Now live: a hybrid process digital twin for amine gas sweetening — it names the fault and the root cause, not just the alarm. Request a demo →

QEM & OPS logo

Getting started

Your Plant, Your Data

Four things to start, and none of them is a new sensor.

Calibrating the model to a new unit is engineering work — we will not pretend otherwise. It is also short, it runs entirely on records your plant already has on file, and it needs no connection to your control system. The whole integration is a file.

Getting the platform onto your unit, in four stagesAn animated diagram. Equipment datasheets, a tag list with a P&ID and a history export are handed over; one calibration pass fits the model to the unit and bakes it to a file. Nothing else crosses the boundary — no control-network agent, no OPC server, no historian migration — and what you are left with is the calibration, the fault library and the causal map for your own plant.datasheetstags + P&IDhistory01What you sendFour things already on filebaked to a file02One calibration passFitted offline, oncenever crossescontrol networkOPC serveroutbound call03The boundaryA file crosses it. That is all.YOURSYour calibrationFault library, causal mapBlind spots, declaredNo per-tag licence04What you keepYours, describing your unit
Twelve seconds, on a loop. The one thing the drawing leaves out is Ask: the assistant calls a language model, and nothing else on the platform makes an outbound call at all.
Open the platform See what it does

What we ask for

Four things, all of them already on file.

  1. 1

    Equipment datasheets

    Design duty, surface areas, passes, clean U, solvent specification. This is what anchors the physics to your unit instead of to a generic flowsheet, and it is the single most valuable thing you can hand over. Without it there is no grey box — only a model that has to guess.

  2. 2

    A tag list and a P&ID

    The tag register with units and descriptions, and the drawing. The drawing becomes the screen your engineers work on, so every tag needs a place on it.

  3. 3

    A history export

    Routine operating history at one-minute resolution, plus the slow laboratory series for solvent strength. The reference train runs on 85 channels over thirteen days — no special test period, no step tests, just normal operation.

  4. 4

    One calibration pass

    Block parameters are fitted offline against that history and baked to a file. After that, queries run on unseen data without refitting, which is what keeps a single-point diagnosis fast enough to sit in front of a board operator.

The boundary

What crosses it, and what never does.

There is no agent on your control network, no OPC connection, and no requirement to move your historian into a vendor-operated service. Starting a pilot is usually a file transfer rather than a firewall change — which is often the difference between a project that starts this quarter and one that starts next year.

What we do not need

  • No step tests and no special operating campaign.
  • No new instrumentation anywhere on the unit.
  • No connection to your control system, and no OPC server.
  • No migration of your historian into anyone’s cloud.

What the engine actually does

  • Reads a history file. That is the whole integration.
  • Runs the reference model, the 54 indicators, the fault signatures and the consequence costing in one process against local files.
  • Produces a diagnosis without a single outbound call.
  • Leaves the calibration, the fault library and the causal map describing your unit — yours, with no per-tag licence.
  • One exception: Ask, the assistant, sends about two thousand tokens of on-screen values and knowledge cards to a language model — never the history file. The provider is a setting, so it can be swapped or hosted by you.

What we do not claim

The part a procurement team will ask about.

  • We hold no security certification today, and we will not imply one. If your site requires one, that is a conversation to have before a pilot rather than after.
  • 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, and we are not going to describe it as if we had.

The method is easiest to judge from the output.

The platform runs on a real amine train with thirteen days of one-minute history behind it. Open a diagnosis and read the evidence for yourself.

Open the platform