Edge computing use cases, success stories

While “the edge” is often discussed as if it is a single point in an architecture, in a real system there are actually many cascaded levels of edge devices lying between the physical world and the enterprise.

“It is exciting that we are now seeing innovation in each of these different levels,” said Jim ten Broeke, Business Development Manager IIoT Europe at Advantech.

Broeke discussed the increasing trend towards the use of edge servers, where enterprise cloud functionality is transparently and seamlessly brought on-premise, reducing communications overhead, accelerating responsiveness and increasing resilience.

“This is happening with both content delivery server applications and in the move towards high-level application edge servers, running at area, building or even individual process level,” he said.

At the communications edge, there is much more awareness today of the threats posed by cyber-crime. “While we are still getting a lot of enquiries for traditional edge gateway functionality, such as protocol conversion, communications media translation, data aggregation and event detection, the market is now equally concerned with the security features of the edge devices,” he said. “How they protect against unauthorized access by individuals, or prevent unauthorized or hacked code being installed and run. At the same time, users are realizing that remote management of these devices is critical to enable the fastest possible response to security patch rollouts across an installed base, as well as offering cost of ownership benefits by reducing truck rolls.”

Broeke believes that the most exciting area is at the lower edge levels, closest to the physical assets and operations. Here, the big trend is towards embedding artificial intelligence (AI) and machine learning (ML) within edge devices. Until recently, AI implementations relied on expensive, high bandwidth computers, and teams of specialist data scientists to create and refine the data models needed. Recent advances have brought the price of implementation down, and corresponding advances in ML, and the availability of pre-trained models for many common applications mean implementation times are often measured in days or weeks, rather than months or years.

“AI and ML embedded in the edge uses connected cameras to automate optical inspection, providing faster and more accurate detection of non-conformance as well as addressing the problem of an aging workforce. On production lines, AI can optimize efficiency across multiple machines, analyzing yields across different operating scenarios. Again, using connected cameras, edge AI provides worker protection through intelligent, 24/7 monitoring of safety zones around moving machinery,” Broeke said.

He pointed out the explosion in the adoption of edge computing at all levels within a system architecture means that a diverse range of hardware and software solutions are required.

“The perception remains that AI and ML require large amounts of processing power, but for some applications, we have implemented them in something as small as a cellular router. As in other areas of edge computing, there is no one size fits all solutions, even in the same application the topology of the installation may determine if it’s better to fit several small distributed edge devices, or to bring signals and data into a larger, more centralized unit.”

Patterns of edge computing adoption

According to Hermann Berg, Head of Industrial Automation Segment at Moxa Europe, when it comes to the Industrial Internet of Things (IIoT) and the digitalization of the industrial edge, most customers follow a similar pattern of adoption. He said: “Early projects focus on creating transparency — how to connect important assets and get relevant data out of automation silos and make it available wherever it is needed. Later stages focus on prediction, and adaptability — what does the data tell us and what actions can we derive from it? And finally, it is about new services and business models — what additional services and revenue streams can be generated with more data?”

Some industry segments have moved through these stages earlier than others. Renewable energy companies with their wind parks and solar farms, for example, responded to the need for transparency and adaptability for their unmanned, often remote sites with dedicated software packages developed in-house or developed by specialised software companies many years ago. Other sectors, with more centralized sites or super-distributed sites — such as low voltage power grid operators with their thousands of transformers — have been slower. Either the need for transparency and adaptability has not been as strong in areas where personnel are available — like in a factory or power plant — or putting IIoT gateways or edge computers next to each asset like an inexpensive low voltage transformer does not justify the investment commercially.

“So, in a number of specific and contained use cases, edge technology has taken off,” said Berg. Those use cases typically revolve around assets that are distributed, that are relatively high cost or where specific actions can be derived easily from data that either offers cost savings or increases revenue.

“Traditionally, the focus of early adopters has been to reap low hanging fruits with little attention to openness, extendibility and integration with other systems and often with no thought of managing, maintaining, patching, and upgrading devices in the field,” Berg said. “With more readily available services and software packages from cloud vendors like Microsoft, AWS and their ecosystem partners, many followers are now shifting their efforts towards building smarter, easier to manage solutions that make use of hardware, edge software and cloud services that integrate smoothly into the existing automation and control networks and systems.”

Berg believes there are some key challenges that need to be resolved. “The obvious challenge is about creating a secure and high-quality data connection from industrial asset to the cloud. This often includes physically connecting legacy equipment and transforming and cleansing industrial data. Edge devices should be easy to manage, industrial-grade hardware with a long product life that covers security patches for years to come.

“The less obvious challenge is about integrating those edge devices into an industrial network on site that has not been designed to serve such purposes. Smart automation engineers and managers will find synergies between the rising need to address cyber security requirements and the need to support more secure data paths, as they upgrade their industrial networks,” concluded Berg.

New application areas for edge computing

“Edge solutions are already being deployed in a broad range of industries and applications – such as within the oil and gas industry to collect data relating to the operation of wellheads and similarly in the water industry to collect data from remote pumping stations,” said Steve Ward, director, application engineering EMEA at Emerson. “These solutions are helping provide actionable information about remote or distributed assets, which can be used to optimize performance, reduce maintenance and eliminate unplanned downtime.”

