Digital twin simulation can connect the changing conditions on a factory floor with a virtual model of production. By bringing Internet of Things (IoT) sensor data into a simulation, manufacturers can move beyond simply displaying equipment status: they can synchronize a model with physical operations and use it to evaluate or test decisions. The essential challenge is not only building a simulation, but also deciding what data to collect and how to translate it into model inputs.
This approach is relevant to cyber-physical systems (CPS), where information gathered in physical space is analyzed in cyberspace and can inform action in the real world. A practical framework separates the digital twinâs work into synchronization, evaluation, and experimental modes, each serving a different operational purpose.
What makes a digital twin different from a monitoring system?
A monitoring system collects and displays information about real-world conditions. That visibility is useful, but monitoring alone does not provide a way to test hypothetical changes or assess what might happen next. A digital twin combines a virtual representation of a production system with data from connected equipment and a simulation model that can be used for analysis.
In manufacturing, the model can represent elements such as machine states, part movements, and process times. IoT devices supply observations from the physical system; those observations update the model so its representation stays aligned with current operations. The model can then support prediction, evaluation, and experimentation rather than merely visualizing events that have already occurred.
The distinction matters because manufacturing conditions can change. A simulation based on fixed assumptions may become less representative as equipment or operating conditions shift. Synchronizing a model with current data helps keep its representation closer to the behavior of the physical system. Historical observations can also inform probability distributions used in separate simulation analyses.
Three modes for digital twin simulation
A useful way to organize a digital twin is to distinguish its real-time role from its analytical roles. Each mode uses simulation differently:
- Synchronization mode: Receives current IoT data and updates model parameters to mirror the physical system. This mode supports real-time monitoring and records operating history.
- Evaluation mode: Uses historical data to estimate stochastic parameters, then evaluates system behavior or operational strategies.
- Experimental mode: Uses the simulation to test scenarios and explore possible changes without treating the experiment as a live representation of current production.
Evaluation and experimental modes are familiar uses of conventional simulation. Synchronization is the distinctive addition: the model is continually updated from the physical system. Keeping these roles separate helps clarify whether a model is representing current operations, assessing performance, or testing a scenario.
Turning IoT data into simulation inputs
Data integration is a central implementation issue. Conventional simulation typically represents uncertainty with probability distributionsâfor example, a distribution of processing timesârather than feeding each observed measurement directly into the model. Repeated simulation runs can then support statistical analysis. A synchronized digital twin has a different requirement: it aims to represent the physical system as it operates, so its simulation must progress in real time and avoid introducing random outputs that do not correspond to observed events.
One proposed approach uses direct, deterministic values for synchronization. Sensor observations are recorded in a data store or digital sheet, and the simulation reads the latest status data as input. For evaluation and experimental runs, the recorded history can be analyzed to estimate probability distributions, which are then used as stochastic inputs.
This creates a feedback loop between live operation and analysis:
- Connect equipment and sensors to the data-collection system.
- Record relevant events or status information as they occur.
- Use current values to update the synchronization model.
- Analyze stored history to develop stochastic inputs for evaluation or experimentation.
- Use the results to inform assessment of the physical system.
The method makes the choice of data important. Teams need to identify which machine states or process events the model requires, how those observations map to model parameters, and how historical data will be handled separately from live inputs.
A factory model demonstrates the approach
The proposed framework was applied to a factory simulation model representing material storage, handling, processing, and sorting. Its operations included a high-bay warehouse, conveyors, gripper robots, an oven, a saw, and a color-sorting line. Sensors were added to the factory model to capture operating status and timing information.
Sensor signals were transmitted wirelessly to a computer, where a program recorded reception times and calculated intervals between selected events. The simulation read the resulting data and updated parameter values in synchronization mode. Historical observations were also used to support stochastic inputs in the evaluation and experimental models.
To check synchronization, the modelâs output was compared with sensory data from the factory model. The reported comparison found that the workpiece interval times shown by the simulation matched the intervals received from the sensors. This demonstrated the proposed data-input scheme in that modeled system; it does not, by itself, establish performance for every manufacturing environment.
Practical considerations for implementation
For engineers and system integrators, the framework highlights several design decisions. First, define the purpose and scope of the twin before selecting data. A device-level twin and a process- or system-level twin involve different modeling scope and data needs. Second, distinguish live synchronization data from historical observations used to characterize variability. Mixing these purposes can undermine the modelâs ability to represent current behavior or support meaningful scenario analysis.
Model verification and validation also remain necessary. A connected simulation is not automatically an accurate one: its representation must be checked against the physical systemâs observed behavior. Sensor placement, event definitions, data recording, and the mapping between measurements and simulation parameters all contribute to whether synchronization is meaningful.
Finally, a digital twin is more than a 3D model or status dashboard. Its value in this framework comes from linking IoT data to a simulation that can represent current conditions, evaluate system behavior, and test alternatives.
Conclusion
Digital twin simulation offers a structured way to connect manufacturing data with operational analysis. Separating synchronization, evaluation, and experimentation helps teams use live sensor values to mirror production while drawing on historical data for stochastic analysis. For plant leaders and technical teams, the practical priority is to define the modelâs purpose, select relevant data, establish a reliable input path, and validate the simulated behavior against the physical system.
Access Official Technical PDF