A procedure with stepsize control for solving \(n\) one-dimensional IVPs (Q955284)

From MaRDI portal
scientific article
Language Label Description Also known as
English
A procedure with stepsize control for solving \(n\) one-dimensional IVPs
scientific article

    Statements

    A procedure with stepsize control for solving \(n\) one-dimensional IVPs (English)
    0 references
    0 references
    0 references
    19 November 2008
    0 references
    Let \(w(x;h)\) be an approximate solution to a first-order ordinary differential equation (i) \(y'(x)= f(x,y)\), \(y(x_0)= y_0\), generated by a stepwise algorithm where \(h\) is the stepsize. An automated program is devised to approximate \(y(x)\) by \(w(x;h)\) by control of the parametor \(h\): (1) Given a desired number of signifigant figures the parameter \(h\) is chosen so that \(w(x;h)\) and \(w(x;h/2)\) agree to the same number of signifigant figures. (2) The values of \(w(x;h)\) and \(w(x;h/2)\) are computed repeatedly using the CESTAC stochastic arithmetic method until they arrive at the value \((w(x;h)- (x;j2)=0)\). Validity of (1) is justified by asymptotic expansion of \(w(x;h)\), \(h\to\infty\). Use of (2) removes roundoff errors. Numerical results are presented for five test problems employing second and fourth-order, Runge-Kutta routines and a fourth-order predictor-corrector routine. The method extends to first-order systems.
    0 references
    initial-value problem
    0 references
    multi-step method
    0 references
    single-step method
    0 references
    CADNA library
    0 references
    stepsize control
    0 references
    Runge-Kutta method
    0 references
    CESTAC stochastic arithmetic method
    0 references
    roundoff errors
    0 references
    numerical results
    0 references
    predictor-corrector routine
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references