CFD · Research · 2020 — 2024 · UNSW
Parallel turbulent Lattice Boltzmann solver
A high-fidelity turbulent LBM solver, built from formulation to parallel implementation.
Solver output — figure to be replaced with a result from the thesis.
Context
High-fidelity simulation of turbulent flow demands numerical methods that stay accurate and stable at scale. The Lattice Boltzmann method offers a route to massively parallel performance — if the formulation, boundary treatment, and refinement strategy are built with the target hardware in mind from the start.
My role
I was the sole developer — from mathematical formulation through validated, parallelised implementation, as the core of my PhD in mechanical engineering.
Approach
- Lattice Boltzmann formulation suited to turbulent flow regimes.
- Parallelisation across CPU and GPU architectures.
- Adaptive grid refinement to concentrate resolution where the flow demands it.
- Systematic validation against canonical benchmark flows.
Outcome
A working research solver underpinning the doctoral thesis and journal publications, and the foundation for later coupled-physics work.
- LBM
- GPU parallelisation
- Adaptive refinement
- Turbulence
- Fortran / C++