All projects
Synthetic Data for Mechanical Vibrations
Industrial computer vision
100,000 synthetic 5-second structural-vibration videos generated with physics-based Python finite element analysis, structured for computer-vision models.
What it is
A synthetic dataset of structural-vibration videos for training industrial computer-vision models.
Problem addressed
Computer-vision models for vibration analysis need large amounts of labelled video across many machine designs, frequencies and vibration levels, which is difficult to collect from real equipment.
What has been built or demonstrated
- 100,000 synthetic 5-second structural-vibration videos.
- Physics-based Python finite element analysis for mechanical accuracy.
- Frequency range 0–15 kHz; vibration magnitude 0.001g–10g.
- Raw videos and motion-amplified videos at 10×–1000×.
- Industrial model designs spanning steel, cement, oil, pharma and paper mills.
- Dataset structured for C-RNN and Transformer-based computer-vision models.
Limitations
- The dataset is synthetic: generated from physics-based models, not recorded from real equipment.
- Delivered client R&D, not a product for sale.
Planned next steps
- Apply physics-based synthetic data methods to further engineering AI work.
Interested in a pilot or technical collaboration?
We are open to pilot partners. Tell us about the engineering problem you are working on.