Unity: A Game Engine Powering Digital Twin Technology

Why this article is important

This article outlines how to implement a digital twin using Unity, which is best-known as a gaming platform.
It explains many of the technical aspects of implementing a Unity-based digital twin.
It outlines how real-time data flows from the production line into Unity.
The article details system performance requirements for a large-scale digital twin and describes the efficiency gains already achieved by the manufacturer.
The reader will understand how the technology evolved from an initial proof of concept into a globally-deployed solution.
The article explains how upcoming enhancements will continue to push the boundaries of digital twin innovation.

In the fifth volume of his famous series In Search of Lost Time, published in 1923, Marcel Proust wrote: “The real voyage of discovery consists not in seeking new landscapes, but in having new eyes.”

Proust wanted readers to understand that breakthroughs often arise from reimagining how existing tools and knowledge can be applied in different contexts. More than a century later, his insight remains highly relevant in industrial manufacturing, where the continued pursuit of increased operational efficiency perpetually drives innovation.

A project completed for a manufacturer recently demonstrated the merit of Proust’s principle. A remarkable unique advantage was obtained from an unexpected tool outside the traditional industrial toolkit for digital twin implementation. Rather than relying on industry standard simulation software, E Tech Group selected Unity, a video game development platform known for its advanced 3D graphics, powerful physics simulation engine and cross-platform deployment capabilities.

Over six years of implementation, these Unity-based digital twins have evolved from a novel solution into a mission-critical system, reducing downtime through simplified troubleshooting and improving operator training, becoming so integral that operators access them as part of their regular workflow. The next iteration will introduce two-way communication with physical equipment to enhance existing line technologies. It will also incorporate machine learning-based analytics for advanced predictive maintenance.

What is a digital twin?

A basic digital twin visualizes the behavior of the physical system it represents in real time—a virtual display of what’s happening right now with the real equipment on the manufacturing floor. From a high level, data from sensors and other equipment in live machinery is continually collected. This telemetry information is then sent to and visually interpreted by a virtual representation of each component of the system in the simulation environment. As a result, the twin’s motion and state are directly tied to that of the equipment it represents.

More generally speaking, a digital twin can be connected to any data source that describes equipment operation. This broader capability opens the door for other off-line uses of the digital twin application:

Operator training: Past failure events recorded in a time-series database can be replayed for training purposes. Simulated failure scenarios can be used similarly.
Software virtual commissioning: A physics-based engine can simulate system behavior. By integrating PLC logic and other control sequences, the application can be used to virtually debug programming logic and verify system performance before deployment.
System optimization: CAD models, physics-based simulations and performance data can be used to optimize the system layout for better efficiency and throughput.
Predictive maintenance: Historical data can be modeled using machine learning to detect anomalies and predict upcoming system failures.
Process control: An advanced digital twin application can incorporate two-way communication with the equipment allowing it to send control commands and dynamically adjust system parameters based on real-time data.

Unity as the core of a real-time digital twin

For a digital twin model to realistically express the state of operation and motion of a physical system, it must receive accurate datasets that describe its next state. While an active digital twin obtains this information from the physical equipment to which it’s connected, offline applications of the twin require a different source, and the accuracy of its motion depends on the quality of the predictive model or simulation engine. Unity is exceptionally well-suited for this task, as its mature physics engine can generate the next states by applying classical mechanics principles, including kinematics, dynamics and force interactions, to simulate real-world behavior. For instance, it can accurately simulate the effects of gravity, friction, collisions and inertia. The physics engine enables Unity to predict real-world behavior, which expands the digital twin’s usefulness to offline applications including development, testing and optimization.

As a game development platform, Unity excels at graphics rendering and scripting capabilities, providing a robust 3D environment for visualizing virtual models of real components. Within Unity, a user can maneuver freely in virtual 3D space, zooming in and out with ease around moving components. For troubleshooting purposes, this movement capability allows operators and support engineers to visually explore the state of the digital twin in real-time, aiding in the identification of production issues as they arise.

Finally, Unity is platform-agnostic, allowing digital twin applications to be deployed on various devices including PCs, laptops and mobile devices.

Data flow and real-time considerations

For a Unity-based digital twin to reflect the current state of physical equipment, there must be a clear path for data to flow from the production line into Unity. Real-world production line data can either be fed directly into the model or be routed through a database or a unified namespace (UNS), where it is then accessed or pushed to Unity to drive the digital twin.

Direct data feed—live data

In some basic implementations, sensor data from the production line and programmable logic controllers (PLCs) can be sent straight to Unity without first being stored in a database. This method often uses industrial protocols like OPC UA, message queuing telemetry transport (MQTT) or Ethernet/IP to transmit the data in real time. Unity then updates its 3D simulation based on each new data packet, creating a highly responsive twin. However, a direct feed can be vulnerable to temporary network delays or downtime. If the network link goes down, the twin might not receive any new updates until the connection is restored.

