A vibration spectrum looks intimidating the first time you see one: a jagged line of peaks, a frequency axis and a few unfamiliar labels. But most common faults in rotating machinery leave distinct fingerprints, and five of them cover much of what a general-purpose monitoring program flags on motors, fans, pumps and gearboxes: unbalance, misalignment, mechanical looseness, rolling-element bearing defects and electrical faults. You don’t need to be an analyst to recognize them. Knowing them lets you read a vibration report critically, ask the right questions and send the right technician with the right tools.
Reading a spectrum in two minutes
A spectrum is produced by a fast Fourier transform (FFT) of the vibration signal. It shows amplitude on the vertical axis, usually velocity in in/s or mm/s, against frequency on the horizontal axis. Analysts usually label frequency in orders of running speed: “1×” is the shaft’s rotational frequency, “2×” is twice that, and so on.
For example, a 4-pole motor on a 60 Hz supply running at 1,770 rpm turns at 29.5 Hz, so 1× is 29.5 Hz and 2× is 59 Hz. Measurements are taken in three directions at each bearing: horizontal, vertical (both radial) and axial, along the shaft. Which direction carries the energy is often as revealing as which frequency does.
Two more tools come up below. Phase compares the timing of vibration at two points, and the envelope spectrum extracts repetitive impacts that are buried in high-frequency noise.
1. Unbalance: a dominant 1× radial peak
What it looks like. A large peak at 1× in the radial directions, with small harmonics and relatively low axial vibration. The exception is overhung rotors, such as a fan impeller on the end of a shaft, which often show high axial 1× as well. Phase is steady, and horizontal and vertical readings at the same bearing typically differ by about 90°. Because the unbalance force rises with the square of speed, amplitude climbs sharply as the machine speeds up.
What causes it. Material building up on fan blades, erosion or corrosion of an impeller, a broken blade, a lost balance weight, or a replacement rotor that was never balanced.
What to do. Inspect and clean first; buildup is the most common cause on fans. Then field balance in place, in one or two planes. Before balancing, rule out resonance (a structural natural frequency near running speed amplifies 1×) with a bump test or a coast-down. Rule out looseness too, because balancing a loose machine rarely holds.
2. Misalignment: 1× and 2×, with strong axial vibration
What it looks like. It depends on the type, and most real misalignment is a mix of both:
- Angular misalignment, where the shaft centerlines meet at an angle, produces high axial vibration at 1×, often with some 2×. Axial phase shifts by about 180° across the coupling.
- Parallel (offset) misalignment, where the centerlines are parallel but displaced, produces high radial 2×, often larger than 1×, with radial phase shifting by about 180° across the coupling.
What causes it. Installation without precision alignment, thermal growth that wasn’t allowed for, soft foot, pipe strain pulling a pump casing, a settling base, or a worn coupling.
What to do. Laser-align to the coupling manufacturer’s tolerance, with thermal growth offsets for machines that run hot. Check and correct soft foot before aligning, and disconnect suspect piping to see whether the casing moves. A flexible coupling tolerates misalignment, but the bearings on both sides still pay for it.
3. Mechanical looseness: a family of harmonics
What it looks like. A series of running-speed harmonics, sometimes out to 10× or beyond, often with half-order peaks at 0.5×, 1.5× and 2.5×, and a raised noise floor. The vibration is usually strongly directional, and the time waveform looks erratic or clipped rather than smooth. Analysts distinguish structural looseness of a base or frame, which shows mostly 1×, from rocking or a cracked foundation (2×) and component looseness, which produces the full harmonic family.
What causes it. Loose hold-down bolts, cracked grout or welds, a bearing loose on its shaft or in its housing, or excessive internal clearance.
What to do. Check bolt torque and inspect the base and grout. Take readings on either side of each joint (bolt head, foot, baseplate, foundation) to find where the motion changes abruptly. During the next repair, measure the bearing fits. Treat looseness as urgent: it amplifies every other fault, so an unbalance that would be tolerable on a rigid machine becomes destructive on a loose one.
4. Rolling-element bearing defects: non-synchronous peaks
What it looks like. Peaks at the bearing’s defect frequencies, which depend on its geometry and shaft speed:
- BPFO, ball pass frequency outer race, roughly 0.4 × the number of rolling elements × shaft speed.
- BPFI, ball pass frequency inner race, roughly 0.6 × the number of rolling elements × shaft speed.
- BSF, ball spin frequency, and FTF, the cage frequency, which is a little under half of shaft speed.
The key clue is that these are non-synchronous. They are not whole-number multiples of running speed, which separates them from the four other patterns in this article. Inner-race defects usually carry sidebands spaced at 1×, because the defect rotates in and out of the load zone. Rolling-element defects often carry sidebands at FTF.
