ABS wheel sensor control with AI — the active safety part where PPM tends to zero.

The problem

The part where a defect is not scrap — it is product liability.

The ABS sensor is not just another brake component: if it fails in the field, ABS is disabled and ESP/ESC goes into degraded mode. For the OEM, that means this part lives with a different level of scrutiny:

1. Overmolded geometry — the air-gap area has tight tolerances. Minimal flash on the overmold changes the sensor's response to the tone wheel.
2. Encapsulation and sealing — a pore invisible to the eye lets water and salt in out in the field. The sensor dies six months later, in the middle of winter, far from your plant.
3. Connector polarity — a reversed assembly passes the static electrical bench, but throws an error code the first time the car is started.
4. Magnet and Hall traceability — a magnet batch change or a Hall chip wrongly assigned turns an 8D into a recall call.

The operator knows it. So does the quality manager. And even so, the classic system (visual + electrical end-of-line) works 99% of the time. That 1% is what the OEM does not forgive in active safety.

How it fits the IRIS system

iLEAN does not replace electrical end-of-line — it takes load off it and gives it context.

The typical ABS sensor defect is not electrical, it is geometric or in the encapsulation. And the information the control needs (BOM, magnet batch, overmolding window, variant changeover) lives on islands and does not reach the decision-maker in time. iLEAN acts as the putty between the line, the ERP, the MES and end-of-line.

Edge sees the part before the electrical bench. Connect reads the BOM and the magnet batch wherever they live. The agent cross-references with the variant and builds the dossier. The person signs — never the other way round.

The three iLEAN pieces applied to ABS sensor control:

• Edge — a terminal with machine vision (CNN) over the carrier leaving the overmolding station. It verifies body geometry, cable length, and the presence and orientation of the connector. Where applicable, dedicated lighting to detect pores in the encapsulation. It triggers the actuator in milliseconds. It works without a network.
• Connect — captures the BOM, the magnet batch and the overmolding window whether they come from the ERP, the MES or the sheet for the new SKU. And it captures what arrives from outside (an OEM email with a tolerance change, a note from the magnet supplier about a dimensional change) at second zero.
• Agent — cross-references the BOM, the line's history, the image of the sensor and the traceability data. If there is a deviation, it holds the part and builds the file — photo, batch, shift, magnet and Hall batches, operator. The person validates and signs; the carrier does not move on its own.

See the full IRIS architecture →

Before and after

Classic control vs. cross-referenced control with iLEAN

Aspect	Visual + electrical EOL	With iLEAN Edge + Connect + Agent
Flash in the air-gap area	Passes if the electrical response is within band	Detected by geometry before EOL
Encapsulation pore	Invisible to the eye, shows up in the field	Dedicated lighting + CNN
Connector polarity	Found in the field (too late!)	Verified in line by image
Magnet / Hall batch	Manual traceability, an Excel sheet	Linked to the batch by the agent
PPAP/IATF dossier	Rebuilt by hand, weeks of work	Automatic per batch, with photos
Response to the OEM's RDR	Days	Seconds

Impact estimate

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.

• Tier-1 brake sensor plant, 2-4 active variants, overmolding + assembly + electrical EOL on the same line.
• Edge pilot over the carrier leaving the overmolding station (camera + lighting + actuator + integration with the ERP/MES BOM and the magnet batch). First value expected within a few weeks.
• Indicative payback between 4 and 9 months, driven by the avoided cost of a single 8D return and by the reduction in field risk (on an active safety part, that lever carries a lot of weight).
• Expected reduction in geometric and encapsulation defects of ≥30% in the first quarter and ≥80% by the sixth month, according to the catalog Pareto.

The industry standard and the reliability of anchored AI

In automotive the standard is on the order of 25 parts per million defective ^[1]. On an active safety part such as the ABS sensor, the OEM pushes even lower. A CNN-equipped camera reaches where the eye cannot.

On the reliability of AI: hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely recontextualizes a piece of data (reading the part and comparing it with the BOM), the best models brought error below 1.5% ^[2]. And even so, on an active safety part, what is critical is never decided alone: iLEAN holds the part and the person signs. The three rings exist precisely for this.

^[1] The 25 PPM standard in automotive quality — Symestic.

^[2] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.