A bearing is already a mechanical computer. It integrates load, speed, lubrication and alignment into one output, vibration, and a good analyst reads that output to tell health from harm. A sensorised bearing makes the measurement explicit by embedding the instrument in the component itself, turning a passive part into a data node on the industrial network. The category spans simple integrated temperature elements, full condition-monitoring units with vibration and speed, and edge devices that compute their own alarms.
This guide covers what a bearing can measure from the inside, the commercial families and how they differ, the integration and data questions that belong in the specification, and the cases where a standard bearing with a bolt-on sensor remains the better engineering decision.
What a bearing can measure from the inside
Three quantities carry most of the diagnostic value, and each benefits from being measured at the ring rather than on the housing.
Temperature taken at the outer ring
Temperature is the oldest and simplest measurement. A rising bearing temperature tracks friction, load and lubrication state, and a sensor at the outer ring sits closer to the heat source than any infrared gun or housing-mounted probe, so it responds earlier and with less lag. On a grease-lubricated bearing, temperature and grease life are linked, and a rise of 10 °C can roughly halve the grease life, which makes the ring temperature a direct input to a re-lubrication decision.
Vibration before the housing filters it
Vibration measured at the ring arrives before the surrounding structure filters and attenuates it. The outer ring is the best accelerometer position on the whole machine, which is why handheld route programmes already measure there, and a 3-axis sensor built into the ring can pass the full 10 kHz envelope that a housing probe loses. A sensor built into the ring removes the mounting question, the cable routing and the variation between fitters that makes bolt-on programmes inconsistent.
Speed, slip and revolution counts
Speed and rotation counting from an integrated encoder or a magnetic target add rotation-slip detection and lifetime revolution counting, even on shafts turning at 100,000 rpm. Grease-life calculations and warranty records consume those counts directly, and slip monitoring catches the skidding that ruins lightly loaded high-speed bearings before any vibration signature appears.
The four commercial families
Commercial offerings fall into tiers, and the choice between them is usually a question of where the data is processed and how the power and the signal leave the rotating part.
| Family | What it measures | Power and signal path | Best fit |
|---|---|---|---|
| Measurement bearing | Temperature, vibration | Wired to the machine's monitor | OEM spindles and critical fixed positions |
| Self-powered sensor unit | Temperature, vibration, speed | Harvested energy, radio link | Retrofit and hard-to-reach positions |
| Edge-monitoring bearing | Spectral features, alarms | Local compute, network out | Large fleets without analyst staff |
| Instrumented R&D bearing | Film thickness, forces, wear | Laboratory telemetry | Tribology research and development |
The families differ less in the sensor than in where the measurement is processed and how it leaves the bearing. A practical node reports every 1 to 60 minutes and carries 12 months of trend.
Energy harvesting and the self-powered node
Removing the cable is the feature that makes a sensorised bearing attractive for retrofit work. A self-powered unit harvests energy from rotation or from the machine's own vibration, stores it, and transmits short bursts of data by radio, so it needs no wiring and no battery change on a schedule. The engineering is a genuine constraint as well as a convenience: the harvested power sets how often the unit can report and how much processing it can do locally, so a self-powered node typically ships features and alarm states rather than a continuous waveform.
Edge computing: shipping alarms, not waveforms
A full-bandwidth vibration signal is far more data than a plant can transmit from thousands of bearings, so the higher tiers process on the spot. An edge-monitoring bearing computes 5 to 20 spectral features, compares them with a stored baseline, and transmits an alarm or a health index rather than the raw signal. That saves bandwidth and, more importantly, saves the analyst labour that a raw-signal programme demands. The trade is transparency: an alarm without the underlying spectrum is harder to diagnose when it is genuine and harder to dismiss when it is not.
Integration realities: comms, heat and identity
Comms that survive rotation
Wiring or a radio must survive rotation at up to 100,000 rpm, usually solved with stationary sensing elements on the outer ring or a contactless link, and a 4 to 20 mA loop or a 0.5 W radio is typical. The temperature limit now applies to the electronics as much as to the grease, so a hot position narrows the choice and a 125 °C ceiling is common. The temperature guide explains how heat already limits a conventional bearing, and electronics tighten that budget further.
Identity, spares and lock-in
Replacement changes too: a smart bearing is a configured part, and a maintenance swap must preserve or re-provision the sensor identity, calibration and network address, often within 24 hours. Data ownership and protocol deserve the same contractual attention as the hardware, because a sensor that speaks a proprietary language to one vendor's cloud is a lock-in decision, not just a purchase. None of these issues is prohibitive, but all of them belong in the specification rather than the post-mortem.
Pricing the whole loop, not the sensor
Sensorised bearings buy the same information a disciplined bolt-on programme buys, with better data quality and lower installation labour. The value concentrates in fleets of 20 or more similar machines, in dangerous or inaccessible positions, and in OEM products where consistent factory-installed monitoring differentiates the machine. What the sensor does not do is create information that does not exist, or remove the analysis question: a temperature spike still needs someone to decide whether it means load, lubrication or a dying seal. Buyers should price the whole loop, sensor, network, data pipeline and decision process, because the sensor is the cheapest part of it whether it lives in the bearing or on the housing.
When a stock bearing plus a sensor still wins
Honest engineering keeps the alternative on the table. For a plant with a mature vibration programme, experienced analysts and standard machinery, bolt-on accelerometers on healthy stock bearings remain cheaper and more flexible: the sensors outlive the bearings, and the analysis practice stays uniform across the plant. Sensorised bearings pull ahead when the position makes bolt-on measurement impractical, such as inside a gearbox or on a rotating structure, when the fleet is large and identical, and when a new machine can integrate monitoring from the design stage, wiring included. The decision parallels condition monitoring itself: buy data where the failure cost is high and access is poor, and buy routine where the machines are forgiving. The failure-mode guide covers the diagnostic logic, built on the ISO 15243 vocabulary, that a monitored position feeds. The four conditions below are the ones that tip the decision toward an embedded sensor.
