Thursday 23 Nov 2017: Multi-Disciplinary Surrogate-Assisted Optimization Activities at Cenaero
Dr. Charlotte Beauthier - Senior Research Engineer at Cenaero
Harrison 170 15:30-16:30
Cenaero, located in Gosselies (Belgium), is a private non-profit applied research center providing to companies involved in a technology innovation process numerical simulation methods and tools to invent and design more competitive products. Internationally recognized, Cenaero is mainly active in the aerospace (in particular turbomachinery), process engineering, energy and building sectors. Cenaero operates a supercomputing infrastructure with 14,000 computing cores, and experimental facilities in composite manufacturing and prototyping. Cenaero’s researchers provide expertise in multidisciplinary simulation, design and optimization in the fields of mechanics (fluid, structure, thermal and acoustics), manufacturing of metallic and composite structures as well as in analysis of in-service behavior of complex systems and life prediction.
Minamo is the design space exploration and optimization platform developed by Cenaero. It implements an online surrogate-assisted optimization framework. Smart adaptive search algorithms allow to efficiently handle high-dimensional design spaces as well as highly constrained optimization problems involving high-fidelity simulations. This surrogate-based paradigm, coupled to a dynamic steering of the search process, allows to rapidly explore the design space and to optimize products in order to make better and enlightened innovative decisions. It brings benefits at different stages of a development process, from the early concept definition to the final design phase, and speeds up the development cycle of complex products by reducing computational costs related to high-fidelity simulations. The data analysis tools, such as analysis of variance or self-organizing maps, help to identify key factors and trades, to discover patterns and correlations within the data and to better understand the impact of antagonistic goals on the search process, offering thorough understanding of the conception space.
The presentation will focus on some important features of Minamo :
- Exploitation of different kind of surrogates (Tuned RBF, improvement of Kriging, multi-fidelity Non-Intrusive POD surrogates, selection and aggregation of surrogate models)
- Mono and multi-objective surrogate-based evolutionary algorithms (online approaches)
- Management of highly constrained problems : multi-point infill criteria based on Probabilistic SVM
- Parameter control in evolutionary algorithms
- Uncertainty propagation and reliability
Some applications of these techniques will be illustrated on industrial applications.