Acceleration of Runge-Kutta integration schemes.pdfVIP

ACCELERATION OF RUNGE-KUTTA INTEGRATION SCHEMES PHAILAUNG PHOHOMSIRI AND FIRDAUS E. UDWADIA Received 24 November 2003 A simple accelerated third-order Runge-Kutta-type, fixed time step, integration scheme that uses just two function evaluations per step is developed. Because of the lower number of function evaluations, the scheme proposed herein has a lower computational cost than the standard third-order Runge-Kutta scheme while maintaining the same order of local accuracy. Numerical examples illustrating the computational efficiency and accuracy are presented and the actual speedup when the accelerated algorithm is implemented is also provided. 1. Introduction The behavior of many systems in nature and society is described by sets of differen- tial equations that are nonlinear in character. Obtaining closed-form solutions for these equations is therefore, generally speaking, impossible, especially for systems where the state vector has a high dimension. Thus numerical methods for solving such equations become very important. One of the most common methods for numerically integrating sets of nonlinear dif- ferential equations is the Runge-Kutta (RK) family of single-step methods. The RK in- tegrator of order n has a local error over a time step h of O(hn+1 ). For the third- and fourth-order (RK3 and RK4, respectively) methods, the number of function evaluations required over a single time step are three and four, respectively. In this paper, we pro- pose a simple scheme that uses an RK-type third-order procedure [ 1, 2] but which uti- lizes just two function evaluations at each step. The scheme has the same order of local accuracy as the standard RK3 method and hence appears to provide improvement in computational efficiency without any significant effect on computational accuracy. The drawback of the scheme, however, is that it is, in reality, a two-step scheme and therefore needs to be initially “started” using an accurate o