Every rotating machine fault has a frequency signature. An imbalanced rotor vibrates at exactly 1x shaft speed. A bearing with a defect on its outer race vibrates at a predictable multiple of shaft speed determined by bearing geometry. Vibration analysis finds these signatures in the measured spectrum before the fault is visible, audible, or thermal.
What Is Vibration Analysis?
Vibration analysis is the measurement and interpretation of mechanical oscillation in rotating and reciprocating machinery. An accelerometer mounted on a bearing housing captures the vibration signal; a data collector or monitoring system converts it to a frequency spectrum using the Fast Fourier Transform (FFT). Faults in a rotating machine produce vibration energy at predictable frequencies. Identifying which frequencies are elevated, and by how much compared to a baseline, is the basis of fault diagnosis.
The technique applies to any machine with rotating elements: motors, pumps, fans, gearboxes, compressors, turbines, and conveyors. It works while the machine runs under normal operating load, which means no production disruption and no disassembly. That combination of non-intrusiveness and diagnostic specificity makes it the dominant condition monitoring method in industrial plants.
Vibration analysis ranges from simple overall level measurements against ISO severity thresholds to detailed spectrum analysis identifying specific bearing defect frequencies, gear mesh harmonics, and resonance conditions. The depth of analysis matched to each asset's criticality determines how much a program costs and how early it detects developing faults.
How Vibration Analysis Works
Sensor Placement
Accelerometers mount directly on bearing housings, as close to the bearing as possible without a flexible mounting pad between the sensor and the metal. Three measurement directions capture different fault types. Radial measurements in the horizontal and vertical directions detect imbalance, misalignment, and bearing defects. The axial direction, parallel to the shaft, detects angular misalignment and axial bearing loads. Most route-based programs take all three directions at each bearing location.
Signal Conditioning and Digitization
The accelerometer produces an analog voltage proportional to acceleration. A data collector or permanently mounted monitor conditions the signal, applies anti-aliasing filters, and samples it at a rate high enough to capture the highest frequencies of interest. For bearing defect detection, sample rates of 25,600 Hz or higher are common. The raw time-domain waveform records the machine's vibration as a function of time.
FFT Processing
The Fast Fourier Transform converts the time waveform into a frequency spectrum. The spectrum shows vibration amplitude at each frequency, expressed in the selected parameter (velocity, acceleration, or displacement). Peaks in the spectrum at specific frequencies indicate vibration sources. Comparing those frequencies against known fault frequencies for the machine's shaft speed, bearing geometry, and gear tooth count reveals which component is generating the energy.
Time Waveform Analysis
The time waveform, before FFT processing, is useful for detecting impulsive events: the sharp repetitive impacts produced by a bearing defect striking the outer race, or a chipped gear tooth contacting its mesh. These impacts appear as periodic spikes in the waveform at the fault frequency. High-frequency enveloping techniques extract the modulation pattern from these impacts to improve detection sensitivity, particularly in the early stages of a bearing defect before the impulse energy is strong enough to appear clearly in a standard velocity spectrum.
Vibration Parameters
Displacement (measured in micrometers or mils) represents the physical distance the machine surface moves. It is most relevant at low frequencies, below 10 Hz, and is used for large slow-speed machines such as paper machine rolls and some turbines.
Velocity (mm/s or in/s, RMS) is the standard parameter for overall machine health assessment. ISO 10816 severity zone classifications are based on velocity. Most general-purpose rotating equipment falls into the frequency range where velocity provides the best sensitivity to developing faults.
Acceleration (g) is proportional to the square of frequency, which makes it increasingly sensitive as frequency rises. Bearing defects and gear mesh faults generate energy at high frequencies, typically above 1,000 Hz, where acceleration is the preferred measurement parameter. High-frequency acceleration measurements (2,000 to 20,000 Hz) using enveloping or demodulation techniques detect early-stage bearing defects that would not yet appear in a standard velocity reading.
