Playing With Particles

No, this is not an article about particle physics. It is about modeling fluids using particles.

Part of my day job is to work on numerical methods using computer simulations of fluid flow problems. This field of research is sometimes referred to as computational fluid dynamics or CFD.

The methods I'm working on are finite element methods (FEM). Although in this particular class of methods many different formulations exist, typically they use a grid or mesh to discretize a computational domain. This results in a set of finite elements which are used within a FEM formulation. An example can be seen below:

While I typically work with this approach, the past weekend I dove into another class of problems that step away from the concept of grids and meshes. These so-called meshless methods approach the discrete problem from a different angle. In smoothed particle hydrodynamics (SPH) instead, the fluid domain is discretized by a set of particles that interact with each other based on a set of relations. The nice thing is that the tedious task of mesh generation can be skipped. One simply has to create a set of particles and distribute the particles inside the computational domain and keep track of where they are during the simulation.

Typical SPH simulation. Image source: SPHinXsys Documentation

Although I didn't have the intention to work on such methods, I stumbled upon a nice open-source SPH library and decided to play around with it. The library is called SPHinXsys, and is mainly developed by a group of researchers at the Technical University of Munich in Germany. If you are interested, you can simply download the API from GitHub and start playing with it yourself!

Although the video is not made by me, here's an example of a simulation created using SPHinXsys:

This video shows a block of fluid hitting an elastic wall. The fluid has a free surface and is able to move around freely within the square domain. This simulation is still quite coarse. I.e., the number of particles is small and you can still easily distinguish individual particles moving around the domain.

SPH is a totally different approach to what I am used to. Now that I've been playing around with just the simulations, I'm excited to learn more about the method itself and how it operates under the hood!