A spectral band alarm watches the vibration energy inside a specific slice of the frequency spectrum rather than the whole signal at once. By dividing the spectrum into diagnostic bands aligned with where particular faults appear, it can detect a developing problem in one band while the rest of the machine, and the overall level, look normal. This targeted sensitivity is what makes band alarms the early-warning layer that a single broadband number cannot provide on its own.
Spectral Band Alarm in one line: A spectral band alarm computes the vibration energy within a defined frequency band, such as a running-speed band or a high-frequency bearing band, and alarms on that band value rather than on the whole signal. Because a small defect confined to one band raises that band's energy well before it moves the overall level, band alarms catch faults early, and each band can be exposed as its own SCADA tag.
Different machine faults announce themselves at different frequencies, so the spectrum is carved into bands chosen to match those fault families. A sub-synchronous band below running speed catches phenomena such as looseness, rub, or oil whirl. A band at running speed, often called the 1x band, tracks the imbalance and misalignment energy that dominates most machines. Harmonic bands above running speed pick up misalignment, looseness, and other faults that generate multiples of the turning frequency.
Above the harmonics, a high-frequency or bearing band captures the energy from rolling-element bearing defects and gear meshing, which appear well above running speed and often at non-integer frequencies. This band is where early bearing deterioration first shows, because a spalling bearing rings at high frequencies long before it produces significant low-frequency vibration. Some systems add an envelope or demodulation band specifically tuned to draw out the repetitive impacts of bearing defects from the broadband noise around them.
Each band is defined by a lower and an upper frequency, and the monitoring system computes a single amplitude, usually an RMS energy, for the signal falling within it. The band boundaries are set with the machine's running speed and known fault frequencies in mind, so that the energy of a particular fault falls cleanly inside one band. Well-chosen bands turn a raw spectrum into a small set of numbers, each of which points at a specific class of problem.
The overall level sums all frequencies into one number, so a small defect is diluted by the machine's dominant vibration and can grow substantially before it shifts the total. A band alarm removes that dilution by looking only at the frequencies where the defect lives. A bearing fault that produces a modest amount of high-frequency energy is negligible against the whole signal but can be a large fractional increase within the narrow bearing band, so the band value rises sharply while the overall barely moves.
This is the essence of early warning. A rising bearing band trips its alarm and flags the machine for attention weeks or months before the defect grows enough to disturb the overall level and threaten the machine. The band has effectively increased the sensitivity of the measurement to exactly the fault it is watching for, without waiting for that fault to become large enough to dominate the total energy.
Bands also localize the fault, not just detect it. Because each band corresponds to a fault family, the band that alarms points toward the likely cause: a 1x band rise suggests imbalance, a harmonic band rise suggests misalignment or looseness, and a bearing band rise suggests a bearing defect. This gives the operator a head start on diagnosis, turning a bare alarm into a directed hint about where to look, which the overall level alone can never provide.
In an online or SCADA monitoring architecture, each band value is exposed as its own tag alongside the overall level, so that the historian trends and the alarm logic act on every band independently. Rather than a single vibration number per point, a bearing point might publish a sub-synchronous tag, a 1x tag, a harmonic tag, and a bearing-band tag, each with its own alert thresholds. This multiplies the diagnostic content flowing into the monitoring system without requiring an analyst to interpret a spectrum in real time.
Individual band tags also let early-warning logic be built directly into the control and monitoring layer. A rising bearing band can drive its own alert, escalate a maintenance notification, or feed a rule that combines several bands to distinguish one fault from another. Because the bands are ordinary tags, they participate in the same trending, alarming, and notification machinery as any other process value, so vibration early warning becomes part of routine operations rather than a separate specialist activity.
For cloud monitoring across many machines, band tags make it practical to survey a fleet for the specific fault each band represents. A supervisor can look across sites for machines whose bearing bands are climbing, targeting inspections where they will pay off, all from trended tag data rather than from opening spectra one by one. The full spectrum remains available for detailed diagnosis when a band alarms, but the bands are what let a monitoring system catch and route early defects at scale.
An overall alarm watches the total vibration energy across the whole frequency range, so a small defect is diluted by the machine's dominant vibration and can grow a lot before it trips. A band alarm watches only a chosen slice of the spectrum, so a defect confined to that slice raises the band value sharply while the overall barely moves. That focus is why band alarms catch early bearing and gear defects the overall would miss.
Common bands include a sub-synchronous band below running speed for looseness and instabilities, a 1x band at running speed for imbalance and misalignment, harmonic bands above running speed for misalignment and looseness, and a high-frequency or bearing band for rolling-element bearing and gear defects. Some systems add an envelope or demodulation band tuned to draw out bearing impact energy. The exact boundaries are set from the machine's running speed and known fault frequencies.
The monitoring system computes an energy value for each defined band from the spectrum and publishes each as its own tag, alongside the overall level. Those tags are trended in the historian and given their own alarm thresholds, just like any process value. This lets early-warning logic act on individual bands, and lets a fleet be surveyed for a specific fault by looking across the corresponding band tags.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.