From a vibration signature to a work order, closing the loop
A worked example of the OT→IT thread: a coolant pump degrades, and the platform turns the physics into a governed, verified business action.
Abstractions are easy to nod along to. Here's the concrete thread, one condition to one action.
1. Sense
A coolant pump's vibration RMS has been climbing for weeks. Ingested live over MQTT Sparkplug, the signal crosses the ISO 20816 alarm zone. In the frequency domain, the dominant peak and its 2× harmonic are drifting upward, a classic outer-race bearing-defect signature.
2. Diagnose
The visible symptom often isn't the cause. Here it is, but the platform confirms it by traversing the asset twins in the knowledge graph, ruling out downstream effects and localizing the failing pump with a graph-grounded confidence score.
3. Propose & approve
The optimizer proposes a specific, bounded action: a bearing-service work order in the CMMS, at medium risk. Because it's consequential, it waits for a human, who sees the agent's track record and a recommendation alongside the request.
4. Execute & verify
On approval, the work order is filed through the proper IT channel, never a direct write to the floor. The outcome is measured; if it helped, the result is recorded to the ledger that earns the agent its next increment of autonomy. If it didn't, it's rolled back.
Sense → diagnose → propose → approve → execute → verify. One thread, fully governed, with a safety line it can't cross.
No IT-only tool can start this thread, it never sees the pump. No OT-only tool can finish it, it never reaches the CMMS. Running the whole loop, governed, is the point.