PhD in Modeling of dynamic systems using advanced scientific machine learning and hybrid modeling techniques

With the recent advances in computing hardware and the democratization of machine learning tools, the race to generate knowledge using machine learning algorithms has started. Thus, the scientific community is creating novel ways to enhance existing physics-based models with machine learning techniques. In the framework of dynamic systems, the challenges are on the rise, as machine learning techniques lag the stability and convergence criteria defined through years of research for classical computation methods.

In the framework of the research chair between Arts et Métiers institute of technology, and SKF Magnetics Mechatronics, a tight collaboration is mounted, aiming the advancement of knowledge in the field of scientific computing, and to leverage this knowledge in solving practical engineering dynamics problems. To cite an example, complex dynamical problems appear in magnetic bearings, as they require an extremely high control frequency, of the order of several kHz, while achieving a positional tolerance of few micrometers. Such a high frequency and a low tolerance require advanced control strategies, often in highly nonlinear systems.

The work aims to develop a modeling and simulation framework for the dynamic systems of interest, using advanced scientific machine learning methods, as well as hybrid modeling. The work will use, among others, neural ordinary differential equations, spectral networks, reservoir computing, model reduction techniques and possible combinations of the aforementioned methods. The applications tackle mainly the dynamic processes appearing in magnetic bearings, in the aim of achieving real-time control.  The tools will be mainly developed in Julia, Python and Matlab programming languages. The successful candidate should be knowledgeable in machine learning and model reduction techniques, programming tools and optimization algorithms. The work is expected to start by the beginning of September 2023.