Ward identified an expanding application for edge technology as being for monitoring and control of distributed renewable power generation, including solar farms, wind turbines and microgrids. “To operate microgrids efficiently, local intelligence is required, along with knowledge of upcoming weather and electricity prices to help decide whether to run the various power sources available, use power from the grid or to export power to the grid,” he said. “An edge controller can provide this intelligence by combining high-speed industrial control with internet connectivity.”

In addition, edge technology is also finding applications in energy management and monitoring of other industrial functions.  Ward pointed out that compressed air systems are now adopting edge solutions to monitor pneumatic air pressure and cylinder efficiency and wear. “This is helping to identify leaks and support predictive maintenance strategies that contribute to reduced energy consumption and greater machinery availability. The information provided can also support continuous improvement projects.

“Edge technology is also being used to manage fleets of assets, such as gas turbines and generator sets, where the equipment owner – which may not necessarily be the user – would like to monitor their status, and possibly also performance and condition. Generator sets are mobile units that typically are rented individually or in groups on a temporary basis. Edge technology can enable the owner to know where the unit is currently located. The functionality provided by edge solutions is not just helping typical industrial applications. Emerson has provided solutions to support the management and tracking of agricultural equipment, and also engine management and ballast control within the marine industry.”

Ward sees edge technology evolving and being implemented in several ways – Edge of the network gateways allow difficult to connect facilities and equipment to be connected and integrated into the corporate network, which allows asset usage and performance data to be collected in real time to support digital transformation and continuous improvement efforts. From this type of assets, centralised reporting can be used to inform users of operational parameters, but for exceptional events, edge devices can send an SMS message or e-mail direct to a user or group of users to enable immediate action.

“Edge solutions provide users with local visualization, supported not only by built-in screens with HMI functionality, but also by allowing local connection to a phone or tablet,” Ward said.  “The ability to connect a portable device via a local Wi-Fi network allows for better security and a more intuitive user interface compared to traditional operator interfaces.

“Local analytics can also be offered by edge solutions, enabling local real time control utilizing more data than previous control strategies. This can include optimization, based on past operation, as well as using external data. This will require new business roles, such as data engineers able to understand what affects machine or process operation and how to implement. ML and AI are developing quickly and we should expect to see ML/AI models being implemented at the edge to optimize performance very soon.”

Edge computing fundamental to manufacturing operations

Smart manufacturing relies on industrial edge sensors and actuators to increase production efficiency, manage production performance and ensure high quality output. “As production environments become more intelligent, they require the use of more sensors and actuators and more types of sensing modalities,” said Fiona Treacy, senior director industrial automation at Analog Devices.

These are fundamental to factory operation, as they measure critical physical parameters and provide outputs to machines, robots, PLC’s or other control devices. New contactless magnetic sensors are enabling the position of a robot arm to be known even in unexpected power down situations, removing the need for multiple cycles of calibration on power up,” she said. “Vision sensors are being deployed on cobots to enable humans and robots to work safely together and this vision sensing technology, combined with sophisticated algorithms, enables accurate depth sensing for more efficient operations and safer factory floors.”

Process and analytical sensors are being used to measure pressure, temperate, humidity, PH, flow and level. “These devices are all now being connected with Ethernet or IO-Link technology back to central control systems to enable real time decision making and adjustment of production parameters,” Treacy said. “With seamless Ethernet connectivity these devices can now be interrogated in real time, configured and reconfigured to optimize performance.” This, combined with localized intelligence, enables faster production and optimization of raw material management. It also means that factory assets can now be controlled remotely and it enables real time management of production flows.

Condition monitoring of the vibration signature on motors, combined with local AI/ML, is being used to detect anomalies in performance and signal the need for maintenance in a bid to eliminate unplanned downtime. Other examples include wirelessly connecting tooling to robot arms to enable robust tool interchange allowing manufacturing lines to be reconfigured quickly.

Power quality monitoring is also benefiting from edge intelligence. “By monitoring the power quality of devices on the factory floor we can now detect in real time the effect of harmonics on the power sources from other machines operating within the factory. Power supply disturbances and harmonics can adversely affect the precision of the factory control system and create maintenance issues,” continued Treacy.

“Finally, one key point worth mentioning is that, as we have an explosion of data at the edge, it is no longer efficient or desirable to communicate all that information to central control. Systems need to process that data locally, at the edge, and provide the insights to the control system. This requires the deployment of edge AI/ML which offers a much more power efficient architecture than mining all of the data centrally. Hence all devices on the factory floor can become connected and intelligent.” However, Treacy leaves us with a last, open question — how long will retrofitting take?

– This originally appeared on Control Engineering Europe’s website. Edited by Chris Vavra, web content manager, Control Engineering, CFE Media and Technology, cvavra@cfemedia.com.

Do you have experience and expertise with the topics mentioned in this content? You should consider contributing to our CFE Media editorial team and getting the recognition you and your company deserve. Click here to start this process.