Digital twins: CAD design through the looking glass
Image credit: Florida Institute of Technology’s Center for Lifecycle and Innovation Management
Products are set to live twice: once in real life, once in digital form.
“What they do is test it virtually, in order to tell the doctor what he needs to do for the patient. You don’t want the doctor to open you up and say, ‘OK, what do I need to do here?’. They provide the virtual capability, so a doctor operating on your knee, before he goes in, knows exactly what hardware he needs to put in,” Michael Grieves describes the use in healthcare of a digital twin, the concept he developed more than a decade ago. Although he first proposed the idea in 2003 as a way of developing products more cheaply with the help of computer models, it has since found wider uses.
Grieves is a research professor and assistant director at the Florida Institute of Technology’s Center for Lifecycle and Innovation Management. In his 2003 paper, Digital Twin: Manufacturing Excellence through Virtual Factory Replication, Grieves wrote: “The use of the digital twin extends throughout the product’s life to provide value to its user and information on how it actually performed for its manufacturer.”
The term ‘digital twin’ he credits to an engineer he worked with at Nasa. Whether it is an artificial knee joint or an aero engine, Grieves believes the approach can deliver major savings in design, manufacture, operations and any subsequent servicing. It is an idea that a number of companies working in the computer-aided design (CAD) space have enthusiastically embraced.
Industrial software firm PTC has put a lot of weight behind the idea of using a digital twin of a physical product to guide servicing and maintenance with the help of virtual-reality (VR) and augmented-reality (AR) technology. For example, by holding a tablet armed with a camera in front of a machine that needs repair, the software can interrogate the digital twin to work out what action needs to be taken.
“The CAD data [produced during the design process] can be used to build augmented-reality service instructions that are informed by real data coming off of the product. So, you can see where the failure is; you understand what caused it,” says PTC’s CAD segment senior vice president Brian Thompson.
“This is the killer enterprise industrial use case for Internet-of-Things technology and VR/AR mixed with CAD data that we think will be amazing. It’s a super-informative process when you as a service engineer can see the digital representation of what you need to do to the product.”
Long before the service engineer reaches the product, the virtual twin is born as a CAD model that, in order to be considered a digital twin, has increased in sophistication over the years. Grieves originally set out three tests of what a model must have to be a true digital copy. “I proposed three tests for whether there is a virtual twin and the first test is the vision test, but I have expanded that to all the senses.”
Importantly, the first test is not only about the exterior appearance. The virtual visual inspection would include disassembling the product and seeing completely realistic representations of its component parts.
Grieves’ second test is showing the virtual product react completely realistically to testing, such as a digital wind tunnel or simulated destructive actions.
His third test is about getting information from the virtual product through a physical inspection, just as a user would from an actual product when, for example, inspecting an engine’s fuel usage.
Grieves reflects on industry’s growing ability to meet his tests: “Where the digital twin is today, at the point I started, you couldn’t pass the visual test, but now we’re pretty much there. You can’t tell if it’s a real car or whether it’s a virtual car.”
Dassault Systèmes’ EuroNorth managing director, Stephen Chadwick, agrees. Photorealistic versions of Audi cars have been shown in television adverts. “Using our 3D Excite product it can look as though it was a movie. I can tell you that advertisements you see for Audi on the television or in the cinema are produced purely inside our software platform… What you see as a perfectly beautiful Audi is actually a virtual representation of a real car based on the CAD model.”
A digital copy that largely meets Grieves’ first two tests can travel through a virtual factory and undergo an operation at each simulated station. This is where the digital twin meets design for manufacture and computer-aided engineering (CAE). Each digital factory station, where operations are carried out on the detailed virtual product, sub-assembly and then final assembly, has been modelled by industry for some time.
“We can practically take the digital product and assemble it in the digital factory, validating that the product can be assembled using the manufacturing process we have designed,” explains Siemens’ manufacturing engineering software executive vice president Zvi Feuer. Siemens can simulate a group of machines working together, with a conveyer and a robot, for example, and failures can also be recreated virtually.
Feuer points out how the digital twin can save money for companies through virtual manufacturing engineering. “In the first 10-20 rounds [of starting to manufacture a new product] you will face problems. The information supplied to the supplier was not accurate enough, the supplier has not used the right material or manufacturing process. These parts are not fitting the right way, the tolerance is bigger than expected; we see this, it happens again and again. Today we have the ability to predict the assembly-ability, the tolerances, which have the biggest impact on the process.”
Grieves has a stark warning for companies, especially in the aerospace industry, that cannot provide a virtual product for the manufacturer’s production simulation. “Boeing and Lockheed Martin require companies to send in their virtual models so they can run tests on them. I think that that really is a big concern of the Boeings and Airbuses, that they are getting the virtual information as well as the physical products. They are selecting their suppliers on the basis of, ‘you need to be able to provide this virtual model,’ because they can’t afford the delays they’ve gotten from people that can’t do that.”
