Pipeline integrity intelligence
PipeLens turns drone, satellite and SCADA data into findings an engineer can verify, a risk score they can defend, and a work order the field can act on — with the evidence for every step kept next to it.
Drone RGB · change against the last flightThermal · radiometric, in °CSatellite · corridor revisitsSCADA · pressure and flow
How it works
Nothing jumps a stage, and nothing is actioned without a human. That is the whole design: the machine measures and ranks, the engineer decides, and the system learns from the decision.
Drone RGB and radiometric thermal frames, satellite revisits and SCADA tags land through connectors — or a CSV, when that is what the field has.
Detectors measure rather than guess: hotspot delta in °C, change against the previous flight, vegetation encroachment, corrosion, objects in the right-of-way.
Each finding moves a segment's score across five weighted factors, four for probability of failure and one for consequence.
A verified finding raises a work order with a recommended action and a due date, then follows a state machine the field cannot skip steps in.
Every verify and reject is a label. Precision per detector is measured, and the site classifier is retrained on your engineers' own decisions.
Why it holds up
An integrity decision has to survive a regulator, an incident review and the engineer who disagrees with it. Every number here can be traced back to what produced it.
A finding is not a row in a table. It carries the frame it came from, the overlay, the cropped evidence and the measured value that triggered it — so a reviewer can disagree with it on the merits.
Overlay · crop · measured delta · model version
Risk is 100 · Σ wᵢ·fᵢ over five named factors. Every score stores its model version and each factor's contribution, so it can be defended in an audit years later.
Versioned weights · stored contributions · full history
An uninspected segment carries risk for being uninspected. A detector without its model says so instead of failing quietly, and a connector missing its credential names the variable it is waiting for.
No silent degradation
5
stages, from raw frame to closed work order
6
connectors: drone, GIS, CMMS, SCADA, satellite, AI
100%
of findings carry their own evidence
1
outbound path, and it is optional: the AI gateway
Where it runs
PipeLens runs on your own infrastructure: the API, the vision worker, the database and the storage volume are containers you host. The map draws from your own geometry rather than a tile server, and the detectors work with no network at all. The one outbound path is the AI gateway that writes plain-language explanations — leave its key empty and the explanations become deterministic instead of missing.
A new account can read the network — segments, findings, risk and the maintenance board. An administrator grants engineer or field access from there.