User Project Details

SuperResTest

Advance deconvolution methods for turbulence and reacting fields using deep-learning techniques

Reacting Flows

Universität Stuttgart

Institut für Technische Verbrennung

The combustion of CO2-neutral pulverised biomass (PBC) is a way to mitigate global warming and fulfil constantly rising energy demands. While direct numerical simulation (DNS) is still prohibitive, advanced modelling tools such as large eddy simulation (LES) can be utilised in the ongoing research on PBC. Closures for the unresolved sub-grid scales are required in LES. Recently, deep neural networks have been utilised to perform super-resolution tasks on images that are generating high-resolution images from low-resolution/noisy images. Such a mapping is mathematically equivalent to the deconvolution of filtered LES fields to high-resolution DNS-like fields which can then be used to provide closure models for LES. The proposed project aims to move a step forward in this direction by the development and application of deep learning as LES sub-grid closures in particle-laden reacting flows, in particular pulverized biomass combustion.