Skip to content

RDDAC

Dataset License: CC BY 4.0 Code License: MIT Python 3.10+ DaRUS Repository DOI

3D laser scan point clouds of one experiment after OP10 (deep drawing) and OP20 (cutting), colored by their deviation from the matching DDACS simulation

Measured point clouds of one experiment after OP10 (left) and OP20 (right), colored by the deviation from the matching DDACS simulation.

A large-scale experimental dataset of 9,000 physical deep-drawing and cutting experiments — the real-world counterpart to the DDACS FEM simulations. Each experiment forms a modified quadratic cup from DP600 dual-phase steel (deep drawing in OP10, cutting in OP20) and records press force signals, sheet-thickness and oil-film traverses, and high-resolution 3D laser scans of the part after each operation. Use it to quantify the simulation-to-reality gap, train models on real process data, or validate DDACS-trained surrogates against physical measurements.
Experiments 9,000
Total size ~87 GB (HDF5, lossless)
Process steps per experiment 2 (OP10 deep drawing, OP20 cutting)
Parameter space 2 geometries x 3 blankholder forces x 3 oil types (18 categories)
Repetitions up to 500 per category
Train / val / test 7,200 / 900 / 900 (predefined, seed 42)
Matching simulations DDACS rddac.zip (~9 GB), fetched by rddac download

The rddac package ships with the dataset and provides a Croissant native interface: one CLI for the download, one Python module for access, and an optional PyTorch IterableDataset for training.

Get the data

pip install rddac               # install the package
rddac download --small          # ~174 MB sample bundle into ./data

A first read

rddac.open_h5(experiment_id) opens one experiment and returns an h5py.File. Each file carries the press force time series, the two sensor traverses, and the OP10/OP20 laser scans.

import rddac

experiment_id = 0                # one experiment in the small sample bundle

with rddac.open_h5(experiment_id) as f:
    force = f["force/data"][:]                 # (n, 8): time, load cells, temp, position, total force
    sheet = f["sheet_thickness/data"][:]       # (n, 2): sensor position, thickness
    z10   = f["pointcloud/op10/z"][:]          # (6400000,) flat scan buffer

print("force samples:", force.shape[0], "| scan pixels:", z10.shape[0])
# force samples: 1140 | scan pixels: 6400000

For training the PyTorch tutorial wraps the same data in RDDACDataset. For a guided tour start with Getting Started.

Drop-in DDACS compatibility

The rddac public surface mirrors the ddacs package one to one: load, add_view, open_h5, inspect_h5, streaming.iter_view / export_to_numpy / load_export, and the PyTorch IterableDataset all share names, signatures, and semantics. Code written against DDACS ports by swapping the import:

# import ddacs as dataset_pkg                    # simulations
import rddac as dataset_pkg                      # real experiments

ds = dataset_pkg.load(data_dir="./data")
for record in dataset_pkg.streaming.iter_view("force-curve", data_dir="./data", dataset=ds):
    ...

Because RDDAC is the experimental counterpart to the DDACS simulations, rddac download also fetches the matching FEM simulations (rddac.zip from the DDACS dataset) into ./data/simulation — skip them with --no-sim. See Dataset Overview for the pairing.

License

The dataset on DaRUS is licensed under CC BY 4.0; the rddac software is licensed under MIT.

Citation

@dataset{baum2026rddac,
  title={Real Deep Drawing and Cutting Dataset},
  author={Baum, Sebastian and Heinzelmann, Pascal},
  year={2026},
  publisher={DaRUS},
  doi={10.18419/DARUS-5589}
}

@article{baum2026deviation,
  title={Statistical Analysis of Simulation to Reality Deviation in Deep Drawing with a Benchmark Dataset},
  author={Baum, Sebastian and Heinzelmann, Pascal and Clau{\ss}, P. and others},
  journal={Transactions of the Indian Institute of Metals},
  volume={79},
  pages={176},
  year={2026},
  doi={10.1007/s12666-026-03870-5}
}