Brake disc and caliper control with AI — the part that cannot afford a single escape.

The problem

Every disc and every caliper is the Tier 1's signature in front of the OEM.

Controlling the disc and the caliper is not an average-quality problem: it is a long-tail problem. The line runs fine 99.99% of the time. The problem is part number 73,842 in the batch, the one that goes through with a micro machining defect in the run-out, or the one that comes off the nutrunner with torque two tenths below the window because the bolt batch changed and nobody adjusted the expected curve.

1. The disc — the turning/grinding line holds tight tolerances (DTV, run-out, minimum thickness of the braking band). The operator inspects by sampling. What happens between one sample and the next is unknown — and the line takt leaves no room to stop and count.
2. The caliper — assembly tightens guide bolts, knuckle fixing bolts and, on hydraulic calipers, the fitting. Each operation has a target torque and angle set by the carmaker's manual. The nutrunner knows the value; the ERP never receives it, or receives it in a spreadsheet that lands at the end of the shift.
3. Traceability — the OEM asks for history by VIN for every active safety component. If there is a recall tomorrow, you have to reconstruct which disc, from which melt number, ended up in which vehicle. With no system, that means a spreadsheet and a week of the quality team's time.

The classic system (operator + sampling + checklist + a nutrunner with its own screen) works 99.99% of the time. The 0.01% is what reaches the end customer and, if it turns into a pattern, the recall. The difference between a Tier 1 that renews its contract and one that goes into OEM escalation is that 0.01%.

How it fits the IRIS system

iLEAN does not add a fourth system — it seals the cracks between the lathe, the nutrunner and the ERP.

Disc and caliper control does not fail for lack of data: it fails because the lathe's data, the nutrunner's data and the OEM specification each live on their own island and nobody cross-references them in real time. iLEAN is the putty that fills that gap without asking you to change lathe, nutrunner, MES or ERP.

Edge sees every disc as it leaves the lathe. Connect captures the nutrunner torque at second zero. The agent cross-references it with the OEM specification and holds the part if something does not add up. The person signs — never the other way round.

The three iLEAN pieces applied to brake disc and caliper control:

• Edge — a terminal with machine vision (CNN) over the disc machining line. It inspects 100% of parts as they leave the lathe: run-out, DTV, porosity, surface damage. It acts on the ejector in milliseconds. It works with no network: if the plant loses WiFi, the detect-and-eject cycle keeps running, because what is critical cannot depend on connectivity.
• Connect — captures torque and angle for every nutrunner operation (Atlas Copco, Stanley, Bosch) through whatever channel already exists (OPC-UA, local file, direct integration). It also captures the supplier's bolt batch (the change-notification email, the delivery note) and the latest revision of the OEM specification that arrived by email from the customer. It unifies all of it at second zero.
• Agents — continuously cross-reference the lathe data, the nutrunner torque, the bolt batch, the disc melt number, the VIN of the destination vehicle and the OEM specification in force. If there is a deviation, they hold the part before it moves to the next station and alert the quality engineer. And they build, without anyone asking, the file by VIN that the OEM requests in any IATF 16949 audit.

See the full IRIS architecture →

Before and after

Sampling-based control vs. 100% in-line control with iLEAN

Aspect	Classic control (sampling + checklist)	With iLEAN Edge + Connect + Agents
Disc coverage	Sampling every N parts, gauge/projector	100% inspected in line, in milliseconds
Critical caliper torque	Nutrunner screen, spreadsheet at end of shift	Captured at second zero, checked against the expected curve
Bolt batch change	Email notice, risk of not updating the curve	Connect captures the notice, the agent recalibrates the window
Traceability by VIN	Manual reconstruction in a recall, days	Automatic file per part, audit-ready
Operation without network	n/a	Edge keeps inspecting on cabinet power
IATF 16949 documentation	A folder and many hours of the quality engineer	Dossier by batch and by VIN, continuously

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 with one disc machining line + one caliper assembly line, exporting to a European carmaker with active PPM escalation and IATF 16949 requirements.
• Edge pilot at the disc lathe output + Connect integration with the caliper's smart nutrunner + an agent building the file by VIN. First value expected within a few weeks.
• Indicative payback between 4 and 9 months, depending on the historical frequency of OEM returns and the average cost of rework/return.
• The hard lever is not theoretical PPM — it is a single recall campaign avoided or an OEM escalation reversed. Either one pays for the pilot and the next two years of operation.

And the OEM standard

The automotive quality standard for critical components sits in the order of 25 PPM^[1] — 25 defective parts per million produced. Holding that with a human audit at the end of the line is not feasible: the brain gets tired, the takt never stops. AI in anchored tasks (reading a part, comparing it against the specification) has error rates below 1.5% in the best models^[2], and even so it does not decide alone — it holds and asks for a human signature. The three safety rings are there precisely for this.

^[1] Symestic — automotive quality standard in the order of 25 PPM for critical components.

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

Related topics in automotive: Shock absorber and suspension control · Alloy wheel control · Wheel ABS sensor control.