Vibration Analysis in Maintenance:
Principles, Methods and What Modern Technology Is Adding

Unplanned machine downtime is one of the most expensive events in industrial operations. In continuous process environments like chemical plants, paper mills or power generation, a single unexpected mechanical failure can cost several hundred thousand euros per hour once production losses, emergency labour and expedited parts are factored in. Most of these failures do not happen suddenly. They develop over days, weeks or months, generating detectable physical signals at each stage of degradation.

Vibration analysis is the discipline that captures and interprets those signals. Decades of application have established its value: it detects developing faults on rotating machinery, including bearing defects, imbalance, misalignment and structural resonance, earlier and more reliably than temperature monitoring, noise or visual inspection. Understanding how it works and what its limits are is the starting point for anyone building or reviewing a condition monitoring programme.

Why Machines Vibrate: The Four Main Sources

All rotating machinery generates vibration. A rotor in perfect condition, perfectly balanced, perfectly aligned and running in undamaged bearings, will still vibrate from residual manufacturing tolerances, minor flow turbulence and small asymmetries in the rotating assembly. As a fault develops, what changes is the amplitude, frequency content and directional distribution of that vibration. Each type of mechanical fault produces a characteristic signature at a predictable frequency, consistent enough across machine populations to be useful for diagnosis.

The four dominant sources of fault-driven vibration in rotating machinery are well-documented and mechanically well-understood. Identifying which one is responsible for elevated vibration is the central diagnostic task in any condition monitoring programme and the starting point for any repair decision.

How Vibration Is Measured: Instruments, Spectra, and Standards

Route-based measurement: a technician acquires a single-point vibration reading at a bearing housing using a contact accelerometer and portable data collector

The standard instrument for industrial vibration measurement is the piezoelectric accelerometer, a contact sensor that converts surface acceleration into an analogue voltage signal sensitive across a frequency range of roughly 10 Hz to 10 kHz. Accelerometers are fixed to bearing housings, gearbox casings or other structural measurement points with threaded studs, adhesive or magnetic bases. The signal is then processed either as an overall velocity level in mm/s rms, or converted via Fast Fourier Transform (FFT) into a frequency spectrum showing how vibration energy is distributed across frequencies.

The FFT spectrum is the primary diagnostic tool in conventional vibration analysis. It decomposes the time-domain waveform into individual frequency components, letting an analyst identify fault sources from the pattern of spectral peaks and their harmonic relationships. ISO 10816 and its successor ISO 20816 define overall vibration severity zones, A through D, based on velocity rms for different machine classes and mounting conditions. Together they provide a standardised framework for comparing machine state against acceptance criteria and triggering maintenance decisions at defined thresholds.

The Limits of Sensor-Based Vibration Monitoring

Point-sensor monitoring has clear diagnostic limits. A sensor measures what is happening at the location where it is mounted, in the direction it is oriented, at the moment of measurement. Everything else is invisible. On a typical industrial machine, there are one to four accelerometers, usually at the bearing housings. The coupling, support frame, baseplate, connected piping and adjacent structural elements are generally not instrumented. Faults that develop in these locations may not produce any usable signal at the monitored points until the defect is already well advanced.

Three categories of problem come up repeatedly in programmes built around point-sensor monitoring. First, structural resonance: because resonance is distributed across the whole assembly, a resonating support frame can cause fatigue damage at stress concentration points far from any sensor, while the bearing reading stays within normal limits. Second, slow-rotating machinery: assets running below roughly 600 rpm, such as cooling tower drives, large gearboxes and agitators, generate vibration at frequencies where standard piezoelectric sensors lose sensitivity. And third, the spatial gap: when a severity alarm fires, the spectrum confirms that vibration at that location has exceeded a threshold. It says nothing about what the machine is physically doing, where the energy is concentrated or what mode shape the structure is taking.

Full-Field Vibration Analysis: Measuring the Whole Structure

A newer approach addresses the spatial gap directly. Vibration amplification is an optical measurement technique based on phase analysis of successive video frames. Instead of measuring at a single point, it estimates the surface displacement of every visible pixel in the image simultaneously, at each operating frequency. The output is an amplified video where structural motion that is normally sub-pixel, invisible to the naked eye, becomes visible: the operational deflection shape of the whole structure at any frequency of interest.

No physical contact with the machine is required. A standard video camera is set up at working distance from the running asset, and the processing is applied after acquisition. This makes the technique useful wherever conventional sensor placement is difficult: machinery in ATEX-classified zones, hot surfaces, assets requiring a permit to access, large structural assemblies where full sensor coverage would be impractical, and equipment running too slowly for standard accelerometers to be reliable.

Combining Both Approaches in a Complete Diagnostic Programme

The two approaches answer different questions at different stages of the maintenance workflow, and they complement rather than duplicate each other. Accelerometers provide the continuous quantitative signal that detects developing faults months in advance and tracks degradation over time. Full-field vibration analysis provides the spatial picture: what the machine and its surrounding structure are physically doing, where the vibration energy is concentrated and what mode shape is present. These are questions that arise naturally from sensor data but that sensor data alone cannot answer.

For a maintenance programme built around rotating machinery, the natural starting point is accelerometer-based monitoring on critical assets: this catches developing faults early and provides the trending data that condition-based maintenance depends on. Full-field analysis becomes most useful when alarms fire without a clear spectral cause, when repairs are not holding, when structural elements fall outside sensor coverage, or when slow-rotating assets sit below the accelerometer sensitivity floor. Together the two approaches cover the time dimension and the spatial dimension of machine vibration, which is the full picture.