Gartner defines edge computing as a part of a distributed computing topology in which information processing is located close to the edge – where things and people produce or consume that information. For manufacturing, which deals in huge amounts of data that ability to collect, analyse and act upon data close to the process can be crucial in improving performance.

At its core, edge computing is a distributed computing framework that brings enterprise applications closer to data sources such as IoT devices or local edge servers. This proximity to data at its source can deliver substantial business benefits: faster insights, improved response times and better bandwidth availability.

Gartner estimates that by 2025, 75 per cent of data will be processed outside the traditional data centre or cloud. From connected vehicles to intelligent bots on the factory floor, the amount of data from devices being generated in our world is higher than ever before. Yet, most of this IoT data is not exploited or used at all. For example, a McKinsey & Company study found that an offshore oil rig generates data from 30,000 sensors — but less than one per cent of that data is currently used to make decisions.

There are generally recognised to be four sectors driving the adoption of industrial IoT and edge computing in manufacturing, power generation, oil and gas, and utilities. Edge computing harnesses growing in-device computing capability to provide deep insights and predictive analysis in near real time. This increased analytics capability in edge devices can power innovation to improve quality and enhance value. It also raises important strategic questions: How do you manage the deployment of workloads that perform these types of actions in the presence of increased compute capacity? How can you use the embedded intelligence in devices to influence operational processes for employees, customers, and business more responsively? To extract the most value from all those devices, significant computation volumes must move to the edge.

“Edge has many definitions, and it depends on people’s perspective and their use case of it,” Sahil Yadav, product manager – Predix Edge, GE Digital, says. “However, the way we think about edge at GE Digital is anything that is close to the asset or is within the asset network and not located either in the cloud or in a separate location. Edge could be a device connected to five of your assets on the factory floor, or it could be an embedded device within a robotic machine that sits inside the machine itself. The definition can vary, but how we define edge is in terms of its capability and application. For us, what edge means is its ability to process data close to the assets then provide results in near real time.”

Three scenarios for edge

For manufacturers, there is a myriad of use cases. One of the most prevalent ones is on a factory floor, where you have an assembly line or assets dependent on other assets that you need to make decisions in real time. What that means is the output of the previous machine is the input to the next machine, and the wellbeing and functioning of the previous machine is essential for the functioning of the next machine.

“In this situation, you need to be able to make decisions in real time about the previous machine and if it is about to break,” Yadav adds. “Obviously, you have control system software in there as well that has been used in the past, but how edge computing helps our customers is through the ability of it being flexible enough to work with all kinds of machines. This also includes the ability to work with legacy machines and still be able to provide those modern drivers and controls. You have the same technology controlling your new machines, as well as your old machines.

“To allow new devices to work with legacy devices, edge is one of the only solutions out there. It gives you the flexibility to talk to old protocols that you do not support in the new infrastructure you are using. With edge, you get the compute space, and you can run protocol translators. This transfers or translates data from port protocols and brings them up to the same level, such as some of the modern formats such as MQTT and OPCUA, so that you can visualise and use this data in a single pane of glass.”

Yadav explains that GE Digital thinks of edge in three main terms or three prominent use cases. “There is the intelligent data pump, which is connecting to the asset and doing protocol translation, to then being able to consolidate that data and send it to the cloud or forward to a central repository,” he says. “Most of our customers are worried about the data bandwidth that is required to support the data that has been sent over because that is driving cost for customers. So, running some of the more straightforward applications such as filtering or aggregation is processing the data so that you do not have to send the raw data forward. You are sending a result ahead for further analysis, which is the intelligent data pump use case, and there is minimal computing power.

“This may align with a definition some vendors use for edge, but the next level for us is the edge computing use case where it is not just about sending data to the cloud, but also analysing the data at the edge itself. This could mean running custom applications, and these are dependent on the type of asset that is being supported. You can run those custom applications and apply machine learning analytics on all sorts of things depending on the amount of computing you have at the edge and perform that analysis.

“The result of the second use case is much more usable directly than the first use case. Manufacturing customers see intelligent edge control as an important use case as they build their solutions, which essentially means that they have done all the computing at the edge after they have analysed the data at the edge. The insights you get out of that data can be utilised by edge itself to send control signals back to the machine. This is a use case that nonlinear customers are asking of us today.”

Dealing with the data flood

Companies are grappling with the enormous challenges that data and automation present in today’s modern manufacturing environment. “Organisations know that the faster they can act to improve processes and lower costs, the more competitive they can be,” Przemek Tomczak, SVP, IoT and utilities, KX says. “The question is how to manage the ever-increasing volumes, variety and velocity of data being created, much of it from connected devices at the edge.”

And the numbers are enormous. A modern connected factory will create millions of data points a second from sensors in machinery. A recent report by Statista estimates data volumes from IoT devices alone will hit 79.4 zettabytes by 2025.