Get your subscription to Control Design’s daily newsletter.

Database-driven feeds

A typical implementation is to send sensor and control data to a high-performance database before Unity retrieves it. E Tech Group’s development team uses InfluxDB for this purpose. This near real-time approach involves time-series storage of data at configurable intervals—for example, every 100 milliseconds or 1 second. Data can either be pulled by Unity clients or pushed via an MQTT broker. E Tech Group prefers the latter and uses HiveMQ for this purpose. An MQTT broker implementation reduces network load since it does not require Unity to poll for updates. Rather, data is delivered to Unity as soon as it changes, so that the twin is fully synchronized with the current state of the real equipment.

While a database-driven approach may introduce a small delay—milliseconds to seconds—it offers significant advantages in comparison to a direct data feed implementation:

Data buffering and reliability: If a connection to Unity is lost, the database still collects the data for historical record and picks up where it left off once Unity reconnects.
Historical analysis: Storing data chronologically enables engineers to look back at previous events and trends.
Scalability: Multiple Unity instances or other applications can simultaneously access the same data source, supporting enterprise deployments.

Unified namespace (UNS)

A more recent innovative connection technique involves the utilization of a unified namespace (UNS), which provides a centralized, standardized way of collecting and contextualizing data from various sources. In digital twin implementations, the UNS can serve as a structured data hub, feeding real-time data from production systems into downstream applications, including SQL, InfluxDB and the Unity-based digital twin itself.

The UNS replaces multiple direct connections, which reduces network congestion, improves data consistency and ensures the digital twin receives accurate, up-to-date system information. Additionally, it simplifies the integration of new data sources or data consumers over time. As production systems change, new machines, sensors or analytics tools can be added to the UNS, allowing the digital twin to easily maintain an accurate representation of the physical environment.

Regardless of which method is chosen, the goal is to keep the digital twin synchronized with the physical equipment. In true real-time implementations, Unity reads data as soon as it becomes available, matching the physical system’s behavior with only minor latency. In near real-time systems, buffering may mean the twin is a fraction of a second behind live events, which is typically acceptable for most troubleshooting and monitoring needs.

CAD models as the basis of virtual representation

To create a fully visualized digital twin, all elements of the real system must be represented within Unity’s virtual environment, including the following:

Physical components such as robots, conveyors, machinery, safety features and general-purpose areas. The goal is to accurately represent the layout of the physical system.
Control systems: Modeling all control and automation logic—PLCs, human-machine interfaces (HMIs) and supervisory control and data acquisition (SCADA) systems—within the digital twin will ensure it replicates the behavior of the physical control system.

A CAD file can be used as the basis of each component in the digital twin. Unity allows CAD files to be imported directly into a project, although they must first be converted into a supported file format. Unity Pixyz is a plugin tool that accomplishes this task for most proprietary CAD file formats, including Solidworks, AutoCAD and CATIA. Within Unity, each component representation should be configured to connect to the appropriate data stream from its associated live equipment.

As an extension, virtual sensors can also be incorporated into the digital twin as an innovative means of capturing real-time data for monitoring, analysis and production optimization.

Key takeaways

Unity, a platform best known for video game development was selected to build digital twins of production lines and transformed the manufacturing process.
By simplifying troubleshooting, the digital twins reduced downtime and improved efficiency, resulting in cost savings.
The ongoing project highlights how tools from outside of the manufacturing industry can propel innovation forward, solving problems in new ways that result in substantial increases in overall equipment effectiveness (OEE).

System performance requirements

The following list outlines the performance requirements for different components of the digital twin system.

Control system: The actual equipment, including PLCs must be connected to a reliable OT network designed for high-speed data transfer. Legacy controls may need to be upgraded to meet the processing demands necessary for running real-time control operations and transmitting data to the database.
Simulation: Unity’s physics engine requires significant computational resources for accurate and timely simulation. This includes robust central processing unit (CPU) and graphics processing unit (GPU) capabilities to handle complex, time-sensitive calculations without lag.
Database server: The server should be fast and reliable, optimized for handling large volumes of time-stamped data. InfluxDB is recommended for its efficiency in writing and querying large datasets, which is important for maintaining high performance in industrial settings.
Visualizing clients: The visualizing clients can run on a wide range of laptops or devices, as long as they have sufficient graphics and processing power. Each client establishes a connection to the system to display real-time data and view the digital twin.
Network and back-end infrastructure: Because multiple clients may connect simultaneously, the network and back-end infrastructure must be designed to support concurrent sessions without performance issues. E Tech Group uses cloud hosting solutions including Microsoft Azure and Amazon Web Services (AWS) for this purpose because they offload much of the networking, security, and IT maintenance responsibilities away from the manufacturer. This implementation also provides easy scalability of resources as demands change, ensures multiple users can connect concurrently without performance drops and allows users to securely connect to the digital twin application over the internet.

