A NEURAL NETWORK BASED APPROACH TO COMPUTATIONALLY INTENSE DISTRIBUTED AEROSPACE SIMULATIONS

摘要

Extremely complex dynamic aerospace systems can be represented by distributed simulation via a realtime architecture. Although inter-component communication channels exist to facilitate high bandwidth data transfers, computationally intensive models compromise the overall simulation performance. Modelling techniques such as Finite Element (FE) analysis and Computational Fluid Dynamics (CFD) require large computational resources, and depending on the complexity of the modelled system, require long periods of time to converge to a solution. In this paper, Radial Basis Function (RBF) networks are applied to the CFD analysis of an aircraft wing in a transonic airflow. It is desired that the analysis form a part of a distributed aircraft simulation environment. The paper investigates to what extent these systems can be approximated by neural networks.