The digital twin is nothing new. A CAD drawing is a digital image of a potential object. That goes back a few decades. Yet the term “digital twin” now takes on a larger picture that can include all of the data on a product, from initial 2D drawings on to 3D CAD, all of the various iterations, manufacturing data, and even data coming back from the field as users interact with a product.
Siemens defines the digital twin as a virtual representation of a physical product or process, used to understand and predict the physical counterpart’s performance characteristics. The description goes on to say that digital twins are used throughout the product lifecycle to simulate, predict, and optimize the product and production system before investing in physical prototypes and assets.
Is a Digital Object More Accurate than A Physical Object?
To take it a step further, digital twins incorporate multi-physics simulation, data analytics, and machine learning capabilities. These virtual constructs can demonstrate the impact of design changes, usage scenarios, environmental conditions, and endless variables. The goal is to eliminate the need for physical prototypes. That reduces development time and potentially improves the quality of the finalized product or process.
To step even further out, the digital twin may provide a more accurate product simulation than the physical product itself can provide. After all, you can put a physical product through all possible contingencies. Yet you can put a digital version through virtually every contingency.
So how do you create an effective digital twin? We turned to Alastair Orchard, VP of digital enterprise at Siemens Digital Industries Software to explain what it takes to build a better digital twin.
Design News: What is the process of creating a digital twin? Is it a matter of collecting existing data, such as design CAD files and simulation data? Or is there something unique about the digital twin that requires generating additional data?
Alastair Orchard: Any files containing data related to the physical item in question are part of its digital twin. So yes, a CAD file… simulation data… the Bill of Process… the Execution Record… the IoT trends, are all part of a digital twin.
Any part of this dataset will be useful: a thermal simulation of the item will predict how hot it will get in operation; an IoT trend will tell you if your predictions were correct and if maintenance is required. Building a separate digital twin when you already have all the data would be a waste of time, however. The real value of the digital twin comes when you can begin accessing it cross-domain. Some examples would be:
- Use product performance data to inform product design
- Simulate the shop floor with a proposed product design to check for manufacturability
- Use a process simulation in real-time during production for predictive control
- Compare a predicted performance to actual performance to schedule maintenance
To do this using separate files from different systems would cost more than the benefit it provides. Siemens’ Xcelerator portfolio offers an avenue to host the digital twin and a set of plug-in platforms for interacting with them. Our Design Suite defines the mechanical and electrical facets of the digital twin; our Simulation Suite validates the design by checking kinematics, thermodynamics, aerodynamics, etc; our Digital Manufacturing Suite checks the manufacturability of the design by attempting to pass it through the production digital twin; our Automation Suite uses a compiled version of the digital twin at the edge to enhance predictive control of the process, and our IoT Suite compares predicted response of virtual sensors with actual values to shift from preventative to predictive maintenance. These pieces make up the comprehensive digital twin. By working across engineering boundaries, the comprehensive digital twin enables cross-domain engineering, virtual validation, and continuous product and process improvement through a closed-loop feedback system supported by cloud-based analytics.
Design News: What are the processes of creating a digital twin?
Alastair Orchard: Most companies have probably already taken the first steps. If we take a robotic cell as an example, you almost certainly have access to a digital representation of the cell. If you are an end-user and you’re purchasing the cell from a machine/line builder, then do yourself a favor and make the CAD and simulation file a part of the deliverable. If it’s a brownfield shop, then you’re going to have to retrofit the existing equipment. In that case, I would start with a point-cloud scan of the cell (and the whole factory while you’re at it).
You need to call a specialist with the appropriate scanning equipment, but they will only take minutes to deliver you a photorealistic 3D representation of your cell. Next, in terms of value, I would look at the logistics around the cell: buffer levels, availability, operator activity, etc. This is achieved by dropping the point cloud scan into a line designer, replacing the static 3D images with equipment from the palette, and configuring runtime information such as speeds, buffer sizes, mean time between failures, and so on. All this data can be picked up from IoT sources if available.
Once we’re confident that the cell is well organized, we can focus on the detailed robot and any human movements – avoiding collisions, maximizing safety, and minimizing wasted time.
It’s important to understand that the results of the digital twin are both a validation of concepts/layouts but also a set of configurations that can be pushed into production. Those would include robotic programs, electronic work instructions for operators, machine setups, and so on. Predictive parts of the digital twin may also be compiled and published to the Industrial Edge platform to support real-time operations.
Design News: Does the creation of a digital twin require the whole sequence from product design and production to field performance? Or can the digital twin be created in stages, i.e., collecting the design and simulation data for developing the manufacturing process?
Alastair Orchard: It’s absolutely a gradual, stepwise process. Each stage refers to the model hosted in the digital twin platform, doing its part to improve its functionality/predictive accuracy. You can also start at any point: from the detailed design of the equipment, from the IoT visualization, or the production process in the middle.
Design News: Can you create a digital twin for services? If so, would its value be characteristically different from a digital twin for a product?
Alastair Orchard: Yes, you can, but it’s essentially the same digital twin. I could be servicing the equipment in the factory or an aircraft engine. One sits in the factory lifecycle, the other in the product lifecycle, but in both cases, I am referring to the same digital twin that was used to design and operate the asset. This is a good example of why it’s best to think of a single comprehensive digital twin as opposed to a service twin. I can use IoT to visualize what the asset is doing, and whilst it’s perfectly OK to configure thresholds for alarms manually, it is much more effective to compare the real-time data streams to the predictions made during design. This is not only to reduce configuration effort but also to improve accuracy because conditions (say temperature) are not absolute values but depend on operating speed and other factors.
To summarize, the service element of the digital twin is enhanced by leveraging the product and production facets, but it’s a valid low-hanging fruit and a common place to start.
Rob Spiegel has covered manufacturing for 19 years, 17 of them for Design News. Other topics he has covered include automation, supply chain technology, alternative energy, and cybersecurity. For 10 years, he was the owner and publisher of the food magazine Chile Pepper.