Industries / Industrial equipment and services / Industrial services / Thermography
Mechanical surveys
A mechanical survey points a heat camera at motors, bearings, couplings, belts and gearboxes, where friction, poor lubrication, misalignment and wear turn into heat.
6
Milestones achieved, from 1985 to 2023.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1985
Belt drives filmed in heat
At SPIE's Thermosense VIII meeting, part of the 1985 Cambridge Symposium and published in March 1986, Tim Carmody of Uniroyal described tests on V-belt drives and on HTD synchronous (toothed) belt drives watched with a thermal video system, and how infrared viewing improves temperature readings in tests of drive efficiency and belt life.
Credit
Tim Carmody, Uniroyal
2008
A standard for machine heat
ISO 18434-1, published in 2008 by the committee that writes the international standards for machine vibration and condition monitoring (ISO/TC 108/SC 5), gives general procedures for thermography of machines. It names the usual causes of abnormal heat as how the machine is run, improper lubrication, misalignment, worn parts and mechanical loading; it sets out two kinds of severity criteria, a temperature difference and a maximum permissible temperature; and it warns that readings on most low-emissivity surfaces, shiny ones that give off little heat radiation of their own, can carry "unacceptable error". Part 2, on reading the images and diagnosing faults, followed in March 2019 and was confirmed in a 2024 review.
Credit
ISO/TC 108/SC 5
2011
Bands for bearing heat
The Bureau of Reclamation's 2011 manual gives severity bands for bearings measured against the surrounding air: 10 to 24 °C above ambient is normal; 25 to 39 °C, investigate and repair when possible; 40 to 69 °C, repair at the next opportunity; 70 °C or more, repair immediately. It credits the table to "Mechanical Applications, InfraMetrics 2004 IR Clinic".
Credit
Bureau of Reclamation
2014
Heat finds what vibration misses
Wael Moussa built a bearing test rig with both vibration and temperature monitoring for his doctoral thesis at the University of Ottawa. He noted that many references held that temperature changes only near the end of a part's life, while vibration analysis is "much less effective" for lubrication problems and gradual wear and gives no reliable sign of how severe many bearing faults are. He seeded real faults at several severity levels and studied the curve of temperature rise before it levels off, rather than the final steady temperature.
Credit
Wael Moussa, University of Ottawa, April 2014
2015
Couplings warm when misaligned
A shaft coupling joins a motor to the machine it drives; when the two shafts are out of line, it flexes every turn and warms. A. R. Mohanty and S. Fatima filmed couplings with a thermal camera at several loads, speeds and amounts of misalignment, alongside accelerometers and a laser vibrometer, and found that misalignment shows in how fast a coupling warms within its time constant, the time it takes to approach a steady temperature, which matched their simple heat-flow model.
LessonA machine checked only after it reaches a steady temperature can hide misalignment that its warm-up shows, so readings have to be taken before steady state.
Credit
A. R. Mohanty and S. Fatima, in IFAC-PapersOnLine, 2015
2023
Motor heat pictures shared
M. Piechocki and colleagues at Poznan University of Technology published thermal images of a squirrel-cage induction motor, the common workhorse motor, running healthy, misaligned with its load by set shifts, or with broken rotor bars, recorded on a test bench at Wrocław University of Science and Technology with a 640 by 512 pixel Workswell camera and a 160 by 120 pixel FLIR Lepton 3.5. They ran their fault classifier on small, low-power computers of the kind that could sit by a machine. The images are free for non-commercial use under a CC BY-NC-ND license, and the record says vibration, current and sound data will follow.
Credit
Piechocki, Pajchrowski, Kraft, Wolkiewicz and Ewert, in Eksploatacja i Niezawodność (Maintenance and Reliability), 2023
Watch for
Heat before vibration
Writing in Canadian Mining Journal, Heidi Schmidt, global sales manager at Opticom Tech, which sells fixed thermal cameras with alarm thresholds for mines, says heat anomalies "often surface early, sometimes days or even weeks before a mechanical failure", ahead of noise or vibration. This runs against the view, summarized in Moussa's 2014 thesis, that temperature changes late in a part's life.
