The short answer to digital twin vs simulation: a simulation is a model you run to explore scenarios, usually offline and on its own. A digital twin is a model that stays connected to a specific physical asset, updated with real data from it, and can feed results back. Every digital twin contains simulation, but most simulations are not digital twins.
IBM puts the distinction neatly: a digital twin uses real-time data from its physical counterpart, whereas simulations often exist entirely in the virtual world without an immediate connection to real systems. That one difference, the data link, drives almost everything else: cost, accuracy over time, who owns the model and what you can do with it.
Digital twin vs simulation side by side
| Question | Simulation | Digital twin |
|---|---|---|
| Connected to a real asset? | No, or only through imported data | Yes, continuously or at set intervals |
| Represents | A design or a class of system | One specific machine, line or plant |
| Typical question | What if we add a third packing station? | Why is line 2 slower than last week? |
| Lifetime | Weeks, for a study or project | Years, alongside the asset |
| Main cost | Building and validating the model | Integration, data quality and upkeep |
When a simulation is enough
For many engineering decisions you do not need a live link at all. A discrete-event model of a warehouse, a throughput study of a bottling line before investment, or a robot reach study during cell design are all simulations. They answer a bounded question, then they are archived. Building a twin for these is waste.
- Capacity and layout studies before capital approval
- Virtual commissioning of control software before start-up
- Robot reach, cycle time and collision checks
- Training scenarios that do not need to mirror today's plant state
When you need a digital twin
A twin earns its integration cost when decisions depend on the current state of a specific asset. Typical cases are energy optimisation across a site, condition monitoring of critical rotating plant, scheduling a line that changes product many times a day, or giving remote experts the same view as the operator. The NIST digital twins programme describes them as synchronised virtual models that help manufacturers represent, diagnose, predict and optimise operations.
What is a Digital Twin?
IBM Technology
This video is hosted on YouTube. Playing it loads content from Google, which may set cookies.
Watch on YouTube
One model, four jobs: how a simulation grows into a twin
The most cost-effective route is rarely to build a twin from scratch. It is to build a good simulation for a project, then keep it. A line model built for virtual commissioning already contains the geometry, the I/O map and the control logic. Add a data connection to the live PLCs and historians and it becomes an operational twin. Add VR and it becomes a training twin.
| Stage | What you add | What you can then do |
|---|---|---|
| Planning model | CAD layout at 1:1 scale | Walk-through, layout and ergonomics checks |
| Virtual commissioning model | Behaviour model and PLC connection | Test control software before start-up |
| Operational twin | Live data from PLC, SCADA and ERP | Monitoring, anomaly detection, what-if runs |
| Training twin | VR interface and scenarios | Operators practise procedures and faults safely |
What data an operational twin actually needs
Most twin projects stall on data, not 3D. Before you budget, check that the following exist and are reachable from a secure network segment:
- PLC tags for states, counts, faults and key analogue values, with consistent naming across lines
- A historian or time-series store with enough resolution for the decisions you want to make
- Energy meters at the level you want to optimise: site, line or machine
- Maintenance records and asset IDs that match the tags
- Production orders and product codes from MES or ERP, so the twin knows what was being made
If half of this is missing, the first phase of a twin project is really a data project: adding meters, cleaning tag names and choosing an interface such as OPC UA. That is not wasted money; it is what makes any later analytics possible.
Is the market moving from simulation to twins?
Spending suggests so. Grand View Research valued the global digital twin market at about USD 35.8 billion in 2025 and forecasts 31.1% annual growth to 2033, with North America the largest region (Grand View Research via PR Newswire). Standards are catching up too: ISO 23247 sets out a digital twin framework for manufacturing, and the Asset Administration Shell is the standard twin structure in Germany's Industrie 4.0 projects.
Questions to ask before you commission either
If questions 1 and 2 are a clear yes, plan for a twin. If not, buy a good simulation and design it so the data link can be added later: consistent tag names, a clean I/O map and the real control logic rather than a simplified copy.
How we build them
At Wartens our iUNI team builds models from CAD at 1:1 scale and, because we are automation engineers first, runs them with the same control logic your plant will run. Operational twins add a web-based 3D viewer, analytics, anomaly detection and what-if simulation, integrated with SCADA, PLC and ERP. A typical digital twin project takes 12 to 24 weeks.
To scope the right level of model for your plant, see our digital twins service, read how virtual commissioning fits in, or talk to an engineer.
