Five pipelines. One screen. Zero alarm fatigue.
Each panel shows the most recent six seconds of 100 Hz pressure data for a monitored segment, the rolling deviation from the model's prediction, and the segment's current state. Every ~12 seconds, a leak signature is injected into one pipeline so you can see the full alert path — trace deflection, control-state change, map pin, alert feed entry, with a plausible GPS coordinate.
This is a runnable demonstration. The five waveforms are real 100 Hz pressure data from anonymized customer pipelines (segment names are fictional). The injected leak signatures, GPS coordinates, and alert feed entries are synthetic — designed to show the operator workflow without exposing any real customer data.
The operator workflow, walked through.
Continuous evaluation
The model evaluates each pressure sample against its learned envelope on every cycle. Operational transients (pump starts, valve openings) get absorbed by the envelope, not flagged. The dashed green band is visible: deviation inside it = nominal.
Signature break
When a pressure-signature event breaks through the envelope, the segment immediately changes state to PCR · LEAK SIGNATURE. The trace marker pins the lowest point reached.
Localized + logged
The NPW arrival-time differential between upstream and downstream sensors resolves a GPS coordinate. The map highlights the affected segment. The alert feed gets a permanent timestamped entry, ready for the audit binder.
Send us your trace. We'll show you the same view with your data.
The dashboard above is the operator UI — the same view your control room will see, populated from the same model architecture, the same Modbus / OPC UA register set, the same alert schema you'd integrate.
We'll take a real pressure trace from any segment you operate and return a tailored playback inside 48 hours — the leak event timeline, the model's deviation curve, the predicted GPS, and a recommendation on threshold tuning for your line. No commitment. No NDA needed unless you require one.