Rotary Tube Bending Process Dataset

This dataset contains real-world data from 318 rotary tube bending processes, including 3 failed cases (IDs 1, 48, and 166). It is stored in a serialized Python pickle file named experiments_process_and_results.pkl (~800 MB) and is intended as a benchmark for machine learning and signal analysis methods.  The dataset includes:

A. Bending Setups:

• Tube: Outer Diameter, and Wall Thickness.

• Machine: Bending Target Angle, Wiper Die Shortening, Pressure Die( Lateral Position, Distance, Boost), Mandrel (Position, Retraction Timing), Collet Boost, and Clamp Die Lateral Position.

B. Process Parameters:

• Loads

  • Machine: Bend Die (Lateral, Rotating, Vertical), Clamp Die Lateral, Collet (Axial, Rotating), Mandrel Axial, and Pressure Die (Axial, Lateral, Left Axial).

  • Sensor: Mandrel Axial, Pressure Die (Lateral 1, Lateral 2).

• Movements

  • Bend Die (Lateral, Rotating, Vertical), Clamp Die Lateral, Collet (Axial, Rotating), Mandrel Axial, and Pressure Die (Axial, Lateral, Left Axial).

C. Geometry Data:

• STL-Suitable:

  • Linear 1: 50 series of datasets from one endpoint of the arc.

  • Arc: Datasets ranging from 0 to the bending target angle, going up every 1 degree.

  • Linear 2: 25 series of datasets from another endpoint of the arc.

• Key Characteristics:

  • Linear 1: Secondary Axis, Main Axis, and Out of Roundness.

  • Arc: Secondary Axis, Main Axis, and Out of Roundness.

  • Linear 2: Secondary Axis, Main Axis, and Out of Roundness.

To load, explore, and visualize data from this dataset, follow the steps in the GitHub repository:  https://github.com/zeyneddinoz/tubebend

Affiliation: University of Siegen, NRW, Germany

Funding: This work was funded by the Deutsche Forschungsgemeinschaft (DFG) – Project-ID: 520256321.