Healthy vs. Faulty CNC Milling Vibration Dashboard
Feature distributions, waveform overlays, and frequency-domain contrasts across 8 milling cuts show which vibration signatures distinguish stable machining from faulty behavior.
Cuts compared
8
4 healthy / 4 faulty
Balanced sample split across the comparison set.
X-axis kurtosis gap
6.19 vs 2.87
+115.9% healthy
The strongest feature separator in the dataset.
Z-axis RMS shift
1.08K vs 1.03K
+4.9% faulty
Faulty cuts lift the vertical vibration baseline and spread.
Z-axis variance lift
5.49×
faulty / healthy mean
The largest variance amplification appears on the Z axis.
Feature spotlight: where healthy and faulty runs separate most clearly
Box-and-strip distributions highlight the clearest class split on X-axis kurtosis and the elevated Z-axis RMS in faulty cuts.
Variance inflation in faulty runs by axis
Log-scaled variance distributions show broader and higher-spread faulty behavior, especially on Z.
What separates the healthy cuts from the faulty cuts
The healthy runs cluster tightly in feature space, while the faulty runs fan outward with larger spread. The most pronounced shift is on X-axis kurtosis, where healthy cuts stay near 6.19 and faulty cuts drop toward 2.87 with much wider dispersion.
The Z-axis RMS change is smaller in mean terms, but the faulty runs are far less stable: their RMS spread rises from roughly 0.26 in healthy cuts to 24.35 in faulty cuts, creating a clean diagonal split when X kurtosis and Z RMS are viewed together.
Strongest class separator
X-axis kurtosis produces the largest effect size in the extracted feature set.
Largest instability signal
Faulty mean variance is 3.25× higher on average across axes, with the biggest lift on Z.
Signal-domain confirmation
Waveform bands and FFT overlays show the faulty class carrying broader envelopes and higher low-frequency to broadband spectral energy.
Selected feature contrast summary
| Feature | Healthy mean ± σ | Faulty mean ± σ | Faulty / Healthy | Spread ratio |
|---|---|---|---|---|
| X-Axis Kurtosis | 6.19 ± 0.06 | 2.87 ± 1.29 | 0.46× | 22.11× |
| Z-Axis RMS | 1.03K ± 0.26 | 1.08K ± 24.4 | 1.05× | 92.08× |
| X-Axis Variance | 168.32K ± 1.08K | 273.39K ± 63.80K | 1.62× | 59.27× |
| Y-Axis Variance | 35.86K ± 2.08K | 94.50K ± 44.21K | 2.64× | 21.23× |
| Z-Axis Variance | 19.39K ± 497.2 | 106.42K ± 71.29K | 5.49× | 143.38× |
File-level separation using X-axis kurtosis and Z-axis RMS
Marker size encodes average variance across axes, making the faulty points visibly larger as well as more dispersed.
Overlaid waveform snippets for healthy vs. faulty cuts
Shaded envelopes capture the min-to-max band across runs; solid lines show the class mean for each axis.
Overlaid FFT spectra for healthy vs. faulty cuts
Frequency-domain envelopes show the faulty class carrying consistently higher spectral power across the observed band.