Solving Fluid Flow Problems Using A Real-codedGenetic Algorithm With Uniform Refinement
Raed I. Bourisli, Deborah A. Kaminski · WIT transactions on engineering sciences · 2004
A Genetic Algorithm with real-coded chromosomes is used to solve a potential flow problem in a contraction channel. A new uniform, successive-refinement strategy is presented wherein the algorithm starts with a coarse solution and works its way throughmulti-levelsof refinementsto arbitrarysolution resolution.Ateach level of refinement, a line-by-line sweep-through method is used to obtain incremental solutions for various parts of the flow domain, thus reducing the number of GA constraints. A finite difference grid was used as the objective function to evaluate the fitness of the chromosomes. Results are very promising for five levels of refinements totalling more than 12,000 nodes. Small population sizes were found to be most effective, often yielding solutions in less than 20 minutes. The mutation operator was also found to be of paramount importance for the speedy evolution of the algorithm.