Digital twin implementation for a manufacturer

The first Unity-based digital twin was introduced six years ago to visualize a manufacturer’s magnetic conveyor-based production lines and integrate with a specific PLC. Although this initial design remains in active use at multiple facilities worldwide, the latest version has been upgraded to be control system agnostic. This means it can gather data from any source, including various PLCs, program architectures, computers, application programming interfaces (APIs) or databases. Each data stream is mapped to its corresponding component within the digital model, ensuring accurate visualization. The current revision is already running on nine major production lines, with three more in development.

Results and ongoing improvements

The combination of Unity’s 3D visualization and movement capabilities alongside its mature physics engine has resulted in many improvements to the production processes for the manufacturer.

Production lines are easier to understand and operate

Because the digital twin visually displays the equipment’s current state, it is inherently easier for operators and engineers alike to understand and troubleshoot it. In addition, the digital twin has been programmed to clearly communicate actions required to address an issue. The overall result is that operators better understand the line and therefore can often identify and address issues themselves without additional support. The production lines with digital twins have proven efficiency improvements because they remain operational a greater percentage of time compared to those without.

Remote monitoring and troubleshooting

A Unity-based digital twin simplifies remote monitoring and troubleshooting. The digital twins are essentially a live feed of the current state of each line, so when an issue arises, engineers viewing the digital twin can easily detect and mitigate the issue. In fact, this particular client has created a centralized control room that displays the digital twins of all their major production lines so that support engineers can monitor all of them from this singular location. It’s not uncommon for the support engineers to notice an issue before the line operators. When this occurs, a phone call is made to make them aware of the issue. The digital twins have resulted in a tremendous savings through reduced downtime and travel costs for support engineers.

A second server for offline applications

Once a digital twin is fully built and connected to a telemetry data source, a second copy of the twin can be implemented to function offline, fully disconnected from the live equipment. Unity’s sophisticated physics engine can generate the data inputs to determine the motions of the system based on realistic modeling of classical mechanics principles. The current revision of the manufacturer’s digital twin is used for several interesting applications of offline functionality.

Software virtual commissioning: The manufacturer is using the offline digital twin to debug control programs prior to testing them on the physical equipment.
System design and optimization: The manufacturer is using the offline digital twin for line design including optimizing the physical layout to maximize production rate, minimize overall footprint and select best-suited equipment.
Training simulation: This manufacturer uses the offline digital twin for training purposes. By recreating failure scenarios, operators can interact with the system to learn how to address them.

Next-generation digital twins will support direct process control

The next revision of the digital twin, currently in development for the manufacturer will introduce the following additional functionality:

Two-directional communication allowing the digital twin to control the line: While the current revision of the digital twin only reads and interprets data from the line, the next revision will allow the digital twin model to send control information directly to the line, offering increased flexibility for production line control.
Failure prevention and analysis: Machine learning algorithms can be applied to historical Internet of Things (IoT) data to analyze the trends over time and train a model. Initially, this model can help identify and diagnose points of failure. As it continues to learn, the model can be set up to alert engineers before a critical failure occurs, allowing them to take immediate action. Once two-directional communication is completed with the next revision, plans are in place to incorporate allowing a trained model to adjust the production line parameters—for instance, conveyor speed—based on learned acceptable ranges.

Final note

As manufacturers continue seeking innovative ways to boost efficiency and reduce costs, digital twin technology is emerging as a powerful driver of industry growth. For this manufacturer, the unconventional decision to adopt Unity, a platform originally designed for video game development, has delivered significant gains in overall equipment effectiveness while creating additional possibilities for future advancements. With each revision, the digital twin continues to become ever more mission-critical to daily operations including advancing toward machine learning-based production line control and implementation of predictive analytics. Indeed, it proves that breakthroughs are often found not in entirely new tools, but in reimagining how existing ones can be applied in transformative ways.

Next steps

Evaluate existing plant data infrastructure to determine compatibility with a unified namespace (UNS) implementation for improved data flow to a potential Unity-based digital twin.
Research and document the specific CAD file formats used for current equipment to assess the feasibility of importing them into Unity via the Unity Pixyz plugin.
Conduct a resource assessment of available server and client hardware to ensure they meet the computational requirements for real-time Unity simulation and database handling.
Develop a proof-of-concept Unity digital twin model for a selected production line segment, focusing on establishing a reliable data stream and basic real-time visualization.