Asset Care 5.0 · Aluminium can line

Reliability decisions, not just dashboards.

AURA turns condition data from the line into what a Reliability Engineer actually delivers: prioritised action plans, failure predictions with confidence intervals, root-cause analyses and a weekly report — every number traceable.

Open the platform → Explore the modules

OEEMTBFMTTRPM Compliance RULWeibullFMECARCA

PLANT · 3D STORY

3D preview unavailable in this browser.

Illustrative walkthrough on a schematic layout — live figures are inside the platform.

IEC induction motor · TEFC Illustrative fields — not an FEA result. Seals modelled in Autodesk Fusion.

drag to rotate

How it works

From a vibration spike to a planned intervention.

One closed loop across Bodymaker, Washer, Decorator, Hélio digital printing, Necker/Flanger and Palletizer. The AI proposes; the engineer decides.

Sense

Vibration, temperature, current and throughput per asset — simulated now, OPC UA later.

Detect

Health Score with contributing factors; Isolation Forest × Mahalanobis anomaly detection.

Predict

Remaining Useful Life with a 95% interval and P(fail ≤ 7 days); Weibull survival per fleet.

Decide

Prioritised plan: what, when, cost vs benefit, parts go/no-go, repair vs replace.

Act & report

Work orders tracked against CMMS/SAP; the weekly reliability report is exported.

Modules

Everything a reliability team runs on — in one workspace.

Explainable by design: each figure shows the data and the formula behind it.

◧

KPIs vs target

OEE (availability × performance × quality), MTBF, MTTR, PM compliance and breakdowns — per asset and for the site.

/api/kpis · /api/site

◔

Health & prediction

0–100 health score with its top factors, and multi-signal RUL that picks linear or exponential degradation by fit.

/api/health · RUL 2.0

✦

AI / ML Lab

Unsupervised anomaly detection, right-censored Weibull MLE (β, η, B10) and a data-driven PM interval.

scikit-learn · numpy

◎

Reliability Copilot

Ask questions in plain language. Answers are grounded in plant data; unverified numeric claims are withheld.

Claude · offline fallback

⬡

Quality Lab

Incoming AA3104-H19 coil gated on the work-hardening exponent n and Lankford anisotropy before it reaches the bodymaker.

Hollomon · r̄ / Δr · FLD

▦

FMECA / RCM & RCA

Failure modes scored S·O·D → RPN drive criticality-based PM; every failure gets a probable cause, evidence and countermeasure.

RPN → PM strategy

◈

Action plan & business case

Weekly priorities ranked by risk and value, with cost-of-downtime and repair-vs-replace economics.

/api/plan · business case

⬢

Digital twin

3D line view with a KUKA KR6 robot model to put asset condition in physical context.

URDF · three.js

✓

Continuous audit

Every source change is checked by static rules and isolated regression tests, with a persistent run history.

/config/audit

Principles

Industry 5.0: the AI amplifies the engineer.

Defensible numbers

Transparent models over black boxes. Each prediction carries a SHA-256 of the exact telemetry it was computed from.

Human in the loop

When the evidence is not clean, a prediction is held for review instead of being reported as healthy. AURA recommends; the engineer decides.

Built to plug in

Runs today on a realistic line simulator; the same logic connects to real sensors and enterprise systems.

Deploy

Mock today, real plant tomorrow — two switches.

KPIs, prediction, RCA, decisions and reports don't change. Only the data origin does.

Sensors

Line simulator with staggered, realistic wear — or live OPC UA tags.

DATA_SOURCE=mock   # simulator
DATA_SOURCE=opcua  # real sensors

Enterprise systems

Simulated MES / CMMS / SAP — or real SAP PM/MM, Maximo and MES connectors.

INTEGRATIONS=mock  # simulated
INTEGRATIONS=real  # SAP PM/MM · MES