, the original
Galerkin method, collocation, subregions, etc.
The method of moments consists in taking moments of the error
function. For instance for a one-dimensional problem this implies
taking moments of ε with respect to the following function:
1,
x, x
2
, . . . (1.18)
These are the φ
(
functions of equation (1.8). Hence we can write a
weighting function such that,
χν
= β
1
ψ
1
+β
2
ψ
2
+
β
3
ψ
3
+
...
= β
1
1+β
2
χ +
β
3
χ
2
+
...
(L19)
The method is illustrated with the following example.
Example 1.2
Assume to have the same equation and boundary conditions as seen in
Example 1.1. The error function is
d
2
u
ε = —
T
+
u
+ x (a)
ax
z
We can propose an approximating function that satisfies the boun-
dary conditions and such as,
u
=
α
ι</>ι + a
2
02+ . . . = ajx(l -x) + a
2
x
2
(l -x) + ... (b)
Note that the φ
ί
,φ
2
, ··· functions satisfy the homogeneous
boundary conditions at x = 0 and x = 1. The a, coefficients are
unknown and are not directly associated with nodal values of the u
functions as we have not defined any nodes over the domain. The
error function for two a, coefficients results