This can be done using q x jq qf r n1 1 j fn1j 538a j

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: X j fn;j (5.3.8c) j;1 : (5.3.8d) j =1 i The formula (5.3.8d) now has the form of (5.1.2a). The parameters j of (5.3.6a) and j of (5.3.8d) are independent of the problem and can be evaluated in advance of the computation. Thus, using (5.3.6a) 0 = Z 1 0 ; 0 16 d =1 Z ;; 1 Z d =1 2 0 Z1 0 Z 1 ; (; ; 1) 5 ; d= d = 12 : 2= 2 2 0 0 Continuing in this manner and using the results in (5.3.7a) yields 5 95 yn = yn;1 + h(1 + 1 r + 12 r2 + 3 r3 + 251 r4 + 288 r5 + : : : + 2 8 720 1 = 1 d= 1 k;1r k ; 1 )f n;1 : (5.3.9a) Additional results, given in Hairer et al. 12], Section III.1, are reproduced in Table 5.3.1. j0123 4 5 6 7 8 1 5 3 251 95 19087 5257 1070017 j 1 2 12 8 720 288 60480 17280 3628800 Table 5.3.1: Coe cients j for Adams-Bashforth methods ( 12], Section III.1). You may recall the second mean value theorem for integrals which states that if p( ) 2 C 0 a b] and q( ) is integrable and does not change sign on a b] then Z b a p( )q( )d = p( ) Examining (5.3.4b), reveals that ; Z b a q( )d 2 (a b): : = ( + 1) : : k!( + k ; 1) k does not change sign for 2 0 1]. Thus, we can use the second mean value theorem with (5.3.7b) and write the local error in the more explicit form Z1 ; d = k hk+1y(k+1)( ) k hk+1 y (k+1) ( ) dn = (;1) 2 (tn;1 tn): (5.3.9b) k 0 Thus, the error coe cient k is also known from, e.g., Table 5.3.1. Contrast this with discretization error formulas for Runge-Kutta methods that were extremely complex. Formulas and their local discretization errors for k = 1 2 3 4, follow. k = 1 : Euler's method yn = yn;1 + hfn;1 n = h y00 ( ) 2 17 (5.3.10a) (5.3.10b) k = 2 : Two-step Adams-Bashforth formula yn = yn;1 + h(fn;1 + rfn;1=2) or yn = yn;1 + h (3fn;1 ; fn;2) 2 5 = 12 h2y000( ): n k = 3 : Three-step Adams-Bashforth formula h yn = yn;1 + 12 (23fn;1 ; 16fn;2 + 5fn;3) n = 3 h3yiv ( ): 8 k = 4 : Four-step Adams-Bashforth formula h yn = yn;1 + 24 (55fn;1 ; 59fn;2 + 37fn;3 ; 9fn;4) n = 251 h4 yv ( ): 720 (5.3.11a) (5.3.11b) (5.3.12a) (5.3.12b) (5.3.13a) (5.3.13b) The coe cients of the Adams-Bashforth methods (5.3.8c) and the local error coe cient according to (5.3.9b) are repeated in Table 5.3.2 for k = 1 2 : : : 6 1]. k j=1 2 3 4 5 6 k 1 1 1 j 2 5 2 2j 3 -1 12 3 3 12 j 23 -16 5 8 251 4 24 j 55 -59 37 -9 720 95 5 720 j 1901 -2774 2616 -1274 251 288 6 1...
View Full Document

This document was uploaded on 03/16/2014 for the course CSCI 6820 at Rensselaer Polytechnic Institute.

Ask a homework question - tutors are online