Vibration measurement methods: the right sensor for the job

Vibration measurement methods: the right sensor for the job

Most engineers reach for an accelerometer first. It's the default, the familiar tool, the one that's always in the kit. But defaulting to a sensor before deciding what you actually need to measure is how you end up with real numbers that lead to wrong conclusions. Choosing the right vibration measurement methods starts with the parameter, not the hardware: whether acceleration, velocity, or displacement, that choice shapes every downstream decision, fault diagnosis, severity assessment, and whether the isolation solution you specify will actually work.

Vibration measurement isn't only about condition monitoring. It's also the engineering input that determines which anti-vibration mount, isolator, or shock absorber will perform correctly in your application. When customers bring measured vibration profiles to specialist manufacturers like Anti Vibration Methods (Rubber) Co Ltd (AVMR), that data is the foundation for engineering a precisely matched solution. Without it, mount selection is guesswork dressed up as engineering.

Vibration measurement methods: picking the right parameter first

Before you pick up a sensor, decide what physical quantity is worth measuring. This is not a minor preliminary step. It determines whether your data will be meaningful or buried in noise before you've even started analysis.

The relationship between frequency and signal strength differs for each parameter. Displacement amplitude is largest at low frequencies, making it the right choice for slow-moving machinery. Velocity stays relatively consistent across a broad mid-range frequency band, which is why it's the standard metric for overall severity assessment on most rotating equipment. Acceleration grows strongest at high frequencies, giving you the signal clarity you need for gearboxes, rolling element bearings, and anything running above roughly 3,000 RPM.

Why frequency range governs everything

At very low speeds, an accelerometer reading is so small it sits in the noise floor. A displacement measurement at the same point gives you a clear, meaningful amplitude. Flip the scenario to a high-speed gearbox and that relationship reverses entirely. A displacement reading becomes vanishingly small while acceleration delivers a strong, readable signal. The practical threshold rule: use displacement below approximately five times running speed, use acceleration above roughly 60,000 CPM (cycles per minute), and use velocity for the broad middle ground.

Getting this wrong doesn't produce an obviously bad result. It produces a quietly bad one: clean-looking data with the wrong sensitivity for the fault you're hunting, leading to missed defects or false confidence.

Matching the parameter to your machinery type

For shaft position monitoring in journal bearing machines, use non-contact proximity probes measuring displacement. These machines rely on a fluid film between shaft and bearing, and it's relative shaft position that matters, not casing motion. For general rotating equipment condition monitoring, casing velocity is the standard. It correlates well with destructive energy and underpins ISO severity limits. For gearboxes and rolling element bearings, switch to acceleration. High-frequency fault signatures from bearing defects and gear mesh frequencies yield strong acceleration signals that velocity simply won't resolve cleanly.

One specific case worth noting: occupational hand-arm vibration monitoring uses frequency-weighted acceleration under ISO 5349, not velocity or displacement. If you're assessing worker exposure from handheld tools, that standard is the starting point and the measurement must be configured accordingly.

The four main sensor types and where each one works

Once you've settled on your parameter, the sensor choice follows logically. There are four main types used in industrial vibration measurement methods, and each has a specific domain where it earns its place. For practical guidance on when displacement, velocity, or acceleration should be used, the Vi Institute publication is a useful reference for matching parameter to fault type.

Accelerometers: the workhorse of vibration testing

Piezoelectric (PZT) accelerometers are the default general-purpose tool, and for good reason. They offer a wide dynamic range, work well across the frequency bands most relevant to bearing fault detection and modal analysis, and pair directly with FFT spectrum analysis. MEMS variants are available for lower-cost continuous monitoring applications. The limitation is real: at very low frequencies, the signal-to-noise ratio degrades. Integrating acceleration to derive velocity or displacement also attenuates high-frequency content, so if velocity or displacement is what you actually need, measuring it directly with the right sensor is cleaner than deriving it from acceleration.

Velocity sensors and non-contact displacement probes

Many modern velocity sensors are solid-state and integrate the acceleration signal to produce velocity output internally, minimising external noise pickup. Some legacy geophone-type velocity sensors do have moving elements, so confirm the sensor type before specifying for a permanent installation. Either way, velocity sensors are a sound choice for continuous monitoring on fluid film bearing machines where casing velocity is the relevant metric.

Proximity probes work on the eddy-current principle and are non-contact, high-resolution, and essential for any machine where shaft position relative to the bearing is what you need to track. They're the correct tool for journal bearing machines precisely because they measure relative displacement between shaft and housing, telling you what's happening inside the bearing clearance, which a casing-mounted accelerometer simply cannot.

