It’s been more than a decade now since the terms edge and cloud computing became commonplace in the vernacular of manufacturing automation professionals. Since then, the hybrid computing environment—in which both edge devices and private/public cloud infrastructure are used to manage operations data—has become the norm across the manufacturing industries.
Now that industry’s gained a great deal of experience with edge computing, we wanted to know more about where manufacturers are benefitting the most from this technology. So, we connected with Steve Mustard, former executive board member and president of the International Society of Automation (ISA) and current president and CEO of National Automation.
Q: Time and costs are always two of the most prominent issues in manufacturing when it comes to implementing any new technology. So where does edge computing stand here? Is it difficult or costly to integrate edge computing with existing manufacturing systems?
Mustard: For many years now it's been very difficult for manufacturing organizations to implement the kind of processing capability they need because of the cost of the hardware and the software. One of the nice things about technologies like edge computing is that companies that produce this technology have been able to use the reduction in cost of silicon and other things to make products which are very cost effective. So, today, you can get a lot of edge processing for your dollar.
From a cost of the product point of view, it's not really a big challenge. I would say the biggest challenge from a cost and complexity point of view is how manufacturers choose to integrate edge computing with their legacy systems.
Q: Can you speak to the key advantages to processing data at the edge rather than in the cloud?
Mustard: One of the key advantages with edge computing is that it relieves the need for centralized computing, whether that’s done on premises or in the cloud because of the intense processing nature that was once required [to handle the amount of data generated in the manufacturing industries]. Now, with edge computing and its processing capabilities, we can deploy that kind of compute capability at the edge near the process itself and actually do a lot of that stuff that was formerly done in the cloud or on premises locally in a large data center.
So, why is that good? For a start, consider latency. Instead of having to capture some sensor information, send it somewhere for analysis, get the results and then make a decision based on those results and then send that information back to a central source, you can now do a lot of that at the edge.
This allows manufacturers to respond very rapidly, which is obviously very beneficial from an operations capability.
There are also benefits from a resiliency point of view, as it's really nice to be able to have all that processing capability local to the process. Because if you have some kind of disruption outside of the facility, you're not as likely to be impacted by that.
Then there are the benefits related to communication costs. Instead of having to send gigabytes of data to the cloud somewhere and then back again, edge allows you to minimize the amount of data transfer.
Q: The advantages of predictive maintenance versus reactive and even scheduled maintenance have been discussed for years now, but it's still not yet a common practice largely due to the cost of implementing related technologies as well as the associated procedural changes. Considering that, how do you see edge computing helping to advance predictive maintenance in manufacturing?
Mustard: Take rotating equipment vibration monitoring for instance. You can collect vibration data, but to analyze it you need a lot of processing capability to look at all of that information in real time. Now, with edge computing, the cloud, IIoT and IP infrastructure you can deploy vibration sensors on devices—even wirelessly—to collect all that data and do some processing locally. You can do more processing in the back end in the cloud, if necessary, and the cost of doing all this is much lower than it used to be.
This makes it easier for manufacturers to deploy predictive maintenance strategies to identify failures before they occur. They can look at trends and start to see the things that might go wrong before they go wrong.
Q: Beyond the vibration monitoring example you gave, what other types of data are typically collected by edge devices that would be most valuable for a predictive maintenance application?
Mustard: You can also measure noise that might be generated by machinery to help determine when things are going wrong. Then there are things like temperature and the condition of oil in machinery. There's a lot of sensing technology that's been available to do this, of course, but a lot of that analysis was done offline in labs and you wouldn’t hear back until weeks later. But when you can get that kind of information on site in real time or near real time, then you can start to really get some benefits even with simple things such as usage data around the amount of time certain pumps run or the flow rates through pumps. You can also start to detect potential blockages or cavitation on pumps, as well as reductions in equipment performance.
With these kinds of insights, not only are you identifying things that could fail, but you're also identifying areas where you're losing performance capability. So, it's not just about maintenance to prevent failures, it’s also about maintenance to maintain performance at the highest levels possible.
Q: I'm glad you brought up resiliency because that's another topic like predictive maintenance that's become a big issue for manufacturers. Can edge computing help with this?
