An adaptive grid method and its application to CFD learning and prediction

A knowledge based engineering system, CFDLearner, which automatically learns results from a number of CFD simulations with a neural network is enhanced to automatically determine the optimum grid resolution where grid independence is found. Simulations are thereafter able to be run at a combination of the optimum grid resolution and lower grid resolutions to decrease the time spent on some simulations at a minor loss of accuracy. The system is explained and tested using single compartment fire simulations.