Two/three-dimensional flow-field parameters, such as species concentration, temperature and velocity are key to understand the physical and chemical behaviors of flows. The measurement data not only indicate the performance of the energy systems in many industrial sectors, such as aviation and manufacturing, but also provide experimental validation of the CFD models for R&D of the energy systems.

Laser absorption tomography (LAT) provides the unique capability of imaging thermophysical parameters of reactive flows, e.g., temperature, species concentration and velocity. LAT is implemented in a manner analogous to x-ray tomography, with the difference that, wavelength-specified incident laser beams are used rather than x-rays to acquire the absorption measurements, i.e., projection data in tomography. In industrial practice, the projection data can only be sparsely obtained from limited projection views, resulting in rank-deficient ill-posed inverse problem and thus leading to errors in image reconstructions.

This project will develop data assimilation for high-fidelity flow-field reconstruction using LAT. Data assimilation (DA) algorithms seek to solve the equations governing fluid motion subject to databased constraints. Instead of training an end-to-end neural network using simulated data, DA algorithms integrate ground-true physics, into the network, so call physics-informed neural network. Therefore, high-fidelity flow-field images are expected to be reconstructed by DA-assisted LAT. The student who joins our group will learn the fundamentals of laser absorption tomography, computational and data-driven solutions for inverse problems. The student will have a high chance of working with renowned international researchers and industrial collaborators.

Primary objectives:

1. Develop LAT model to integrate DA algorithms

2. Develop physical-informed neural networks for LAT and numerically test the network.

3. Design lab-scale experiment to validate the developed neural networks and LAT model.

4. Interpret flow-field behaviors from high-fidelity image reconstructions.

5. Improve communication and writing skills via conference presentations and journal publications.

Required skills:

1. Matlab/Python/Tensor-flow coding (mandatory)

2. Machine learning algorithms and coding (mandatory)

3. Tomography and inverse problems

4. Flow-field mechanics and dynamics

5. Optical/electronic experimental skills.

Please note that this advert will close as soon as a suitable candidate is found.