Vibration Analysis Basics for Rotating Equipment

Why Vibration Analysis Matters

Every rotating machine — pumps, motors, fans, compressors, gearboxes — generates vibration as it operates. When the machine is healthy, the vibration is low and follows predictable patterns. As components wear, the vibration signature changes, and these changes occur long before the machine fails. Vibration analysis is the process of measuring, interpreting, and acting on these patterns to detect faults early.

A well-implemented vibration programme can detect bearing faults up to 6 months before failure, unbalance and misalignment within weeks of onset, and gear tooth defects at the earliest pitting stage. This lead time enables planned maintenance during scheduled downtime rather than emergency repairs during production.

Vibration Measurement Parameters

Vibration can be measured in three ways, each sensitive to different fault types and frequency ranges:

Frequency Analysis & FFT

A vibration signal is a complex waveform — a mixture of many frequencies superimposed on each other. To identify specific faults, we use a Fast Fourier Transform (FFT) to decompose the time-domain signal into its frequency-domain components. The resulting frequency spectrum shows the amplitude of vibration at each frequency, allowing us to identify which component is generating the vibration.

Key frequencies to identify on the spectrum:

ISO 10816 Vibration Severity Chart

ISO 10816 provides vibration severity guidelines for machines grouped by class. The standard measures overall RMS velocity (mm/s) in the 10–1,000 Hz band. Here's a simplified summary for common machine classes:

Values shown are RMS velocity in mm/s, measured on non-rotating parts (bearing housing).

Common Machine Faults and Their Signatures

• Unbalance: Dominant 1× peak. Phase angle is consistent across the rotor. Amplitude increases with speed squared. Corrected by balancing.
• Misalignment (angular): Dominant 2× peak (sometimes 1× as well). Phase difference of 180° across the coupling. Corrected by realigning.
• Misalignment (parallel/offset): Strong 2× peak with significant 1× component. Phase difference of 180° across coupling radially.
• Mechanical looseness: Multiple harmonics of 1× (2×, 3×, 4×, 5×). Often directional — worse in one direction. Check for loose mounting bolts, soft foot, or cracked foundations.
• Bearing defect (outer race): BPFO frequency and its harmonics. As the defect worsens, sidebands at 1× running speed appear around the bearing tones. Envelope (demodulation) analysis detects this earliest.
• Bearing defect (inner race): BPFI frequency with sidebands at 1× running speed. More complex spectrum than outer race faults. Often accompanied by an increase in the high-frequency impact (SPM/SE) readings.
• Cavitation (pumps): Broadband high-frequency vibration (5–20 kHz) that sounds like gravel passing through the pump. Not at a specific frequency — the key is the broadband nature and the sound.
• Gear tooth defect: Gear mesh frequency (teeth × speed) with sidebands at running speed. As the defect worsens, sideband amplitude increases. Localised tooth damage shows as impulses in the time domain.

Setting Alarm Levels

Vibration alarm levels should be set based on baseline data and statistical analysis, not just the ISO chart. A proven approach:

1. Baseline: Collect 30+ readings over several weeks of normal operation to establish a statistical baseline.
2. Alert level: Set at baseline mean + 2× standard deviation (95% confidence). This triggers investigation but not shutdown.
3. Alarm level: Set at baseline mean + 3× standard deviation (99.7% confidence) OR ISO 10816 "Unsatisfactory" limit — whichever is lower.
4. Danger/Trip level: Set at ISO 10816 "Unacceptable" or baseline mean + 4× standard deviation. This triggers immediate action or shutdown.
5. Rate-of-change alarm: Set a trend alarm that triggers if vibration increases by more than 25% in 24 hours — this catches rapid degradation faster than absolute thresholds alone.