Advances in Engineering Software
| Hi-index | 0.00 |
A geometric primitive for CAD implementation ispresented in this paper (Bèzier Neural Network BNN). Itis specifically designed to reproduce geometric shapeswith functional requirements such as aerodynamic andhydrodynamic profiles. This primitive can be useful whena known and well defined map between functionalrequirements and geometric data does not exist, and ithave to be deduced by a physical or numericalexperimental analysis. BNN gives rise to a typical CADrepresentation, a Bèzier curve, of a functional profile,once the functional parameters are supplied. In BNN thecapability of neural network to approximate very complexand non-linear function has been combined with thecapability of Bèzier functions to describe geometry, in aunique neural network. In this work BNN is used in therepresentation of aerodynamic profiles starting to theirtypical functional parameters: lift and drag coefficients,Reynolds number and angle of attack. BNN is tested inreproducing the wing profile of the 4-digit NACA series.The output of BNN is compared with the results of a fluid-dynamicanalysis performed by commercial software.