Powerful trends will push manufacturing close to complete automation by 2050, while the people still working in the industry will be empowered to rapidly innovate like never before.
Years ago Warren Bennis predicted that, “The factory of the future will have only two employees, a man and a dog. The man will be there to feed the dog. The dog will be there to keep the man from touching the equipment.”
We’re not quite there yet. But a number of powerful and interconnected trends will push us close to that state by 2050, while the people still in manufacturing will be empowered to rapidly innovate and build like never before.
Multitasking and Automation
It’s a safe bet that by 2050 the average machine tool will be fully automated and more capable. Multitasking will be common, perhaps nearly universal. The trend is well established.
As industry veteran Scott Walker, chairman, Mitsui Seiki USA (Franklin Lakes, NJ) observed: “In the early 2000s, the North American market for five-axis machines was 150. Today it’s 3,000. Machines are also combining grinding and milling, or laser metal deposition and milling, or grinding and work hardening.” He added that while the benefit is the ability to accomplish more inside the work envelope, the “nightmare” has been getting all these functions to operate properly and consistently. “But that will change as technology, monitoring and software all get better.”
The elephant in the room for many manufacturers is the extent to which 3D printing will alter the technology mix, and beyond that, its implications for product design and a host of other issues. So far, the speed limitations and high raw material costs of additive manufacturing have severely limited its viability beyond prototyping. But Terry Wohlers, principal consultant and president, Wohlers Associates (Fort Collins, CO), said speed “won’t be an enemy” by 2050.
Take a powder bed system: The bulk of the production time is in tracing the surface with the laser to fuse the material. “But systems are now available with many lasers working simultaneously on a build platform,” Wohlers said. “The energy from an electron beam can be split into as many as 100 beams to help speed the process.” On the other hand, these approaches require a lot of energy, which is expensive. Wohlers thinks we’ll overcome those limitations, perhaps by “harnessing the energy of the sun directly to melt material, rather than plugging into a 440 outlet.”
Wohlers added that directed energy deposition is inherently faster than the powder bed method for building metallic components, but “users are limited in the objects they can create and there is a trade-off in resolution, generally requiring machining, and sometimes a significant amount.” This brings us back to hybrid systems that combine additive with CNC milling. Like Walker, Wohlers believes the problems in getting these two approaches to work harmoniously will largely be solved in the next 32 years.
Another factor arguing for greater use of additive techniques is an expected drop in material costs and a wider array to choose from. “Today’s machines work with only a few dozen thermoplastics, for example,” Wohlers said, “yet thousands are available for conventional manufacturing.”
Perhaps more important, the polymers currently used in 3D printing cost up to 50 times as much as similar polymers for conventional manufacturing. That puts the breakeven point in the hundreds to thousands of units depending on the size of the part. But Wohlers said many of the patents on machines that produce parts with polymers have expired, leading to new machines that use lower-cost materials. “The breakeven point will improve dramatically so that additive will challenge injection molding for a much wider range of products, including higher volume applications.”
There’s a similar case to be made for metal, but Walker, for one, is skeptical that additive will achieve a cost profile in metals that justifies replacing traditional methods. “It’s so much easier to heat 60 tons and roll sheet metal than building something with powdered metal or cladding,” Walker said. “I look at additive as a function you can put in a machine envelope to add value to the process. But I don’t look at additive as a replacement process for making steel, unless the technology changes and we get to the point of molecular manipulation using a different kind of energy source.”
Additive manufacturing does have one ace left, at least for some players: The ability to create forms that would otherwise be impossible. This not only opens up the potential for new products and features, it also helps alleviate 3D printing’s speed problem. That’s because an open lattice structure made possible by 3D printing can achieve the required strength and stiffness needed for many applications with far less material than a solid structure. And 3D printing’s production speed is directly proportional to the cubic volume of material. Strange new structures overlap nicely with our next topic.
Automated, Creative Design
In Walker’s view, manufacturing is poised to achieve its biggest productivity gains in two areas, one of which is in digitizing all the work that’s necessary to prepare a manufacturing process. “Today a designer starts out with a digital model, and then generates a tool path … then someone designs a fixture … then you get a forging … then an application engineer drip feeds the program to the machine and goes tool by tool and monitors how the cut sounds and how it looks … and eventually he gets the machine to make the part … and then he refines all the motions so he can reduce the cycle time.” It’s actually even worse because getting the initial design is also cumbersome. Luckily, many bright minds are working hard to ease and speed every part of this process.
