Bearing temperature monitoring measures the heat at or near a bearing to detect lubrication loss, overload, misalignment, or cooling failure before the bearing is destroyed. It is a lagging indicator temperature climbs only after friction is already elevated, so its real value is as a protection layer with staged alarm and trip limits, not as an early-warning tool.
That distinction is the whole point of this guide. Plants that treat a temperature sensor as their primary bearing-health tool are usually disappointed: by the time the housing is hot, a vibration analyst reading defect frequencies would have flagged the fault weeks earlier. Temperature earns its keep differently, as a reliable, cheap, last-line protection signal that trips the machine before a bearing seizes and destroys the shaft. This guide covers why it lags, which sensor to use, what is normal, and how to set limits that protect without nuisance-tripping.
Why is bearing temperature a lagging indicator?
Because heat is the result of friction, and friction is already well underway before the housing warms measurably. A developing spall, a thinning oil film, or a small misalignment first shows up as changes in vibration, new impact frequencies, rising energy, while the bulk temperature of a large steel housing barely moves. The housing has thermal mass; it averages and delays. By the time the sensor reads 15°C over baseline, the mechanical damage is often advanced.
None of this makes temperature useless, it makes it a different tool. Some failures announce themselves mainly through heat: a lubrication starvation event, a blocked cooling line, or a suddenly seized component can spike temperature fast. And temperature monitoring is cheap, robust, and easy to wire into a trip circuit, which is exactly what you want for the last line of defense. The mistake is expecting it to do vibration's job.
What makes a bearing run hot?
A temperature rise is a symptom, and reading it well means knowing the short list of things that produce it. When a bearing heats up above its baseline, the cause is almost always one of these:
- Lubrication problems. The most common driver and the fastest. Too little lubricant, too much (over-greasing churns and heats), the wrong grade, a degraded or contaminated film, all raise friction and heat. This is why a temperature rise so often traces back to lubrication.
- Overload or excessive preload. More load than the bearing was sized for, or a fit and axial setting that clamps the bearing too tightly, increases rolling friction and heat generation.
- Misalignment. A shaft or housing out of alignment loads the bearing unevenly and forces the rolling elements to skew, generating heat and uneven wear.
- Cooling or ventilation failure. A blocked cooling passage, failed fan, or lagging that traps heat can raise temperature with nothing mechanically wrong inside the bearing yet, though it will not stay that way for long.
- Advanced mechanical damage. Late-stage spalling, cage failure, or a seizing element finally shows up as heat, by which point vibration has been shouting for a while.
Notice that most of these are conditions a bearing can survive if caught, which is exactly why the alarm tier matters: it buys time to find and fix the cause before heat becomes damage.
RTD or thermocouple: which sensor should you use?
For bearing monitoring, the platinum RTD (Pt100) is the default choice for accuracy and long-term stability; thermocouples are used where wide range or fast response matters more than precision. The difference matters because bearing protection depends on trustworthy trending, and a sensor that drifts undermines every setpoint built on it.
| Property | RTD (Pt100) | Thermocouple |
|---|---|---|
| Typical accuracy | Higher (about ±0.5°C) | Lower (about ±1–2°C) |
| Long-term stability / drift | Excellent, good for trending | Drifts more over time |
| Temperature range | Moderate (ideal for bearings) | Very wide |
| Response speed | Slower | Faster |
| Best fit | Bearing housings, protection trending | Very high temps, fast transients |
Placement matters as much as the sensor. The gold standard is a sensor embedded in the housing so the tip contacts, or sits as close as possible to, the bearing outer ring, in the load zone. A sensor reading oil-sump or ambient air temperature will lag the bearing metal by even more, widening the delay that already makes temperature a late signal. Good contact pressure, protected leads, and consistent placement across identical machines are what let you compare readings and set meaningful limits.
What is a normal bearing temperature, and where are the limits?
There is no single normal number, it depends on bearing type, load, speed, lubricant, and ambient. As a rough orientation, many rolling-element bearings run steadily somewhere below about 70–90°C and fluid-film (sleeve) bearings are commonly monitored with metal-temperature alarms around 90°C and trips around 100°C because babbitt starts to soften as temperature climbs further. Treat these as orientation, not gospel: the useful reference is always the machine's own established baseline.
The single most useful rule: a sustained rise of roughly 10°C above a machine's normal baseline is a warning even if the absolute temperature is still well under any fixed limit. A pump that always runs at 55°C and now holds at 68°C is telling you something changed, lubricant, load, alignment, or cooling, long before it approaches a 90°C alarm. Rate-of-change and deviation-from-baseline catch problems that absolute thresholds miss.
How do you set alarm and trip setpoints?
Setpoints have two jobs that pull in opposite directions: catch a real problem early, and avoid nuisance trips that teach operators to ignore the system or bypass it. The balance comes from baselining, staging, and time delays.
- Establish the baseline first. Record normal operating temperature across the real range of loads, speeds, and ambient conditions for that specific machine. Without a baseline, every limit is a guess, and deviation alarms are impossible.
- Respect the hard physical ceilings. Check the bearing manufacturer's maximum operating temperature and the lubricant's limit (grease and oil degrade with heat). Your trip must sit below the temperature that damages the bearing or cooks the lubricant, those are the real red lines.
- Stage the limits: alarm below trip. Set an alarm that gives a human time to investigate and a higher trip that protects the machine automatically. A common pattern is alarm well above baseline but below the physical ceiling, with trip a defined margin higher.
- Add deviation and rate-of-change triggers. A sustained ~10°C rise above baseline, or a fast rate of climb, should alarm even when the absolute value is still moderate. These catch the developing problems that fixed thresholds miss.
- Use time delays to reject transients. Require the limit to be exceeded for a short, defined period before tripping, so a momentary spike or a bad reading does not stop production. Tune the delay short enough to still protect against a genuine fast event like lube loss.
Set well, this gives you a two-tier response: the alarm routes a person to look while there is still time to plan, and the trip is the automatic backstop when nobody catches it. Both belong in your condition-based maintenance triggers and on the maintenance KPI dashboard alongside vibration.
What the standards and numbers say
- For rotating machinery, bearing temperature is treated as a protection signal with staged alarm and shutdown logic under machinery-protection practice such as API 670 (machinery protection systems), and RTDs are the specified sensor for accuracy and stability. Industry guidance for large electric motors puts sleeve-bearing monitoring near alarm 90°C / shutdown 100°C with rolling bearings generally lower (Stamford AVK, AGN 027: winding and bearing temperature sensors).
- Pt100 RTDs offer roughly ±0.5°C accuracy versus ±1–2°C for thermocouples and far better long-term stability, which is why they are the standard for bearing trending and protection (Industrial Monitor Direct, RTD monitoring for anti-friction bearings). A sustained ~10°C rise above baseline is the widely used early-warning heuristic.
Temperature protection is only as good as the trending behind it, and the trend is only useful if it lands next to the vibration, the lube records, and the failure history. Harmony pulls sensor readings, machine signals, and maintenance data into one operational data layer, so a slow baseline drift becomes a visible pattern instead of a number nobody watched, and it can flag the anomaly and draft the work order for a human to approve. It layers onto the machines and controls you already run, no rip-and-replace; see how it works or the CLS case study. To place temperature in a full program, see predictive maintenance equipment reliability how a rise ties back to lubrication and how heat maps to the bearing failure modes it warns of.