In December 2021, a strange creature began running around Audi factories, helping them to prepare for the future.
Spot, a robotic dog developed by Audi and NavVis, a German startup, carries a 3D laser scanner that records images of every tool and square meter. The images will be fused to create a digital twin of the factory that can be used to plan production, change tooling, and realize a host of other efficiency gains and cost savings.
Because Spot can work weekends without complaint and doesn’t need breaks, he can scan an entire factory in 48 hours, a task that would take a team of people three weeks.
“Merging all the planning data in our digital twin has given us a holistic look at our future production plans years ahead of time,” said Andres Kohler, head of Audi’s virtual assembly planning team.
Spot, a test project, may be a few years ahead of his time — most automakers are relying on other scanning methods to digitize their factories — but digital twins are emerging as a key to fast, safe and efficient production, especially as automakers race to keep up with trends such as electrification and the software-defined vehicle.
Using twins, automakers can game out the conversion of an internal-combustion factory to build electric cars. They can test-fit a new piece of tooling, train workers in new processes, and modify those processes on the fly to improve through-put rates and even vehicle quality.
“When you have a digital twin, you can do things in the virtual world before the costly commitment to doing it in the physical world,” Richard Kerris, vice president Omniverse platform development at Nvidia, told Automotive News Europe. A true-to-reality digital twin, down to the last nut, bolt, process and worker of a production network can give an automaker “superpowers,” he said.
Nvidia is working with BMW, General Motors, Mercedes-Benz and others on its Omniverse platform, which Kerris describes as the operating system of an industrial metaverse. He quickly added that “metaverse” in an industrial context “is not everybody floating around as avatars,” but rather a network that allows seamless connectivity between 3D virtual worlds.
“The key thing about a digital twin is that it doesn’t stop when you’ve made it,” Kerris said. “Over time you can study and learn things — maybe there are ways to train robots to move more efficiently or safer ways to move around equipment.”
Digital twins and the broader concept of an industrial metaverse build on the principles of what is known as Industry 4.0, which aims to build better processes through data collection (the Internet of Things) and processing.
“The first wave of Industry 4.0 brought transparency and some predictiveness, but it did not really combine real-world data and the capability to simulate,” Pierre Bagnon, vice president, global head of intelligent industry accelerator at Capgemini, told Automotive News Europe. “There are still a lot of silos.”
In the automotive industry, digital twins can break down the walls between product development — the domain of designers and engineers — and process engineering, where cars go from prototype to production, said Bagnon, who worked at Robert Bosch for more than 15 years.
“What we dreamed of was an ideally production-friendly product design,” said Bagnon’s colleague Matthias Eisenschmid, vice president, head of automotive production and supply chain management, Capgemini Invent Germany, and a veteran of production planning at Mercedes-Benz. Digital twins are a “game changer” for this use case, he said, allowing product and process to develop in a synchronized, parallel way.
“You get a production line that is mature right from the start,” Eisenschmid said. “In the past you had to spend a lot of time and money in physical production to ramp up and optimize the processes. Using an industrial digital twin, you can identify and solve most of your problems upfront in the virtual phase.”
Digital twins can shorten time to market by 20 to 30 percent, improve quality by 20 percent and resource efficiency by up to 40 percent, Bagnon and Eisenschmid said. “That’s the magnitude of ambition that our clients are putting in place to justify the investment,” Bagnon said.
“It’s a multimillion-dollar investment, but the more important question is what will happen if you don’t do it?” Eisenschmid said. “There are new competitors who are already taking significant advantage of these new technologies as common tools.”
Renault Group is creating its own industrial metaverse including digital twins, a four-step, years-long process. “We want to have a fully actionable universe with a full replica of what is happening in the real world,” Eric Marchiol, director of digital transformation at Renault Group, told Automotive News Europe.
“We have learned that massive data usage is the ‘killer app’ to achieve our targets,” Marchiol said.
Renault has ambitious targets: By 2025, it expects to see overall savings of 320 million euros, inventory savings of 260 million euros, a 60 percent reduction in vehicle delivery time and a 50 percent cut in manufacturing CO2 emissions. On the quality side, it will be a major contributor to a 60 percent cut in warranty costs.
But Renault’s path to a full industrial metaverse points to the scope of such a project.
The first step, data collection and standardization, started in 2016 when Renault built its own apps to be able to visualize existing data on the shop floor on tablet computers, which led to some “quick wins,” Marchiol said. It then created an internal team to “crack what is inside” programable logic controllers, robots and “black boxes” in other equipment.
With 10,000 robots across the group from different suppliers such as Comau or Kuka, as well as machinery that dates back decades, “we had a huge diversity of data,” Marchiol said.
Now, he said, Renault can gather data from its equipment in a standardized way — and it is receiving a billion data sets a day. “But it’s not enough if you only collect from equipment because we still have a lot of manual operations,” Marchiol added. To capture this data, Renault has built digital workstations that track human movements.
