Mechanics & Industry
Volume 18, Number 5, 2017
|Number of page(s)||7|
|Published online||01 November 2017|
β-NTF reduction and fast kriging simulation of optimal engine configurations
2 Université de Bordeaux, I2M CNRS, UMR 5295, 33607 Bordeaux, France
3 UVHC, LAMIH CNRS UMR 8201, Le Mont Houy, 59313 Valenciennes cedex 9, France
4 UVHC, ENSIAME, Le Mont Houy, 59313 Valenciennes cedex 9, France
5 AKIRA Technologies, ZA Saint Frédéric rue de la Galupe, 64100 Bayonne, France
* e-mail: firstname.lastname@example.org
Accepted: 18 August 2017
In an optimization process, models are applied to simulate different design behaviors in order to determine the most suitable one. However, this requires the use of a structured methodology to correctly explore the design space and truly converge to the best solution. It is therefore necessary to test and validate the optimal design. For engines, two ways are essentially used: building and testing a real cylinder, or simulating the new design with Computational-Fluid-Dynamics (CFD) models. These two techniques are both expensive and time consuming. An alternative way is proposed to test new designs with a fast simulation based on a kriging method. The exploration of the design space is based on 27 cylinder configurations and the results of their CFD models. It converged to an optimal design depending on the objective function. A kriging method was used to interpolate the behavior of the optimal design just found. In this paper we present the β-NTF model reduction (to define the data set used by the kriging method) and the principle of the kriging technique. We then briefly discuss the results. The results underline the method's advantages despite the small gap between the expected results and those for kriging.
Key words: kriging / fast simulation / β-NTF reduction / design space / 2-stroke engine optimization
© AFM, EDP Sciences 2017
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.