INSICK
ENUA
ua
Platform

Pulse. Predict. Trace.

Three modules that turn flight logs into a fleet that tells you what hurts — and a methodology you can read.

Pulse

A live health score for every airframe.

Flight logs from ArduPilot, PX4, DJI or your own flight controller become a 0–100 Health Score built from vibration, motor current, battery curves and sensor variance. Fleet dashboard: ready, degraded, grounded — at a glance.

312 airframes · 274 ready · 31 degraded · 7 grounded→ See it in Hangar

Predict

Know which machine fails next week — today.

Remaining-useful-life models for motors, ESCs, batteries and propulsion. Alerts ranked by mission impact. Battery intelligence: cycle-level degradation, cold-weather performance, counterfeit and defective cell detection.

9 airframes to pull today · RUL < 5 flights→ See it in Hangar

Trace

Every component batch, held accountable.

Batch-to-field traceability, automated flight-test QA with statistical pass/fail, supplier scorecards and evidence-grade reliability reports for procurement and investors.

Batch LT-0931 · 4.2× fleet median failure rate→ See it in Hangar
The airframe record

One record per airframe, written by every product.

The engine keeps a single history for each machine. Each product adds what it sees at the moment it sees it, and reads everything the others added.

  • BenchAt the factoryFactory passport: reference values and the serial of every component.
  • ProbeIn flightHealth log: per-motor current and RPM, vibration spectrum, cell voltages.
  • PreflightBefore each sortieGo / no-go checks, with the reason and who ran them.
  • HangarAcross the fleetFlights, alerts, work orders, repairs and what they found.
The airframe recordAirframe A-1042
  • Baseline and Health Score history
  • Components, serials and batches
  • Flights and pre-flight checks
  • Alerts, work orders and outcomes

Because every product reads the same record, nothing is measured twice: Preflight checks against the baseline Hangar built, Hangar starts from the passport Bench wrote, and Trace links a field failure back to the batch on that passport.

Which product uses which module

PulsePredictTrace
HangarFleet-wide Health Score and alertsRemaining life and the pull-today listBatch tracing and supplier scorecards
PreflightChecks against the airframe’s own baselineWarns when remaining life is shortFlags components from recalled batches
BenchWrites the first baseline: the factory passport—Records batches; holds lots with flagged components
ProbeFeeds higher-quality signals from the first flightRicher inputs for remaining-life modelsConfiguration and firmware fingerprint
Process

From raw logs to a fleet that tells you what hurts.

  1. 01

    Upload or stream flight logs

    Batch uploads or live MAVLink. Offline-first — connectivity is optional at the edge.

  2. 02

    INSICK builds per-airframe baselines

    Each machine gets its own normal: vibration signature, current profile, discharge curve.

  3. 03

    Anomalies and predictions surface

    Deviations flagged automatically, ranked by mission impact, with a remaining-life estimate.

  4. 04

    Dashboard and reliability reports

    Fleet readiness in one view. Exportable, evidence-grade reports for QA, procurement and investors.

Methodology

How the Health Score is computed. No magic.

Every number INSICK shows traces back to a signal in your own logs and a rule you can read. The model ranks; the physics explains.

→ Weights and thresholds are editable in Hangar
  1. 01

    Signals extracted per flight

    From each log we derive about 40 features: motor current and PWM-to-thrust ratio per motor, vibration RMS on three axes, battery discharge slope and per-cell spread, ESC and motor temperatures, GPS/compass variance, loop-time jitter.

    • Motor current
    • Vib RMS
    • Cell ΔV
    • Temp
    • HDOP
    • Loop jitter
  2. 02

    A baseline for every airframe

    The first 10–15 flights establish this machine's own normal. Baselines are per airframe, not per model — two identical quads fly differently after a week in the field.

    • Per airframe
    • Rolling window
    • Seasonal adjust
  3. 03

    Deviation, not thresholds

    A feature counts as anomalous when it drifts from its baseline over several flights in the same direction. One hot flight is weather; six is a bearing.

    • Trend test
    • Multi-flight
    • Directional
  4. 04

    Health Score 0–100

    Component scores are weighted by mission criticality (propulsion > power > sensors > payload) and combined into one airframe score. 80+ ready, 60–79 degraded, below 60 grounded. Weights are visible and editable per fleet.

    • Weighted
    • Explainable
    • Configurable
  5. 05

    Prediction with uncertainty

    Remaining-useful-life models are trained on labelled failures from the fleet itself and return a range, not a date. Every prediction shows its confidence and the signals that drove it.

    • RUL range
    • Confidence
    • Fleet-trained
The learning loop

Every repair makes the next prediction better.

  1. 1SignalA component drifts from its baseline over several flights.
  2. 2AlertHangar raises an alert with the evidence and a remaining-life range.
  3. 3Work orderA technician inspects or replaces the part.
  4. 4FindingThe technician records what was found: confirmed, not confirmed, or something else.
  5. 5RetrainingConfirmed and rejected alerts become labelled data for the next model release.

Models ship as versioned, signed bundles. A fleet can stay on a model version until its own team approves the next one.

Measured, not claimed

The numbers we report about ourselves.

Every fleet sees in Hangar how well INSICK is doing on its own machines, computed from its own data.

  • Alert precisionShare of alerts a technician confirmed on inspection.
  • Lead timeHow many flights before a failure the first alert came.
  • Missed failuresFailures with no alert beforehand, reviewed one by one.
  • False positivesAlerts closed as false positives, by component class.
Next · HangarSee it in Hangar, the console your operators will open.