Digital Twin: What Does It Actually Represent?

A twin may represent a piece of equipment, accompany a product through its lifecycle, or describe operations across a facility. Its value depends on which part of reality it captures and how that representation connects to work. A cyber-physical production system, by contrast, also encompasses physical sensing, communication, and control.

Product, equipment, and operational twins represent different things

Three perspectives provide a useful practical distinction. They classify twins by the object being considered, not by an exhaustive taxonomy established in a standard.

Product twin: A product or individual product instance is central. Origin, manufacturing history, materials, characteristics, and changes can be associated with it.

Equipment twin: The focus is a machine or plant asset. The representation covers components, operating states, measurements, and, where applicable, technical models of its behavior.

Operational or process twin: The focus is on work performed and the relationships among tasks, materials, batches, people, instructions, measurements, and results.

Consider a fictional manufacturing case. Batch CH-742 is processed on equipment AN-06. A measurement returns an unexpected value, and the responsible person must determine what it means.

The product twin can describe which materials and manufacturing steps belong to CH-742. The equipment twin can show recorded states of AN-06 during processing. The operational twin shows which instruction applied, who carried out the measurement, what was recorded, and which review or decision followed.

The three perspectives complement one another. They answer different questions about the same operation.

Digital models, digital shadows, and digital twins differ in their connections

Kritzinger and colleagues distinguish three levels of data exchange between a physical system and its digital representation. ^[2]

A digital model has no automatic data exchange. A digital shadow receives data automatically from the physical system. A digital twin, under Kritzinger's narrower classification, automatically exchanges data in both directions.

This classification describes technical coupling. It should not be treated as identical to every broader use of the term digital twin.

An operational twin can also influence execution through tasks, reviews, process-dependent blocks, and subsequent steps. New results update the representation; defined process rules determine which tasks or checks follow.

The return path may run through a person guided by the system. That person receives relevant work context, enters new information, and makes the required professional decisions. The system advances the workflow according to the configured process.

This is an operational twin with human-guided feedback. It does not, by itself, meet Kritzinger's narrower condition of automatically bidirectional coupling. Nor does it imply autonomous machine control.

Objects, events, and relationships form the operational twin

An operational twin needs more than data fields. Objects, recorded events, and the relationships among them must remain distinguishable.

For CH-742, these include the material batch, equipment AN-06, the measurement task, the applicable work instruction, and the result.

A context model defines which relationships matter to a particular task. The relationship between a measurement and its sample is different from the relationship between equipment and a production order.

Process events record performed actions or observed state changes. A planned measurement is not the same as its confirmed execution.

During a subsequent investigation, data lineage can help explain how a result arose from source information and processing steps and which downstream information depends on it.

For CH-742, these perspectives create a connected view of the unusual measurement and the work that followed. The sequence does not have to be reconstructed from unrelated lists.

Benefits include traceability, process comparison, and focused review

An operational twin supports three particularly useful tasks.

Trace origins. A recorded result can lead back to documented inputs, materials, and preceding steps. For CH-742, investigators can follow the measurement back to the relevant task and its supporting context.

Compare executions. When batches are recorded using comparable events, differences in sequences, waiting times, reviews, and results become observable. Process variants distinguish the paths actually followed. Process mining can examine those differences systematically.

Prepare reviews. Someone deciding on a deviation or release needs the affected object, applicable instructions, results, and unresolved questions. The operational twin brings the recorded relationships together without replacing the decision.

For CH-742, the record can indicate whether a repeat measurement was ordered, who performed it, and which further task resulted.

Costs depend on data scope, integration, and maintenance

The cost of a digital twin follows from its intended scope.

A representation of a few equipment states requires different data and interfaces than an operational twin spanning materials, batches, instructions, devices, and decisions.

Key effort drivers include selecting and identifying relevant objects, deciding how much detail to capture, integrating existing systems, configuring work processes, and maintaining rules and interfaces.

Migrating existing records can also require work. The key questions are which historical data remain necessary and whether their origins can be traced sufficiently.

For CH-742, the first design decision would be which material, equipment, and task relationships are necessary to investigate a measurement finding. Those requirements provide a more useful basis for estimating effort than an abstract price per twin.

The representation ends where the recorded evidence ends

An operational twin can display only relationships that were recorded or derived from an explicitly identified basis.

An unrecorded measurement must not be presented as documented work. A stored number alone does not establish that the measurement procedure was performed correctly or that the result is professionally sound.

Comparing executions has limits as well: a temporal or statistical association does not prove causation.

An operational twin does not replace professional review, required validation, or responsible release decisions. The human decision remains a separate, visible step.

Digital twin vs. simulation is another important distinction. A digital representation of work performed is not necessarily a model that predicts future machine states or production results. The two capabilities may coexist, but neither automatically establishes the other.

The operational twin is most useful when both the recorded history and the boundaries of its evidence remain clear.