PhD Scholarship in Data-driven modelling in fluid mechanics

The FLOW research group is a young, dynamic group working in the fields of thermodynamics, fluid mechanics, and data-driven modelling. At the Department of Engineering Technology (INDI) — Thermo and Fluid Dynamics (FLOW), there is now a vacancy for a PhD research position starting 1 October 2021.

At FLOW, we have a unique expertise in both physical and data-driven modelling of thermal-fluid systems. In this project, the focus will be on the data-driven modelling of fluid flows.

Many applications in engineering rely on mathematical models for the design, optimisation, control, and monitoring of systems and subsystems. When high-fidelity models are required, numerical (CFD) models are often the first choice. When simple (low-order) models are needed, then either the CFD models are reduced, or empirical models are used that are typically fine-tuned using experimental data. However, such empirical models are often overly simple and therefore cannot be used in many cases (often when unsteadiness, nonlinearity, and/or turbulence play a major role).

Recent advances in the nonlinear system identification and machine learning communities have resulted in the development of a number of tools that show great potential for the modelling of fluid flows. In particular, our group has pioneered the use of nonlinear data-driven state space models for fluid-dynamics applications, with proof-of-concepts such as oscillating cylinders and pitching aerofoils.

In this project, the objective is to compare data-driven modelling techniques (from system identification and machine learning) to achieve improved design and control of fluid-dynamics systems.

The candidate will first need to become proficient in the construction of reduced-order and data-driven models. Throughout the project, the candidate will also need to follow up recent advances in these fields.

The research will entail:

1) the development of data-driven models using system identification and machine learning techniques, starting from CFD data or experimental data available in the research group

2) the comparison of different modelling approaches,

3) the reduction of data-driven models, using techniques developed in our group, with the aim of making the models more compact and interpretable

Furthermore, the candidate will be expected to apply for external funding, to write, submit, and follow up on articles, and actively participate in scientific conferences.

The candidate will be expected to assist in the teaching of lab work and exercise sessions of undergraduate courses in mechanical engineering, as well as the supervision of undergraduate student projects.