Skip to content
A dark industrial control panel with lit digital readouts showing live zone temperatures and setpoints

Division 02 · Digital Twin

A model is only as good as what you feed it.

A working copy of your plant that you can experiment on. Try a change, push it until it breaks, and see what it costs before you touch the real thing. Two things feed it: an agreed equipment list, and live readings from the machines themselves.

Test changes safely · Built on an agreed equipment list · Kept up to date

No risk

Try the change on the model, not on the plant

Compared

Two options, side by side, with numbers

Up to date

Kept up to date by real readings

75%

Of projects like this fail because of the data going in

Three quarters of these projects fail, and almost never on the modelling.

The modelling is the part everyone sees: the shapes, the dashboards, the spinning picture in the board pack. That is rarely where it goes wrong.

It goes wrong underneath. The equipment list was wrong, or the readings were averages from a sensor nobody has checked since the day it went in. The model was fine and the answer was still useless.

What we build

A copy of your real machines and processes, so you can try things without touching production. Change a setting, run it forward, and put a cost on the result before you commit.

  • Run a proposed change and compare it against doing nothing
  • Spot trouble early, by watching where the plant and the model disagree
  • 3D views of places too big or too dangerous to walk around
  • A place to test ideas that saves rounds of real world trials

Kept up to date

A model is only right while something keeps feeding it. We take readings from your existing sensors and control systems, and from our own sensors on the machines worth listening to. Real measurements correct the copy.

We learned this on plants that never stop

Most of our experience is in plants that never stop, which is the hardest version of this. Shutting down costs a fortune, the rules are strict, and the equipment has been changed twenty times since the drawings were signed. Any plant old enough to have been altered is the same problem in a different building.

How it runs

Build it, argue with it, then run it.

A model earns its keep the first time it disagrees with the plant and turns out to be right.

01 · Build it

From your equipment list

Built on top of an agreed equipment list, so the model matches the plant as it actually is, not as the drawings say.

02 · Test it

Changes and breakdowns

Setting changes, output targets, breakdowns and layout options all run on the model. You see the results side by side before anything moves on the plant.

03 · Run it

Kept up to date, and watched

Live readings keep the model up to date. The warning you get is the gap between what the model expects and what the plant is doing.

The data decides the outcome. Around three quarters of these projects fail on the quality of the data going in. That is why the equipment list and the sensors sit in the same company as the model. The Sensors division

“The twin was beautiful. It was fed by an hourly average from a sensor last checked at commissioning.”
Reliability engineer, midstream operator.

Next step

Model one unit, then argue with it.

We model one unit, connect whatever live data you already have, and show you where the model and the plant disagree. That gap is usually the most valuable thing we hand over, and you see it before committing to anything more.