Permanent-magnet motor control with AI — a rotor defect does not stop the motor from being assembled; it shows up as vibration at the EOL test, when it is already too late.
A cracked, offset or badly magnetized magnet gets assembled all the same — and is discovered as noise or irregular torque at the end of the line. iLEAN inspects insertion at 100%, ties magnetization and balancing to the rotor serial and correlates the EOL signature with its upstream cause. The person decides what to correct.
The bad rotor gets assembled all the same — and the EOL test only shows the symptom.
On a permanent-magnet motor line, a rotor defect has an uncomfortable property: it does not stop assembly. A magnet with a fine crack, a position off by a few hundredths or a magnetization that fell short pass the station without a sound, the rotor is married to the stator, the motor is closed — and the problem shows up at the end, as noise, vibration or irregular torque in the EOL test. Or, worse, it does not show up at the EOL test and the vehicle finds it.
Three things happen at once on these lines and almost nobody puts them together:
- Magnet insertion is controlled by sampling — manual visual inspection or an occasional camera. The magnet cracked during handling, the rotated magnet, the adhesive missing on half a face are not seen at 100%, and consistency changes with every shift.
- Each station's data lives in isolation — the magnetizer keeps its curves on its own equipment, the balancing machine keeps its own, the EOL test keeps its own. Nobody can answer "what did the process measure on this particular rotor?" without disassembling the motor and digging through three systems.
- Unbalance is corrected too late — the balancing machine removes material or adds mass at the end, motor by motor, without anyone asking why the initial unbalance has been climbing for weeks. You pay for the symptom on every part instead of attacking the cause once.
The result is always the same: the EOL test rejects, the motor goes to the analysis bench, somebody strips it and searches by hand — and by then the station that caused the defect has been producing the same way for a whole shift. The knowledge of what to look at first lives in the veteran technician's head; the day he rotates to another line, it leaves with him.
iLEAN does not replace your stations — it stitches together what your stations never join.
The rotor line's problem is not a lack of measurement: the magnetizer measures, the balancing machine measures, the EOL test measures. The problem is that every measurement lives on its own island and no system joins them by rotor. iLEAN acts as the putty that stitches insertion vision, the magnetizer, balancing and the EOL test to each rotor's serial, without asking you to change a single piece of equipment on the line.
Edge inspects insertion at 100%. Connect ties every measurement to the rotor serial. Agents correlate the EOL signature with the upstream process. The person decides what to correct.
The three iLEAN pieces applied to permanent-magnet motor control:
- Edge — machine vision running locally at the magnet insertion station: position of every magnet in its pocket, integrity (cracks, chips) and presence of adhesive, on 100% of rotors and inside the cycle time. It runs locally: if the plant loses its network, Edge keeps inspecting, recording and stopping the defect at the station. What is critical does not depend on WiFi.
- Connect — integration with the magnetizer (curve of every pulse, flux verification) and with the balancing machine (initial unbalance, correction, residual), tying every measurement to the rotor serial. An industrial protocol where one exists; reading the PLC or the results file where it does not. Every motor's complete genealogy is built on its own, station by station.
- Agents — agents that correlate the EOL vibration signature with the upstream process measurements: what the motors that sound alike have in common — a worn insertion tool, a short magnetization cycle, an initial unbalance that has been climbing for days. They point to the station and the specific drift so you attack the cause, not the symptom. Agents never act alone on anything critical — they propose, the line manager decides.
Classic rotor line vs. a line with iLEAN
| Aspect | Isolated stations + sampling | With iLEAN Edge + Connect + Agents |
|---|---|---|
| Magnet insertion inspection | Sampling or manual visual, varies by shift | 100% vision, in cycle, with evidence by serial |
| Magnetizer and balancing data | Each machine keeps its own, isolated | Tied to the rotor serial, queryable instantly |
| Where the rotor defect is caught | At the EOL test, as noise or vibration | At the station where it originates |
| Analysis of an EOL reject | Strip the motor and search by hand | Automatic correlation with the upstream process |
| Balancing rework | Recurring, corrects the symptom part by part | Reduced, the cause is attacked upstream |
| Genealogy per motor (audits, returns) | Manual reconstruction across three systems | Complete and automatic, by serial |
Impact estimate for your plant — to be validated with your numbers.
