Edge Computing Explained: Solving Production Problems with IIoT
Isaac Maw posted on March 07, 2019 |

Edge devices are an essential technology for connecting industrial and manufacturing assets to the IoT, especially for older equipment without connectivity features built-in. Aspen Technology, also called AspenTech, is an IoT solutions provider with plenty of experience putting edge devices into place. Engineering.com recently reached out to Keith Flynn, senior director of product management at AspenTech, to explain the role of the edge within industrial IoT and what the future holds for these devices and the technology in general.

Keith, what’s standing in the way of widespread adoption of industrial internet of things (IIoT) implementations?

The biggest factors, in my mind, are these:  First, a fundamental understanding of what IIoT is. A lot of people see the words IoT and they're thinking cloud strategy and MES. I've talked to customers personally who say they have an IoT initiative, and they're looking to move 40 thousand tags from their infrastructure into the cloud. In my mind, that's not IoT.

Some of the biggest challenges today have to do with understanding what IoT truly is. For example, the fact that it was originally conceptualized from sensors, and it's not about sending everything to the cloud because cost can become a quick barrier when they try to go ‘big bang’.

There is a huge difference between a cloud strategy and an IoT strategy. When you're talking about IoT, you're talking about instrumenting assets and leveraging sensors to gain new information, and then using that new information to leverage the compute power of the cloud and all of the applications. A lot of people come at it and think they need to move their entire historical data content into the cloud in order to achieve something. So, providing guidance to the customers is an important part of the IoT business.

For example, can an edge device handle 50 thousand tags a second? Yes, but you'll need several edge devices to do that. You can't just expect a small edge appliance to handle some of the volumes customers are asking for. This is important knowledge that helps us match the technology, the software, and the hardware to the customer’s use case. But in some situations, we have to help customers adjust their expectations, and will tell them that their use case isn't really an IoT solution, their use case is really an MES solution, which is really more of a cloud strategy than it is IoT.

Now, according to your company, assets that are not connected currently make up about 40% of all industrial assets in the field. Edge devices are used to connect those stranded machines. However, on new machines, connectivity features are becoming more and more ubiquitous. Would you say edge computing in this industrial context is a transitional technology, or is it really here to stay?

I honestly think it's here to stay. When you say, ‘transitional,’ you kind of hit the nail on the head. I'll give an example from a company who'll remain unnamed:

We were working with a company who have about 22 thousand existing assets in the field, and they produce about 500 new assets a year. Their intention is to digitize their service offering as an OEM from a maintenance perspective. Whatever strategy they come up with has to offer the same information from these 22 thousand assets in the field, which may range in age of over 20 years to one year old. Edge is really the only way to tackle that with a uniform strategy. It may not be the same exact edge solution for each, because the older you get the more additional sensors you may need Meanwhile the newer devices that go out there may have better capabilities from an instrumentation perspective. Edge becomes a very important tool because you have to be very strategic in terms of which edge device goes with which legacy machine, and you must consider what the machine provides natively.

There are edge appliances that can connect via cellular networks, wireless networks, or hardwired connections. This combination of networking and compute power is essential to move forward, because it's the only way the industry can solve some of these problems. Ten or twenty years ago, the cost of even thinking about doing something like this was so prohibitive, you just kind of ‘rolled with it’ and went with the old clipboard. People were emailing in data numbers, or manually checking them. Now we can tie all that together under one umbrella.

For companies that have looked at expensive, prohibitive solutions such rip and replacing equipment who are now considering the opportunities of edge devices, can you give a starting point or a few tips for taking this route? What information do you need to get started?

There is a core set of information: Things like run status, running hours, product count, vibration. If that core information is not fully available, in some cases you might be able to pull a run status off an auxiliary contact into a PLC that's sitting close by, but that's just providing enough information to indicate status.

If you would've asked me this question even one year ago, it was more complex. A year ago, even adding a wireless vibration sensor to a bearing housing required local power. Twelve months later, we're now installing magnetic, completely wireless battery powered vibration sensors.  You just find a spot, mount it, make sure the battery's okay, and you walk away. You don't need to pull a cable. You don't need to look for cable trays or conduit drops, or anything like that, or even local power.

It's evolving so quickly that year on year we're starting to see the sensor technology catching up now to the gateway technology. This idea of ripping and replacing or dropping things in is going away, and it's really now about finding out about which core base information is important. Maybe there's a PLC on it but it doesn't have vibration. Or maybe it doesn't have the rotation or zero speed switch or belt break indicator. You can now go in with edge technology and augment what's already there or implement something completely new. It's becoming so cost effective, and the sensors are catching up the edge devices.

From AspenTech's perspective, we try to make sure that we can accommodate the rip and replace scenarios by embedding the proper software. Our Edge Connect software that we imbed actually supports new protocols for wireless sensing. It supports older protocols, or PLC, and we'll call them metered communications. So we do the mod bus. We do ethernet/IP, but we also support the newer OPC-UA, and then we support things like MQTT, in one instance. It becomes better and better as time evolves.

You’ve mentioned large rotational assets. What about machine tools, industrial robots, things like stamping presses or automation and conveyors—things on the manufacturing side. Doable?

Absolutely doable. With Edge Connect, we actually implement protocols from manufacturing, such as MT-connect, for instance, a protocol ‘spoken’ by CNC equipment. A piece of CNC equipment might be giving us spindle speeds, or what process stage it's at, or product counts. We can also get things like rejects, accepted parts, cycle times, so that we can look at not just production type numbers, we can also start calculating things like output. We can start looking at the differences in cycle times that are indicators to a problem. So: yes, when it comes to discrete manufacturing, we have the protocols to support that as well.

How long before all manufacturing operations are full of connected assets and the idea of a stranded asset off the network is a thing of the past?

Well, as I said above, I've seen a change in a year since joining Aspen Tech. It was 2014 when we took on the concept of IIoT, and here we are just over four years later focusing more on sensors. Five years out, in the 2025 time frame, I think the industry will be focusing more on what we do with the data, rather than on the problem of connecting assets. And it's not just the edge devices, either. The telecoms and the service providers are catching up. If you were to pick a convergence point where all those stars align, I'm thinking 2025 and onwards you're going to be focused more on what to do with it, instead of how to connect it.

For a prediction, this is quite near-term. You're talking about things like AI and AI-enhanced analytics in the future as well, right?

I am. At 2025, we’re going to have more bandwidth on the cellular providers using 5G and IIoT purpose networks. We’ll need gateway devices, from a technology perspective, with the appropriate GPUs. We're seeing GPUs now appearing in devices that have been typically ARM-based, allowing us to do more AI on the edge. We’re also starting to see the sensors catch up. Within the next half a decade we will basically be looking at full edge computing with the appropriate hardware and network in place to achieve that.

However, again, there's a learning curve associated with the two. You're starting to see a shift in the service field, as well. These big service providers and system integrators need some time to absorb that learning curve where IT and OT are part of the same system integration service offering. From 2025 onwards, I think we will be focusing more on the problems to solve, with the connectivity part behind us.

For more information on using IIoT in your factory, check out The Connected Factory and More: 5 Examples of How IIoT is Changing Manufacturing

Recommended For You