Factory Digital Twin Reference Architecture#
During facility design, construction, and commissioning, factory digital twins help visualize designs in the context of the entire facility and/or production process.
Workflow Diagram
The overview diagram illustrates the workflow for creating a factory digital twin. Each portion of the diagram is explored in more detail in the sections that follow.
Diagram of the roles of different departments and groups in the sample workflow for a product configurator. Each department/group is explored in more detail in its own section of this documentation.#
Information Technologies (IT)
Infrastructure
Infrastructure design and deployment to be able to bring in various parts of the company. Plan according to what the factory and development teams are utilizing, and ensure that the flow of data to and from the digital and physical factories are seamless.
Security (Streaming / Data)
In today’s interconnected business landscape, safeguarding sensitive data, whether belonging to customers or employees, is paramount for maintaining trust and integrity. Cyber threats, ranging from hackers to fraudsters, pose significant risks, potentially resulting in reputational damage and financial losses. Compliance with IT security regulations is not only a best practice, but also a legal obligation in many jurisdictions, with non-compliance carrying severe penalties. Core principles like Confidentiality, Integrity, and Availability (CIA) underpin effective security strategies, with availability being particularly critical in collaborative environments reliant on constant data access. Measures such as network security, endpoint security, and firewalls play pivotal roles in mitigating threats, while understanding the motives behind cyber attacks is essential for devising robust defense mechanisms.
Workstations (local/virtual)
Workstations are critical for many parts of the enterprise to connect and contribute into the digital twin of the factory. Workstations can come in two forms: local (physical under a desk or laptop) or virtual (via a Virtual Machine). The critical element for Workstations in the digital factory is having all of the system resources needed to run what workflow is required.
Nucleus
To collaborate effectively across a multitude of teams, a central database is needed. With Nucleus, developers take advantage of a powerful database and collaboration engine that allows them to deliver live and collaborative workflows in their tools and applications. Nucleus also offers developers a set of fundamental services that allow a variety of client applications, renderers, and microservices all to share and modify representations of virtual worlds together.
Proof of Concept (POC)
A PoC is a pilot project that reflects a real-world scenario. Since developing and implementing solutions in an enterprise environment are often not bespoke, it is critical to run a PoC without impacting the run of business , providing a way to test, validate and stress test before before expanding and deploying to the rest of the enterprise with IT as the connective tissue across all.
See the Internet Technologies (IT) Diagram for a further breakdown of this portion of the overview diagram.
Geometry Creation
USD as collaborative geometry and data schema
OpenUSD is a crucial framework for creating digital twins. USD is a representation of an asset’s geometry, visuals, and data in the digital twin of a factory. Assets that are ready for simulation can be used to design and deploy multiple factories, while also allowing for unique data connectors that can visualize the current or predictive state of a machine or process in specific factories.
Geometry Interoperability
Digital twins of factories require geometry models that can be categorized into two types:
Factory building models including structure, architecture, Mechanical Electrical and Plumbing (MEP) and other static assets, and
Factory production models including machines, production lines, industrial assets, and related elements.
Both categories of geometry models are created using multiple applications generating USD assemblies such as Autodesk Revit and Flexsim, Bentley Microstation, Emulate 3D by Rockwell Automation, Siemens Teamcenter and NX, and more.
End-to-end optimization of production lines
By taking advantage of the USD bidirectionality unlocked between Omniverse powered tools and applications and simulation modeling and analysis applications like Autodesk Flexsim, developers unlock powerful workflows for their users, allowing them to, for example, simulate, test, and optimize end-to-end processes for maximum production cycle throughput. By collaborating in the larger digital twin assembly, users of these tools and applications can leverage the applications and functions of their choice to understand the impact of production process optimization.
Product Assembly Optimization
Optimizing product assembly sequences allows for testing the most efficient and clash-free sequences for thousands of components, improving machine cycle times.
Simulation / SDG
Simulating and validating the function and performance of autonomous machines such as industrial manipulators and autonomous mobile robots (AMRs) is key to improving processes from material handling to production to final packaging and delivery.
Isaac Sim is an extensible robotics simulator that gives you a faster, better way to design, test, and train AI-based robots. It’s powered by Omniverse™ to deliver scalable, photorealistic, and physically accurate virtual environments for building high-fidelity simulation.
NVIDIA cuOpt™ is a world-record-breaking accelerated optimization engine. cuOpt helps teams solve complex routing problems with multiple constraints and deliver new capabilities, like dynamic rerouting, horizontal load balancing, and robotic simulations, with subsecond solver response times.
cuOpt can be used in factories to automate several tasks for all types of robotics processes. For example, autonomous mobile robots (AMRs) that are used to deliver payloads to different stations or cells by dynamically routing them based on priorities, aisle congestion and human presence and many other parameters.
Omniverse Replicator, a framework within the Omniverse platform, accelerates AI perception network training by generating physically accurate 3D synthetic data. It equips deep learning engineers with tools to bootstrap model training, enhance existing ones, or create new models overcoming dataset limitations. Its extensibility allows developers to build custom synthetic data generation pipelines to address a multitude of computer vision use-cases, including visual inspection, training robots, inventory management and more.
