Digital twin · engineering

The plant as a model — before you touch it.

A digital twin is not a diagram but a computing model of your plant: fed from the historian, simulated on the real structures of your controller, coupled via OPC-UA. You test parameters, predict wear and run commissioning virtually — without risking a single real part.

OPC-UABeckhoff TwinCATTimescaleDBSoft-PLCPython · scikit-learnR Shiny

1 · Mirror

Live image from the historian — millions of data points, every machine mirrored in real time. The twin knows what the real plant is doing right now.

2 · Simulation

Runs on the real structures and tags of your controller (soft PLC) — structure real, values simulated. Parameters can be tried safely, without risking a single part.

3 · Optimisation

What-if computation and ML suggest better parameters and detect drift before scrap occurs — prediction instead of reaction.

Interactive live simulation

Simplified model of a CNC mill for illustration — the full simulation runs in the demo.

Real operation stable

Override

Move the parameters — the plant reacts live.

Process X-ray

Workpiece NW-1001 Bearing Y

Ghost compare · vibration

Ghost = ideal reference run. When the live curve drifts, the twin sees it early — before scrap happens.

Shift log · events

Ask the optimizer AI assesses the current state — improvements & risks.
What the twin runs on

Historian (TimescaleDB)

Millions of data points as the factual base — the twin doesn't guess, it computes on real histories.

OPC-UA (bidirectional)

Vendor-neutral coupling to PLC and control system — read and, where authorised, write back.

Soft-PLC sandbox

A clone of the control logic with real structures and tag names — experiment without touching the line.

Beckhoff TwinCAT

PC-based control as an open target platform — connectable instead of a proprietary island.

ML & R/Python

scikit-learn and R Shiny for what-if, correlation and remaining-life prediction.

What it's used for

Virtual commissioning

Test control logic against the model before the hardware exists — shorten commissioning, find errors up front.

Throughput & energy

Optimise parameters on the model instead of on live production — no scrap while searching for the optimum.

Predictive maintenance

Remaining life from vibration and temperature — act before a bearing stops the line.

Risk-free training

Operators practise edge cases on the twin that you'd never provoke on the real plant.

First in the model. Then for real.

Every parameter change, every changeover, every new logic — first safely on the twin. That saves scrap, downtime and risk.

See the full simulation ↗ The data base behind it → More process simulations →