Lombardi G. - Department of Aerospace Engineering, University of Pisa
Vannucci S. - Department of Aerospace Engineering, University of Pisa
Ciampa A. - Istituto Nazionale di Fisica Nucleare, Pisa
Davini M. - Department of Physics, University of Pisa |
The aerodynamic of a keel of an America’s Cup yacht is quite complex: in fact, the geometry is characterised by a “flapped wing”, with a large slender body (the bulb) and two winglets placed at the tip. Therefore, a systematic analysis taking into account the effects of all the parameters appears to be difficult.
In order to tackle this problem resort was made to a direct numerical optimization technique; this approach was attractive, as it made it possible to address the problem systematically, and offered flexibility in the choice of the design variables.
In the analysis through direct numerical optimization, an aerodynamic code is coupled in a loop with an optimization routine, so as to automatically manage the values of the design variables – typically concerning geometry modifications – with the aim of minimizing a given scalar quantity (objective function). This approach is extremely flexible, and capable of meeting multi-disciplinary requirements. The choice of the design variables is of fundamental importance, as they define the set of solutions within which the optimal one is sought. Finally, the design problem formulation requires the definition of the constraint functions. All these aspects will be highlighted in the paper.
Because the flow complexity, the aerodynamic solutions require a RANS solver, and FLUENT 6.2.16 was used. For he parametric CAD model, used for the optimisation loop, CATIA V5R16 was used, and an automatic procedure for the grid generation was developed by Gambit 2.3.16. The procedure was driven by the multidisciplinary optimisation code modeFRONTIER 3.2.
The objective function was a weighted combination of the drag in different conditions, with fixed total lift values. The effect of the centre of gravity position and the manoeuvring performance reduction with the bulb length were taken into account, and several constraints were considered.
The procedure requires very high computational capabilities, in order to obtain the optimised configuration in a reasonable time. An AMD Opteron 280 Dual Core processor cluster has been used; as far as the cluster fabric is concerned an high performance Myrinet network protocol has been the choice. In the paper a brief performance analysis of the complete system will be provided. |
