Optimal Selection of Rotor Sections Using CFD and Neural Networks

This work aims to develop a framework for the optimisation of various aspects of rotor blades. The proposed method employs CFD combined with artificial neural networks and optimisation methods based on genetic algorithms. To demonstrate this approach, two examples have been used, one is the optimal selection of 4-digit NACA aerofoils for rotor sections and the other is the optimisation of linear blade twist for rotors in hover. For each case, a specific objective function was created and the metamodel was subsequently used to evaluate this objective function for optimisation. The obtained results agree with real world design examples and theoretical predictions. For the selected cases, the artificial neural network was found to perform adequately and the results were sensitive to the data used for training. The genetic algorithm was also effective in identifying a set of near-optimal parameters. The main CPU cost was associated with the population of the database necessary for the metamodel