Post Doctorant Mécanique H/F

Sujet de recherche : 

Making fast and accurate predictions becomes compulsory in sciences and engineering when many scenarios must be analyzed in order to evaluate risks, optimize designs or propagate uncertainty. To achieve the required computational performances model order reduction enabled almost real-time responses. Such a reduced model enables operation under stringent real-time constraints and, more particularly, evaluating numerous scenarios in probabilistic settings. However, many times, the models are not accurate enough. In these circumstances, a gap between simulated values and measurements is present. Two main procedures address the just referred scenarios with poor predictability capabilities. The first relies on the use of physics-informed learning, the other alternative consists of making a physics-based solution augmentation by adding the model of the gap that can be learnt from the use of any regression technology