When you think of an “emerging” technology, it may evoke the idea of something developing and on the edge of breaking through, but the reality is a little more complicated than that. In the case of additive manufacturing (AM) – the equipment, the material, the processes – its emergence was more along the lines of something evolving. And while emerging technology may be thought of as breaking out of beta testing to the workaday world, AM has been around for decades.
AM is generally accepted as having gotten its start in 1984, when Chuck Hull developed stereolithography – a photochemical process that uses a digitally directed laser beam to build up, layer by layer, a structure from a liquid polymer. Hull went on to found 3D Systems and released his first production machine, the SLA-1, in 1987. Whichever date you accept, this isn’t what might be considered “emerging.”
Like a Horse Race
Peter Zelinski, editor-in-chief of Additive Manufacturing Media, has a different perspective on the technology. For the past several years, he’s covered the global AM scene, creating YouTube videos, podcasts, websites, and a print publication – all of which means he’s been on the ground, talking with people in the industry to an extent that others can only imagine. While the array of processes encompassed by the moniker “additive manufacturing” are all advancing, he says that this is happening at different rates. Consequently, there are “jockeying moves,” as different processes come to the fore while others fall back.
Zelinski says that, currently, processes like the directed energy deposition of metals and the use of granulated polymers – in lieu of spools of filament – are making strides in AM.
But he points out that there is a wide array of AM technologies available to manufacturers, depending on the application – which drives process choice and material requirements.
A Wide Array
There is vat photopolymerization, which encompasses the original stereolithography as well as things like direct light processing.
There is powder bed fusion, which has several subsets, including selective laser sintering, selective laser melting, and electron beam melting.
There is binder jetting, which uses an adhesive (i.e., the “binder” that is jetted).
There is material jetting – plastic or metal – where molten materials are printed and then harden as they cool.
There is direct energy deposition (DED), which uses a laser to melt the powder material as it is deposited.
And there are more.
DED, Zelinski notes, has improved to the point that it is now capable of generating finer details and is being used to build aircraft and space components with complex inner channels. And there are cases where the build material is provided by a wire feeder, which facilitates making large “castings.”
(One area where there is considerable growth in the AM field is the production of larger parts.)
And along with process development goes material development, so a wide array of metals, plastics, and composites are now available for use with AM.
Zelinski points out, however, that more than a small amount of material engineering is necessary; while the same material may be used to additively produce a part as in, say, molding it, microstructural changes due to the nature of the additive process may not meet the part requirement.
He says this is more pronounced in metal applications. “Is the Inconel 718 that has gone through a laser powder bed fusion the same as one that has been cast?” Given the applications where that material is commonly used, slight differences in metallurgy are not likely at all acceptable.
But Zelinski makes an important point: “3D printing systems for industrial applications can do industrial work. The barriers holding back additive are numbers – for a given application, it might not be fast enough. Or the cost per part you need to get to isn’t achievable. Or the part produced isn’t precise enough without downstream steps that would add more cost to it.”
He adds, importantly, “Additive technologies are advancing quickly. Numbers for it are changing way faster than the capabilities of more established technologies.”
It’s what happens with an emerging technology.
Still, the numbers may not work for everyone. But the number of companies for which AM is becoming more and more efficient and effective is growing.
Although the actual Apollo 13 mission occurred in April 1970, most of us probably became familiar with it during the summer of 1995, when the Ron Howard movie starring Tom Hanks was released.
Among the many interesting plot points of “Apollo 13” – a testament to clever MacGyvering – is NASA’s creation of a digital twin of the spacecraft to facilitate the ground crew’s understanding of what was going on thousands of miles away in space.
So, again: an emerging technology that has been around for a while but has significantly improved thanks to recent advancements in sensors, software, and data handling.
Twin Types
In a sense, if we consider digital twins in the context of human twins, digital twins are both identical and fraternal.
That is, identical twins come from the splitting of a single egg, which results in both embryos having the same DNA. Fraternal twins come from two eggs and two sperm cells, created at the same time but lacking identical DNA.
A digital twin, created with CAD and CAE tools, is an identical model of a physical object, whether that object is a part or a system. So, the digital twin is fundamentally different than its counterpart but is engineered the same.
