An Intelligent Agent Architecture for Concurrent CFD Feature Extraction

CFD simulations are advancing to correctly simulate highly complex fluid flow problems that can require weeks on expensive computing clusters. These simulations can generate terabytes of data and pose a severe challenge to a researcher analyzing the data. Presented here is a solution to drastically reduce researcher post-processing time by extracting fluid flow features concurrent with a running CFD simulation using intelligent software agents. The software agents are designed to work inside the CAFÌE1 concept and operate efficiently on high performance computing clusters. Three types of agents are given and their belief tuples defined. A simulation of a blunt-fin is run showing convergence of the horseshoe fin line to its final spatial location at 540 iterations, or 60% of solution convergence. The agent architecture correctly selects between two vortex feature extraction algorithms and correctly identifies the expected probabilities of core lines throughout solution convergence.