Accelerometer Mounting Methods and Frequency Response

How you attach a vibration accelerometer to a machine sets the highest frequency you can trust. Stud mounting reaches the sensor's full rated range; adhesives come close; magnets cut the usable ceiling to a few kilohertz; and a handheld probe collapses it to roughly 1 kHz or less. Mount wrong and the bearing fault you were hunting never shows up in the spectrum.

This matters because the whole point of vibration monitoring is catching high-frequency defects early, bearing tones, gear mesh, blade-pass, and those live in exactly the band a sloppy mount throws away. The mounting choice is not a convenience decision. It is a measurement decision that quietly decides which failures your program can and cannot see. This guide covers the four common methods, the frequency ceiling each imposes, and how to pick the right one for the reading you actually need.

Why does accelerometer mounting affect frequency response?

Mounting affects frequency response because the accelerometer and whatever holds it to the machine form a spring-mass system with its own resonance. Anything soft or loose in the coupling, a thin air gap, a magnet's contact face, a fingertip on a probe, lowers that mounting resonance. Above roughly a third of the mounting resonant frequency, the sensor stops reporting the true motion of the surface and starts adding its own.

The reference condition is a sensor screwed to a smooth, flat, machined surface with a film of coupling fluid between the two faces. That is how accelerometers are calibrated, and it is the stiffest, most repeatable coupling you can build. Every other method trades some of that stiffness for speed or convenience, and the price is always paid at the top of the frequency band. Low-frequency readings, imbalance at 1x, misalignment at 2x, survive almost any mount. High-frequency readings do not.

What are the main accelerometer mounting methods?

There are four methods in common industrial use: threaded stud, adhesive, magnet, and handheld probe. They trade off in a predictable order, the faster and more portable the mount, the lower the frequency ceiling and the less repeatable the reading. Mounting practice for accelerometers is codified internationally in ISO 5348 which describes each technique and its effect on usable bandwidth.

• Threaded stud. A stud screwed into a spot-faced, tapped hole, with a thin film of grease or coupling fluid at the interface. Reference-grade: it duplicates calibration conditions and gives the widest usable frequency range and the best repeatability. It is permanent and slow to install.
• Adhesive. The sensor, or an adhesive mounting pad, bonded with a stiff epoxy or cyanoacrylate. Nearly as good as a stud when the adhesive is thin and hard; soft adhesives and thick glue lines cost high-frequency response. Good for surfaces you cannot drill.
• Magnet. A two-pole (rail) or flat magnet base. Fast, portable, and the workhorse of walk-around route collection, but the magnet's contact face is a soft spring that pulls the ceiling down to a few kilohertz. Curved and rough surfaces make it worse.
• Handheld probe. A stinger pressed against the machine by hand. The most convenient and the least trustworthy: hand pressure and body motion drop the usable range to roughly 1 kHz or below, and it is not repeatable point to point.

Stud, adhesive, magnet, probe, how do their frequency ceilings compare?

The usable ceiling falls in that order, and the drop from stud to probe is roughly an order of magnitude. Exact numbers depend on the specific sensor, its resonant frequency, and the surface, so treat the figures below as typical ranges published by sensor manufacturers and in ISO 5348, not guarantees for your hardware.

What frequency do you actually need to capture?

The mount you need is decided by the highest defect frequency in the machine, not by habit. Work out what you are hunting before you pick a base. Rolling-element bearing defect tones and their harmonics commonly reach several kilohertz; gear mesh frequency is the tooth count times shaft speed and climbs fast on high-ratio boxes; blade- and vane-pass tones scale with element count. A 30-tooth gear on a 1,780 rpm shaft meshes near 890 Hz at the fundamental, and the useful diagnostic content sits in the harmonics well above that.

Run the arithmetic and the mounting choice makes itself. If your alarm frequencies for a machine top out at 800 Hz, a magnet is fine and a probe might do for a spot check. If you need clean bearing harmonics at 5–6 kHz, only a stud or a hard adhesive bond will get you there, a magnet will smear or hide exactly the tones that give early warning. This is why the same walk-around route that catches imbalance can completely miss an early bearing defect: the mount was never rated for the frequency the defect lives in. For the wider strategy this feeds, see condition-based maintenance and predictive maintenance.

How do you choose and install the right mount?

Match the method to the frequency you need, then prepare the surface as if the reading depends on it, because it does.

1. Set the frequency target first. Calculate the highest defect frequency you need, bearing tones, gear mesh, blade pass, and add margin. That number, not convenience, decides the method.
2. Pick the method that clears the target. Above about 2 kHz, rule out magnets and probes and move to stud or hard adhesive. Below it, a two-pole magnet on a prepared spot is a reasonable route choice.
3. Prepare the surface. Spot-face a flat, smooth pad. Paint, rust, curvature, and roughness all lower the effective ceiling, a magnet on painted, curved pipe performs far below its rating.
4. Use a coupling film for studs. A thin layer of grease or coupling fluid at the stud interface removes the last air gap and pushes the mounting resonance to its highest.
5. Fix permanent measurement points. For trended data, install studs or bonded pads at defined locations and axes so every reading comes from the same spot. Repeatability is what makes trends comparable month to month.
6. Keep the axis and location constant. Radial-horizontal, radial-vertical, and axial readings differ; a wandering probe point mixes them and ruins the trend.
7. Record the method with the data. Note the mount used at each point so an analyst knows the trustworthy band of every spectrum and does not chase a ghost peak that is really a mounting artifact.

How does ISO 5348 frame good mounting?

ISO 5348 is the international standard for mechanically mounting accelerometers, and its core message is the one above: the mounting method determines the usable frequency range, and stud mounting on a clean, flat, coupled surface is the reference against which every other method is judged. It describes stud, adhesive, magnetic, and probe mounting and the bandwidth trade-off each carries.

The evaluation side, what the numbers mean once you have a clean spectrum, is governed by the vibration-severity standards. ISO 20816 is the current series for measurement and evaluation of machine vibration; it consolidated and replaced the older ISO 10816 and ISO 7919 standards, combining housing and shaft vibration criteria into one framework. Good data starts at the mount and ends at a standard-based judgement of whether the machine is in an acceptable zone.

Where mounting fits your reliability program

Mounting discipline is the unglamorous foundation of every vibration program. The most expensive sensor in the world reports garbage above its mounting ceiling, and an analyst who does not know how a point was mounted cannot tell a real fault from an artifact. If you are standing up condition monitoring on rotating equipment, decide the mount per point, prepare the surfaces once, and record the method, that groundwork pays off every route for years.

Vibration is one signal among several. For the stress-wave band above where accelerometers stay trustworthy, see acoustic emission monitoring; for how condition data feeds uptime math, see asset availability and the reliability metrics behind it, MTBF. The harder problem is usually not collecting readings but connecting them to everything else the plant already knows about the machine. That is the layer machine monitoring and platforms like Harmony provide, pulling sensor, PLC, and maintenance data into one operational view so a rising vibration trend lands next to the work-order history and the production schedule, not in a silo. It layers onto the systems you already run, with no rip-and-replace. See how the platform works or read the CLS case study.