The path to digitize a factory is both closer and cheaper than most engineers realize.

Reshoring has been a buzzword for a few years now. But when supply chains are undergoing dramatic disruption and inflation is raging worldwide, what is the reality?

According to research firm IDTechEx, it’s only a matter of time before an array of sensors and cobots spur far greater automation and flexibility. The firm recently published a white paper titled “Factory of the Future” that summarizes the expected advancements. Indeed, some of these changes are both relatively inexpensive and simple in scope yet open a realm of possibilities for greater process control.

IDTechEx senior technology analyst Matthew Dyson, Ph.D., who co-authored the paper, discussed the key trends in industrial manufacturing and the timeline for adoption with PCEA president Mike Buetow in late July. The following is lightly edited.

Mike Buetow: You just co-authored a white paper titled “Factory of the Future.” Lots of people, of course, are considering what that looks like. What spurred your interest?

Matthew Dyson: It’s the combination of technologies that we see being developed. The white paper is a compelling use case for them. It’s about how you can make manufacturing more efficient to address concerns like reshoring, inflation and so on.

It’s about incorporating sensors, incorporating cobots, incorporating predicted maintenance, and is one of these application areas that is both useful to society and includes a lot of different and separate technologies, each interesting in its own right.

Matt Dyson Headshot 2022

Matthew Dyson

MB: In the white paper, you boiled down the next wave of industrial manufacturing to a few key pivotal trends: sensors, additive manufacturing, automation and flexibility, which in your research you talked about AGVs or cobots. How did you arrive at those?

MD: At IDTechEx, we try to work out how technologies are going to evolve over the next five to 10 years. That involves talking to a lot of people and attending a lot of conferences around the world.

What we see is basically these trends that we mentioned in the white paper. There are lots of companies talking about how they can integrate more and more sensors into their equipment. Not just for things like process control, which is pretty standard by now, but also monitoring the machines themselves, which is somewhat less common. These can be all different types of sensors: they can be monitoring temperature or vibration, for example, and feeding back that data continuously so that you can identify this machine is vibrating more than expected, and maybe we need to have a look at that before the whole production line goes down and costs a lot of money.

We’ve seen lots of other examples of people aiming to integrate sensors into manufacturing processes, even the things that seem kind of prosaic, like keeping track of the number of parts in a parts bay. This can be done by putting a little pressure sensor in there that then automatically orders more parts when the level drops to a certain point. There is a company that makes gaskets that’s incorporated an RFID tag so that they can check continuously whether it’s been installed correctly, and whether it’s been subject to the appropriate amount of pressure. These are all examples of not just monitoring the quality of the product that is produced but monitoring the manufacturing process itself in real time.

I think that’s the key kind of stat when you hear people talking about Industry 4.0 and the digital factory and mass digitization and all these kinds of buzzwords. What it’s about is continuously monitoring not just the product but the processes involved in producing the product so that you can identify problems early; so you can make small adjustments; so the quality of the product is maintained; so that your supply chain can be as efficient as possible; so that you can check the installation has been done correctly, or repair has been done correctly without having to pay somebody to go around and check manually. All these things will be communicating wirelessly. You will be able to build up these huge datasets that will enable the monitoring of any other factors that could be influencing actual quality.

It’s that mass digitalization that is the key trend.

9 dyson figure 1

Figure 1. Sensors, cobots and AGVs make up some of coming automation features.

MB: Did your research get into the higher-level systems that must be implemented to be able to access that data in a timely fashion? We could be looking at billions and billions of data points in a typical factory that need to then be processed very quickly to adapt in real time.

MD: It’s a very interesting point. Our research is focused mainly on the hardware, so I’m coming at this from the materials, sensors and antennas side, but I have looked into the software to some extent. This touches on the idea of AGI, so how can you reduce the amount of data that need to be transmitted, and of course one way is to do some of the processing at the relevant location before they get transmitted.

I suggest that in some cases you don’t necessarily need to take readings every 10th of a second or something. If you have a manufacturing line and just want to monitor the vibration or frequency or amplitude of some machine, detecting that every few minutes is probably completely adequate.

If there’s a problem with the machine, the vibrations are gradually going to get worse over days. So relative to, say, many of the other applications where continuous data monitoring is needed, such as autonomous vehicles, in many cases a factory is more straightforward. In some cases, it can just be a check just after installation. The gaskets I mentioned, someone who installed them could scan them to ensure they are seated correctly, which could be hard to tell once they’ve been installed.

There’s an interesting point in the business models that people are adopting. Some of the manufacturers of large-scale equipment are keen to offer a full solution with all the digital services in there, and they have their own teams of software engineers doing the data management. The other approach is to outsource all that to a separate provider that would provide just the sensors to stick on the machines, and they’ll monitor the data for you and give you a dashboard on your phone or computer with a few of the tangible insights.

