An overall vibration number tells you a machine is vibrating too much, but not why. The vibration velocity spectrum is the tool that answers why, by breaking the vibration signal into the individual frequencies that make it up. Instead of one lumped figure, you get a chart of amplitude versus frequency, and the frequencies where peaks appear point to specific mechanical faults. Because those fault frequencies are tied to the machine's running speed, a spectrum turns a vague vibration complaint into a specific diagnosis.
Vibration velocity spectrum in one line: A vibration velocity spectrum is the frequency-domain view of a machine's vibration, produced by applying a Fast Fourier Transform to the vibration signal and plotting amplitude against frequency in velocity units. Peaks in the spectrum appear at specific frequencies, such as one and two times running speed, and their location identifies which mechanical fault is present.
A vibration sensor produces a time waveform: amplitude plotted against time as the machine shakes. That waveform contains all the information, but it is a tangled superposition of every vibrating source in the machine, and the eye cannot usually pick the causes out of it. The Fast Fourier Transform, or FFT, is the mathematical operation that untangles the waveform into the set of pure frequencies that add up to make it. The result is the spectrum: the same energy, but organized by frequency so that each vibrating mechanism appears as its own distinct peak.
The value of that reorganization is that different faults vibrate at different, predictable frequencies. Imbalance, misalignment, looseness, bearing defects, and gear problems each excite characteristic frequencies related to the machine's geometry and speed. In the time waveform these all overlap into one messy trace, but in the spectrum they separate into peaks at distinct locations. Reading a spectrum is largely a matter of noting which frequencies carry the energy and translating those locations back into the mechanisms that produce them.
The two views are complementary rather than competing. The time waveform is best for impulsive, transient events, a periodic knock or impact reads clearly as a spike in time even when it is smeared across many frequencies in the spectrum. The spectrum is best for steady, repetitive vibration, where separating the constant sources by frequency is exactly what diagnosis needs. Skilled analysts keep both, but for routine fault identification on rotating equipment the frequency spectrum is the primary tool.
The same vibration can be expressed as displacement, velocity, or acceleration, and the choice of units shapes what the spectrum emphasizes. Displacement weights low frequencies, acceleration weights high frequencies, and velocity sits in the middle. For the mid-frequency range where most rotating-machinery faults live, from around running speed up through the first several harmonics, velocity gives a flatter, more even response, so a moderate fault at low frequency and one at higher frequency are represented on comparable terms. That is why velocity in millimeters per second is the workhorse unit for general rotating-machine diagnosis, while acceleration is reserved for the very high frequencies of bearing and gear defects.
The reference point in any spectrum is 1x, the frequency equal to the machine's running speed, one cycle per shaft revolution. A dominant peak at 1x is the classic signature of mass imbalance, since an unbalanced rotor pushes once per revolution. A strong peak at 2x, twice running speed, points toward misalignment or a bent shaft, and its presence alongside 1x often distinguishes those faults from pure imbalance. Because these frequencies are defined relative to shaft speed, analysts often think in orders, multiples of running speed, so the same pattern reads the same way whether the machine turns fast or slow.
Higher harmonics and their patterns add more resolution. A whole series of harmonics, 1x, 2x, 3x and beyond, typically indicates mechanical looseness. Sidebands clustered around a central peak indicate a modulating fault such as a gear or bearing problem. Peaks that are not integer multiples of running speed point to sources that are not locked to the shaft, such as rolling-element bearing defect frequencies or belt frequencies. Learning the vocabulary of where peaks land and what patterns they form is what converts a velocity spectrum from a chart into a diagnosis.
Because fault frequencies are predictable from the machine's speed and geometry, they can be monitored automatically rather than by an analyst inspecting every chart. A condition-monitoring system can define narrow frequency bands around 1x, 2x, and the known defect frequencies, then track the amplitude within each band over time. A rising 1x band flags developing imbalance; a growing 2x band flags emerging misalignment; energy appearing at a bearing defect frequency flags a failing bearing. This turns spectral diagnosis from an expert-only, one-machine-at-a-time task into something a monitoring platform can scan across many machines.
Feeding those band values into a SCADA platform lets the spectral picture live alongside the rest of the machine's operating data. Merobix can trend the amplitude at 1x, 2x, and defect-related bands for each monitored point, so an operator sees not just that overall vibration rose but which frequency band drove the increase, which points straight at the likely cause. For remote and unmanned sites, this is the difference between knowing a pump is vibrating and knowing it is probably misaligned, without dispatching a specialist to interpret raw spectra on site.
The real strength is trending each band over weeks and months to catch faults while they are still small. A single spectrum is a snapshot; a slowly rising 1x band is a story about imbalance developing, perhaps from erosion or fouling on an impeller. Historized band trends establish each machine's normal spectral fingerprint and make departures from it stand out early, so maintenance can be planned around a growing peak rather than reacting after the overall vibration alarms. Combined with correlated process data, the spectral trend often explains not just what is failing but why.
A time waveform plots vibration amplitude against time and contains every vibrating source overlapped into one trace, which is hard to interpret directly. A spectrum applies an FFT to that waveform to plot amplitude against frequency, separating each source into its own peak. The waveform is best for impulsive events, while the spectrum is best for identifying steady, repetitive faults by the frequencies where their energy appears.
Velocity gives a relatively flat response across the mid-frequency range where most rotating-machinery faults occur, so faults at low and higher frequencies are represented on comparable terms. Displacement emphasizes low frequencies and acceleration emphasizes high frequencies. Velocity in millimeters per second is therefore the standard unit for general rotating-machine diagnosis, with acceleration used specifically for very high-frequency bearing and gear defects.
1x is the frequency equal to the machine's running speed, one cycle per shaft revolution. A dominant peak at 1x is the classic signature of mass imbalance, because an unbalanced rotor exerts a force once every revolution. A strong peak at 2x, twice running speed, instead points toward misalignment or a bent shaft, so comparing the 1x and 2x peaks helps separate those common faults.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.