Research / Predictive maintenance results / Sample report
Sample report: a bearing back-test on public data
This is the report a client receives after a first predictive-maintenance job, written here for bearings from public test rigs.
- Data
- FEMTO PRONOSTIA and XJTU-SY, public
- Failure labels used
- None
- Updated
- 11 Nov 2026
The data in this report is public test-rig data, so anyone can check it. A client's results report follows the same pattern on the client's own machines.
What was tested
32 bearings run until they failed, from two independent public test rigs: FEMTO PRONOSTIA (17 bearings) and XJTU-SY (15 bearings), under 6 operating conditions.
The indicator starts from each machine's own early recordings in healthy condition. It uses no failure labels.
Headline result
The median degradation monotonicity is +0.96 on FEMTO and +0.98 on XJTU. Monotonicity says how steadily the indicator moves in one direction; 1 is perfectly steady.
The classic indices that most monitoring uses (RMS and kurtosis) trend less steadily on the same bearings; a standard statistical test (Wilcoxon, p < 0.001) shows the gap is not chance. The classic RMS index also changes from rig to rig (−0.12 on FEMTO, +0.90 on XJTU); the indicator stays at +0.96 and +0.98.
Run by run
Each row is a single run. The strong and the weak runs are listed together.
| Run | Monotonicity | Note |
|---|---|---|
| FEMTO, condition 1 | 0.996 | Alarm with about 92% of life still ahead |
| XJTU-SY, 35 Hz / 12 kN | 0.988 | Strong |
| XJTU-SY, 37.5 Hz / 11 kN | 0.986 | Strong |
| FEMTO, condition 3 | 0.545 | Weak |
| FEMTO, condition 2 (bearing 2_3) | 0.411 | Weak |
| FEMTO, condition 2 (bearing 2_1) | 0.305 | Weak |
Checks that the result is real
The bearing results were reproduced bit for bit, and again after a rebuild from the raw data files.
In one control, the indicator was given random noise instead of the vibration data. In another, the time order of the indicator was shuffled. Both controls score far below the real indicator.
| Indicator | FEMTO | XJTU-SY |
|---|---|---|
| Real indicator | +0.963 | +0.976 |
| Random noise unrelated to the signal | +0.134 | +0.127 |
| Time order shuffled | −0.020 | +0.057 |
Each point is scored using only data recorded before it.
Replay through the production pipeline
A run-to-failure test stream was replayed through the production pipeline, with a simulator sending the same packets a field sensor would. Monotonicity was +0.994 and the alarm came at about 31% of life, with about 69% still ahead.
What this report does not cover
Both rigs are accelerated lab tests, not machines in service; a field pilot is the next milestone and the deciding test. On the bearing benchmarks the indicator is not better than one built from the raw FFT spectrum.
What your report would add
On your machines the indicator runs next to your current monitoring. At the end we compare its trend with your maintenance log, machine by machine, and report what it caught, what it missed and where it raised false alarms.
Get this report for one machine
Send recordings from one machine in good condition, and its maintenance log if you have one.