BubbleML is hosted on AWS and can be publicly downloaded. Each boiling study can be downloaded separately:
| Study | Size |
|---|---|
| Single Bubble | 503.0 MB |
| Pool Boiling Saturated | 24.4 GB |
| Pool Boiling Subcooled | 10.5 GB |
| Pool Boiling Gravity | 16.5 GB |
| Flow Boiling Inlet Velocity | 11.4 GB |
| Flow Boiling Gravity | 10.9 GB |
| Pool Boiling Subcooled 0.1 | 155.1 GB |
| Pool Boiling Gravity 0.1 | 163.8 GB |
| Flow Boiling Gravity 0.1 | 108.6 GB |
| 3D Pool Boiling Earth Gravity | 122.2 GB |
| 3D Pool Boiling ISS Gravity | 62.6 GB |
| 3D Flow Boiling Earth Gravity | 93.9 GB |
Each download is a .tar.gz. They can be unzipped using the command
tar -xvf <study>.tar.gz -C /path/to/BubbleML/<study>/
After unzipping, you will see a collection of hdf5 files in /path/to/BubbleML/<study>/.
Each hdf5 file corresponds to one simulation. The hdf5 files can be loaded with common libraries,
such as h5py`.
We provide documentation describing the different hdf5 datasets in each simulation file. There are also examples showing how to load a BubbleML simulation, list out it’s datasets, visualize the different simulation fields, and access the metadata.
We provide a separate reproducibility capsule for running Flash-X simulations. This repo includes all of the submission files BubbleML used. Modifying these will be a straightforward way to generate new data and extend BubbleML. Note, Flash-X is designed for large-scale, long-running simulations so there are dependencies on other projects like MPI.
The studies can also be downloaded in bulk by running the bash script
bash download_all.sh
This will download all datasets listed above. Note: the full dataset is over a terabyte in size.