Once the CAD software’s virtual product has been improved for manufacture with the digital factory, simulated testing is another area of potential savings. Instead of expensive testing facilities everything is done inside the computer.
“The virtual tests are better than the physical tests because when you test one or two of them in aerospace that is expensive, while I can do a whole lot of testing virtually at virtually no cost at all,” says Grieves.
Chadwick adds: “We find a lot of our customers utilise the [Dassault software] platform to be able to do that type of testing and indeed many have removed the physical test because the certification is no longer required from a physical model; the digital data driven model is far more accurate.”
Once approved and in the field, sensors are expected to provide feedback to help enhance that original design and inform future versions of the product. Last year, PTC gave a demonstration of how a virtual product, in this case a bicycle, can be improved by data from the field. As Thompson explains: “Customers see value in tracking an asset once it is in the field, understanding what is happening within its service life. The capabilities we tried with the bike demo are going to be released in the autumn.”
However, Thompson sees an obstacle with a requirement to feed back data about every aspect of a product. “If you have 10,000 unique CAD models that are real-time tracking 10,000 bikes out in the field that is a lot of data.” His solution is for key design elements to be the focus of data collation, for example, the bike’s forks. He points out that a business may not want a high-fidelity complete model because the cost associated with maintaining that, and the vast amount of data being returned from the field, cannot be justified. Instead, key aspects to the product, like a bike’s forks, reduce the data handling costs while providing an insight for specific product improvement.
Managing the potentially vast amounts of data is not a concern for Grieves. “I am not so concerned about data structures. We have a good ability to move things around,” he says. What does concern him is that he sees an IT industry that needs more cooperation because it is moving too fast for standards. “The best we can hope for is harmonisation - that all the people play nicely together in the sandbox, rather than having proprietary data structures that people can’t get their hands on.”
The elephant in the room with all IT infrastructure is cyber security. “That is one of my big concerns,” says Grieves. “When we start to have smart products that reflect what they look like, we need to make sure that only the authorised agents can gain access to that and not everyone. One of the things I propose is to build into these products a high degree of paranoia. They need to be concerned with who it is that is interrogating their information. We need to be developing very paranoid products.”
Cyber security is also fundamental to another IT element supporting the digital twin concept, the Internet of Things (IoT). A company strategy for IoT is inextricably linked with a virtual product for Thompson. He says: “I find it strange if you say you have a digital-twin strategy without having an IoT platform strategy. You have to have a way to connect those two elements.”
The IoT approach is already being used by aero engine manufacturers, as Grieves explains: “A company doing a good job is Rolls-Royce. If you speak to them, they collect data off of every single engine and every single flight and they have the ability to predict product failure having run this giant correlation engine. When they see this status and that status they know they will have a problem with this component.”
For Feuer, such real-time IoT data collation can help with a company’s business model and not just design improvements or repair servicing. He gives the example of a printing press business. “Take this press: why is it smart, it’s because technology today allows us to add sensors to transmit information back to the manufacturer and so they can offer a new business model, for example, pay-per-print or pay-per-ink.”
These uses for the virtual product sound like aspects of product lifecycle management (PLM). This is because the digital twin concept is now viewed as being part of PLM. The definition of PLM is generally considered to be a process of managing the entire life cycle of a product from inception, through engineering, design and manufacture, to service and its disposal.
Like Grieves’ digital twin, PLM’s degree of implementation differs from industry to industry and company to company, and views vary about the virtual product’s place within it. Feuer says: “Digital twin is a solution within PLM on the way to Industry 4.0,” - the concept of the digital industrial enterprise.
Thompson sees PLM as the genesis of the virtual product and the place where the digital copy will be ‘born’. He also sees the twin as being something that draws together data from the PLM domain and areas such as enterprise resource planning (ERP), which he places beyond life-cycle management.
“Once you start manufacturing the product, most times that is when the connection to the PLM system ends and you’re now in the role of ERP, or something like that, which is where that [twin] gets tracked.”
The twin effectively brings together all the relevant information the company needs from within the PLM activities and others beyond it, such as in-service sensors, into the original CAD model.
Feuer has a wider view of PLM: “It is growing in its domain. If you talk to my boss, maybe 15 years ago he would say PLM is CAD and CAD data management and some CAE tools. And when you talk to him today he is going to talk about PLM covering the entire span of the smart product, which includes requirements system engineering, even before CAD, then domains of CAD, CAE, manufacturing, software, life-cycle management, and integration for electronics.”
Just as industry has developed Grieve’s original idea for the digital twin, PLM has not stood still. As Feuer puts it, “we are constantly realising that there is more to PLM than when this journey began 35 years ago.”
Grieves’ digital twin concept is slowly being fully realised, technologically. In terms of uptake, aerospace is probably the most advanced. What is clear is that in the 13 years since Grieves and his Nasa colleague came up with the idea, the concept is seen as having advantages beyond those originally envisaged and will ultimately have a much wider impact including business models and servicing. Where will it be in 35 years?