- The failure cost is high. A monitored position pays where a stoppage is expensive and the machine is critical to the process.
- Access is poor. Inside gearboxes and on rotating structures, bolt-on measurement is impractical and the ring is the only sensible place to measure.
- The fleet is large and identical. Comparability turns one bearing's data into a baseline for 100 or more of the same machine.
- Monitoring can be designed in. A new OEM machine can integrate the wiring and the data path from the start, when it is cheapest.
Where the technology is heading
The direction of travel is clear: sensing costs fall, energy harvesting matures, and edge computing shrinks until the alarm-not-raw-data model becomes the default for mid-value machinery. Bearings are a natural host for this instrumentation because they are already where machine health concentrates, and every serious manufacturer now offers some version of the instrumented part. Plants that already run a disciplined vibration programme will find sensorised bearings an evolution rather than a revolution, because the frequencies, the staging of damage and the response decisions do not change when the sensor moves into the ring. The underlying diagnostic logic, defect frequencies, envelope analysis and the temperature-grease interaction, is covered in the grease interval guide and the vibration analysis guide. The ISO 13374 data-handling framework gives the vocabulary for the pipeline that carries a measurement from the ring to a work order.
A specification checklist for a monitored position
Work through the following before committing to a sensorised part, because each item changes the cost and the maintenance model.
| Question | What to fix | Why it matters |
|---|---|---|
| Measured quantities | Temperature, vibration, speed, or a combination | Sets the sensor set and the diagnostic ceiling |
| Power and signal | Wired, harvested, or battery-free radio | Decides whether retrofit is practical |
| Processing location | Raw signal out or alarm at the edge | Balances bandwidth against diagnosability |
| Temperature limit | Electronics rating against the real ring temperature | A short-lived node is worse than no node |
| Identity and protocol | Open protocol, documented calibration, spare provisioning | Avoids vendor lock-in and swap errors |
| Data owner | Who stores, sees and acts on the data | Monitoring without an owner changes nothing |
- Confirm the failure cost and access. A monitored position earns its keep where a failure is expensive and access for bolt-on measurement is poor.
- Choose the family from the table above, then fix the measured quantities and the processing location.
- Check the environment. Verify that the electronics, bearings and grease share a temperature and contamination budget.
- Plan the data path. Decide the protocol, the storage, the alarm thresholds and, above all, the person who acts on the alarm.
- Budget the learning period. The first 12 months tune baselines and thresholds, exactly as it did when bolt-on monitoring was introduced.
- The ring is the best sensor position. Temperature and vibration measured at the outer ring respond earlier and suffer less filtering than housing-mounted measurement.
- Four families, one question. Measurement, self-powered, edge and R&D bearings differ mainly in where the data is processed and how it leaves the part.
- Value is in the loop, not the sensor. Price the sensor, network, pipeline and decision process together, because the sensor is the cheapest element.
- Integration is the real work. Rotation-surviving comms, electronics temperature limits, spare identity and data ownership belong in the specification.
- Bolt-on still wins sometimes. For mature programmes on standard machines, a stock bearing with a disciplined sensor programme remains cheaper and more flexible.
The conventional and monitored families that populate a sensorised fleet remain filterable in the bearing selector, and the deep groove and instrumentation-ready ranges are grouped in the deep groove catalogue and the other bearings range. The maintenance context that sensorised bearings feed into is explored in the condition monitoring overview, and the machinery that most often adopts the technology is described on the industry pages. Programme leaders such as SKF and NSK illustrate how instrumented and conventional parts sit side by side, and the engineering background sits in the technical library.
Frequently asked questions
What can a sensorised bearing measure from inside the ring?
Mainly temperature, vibration and speed. Temperature at the outer ring tracks friction, load and lubrication. Vibration measured at the ring avoids the filtering of the surrounding structure. Speed and encoder counts add slip detection and lifetime revolution counting for grease-life calculation.
How do the sensorised bearing families differ?
They differ less in the sensor than in power, signal path and processing. Measurement bearings are wired to the machine's monitor, self-powered units harvest energy and transmit by radio, edge-monitoring bearings compute alarms locally, and instrumented R&D bearings stream laboratory-grade data like film thickness.
Do sensorised bearings remove the need for condition monitoring analysis?
No. They improve data quality and cut installation labour, but a temperature or vibration change still needs someone to decide whether it means load, lubrication, alignment or a fault. Monitoring without an owner changes nothing, whether the sensor lives in the ring or on the housing.
How does energy harvesting work in a smart bearing?
The unit converts a little of the machine's rotation or vibration into electrical energy, stores it, and transmits short data bursts by radio. That removes the cable but limits how often it can report and how much local processing it can do, so self-powered nodes usually send alarms or features rather than raw waveforms.
What integration issues should a buyer specify?
Four stand out: how wiring or radio survives rotation, the temperature limit that now applies to the electronics as well as the grease, the replacement logistics for a configured part with a sensor identity and calibration, and the data protocol and ownership that decide vendor lock-in. All belong in the specification.
When is a standard bearing with a bolt-on sensor still better?
When the plant already runs a mature vibration programme with experienced analysts and standard machinery. Bolt-on sensors are cheaper, more flexible and outlive the bearings, and the analysis practice stays uniform. Sensorised bearings lead where access is poor, fleets are large and identical, or monitoring can be built into a new machine.