Fault Signatures in the Vibration Spectrum
Each fault type concentrates vibration energy at specific frequencies relative to shaft speed and machine geometry. The table below lists the primary fault frequencies and secondary indicators that confirm a diagnosis.
| Fault | Primary Frequency | Secondary Indicators |
|---|---|---|
| Imbalance | 1x shaft speed (dominant in radial) | May appear in axial if angular component present; phase is consistent and repeatable |
| Angular misalignment | 2x shaft speed | 1x also elevated; strong axial component; 180-degree phase difference across coupling |
| Parallel (offset) misalignment | 1x and 2x shaft speed | 2x often dominant; phase difference between drive end and non-drive end; radial direction primarily |
| Bearing outer race defect (BPFO) | BPFO = (N/2) x RPM/60 x (1 - d/D x cosθ) | Sidebands spaced at shaft speed around BPFO; amplitude modulation at shaft speed |
| Bearing inner race defect (BPFI) | BPFI = (N/2) x RPM/60 x (1 + d/D x cosθ) | Sidebands at BPFI +/- 1x shaft speed; modulation visible due to load zone variation |
| Bearing rolling element defect (BSF) | BSF = (D/2d) x RPM/60 x (1 - (d/D x cosθ)²) | Often appears with sidebands at cage frequency (FTF); impulsive in time waveform |
| Gear mesh defect | Number of teeth x RPM/60 | Sidebands at shaft speed of pinion and gear; ghost frequencies from manufacturing errors |
| Mechanical looseness | Integer multiples of shaft speed (1x, 2x, 3x, 4x...) | Half-order frequencies (0.5x, 1.5x, 2.5x) in severe cases; noisy, inconsistent waveform |
| Resonance | Natural frequency of structure or rotor | Excited when an operating frequency (1x, 2x, or gear mesh) coincides with natural frequency; amplitude drops sharply away from that speed |
In the bearing frequency formulas: N is the number of rolling elements, d is rolling element diameter, D is the bearing pitch circle diameter, and theta is the contact angle. All four bearing frequencies (BPFO, BPFI, BSF, and FTF) are published in bearing manufacturer catalogs and can be calculated if the bearing part number is known.
Vibration Severity: ISO 10816
ISO 10816-3 defines four vibration severity zones based on overall velocity (mm/s RMS), measured at the bearing housing. The zones apply to complete machine assemblies running at normal operating speed and load. They do not apply to individual components or to frequency-specific fault amplitudes.
| Zone | Description | Group 1 (>300 kW) | Group 2 (15-300 kW) |
|---|---|---|---|
| Zone A | Newly commissioned, acceptable | Up to 2.3 mm/s (0.09 in/s) | Up to 1.4 mm/s (0.06 in/s) |
| Zone B | Acceptable for continuous operation | 2.3 to 4.5 mm/s (0.09 to 0.18 in/s) | 1.4 to 2.8 mm/s (0.06 to 0.11 in/s) |
| Zone C | Alarm; schedule corrective action | 4.5 to 7.1 mm/s (0.18 to 0.28 in/s) | 2.8 to 4.5 mm/s (0.11 to 0.18 in/s) |
| Zone D | Danger; consider shutdown | Above 7.1 mm/s (0.28 in/s) | Above 4.5 mm/s (0.18 in/s) |
Group 1 covers large machines above 300 kW with shaft heights above 315 mm, typically rigidly mounted. Group 2 covers medium-size machines from 15 kW to 300 kW. ISO 10816-3 also includes Group 3 (pumps above 15 kW) and Group 4 (pumps below 15 kW) with their own boundaries. Zone boundaries are guides, not absolute thresholds; a machine that enters Zone C is not necessarily about to fail, but it warrants investigation to understand the source of the elevated vibration.
Check your machine's vibration level against ISO 10816 zone boundaries instantly.
Vibration Severity CalculatorContinuous vs Route-Based Vibration Monitoring
Route-Based Monitoring
In a route-based program, a technician carries a portable data collector and walks a defined route through the plant, stopping at each measurement point to record readings. Routes typically run monthly for critical equipment and quarterly for less critical assets. The portable analyzer stores measurements and transfers them to a database where software trends the data over time.
Route-based programs work well for assets where fault progression is slow relative to the measurement interval. The main limitation is the gap between visits: a bearing that develops a defect two days after a monthly reading and progresses to failure within three weeks will not be caught until the next route, when it may already be in an advanced state. Route-based programs also miss intermittent faults that occur under specific load conditions that don't coincide with the measurement visit.
Continuous Online Monitoring
Permanently mounted sensors stream vibration data to a monitoring system in real time. Alarm thresholds trigger alerts when overall levels or specific fault frequencies exceed defined limits, giving maintenance teams immediate notification of developing problems regardless of when they occur. Data is available for analysis at any time without requiring a technician to be present at the machine.
Continuous monitoring eliminates the window-of-vulnerability problem in route-based programs. It also captures load-dependent or intermittent faults that would be missed by periodic measurements. The trade-off is higher hardware cost per measurement point: a permanently mounted wireless sensor or wired transmitter at every bearing location costs significantly more than a portable analyzer shared across hundreds of points.
