Wondering what is a digital twin in manufacturing? It is a live virtual copy of a machine, line or process, kept in step with the real thing, so you can test and monitor changes without touching production.
This guide is part of our series on digital transformation in manufacturing.
| Item | Value |
|---|---|
| Main standard | ISO 23247 (Digital Twin Framework for Manufacturing) |
| Part 1 title | ISO 23247-1:2021, overview and general principles |
| Part 1 publication date | Oct 22, 2021 |
| NIST paper use case scenarios | 3 |
| Part 1 scope | Terms and definitions, plus requirements of the framework |
Figures come from the NIST publication page and the Genorma listing for ISO 23247-1. Standards and their status can change, so check the source before you rely on them.
A digital twin in plain words
A digital twin is a software model of a physical thing in your plant. The model is not a one-off drawing: it is linked to the real asset and updated as the asset changes.
That link is what separates a twin from an ordinary simulation. A simulation is usually run on demand; a twin stays connected to the machine, so the answer stays relevant.
The term is used loosely. NIST says that “there remains a lot of confusion about the digital twin concept and how it can be implemented in real manufacturing systems, especially among small-to-medium-sized enterprises.” Expect vendors to use the same words for quite different products.
What a twin is made of
Most manufacturing twins combine the same few building blocks. The names vary by vendor, but the roles are consistent.
- The physical item: a CNC machine, a robot cell, a pump, or a whole line.
- Data collection: sensors and controllers that report what the item is doing.
- The model: software that represents the item’s behaviour and state.
- The user layer: dashboards or tools where people review results and decide.
The data often comes from the same control layer you may already run. If that is unfamiliar, read about what is a scada system to see where plant data starts.
Where the standard comes in
Because definitions differ, a shared framework helps. ISO 23247 is titled “Digital Twin Framework for Manufacturing”, and Part 1 was published on Oct 22, 2021, according to the Genorma listing.
Its scope covers an overview and general principles, including terms and definitions and the requirements of the framework. In short, it gives buyers and engineers common vocabulary.
NIST’s paper introduces the ISO 23247 series in detail. It explores three digital-twin-in-manufacturing use case scenarios, and says the designs can be implemented at the NIST Test Bed or on any manufacturing shop floor.
Practical uses on the shop floor
NIST’s own examples are a useful guide. The three use case scenarios in that paper are meant as practical guides, and the designs can reportedly be put in place at the NIST Test Bed or on any manufacturing shop floor. The steps below are a simple way to start.
- Pick one asset that causes pain, such as a machine with frequent stops.
- Decide what you need to see: state, condition, or performance.
- Connect the data you already have before buying new sensors.
- Review the twin’s output with the people who run the machine.
Maintenance is a natural partner, because twins reflect asset condition. Your maintenance records, often kept in a CMMS, are a good data source; see what is a cmms if you are new to it.
How a twin differs from related tools
Plants already run many systems that touch the same data, so it helps to separate them. Each answers a different question.
- Monitoring dashboards show what is happening now. A twin adds a model that can explain or predict why.
- Simulation software tests a design or a schedule. A twin keeps that model tied to live conditions.
- Maintenance systems record work done and work due. A twin can use that history as context for condition.
- 3D design files describe what was intended. A twin tracks what actually exists on the floor.
None of these is replaced by a twin. In most plants the twin sits on top of them and reuses their data.
Common pitfalls
Some common pitfalls are worth watching for when you plan a first project.
- No clear question. A twin built “to see what happens” rarely produces a decision.
- Poor data. If sensor readings are missing, mislabelled or out of sync, the model drifts away from the machine.
- Too wide a scope. Modelling a whole factory first makes it hard to prove value or find errors.
- No owner. Someone on the operations side needs to keep the model current as equipment is changed.
- Vendor lock-in. Proprietary formats can make it costly to move later.
Using a published framework as a checklist for requirements can reduce the last problem, because it gives you a neutral vocabulary for specifying what you need.
What good looks like for a first project
A sensible first twin is small, boring and measurable. Choose a machine whose behaviour operators already understand, so they can tell quickly whether the model matches reality.
Agree in advance how you will judge success. That might be fewer surprise stoppages, faster changeovers, or simply a clearer picture of why a process varies. Write the target down before any software is chosen.
Involve operators and maintenance staff early. They notice when a model disagrees with the machine, and their corrections are what make the twin trustworthy over time.
Finally, plan the unglamorous parts: who updates the model after a repair, how data access is secured, and what happens when a sensor fails. A twin that nobody maintains soon becomes an out-of-date picture of a machine that has moved on.
Questions to ask before you start
- Which decision will the twin change?
- Who owns the data, and is it reliable?
- Does the product follow a published framework such as ISO 23247, or a private one?
- Can you start with one machine and grow?
If a supplier cannot answer these plainly, treat the claims with caution. This site is independent and does not sell or recommend any product.
Frequently asked questions
Is a digital twin the same as a simulation?
Show answer
No. A simulation is usually run on demand, while a twin stays connected to the real asset and reflects its changing state.
Is there a standard for manufacturing digital twins?
Show answer
Yes. The ISO 23247 series gives manufacturers a shared framework and vocabulary, though products may follow it to different degrees.
Is it only for large factories?
Show answer
NIST’s paper was written because of confusion especially among small-to-medium-sized enterprises, so smaller firms are clearly part of the conversation.
Where should a first project start?
Show answer
With one asset and one decision, using data you already collect, as described above.
