High-Performance Computing for Computational Fluid Dynamics for Free

Anders Dalsgaard Melander: High-Performance Computing for Computational Fluid Dynamics for Free Surface Flows

When designing ships, offshore wind turbines or other coastal and open-water constructions, the impact of water flows and waves is important. When designing an offshore wind turbine, you for example want to be sure that the construction can withstand the impact of waves. When designing ships, many are interested in increasing fuel efficiency through the design of the hull. Being able to simulate such problems is therefore of high interest. The simulation of these things is the field of Computational Fluid Dynamics (CFD). Here the Navier-Stokes equations are used as the mathematical model to describe the physical phenomena’s. However, it is well known that these equations are very computationally expensive, meaning the time required to do simulations are very long. It is therefore of interest to be able to reduce the computational cost by use of modern numerical techniques as well as implementation on modern many-core hardware.

The goal of this project is to research and develop a state-of-the-art massively parallel Navier-Stokes solver for high-fidelity free surface flow modeling using spectral element methods. Free surface Navier-Stokes solvers have high computational cost, therefore compromises are often made in terms of solution accuracy for a lower computational run time. The project will address the high cost and the necessary scientific challenges in utilizing high-performance computing on state-of-the-art many-core hardware to lower cost through careful algorithmic implementations. Moreover, by utilizing high-order numerical methods for the proposed solver, the research will focus on answering how to lower the high simulation time of Computational Fluid Dynamics. High-order numerical methods were shown to achieve better performance for many types of fluid computations compared to low-order methods based on error versus computational cost. However, less research have focused on how to optimally map the methods to many-core hardware.

This kind of solver will allow for highly accurate simulation of kinematics, turbulence, etc., in the vicinity of fixed, forced and free bodies submerged in the fluid. Moreover, this will also improve the speed of simulations, allowing for more testing to be done when designing wind turbines, ships etc., therefore increasing the quality of these products, e.g., more fuel-efficient ships.

PhD project

By: Anders Dalsgaard Melander

Section: Scientific Computing

Principal supervisor: Allan P. Engsig-Karup

Co-supervisor: Sven Karlsson

Project titleHigh-Performance Computing for Computational Fluid Dynamics for Free Surface Flows

Term: 01/12/2020 → 30/11/2023


Anders Dalsgaard Melander
PhD student
DTU Compute


Allan Peter Engsig-Karup
Associate professor
DTU Compute
+45 45 25 30 73


Sven Karlsson
Associate Professor
DTU Compute
+45 45 25 37 54