αλέξανδρος κοντογιάννης

Hopkinson Lab (ISO-40)

CUED Trumpington Street

Cambridge CB2 1PZ UK

ak2239@cam.ac.uk

I am EPSRC research fellow in fluid dynamics and applied mathematics at Cambridge University Engineering Department.

I am currently working on Bayesian inverse problems in fluid dynamics for magnetic resonance velocimetry. In the past, I have worked on aerodynamic modelling, fluid-structure interaction, simulations of ice accretion on aircraft wings, and aerodynamic shape optimisation.

My main research concerns the formulation of new machine learning methods that automatically reconstruct corrupted flowfields. These methods learn the most probable simulation that corresponds to the corrupted flowfield, and, at the same time, infer unknown quantities (e.g. pressure) that are either hard or impossible to measure otherwise.

In a nutshell, I develop algorithms that learn the most probable physical model (aka digital/physical twin) of the flowfield dynamics from data.

selected publications

2023

Physics-informed compressed sensing for PC-MRI: an inverse Navier-Stokes problem

A. Kontogiannis, and M. P. Juniper

IEEE Transactions on Image Processing, 2023

Abs arXiv HTML

We formulate a physics-informed compressed sensing (PICS) method for the reconstruction of velocity fields from noisy and sparse phase-contrast magnetic resonance signals. The method solves an inverse Navier-Stokes boundary value problem, which permits us to jointly reconstruct and segment the velocity field, and at the same time infer hidden quantities such as the hydrodynamic pressure and the wall shear stress. Using a Bayesian framework, we regularize the problem by introducing a priori information about the unknown parameters in the form of Gaussian random fields. This prior information is updated using the Navier-Stokes problem, an energy-based segmentation functional, and by requiring that the reconstruction is consistent with the k -space signals. We create an algorithm that solves this inverse problem, and test it for noisy and sparse k -space signals of the flow through a converging nozzle. We find that the method is capable of reconstructing and segmenting the velocity fields from sparsely-sampled (15% k -space coverage), low ( ∼10 ) signal-to-noise ratio (SNR) signals, and that the reconstructed velocity field compares well with that derived from fully-sampled (100% k -space coverage) high ( >40 ) SNR signals of the same flow.

2022

Joint reconstruction and segmentation of noisy velocity images as an inverse Navier–Stokes problem

A. Kontogiannis, S. V. Elgersma, A. J. Sederman, and M. P. Juniper

Journal of Fluid Mechanics, 2022

Abs arXiv HTML

We formulate and solve a generalized inverse Navier–Stokes problem for the joint velocity field reconstruction and boundary segmentation of noisy flow velocity images. To regularize the problem, we use a Bayesian framework with Gaussian random fields. This allows us to estimate the uncertainties of the unknowns by approximating their posterior covariance with a quasi-Newton method. We first test the method for synthetic noisy images of two-dimensional (2-D) flows and observe that the method successfully reconstructs and segments the noisy synthetic images with a signal-to-noise ratio (SNR) of three. Then we conduct a magnetic resonance velocimetry (MRV) experiment to acquire images of an axisymmetric flow for low (≃6) and high (>30) SNRs. We show that the method is capable of reconstructing and segmenting the low SNR images, producing noiseless velocity fields and a smooth segmentation, with negligible errors compared with the high SNR images. This amounts to a reduction of the total scanning time by a factor of 27. At the same time, the method provides additional knowledge about the physics of the flow (e.g. pressure) and addresses the shortcomings of MRV (i.e. low spatial resolution and partial volume effects) that otherwise hinder the accurate estimation of wall shear stresses. Although the implementation of the method is restricted to 2-D steady planar and axisymmetric flows, the formulation applies immediately to three-dimensional (3-D) steady flows and naturally extends to 3-D periodic and unsteady flows.

2021

Inverse problems in magnetic resonance velocimetry: shape, forcing and boundary condition inference

A. Kontogiannis, and M. P. Juniper

ASME, Fluids Engineering Division (Publication) FEDSM, Oct 2021

Abs HTML PDF

We derive and implement an algorithm that takes noisy magnetic resonance velocimetry (MRV) images of Stokes flow and infers the velocity field, the most likely position of the boundary, the inlet and outlet boundary conditions, and any body forces. We do this by minimizing a discrepancy norm of the velocity fields between the MRV experiment and the Stokes problem, and at the same time we obtain a filtered (denoised) version of the original MRV image. We describe two possible approaches to regularize the inverse problem, using either a variational technique, or Gaussian random fields. We test the algorithm for flows governed by a Poisson or a Stokes problem, using both real and synthetic MRV measurements. We find that the algorithm is capable of reconstructing the shape of the domain from artificial images with a low signal-To-noise ratio.
Adjoint state of nonlinear vortex-lattice method for aerodynamic design and control

A. Kontogiannis, and E. Laurendeau

AIAA Journal, Feb 2021

Abs HTML

The adjoint state of a nonlinear vortex-lattice method, coupled with sectional polars obtained with an infinite swept-wing (2.5D) Reynolds-averaged Navier–Stokes (RANS) solver, is presented as an interactive method for preliminary aerodynamic design and control of moderate-to-high aspect ratio wings in subsonic and transonic flows. First, the nonlinear vortex-lattice method is described, augmented with a Prandtl–Glauert compressibility correction and regularized so that poststall solutions can be recovered. The linearization of the coupled system of equations is formulated using a Newton method, and the adjoint state is obtained for lift, drag, and inverse design functionals. Numerical results show that the aerodynamic model approximates well the forces predicted by three-dimensional (3D) RANS simulations for the Common Research Model wing, and that the adjoint-state gradients agree well with finite difference tests for a plethora of test cases. More interesting examples are shown, ranging from the optimal unstalling of wings to the assimilation of 3D RANS data into the present model, as well as its ability to extrapolate.