At the front end, generative design technology is helping an ever wider group of creatives to quickly explore new geometric possibilities. In the case of Fashion 360 from Autodesk (San Rafael, CA), the software runs on the cloud and uses machine learning and artificial intelligence (AI) to automatically generate hundreds of designs that each satisfy the designer’s criteria for strength, cost, manufacturing method, material and so forth. What’s more, explained Bob Yancey, Autodesk’s director of manufacturing and production strategy, the designs are “not just some impossible-to-use idealized geometry, they are real working CAD models that can be further manipulated in CAD software.”
They are also what Yancey refers to as “manufacturing aware,” which means they started with the desired manufacturing methods built in as a constraint from the beginning. “So if you specify that the part needs to be able to be machined on a five-axis CNC, all of your design options will conform to that constraint,” he said.
It doesn’t eliminate the need for a human designer. As Yancey put it, “Describing the design challenge with precision and expertise is an engineering skill that will not go away. What generative design software does is give you more design options than any human could conceive on their own, so you have greater confidence that you are considering far more options and getting better outcomes. We see this as a future of co-creation between engineer and computer, or human intelligence and artificial intelligence.”
Perhaps more profound, the designs are often startling and superior to what a human would have envisioned. As Diego Tamburini, principal industry lead—Manufacturing Industry Communities / Cloud + AI Division for Microsoft (Redmond, WA) put it, “If I sit in front of my CAD tool to design a part, I already have 1,000 preconceptions about how it should look based on centuries of previous designs. AI has no such notions. And while I agree that automating design is a hard problem and it’s difficult to imagine a computer designing complex items, we have to recognize that conversely there are many instances in which we’re living with sub-optimal human designs.”
Some of our preconceptions stem from our experience with raw materials being limited to blocks, bars and sheets. But 3D printing is not limited to those raw materials. Nor is it limited in the shapes it can create.
If you remove those constraints and let the AI work, generative design often creates shapes that are “completely different from what we’re used to. More organic, like animal bones,” said Tamburini.
Wohlers echoed this and said that nature offers excellent examples of structures with remarkable strength-to-weight ratios. Until recently, 3D printers produced lattice, mesh or cellular structures defined by canned programs with little understanding of their strength properties. “The newest topology optimization tools can produce engineered lattice and mesh structures with certainty of strength,” Wohlers said. “In the future, we could see super lightweight structures in different metal alloys that are lighter than carbon fiber composites, which is time-consuming to produce and expensive.”
Tamburini said he’s seen cases in which the computer comes up with a latticed shape that doesn’t inspire confidence in the human (it looks too light and flimsy), so the designer covers it with something so it looks sturdier. People remain people, after all.
Speeding the Process
Krisztina “Z” Holly, founder and chief instigator of Make It In LA (Los Angeles) underlined the benefit of combining ever smarter software with 3D printing and other new technologies (like virtual reality) to greatly speed the iterative product development cycle. Besides the ability to get more feedback from the consumer earlier in the process, which may lead to much better products, she pointed out that the new tools democratize the design and build process.
“What does that mean in terms of how we innovate and who innovates? I think the world will be a different place if we allow non-engineers to design the kinds of products they want,” she said. “It will also be easier for entrepreneurs to start a manufacturing business. What kinds of products will become available if people can start a physical products business as easily as they can a digital products business?”
One outcome she envisions: “Two sets of skills become incredibly valuable. One is in-depth technological skills for coding these systems and understanding the nitty gritty of what works and what doesn’t.” The other is empathetic understanding of the needs of the customer and market opportunities.
What skills become less critical? Actually running the machines. The process from design to CAM will be more or less automated. As Walker explained, if the design model included information about the material (as is becoming standard), the machine should “have the intelligence to do the rest. Take the right tools out of an 8,000 tool rack and follow the correct toolpaths at the appropriate speeds. The machine should have the vision and audible monitoring capabilities to avoid collisions and also gauge the cutting conditions and adjust speeds and feeds accordingly. That’s what applications engineers do today. How many will we need in 30 years? Hopefully none.”
Having said that, Holly cautioned against a focus solely on jobs. First, there will be jobs—just different, more creative jobs. And second, “The important thing is keeping innovation local,” she said. “There are a lot of negatives to sending design and manufacturing overseas. You lose control of the intellectual property. It’s not good for the environment. And you lose sight of what’s possible unless you have your hands in manufacturing itself.”