The physical object is fitted with sensors, measured with other sensors (e.g., cameras), and set up as an IoT device. The combination of these elements provides real-time information about parameters and consequent changes to the physical object (e.g., what happens when the knob is repeatedly turned to 11?) to a digital model of the physical object. This digital twin then embodies that information to predict results, providing important data to the user, who can then improve whatever object is being modeled.
Heating or Rotating
Consider, for example, the top surface of an automotive instrument panel made from ABS material. For issues related to both cost and mass, there is an effort to make the part as thin as possible. Given the sun load in places like Phoenix, Arizona, the temperature of a dashboard can reach nearly 160 degrees. So, this would be an ideal situation to use a thermocouple to measure the temperature and amount of warpage that occurs in real life from the heat – data that is then sent to the digital twin.
That then allows digital analysis based on real-world information. As a result, the material, thickness of the shell, design of the component, or some other characteristic can be adjusted as needed to improve future instrument panels.
Or consider a machine with rotating elements. A digital twin of the bearings can be created to measure the behavior of the actual bearings over time and use. Companies can then predict the machine’s maintenance needs, allowing users to schedule routine service at their convenience, minimizing downtime and keeping the physical machine running in peak condition.
In fact, McKinsey consultants note that machine building – particularly in the area of customized, special machines – can greatly benefit from digital twins: not necessarily in the context of having models of the machines, which are one-offs, but by creating a library of models of key components (like the aforementioned bearing arrangements) that can be assembled.
Big Benefits
The impacts on both process and product is significant. According to a survey conducted of the aerospace and defense industries by Capgemini Research Institute:
75% say digital twins improve value from the start – when design commences.
81% say there are operational improvements, such as availability and reliability of equipment.
73% say there is improvement in the production rate.
76% say there is an improvement in quality.
And from a competitive point of view, it is worth considering that 73% of companies surveyed said they have a long-term roadmap (more than five years) for digital twins, and 61% said that digital twins are a strategic part of their digital transformation.
Said simply, the aerospace and defense industry perceives digital twins as providing a significant advantage – and there is no reason to think that this isn’t the case for any durable goods industry.
Not Easy
Those of a certain age will remember a series of books for students written to help them with their studies: “Latin Made Simple,” “Calculus Made Simple,” and so on.
Note that they weren’t … made easy.
Because there is a big difference between the two operative adjectives.
Digital twins are simple, but creating them isn’t exactly easy. Doing so is a cross-functional and cross-technical undertaking that involves designers, engineers, product experts, and production personnel – to name a few.
The list of technologies relating to digital twins is rather extensive. They include the aforementioned sensors and IoT devices; edge computing devices, routers and servers; software including digital twin platforms (yes, there are specific software available from all the major vendors), 3D modeling and simulation, data analysis and visualization; machine learning – and the list goes on. How big is this? According to McKinsey, the global market for digital twin technologies is anticipated to reach the order of $73.5 billion by 2027.
Again, the creation and operation of digital twins is not easy. And while vendors can help make it simple, it is still essentially complex. But there should be little doubt that (1) this technology is emerging in a big way, and (2) consequently, those who don’t engage with it – and possibly fully embrace it – are going to find themselves at a competitive disadvantage because others are.
Stephen Laaper is a principal at Deloitte Consulting LLP and a manufacturing strategy and smart operations leader in Deloitte’s Supply Chain & Network Operations practice. He leads the firm’s smart manufacturing services.
Laaper understands manufacturing. Given that, it isn't surprising that he talks a lot about generative AI (GenAI). However, it is surprising that when describing the capabilities of GenAI in the process of making things, he cites tool clearances as an example.
GenAI is trained on text, images, and audio and can create solutions and recommendations related to that content. While Deloitte has identified seven total types of AI, GenAI is all we need to consider for the purposes here.
As has been the case with other emerging technologies, Laaper says that AI has been around for a while but that there is now tremendous interest across the board, from consumers to manufacturers.
“We recently completed a comprehensive AI strategy and road mapping with a large OEM,” he says.
What’s more, the technology has moved beyond just proof of concept and is now adding direct value through deployment.
Which brings us back to the tool clearances.
