Practical Edge Computing for Manufacturing
5 min read
As the manufacturing sector gradually moves towards Industry 4.0, we’re seeing a few key trends that are starting to dominate today’s industrial landscape. While this transition is characterized by a diverse set of technologies that range from machine learning (ML) to augmented reality (AR), at its core we see the common themes of digitization, information, and communication.
In truth, most of these advances boil down to how machines transmit data to other machines and to us.
Edge computing is simply another side of this same coin. There’s recently been a lot of hype around this buzzword, and many marketing teams would like you to believe that it’s more complicated than it actually is. Edge computing enables plenty of sophisticated technologies, but the underlying concept is actually quite simple.
In this article, we’re going to strip away that hype to give you a hard-nosed account of edge computing that’s practical and value-forward.
What is Edge Computing and Why Does it Matter?
To compute “at the edge” is nothing more than to run code within close physical proximity to the data source.
We can think about a computer network like a tree. At the tree’s core is the trunk, and everything connects back to it; the same is true for the cloud or data center. At the tree’s edges, we have leaves and roots, which receive inputs, like water, sunlight, and carbon dioxide, and they produce outputs such as oxygen. Analogously, sensors and actuators are the I/O devices that sit at the edge of an industrial network.
Edge computing replaces the paradigm of sending all data to a PLC or central server for processing with local compute. By either embedding edge devices with microcontrollers or setting up nearby gateways, this amounts to crunching numbers closer to their source. Note, however, that edge and cloud are not exclusive; they exist on a continuum. An edge device, for instance, can enable real-time control on the factory floor while still sending periodic aggregations up to the cloud in order to generate macro insights.
But why move to the edge? Here are the top reasons:
lower latency
better cybersecurity
lower throughput decreases network load and costs
improved reliability
Latency is the most important advantage for industrial use cases. Controlling equipment in real-time isn’t possible when data needs to travel to and from a remote data center. Additionally, keeping this data on the local network improves the organization’s cybersecurity posture by decreasing the attack surface. Lastly, less reliance on the internet decreases load and means your factory can continue to function even during a network outage.
Because of these features, edge computing opens up new opportunities for maximizing overall equipment effectiveness (OEE) alongside reducing downtime and total cost of ownership (TCO). Now let’s look at some examples of the edge in action.
Top Use Cases for Industrial Edge Computing
The first use case we want to highlight is data processing and analytics. This can be as simple as filtering noise out of sensor data or as complex as deploying ML models at the edge for more advanced insights.
One type of analytic that’s currently seeing wide adoption is anomaly detection for predictive maintenance. Essentially, this involves running an algorithm that uses sensor data to determine if and when a machine is likely to break. Of course, more uptime directly translates to more revenue in industry, so the ROI of decreasing unplanned downtime is clear. Additionally, it reduces maintenance costs by optimizing equipment life-cycles and preventing critical failures.
At a higher level, edge computing is also integrating with SCADA to unlock insights into overall performance metrics and to facilitate real-time control. For example, Automation World lays out an architecture where the SCADA system is used “as a source of information for machine learning infrastructure”.
Edge computing can also bolster industrial automation by enabling greater levels of autonomy. For instance, this can enable flexible solutions for material transport, such as AGVs. When we combine the low latency of edge computing with its capabilities for real-time machine-to-machine (M2M) communication, it’s clear that edge computing is a foundational technology for advanced automation systems.
The final use case we want to highlight is protocol translation. Since the average factory in the US is 25 years old and filled with machinery that’s nearly a decade old, enabling M2M communication can be challenging simply because these machines don’t speak the same language. Many legacy machines, for instance, don’t support OPC UA and MQTT, two of today’s leading protocols.
An edge device can act as a translator, thus enabling older machines to each other and to newer ones. Furthermore, edge computing also brings its own suite of adaptive protocols, including HART, which sends and receives digital information across analog wires, as well as IO Link, a short-range protocol that serves as an intermediary between IO devices and a central programmable logic controller (PLC).
Conclusion
Although the future of manufacturing is moving towards the smart factory, and therefore the adoption of a whole range of advanced technologies, we don’t buy into the “all or nothing” maxim. You don’t have to go all the way to leverage edge computing.
There’s plenty of room for a tempered and phased approach to updating and upgrading your industrial automation processes. And that’s where Outlier’s expertise is so valuable: we can help you figure out what to buy, how to implement and integrate it into your existing workflows, and how to see a clear ROI. We’re with you every step of the way.
Find out more by getting in touch today.