Now live: a hybrid process digital twin for amine gas sweetening — it names the fault and the root cause, not just the alarm. Explore the platform →
Data & Security
Where Your Data Lives
The engine reads a file. That is the whole integration.
There is no agent on your control network, no OPC connection, and no requirement to move your historian into a vendor-operated service. A diagnosis is produced without a single outbound call.
What actually crosses a boundary, and what does not.
01
An export, not a pipe
The engine reads a history file. Starting a pilot needs no firewall change, no OPC server and no read access to the control network — which is usually the difference between a project that starts this quarter and one that starts next year.
02
The diagnosis is self-contained
The reference model, the 54 indicators, the fault signatures and the consequence costing all run in one process against local files. No plant data leaves the deployment for a diagnosis to be produced.
03
The assistant is the one exception
qemIntel calls a language model. It sends the retrieved knowledge cards and the values on your current screen — roughly two thousand tokens — never the history file. The provider is an environment setting rather than code, so it can be swapped, or pointed at a model you host.
04
You own what was built
The calibration, the fault library and the causal map describe your unit. There is no per-tag licence and no lock-in that makes leaving expensive.
What we do not claim
We hold no security certification today, and we will not imply one. The platform is currently hosted; because the engine is a self-contained process reading local files, running it inside your own network is a deployment question rather than a rewrite — but it is work we have not yet done for a customer. If your site requires either, that is a conversation to have before a pilot, not after.