Savonius turbine simulation in WebGL
This is a fluid mechanics course project, co-authored by:
- Raúl Álvarez
- Ignacio Peñafiel
- Víctor Tirreau
- Felipe Valenzuela
It consists of a Stable Fluid simulation of a Savonius rotor or turbine that runs on the GPU using WebGL. It's heavily based on Jérémie Piellard's (@piellardj) navier-stokes-webgl. Thank you Jérémie for your support and directions.
You can find the live demonstration here.
Simulation
This is an implementation of the Stable Fluid described by Jos Stam. The simulation is implemented on GPU with the method provided in GPU Gems.
Data storage
This simulation can run in two modes for storing the velocities:
- velocity stored in float textures: each component (x, y)
is stored as a 32bit float.
To do so the following extensions must be available:
OES_texture_float,WEBGL_color_buffer_float,OES_texture_float_linear. - velocity stored in normal textures with four 8bit channels. In this mode, each component is stored as a 16bit fixed point value, encoded in two 8bit texture channels. This mode provides less precision for the computing, and you can see artifacts if you push the display color intensity to the maximum.
Building the app
Make sure to install all the dependencies with npm i, and then
run npm run build for the first time from the project directory.
You can also rebuild the generated docs folder with npm run rebuild,
which is usual in the development cycle. If you only
edit the simulation code (and don't change the page layout),
it's faster to just run npm run webpack.
Known limitations
The current implementation does not include the diffusion term of the Navier-Stokes equation, nor does it consider the force exerted by the turbine on the fluid as it rotates.
The generated power is calculated by the product of the torque by the angular speed. The torque is calculated by integrating the pressure along the blade perimeter. This is done on the CPU, downloading the pressure texture from the GPU. The displayed power is an exponential moving average of the instantaneous power, and lacks strict physical units. It's scaled to look appealing, as its main purpose is to compare the performace of different turbine configurations.
Feel free to clone the repo and create a Pull Request to extend the project.