Working the AI
Laaper points out that during the development of a product, there are “very robust exchanges between manufacturing engineers and design engineers.” One of the issues that must be addressed during this process is whether what is designed can be manufactured – as in whether there are required tool clearances (e.g., for a spotwelding gun to fit into a particular area).
Laaper says that typically requires spending a non-trivial amount of time ensuring that the space for tools is included in the CAD design, space that is necessary when going from digital to physical iterations.
But an alternative is to create an AI agent that can go through the CAD model and assess whether the included tooling clearances are sufficient. If they aren’t, then the agent flags them and provides recommendations to address the issue.
He says that this can take months out of the development cycle.
Note that the system only makes recommendations. The human user is not out of the loop but rather significantly supported in their decision-making.
Fixing the Robot
Another benefit Laaper cites of GenAI is in the area of production maintenance.
Say there is robot, and all of the available information about the robot – from the manuals to the operational data, historic and current, to failure mode and effects analysis to failure codes to faults and resolutions – has been used to train the GenAI system.
Now imagine it is third shift, when a fairly new maintenance crew is working.
“One of the problems that many manufacturing companies have is transitioning the institutional knowledge that has been built up over the years to new people,” Laager says.
People like those working third shift.
Something goes wrong with the robot. The maintenance people can then query the GenAI system with natural language (i.e., the way they ordinarily talk – not some sort of technical language) and get recommendations to fix the problem.
An immediate effect of this is that mean-time to repair goes down. But there are other benefits as well. Consider that when someone is trying to repair something, they may change out a part, then see if it does the job. If it doesn’t, then that new part likely remains, and another is tested. Eventually, the repair is made. But at a cost in time – and parts.
One of the things that Laager emphasizes is that the GenAI systems operate as assistants. This is not a case of decision-making done by a system but rather of a system providing recommendations to the human operator, who then decides a course of action.
Integrating the Existing
Another thing Laaper notes is that this isn’t a case where what has been operational in the past (e.g., a lean production system) is simply replaced. Rather, he says that it is important to integrate the existing systems with the digital systems in order to get a better result.
While he doesn’t minimize the amount of work that must be done to train the AI system – after all, there are probably manuals and notebooks on shelves that need to be inputted and up-to-date information, like spare-parts inventory, must be accessible – he says that, in his experience, he has seen solutions achieved in as little as 10 weeks’ time facilitated by things like pre-built accelerators, technology that Deloitte has invested in.
Software Changes
In the auto world, efforts are underway to develop “software-defined vehicles,” where modifications can be made to the performance of a vehicle through software – such as increasing the range of an electric vehicle by adjusting parameters with an over-the-air update – like updating a smartphone.
Laager says that “software-defined manufacturing” is emerging. Analogously to the vehicle situation, improvements in throughput or enhanced functionality can be achieved digitally.
None of this is merely theoretical. And those who embrace it will undoubtedly be the ones who are the most competitive in their industries, which, arguably, is the case with all of these emerging technologies.
The takeaway? Don’t wait for tomorrow, because this tech is here today.
And the Robots
The thing about industrial robots is that they’re not exactly new.
At least for some companies.
The first application of an industrial robot occurred in 1961. The Unimate robot, which was produced by Unimation Inc., used an arm that had been invented and patented by George Devol in 1954. Devol and his business partner, Joe Engelberger, founded Unimation in 1956.
The application in question was tending a diecasting machine in a General Motors plant.
Chuck Brandt, chief technology officer at the ARM Institute, points out that while there are several companies – like General Motors – that have a long history in robotic deployment, most of the manufacturing firms in the United States are small, and consequently, for many of these smaller firms, robotic technology is emerging.
According to the most recent figures from the International Federation of Robotics, there are 285 robots per 10,000 manufacturing employees in the United States; by contrast, there are 397 in Japan, 415 in Germany, and 1,012 in South Korea.
With that kind of disparity, there is a lot of upside to robots in U.S. firms. Brandt says this is particularly the case, as finding employees is difficult. So, automation makes sense for simpler tasks, like machine tending.
Brandt says there is notable growth in the deployment of collaborative robots, thanks to their simplicity in deployment (“These companies don’t have roboticists.”) and safety, which allows them to work in closer proximity to humans than conventional industrial robots.
To read the rest of the Emerging Technology Issue of MT Magazine, click here.