Practical Approach
Most industrial facilities use both methods. Continuous monitoring concentrates on the 20 to 30 most critical assets: those where unplanned failure causes significant production loss, safety risk, or repair cost. Route-based monitoring covers the remaining assets where the cost of a continuous sensor is difficult to justify against the consequence of failure. The decision point is whether the value at risk from a missed fault exceeds the cost of continuous monitoring for that specific machine.
Vibration Analysis for Specific Equipment
Centrifugal Pumps
Pumps are among the most measured assets in most industrial facilities. The dominant fault modes are mass imbalance on the impeller (appears at 1x), bearing defects on both the drive-end and non-drive-end bearings, mechanical looseness from worn wear rings or loose baseplate bolts, and cavitation. Cavitation produces a broadband high-frequency noise floor elevation that is visible in both the spectrum and audible from the pump. Pump vane pass frequency (number of impeller vanes times shaft speed) is a normal running frequency but becomes elevated when there is impeller wear, recirculation, or insufficient NPSH margin.
AC Induction Motors
Motor vibration analysis targets both mechanical and electrical fault modes. Mechanical: bearing defects at the drive-end and non-drive-end, imbalance from rotor eccentricity or dirt accumulation, and looseness. Electrical: rotor bar defects produce sidebands around the fundamental electrical frequency (line frequency, typically 50 or 60 Hz) and its harmonics. Detecting rotor bar faults from vibration alone requires sufficient frequency resolution in the spectrum to resolve the sidebands from the electrical frequency peak. Motor current signature analysis (MCSA) is a complementary technique specifically designed for rotor and stator fault detection.
Fans and Blowers
Imbalance is the dominant fault mode in fans and blowers, and it is often caused by material buildup on the blades rather than mechanical defects. An ID fan handling ash-laden gas from a combustion process can accumulate hundreds of kilograms of material unevenly across the blades. The imbalance signature at 1x shaft speed will grow gradually until the buildup reaches a threshold or dislodges suddenly. Blade pass frequency (number of blades times shaft speed) is a normal excitation frequency; it becomes elevated when blades are damaged, eroded, or when aerodynamic loading is uneven across the wheel. Resonance is a particular concern in large fans where natural frequencies of the fan structure can be excited by blade pass or shaft speed harmonics.
Gearboxes
Gearbox vibration analysis uses gear mesh frequency as the primary diagnostic frequency: the number of teeth on the gear times its shaft speed in Hz. Sidebands at shaft speed intervals around the gear mesh peak indicate eccentricity or load distribution issues. Hunting tooth frequency identifies specific gear tooth pairs that contact repeatedly; it appears at a very low frequency below shaft speed. For planetary gearboxes, the frequency relationships are more complex because the planet gears rotate and orbit simultaneously. Each gear stage adds its own mesh frequency to the spectrum; a multi-stage gearbox can have several mesh frequencies present simultaneously.
Compressors
Reciprocating compressors have a distinct vibration character: the piston strokes produce cyclic force at shaft speed and its harmonics, and the valve events add higher-frequency impulses. Valve faults in a reciprocating compressor are diagnosed by analyzing the time waveform for changes in the impact timing and amplitude at specific crank angle positions corresponding to valve events. Centrifugal compressors are analyzed similarly to pumps: impeller imbalance at 1x, bearing defects at bearing fault frequencies, and aerodynamic excitation at vane pass frequency. Surge in a centrifugal compressor produces a severe low-frequency vibration event that appears as a broadband energy burst across the spectrum.
Setting Up a Vibration Analysis Program
Baseline Measurements
A vibration analysis program requires a reference baseline for each asset: measurements taken when the machine is in a known good condition, ideally after a rebuild or new installation. Without a baseline, there is no reference against which to judge whether a current reading is normal for that specific machine. Baseline measurements must be taken at the same load, speed, and measurement points as all subsequent readings. Document the operating conditions precisely: motor load in amps or percent, pump discharge pressure, fan damper position.
Alarm Level Setting
Two-tier alarm systems are standard. A first alarm (alert) at 150% of the baseline overall level triggers an investigation to determine the source of the increase. A second alarm (danger) at 250% of baseline requires immediate action. These percentages are starting points; the appropriate levels depend on the machine type, the speed at which faults progress, and how much vibration variation is normal for that specific application. ISO 10816 zone boundaries provide a secondary check: regardless of baseline, a reading in Zone D warrants immediate attention.
