Post-doctoral position in multidisciplinary design optimization (ULB-BATir, Belgium)

Environment

The post-doctoral position will be accomplished within the BATir (Building, Architecture & Town planning) department of the Brussels School of Engineering/´Ecole polytechnique de Bruxelles, at the Universit´e Libre de Bruxelles (http://batir.ulb.ac.be ). Duration of the position: 18 months, starting in April 2012.

Profile

The applicant should hold a PhD in computational mechanics. Additional competences in numerical optimization, and structural finite elements, as well as programming skills (in MATLAB, PYTHON, C++, or any equivalent programming anguage) will be highly appreciated.

Contact

Please send by email: 1. a detailed CV (with the complete list of publications), 2. a summary of your PhD thesis, and
3. a cover letter to: Prof. Rajan FILOMENO COELHO (rfilomen@ulb.ac.be). The closure date for application
is set to the 31st of March 2012.

Description

Numerical optimization in civil and mechanical engineering has emerged as a powerful method to provide lightweight structures accounting for several criteria (lower cost and environmental impact, higher safety, etc.), the usual framework consisting in coupling a simulation (typically: a finite element analysis) to an optimization code. Nevertheless, the computational effort required to assess the performances of a single design can be expensive, which hinders the ability to address large-scale industrial designs. To alleviate this issue, surrogate-based or metamodel-assisted optimization techniques have been proposed to replace the high-fidelity model by an interpolation or regression model [1]. However, most surrogates (e.g. moving least squares, kriging, artificial neural networks, support vector machines) proposed in the literature focus on continuous inputs, i.e. they cannot model discrete/integer numerical values, or technological parameters like the choice of a material (e.g. steel, aluminum).

Consequently, the objective of this postdoctoral position is to investigate a domain still largely uncovered in the engineering optimization community (despite its theoretical and industrial interest), namely the treatment of mixed variables in surrogate-based optimization. Preliminary research studies in surrogate-based optimization have already een done in our department [2, 3], and novel ideas have been recently proposed to tackle mixed variables through a modified moving least square method. Starting from this background and his/her previous experience, the candidate will develop an original metamodel capable of approximating response surfaces depending on mixed variables. Inarticular, the applications (in civil or mechanical engineering) investigated during his/her PhD could fruitfully be used as academic or industrial examples to validate the proposed approach.

References

[1] J W Bandler, S Koziel, and K Madsen. Editorial–surrogate modeling and space mapping for engineering optimization. Optimization and Engineering, 9:307–310, 2008.
[2] P Breitkopf and R Filomeno Coelho, editors. Multidisciplinary Design Optimization in Computational Mechanics. ISTE/John Wiley & Sons, Chippenham, UK, April 2010. 1 volume, 549 pages.
[3] R Filomeno Coelho, J Lebon, and Ph Bouillard. Hierarchical stochastic metamodels based on moving least squares and polynomial chaos expansion – Application to themultiobjective reliability-based optimization of 3D truss structures. Structural and Multidisciplinary Optimization, 43(5):707–729, 2011.