Built to Order…Locally
Everyone seems to agree that manufacturing will become much more geographically dispersed, a process greatly aided by the multiplying capabilities of individual machine tools. Yancey said many manufacturers want to both decrease risk and make products closer to the customer to better adapt them to the market.
Walker agreed and also predicted that transportation costs and environmental impacts will drive companies to produce locally. He added that there are also government mandated offsets, in which a manufacturer must produce a certain number of parts in a country in order to sell products in that country. Another motivator is maintaining profitability despite fluctuations in currencies, a problem exacerbated by tightening margins.
At the same time, there will be a much greater degree of customization and a much tighter supply chain. As Tamburini summed up: “The practice of forecasting demand and mass producing parts to meet expected demand will be turned on its head. It will be closer to the customer telling the manufacturer exactly what they want and the manufacturer making it then and only then. Digitization and automation are making this dream more technically, and even economically, feasible.”
This would not apply to every product, and the lines between standardization, customization and personalization are blurry. But Tamburini is certain that the practice of customizing products via a list of pre-defined options will grow exponentially. Some products, like prosthetics and clothing, may be entirely personalized. Likewise, most manufacturers will be generalized service bureaus and not specialists. The machines will build whatever comes over the cloud, on demand.
As Walker put it, “Today a company needs a three- to five-year contract to make a component because the cost to buy, program and tool up a machine to make the part consistently and accurately is huge.” Make machines automated and multi-functional and eliminate much of the setup effort and manufacturing becomes a more nimble, and perhaps a lower margin business. As Holly of Make It in LA would have it, the design process and the design tools and interfaces become even more important.
Keeping Things Humming
Increased digitization, “hyper-connectivity,” and AI should greatly improve our ability to keep production running with a minimum of manpower and downtime. Tamburini said most of the data now being collected is used to monitor what’s going on in the factory and throughout the supply chain. “But we’re starting to ask ‘why’ certain things happen and using AI to predict what will happen. The next phase in this process is using AI and machine learning to enable autonomous responses.”
In other words, with enough data to analyze, machine learning can accurately predict specific part failures. With good decision algorithms and knowledge about all the production demands on the shop floor, the system can also decide for itself what to do about the pending failure: order the part, schedule the downtime, move certain jobs to alternate machines, and so on. You could even envision a machine fixing itself or ordering the robot that can, though Walker said he doesn’t think we’ll ever get away from the need for human maintenance technicians. However, he does think the machines will communicate audibly about what needs to be done—no need for handheld devices or screens and controls.
Tamburini said Microsoft has a head-mounted product (HoloLens) that allows you to interact with holograms around you. “It overlays digital info on top of reality, giving you super powers, in a sense. People are finding that augmented reality can be used to do things like provide assembly instructions, or QC instructions, or maintenance instructions, thus reducing the need for training.” For example, a remote maintenance expert can assist a local technician by pointing to a part or indicating how to move a part, as if they are both looking at the same thing in the same shop together.
Finally, Tamburini pointed out that one of the beauties of machine learning is that “the moment they get better, that capability or knowledge can be instantaneously broadcast to the entire world, because it’s just software. So everybody gets smarter and better, assuming we can share data.” He contrasted this with relying on an expert in the plant who uses his own years of experience to interpret the sounds of the machines and the like. “It’s very challenging to distribute that kind of knowledge.”
What Won’t Change
To the extent it came up at all, the experts don’t seem to think that manufacturing precision will advance much in the next 30 years. “We’re working at tolerances now where the metrology to determine the accuracy is the bigger problem,” as Walker put it. “The next step to getting better tolerances would be molecular manipulation,” (which no one envisioned). No one seemed to think that machining speed would be significantly faster either. Even the improvement to 3D printing speeds discussed earlier will be more evolutionary than revolutionary—not as significant as the increase in productivity due to software improvements. Likewise, our current ability to produce tiny components is already amazing. Wohlers referred to miniaturization as “partly a solution looking for a problem. One of the few applications are tiny sensors embedded within 3D printed parts.”
If you’re worried about the changes, Walker might comfort you with this thought: “We’ve been tweaking manufacturing since the 1780s. The next 30 years will be more tweaking, unless we come up with something truly revolutionary.” He asked me if anyone said they’d figured out how to manipulate gravity so we could fly to the moon without burning fossil fuels, knowing the answer. Come to think of it, no one mentioned the dog who kept the man from making changes to the machine, either.