An avionics failure in service costs more than any MRO investment — you avoid the AOG beforehand with cross-referenced data, not afterwards with spare-part stock.

The problem

The signal of the avionics failure was there — split across three systems nobody cross-references in real time.

The airworthiness manager at an operator or a Part-145 MRO knows the pattern: when an unscheduled AOG appears, the postmortem almost always reveals that the signal was there. A BIT (Built-In Test) that tripped intermittently on the last few flights, an ACMS parameter that drifted outside the expected envelope, a report from the previous shift's technician noted down in a spreadsheet. The data existed. What was missing was the brain to cross-reference it with the individual aircraft's history and say "this is degrading".

The classic PHM model sits with the OEM, for the entire fleet. The specifics of the individual aircraft — its history, its routes, its landing severity — live with the operator. And the reality of the shop (what was replaced and when, what stock there was, which technician signed) lives in the MRO ERP. Three systems, three owners, and between the three there are cracks that AOGs slip through.

How it fits the IRIS system

iLEAN is not another MRO IT tool — it is the putty between the OEM's BIT, the operator's telemetry and the shop's MRO ERP.

The problem in aerospace predictive maintenance is not a lack of data: it is data on three islands that do not talk to each other in time. iLEAN acts as the putty that fills the cracks between what the aircraft reports (BIT, ACMS, QAR), what the operator knows (routes, landing severity, event history) and what the shop records (interventions, stock, the technician who signs). It replaces neither the MRO ERP nor the OEM's system — it stitches them together.

Edge captures shop-floor sensors during the intervention. Brain keeps the PHM model per individual aircraft. Agent assembles the work order and the airworthiness dossier. The Part-66 person signs — never the other way round.

The three iLEAN pieces applied to predictive MRO:

• iLEAN Edge — a terminal in the hangar and at the inspection points. It captures test-bench measurements (vibration, temperature, pressure), instrument readings during the inspection, and, via OCR, the aircraft's own displays when a direct interface is not possible. It works with no network — if the hangar loses connectivity during an intervention, it keeps capturing.
• iLEAN Brain (PHM model) — a PHM model per component and per individual aircraft, fed by the aircraft's BIT (through the operator's Connect), post-flight ACMS telemetry, and data from every intervention (the shop's Connect). It maintains a remaining useful life (RUL) projection per component, raises an alert when aircraft X diverges from its fleet or when a part approaches a threshold, and proposes a maintenance slot to the shop's planning team.
• iLEAN Agent (Documentation) — an agent that assembles the work order once a predictive intervention is accepted: the ATA chapter affected, the component to replace, the qualified Part-66 technician available, stock confirmed, airworthiness dossier ready for signature. The technician does not assemble paperwork — they review it, carry out the task and sign.

See the full IRIS architecture →

Before and after

Reactive + calendar MRO vs. predictive MRO with iLEAN

Aspect	Classic MRO (calendar + reactive)	With iLEAN Edge + Brain + Agent
PHM model	From the OEM, for the fleet	Per individual aircraft, with real data
Intermittent BIT signal	Ignored until it becomes persistent	Brain cross-references it with history and raises an early alert
Unplanned AOG	Reactive: AOG, aircraft off the roster, delays	Anticipated: intervention in a normal slot
Work order	Manual, the technician hunts for documentation	Assembled by Agent, the technician reviews and executes
Airworthiness dossier	Rebuilt at closeout	Live from minute one, ready for signature
Spare-part pool cost	Over-stocked because of uncertainty	Predicted demand — stock sized to the real risk

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 operation. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Part-145 MRO or operator with a fleet in the order of dozens of aircraft, ATA chapters with a history of unscheduled AOG over the last few years, a significant spare-part pool (rotable avionics components).
• Brain + Connect pilot on one ATA with AOG history (the operator's failure Pareto). First value expected within a few weeks: a PHM model per aircraft running on that ATA, early alerts on the dashboard.
• Indicative payback between 4 and 9 months. Levers: unplanned AOG down ≥30% on the ATAs covered, lower pool cost through better demand prediction, less paperwork for the Part-66 technician, an airworthiness dossier ready without rebuilding it.
• The hard lever: a single mid-sized AOG avoided per year in a mid-sized fleet pays for the whole system.

And the airworthiness Postholder's reasonable doubt

"What if the AI proposes deferring an intervention and the component fails?" — hallucination is a problem of free generation, not of anchored tasks. Brain does not decide to defer anything: it produces a RUL estimate with its confidence interval, raises an alert when it approaches a threshold, and the intervention decision still belongs to the Postholder or the qualified Part-66 technician. In anchored tasks, the best models brought error below 1.5% ^[1]. And iLEAN's three safety rings are designed for precisely this: the European AI Act requires human oversight in high-risk AI, and iLEAN's architecture guarantees that critical airworthiness operations are executed only by a qualified person.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.