Representative raw 3-axis trace overlays paired with run-level feature distributions to show where good and bad machining runs separate — and where they do not.
Use the first chart to compare the raw temporal behavior of the representative good and bad recordings on X, Y, and Z directly. Then move down to the feature charts to see whether that visual difference survives when each run is compressed into RMS, peak, crest factor, and kurtosis features.
features.csv contains 4 run-level feature vectors (2 good, 2 bad), while traces.csv contains 2 representative raw traces. The raw overlays therefore illustrate representative behavior, and the feature charts cover the full set of supplied runs.
The centered raw-distribution view removes each trace’s mean before plotting, so Z-axis gravity offset does not dominate the spread comparison.
| File | Status | Raw trace present | Peak Y | Crest Y | Kurtosis Y | RMS Z |
|---|---|---|---|---|---|---|
| M01_OP01_good_001.h5 | Good | Yes | 2.1K | 11.9 | 7.40 | 1.0K |
| M01_OP01_good_000.h5 | Good | No | 2.1K | 12.2 | 6.98 | 1.0K |
| M01_OP01_bad_000.h5 | Bad | Yes | 888 | 3.85 | -0.72 | 1.1K |
| M01_OP01_bad_001.h5 | Bad | No | 876 | 4.89 | -0.03 | 1.1K |