Modeling and Numerical Methods for Uncertainty Quantification (MNMUQ 2019)
French-German Summer School / Ecole Thématique CNRS
September 2-6, 2019
IGeSA Center, rue de la Douane, Porquerolles Island, France
Jean-Marc Bourinet, Sigma Clermont, Clermont-Ferrand, France
Carsten Proppe, KIT, Karlsruhe, Germany
Scope and objective
Accounting for uncertainties is of utmost importance in many engineering systems and physical phenomena. Quantifying the impact of these uncertainties on responses of interest has become an active field of research over the past few years in many disciplines such as structural mechanics, material science, fluid dynamics, geophysics, electromagnetics, telecommunication networks, chemistry, ...
The main objective of this summer school is to address the important mathematical concepts of uncertainty quantification (UQ) and the related computational / algorithmic challenges arising in these engineering and science disciplines.
Participants will be given an overview of the most recent numerical techniques for uncertainty quantification and will be provided with examples of their use in various fields of application. This summer school will gather young researchers, leading scientists and engineers, and will offer this audience a perfect place to exchange ideas and develop cross-disciplinary discussions.
The program will specifically address:
- the Monte Carlo method and variants thereof for rare event probability estimation. The focus will be put on the most efficient techniques applied in the context of reliability analysis of systems with high safety levels (importance sampling, subset simulation and multilevel Monte Carlo),
- the construction of approximate models used as surrogates of costly-to-evaluate functions. The course will jointly present several techniques: sparse polynomial chaos, low-rank approximations and kernel-based methods (kriging and support vector machines). We will be specifically interested in models trained on sequentially enriched sets of data, with applications in UQ.
The courses will also cover the following topics:
- random processes / random fields,
- global sensitivity analysis,
- optimization under uncertainty,
- structural health monitoring,
- inverse methods and Bayesian updating.
This summer school is mostly aimed at young researchers (PhD students and postdocs) working on stochastic approaches. It also targets researchers and engineers interested in the development and implementation of methods in a UQ context.
As a prerequisite, the candidates are expected to have a basic to good knowledge of numerical linear algebra, numerical methods for PDEs, probability and statistics.
All courses will be given in English.
- Jean-Marc Bourinet, Sigma Clermont, France
- Mathilde Chevreuil, University of Nantes, France
- James-A. Goulet, Polytechnique Montréal, Canada
- Bertrand Iooss, EdF R&D Chatou, France
- Rodolphe Le Riche, Ecole des Mines de Saint-Etienne, France
- Iason Papaioannou, TU Munich, Germany
- Fabrice Poirion, ONERA Chatillon, France
- Carsten Proppe, KIT Karlsruhe, Germany
- Bojana Rosić, University of Twente, Netherlands
- Bruno Sudret, ETH Zurich, Switzerland