A Simple Particle Advection Example

In my last blog, I introduced temporal streaming in VTK. For this, I used a simple unsteady flow-field example and at the end of the blog, I challenged the readers to developing a simple particle advection code built on temporal streaming concepts. Here is my implementation of the particle integrator.

Most of the code in this filter actually takes care of initializing and maintaining data structures. The interesting parts are quite short. The streaming is implemented with something as simple as this:

Note that we don’t do any computing in the first time step (because we need pairs of two). For all other time steps, we use the previously cached dataset together with the current dataset to do the particle integration. Also, we tell the pipeline to keep executing until the last time step is reached. Simple.

The integration code is even simpler:

The ProbeVelocity function takes care of evaluating the velocity field at the given point coordinates. In this case, we evaluate both the previous and current time step at the previous particle locations. Then we average these velocities and advect the particles with this average value to obtain first order integration. In the actual VTK filter, we spent quite a bit of effort optimizing this part. Evaluating a field at a particular coordinates requires finding the cell containing it, which is non-trivial work for any grid types other than rectilinear grids.

The rest of code simply does bookkeeping. We can evaluate the filter with the following code.

We then get the following image.

particles

I also wrote a very simple filter to animate this results. Here it is.

Very simple filter. It takes the entire collection of particles (which is essentially a collection of particle paths) and takes out the particles that belong to the requested time step. By the way, when you look at this filter, you probably understand better why I created the TIME_VALUES meta-data and the “Time Values” variable in the particle filter.\

We can now change our pipeline a bit to generate an animation. Here is the code:

The output looks like this (click to animate):

Animation

That’s it folks. We’ll dig into other types of streaming in the future.

Questions or comments are always welcome!