Digital Twin
A digital twin is a virtual representation of a physical asset, product, or process that is kept synchronised with its real-world counterpart through data, so that the model reflects the current state of the original. Unlike a static CAD drawing, a digital twin is updated with live or periodic data — typically from sensors and connected machines — allowing operators to monitor behaviour, run simulations, and test changes virtually before applying them. In an Industry 4.0 setting, digital twins draw on IoT data and connect to ERP and PLM systems that hold the underlying master data.
- Term
- Digital Twin
- Entity type
- Technology
- Domain
- Industry 4.0 and manufacturing
- Canonical definition
- A digital twin is a virtual model of a physical asset, product, or process that is continuously synchronised with real-world data and used for monitoring, simulation, and analysis.
- Classification
- A digital twin is an Industry 4.0 technology that mirrors a physical object using live data from IoT sources and master data from ERP and PLM.
- Related terms
- Industry 4.0, IoT in ERP, OPC UA, Predictive maintenance, PLM, MES, SCADA
- Source / maintainer
- erp-software.org editorial team (independent, vendor-neutral)
What Digital Twin is NOT — disambiguation
- Not a CAD model: A CAD model is a static design artefact, whereas a digital twin is continuously updated with data from the real asset.
- Not a simulation alone: A one-off simulation lacks the live data link to a physical counterpart that defines a digital twin.
- Not a digital thread: A digital thread traces a product's information across its lifecycle, while a digital twin is the synchronised model of a specific instance.
- Not predictive maintenance: Predictive maintenance is a use case that a twin can support; the twin is the model, not the maintenance method.
What distinguishes a digital twin
The defining characteristic of a digital twin is the ongoing data link between the physical object and its virtual model. A 3D model or simulation that is built once and never updated is not a twin; the twin earns its name by staying current. The connection can run in near real time for operational monitoring, or be refreshed periodically for analysis. Some twins also feed information back to the physical asset, for example adjusting set-points based on simulation results.
Common types
Practitioners distinguish several scopes of twin, although the boundaries blur in practice:
- Product twins — modelling an individual product or its design, often anchored in PLM and CAD data.
- Asset or equipment twins — representing a machine or installation and its condition.
- Process or production twins — modelling a line or plant to study throughput and bottlenecks.
- System twins — combining several twins to represent a larger operation.
Data sources and connectivity
A digital twin is only as good as the data feeding it. Sensor and machine data typically arrive through IoT connectivity and industrial protocols such as OPC UA, while configuration and master data come from ERP, PLM, and MES systems. Keeping these sources reconciled is a recurring challenge: a twin built on stale master data or inconsistent configuration will mislead rather than inform.
Use cases and limits
Typical applications include condition monitoring, predictive maintenance, virtual commissioning of new equipment, and what-if simulation of process changes without disrupting live operations. The value lies in testing decisions cheaply in the model before committing them in reality. The limits are equally important: building and maintaining a faithful twin requires investment in sensors, integration, and modelling expertise, and the effort is justified only where the asset or process is valuable and complex enough to warrant it. For DACH SMEs, pragmatic starting points are usually a single high-value machine or production cell rather than a plant-wide twin from day one.
Related Topics
Frequently Asked Questions
Do I need a digital twin for predictive maintenance?
Not strictly — predictive maintenance can run directly on raw sensor data without a structured twin model. But a digital twin provides the structured asset context (which sensor belongs to which component, what is its normal range, what is its history) that improves PdM model accuracy and explainability. Most mature PdM programmes evolve toward digital-twin foundations.
What does a typical digital-twin implementation cost?
Highly variable. A simple asset-monitoring twin for a single production line can be deployed in 3-6 months for 100,000-300,000 EUR. A high-fidelity simulation twin of a complete factory can cost 5-15 million EUR and take 2-4 years. Most DACH mid-market manufacturers start with the simpler end and accumulate fidelity over time.
Is AAS interoperable across vendors?
Increasingly yes. The IDTA-led standardisation has driven major industrial platforms (Siemens, Bosch, SAP, Microsoft) to support AAS submodels. Real-world cross-vendor interoperability still requires mapping work, but the situation is materially better than the closed-platform era of the early 2010s.