This is Case 5 of the five-part RCA series — and the trickiest of the five, because it’s the case where the cause is the driver, not the vehicle, and the engineering solution has to bend toward driver behavior rather than against it.
The complaint: brake noise and reduced initial bite from the friction brakes in a subpopulation of EVs. The Pareto, predictably, points at a rotor supplier lot. The investigation has to refuse it (again) and look at how the affected vehicles are being driven. The pattern, once you look for it, is clear: the affected vehicles are the ones whose drivers commit to one-pedal driving — using regenerative braking for nearly all decelerations, never touching the friction brake pedal. The rotors, never scrubbed clean by friction-brake engagement, accumulate a layer of rust. The noise and reduced bite are the rust, not a defect.
The fix is a software change, not a hardware recall. The vehicle implements a periodic automated brake-scrub cycle — a brief, low-pressure friction-brake engagement during low-speed coasting that wears the rust off without the driver noticing. Plus a one-time driver-education prompt suggesting an occasional manual brake use in wet weather. The combination drops the complaint rate to baseline within a quarter.
This is the usage/behavior RCA pattern — the cause is a legitimate, even desirable driving style, and the fix is a system-design accommodation rather than a fault-correction. It’s also the case that requires the most discipline to find, because the “obvious” answer (the rotor lot) is institutionally easier than the right answer (“it’s how people are driving”).
What it covers
Eleven sections, about seventeen minutes of careful reading.
§ 1–2 — The complaint. Brake noise, reduced initial bite. The customer-service signals. The first-pass engineering response.
§ 3–4 — The rotor-lot impostor. The Pareto fit. The cross-validation that breaks it (some vehicles in the suspect lot are fine; some outside it are not).
§ 5–7 — The behavior dimension. Looking at usage data alongside the symptom. The one-pedal-driving correlation that emerges when you stratify by how often the friction brake is pressed rather than by which supplier lot the rotors came from.
§ 8–9 — The rust mechanism. Why a never-scrubbed rotor accumulates rust faster than the design assumed. The humidity-temperature interaction that makes it worse in some climates.
§ 10–11 — The software fix. The automated brake-scrub cycle. The driver-education prompt. The validation: complaint rate drops to baseline within one quarter post-OTA.
Read it
The report lives at its own URL with a warm-paper engineering-document layout — plum accent for the usage-behavior motif, classical RCA-report styling throughout.
Code & data
↓ usage-rca-code-and-data.zip — the Python generator (brake_corrosion_rca.py), the synthetic 500-vehicle dataset (brake_fleet.csv), and a README.
The full series
Case 5 of 5. Read the rest in order:
- Battery Degradation Case Study — software / firmware
- Preconditioning-Mode NVH — electromagnetic / structural
- Module Weld Resistance — tooling / manufacturing
- HV Isolation Faults — environmental (condensation)
- Brake Rotor Corrosion — (this post) — usage / behavior
Followed by the agentic framework that automates this method, and its worked walkthrough.
← Back to Autonomy