The University of Luxembourg (UL) and Luxembourg Institute of Socio-Economic Research (LISER) invite applications for a DRIVEN PhD Fellow (Doctoral Candidate) position (m/f) as part of the DRIVEN Doctoral Training Unit (https://driven.uni.lu), consisting of 19 doctoral candidates. DRIVEN is funded by the FNR PRIDE funding instrument https://www.fnr.lu/funding-instruments/pride/.
PRIDE PhD Fellow Ref: R-AGR-3440-13-C
You will be working as part of DRIVEN Doctoral Training Unit (DTU) funded by the FNR PRIDE scheme. The Computational and Data DRIVEN Science DTU will train a cohort of 19 Doctoral Candidates who will develop data-driven modelling approaches common to a number of applications strategic to the Luxembourgish Research Area and Luxembourg’s Smart Specialisation Strategies. DRIVEN will build a bridge between state-of-the-art data driven modelling approaches and particular application domains, including Computational Physics and Engineering Sciences, Computational Biology and Life Sciences, and Computational Behavioural and Social Sciences.
In the research direction “Data-driven identification of governing equations in continuum mechanics” within the field of Computational Mechanics, the Doctoral Candidate will address how the governing model equations can be systematically identified from observations of the system’s spatio-temporal activity. Computational data-driven methods for model identification in continuum mechanics based on (potentially incomplete, non-uniform and noisy) measurement data are expected to deliver the mathematical model of the observed problem as set of parameterised (nonlinear) PDEs. Here, identification of constitutive relations for smart materials in multiphysics energy harvesting problems is a possible focus direction. The PhD project will explore the opportunities and limitations of selectively learning from an existing library of canonical phenomenological constitutive models. Similar to an expert system, the most influential terms (linear or nonlinear) in a pool of governing equations and parameters are selected by regression analysis on available data. Based on an objective function on a selected output quantity the constitutive behaviour is described as a linear combination of phenomenological basis models. Evaluation and classification of the measurement data is followed by conceptualising PDE model-updating given the availability of new and possibly relocated data. The PDE-based model obtained for a specific objective is finally automatically fed into a high-performance predictive analysis tool (FEniCS, https://fenicsproject.org) providing high-level abstraction layers for mathematical problem descriptions.
Your lead supervisor will be Prof. Andreas Zilian. Further supervision will be provided by Dr. Jack Hale and Prof. Stéphane Bordas.
Your primary tasks as a DRIVEN fellow are to: