MiFluids

This project is a virtual lab that I am currently working on which will be used to teach students about fluid flow. The lab was created using Unity for the visualization component and using Python to process the data coming from the fluid simulation software.

This virtual lab is meant to run inside of browsers in real time. However, realistic fluid simulations are very costly and can sometimes take minutes to hours to calculate, and require higher end hardware. In order to resolve these issues, we worked with the client to find a way to pre-bake the data and display it in the browser for students to view.

Highlights:

Displaying Fluid Flow Data

In order to create this video, we have the client create a simulation using proprietary software and send us a CSV file for each timestamp in the simulation. From there we transfer the data into a texture which provides efficient storage. In order to do that, we pack the u,v,w components of the velocity into the r,g,b channels. Then the frames are compiled into a frame sheet. This is then served to the student, and using metadata we are able to unpack the original values of u,v,w within an acceptable percentage of error for the user to probe and read. This is able to be run in realtime, due to the efficient storage, and avoids the computations that are required for realistic simulations. We considered baking out a video for this component but decided against it due to some unwanted effects of compression algorithms. With this current format it is uncompressed and is able to represent the data with a +/-0.39% error.

Vector Field:

Another important part of the lab is the vector field, this is fairly straightforward. It samples the data in the U and V directions and determines the direction and magnitude of the velocity at that point. It then rotates the vectors based on the direction and sets the length based on the magnitude. Here is a demonstration of it running:

NOTE: This is on display in the video at the 9 second mark

Stream Lines:

This works in a very similar manner to the airflow visualization. The streamlines are calculated offline using “matplotLib”, then baked into another frame sheet and provided to the user to display with the other components of the visualization. We will explore turning this into a video file in the near future, as this component will be able to benefit greatly from video compression and does not require direct sampling and accurate values.

Probing the data:

To interact with this data we allow the user to add probes to sample the data that will display the airflow data to the user. We also have implemented the ability to add a rake so that multiple probes can be placed in order. Here is an example of this in action:

NOTE: This is on display in the video at the 18 second mark

I am unable to provide a link to the project as this was a project I am developing at work, I am however allowed to display videos of the work.