In the second step, Renault is creating digital twins of its factories using technology from Siemens and Dassault to simulate new equipment, and it is scanning its existing plants down to the last millimeter. Renault then has an avatar it can use for equipment setup training or ergonomics.
Next, Renault is tracking all outbound and inbound activities from suppliers, “so we know exactly where every truck is,” Marchiol said. This extends Renault’s data ecosystem outside its own industrial footprint.
In the fourth step. Renault will leverage the power of its industrial metaverse across the group through cloud-based, easy-to-use analytics tools. “We are democratizing the use of big data tools inside all our operations,” he said.
Digital twins and other parts of the metaverse are critical to Renault’s shift to electric cars and powertrains, which will be built under the Ampere spinoff in a group of plants in northern France called ElectriCity.
“When we are starting new processes for an electric motor or battery, we can’t take six months to reach full capacity or process optimization,” Marchiol said.
He expects Renault to have a fully functional metaverse within two years, with the final step closing an automatic feedback loop to allow parameters of equipment such as robots or presses to be modified using digital twins.
Such twins are already being used to model new production lines, including programming robots and performing safety checks in virtual reality, “before spending even 1 euro,” Marchiol said.
Renault has found that data collection can lead to unexpected breakthroughs.
“We started with the idea that we could save a lot of money with predictive maintenance, and we did, but it is not the biggest moneymaker,” Marchiol said. “When you start gathering data from all your equipment and develop consistencies in the process, you are able to see the elephant in the room you were not able to see before.”
For Renault, that includes quality issues in the bodyshop that can be traced down to the level of a specific weld, or a process delay by a specific robot on a vehicle configuration that might come down the line only once every three hours.
“The same data that can be used for predictive maintenance can be used for quality or energy usage,” Marchiol said. “Each time we add equipment, we don’t need to ask ourselves, ‘Do we need to collect data?’ It’s an automatic policy.”
Added Marchiol: “We have opened the box on the metaverse, but we need to implement it everywhere with more functions.”
Other automakers are counting on digital twins to drive efficiencies, innovation and cost savings. GM, working with General Electric, has developed what it calls a Virtual Factory Testbed to model processes for build-to-order production; digital twins also test the integration of physical and information systems on the plant floor.
Mercedes, which Kerris described as one of Nvidia’s “lighthouse” customers, will work with the supplier to build a digital twin of its factory in Rastatt, Germany, where it will produce full-electric “entry-luxury” cars on the forthoming MMA platform starting in 2024. Using a digital twin, Mercedes can test the implementation of the platform and its tooling and processes without disrupting existing production.
BMW, like Mercedes, is relying on digital twins (and working with Nvidia) as it prepares its factory network to build full-electric — as well as software-defined — cars. Its iFactory model aims to “make planning and simulation of all processes and the entire production system 100 percent virtual,” the automaker said last June.
“Production planning can integrate the virtual product into the virtual factory at an early stage,” Michele Melchiorre, BMW’s head of production system, planning tool shop, plant construction, said in a release. “This reduces planning effort and capital expenditure and, at the same time, ensures processes are more efficient and more stable during ramp-up.”
BMW’s factory in Regensburg is now fully digitally mapped, and its digital twin is being used for planning future plant structures and configurations. Other BMW plants with digital twins include its U.S. factory in Spartanburg, South Carolina, and its Munich facility.
BMW said it would complete the digital scanning of all its plants by “early 2023,” in partnership with NavVis, which aside from Spot the robot dog, specializes in mobile laser scanners that operators strap to their body as they move around the factory floor.
“We are heading in the direction of a real-time visualization of your factory,” said Phillip Quadstege, senior solution manager digital factory at NavVis.
NavVis scanners capture shop-floor equipment in its exact position, but also the physical space around tooling up to the ceiling. That can determine how much space remains for new tooling, which is crucial if automakers want to expand output.
New tooling comes standard with its own digital representation, Quadstege and other experts said, but older equipment needs to be scanned piece by piece.
“Most of the data the companies have is terribly outdated,” he told Automotive News Europe. “A lot of companies still work in 2D.” Even new tooling needs to be scanned once in place, he said, because a site plan might differ from the actual installation, making the digital twin inaccurate.
Once a plant is scanned, NavVis creates a digital twin using cloud technology.
Quadstege says transparency –the ability for anyone, anywhere within a company to see what is happening in a factory that may be thousands of miles away — is the key enabler of benefits from digital twins.
“There is a lot of potential if you introduce transparency,” he said. “You can reduce investment; product quality goes up; you have efficiency gains in planning.”
Because factories compete against each other for vehicle allocations and investment, such transparency could be seen as a threat, he said. “Plant managers worry about their plant in comparison to others,” he said. “But on the other hand, you can’t avoid digitalization.”