task_categories:
- time-series-forecasting
language:
- en
language: - en pretty_name: "NASA Milling Dataset" tags: - manufacturing - time-series - sensors - tool-wear license: "cc-by-4.0" task_categories: - other
Dataset Card for NASA Milling Dataset
Dataset Summary
This dataset contains measurements from milling machine experiments conducted at NASA Ames Research Center (Goebel, 1996) in collaboration with UC Berkeley.
The experiments investigated tool wear during milling under different conditions. Multiple sensors were used, including acoustic emission, vibration, and spindle motor current sensors, capturing the machine’s behavior during the cutting process.
The dataset has been reformatted into Parquet for easier use in data science workflows. It provides both experimental parameters and sensor signals, along with measurements of flank wear (VB) taken intermittently.
This dataset can be used for research in:
- Tool condition monitoring
- Sensor fusion
- Predictive maintenance
- Time-series analysis in manufacturing
Supported Languages
The dataset contains numeric sensor data only.
Language-agnostic; language: en is used since the documentation is in English.
Dataset Structure
The dataset is provided as a single Parquet file (data.parquet).
Each row corresponds to a recorded experimental run with sensor measurements and metadata.
Features (columns):
case: Case number (1–16), defines experimental setup.run: Counter for experimental runs in each case.VB: Flank wear (mm), measured intermittently.time: Duration of experiment run.DOC: Depth of cut (mm).feed: Feed rate (mm/rev).material: Workpiece material (1 = cast iron, 2 = steel).smcAC: AC spindle motor current.smcDC: DC spindle motor current.vib_table: Table vibration.vib_spindle: Spindle vibration.AE_table: Acoustic emission at the table.AE_spindle: Acoustic emission at the spindle.
Experimental Conditions
The 16 cases represent combinations of:
- Depth of cut: 0.75 mm or 1.5 mm
- Feed rate: 0.25 mm/rev or 0.5 mm/rev
- Material: Cast iron or steel
Each condition was repeated with two different sets of inserts, giving 16 total cases.
Data Splits
The dataset has no predefined splits.
All data is provided in a single file. Users may create their own train/validation/test splits.
Dataset Creation
Curation Rationale
The dataset was collected to study tool wear progression during milling, and to investigate how different sensor signals correlate with wear.
Source Data
- Original source: NASA Ames & UC Berkeley milling experiments https://data.nasa.gov/Raw-Data/Milling-Wear/vjv9-9f3x/data
- Documentation: Goebel, K. (1996). Management of Uncertainty in Sensor Validation, Sensor Fusion, and Diagnosis of Mechanical Systems Using Soft Computing Techniques. Ph.D. Thesis, UC Berkeley.
Collection Process
- Milling performed on a Matsuura machining center (MC-510V) at cutting speed 200 m/min (826 RPM).
- Sensors: Acoustic emission, vibration (spindle & table), spindle motor current.
- Data acquisition: High-speed DAQ at up to 100 kHz, filtered and RMS-processed to ~250 Hz effective sampling.
- Tool wear measured as flank wear (VB) using a microscope, at irregular intervals.
Considerations for Using the Data
- Flank wear (
VB) was not measured after every run; some values are missing. - Signals were preprocessed (filtered and RMS smoothed). Raw AE and vibration were also recorded but are not included here.
- Units vary by signal type (currents in A, vibrations in g, AE in V).
License
This dataset is shared under the CC-BY-4.0 license. Please cite appropriately when using.
Citation
If you use this dataset, please cite:
@phdthesis{goebel1996,
title={Management of Uncertainty in Sensor Validation, Sensor Fusion, and Diagnosis of Mechanical Systems Using Soft Computing Techniques},
author={Goebel, Kai},
school={University of California, Berkeley},
year={1996}
}
Dataset Card Contact
For issues with this dataset card, please open an issue or pull request on the Hugging Face dataset repository.