Global Error Estimation with Runge--Kutta Methods (Q3698181)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 3923989
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Global Error Estimation with Runge--Kutta Methods |
scientific article; zbMATH DE number 3923989 |
Statements
Global Error Estimation with Runge--Kutta Methods (English)
0 references
1984
0 references
global error estimation
0 references
Runge-Kutta methods
0 references
low-degree
0 references
interpolation
0 references
numerical results
0 references
0.97854877
0 references
0.97683537
0 references
0.9567872
0 references
0.95604485
0 references
0.95256484
0 references
0.94258595
0 references
0.9203704
0 references
An analysis of global error (GE) estimation for Runge-Kutta solutions of ordinary differential equations of the initial value type (MP) \(y'(x)=f(y(x))\) with \(y(x_ 0)\) known, is presented. The basic technique is that of Zadunaisky in which the GE is computed from a numerical solution of a neighbouring problem (NP) \(y'_ h(x)=f_ h(y_ h(x))=f(y(x))+d_ h(x)\), where \(y_ h(x_ 0)\), \(d_ h(x)=Q(x)- f(P(x))\) with \(Q(x)=P'(x)\) or \(P(x)=\int^{x}_{x_ 0}Q(t)dt\), related to the main problem (MP) by some method of interpolation. In order to avoid high degree interpolation, the interval of integrating is split into subintervals each containing \(m+1\) points. The \(y_ i\), \(i=0,...,N\), are thus piecewise interpolated with polynomials of degree m and \(d_ h\) is defined as a piecewise function. In the case of the f interpolant one has (*) \(\max \| E_ n\| \leq Ah^ p\), if \(m\leq p- 1\), \(\max \| E_ n\| \leq Ah^{\min [m+1,2p]}\), \(m\geq p\), \(A=const\), where p is the order of the RK process, \(E_ n=\epsilon_{hn}-\epsilon_ n\), \(\epsilon_ n=y_ n-y(x_ n)\) with the numerical \(y_ n\) and true \(y(x_ n)\) solutions of (MP) at \(x=x_ n\). More than one term in the GE may be estimated from (*), and besides the case, avoiding numerical differentiation, permits smaller m for the same order of GE estimation. It is shown that RK formulae which admit valid GE estimation using low- degree interpolation can be developed, thus leading to more accurate and computationally convenient algorithms than was hitherto expected. Some special RK processes up to order 4 are presented together with numerical results.
0 references