Measurement Basics
Vibration is measured with a piezoelectric accelerometer attached to the machine bearing housing. The accelerometer outputs a voltage proportional to acceleration (g or m/s²). Most data collectors process this signal to display:
- Overall vibration level — RMS velocity (mm/s) or acceleration (g) — for trending and alarm limits
- Time waveform — vibration amplitude over time — useful for identifying impacts and transients
- FFT spectrum — vibration amplitude at each frequency — the primary diagnostic tool
- Envelope (demodulation) spectrum — used for early bearing defect detection
Measure at the bearing housing, not on the casing or baseplate. Bolt-mounted or magnet-mounted accelerometers give consistent results; hand-held probes have poor high-frequency response and are only suitable for overall trending.
The FFT Spectrum
The Fast Fourier Transform (FFT) converts the time-domain signal into frequency-domain representation — showing which frequencies are present and at what amplitude. On the x-axis: frequency (Hz or CPM — cycles per minute = Hz × 60). On the y-axis: amplitude (typically velocity in mm/s RMS).
The most important frequencies to identify are multiples of running speed (1× = 1 RPM in Hz). For a 1,450 RPM machine: 1× = 24.2 Hz; 2× = 48.3 Hz; 3× = 72.5 Hz.
Fault Frequencies
| Fault | Primary frequencies | Direction | Phase characteristics |
|---|---|---|---|
| Mass imbalance | 1× RPM dominant | Radial (H+V), low axial | H and V 90° apart; consistent phase at 1× |
| Misalignment (parallel) | 1× and 2× RPM | Radial dominant | 180° phase across coupling at 1× |
| Misalignment (angular) | 1× and 2× RPM; high axial | Axial dominant | 180° axial phase across coupling |
| Looseness (structural) | Multiple harmonics 1×–5×; sub-harmonics (0.5×) | All directions | Asymmetric — H ≠ V |
| Resonance | Dominant peak at natural frequency; amplified by running speed or forcing frequency | Direction depends on mode shape | Phase shift of 90–180° as speed sweeps through resonance |
| Bearing defect (early) | BPFO/BPFI/BSF — often sub-harmonic amplitude, sidebands | Radial for outer race; all for inner race | Impulsive time waveform — detected in envelope spectrum |
| Gear mesh | GMF = number of teeth × RPM; sidebands at ±RPM | All directions | Sidebands indicate load variation or wear |
Setting Alarm Levels
Two approaches for alarm setting:
1. ISO 10816 severity zones: Use the zone boundaries from ISO 10816-3 or 10816-7 as alert and danger limits for overall vibration velocity. This provides a generic starting point. See our ISO 10816 standards guide.
2. Statistical alarms from baseline: Collect 8–12 baseline measurements on a machine known to be in good condition. Set alert at baseline mean + 2× standard deviation; set danger at 3× baseline. This is more sensitive than ISO limits for a specific machine.
Set both overall level alarms and spectral band alarms. Spectral band alarms — e.g. alarm if 1× RPM exceeds 5 mm/s — are more diagnostic and catch specific fault types before overall level reaches the alarm.
Organising Data Collection
Structured collection routes ensure nothing is missed and enable meaningful trending. For each machine in the programme:
- Define measurement points (typically 3 per bearing × number of bearings)
- Mark measurement point locations on the machine with paint or peened reference points
- Collect at defined intervals: critical machines every 2–4 weeks; general machines every 4–8 weeks
- Analyse every spectrum — don't just compare overall levels
- Store data in a condition monitoring database (CMMS or dedicated software)
- Generate exception reports for any machine exceeding alert limits