In the present paper, an evolution strategy has been combined with an extended free form deformation technique. This method has - especially for a design optimisation of fluid dynamical systems - various advantages. For many complex systems the number of parameters can be drastically reduced using free form deformation. Especially, the required additional process of grid generation, which can take several days for complex shapes, can be integrated into the optimization framework. Since the mesh is deformed together with the shape, no re-meshing is needed. For evolutionary optimisation a high degree of automation and parallelization is needed which can easily be realized by running FLUENT via journal files in a cluster environment and via the exchange of computational grids by modifying case files. This method of adapting the representation allows an increase of flexibility during the optimisation process which is integrated in and governed by the process itself. As shown in the test scenario the performance of the design can be increased by introducing new control points which increase the dimensionality of the design space while allowing a higher degree of locality. However, there are also some limitations when using free form deformations with CFD which have to be considered carefully. With respect to the computational grid large modifications of the coordinates of the control points implies consequently large deformations of the cells. Therefore, it has to be guaranteed that the CFD mesh is still valid after deformation. The ratios of cell width to cell height have to be observed and changes in the order of the control points have to be prevented. In the case of very large deformations the design has to be re-meshed and the optimisation has to be re-started based on the updated mesh. Improving the adaptive representation will be one of the goals for future research focus, in particular the question of when to increase the variability of the design during the optimization process requires additional analysis.


    Zugriff

    Zugriff über TIB

    Verfügbarkeit in meiner Bibliothek prüfen


    Exportieren, teilen und zitieren



    Titel :

    Evolutionary design optimisation of complex systems integrating FLUENT for parallel flow evaluation


    Weitere Titelangaben:

    Evolutionäre Design-Optimierung komplexer Systeme mit Integration der Software FLUENT für die Bewertung von Parallelströmungen


    Beteiligte:


    Erscheinungsdatum :

    2005


    Format / Umfang :

    11 Seiten, 8 Bilder, 14 Quellen


    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Print


    Sprache :

    Englisch




    Fluent releases Version 5.0 FLUENT software

    Emerald Group Publishing | 1999



    Multi-Objective Robust Design Optimisation Using Hierarchical Asynchronous Parallel Asynchronous Evolutionary Algorithms

    Lee, DongSeop / Gonzalez, Luise Felipe / Srinivas, Karkenahalli et al. | AIAA | 2007



    Design Optimisation using Distributed Evolutionary Methods

    Doorly, D. J. / Peiro, J. / Spooner, S. et al. | British Library Conference Proceedings | 1999