Rotating equipment failures account for 60% of unplanned industrial downtime. Every motor, pump, gearbox, and compressor has a unique vibration signature when healthy — and that signature changes in predictable ways as bearings wear, shafts misalign, gears degrade, and rotors go out of balance. The difference between catching faults 6-8 weeks early and missing them comes down to sensor quality, mounting precision, and data infrastructure. Vibration sensors need precise mounting on bearing housings — not sheet metal guards. Acoustic sensors need quiet zones and isolation from ambient noise. Retrofit installations compromise both. When you design vibration and acoustic monitoring into your new plant, every sensor mount point is machined, every cable is shielded, and every baseline is captured during commissioning. Design Your Vibration Network
Why Mounting Quality Makes or Breaks Vibration Data
A vibration sensor is only as good as the mechanical path between the bearing and the sensor. ISO 10816 and ISO 20816 specify measurement on bearing housings — every millimeter of metal between bearing and sensor attenuates high-frequency content. Mounting on sheet metal guards or through thick paint can reduce usable frequency response by 50-80%.
Want every mount point in your construction drawings? Design Your Vibration Network
Sensor Technology Selection
| Sensor Type | Frequency Range | Sensitivity | Best For | Limitation |
|---|---|---|---|---|
| Piezoelectric (ICP) | 0.5 Hz – 20+ kHz | 100 mV/g standard | Permanent monitoring on critical rotating equipment; gold standard | Requires constant power; no DC response |
| MEMS Accelerometer | 0 Hz – 6 kHz | Variable; lower than PE | Wireless sensors, embedded monitoring, fleet coverage | Limited high-frequency; susceptible to zero drift |
| Proximity Probe | 0 – 1 kHz (shaft) | 200 mV/mil | Sleeve-bearing machines: turbines, large compressors | Requires bore through housing; shaft measurement only |
| Acoustic Emission | 20 kHz – 1 MHz | Wideband/resonant | Micro-cracks, incipient bearing faults, friction, leaks | Sensitive to mounting quality; needs quiet zone |
| Ultrasonic Mic | 20 – 100 kHz airborne | Non-contact | Leak detection, lubrication assessment, electrical discharge | Ambient noise interference |
Frequency-to-Fault Mapping
| Fault Type | Frequency Signature | Detection Method | Required Sensor | Prediction Horizon |
|---|---|---|---|---|
| Rotor Imbalance | 1× RPM dominant | FFT spectral; trending 1× amplitude | Accelerometer (tri-axial) | Months |
| Shaft Misalignment | 2× RPM dominant; high axial | FFT; 2×/1× ratio trending | Accelerometer (axial + radial) | Weeks-months |
| Bearing Outer Race | BPFO + harmonics | Envelope analysis (demodulation) | High-freq accelerometer; AE | 6-8 weeks |
| Bearing Inner Race | BPFI modulated by 1× RPM | Envelope with RPM sync | Accelerometer; proximity probe | 4-6 weeks |
| Gear Tooth Wear | Mesh freq + sidebands | Time synchronous averaging | Accelerometer on gearbox (stud) | 3-12 weeks |
| Cavitation | Broadband 2-20 kHz | RMS acceleration trending | Accelerometer + AE (suction side) | Days-weeks |
| Lubrication Breakdown | Ultrasonic 20-40 kHz rise | dB trending in ultrasonic band | Ultrasonic or AE sensor | Days |
| Structural Looseness | Sub-harmonics + multiples | Time waveform (impact patterns) | Accelerometer (radial) | Weeks-months |
Sampling Rate & DAQ by Equipment
| Equipment | RPM Range | Min. Sampling Rate | Channels/Machine | Data Volume |
|---|---|---|---|---|
| Motors/Pumps (>600 RPM) | 900-3,600 | 25.6 kHz | 3-4 | ~3 MB/1s capture |
| Low-Speed (<600 RPM) | 30-600 | 6.4 kHz (500 mV/g sensor) | 2-4 | ~1.5 MB/capture |
| Spindles (>10K RPM) | 10K-60K | 51.2 kHz minimum | 2-3 | ~6 MB/1s capture |
| Gearboxes | Varies | 51.2 kHz | 4-6 | ~5 MB/capture |
| Turbomachinery | 3K-15K | 25.6 kHz + DC-1 kHz prox | 4-8 | ~8 MB/capture |
Need DAQ hardware sized for your equipment? Design Your Vibration Network
Vibration vs. Acoustic: Complementary Detection
AI Baseline During Commissioning
Confirm every sensor reads correctly: sensitivity, noise floor, frequency response. Fix mounting problems before production starts.
Record spectra at 25%, 50%, 75%, 100% load over 2-4 weeks. Multi-condition baseline is the "healthy" reference for AI.
ISO 10816/20816 velocity levels plus AI dynamic baselines that adapt to operating conditions. Eliminates nuisance alarms.
Anomaly detection (Isolation Forest, autoencoders) on healthy data. Supervised models pre-loaded with known fault signatures. Continuous improvement from production data.
Key Benefits & ROI
Vibration Data Quality Is Decided at Construction
iFactory designs vibration and acoustic monitoring for greenfield plants — sensor selection, precision mounting, DAQ hardware, signal processing, and AI baselines as construction-ready docs.
Frequently Asked Questions
A Sensor on Sheet Metal Tells Nothing — A Sensor on a Machined Pad Tells Everything
Mounting quality determines whether AI detects bearing faults 6-8 weeks early or misses them entirely. Design it right before construction.
