Loading...
Embedding periodic box direct numerical simulations in two-fluid large eddy simulations
Date
2025
Abstract
Multiphase turbulent flows involve complex multi-scale energy transfer mechanisms that remain poorly understood. This thesis presents a novel multi-scale approach for simulating multiphase turbulence by coupling Periodic Box (PB) Direct Numerical Simulations (DNSs) with a Largef Eddy Simulation (LES) in real time. An in-house Lattice Boltzmann (LB) code optimized for GPUs is developed to achieve this.
Initially, this method is tested on single-phase turbulence. PB DNS boxes are coupled to each lattice node in the LES, with characteristic strain rates matched between the two simulations at the appropriate scale. The PB DNSs are then run until a statistically steady state is reached and eddy viscosity information is extracted and used for Sub-Grid Scale (SGS) modelling in the LES. A database is used to minimize redundant calculations. The components are validated withvarious test cases, and the multi-scale model is implemented with one-way and two-way coupling. The results show the characteristic strain rate to be an effective representation of SGS turbulence in isotropic conditions. The PB DNS-coupled LES results are shown to match well with full DNS solution.
For multiphase flows, two Euler-Euler (Two-Fluid) model implementations on the LB method are proposed: one with a pressure-free LB and a pressure Poisson equation (more accurate but computationally expensive), and another solving a mixture equation and a dispersed phase momentum equation using an incompressible LB with a phase-intensive formulation (less accurate but computationally cheap). Both approaches are validated and shown to be stable for turbulent flows. The multi-scale model is then applied to emulsions using the new Poisson-based TF model with LES and a pseudopotential LB for PB DNS. The dispersed phase volume fraction is also matched between LES and PB DNS. The data from PB DNS is analysed to obtain insights into SGS behaviour in emulsions. The coupled model predicted the turbulence spectra and estimates the droplet size distribution accurately, with strong agreement with full DNS simulations.
Supervisor
Van den Akker, Harry E. A.
Description
Publisher
University of Limerick
Citation
Files
Funding code
Funding Information
Sustainable Development Goals
External Link
Type
Thesis
Rights
http://creativecommons.org/licenses/by-nc/4.0/