In the early stages, the impacts are too small to show in an ordinary velocity spectrum. Each impact rings the bearing and its housing at high frequency. Envelope analysis, also called demodulation, band-pass filters that high-frequency region, rectifies the signal and extracts its envelope. A spectrum of the envelope then shows the defect frequency and its harmonics clearly, often months before they appear in velocity.
| Stage | Where it shows | What you see |
|---|---|---|
| Early | Ultrasound and envelope | Small, repeating impacts; often lubrication-related |
| Developing | Envelope and acceleration spectra | Clear defect frequency with harmonics |
| Advanced | Velocity spectrum | Defect peaks with sidebands, rising noise floor, more heat |
| Late | Overall level and temperature | Discrete peaks can blur into broadband noise as damage spreads |
The late stage misleads people: discrete defect peaks can shrink as the damage becomes general, just as the bearing gets close to failure. Trend the overall level and temperature as well as the peaks.
What causes it. Inadequate or contaminated lubrication, overloading, misalignment or unbalance driving the bearing, poor installation, and on inverter-driven motors, electrical discharge through the bearing, which can leave a fluted pattern on the race.
What to do. Confirm the diagnosis with the envelope trend, check lubrication (ultrasound helps here), and plan a replacement at the next convenient stop rather than waiting for the noise. Then look for the root cause, or the new bearing will follow the old one.
5. Electrical faults: 2× line frequency and pole-pass sidebands
What it looks like. Two common signatures on induction motors:
- Stator and air-gap problems produce a peak at twice line frequency: 120 Hz on a 60 Hz supply. On a 2-pole motor this sits right next to 2× running speed. At 3,580 rpm, 2× is 119.3 Hz, so you need resolution finer than a few tenths of a hertz to separate the two. A 200 Hz span with 1,600 lines gives 0.125 Hz per line.
- Broken or cracked rotor bars produce sidebands around 1× and its harmonics, spaced at the pole-pass frequency: slip frequency × number of poles. Using the motor above, synchronous speed is 1,800 rpm and the motor runs at 1,770 rpm, so slip is 30 rpm, or 0.5 Hz. With 4 poles, the pole-pass frequency is 2 Hz, and the sidebands sit at about 27.5 Hz and 31.5 Hz. Overall amplitude often pulses, or beats, at that rate.
The decisive field test is to cut power and watch. Electrical vibration disappears the instant the motor is de-energized; mechanical vibration decays as the machine coasts down.
What causes it. An uneven air gap from eccentricity or a distorted frame (soft foot again), loose stator laminations, shorted turns, unbalanced supply voltage, and cracked rotor bars or end rings, often after repeated hard starts.
What to do. Check supply voltage balance and soft foot first, since both are cheap to fix. For rotor bars, confirm with motor current signature analysis, which shows sidebands around line frequency in the current spectrum. Then schedule electrical testing, and a repair or replacement before the damage spreads to the stator.
The five patterns at a glance
| Fault | Where to look | Signature | First action |
|---|---|---|---|
| Unbalance | Radial | Dominant 1×, about 90° horizontal-to-vertical phase | Clean, then field balance |
| Misalignment | Axial and radial, both sides of coupling | 1× and 2×, 180° phase across coupling | Correct soft foot, laser-align |
| Looseness | Radial, usually one direction | Many harmonics, half-orders, raised floor | Check torque, base and fits |
| Bearing defect | Envelope spectrum | Non-synchronous peaks and harmonics | Check lubrication, plan replacement |
| Electrical | Radial at the motor | 120 Hz, or pole-pass sidebands | Power-off test, check supply and soft foot |
What the overall number tells you
The overall, broadband level answers “how bad?”. The ISO 20816 series, which has replaced several parts of the older ISO 10816 and ISO 7919 standards, defines evaluation zones for broadband vibration on many classes of machine. The zones run from the level expected of newly commissioned machines up to vibration severe enough to cause damage. Use them to set alarm levels and decide urgency. Use the spectrum to decide what to fix.
From patterns to work orders
Pattern recognition is the easy part. The hard part is measuring often enough, on enough assets, to catch faults while they are still cheap. On a predictive maintenance program, our Pulsenode W wireless sensors measure tri-axial vibration from 2 Hz to 10 kHz along with surface temperature and magnetic flux, which gives running speed and therefore slip. Diagnostics flag the likely fault. With Perflens Care at the Verified or Dedicated tier, an analyst checks each alert before it reaches your crew. The corrugator drive example engagement shows what that looks like across 38 assets in a modeled scenario.
Whatever tools you use, the habit is the same: look at the spectrum, name the pattern, and fix the cause rather than the symptom.