The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.
- Permanent-magnet rotor assembly line with an insertion station, a magnetizer, a balancing machine and an EOL test with vibration signature; recurring EOL rejects caused by the rotor and process data siloed station by station.
- Pilot: Edge vision at the insertion station + Connect integrating the magnetizer and the balancing machine. First value expected within a few weeks: 100% inspection and genealogy by serial work from day one, before the correlation with the EOL test gets sharp.
- ≥30% fewer EOL rejects caused by the rotor — an estimate to be validated — by stopping the insertion defect at the station and attacking magnetization and unbalance drift at their source.
- Less balancing rework: initial unbalance drops once insertion and magnetization stay inside the band, and the balancing machine stops paying part by part for what the line generates upstream.
- Indicative payback between 5 and 12 months, an estimate to be validated with your volume and your cost of rejects. The hard lever: every motor that never goes to analysis or rework is direct cost removed.
- A recurring benefit that does not enter the ROI but carries weight: the complete genealogy per motor stays as a permanent plant capability — audits and returns are answered in minutes, with evidence by serial.
And the quality manager's reasonable doubt
"What if the AI correlates badly and sends maintenance to the wrong station?" — iLEAN Agents do not act on the line by themselves. They propose; the line manager decides; the intervention is carried out or not. Hallucination is a problem of free generation, not of anchored tasks: in tasks where the AI cross-references process measurements with the EOL signature by serial, the best models brought error below 1.5%[1]. And even so, what is critical goes through the safety rings — the Agents live in the outer ring, they propose inward, and any adjustment on a station is signed by a person. Never the other way round.
[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.
What people ask about permanent-magnet motor control with AI
What does vision inspect during magnet insertion?
Three things on every rotor, on 100% of production: the position of each magnet in its pocket (offset, magnet rotated or not fully seated), integrity (cracks, chips, broken corners — sintered magnets are brittle and the damage can come from handling) and the presence and coverage of the adhesive where the process uses one. iLEAN Edge runs the vision locally, inside the station's cycle time, and ties every result to the rotor serial. A cracked or offset magnet does not stop the motor from being assembled — that is exactly why 100% inspection at the station is the only way to keep it from reaching the EOL test.
How are the magnetizer and the balancing machine integrated?
With Connect, the iLEAN piece that talks to the equipment you already have. From the magnetizer it takes the curve of every pulse (current, energy and the flux verification result if there is one); from the balancing machine, the initial unbalance, the correction applied and the final residual. Every measurement is tied to the rotor serial, not to the shift or the batch. There is no need to replace the magnetizer or the balancing machine: Connect integrates with what is already there — an industrial protocol where one exists, reading the PLC or the results file where it does not.
What does it correlate with the EOL vibration signature?
Everything the rotor carries from upstream: magnet position and integrity measured by vision, the magnetization curve, the initial and residual unbalance from the balancing machine, and tightening or press-fit torques if the line records them. When the EOL test flags a motor for noise, vibration or irregular torque, iLEAN Agents look for which process measurements the motors with that same signature share — and point to the station and the specific drift. You attack the cause (a worn insertion tool, a magnetization cycle that falls short), not the symptom motor by motor.
Is it useful if I already do visual inspection and sampling?
Yes, and that is the most common case. Sampling and manual visual inspection catch the gross defect, but a magnet with a fine crack or an offset of a few hundredths gets through — and consistency across shifts is not the same. 100% vision removes the sampling lottery and leaves evidence by serial: every rotor has an image and a measurement for every magnet. That changes two things: the defect is stopped at the station where it originates, and when a motor comes back from the field, the rotor's complete genealogy is retrieved in minutes, not days.
How much does it reduce EOL rejects?
It depends on the starting point — a line with flux verification after magnetizing and well-controlled balancing does not have the same room as a line with sampling and station-by-station data silos. As an order of magnitude, and always as an estimate to be validated with your data: ≥30% fewer EOL rejects caused by the rotor, and less balancing rework once the unbalance is attacked at its source (insertion and magnetization) instead of being corrected at the end. We send you the estimated ROI in 48h with your line's real numbers.
Tell us your case and in 48h we'll send you the estimated ROI of this AI project for your electric motor line.
We work on your plant's real data, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h See automotive