This content is not included in
your SAE MOBILUS subscription, or you are not logged in.
Surrogate Based Optimization for Multidisciplinary Design
Technical Paper
2011-01-2507
ISSN: 0148-7191, e-ISSN: 2688-3627
This content contains downloadable datasets
Annotation ability available
Sector:
Language:
English
Abstract
When designing components of an aircraft, such as wing or nozzle, the physics involved cover several areas of engineering, such as weights, aerodynamics or thermal engineering. They often individually lead to different solutions. This makes multidisciplinary design difficult. The first obstacle is the communication between different dedicated solvers and the definition of a common objective. Once this difficulty is solved, designers have to face costly simulations. Moreover, the classical design procedure involves several iterations between simulation means and technical teams. Because of the strict delay in industrial context, very few adjustments to initial design are allowed. Based on recent work in operational research and numerical optimization, this article describes a methodology and best practice that allows automation of such processes, better efficiency in the search of an optimal design and shorter design time.
This methodology, applied successfully on several topics, is already well known and described in the literature [1]. Design space is sampled, physical simulations are done using sample values, a surrogate model is built on samples, and is optimized towards best design, meeting requirements of all disciplines. Some choices can make this methodology reliable and applicable to industrial context as any other production mean: samples respect a Latin Hypercube (LHS) pattern, optimized along a discrepancy criterion. Then, numerical simulations are run in parallel to build a gaussian process model. A global optimization search of possible minima locations is performed through this costless model. Then optimal designs are evaluated through several measures, involving the standard deviation of this specific surrogate model (kriging), and the relevance of the values with regards to the expected physical phenomena, to assess the accuracy of the model. This could lead to a refinement of the model or to the choice of a unique and satisfactory optimum.
Through that process, an optimal design is found automatically, thus reducing production time. Supporting, and not replacing, the work of designer teams, it does improve efficiency of designs in various domains of aeronautic: air systems, turbo machinery, aerodynamic design… Using numerical optimization, it allows considering more complex phenomena in fields where engineers have decennia of experience and feedback, and helps in acquiring quick experience in breakout technologies.
An application involving turbo machinery cooling is proposed to illustrate this methodology on an aircraft part with complex phenomena.
Authors
Citation
Boussouf, L., "Surrogate Based Optimization for Multidisciplinary Design," SAE Technical Paper 2011-01-2507, 2011, https://doi.org/10.4271/2011-01-2507.Data Sets - Support Documents
Title | Description | Download |
---|---|---|
Unnamed Dataset 1 |
Also In
References
- Iooss, B. Boussouf, L. Marrel, A. Feuillard, V. “Numerical study of algorithms for metamodel construction and validation,” Safety, Reliability and Risk Analysis: Theory, Methods and Applications 2009 2135 2141
- Marrel, A. “Mise en œuvre et utilisation du métamodèle processus gaussien pour l'analyse de sensibilité de modèles numériques,” 2008
- Fang, Kai-Tai Li, Runze Sudjianto, Agus « Design and modeling for computer experiments” Computer science and data analysis series 2006
- Kirkpatrick, S. Gelatt, C. D., Jr. Vecchi, M. P. “Optimization by Simulated Annealing” 1983 Science 4598
- Sacks, Jerome Welch, William J. Mitchell, Toby J. Wynn, Henry P. Design and Analysis of Computer Experiments Statistical Science 4 4 409 435 1989
- Laurenceau, J. “Surfaces de réponse par krigeage pour l’optimisation de formes aérodynamiques,” PhD thesis 2008
- Zitzler, Eckart Laumanns, Marco Thiele, Lothar “SPEA2: Improving the Strength Pareto Evolutionary Algorithm”
- Varlet, F. Millot, G. “Aerothermal Coupling with Integrated Exchange Surfaces,” SAE Technical Paper 2011-01-2508 2011 10.4271/2011-01-2508
- Csató, L. Opper, M. “Sparse representation for Gaussian process models” Advances in Neural Information Processing Systems 13
- Samuelides, M. “Apprentissage statistique” Eyrolles editions