Video game developer Maxis released its hit product SimCity in 1989. In it, players constructed digital replicas of streets, buildings, and cities, virtually funding their computer creations through tax levies and income from public works. It was perhaps the first commercial attempt at digitally replicating not only physical objects, but also their non-physical attributes. As you’ll see, they were on to something big.
Unfortunately, the technology needed to make truly comprehensive computer models was not yet available in 1989. The High Performance Computing Act that funded the National Information Infrastructure—what Al Gore called the “Information Superhighway”—was yet to become law. And the then cutting edge Intel 486DX microprocessors that had recently become available boasted just over one million transistors; by comparison, the Apple M1 Pro chip in the laptop used to write this article has 33,700,000,000 transistors, nearly 34,000 times that number.
Despite these limitations, some futurists saw the potential for advanced simulations. Right around the time that SimCity became all the rage, computer scientist David Gelernter’s book “Mirror Worlds: Or the Day Software Puts the Universe in a Shoebox” predicted a revolution in computer modeling and simulation (and kudos to Gelernter for this article’s title, by the way).
Yet it wasn’t until Michael Grieves, now executive director for the Digital Twin Institute, spoke at a 2002 Society of Manufacturing Engineers conference in Troy, Mich., that this concept was first discussed in the context of manufacturing, when he proposed it be used as the foundation for product lifecycle management (PLM) software.
In 2010, John Vickers of NASA finally gave this new technology a name—the Digital Twin—in his annual Roadmap Report to the space agency. Both experts will tell you that, 12 years later, digital twins are quickly becoming an integral, indispensable part of modern engineering.
Today, automakers use digital twins to lay out assembly lines, simulate manufacturing processes, and eliminate costly road testing. Energy companies validate windmill designs and gather field data with digital twins, while city planners use them to determine traffic patterns and ideal infrastructure configurations. And healthcare providers are already leveraging digital twins to analyze medical products, and might one day use them to evaluate their patients’ treatment plans. Such capabilities are only the tip of the digital iceberg.
But what about machine shops, sheet metal fabrication houses, moldmakers and service bureaus? The owners and managers of these small to medium-sized businesses might ask, “Who cares? We’re struggling to get parts out the door, on time and at a fair price. What good is all this digital twin stuff when I can’t even find decent employees?”
Vance Martin, roles portfolio senior manager for Mechatronics Design Software, 3DEXPERIENCE WORKS at Dassault Systèmes Americas Corp., Waltham, Mass., counters questions like that with the following argument. “I would ask why they’re struggling in the first place. Is it a process-related problem they’ve yet to identify? Are they spending two days programming a job only to find that they’re scrapping parts and have run out of material? By adopting a digital manufacturing strategy, these and many other common issues are easily resolved.”
Note that Martin did not say digital twin, but rather digital manufacturing strategy. This is a fine distinction, and it hits at the heart of Industry 4.0. Digital twin purists might argue that the term refers to complex electromechanical systems—tractors, for instance, or oil rigs—and that the digital twin mirrors and accompanies those objects from cradle to grave, housing all manner of design, procurement, and usage data along the way. The machined, stamped, or printed components that the shops just mentioned might struggle with, on the other hand, represent small pieces of those much larger assemblies. Therefore, they will not be “twinned” to the same depth—or maybe not at all—compared to the products in which they are used.
Yet he’s quick to note that the digital twin concept remains relevant even at this lower scale. “In most situations, the 3D models that these smaller shops work with are simply mathematical volumes that represent a physical object, whereas digital twins have intelligence,” said Martin. “A designer might not use one to design a machined shaft, and no one will bother collecting usage data on a stamped metal housing, but the models for those parts can still contain a wealth of information that provides benefits similar to that of a digital twin.”
Dassault Systèmes’ 3DEXPERIENCE WORKS Roles Portfolio Director Stephen Endersby agreed. “They also improve communication and make collaboration easier,” he said. “Digital twins and digital manufacturing in general help to eliminate the islands of information that so often lead to waste and duplication of data. Everyone’s working from the same integrated platform.”
The information that Martin and others referred to includes raw material properties such as lot numbers, hardness, and tensile strength. There are feed, speed, and depth of cut recommendations to consider, as well as quality control criteria and inspection results. The digital twin could also house process-related data like what cutting tools and toolpaths were used, who made the parts, when, on what machine, and were there any difficulties?
Gathering these and other bits of manufacturing minutia into a single digital part representation—whatever you call it—not only creates a historical record of the production process, but gives machinists and sheet metal workers the tools to make it more efficient.
Rahul Garg, vice president of industrial machinery at Siemens Digital Industries Software, Plano, Texas, sees things much the same way, suggesting that it’s the smaller manufacturers who stand to gain the most from a digital journey. Unfortunately, they’re also often the most reluctant.
“Many times, these businesses fear that digitalization technologies are only useful for larger enterprises, but I would say it’s usually the opposite. Smaller companies are generally dealing with lower margins. They stand to lose the most if there are errors during the manufacturing process, and must make sure that they can meet the customer’s needs in the fastest possible manner and are achieving maximum throughput from their machine tools.”
Here again, the solution is to embrace digital technology, not spurn it.
Whether it’s CNC programming and simulation, job preparation, production planning, operator training, or process optimization, digitization is very easily leveraged, increasing overall profitability and eliminating defects. Said Garg, “Achieving efficiency in all of these areas is important for a shop, whatever their size.”
Such initiatives can bring other benefits to a manufacturing company, he added. For instance, a digital shop with automated quoting capabilities—often online—can effortlessly provide potential customers with an instant price, even for disparate processing methods such as machining and 3D printing.
