1 · Dashboard
Sensor monitoring
History, status, alerts and recommendations for five demonstration sensors. Use the simulation buttons to trigger a leak and watch the analytical module react.
Status
–
Latest ΔI
–
Alerts (sensor history)
–
Recommendation
–
Fluorescence intensity over time
I₀ (baseline) vs I (current) – last 40 readings; dashed lines = risk thresholds
I₀ baseline
I current
Caution
Warning
Alarm
Measurement history
Last 12 readings of the selected sensor
| Time | I₀ | I | ΔI % | T °C | RH % | Level |
|---|
Alerts
Latest non-OK readings across all sensors
2 · Risk calculator
ΔI = (I₀ − I) / I₀ × 100%
Enter a measurement. The threshold class, the analytical module (context rules) and the ML decision tree are compared side by side.
Enter values and press Calculate risk.
3 · AI / ML module
Model report
Hold-out accuracy
Same features, stratified 75 / 25 split
Confusion matrix
Random Forest · rows = true, columns = predicted
Feature importance
Random Forest
Disclaimer. The models were trained on modelled and synthetic data generated by us (
tools/generate_data.py). High accuracy only means the models reproduce the demonstration classification – it says nothing about real-world detection performance. Real measurements from the laboratory stage (in cooperation with Łukasiewicz–PORT) are required for validation.