Variable Communication Rates in a Distributed Simulation

Posted by on Oct 21, 2009 in Distributed Heterogeneous Simulation, Eric A. Walters, Jason R. Wells, Maher Hasan, Oleg Wasynczuk, Power Systems, Publications | 0 comments

E. A. Walters, M. Hasan, C. E. Lucas, PC Krause and Associates, Inc; O. Wasynczuk, Purdue University; P. T. Lamm, U.S. Air Force Research Laboratory

As a result of the increased capabilities (surveillance, directed energy weapons, fuel efficiency, etc.) for both military and civilian aircraft, new architectures are being proposed for integrated power and propulsion systems. To analyze these new designs, large-scale multi-disciplinary (aerodynamic, electrical, mechanical, hydraulic, thermal, etc.) system simulations are required. To this end, distributed simulation has become a vital tool by enabling subsystem models developed using different commercial-off-the-shelf simulation programs to be interconnected to form an end-to-end system simulation that can execute orders of magnitude faster than comparatable single-model (non-distributed) implementations. Although distributed simulation has been successfully applied to numerous military aircraft systems, engineering expertise is required when selecting the fixed-rate communication intervals between subsystems to guarantee that the dynamics are adequately portrayed for all modes of operation. This engineering judgment has typically led to conservative selections wherein needless communications are performed, thereby hindering the overall simulation speed. In this paper, a new variable-communication-interval algorithm is established wherein the communication intervals are selected based upon user-specified error criteria. With this algorithm, the communication intervals for each model-to-model interface are selected independently and determined based upon the dynamics of the exchanged variables. The advantages of this technique include: reduced engineering time required to establish a distributed simulation by simply allowing error criteria to be specified, an increased assurance of system simulation accuracy, and increased simulation speeds by eliminating unnecessary communications. This technique is applied to an example system wherein it is shown that a five-fold increase in simulation speed can be achieved with the same accuracy when compared to the traditional fixed-rate approach, or a six-fold increase in accuracy can be achieved for the same simulation speed.

3rd International Energy Conversion Engineering Conference, August 15-18, 2005, San Francisco, CA.