Achieved whena side-by-side test on machines run to failure, published with its data, shows thermal alarms coming before vibration alarms for named fault types.
Claimed by
Opticom Tech, October 2025
Open
01
Time heat against vibration
A public measurement of how many hours of warning heat gives compared with vibration as a bearing wears out. Step one uses public run-to-failure records that log both: for each run, apply fixed alarm rules (for example, vibration three standard deviations above its first hour, and temperature 10 °C above its first hour) and record how long before failure each one fires. Step two runs new bearings to failure on a guarded rig with a thermal camera filming the housing, to add images to the record.
The briefCredit · unclaimed02
Belt tension read by heat
A bench table showing whether a thermal camera can tell a loose V-belt from a properly tensioned one. Run common V-belt sizes on a guarded test drive at tensions from 40% to 120% of the maker's recommended value, at three loads, and record belt and pulley temperatures with a camera and slip with a tachometer on each pulley.
The briefCredit · unclaimed
Open 01 · the brief
Time heat against vibration
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public measurement of how many hours of warning heat gives compared with vibration as a bearing wears out. Step one uses public run-to-failure records that log both: for each run, apply fixed alarm rules (for example, vibration three standard deviations above its first hour, and temperature 10 °C above its first hour) and record how long before failure each one fires. Step two runs new bearings to failure on a guarded rig with a thermal camera filming the housing, to add images to the record.
Why it is still open
Public run-to-failure sets with a temperature channel exist. The curated Awesome Bearing Datasets list records FEMTO-ST's PRONOSTIA set (2012), a 2024 KAIST set that ran a bearing for 128 hours until it passed 85 °C and 9 m/s² of vibration, and Paderborn University's 2024 set under changing load and speed; the only listed set with thermal images, from the Autonomous University of Querétaro and UPC (2025), holds a single drilled hole, not wear grown to failure. Moussa (2014) compared the two methods on seeded faults, not over whole lives. A Crossref search in November 2026 found a 2026 paper whose title promises early warning from vibration and temperature signatures (Nijhavan, not read) and no study that reports lead times across the public sets.
Built with what exists
The BearingDatasets Python library, which loads PRONOSTIA and other sets in one format, plus the KAIST and Paderborn files; Python with pandas; for new runs, a guarded test rig with a radial load on deep-groove ball bearings, an accelerometer, a thermocouple and a radiometric thermal camera such as a FLIR Lepton 3.5 module; Hugging Face to host the recordings.
First step
Load the KAIST run-to-failure files, plot vibration and temperature on one time axis, and note the hour at which each crosses its alarm.
Done looks like
For every public run that has both channels, the hours between the vibration alarm and the temperature alarm, published with the code and the alarm rules; then at least 10 new runs to failure with thermal video, released under an open license, scored the same way.
Open 02 · the brief
Belt tension read by heat
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A bench table showing whether a thermal camera can tell a loose V-belt from a properly tensioned one. Run common V-belt sizes on a guarded test drive at tensions from 40% to 120% of the maker's recommended value, at three loads, and record belt and pulley temperatures with a camera and slip with a tachometer on each pulley.
Why it is still open
Carmody's 1985 paper used thermal video in belt efficiency and life tests, but only its abstract is indexed. Piotr Krawiec and colleagues at Poznan University of Technology (2020) filmed a polyurethane V-belt at three speeds and several brake loads, with tension held at the belt's listed working tension, and found the drive ran best when the two spans of the belt were at similar temperatures. A web search in November 2026 found no study that varies tension and reports temperature.
Built with what exists
A small electric motor and a brake or generator as the load, on a guarded frame; V-belts and pulleys from a parts supplier; a sonic belt tension meter; two optical tachometers; a radiometric thermal camera; Python.
First step
Run one belt at its recommended tension and at 60% of it, at one load, and film both until the temperatures level off.
Done looks like
Three belt sizes, each at six tensions and three loads, run until temperatures settle, with slip, belt and pulley temperatures and all images published under an open license; and the tension, if any, below which the belt runs more than 5 °C hotter than at its recommended tension.