Category Archives: Uncategorized

Snapshot: Supersonic turbulence simulations with FLUXO

We show a supersonic turbulence simulation in a periodic box fueled by perpetual kinetic energy injection on the largest three modes. Such setups are subject of active research in astrophysics and are believed to play an important role in the dynamics of galactic clouds and star formation, e.g., [1,2,3]. An initially uniform density distribution (left plot in the video) is gradually driven to a turbulent state till an average velocity of Mach 3.5 is reached (about t=0.19). Locally, speeds can spike above Mach 35 (bottom right plot in the video). The simulation was computed on FLUXO (https://github.com/project-fluxo/fluxo) with a robust invariant domain preserving Discontinuous Galerkin scheme (DG) [4] solving the compressible, quasi-isothermal ($\gamma =1.001$) Euler equations on a uniform grid with $128^3$ degrees-of-freedom.

Smooth blending with a monotone finite volume scheme stabilizes the high-order DG at strongly shocked regions (filaments of high density concentration) displayed by the blending parameter plot (top right in the video). Blue encodes regions of 100% high order DG while red dots indicate focused blending with FV. Hypersonic turbulence regimes are very challenging for any numerical scheme considering the stark gradients in density, bubbles of near-vacuum and the extremely high speeds involved in such simulations.

[1] https://academic.oup.com/mnras/article/436/2/1245/1126116
[2] https://adsabs.harvard.edu/full/1981MNRAS.194..809L]
[3] https://academic.oup.com/mnras/article/480/3/3916/5060766
[4] Rueda-Ramírez, A. M., Pazner, W., & Gassner, G. J. (2022). Subcell limiting strategies for discontinuous Galerkin spectral element methods. arXiv preprint arXiv:2202.00576.

Talk on 2021-01-21: Julia for adaptive high-order multi-physics simulations

On Wednesday, 27th January 2021, 3:15pm CET, Michael Schlottke-Lakemper will give an online talk on

Julia for adaptive high-order simulations

To obtain the Zoom link, please contact the organizers via the official meeting announcement.

Abstract

Julia has been touted as a programming language especially well-suited for numerical analysis and scientific computing. However, while its prevalence is steadily increasing, it has not yet seen widespread adoption in the computational science or high-performance computing communities. One of the hurdles is a (perceived) lack of real-world examples that show how Julia can be used to conduct numerical simulations and what its advantages and drawbacks are for scientific applications.
To remediate this, in this talk we discuss the development of a purely hyperbolic method for self-gravitating gas dynamics within our Julia-based open source simulation framework Trixi.jl (https://github.com/trixi-framework/Trixi.jl). In this approach, we reformulate the elliptic gravity problem into a hyperbolic diffusion problem, which is solved in pseudotime using the same explicit high-order discontinuous Galerkin method we use for the flow solution. A key benefit is that in the resulting multi-physics simulation problem, we can reuse existing hyperbolic solvers while retaining advanced features such as non-conforming and solution-adaptive meshes.
Next to presenting numerical results, we will critically examine our experience with building a multi-physics simulation framework with Julia. We will discuss its strengths and weaknesses as a programming language for research software engineering, including an assessment of how Julia’s claimed benefits hold up against scientific reality, and give a live demonstration of Julia and Trixi.jl in action.

To make the shown examples reproducible by the audience, the Jupyter notebook used for the live demonstration is available at https://github.com/trixi-framework/talk-2021-julia-adaptive-multi-physics-simulations. It can be either run from a local Julia/Jupyter installation or in the cloud via Binder (without having to install Julia locally).

Snapshot: 2D PyOpenGL real-time simulation of the forward-facing step test case

WENO methods refers to a class of nonlinear finite volume or finite difference methods which are well suited to approximate solutions of hyperbolic conservation laws with high order accuracy in smooth regions and essentially non-oscillatory transition for solution discontinuities. In the following video we used a 5th order accurate WENO-FD scheme with immersed boundary conditions.

Here we solve the forward-facing step test case in real-time using Python3/PyOpenGL/GLFW with an Intel Corporation HD Graphics 620 (rev 02). As one can see, it is also possible to add pressure peaks to the simulation by simply clicking with the mouse on the desired field.