For frequency-specific fault detection, set separate alarm levels for each identified fault frequency. A bearing outer race defect frequency (BPFO) that is three times the background noise floor at that frequency is an early-stage alert. At ten times the background level, the defect is progressing and a replacement should be planned. These multipliers are guidelines to calibrate against experience with that bearing type and application.
Measurement Documentation
Every measurement needs consistent documentation: asset identification, measurement point designation (drive-end bearing, horizontal direction), sensor mounting method, machine operating conditions at the time of measurement, and the date and time. Inconsistency in sensor placement between readings is one of the most common sources of false trends in route-based programs. A magnetic-base sensor placed in slightly different positions at different visits will produce different readings even if the machine condition has not changed. Drilled and tapped sensor pads, stud-mounted sensors, or permanent sensor mounting eliminates this variable entirely.
Trending and Analysis Cadence
Trend overall vibration levels and key fault frequencies over time. A single elevated reading is less significant than a consistent upward trend over several months. Many faults produce step-changes in vibration (a sudden amplitude increase) that are clearly visible in a trend plot. Others grow gradually. The rate of change matters: a rapidly increasing bearing defect frequency amplitude warrants earlier intervention than a slowly increasing one. Review trend data at a regular cadence, not only when an alarm triggers.
Calculate BPFO, BPFI, BSF, and FTF for any bearing using its catalog dimensions.
Bearing Fault Frequency CalculatorFrequently Asked Questions
What is vibration analysis used for in rotating machinery?
Vibration analysis detects mechanical faults in rotating and reciprocating equipment before they cause failure. Common faults it identifies include mass imbalance, shaft misalignment, bearing defects on the inner race, outer race, or rolling elements, gear mesh defects, mechanical looseness, and structural resonance. Each fault type produces vibration energy at predictable frequencies, making it possible to identify the specific fault from the measured spectrum.
What frequency does imbalance appear at in a vibration spectrum?
Imbalance produces vibration at exactly 1x shaft speed (also written as 1x RPM). This is the most common vibration fault and typically dominates the spectrum in the radial directions (horizontal and vertical). If the imbalance has an angular component, a 1x peak will also appear in the axial direction.
What is the ISO 10816 vibration severity standard?
ISO 10816-3 classifies vibration severity into four zones: Zone A is acceptable for newly commissioned machines, Zone B is acceptable for long-term continuous operation, Zone C indicates a condition that warrants attention and corrective action should be scheduled, and Zone D indicates vibration severe enough that the machine may be considered for shutdown. Zone boundaries are defined in velocity (mm/s RMS) and vary by machine power class.
How is bearing fault frequency calculated?
Bearing fault frequencies are calculated from bearing geometry and shaft speed. The outer race defect frequency (BPFO) equals (N/2) x RPM/60 x (1 - d/D x cos theta), where N is the number of rolling elements, d is rolling element diameter, D is pitch circle diameter, and theta is the contact angle. The inner race defect frequency (BPFI) uses the same formula but with (1 + d/D x cos theta). Bearing manufacturers publish these geometry values in their catalogs.
What is the difference between velocity and acceleration in vibration measurement?
Velocity (mm/s or in/s RMS) is the standard parameter for overall machine health assessment and is the basis of ISO 10816 zone classifications. It is most sensitive in the mid-frequency range (10 Hz to 1,000 Hz) where the dominant faults in most industrial machines fall. Acceleration (measured in g) is more sensitive at high frequencies and is preferred for detecting bearing defects and gear faults, which produce energy above 1,000 Hz. Displacement (micrometers or mils) is relevant at very low frequencies, typically below 10 Hz, and is used for large, slow-speed machines such as paper machine rolls.
How often should vibration measurements be taken?
For route-based programs, measurement intervals typically range from monthly for critical machines to quarterly for non-critical assets. The correct interval depends on how fast a developing fault progresses to failure for that equipment type. Bearings in high-speed, heavily loaded machines can fail within weeks of producing measurable defect frequencies, while a slow-speed machine bearing may show fault signatures for months before failure. Continuous monitoring eliminates this interval problem entirely by measuring in real time.
Continuous Vibration Monitoring for Critical Assets
Tractian's wireless sensors mount on bearing housings and capture continuous vibration and temperature data. The platform runs spectrum analysis automatically and alerts teams when fault frequencies rise above baseline.
See Mechanical Monitoring