“While edge computing – which the Industrial Internet Consortium defines as having the capability to conduct computing at the source of where it needs to take place – is undoubtedly revolutionising the industry, it is also asking questions of legacy technologies and systems that are struggling to cope with the data deluge,” Tomczak adds.

Indeed, the case for edge computing has never been stronger. In a sensitive industry such as manufacturing, where even the smallest degrees of latency can disrupt machine lines, reduce quality and output, and cost millions in lost production time, being able to act on insights from the data in the shortest possible time frame is vital.

Cloud computing has also proved to be a transformational technology in manufacturing, providing significantly cheaper storage and processing capabilities while also giving rise to a vast marketplace of software and services and that manufacturers can further enhance their processes and productivity. “However, concerns around security, reliability and latency remain; a data management architecture which contains a hybrid of edge computing and centralised data management, either on-premise or the cloud, is therefore gaining prominence as a preferred manufacturing use case,” Tomczak explains. “Together they bridge the worlds of big data and fast data; allowing manufacturers to manage the massive volumes of data being created while providing a means to bring historical and real-time data together for deeper levels of insight and understanding.”

Automation driving adoption

Deeper understanding is critical to implementing what many see as the ultimate aim of edge computing, machine-to-machine communication, and autonomous computing to achieve a state of self-awareness and ultimately create self-learning machines to better control manufacturing processes, increased uptime, yield and useful life of assets. “Streaming analytics is the technology that enables automation, allowing manufacturers to collect and analyse data in real time at the edge of their network as well as at the data centre while comparing it to historical records and context,” Tomczak says. “This process also takes advantage of the three critical needs of data management. The ability to ingest the enormous amounts of data, the ability to analyse it in a low latency environment and the ability to utilise machine learning to link or correlate multiple new and historical data sets together to generate valuable actional insight in real time.

“Together the rise of edge computing coupled with a streaming analytics software solution is enabling microsecond decision making that ensures high-tech manufacturing organisations can keep their equipment and product lines constantly monitored and running efficiently. As we head towards Industry 4.0 and an ever-increasing number of connected devices within the manufacturing sphere, those that do not have the right data processes in place will be unable to continue to improve productivity and efficiency through better use of analytics.”

It all begins with the network

Cloud, edge technologies and the proliferation of IoT devices represent a clear way forward to enable the concept of smart factories in manufacturing. Smart factories rely upon autonomous devices to produce and interpret a constant stream of data from connected operations and production systems.

“The devices that comprise this system learn and adapt to new demands, communicating both via the network edge and into the cloud,” Simon Pamplin, technical director at Silver Peak, says. “Underlining the significance of this transition, the changes this technological evolution brings are thought to be so dramatic that it has been heralded as Industry 4.0, or the ‘Fourth Industrial Revolution.’

“If manufacturing organisations are to fully realise the benefits of this cloud and edge revolution, they will need to have a wide area network (WAN) capable of securing and delivering the huge data volumes traversing the network edge. As factories become ‘smarter,’ thousands of connected devices will generate and send data, placing huge demands on the network. Simply put, a secure and reliable WAN is the foundation upon which edge devices can communicate at optimal speeds – it, therefore, determines the success or failure of Industry 4.0 IoT efforts. This being the case, any transformation plans need to address the WAN first and foremost.”

Manufacturing organisations that leap into digital transformation before rearchitecting the underlying network will find that the expectations they have formed of smart factories may be hampered. This is because traditional, data centre-centric private networks were built to support a bygone era when applications were centrally hosted – the massive and complex network demands of digital transformation and the cloud was never envisioned.

“From a manufacturing perspective, as IoT devices proliferate and flood the network with data via the edge, it will result in latency issues on the traditional WAN, resulting in potential delays across the production line,” Pamplin continues. “Therefore, any move towards digital transformation before the network has been adequately architected may well introduce new challenges and additional costs for an organisation.

“This being the case, Industry 4.0 is an excellent use case for advanced networking solutions, such as software-defined WAN (SD-WAN), which can intelligently and securely segment, secure and prioritise application and device traffic by business criticality. This is essential for eliminating the latency issues common with legacy WANs and ensuring operations are kept at top speed. Furthermore, SD-WAN can utilise 5G transport, which can expand the number of IoT devices that can be reliably connected to the network, enabling a virtual network of ultra-high-speed connections across multiple devices.”

These attributes make new networking solutions – and advanced SD-WANs in particular – catalysts in accelerating manufacturing digital transformation. After all, while expectations are soaring, technology can only perform within the limitations of the network.