Mustard: When you’re able to do a lot of data processing at the edge, if you think about the availability of your information network, you’re maximizing the capabilities of what's there. For instance, if there is a disruption in, say, the wide area network or in the cloud, you're much less likely to have that impact your operation. A good example of this is what we all recently saw with the CrowdStrike incident, which took out the command of infrastructure for many operations in one fell swoop. And if you localize your infrastructure, especially in an operations technology environment where you're protecting against automated patches, then the likelihood of something like that affecting your operation is going to be reduced.
And then, of course, there’s OEE (overall equipment effectiveness) which addresses quality and performance. This can be impacted positively with edge computing also.
Take, for example, image processing on production lines to look at the quality of a product as it’s being produced. That image processing, however, requires a lot of compute power. But now you can do that at the edge and you can apply artificial intelligence to look at the images to identify quality defects as they start to appear.
Q: Speaking about the CrowdStrike incident you mentioned, can edge computing be used to support operations continuity during network outages or other types of connectivity issues?
Mustard: Yes, if you can arrange the functions in your manufacturing processes such that you can organize it and localize the key processes that need to be localized,
then you can be more tolerant of disruptions. If you imagine a conventional kind of manufacturing execution system—all the way from the sensor level to the enterprise level where everything is dependent on everything else—if one level fails, then everything else fails. In that setup, a CrowdStrike-type incident is going to basically take out your entire manufacturing capability in one go.
But if you have an edge-enabled environment where your manufacturing zones can operate autonomously, even though they have interactions with other systems but aren’t required to, then if you have some kind of enterprise level disruption you can keep your critical operations running and restore everything in the background.
Q: This topic of operations resiliency really became a big issue in manufacturing following all the COVID-related supply chain disruptions. Can you explain what role edge computing plays in optimizing a manufacturer’s supply chain and production scheduling to mitigate those risks?
Mustard: COVID was that moment where I think we all realized that the supply chains for everything were working until they didn't work and then, when they didn't work, it was a disaster. But on the plus side, I think what we also learned from COVID is that if you look at, for instance, vaccine manufacture, we stood up manufacturing plants to make COVID vaccines in next to no time because people thought, well, we can't build them in the conventional way of where we go and build this facility and it takes 4 years to construct it all. So, we're going to have to build it in a modular way, and then we can test out the modules and then assemble it all on site to get everything up and running much more quickly than we would have done otherwise.
The lesson here is that manufacturers should be thinking about how they can be able to rapidly transition some manufacturing from one location to another in the event that something like that happens again so that you can keep things going. And the only way you're going be able to do that is if you've got a modularized architecture in your infrastructure. With that modular structure you can download the new recipes, reconfigure the machinery and be up and running in a few days or weeks instead of six months or years. I think edge computing plays a big part in being able to achieve that level of flexibility.
Q: Is that due to all the data housed in edge computing or are there other specific aspects about it that enable the kind of modular infrastructure flexibility you’re talking about?
Mustard: Yes, it allows you to properly modularize some of the functionality. Without that, you're stuck with conventional control system and enterprise system architectures where everything has to be located in certain places. Edge computing allows you to deploy quite compute intensive capabilities anywhere; therefore, it allows you to kind of reconfigure things much more easily.
Q: We can't discuss any connected computing technology without also discussing cybersecurity, which has become an increasingly important issue in manufacturing over the past few years as more groups are targeting specific devices on industrial networks. Can you explain how edge computing be used here to improve cybersecurity in manufacturing environments?
Mustard: Cybersecurity is one of my passions, and so I think the first thing about how edge computing can help here is through the ability to minimize your attack surface by reducing the number of ingress and egress points to your facility.
Minimizing the communications you need to communicate with the cloud or even across the enterprise itself is a great thing. If you can do processing locally, that's not only good from a resilience point of view, but also to protect key connection points from unauthorized users.
Q: Based on your experience in industry, what are the key challenges you typically see when it comes to manufacturers implementing edge computing? What are the typical obstacles they run into and how have you seen them be successfully overcome?
Mustard: Probably the most common thing I see people do wrong is that you can buy a lot of technology and implement new systems, but unless you change your processes and procedures as well as the culture around how you do things, you're not going to get the maximum benefit out of the technology.
For example, we talked about predictive maintenance, but unless you change how you do things, people are just going to continue to do scheduled maintenance as well as the reactive and proactive maintenance. So you have to train people in the fact that we're doing things differently now. Otherwise, you’re not going to get the maximum benefit.