It depends on the company strategy and software resources they have available as to whether they want to bundle all the digital factory ideas into their products or just work with somebody else and do it afterwards. I think the data challenge is significantly easier than, say, autonomous vehicles.

MB: If I extrapolate then, the combination of sensors everywhere plus some degree of artificial intelligence would learn and anticipate how often sampling would be needed.

MD: Absolutely. You would identify when your sampling would be needed by how quickly these problems evolve. You would also use artificial intelligence, and it doesn’t need to be that sophisticated. To take the vibration example, you’ve got a threshold where you’re happy if the vibration is within that range of amplitude and frequency. As soon as it moves out of that frequency or amplitude for a specified period, then some kind of alarm sounds, and a manager or maintenance person is alerted. These kinds of algorithms are not particularly different from what you might get in an intensive care unit where they monitor the patient’s heart rate and if it’s too much outside the parameters, it will give some sort of alert. It’s been around for years, and that’s a higher-value use case, but you’re applying fairly similar ideas but now much more widely across factories.

MB: I’m guessing you talked to a range of companies, OEMs, product manufacturers and so forth in the course of preparing your white paper and perhaps across a variety of regions as well.

MD: Absolutely. We’ve spoken with people who make manufacturing equipment, some of which are quite big; sensor providers; research institutes that set up pilot lines that highlight this kind of technology built into them so that companies could see what’s possible. Those are happening everywhere, and often SMEs [small- or medium-sized enterprises] are not necessarily aware of what is possible and what the technology can offer them. There’s quite a lot of interest from government to facilitate that kind of process by having these pilot lines where you can see all that’s automated and all this predictive maintenance that’s happening.

MB: Did you uncover any regional differences in what I would say is "current" implementation?

MD: There’s obviously a kind of trade-off with wages. Broadly speaking, it’s the more developed economies, where wages are higher, where – probably unsurprisingly – doing more of this kind of factory automation makes commercial sense. That is part of the aim of reshoring, as I just discussed, in that if you’re going to bring manufacturing back from lower-wage economies, you’ve got to automate an awful lot of it to keep it cost-effective.

That’s one of the motivations, so you see a lot of interest in high-wage economies like Germany and Scandinavia. A lot of effort is going into it there. I think this reshoring trend is sort of happening in general, and these technologies, at least ideally, should be able to do that without doubling the price of everything. The wages are higher, but you need fewer people to run the lines; it needs repairing less often; maintenance less often; and checked less often, because all of that’s being done digitally.

MB: Cost, of course, is a big limiter. Does your research attempt to assess the capex for various-sized operations?

MD: We haven’t looked yet at the kind of capital that is required. There are so many different types of factories manufacturing different types of things. We primarily focused on the hardware technologies that would facilitate this transition. To give a few examples, we’ve seen a Japanese manufacturer with a novel method of making PCBs, and it has a lot of interest from Toyota, which is keen to bring its PCB manufacturing back to Japan. There are all sorts of these supply-chain security concerns, geopolitical tensions and that sort of thing, and there’s a price that people want to pay. And these technologies are a way of being able to meet those supply-chain security requirements and geopolitical considerations without having to pay a lot more for everything.

How much does it cost to get all this stuff installed, to digitize a factory, is a difficult question. All factories are different. But technologies are being developed to make that cheaper than you might imagine. You don’t have to buy completely new machines. For example, there are some new projects at some German research centers like the Fraunhofer Institute using 3-D printed electronics to make sensor housings that can then be integrated into fully depreciated old equipment that still works well. So if you’ve got a bit of heavy industrial equipment that is old but works fine and you want to digitize it, you can add sensors to it without having to spend all the extra money on an entirely new piece of equipment.

A lot of these sensors can be quite simple, and people talk about things like sticker electronics, which aren’t necessarily quite there yet. But people use these kinds of energy harvesting approaches so that this doesn’t need to be wired in. You could get a magnet and stick something on it that will harvest energy from the vibrations and use AGI to send the data back. That whole unit might be a few hundred dollars.

A digital factory doesn’t necessarily mean spending millions on new equipment. It can mean: make judicious investments in instruments; have some subscription to some kind of software; and install a selection of relatively low-cost sensors. It certainly shouldn’t mean we need to fully refit the factory.

MB: What I’m hearing is, just like there’s a wide variety of factory layouts and processes and focuses today, that won’t change with Industry 4.0. These sensors and cobots and other things you’ve mentioned will be implemented in a specific and individual way. We won’t see a generic factory where everyone has the same kind of line and information collection and so forth.

MD: I absolutely agree. Just gradually like at your home, more and more things will become digital. Five years ago, I didn’t have a smartwatch. Now I do. These things kind of gradually progress, but there’s no need to go for a wholesale readjustment. You can identify, “OK, I have this machine. I don’t really know why it keeps going wrong. Maybe I could get some advanced warning if I installed some sensors.” And probably without knowing it, you’ve become part of Industry 4.0 because now you [now] have continuous monitoring enabling predictive maintenance on this particular machine.

