Vibration monitoring has a credibility problem. Too many dashboards present a single “overall vibration” number and an amber/red indicator, and too many asset owners have learned the hard way that this is not enough to make a maintenance decision. The sensor is rarely the problem. A modern industrial accelerometer has more than enough bandwidth. The analysis is the problem. A rotating asset produces distinct fault signatures at distinct frequencies, and collapsing those into a single RMS number throws away most of the diagnostic value.
At PRAXIS we structure our vibration analysis around seven frequency bands, labelled B1 through B7. The band boundaries are chosen to isolate the physical mechanisms that actually fail on the assets our clients care about: motors, gearboxes, pumps, cranes, conveyors, rail infrastructure. The bands are not proprietary mathematics. They are a disciplined way of grouping the physics.
B1 · Sub-1× running speed
This band captures looseness, misalignment at low frequency, structural flex, and, on rail assets, sleeper voiding and ballast dynamics. A rising B1 on a pump is almost always foundation or coupling related, not the pump itself.
B2 · 1×–3× running speed
The classic unbalance, misalignment, and bent-shaft band. Energy here is highly diagnostic: unbalance concentrates at 1×, misalignment produces a strong 2× component, and certain coupling faults generate 3×. The ratio between these peaks tells you which fault is active.
B3 · 3×–10× running speed
Blade-pass, vane-pass, and belt-drive frequencies. On a centrifugal pump this band will tell you about impeller damage, on a fan it will show blade fouling, and on a belt-driven drive it will pick up pulley run-out well before the operator hears anything.
B4 · Gear mesh and harmonics
Typically tens of Hz to a few kHz depending on the train. Mesh frequency amplitude alone is often steady; the diagnostic signal is sideband energy around the mesh frequency at shaft rotation rate, which is the fingerprint of tooth wear, chipping, or backlash.
B5 · Bearing defect frequencies
Inner race, outer race, cage, and ball spin frequencies are computed from bearing geometry. Early defects appear here as discrete peaks well above background; they rarely change the overall RMS before they are several months into development.
B6 · 2–5 kHz envelope range
This is where incipient bearing damage first becomes detectable by demodulation (envelope) analysis. A bearing can show a clean spectrum in B5 but clear modulation in B6 for a considerable period before the defect frequencies themselves rise.
B7 · 5 kHz–10 kHz and above
Friction, lubrication distress, and surface asperity contact. A rising B7 on a rolling-element bearing is frequently the earliest indicator available and will typically precede any B5 peak by weeks to months.
Why it matters for the work order
The discipline this imposes is simple and useful. When we alert on an asset, we alert on which band has moved, by how much relative to the asset's own baseline, and for how long. “Vibration is high” is not actionable. “B6 envelope energy at the outer race defect frequency is three standard deviations above the 30-day baseline and trending upward over the past 11 days” is a work order.
This structure also explains why we are conservative about the word “prediction.” Sensored does not forecast the hour of failure. It detects when an asset deviates from its own validated behaviour, and it names the physical mechanism most likely to be behind the change. In our experience that shift, from overall level to band-resolved diagnostics, is what separates condition monitoring that gets used from condition monitoring that gets ignored.
