This article is part of a series covering edge computing. Read the rest of the series now.
Before you can homestead in the wilderness, or live and work in any new place, you have to define its characteristics, trace its boundaries, understand its constraints—and stake out a claim. The same goes for edge computing. Luckily, there's plenty of advice.
"Some believe the edge is between operations technology (OT) that runs production and information technology (IT) that runs the cloud, but we say the edge is computing that sits on machines or manufacturing floors, and talks to distributed control systems (DCSs), programmable logic controllers (PLCs), instruments, sensors and relays," says Yuri Chamarelli, lead product marketing specialist for control systems and edge computing at Phoenix Contact. "In fleet management, there are devices that collect data and send it to the cloud, but where is the edge there? In the trucks, collectors, network or the cloud? It isn't there. So, how the edge is defined is conceptual, not physical. The edge is just an idea about how to collect and process, and send it securely to systems onsite or in the cloud."
Because edge computing typically involves turning sensors and field components into Internet-capable devices, Chamarelli reports Phoenix Contact's PLCNext controller can run Linux operating system and use other internal components similar to a Raspberry Pi board, and implement them in an industrially approved housing to perform PLC runtime functions. "Linux in an edge device like this can do data processing in the field, which is getting increasingly inexpensive, even as data processing in the cloud also gets cheaper. This means users can pick whichever industry-specific method is best for them and their application," explains Chamarelli. "The advantage of processing data on the manufacturing floor is it allows real-time monitoring and immediate feedback even if the overall network goes down. Plus, it's less costly if all data collected doesn't have to go to the cloud."
For example, pumps at an isolated water/wastewater plant without much communications infrastructure may only use a dial-up or 4G modem. This means it costs more to send more data, Chamarelli explains, so the plant may only send back feedback inputs when an action needs to be taken. "Even so, if you're running several hundred or a couple of thousand pumps, this can get very expensive, especially because many users also want temperature, vibration, current draws and other information. This is why some users are exploring fog computing, which is basically having cloud-computing on premises, and send raw data to it via their local Internet protocol (IP) infrastructure and servers. In addition, even though we don't have much big data analysis and performance algorithms in the field yet, it's going to be a bigger issue as users decide what and how much they want to analyze."
Shifting landscapes
Because edge computing is a relatively new field, its definitions and boundaries also don't remain static for very long. Shane Hale, global business development director for Pervasive Sensing at Emerson, reports the edge is the first location where data can be analyzed and manipulated close to the production equipment that generates it but still remain separate from controls and larger systems.
"Because edge technologies are evolving so fast, it's still a wild west environment, with users creating all kinds of different definitions for edge computing and edge analytics as they try to figure how to use them. "We say the edge is where we can do some analytics in the field or close to it for specific functions, combine multiple data sources, and process, prioritize and prepare data to be sent elsewhere for further analysis," explains Hale. "For example, an application may not need to send up all seven or eight pressure and temperature readings. Instead, we can do some simple analytics, where we first collate that data at the edge, and then send just one or two more meaningful data points to develop a useful insight."
Hale adds that Emerson's Plantweb Insight software platform can monitor and manage process equipment, such as pumps, heat exchangers, pressure relief valves (PRV), steam traps and cooling equipment, and can run on edge computing devices and apply preconfigured analytics to monitor specific assets. "PLCs and DCSs use collected data to control their processes, but they can also take in algorithms to adjust operations based on recommendations coming back from those analytics," says Hale. "For example, we recently did a PRV monitoring project running on edge devices in a refinery, and found that multiple valves were lifting off and flaring early because they weren't set correctly. So, by identifying the valves from insights calculated in their edge devices and resetting them properly, they were able to run more efficiently and closer to their correct pressure points."
Bridging OT and IT
Even though edge computing calculations can be performed anywhere, Nate Kay, PE, senior project engineer at Martin CSI, reports it defines the edge as the boundary between operations technology (OT) and information technology (IT) because it's where its edge devices are located, along with its firewalls and network address translation functions for establishing interfaces between OT and IT. Martin CSI is a system integrator in Plain City, Ohio, and a certified member of the Control System Integrators Association (CSIA).
"Edge devices straddle the OT side's processes, sensors, machines and PLCs, and bridge to the IT side's networks and business systems," he says. "We process data at the edge, such as turning revolutions-per-minute and other raw measurements into meaningful insights, and pumping them up to the cloud. We use a variety of hardware for this, including Phoenix Contact's mGuard, Red Lion's Data Station Plus, Opto 22's groov EPIC and Inductive Automation's Ignition Edge platform on a small PC. Each client decides what information must go to the cloud, but calculating and getting data into a consistent format and standardized database on the edge can save processing power and revenue because users aren't simply dumping all their raw data onto a cloud-based database before sorting it."
Because edge devices can employ MQTT or other standard protocols, handle different languages and translations, and process more data locally, Kay explains that Martin and its clients can advance beyond their initial OEE projects. "We can perform more complex calculations, use algorithms to look at data in different ways, and extract intelligence we never had before," says Kay. "For example, vibration data or current draws can be collected from all the lines and equipment in a plant, performance and failures can be correlated, and predictive models that look much further in advance can be developed more easily. We even have a IIoT interest group working on this now."