When an LDV earns its cost

Laser Doppler vibrometry is non-contact, optically precise, and capable of detecting localised vibration that a surface-mounted sensor would average away or miss entirely. It measures velocity directly without adding any mass to the structure, which matters in modal testing on lightweight or delicate components where sensor mass loading would shift the natural frequencies you're trying to measure.

The limitations are genuine. LDV requires a clear line of sight. Surface reflectivity affects signal quality. The cost is significantly higher than any contact sensor. For R&D testing, modal analysis on complex structures, or applications where physical sensor attachment is not possible, an LDV justifies the investment. For routine condition monitoring on standard rotating plant, it doesn't.

Vibration measurement methods: sensor mounting and setup done properly

This is where good sensors produce bad data. Mounting method is not a secondary consideration. It directly sets the upper limit of your usable frequency range, and no amount of signal processing recovers information that mounting resonance has already destroyed.

How mounting method caps your usable frequency range

Stud mounting is the benchmark. It matches the sensor's factory calibration configuration and, per manufacturer calibration data, delivers the widest usable frequency response, reliable beyond 10 kHz for most standard accelerometers, though the precise upper limit depends on the sensor model and calibration specification. Surface preparation matters: the mounting surface should be smooth and flat, the hole drilled perpendicular to the surface. A coupling fluid such as silicone grease or light machine oil further improves interface stiffness and extends high-frequency transmissibility. For an overview of common sensor mounting methods and their trade-offs, see the Wilcoxon guide.

Adhesive mounting is usable up to roughly 7,000 to 10,000 Hz when the adhesive layer is minimal and surfaces are properly prepared, though the achievable upper limit varies with adhesive type, sensor mass, and surface finish. The most common mistake is applying too much adhesive, which introduces compliance at the interface and drops the effective upper frequency limit. Manufacturer-specific recommendations on fastener torque and surface finish can be found in detailed accelerometer mounting technical information.

Magnetic mounting is convenient for walkaround surveys. It is not suitable for bearing fault detection. The magnetic base adds mass to the sensor assembly, lowers the mounted resonant frequency, and creates a filtering effect that rolls off high-frequency content. Depending on magnet type and sensor mass, the usable range can drop to as low as 500 Hz to 6,000 Hz, compared to beyond 10 kHz for stud mounting. For low-frequency checks on large, slow-moving plant, that trade-off is acceptable. For any measurement where you need to see bearing defect frequencies, it isn't.

Sampling rates and anti-aliasing: the numbers that protect your data

The Nyquist rule is non-negotiable: your sampling rate must be at least twice the highest frequency of interest. For general vibration analysis up to 400 Hz, a 1,000 Hz sampling rate is sufficient. For bearing fault detection where fault frequencies can reach 6 kHz, industry practice recommends 25.6 kHz sampling or higher, with margin for the accelerometer resonance and expected carrier bands.

Anti-aliasing filters must be applied before analogue-to-digital conversion. Set the low-pass filter cutoff at or below 0.5 times the sampling rate, with a roll-off of at least 60 dB per octave. Aliased frequencies fold back into your spectrum without any visible indication that they don't belong there. They look like real peaks. Without proper anti-aliasing, you can diagnose faults that don't exist, or miss ones that do. For windowing in FFT analysis, a Hanning window is a widely used choice for vibration signals; it reduces spectral leakage from non-periodic content and keeps the spectrum readable, though alternatives such as Blackman-Harris may be preferable depending on the specific analysis requirements.

Reading your data: FFT analysis and spectrum interpretation

Raw vibration data is just a time-domain waveform until it goes through analysis. FFT (Fast Fourier Transform) converts that waveform into a frequency-domain spectrum, showing amplitude against frequency. This is where the diagnostic information lives.

What FFT analysis reveals in a vibration spectrum

A peak at 1X running speed (one times the shaft rotational frequency) indicates unbalance. A peak at 2X points to misalignment or looseness. Gear mesh frequencies appear at the tooth count multiplied by shaft speed. Bearing fault frequencies, the ball pass frequency outer race (BPFO), ball pass frequency inner race (BPFI), and ball spin frequency (BSF), are geometrically predictable from bearing dimensions and running speed. These are the signatures you're looking for, and a clean, properly acquired spectrum makes them readable.

Envelope analysis for early bearing fault detection

Standard FFT can miss early-stage bearing deterioration entirely. The fault energy at that stage is too small relative to the broadband noise floor to appear as a distinct peak. Envelope analysis, sometimes called the high-frequency resonance technique, addresses this by demodulating the high-frequency carrier signal excited by bearing impacts. The result is a spectrum of the envelope, showing the low-frequency fault repetition rate clearly, even when the raw spectrum shows nothing unusual.