AI Integration
The manufacturing process involves various stages of inspection and analysis to ensure product quality and reliability. Automated optical inspection (AOI) techniques such as surface-mount technology (SMT) inspection, detecting missing or incorrect components, and X-ray inspection play crucial roles in identifying defects early on. Final quality inspection, RMA analysis, metal and paint inspection, and finished product inspection further validate product integrity.
Additionally, advanced AI agents, including virtual line managers (VLMs), facilitate data analysis, while ChatBots and LLMs interact with data for enhanced decision-making. Integration with robotic arms enables automated product inspection and quality control, streamlining processes and improving efficiency.
Moreover, the utilization of Modulus technology enhances existing computational fluid dynamics (CFD) workflows, enabling faster and more accurate simulations to optimize manufacturing processes. This comprehensive approach to inspection and analysis ensures high-quality products and enhances overall operational effectiveness in manufacturing environments.
Data / IOT
Teams can leverage massive amounts of factory production data — enterprise, live IoT/machine-level, historical data — to build operational digital twins with Omniverse. This enables use cases ranging from remote monitoring to situational awareness, and allows teams to run simulation scenarios.
By combining production data with 3D data, developers can build a ‘single pane of glass’ view of all factory data with context empowering cross-functional stakeholder groups to collaborate, plan, retrospect, and troubleshoot at factory scale. All data with context helps solve high-value problems like bottlenecks, cycle time and throughput improvements.
Developers can quickly and easily take advantage of IoT data in their 3D apps by using IoT samples (beta) and guides on how to:
Connect IoT data sources to the tools and apps they build with Omniverse
Incorporate IoT data in USD
Visualize IoT data, using an OmniUI and Extension
Perform transformations of USD geometry using IoT data
Incorporate OmniGraph/ActionGraph, a visual scripting language, providing the ability to implement dynamic logic in response to the IoT data
Omniverse Platform
Omniverse SDKs - Leveraging Omniverse SDKs enables developers to build custom applications for specific needs such as review and monitoring applications, optimization, and testing application or task-specific simulation applications.
Digital Twin of Factory (Structure, HVAC, machine libraries and assembled production lines) - Collaborating and coordinating various disciplines involved in factory building can be seamlessly achieved by developing applications with Omniverse. By structuring and implementing component-based simulation-ready assets into their applications for production layouts, full scalability to individual machines and assets can be achieved, with changes being updated across all representations in all factories. Moreover, creating simulation-ready assets allows unique identifiers to be assigned to each asset, enabling data ingestion for data visualization and simulation across all factories.
Custom Extension development for advanced and proprietary functions - Omniverse offers developers the ability to create custom extensions for various functions and interfaces, including data ingestion and visualization from edge devices, as well as operational production prediction like work in progress (WIP) count prediction, as examples. By doing so, digital twins become a review and validation platform, as well as a simulation and prediction engine.
AI enablement - By integrating AI frameworks into the tools and applications they develop with the Omniverse platform, developers enable their users to train and deploy models for enhanced simulation, predictive functions, AOI, and related computer vision functions. The ability to train and deploy on the same platform ensures seamless integration of AI frameworks, providing a single view to the digital twin.
Delivery
Accessing a design application or AI framework model is not always as simple as a system like a laptop or physical workstation sitting at someone’s desk. It’s easy to overlook that all the services required to run and connect disparate frameworks come with complex networking, complex connectivity and bandwidth requirements. That’s why understanding the delivery mechanism of each critical component and ensuring access to all stakeholders are important for a smooth delivery of a Factory Digital Twin.
Omniverse Cloud - Cloud services platform enabling development, deployment, and management of advanced 3D applications and pipelines. Developers have access to a range of powerful Universal Scene Description (OpenUSD), RTX, and generative AI-enabled service-level cloud APIs to power next-generation tools.
Omniverse Cloud is the optimized platform perfect for ensuring that all teams are connected with their individual pipelines to the main workflow of designing and building the digital factory.
Graphics Delivery Network (GDN) - As a part of Omniverse Cloud, GDN takes advantage of NVIDIA’s global cloud streaming infrastructure to deliver seamless access to high-fidelity 3D interactive experiences with frictionless Click-to-Launch streaming and other key benefits.
GDN is the easiest way to access Omniverse Cloud all from a shared URL to your steamed application.
NVIDIA Certified Systems - The NVIDIA-Certified Systems program has assembled the industry’s most complete set of accelerated workload performance tests to help its partners deliver the highest performing systems, evaluated by NVIDIA engineers for performance, functionality, scalability, and security.
NVIDIA XR Solutions - CloudXR is NVIDIA’s solution for streaming virtual reality (VR), Augmented reality (AR), and mixed reality (MR) content from any OpenVR XR application on a remote server - cloud, data center, or edge.
Summary
Omniverse Cloud provides APIs, SDKs, and services to integrate Universal Scene Description (OpenUSD) and RTX rendering technologies into 3D applications. The Omniverse Graphics Delivery Network (GDN) enhances real-time 3D content delivery across devices. NVIDIA CloudXR enables high-quality VR and AR experiences over networks, allowing remote access without high-end local hardware. On-premises deployment involves setting up infrastructure and configuring systems. Access methods include web interfaces, virtual workstations, local workstations, and VR/AR devices. NVIDIA-Certified Systems ensure top performance, having undergone rigorous testing by NVIDIA engineers.
For In-Depth Information, See the Following Sections: