Unformatted text preview: 0 j Z E= k Z 0 Let L ( )d + h 0 Evaluating (9.2.4) at the collocation points Z X
J i
k 15 =1 y ( )b + h E:
0 ik k i (9.2.6c) If the errors E , j = 1 2 : : : J , and E are neglected, we see that the collocation
solution is obtained from a J stage implicit RungeKutta method. Thus, letting Y(x)
denote the approximate solution, we have
j Y(x ) = Y(x 1) + h
i i; X
J bk (9.2.7a) ak) k = 1 2 : : : J: (9.2.7b) i =1 k where k = Y ( ) = f (x 1 + h Y(x 1) + h
0 ik ik i; k i i; X
J i
j =1 kj k ik ij Solution continuity, required for rstorder BVPs, y(x ) = y(x+) i = 1 2 ::: N ;1 ; i i is automatically satis ed.
Solutions at any point x 2 x 1 x ] are de ned by the interpolation polynomial (9.2.4)
and (9.2.3a) as
i; i X
J Y(x) = Y(x 1) + h
i; i
k =1 Y( )
0 Z ik 0 L ( )d : (9.2.7c) k Some common choices of the collocation points are:
1. GaussLegendre points. The points , k = 1 2 : : : J , are the roots of the Legendre
polynomial of degree J mapped to the interval (0 1). The roots of the Legendre
polynomial are normally prescribed on (;1 1). This may be done by the linear
transformation = (1 + )=2, 2 ;1 1]. For all J , 1 > 0 and < 1 thus,
s...
View
Full Document
 Spring '14
 JosephE.Flaherty
 Numerical Analysis, The Land, Collocation, Ly, collocation points

Click to edit the document details