The process involves band-pass filtering to isolate the high-frequency resonance band, computing the signal envelope, then applying FFT to that envelope. This technique often detects bearing deterioration earlier than broadband velocity measurements and is widely used in predictive maintenance programmes. If your condition monitoring system doesn't include envelope analysis, early-stage faults may go undetected until they progress to a point where broadband measurements pick them up.

Standards and compliance limits that industrial measurements must meet

Any vibration measurement used to support engineering decisions, condition monitoring programmes, or occupational health assessments needs to sit within a recognised standards framework. Using the right method but ignoring the applicable standard leaves you with data that can't be acted on with confidence.

Machine vibration severity: ISO 20816 and ISO 10816

ISO 20816 supersedes ISO 10816 and sets procedures and operational limits for evaluating vibration on rotating machinery, expressed in velocity RMS in mm/s. The standard defines four zones, A through D, from newly commissioned machinery in good condition through to severity levels requiring immediate action. For industrial motors above 15 kW, the ISO 10816-3 thresholds still provide a widely used practical reference: below 2.3 mm/s RMS represents unrestricted operation; above 4.5 mm/s RMS signals damage risk. For direct reference to the formal document, see the ISO standard covering vibration measurement procedures and limits. ISO 16063 governs calibration of measurement instruments to ensure traceability back to national standards, which matters whenever your data is being used to make critical decisions about plant safety or maintenance scheduling.

Human exposure limits: ISO 5349 and ISO 2631

ISO 5349-1 and -2 cover hand-arm vibration measurement and evaluation using frequency-weighted acceleration, expressed as an eight-hour energy-equivalent value A(8). ISO 2631 covers whole-body vibration, with a daily action value at 0.5 m/s² and a limit value at 1.15 m/s². Both standards are referenced directly in UK HSE guidance and the EU Physical Agents (Vibration) Directive 2002/44/EC.

If you're running an occupational exposure monitoring programme, your measurements must be structured around these standards from the outset. Collecting general vibration data and applying exposure limits retrospectively doesn't satisfy compliance requirements and won't hold up under scrutiny from enforcement bodies.

From vibration data to the right isolation solution

Clean, well-interpreted vibration data has a use that goes beyond fault diagnosis. It's the direct engineering input for anti-vibration mount selection, and this is where measurement discipline pays off in a tangible way. If you want a deeper primer on the mechanics, see How does vibration isolation work?.

How measured vibration profiles drive mount selection

The natural frequency of the disturbing vibration force, measured in Hz, sets the primary constraint for the isolation system. For effective attenuation, the isolator's own natural frequency should be less than a third of the excitation frequency. That relationship comes directly from the transmissibility equation: a frequency ratio of three or above gives better than 90% isolation above resonance. Load per mount, deflection under static load, and the full frequency range of the disturbance all feed into the specification. None of these inputs can be reliably estimated without measured data. Without accurate vibration profiles, mount selection becomes an informed guess, and informed guesses fail in service.

Why measured data matters to precision rubber manufacturers

When an OEM design engineer or plant maintenance manager brings real vibration data to a specialist manufacturer, the engineering conversation becomes specific rather than approximate. AVMR engineers anti-vibration solutions against real application data: frequency, amplitude, load, and environmental conditions all feed directly into specifying rubber compound hardness, mount geometry, and natural frequency targets for bespoke components. For practical help on selecting the right hardware, see How to select an Anti Vibration Mounts, it explains the basics designers and maintenance teams need to know. This applies to standard catalogue anti-vibration mounts, bobbin mounts, and fully custom shock absorption solutions alike.

Generic catalogue selection based on rough load estimates and a nominal running speed is how under-specified mounts end up failing at six months when they should last six years. Measured data is what separates a solution engineered to match your application from one that's close enough until it isn't.

The measurement comes first

Every decision downstream of vibration data is only as good as the data itself. Choose the wrong parameter and you measure the wrong thing. Mount the sensor incorrectly and you destroy high-frequency content before you've even started. Skip the anti-aliasing filter and your spectrum contains artefacts that look like faults. Ignore the applicable ISO standard and your results have no defensible basis for action.

The logic of sound vibration measurement methods runs in one direction: parameter first, sensor second, mounting discipline third, sampling hygiene fourth, standards compliance throughout. If you're currently selecting isolation mounts without a proper vibration measurement baseline, you're making engineering decisions on assumptions. That's a straightforward problem to fix, and the fix starts before you pick up a sensor.

If you have vibration data and need a manufacturer who can use it properly, Anti Vibration Methods (Rubber) Co Ltd (AVMR) manufactures anti-vibration solutions in the UK and engineers them against real application data, vibration, shock, sealing, and friction. Bring your data. The engineering conversation will be specific. For readers who prefer guidance in French, see How does vibration isolation work? (French).