Combining computational fluid dynamics and magnetic resonance imaging data using lattice Boltzmann based topology optimisation

This thesis presents the combination of magnetic resonance imaging (MRI) measurements and computational fluid dynamics (CFD) to reduce statistical measurement noise and identify objects and finer structures in the MRI data. Using a lattice Boltzmann based topology optimisation approach, the method allows those solutions that best match the measured flow field but satisfy the macroscopic conservation laws of fluid flow, here mass and momentum conservation. This combination is formulated as a distributed control problem that minimises the distance between measured and simulated flow field, the latter being the solution of a parametrised Boltzmann equation with Bhatnagar-Gross-Krook collision operator, where the controls represent the porosity distributed in the domain. The problem is solved with an adjoint lattice Boltzmann method using the open source software OpenLB.