It’s not: “Oh, I have to start with a blank piece of paper and a thick checkbook.” It’s: “OK, I’m going to make these incremental improvements to take advantage of new, low-cost sensing technology, improvements in energy harvesting, and particularly the lower cost of client services processing, to solve particular problems within my factory.” “Oh, I keep running out of this equipment because its use isn’t linear, and I don’t know when to order. Let’s put a sensor in here to track when that’s about to run out.” “This equipment keeps breaking down. I don’t really know why. Can I check that?” “My maintenance people keep making mistakes installing the stuff. Can I put something in to check?”

On the maintenance side, there’s also an interesting kind of placebo effect. If you start adding the ability to monitor how effectively maintenance has been performed, you apparently see the maintenance being performed better, even if you don’t go and check afterwards. Go back to the gasket example: if you make the gaskets able to record whether they were installed correctly, then a much higher proportion will be installed correctly because people now know they’re being monitored. You end up with these unexpected benefits, but by better monitoring how various tasks are performed, perhaps unsurprisingly, people on average will perform them better. And you can still identify the odd mistake.

9 dyson figure 2

Figure 2. Digital dashboards monitoring machines and processes and engaging in predictive analytics will become the norm.

MB: Is that something that you observed yourself, or something that over the course of your research and interviews was brought to your attention?

MD: This is something I have been told or have seen in presentations from multiple people from different, unconnected companies. In one of the more interesting ones – which isn’t quite a factory – a Swedish company developed a device to monitor moisture levels behind tiling in showers. You scan it with an RFID scanner; it’s just a coil, and the frequency changes with the amount of humidity. And then they discovered their showers were being installed much better once people started putting the stuff in. The building company was getting far fewer callbacks from people saying “my shower is leaking” once they obliged their staff to install this device. And of course, the device doesn’t actually need to work. It could just be a piece of plastic. But if you tell everyone this will enable us to monitor whether your bathroom leaks, you find that people will improve the quality of their work, on average. So there are all kinds of other benefits you wouldn’t necessarily anticipate. You can certainly get those in a factory setting as well. The more things that are monitored, the more maintenance tasks that are easy to check, just like you would like in quality control on the product itself, which is already well established; now you get quality control on all these kinds of repair tasks as well.

MB: Never forget: Big Brother is watching.

MD: That is the downside. [laughs]

MB: What role will AI play, and do you think it will vary by region, perhaps in part due to governmental constraints on what algorithms are going to be permitted to do?

MD: I’m not an AI expert, but I think you can get it quite a long way by relatively simple data processing of the kinds that I’ve described by checking the parameters within thresholds and things like this. When it comes to doing more sophisticated things, you’re looking at object recognition and that kind of thing. I don’t think that there will be, because you’re monitoring an industrial process with employees who are all consenting to be there. I think you’re going to get a lot fewer issues than you do in more of a consumer context. It’s not quite the same as monitoring people and leading them around to see if they do something. Everyone there is an employee. You’re not really monitoring them; you’re monitoring the equipment. How that evolves and how automation works its way through the workforce is a much bigger topic than just talking about factories of the future, and how that progresses politically is an open-ended question.

I think in most cases this is technology that can facilitate reshoring because it will enable you to manufacture more efficiently. Given that there’s kind of a political tailwind across the West and Japan to bring high-value advanced manufacturing back to the West, these are the kind of technologies that can facilitate that.

If you’re a labor union, I wouldn’t say this is a threat to our workers’ jobs. I would see it as an opportunity to have maybe more jobs, or more manufacturing, because people can do higher-value tasks, and more of the manufacturing currently taking place elsewhere can take place close to home. And we get the environmental benefits of less logistics. I think the pros outweigh the cons. I think data protection issues are less significant than in a consumer setting.

But I don’t think it will be this Big Bang moment. I think it will happen gradually; the equipment will become cheaper; it will become more established; people will become more comfortable with data processing and data dashboards, and how they can use this to improve the output and have less downtime, and I think it will just kind of gradually become the norm, like mechanization did one hundred years ago. It will become standard.

When people think of the digital factory, they picture a lot of robots and cobots … maybe an Amazon factory with a lot of stock pickers. That’s already started, and that’s running in parallel to the things I’ve described, which are smaller scale. But what I’ve been describing is also monitoring the health of the cobot – for instance, are any bits wearing out too early? It’s about monitoring as much as you can to either stop problems from happening in the first place or identify changes before they become problems. •

Au: For a copy of the paper, visit

Mike Buetow is president of PCEA (;

PCB West: The leading technical conference and exhibition for electronics engineers. Coming Oct. 4-7 to the Santa Clara (CA) Convention Center.

Submit to FacebookSubmit to Google PlusSubmit to TwitterSubmit to LinkedInPrint Article
Don't have an account yet? Register Now!

Sign in to your account