Industry 4.0: Advantages of edge computing

Experiencing the best of both worlds—centralized and decentralized—is the idea behind edge computing, which allows data processing at or near the source of data generation. Proponents say edge computing can make industrial data handling and analysis more efficient, secure and cost-effective than a completely centralized approach.

In conventional industrial internet of things systems, plant data is collected and sent for processing to the cloud, a data center containing a group of servers connected to the internet. This centralized data handling system gives users a global view of all connected equipment, which may be in a number of different locations. The system also allows users to quickly and easily update software in far-flung machines, said Michal Skubacz, head of industry software for machine tool systems at Munich-based Siemens AG, which develops technology for industrial edge computing.

Cloud Downsides

However, “if you push everything to the cloud, you are dependent on network connectivity up to a cloud system,” Skubacz said. If the network breaks down, so do critical cloud-based production applications.

In edge computing systems, data processing is performed chiefly on the periphery rather than in the cloud, where Siemens’ MindSphere internet of things operating system functions. Image courtesy of Siemens

Cloud-based data management also introduces latency into industrial applications, he noted. Because it takes time for data to travel back and forth between the cloud and the plant floor, applications that require real-time responses cannot work properly.

Then there’s the issue of data load. Consider a shop floor with 50 CNC machines, each needing to be monitored during the entire production process. During production, information on machine movements could be collected hundreds of times per second by a large number of sensors. “The volume of data from those 50 machines that needs to be transferred would be enormous,” Skubacz said. “You would need networks with throughput that you probably don’t want to pay for.”

In addition, he said, privacy and security concerns may make firms reluctant to push secret and essential data out of their shops, where it could be more vulnerable to theft.

Moving to the Edge

These factors support the move toward edge computing, in which plant equipment analyzes most of its own data in real time. In many cases, “it makes a lot more sense to do the processing and analytics locally than sending it all up to the cloud,” said John Crupi, vice president of edge analytics at Greenwave Systems Inc., Irvine, California, a provider of edge analytics technology.

According to Skubacz, some companies think edge computing simply involves connecting an industry PC to a machine, but there’s more to it than that. “I understand edge computing to be an integrated system allowing us to run different applications very close to production. At the same time, it’s connected to some sort of system normally in the cloud that enables management of applications, remote updates” and other functions initiated away from the plant floor.

The edge does what Crupi describes as the first stage of processing. “Say you had a pattern detector that triggers an alert if there’s a rapid rise in (equipment) temperature,” he said. “You can handle that locally by going into the control and changing some parameters or notifying some other system on the plant floor.”

Axon Predict is an analytics platform that can be used in edge computing systems. Image courtesy of Greenwave Systems

The cloud, meanwhile, handles the second stage of processing, which Crupi calls aggregate analytics. For example, he said, “If you want to compare the average temperature of every red machine in a plant to red machines located everywhere in the world, each red machine at the edge computes its average temperature, sends that result to the cloud, and then the cloud sends down the average of all the machines to the edge” so the comparison can be made.

The cloud can also serve as a repository for historical data. For instance, Crupi said, the cloud can store several weeks of data that will allow calculations to determine whether machine speeds are within certain limits at a specified temperature.