Insights / Digital twins
Digital Twin vs Factory Simulation: What Is the Difference?
A factory simulation is a model of how a plant behaves, run under a set of assumptions to answer a question such as whether a new line will hit its target. A digital twin, in the strict sense, is a model of the plant that is kept up to date automatically from live production data and can feed decisions back to it. Simulation answers what would happen if something changed. A twin also answers what is happening now and what is likely to happen next. In everyday use the two terms overlap heavily, and most manufacturers asking the question need the first before they need the second.
Where the terms come from
Simulation has been used in manufacturing for decades to test layouts, line balancing and capacity before money is spent. The digital twin idea is usually traced to Michael Grieves, who described a virtual counterpart linked to a physical product in the early 2000s. It became a manufacturing term as plants gained the sensors and data systems to keep such a counterpart up to date.
A widely cited academic classification (Kritzinger and colleagues, 2018) separates three levels by how data flows between the plant and the model:
| Level | Plant to model | Model to plant | Example |
|---|---|---|---|
| Digital model | Manual | Manual | A simulation built from plant data to test an investment |
| Digital shadow | Automatic | Manual | A model that reads live production data; people decide what to change |
| Digital twin | Automatic | Automatic | A model that reads live data and sends schedules or set points back to the plant |
There is also an ISO framework for manufacturing digital twins, the ISO 23247 series. In practice, software vendors and consultancies use “digital twin” much more loosely, often for any detailed model of a plant, and sometimes for a 3D visualisation with no behaviour behind it at all.
The practical difference
| Factory simulation | Digital twin (strict sense) | |
|---|---|---|
| Data | Plant data, history and assumptions, loaded when the model is built or updated | Live feed from control systems, MES or sensors |
| Lifespan | Built for a decision, then kept and reused as the plant changes | Runs continuously alongside the plant |
| Typical questions | Should we invest, where is the bottleneck, can we change the shift pattern | What should we run next, which order is at risk, why did output drop this shift |
| What it needs | Line data, cycle times, changeovers, downtime, product mix | All of that, plus reliable data integration and someone to act on its output |
| Cost profile | A modelling project or subscription | Ongoing, including integration and data maintenance |
Which one do you need?
Start from the question.
If the question is a decision, a simulation is the right tool. Should we buy this line? Will this layout reach the throughput we promised? Can we close the weekend shift? How big should the buffer tanks be? These questions have an answer. Once you act on it, the model has done its main job, and it can be updated and reused for the next decision.
If the question repeats every day, a connected model starts to earn its keep. What should we run next? Which order will be late? Where is the plant losing time this shift? That value depends on the plant having data good enough to feed the model, and on people who will act on what it says.
Buying a fully connected twin to answer a one-off capital question costs far more than the question needs. Building a sound simulation first is rarely wasted, because a model that represents the plant correctly is the foundation any twin needs anyway.
Questions to ask a vendor
- Does the model represent behaviour (flow, stoppages, changeovers, tank levels, crews) or only geometry?
- Where does its data come from, and how often is it updated?
- Has it been checked against the plant’s real output for a past period?
- Can it test changes that do not exist yet, such as a new line or a different shift pattern?
- Who builds and maintains the model: your team or theirs?
- What does it cost to keep the data connections working?
Where WolfWyse fits
We call our models factory digital twins, as much of the industry does. Under the strict classification above, the models we build for investment decisions sit at the digital model level: we build them from your plant’s own production data and keep them up to date as the plant changes. They represent behaviour, including tanks, batch processes, changeovers, crews and stoppages, so they can test equipment and layouts that do not exist yet.
The same model can then be used for scheduling. At a vinegar plant, the model of milling, fermentation and bottling was used to optimise production scheduling and shift patterns, giving 15% more throughput and 56% less idle time. At a pickles plant, the model of the full facility supported the move from 24/7 to 24/5 operation.
Common questions
Is a 3D model of the factory a digital twin?
Not on its own. A 3D model shows what the plant looks like. A simulation or twin has to represent how it behaves: how product flows, where it queues, what happens when a machine stops. Some tools combine the two.
Does a digital twin need IoT sensors?
Under the strict definition it needs an automatic data link from the plant, which can come from sensors, the control system, MES or a data historian. Many plants already hold enough data in these systems without new sensors.
Is discrete-event simulation the same as a digital twin?
Discrete-event simulation is a modelling method: it moves the model forward from event to event, such as a batch finishing or a machine stopping. Many factory simulations and many digital twins use it underneath. The difference between them is how the model is connected to the plant.
What is a digital shadow?
A model that receives data from the plant automatically but does not send anything back. People read its output and decide what to change.
Related guides
- Capacity planning without spreadsheets: where a single-figure plan overstates what the plant can make.
- How to build the business case for a new production line: one of the commonest reasons to build a simulation.
- Case studies: models built for food, drink and healthcare manufacturers.