The concept of the Digital Twin is still evolving but it is powerful and self-evident enough that most manufacturers believe that they need it. Leading organizations expect that digital twins will help them deliver better products, services, and experiences to their customers, at lower costs than are currently possible. As digital replicas of physical (or cyber-physical) products, digital twins should act as crystal balls allowing engineering teams to understand how their products will behave and respond to real-world use and abuse, long before that information is needed.
As manufacturers move from selling products to increasingly offering product-based services, the lifespan and total lifecycle costs of their devices become more critical. If you can produce less expensively, then naturally your profit margins are higher. And here is the key to using a digital twin to improve your competitive position.
Be sure you have time and resources to devote to this. Before moving to adopt a digital twin methodology in your company, it’s a good idea to make sure you have a complete understanding of what it might look like.
At its heart, a digital twin a mathematical model, and the “digital” part of the name implies that this mathematical model is represented in binary form, which can be calculated and manipulated by computers. But a model for what purpose? There are many kinds of models, and most people think of 3D models first. A 3D geometric model is a good start and an important foundation for the digital twin.
But to be truly useful, the digital twin should express other aspects of the physical artifact—like behavior. Digital twins can include models of materials, coatings, embedded software, embedded control systems, power sources, internal chemical reactions, reactions to environmental conditions (such as temperature, electrical fields, weather, etc.), and more. All of these aspects of behavior combine to simulate the real-world behavior of the physical device.
There is one more wrinkle that digital twins (can) consider: the uniqueness of each physical instance. The old wives’ tale claims that you shouldn’t buy a car built on a Monday or a Friday because manufacturing workers produce poor quality results at the beginning and end of a week. Although I am confident that this is not true, it is a reminder that serial number 1 and serial number 100 of a product are built differently, perhaps by different people under different conditions.
Digital twins incorporate not only virtual modeling of the theoretical performance of any particular serial number but include the instance-specific details for individual physical products in the series. There is a unique digital twin for every serial number that rolls off the manufacturing line. Therefore, all of the data collected during manufacturing (temperature was low on the paint-baking machine that day), and all of the data collected by the device during its use (IoT sensors) combine to enhance the picture we have of that particular instance of the product.
Each digital twin includes the relevant details of that specific physical instance and can predict its unique behavior in response to changing environmental and user-driven conditions in the future. That assumes that we have put the right capabilities in place to capture, track and manage this information.
Many of the capabilities mentioned so far exist today. There are software tools for 3D modeling, control system modeling, static/dynamic analysis, chemical reaction modeling, fatigue analysis, IoT data capture, manufacturing execution (IIoT) data capture, and more. But what is the system that brings all of that data together in a meaningful way, so that you can ask questions and run scenarios? That system is what does not exist. Let’s consider for a moment how we might build this digital twin system.
If we start with systems modeling, 3D modeling and some basic finite-element analysis models, we can create a Virtual Twin that is a good foundation for digital twins. To connect, configure, control and manage all of these models, we need a system like PLM (Product Lifecycle Management). To dive deeper into your virtual twin’s capabilities, you might add things like control system modeling with Hardware-in-the-Loop test capabilities, MDAO (Multi-Discipline Analysis & Optimization) tools, and FMI (Functional Mockup Interface) capabilities.
Somewhere along the way, you will likely choose to enhance your product definition data and become more of a Model Based Enterprise (MBE). This activity may lag a bit, but proceed in parallel with a PLM system deployment.
Integration with ERP/MRP/MES is another important step to build a bridge from the virtual to the physical. Assuming you already have a modern, robust MES system in place, integrating this data back through ERP/MRP and into your virtual models to deliver enhanced analysis results is a great step. At this point, your first digital twins will start rolling off the line at the same time as their physical counterparts.
Many companies today are busy learning how they will leverage IoT sensors, what data they will capture, and how they will use that data for understanding how their products behave during real-world usage. Assuming an IoT project like that is already underway and proceeding in parallel with all of the activities above, your initial digital twins can be enhanced to track and update alongside your products out in the field.
While much of the necessary technology for digital twins exists today, most organizations do not have all the necessary skill sets in place to take advantage of digital abilities. Employees will need new skills in data science, modeling, and next generation configuration management. New organizational roles will be required to support all of this as well. But if your organization is nimble, and the culture is flexible, the organizational changes needed to adopt these tools, skillsets, and new roles shouldn’t add significant time to the projects I’ve described above.
If your business is trying to pivot into selling products-as-a-service, or you just understand the long view of serviceability (and profitability), then it’s time to get started laying the foundation of your digital twin factory—today.
Jonathan Scott is Chief Architect at Razorleaf.