Are you inspired by combining physics-based models with machine learning techniques?
Are you fascinated by fundamental flow physics?
Are you passionate about programming and high-performance computing?
Are you eager to collaborate with industrial partners?
We are looking for two motivated PhD candidates that, combining model-based (physics) and data-driven (machine-learning) approaches, will develop innovative, highly accurate and highly efficient solvers for rarefied gas flows.
Computational fluid dynamics is an essential enabler for science and for many outstanding societal challenges. Many key advanced and emerging technologies require unprecedented control of heat and mass transfer in flows, from continuum to highly-rarefied conditions, often in presence of electromagnetic fields, chemical reactions, and complex interactions with boundaries. Due to the non-equilibrium nature of rarefied flows and the pronounced influence of molecular effects, these transport processes are highly complex and occur in non-standard circumstances.
Contemporary understanding is currently too incomplete to support the development of emerging technologies. Computational modeling is extremely demanding and, in most situations, well beyond foreseeable computing capabilities.
In this project you will break ground on the way rarefied flows are modeled for emerging technologies, by developing innovative approaches that blend data-driven (machine learning) and model-driven (physics-based) methodologies. In this way, you will contribute incorporating the accuracy of computationally expensive atomistic models into macroscopic approaches, while simultaneously severely cutting back the computational cost.
You will work in a consortium consisting of university groups with complementary skills in fluid dynamics and statistical physics (Fluids and Flows, TU/e) and machine-learning techniques (AMLab, UvA), and commercial partners with a need for these new methodologies. ASML, leader in high-resolution lithography solutions, Flow Matters, specialized in consultancy/licensing of rarefied flow solutions, and Carbyon, developing direct-air-capturing systems, contribute with expertise, data and experiments and will be the prime validators and first users of the novel solutions.
The two PhD projects will specifically focus on:
- Development of accelerated Direct Simulation Monte-Carlo (DSMC) algorithms by learning from data.
- Development of fast Particle-in-Cell (PiC) algorithms combining data-driven approaches.