A Heuristic Approach to Dynamic Design Space Definition for NLF Aerofoil Optimisation
Richard McRoberts, Juliana M. Early, Lee Cameron, Stephen W. T. Spence · 29th AIAA Applied Aerodynamics Conference · 2011
A hybrid genetic algorithm has been developed and is benchmarked against a gradientbased and standard genetic algorithm (GA) to test the ability of each to generate natural laminar flow (NLF) aerofoil profiles. The performance characteristics of the GA and hybrid GA are assessed on a number of analytical test functions to illustrate the efficiencies gained from including a local search technique. A “meme pool” constructed from a variety of local searches is important for the algorithm to be robust as different local searches have inherent biases; however a trade-off between efficiency and robustness is seen to arise. For the engineering design problem it is shown that the hybridised method offers improvements in the total runtime of the algorithm and also generates superior designs than compared with a traditional GA. The capability to include problem specific knowledge in a hybrid algorithm is demonstrated along with the enhanced efficiencies that can be achieved.