The legacy conundrum

There is no doubt that the growth in industrial edge applications presents new opportunities for many manufacturers. Enterprises are moving more of their applications and compute capabilities to edge devices to improve latency, reduce bandwidth usage, enhance security, and reduce disruption. But, as Charlene Marini – chief product and marketing officer, Pelion cautions, in a rush to add slick new IoT technology, it is easy to overlook the legacy operational technology (OT) that has been quietly gathering that all-important sensor and actuator data. “Manufacturers undergoing digital transformation are turning to on-site edge computing solutions that enable a multitude of use-case applications, interface with various systems, monitor for suspicious or anomalous events, and maintain operating software across a range of downstream system components,” Marini explains. “Edge gateways are becoming the focal point for operators who expect more from their gateways than just the traditional OT/IoT device management role and demand significantly more functionality.

“They can act as the point of unification for a range of new and old devices on-site, including device management, integration, protocol translation, and managing new and existing applications. Importantly, they also enable the evolution of system functionality over time. These gateways utilise Linux-based containers to securely host multiple apps, gateway devices configured as single nodes or clusters. This range of services form the backbone of a CIM (Computer Integrated Manufacturing) framework.

“Any serious offering needs to combine extensive downstream device lifecycle management, comprehensive gateway system management, and, increasingly, the all-important secure applications enablement capability.”

The open-source nature of gateways means that this is just the beginning of what is possible. The introduction of a secure on-site applications ecosystem will drive innovation for third-party apps, including the potential for smartphone-like app stores, enabling new use–cases, business models, and, ultimately, more value. “A vibrant marketplace for third-party innovation empowers owners and operators to vastly expand the scope of their offerings, unshackled by the limits of existing in-house application development resources,” Marini concludes. “The prospect of integrated, federated apps that abstract complexity will allow the manufacturer to combine rich data insights from a range of new and legacy devices while optimising network usage and compute response times.”

Moving to the intelligent edge

The intelligent edge is an essential step to some autonomous operations because the decisions are still made away from the edge. They are made out at some central location or by somebody else. When you get into intelligent edge control, there is no-one in this operation.

What is needed for companies to implement that sort of strategy? Is it understanding, is it trust, or is there still some technology missing? “There are a couple of factors, and the most important is security,” Yadav adds. “The reason why some manufacturing companies are not moving to cloud is because of security reasons, and edge solves that problem, which is a benefit for edge. However, even when your assets and network are secure, putting a new device into the network itself raises security concerns. There is that trust factor that impacts their ability to utilise edge as rapidly.

“The second reason is the cost. Today, because the technology is new, the cost of implementing edge is growing out of proportion as companies are scaling. That is creating challenges for companies to sustainably use edge technology or even adopt edge technology to some extent. So, trust and cost are two main reasons why the customers will not move as quickly.”

Edge will not be required everywhere, but over time it will become more pervasive throughout the sector. “For example, self-operating assets such as an aeroplane or on a submarine would probably use edge technology to drive some of those intelligent edge controls,” Yadav concludes. “For assets that are not time-dependent or can survive for a longer time without needing real-time attention, intelligent edge control may not be the path forward. For those assets, the major use case is essentially fleet management, which is collecting data from hundreds of these assets and being able to analyse that data and visualise it, then making decisions based on the completed information that you have.”

Monitoring drives the growth of the industrial edge 

Manufacturers currently face a backdrop of heightened tensions brought on by long-term global competition and the more recent COVID-19 pandemic. As such, manufacturers are under increased pressure to maximise their productivity and output. Covid-19 has acted as a wake-up call to many manufacturers, who had typically been late adopters of information technology.

“Lockdowns and resulting restrictions on workforce mobility have pushed manufacturers to assess the value proposition of industrial IoT solutions in a new light,” Galem Kayo, product manager at Canonical – the company behind Ubuntu, says. “The demand for production monitoring solutions, for instance, is growing because they enable remote work. Similarly, the need for solutions that automate labour-intensive tasks like quality control is also sought after. These circumstances result in an acceleration in the adoption of IoT at the edge.

“To focus on real-time monitoring, the restrictions put on people’s movement was pivotal in elevating the use of IoT, as manufacturers needed to track and monitor productivity at their plants without human involvement. IoT, encompassing sensors and cameras, enables outputs to be tracked, providing data-informed visibility on what is going on within factories and on production lines. Sensors directly on machines allow the analytics to inform how productive the plants are and whether their overall equipment effectiveness is up or down.

“The same applies to quality control, the traditional processes of which are dependent on human labour. The human eye cannot detect microscopic defects on small parts, and in large-scale manufacturing, human error is a significant risk. Using deep learning and computer vision to automate quality control is hugely advantageous, as it is possible to train algorithms to do this specific task – saving both money and time and boosting reliability in the process.”

A key advantage of IoT solutions is that they can be deployed in an agile manner. Manufacturers can start small, learn, and then scale up. They can deploy an IoT platform on one machine by adding sensors and then connect the data this generates to the cloud to expand the intelligence. This allows manufacturers to first measure the ROI before scaling, which is all easier with edge technology. The process can happen in weeks and at a cheaper cost than any legacy solution, which is a huge advantage for manufacturers.

Read more of our exciting features here!