Stuart D. Perry

Evaluation of Parallel Algorithms for the Real-Time Simulation of a Stiff Nonlinear System

Yen-Cook, Miranker-Liniger, and Stability Region Placement (SRP) methods are evaluated on the basis of their applicability to real-time simulation using the VU68K-HDG parallel processing system as the target machine. Because of stability-computation time trade-offs, the SRP method, which allows selection of step size independent of the system eigenvalue, is shown to be the only one of the three methods applicable to the real-time simulation of a particular stiff, nonlinear, third-order test system.

Parallel Simulation Using Stability Region Placement Methods

Using the VU68K-HDG, a 68000-based parallel processing computer, a scaled Stability Region Placement (SRP) method is applied to a stiff, nonlinear, third-order differential equation. An optimization technique is used on the scaled SRP integration coefficients and, using an emulation of the scaled system on a minicomputer for verification produces an rms state error characteristic that compares favorably with the rms data from a real arithmetic simulation. The advantage of the SRP method over conventional simulation methods in choosing the step size is maintained during optimization. The combination of the VU68KHDG hardware and the SRP design method provides a capability for the real-time simulation of nonlinear, stiff, dynamic systems.