A new methodology which uses a neural network to evaluate the quality of an airfoil by interpreting the distribution of flow parameters on a quasi-3D surface has been developed. The training data was generated by Computational Fluid Dynamics (CFD) analysis of randomly perturbed airfoil. Heuristic rules based on Fluid boundary layer analysis were used for evaluating the flow parameters on the airfoil. Two different heuristic measures of quality are used 1) Diffusion which is a measure of losses due to local pressure changes in the flow distribution, and 2) Deviation which is a measure of losses due to non-uniformity of acceleration of the flow. These two parameters were used as the outputs of the neural network.
展开▼