Here, visitors to the company’s website might easily kick the tires on different manufacturing approaches, determine which is the most cost-effective or viable, then order parts. “Though not necessarily tied to the digital twin, applications like these are gaining steam, and provide a lot of value to the end-user and manufacturer alike,” he said.
Applications are also becoming both easier to deploy and more common. Dassault Systèmes and Siemens each have numerous software platforms within their portfolios, covering the gamut from information intelligence and content creation to electronic design and software development. So does Hexagon’s Manufacturing Intelligence division of North Kingstown, R.I., where Digital Twin Product and Market Manager Silvère Proisy doesn’t get too wound up about the terminology used to describe the digital twin. His biggest concern is that people use it.
“How you define digital twins depends on your requirements,” Proisy said. “A large OEM might develop a digital twin early on in the design cycle and use it all the way through to the product’s end of life, gathering data the entire time. At the other end of the spectrum, a five-person machine shop might need a digital twin that only supports toolpath simulation and process optimization. In either case, it brings significant value.”
Proisy spends much of his time here, at the shop floor end of the digital twin spectrum. He predicts that digital twin technology will become increasingly capable as machine controls become more intelligent and the Industrial Internet of Things (IIoT) grows more widespread. He noted that Hexagon and others are already creating kinematically accurate representations of CNC machine tools, and have begun extending that concept to mirror the control along with the sensors that more advanced machine tool builders are now integrating into their wares.
“There are multiple levels to consider here,” he said. “There’s the asset management aspect to it, where a machine tool builder might deploy a digital twin for the same reasons as any other capital equipment manufacturer—for design purposes, and later, to remotely monitor their products for performance or predictive maintenance reasons. But there’s also the more fundamental needs of their customers, such as machine tool crash avoidance, setup time reduction, and CNC program optimization. All are enabled by the digital twin, and are every bit as important as their more comprehensive counterparts.”
Gene Granata seconds that view. The director of product management at CGTech Inc. of Irvine, Calif., he explained that, in the world of simulation, you’re always trying to get as close as you can to reality. “Every step you can take in that direction gives the user an advantage,” he said. “It means that you’re going to catch everything that you can catch, and that you can rely on the simulation matching what’s really going to happen on the machine when you push the cycle start button.”
Achieving this requires clean, accurate, and complete data. It’s not enough, he said, to construct blocky models of machining centers and use cylinders and cones to mimic cutting tool assemblies. These blocky representations come nowhere close to a digital twin and provide little clarity to the machining process. Nor is the difference between “good data” and bad always understood, or where the best information can be found.
“When it comes to generating intelligent toolpath simulations, there are a lot of loose ends out there,” said Granata. “Does the software know how the machine interprets control commands? Does it know what G-code formats are acceptable? Does it know the axis limits, the available spindle power, and how the machine will react under certain conditions? To get the best results, you need the most complete and accurate digital twin of the CNC machine as possible. For example, rather than depending on someone’s opinion about how they think a machine behaves, our machine building specialists use data from the parameter files that reside in the CNC control. Incorporating this ‘DNA’ from the machines removes guesswork by customers, and provides exact answers about how CNC machines behave so users get the most accurate digital twins possible.”
Cutting tools are another good example of this. Depending on the source, a 3D model of a tool assembly might be dimensionally accurate and pleasant to look at, but in terms of digital data, is quite simply dumb. By contrast, some tooling manufacturers provide models along with information about cutting parameters and limits of use in specific materials. According to Granata, this is extremely valuable in a software system that recognizes these values and incorporates them into the simulation, and for optimizing NC program feedrates.
Such capabilities go farther than crash avoidance or shorter cycle times. Given that the industry faces a chronic shortage of skilled workers, and even more so of experienced programmers, intelligent machining simulations—and accurate simulations of any kind, for that matter—serve to keep those less skilled “between the guardrails” for safe manufacturing.
Said Granata, “It also makes the machining process more efficient. There are fewer worries over catastrophic tool failures because the programmer strayed outside the recommended boundaries. And when you’re driving the tool the way it was designed to be used, part finishes improve, chip flow and heat dissipation will be better, tool life increases and cycle times fall. That’s the very definition of toolpath optimization, but it requires that the digital twins representing the CNC machines and cutting tools are as accurate as possible.”
San Rafael, Calif.-based Autodesk Inc. is another software provider with a corporate eye on the digital twin. Product manager George Roberts works in Autodesk’s U.K. office, where he is responsible for supporting customers with post-processing, simulation, and shop floor activities. Roberts noted that the term digital twin “gets thrown around almost everywhere” and has become common in architecture, civil engineering, media creation, and even game development (hello, SimCity). His purview, however, is manufacturing.
Roberts concurred with Granata and Proisy on the importance of accurate process simulation, but suggested that machine tool monitoring and other forms of shop floor management are similarly important. “There are two sides to that,” he said. “The first is what does the operator need to know? This includes real-time information like whether the machine is running and how much time remains in the current cycle. Then there’s the more analytical data—usage statistics, and how long until the next preventive maintenance cycle. All of that might soon be wrapped up into a digital twin of the entire production floor.”
Like CGTech and Hexagon, Autodesk is also very interested in gathering configuration data from the machine tool control. According to Roberts, this will help predict how the equipment will perform in certain circumstances, and give everyone involved the information needed to make better decisions.
“That’s why we’ve been in talks with other vendors about possibly bringing relevant machine and control data back into our systems,” he said. “It’s preliminary, but we’ve investigated ways to interrogate various parameters and sensor data on CNC equipment and then load that into the manufacturing software. Whatever helps to simulate the machining process as accurately as possible before executing the program, that should be everyone’s goal.”
Connect With Us