Maybe the greatest benefit of a new technology is its lack of historical baggage, thereby enabling industry professionals to approach it with an open mind and a clean slate. Edge computing currently enjoys this status in manufacturing. For how long, depends on how well manufacturers understand, embrace and successfully implement it on their assembly lines and throughout their plants.

In simple terms, edge computing decentralizes the data processing load in a plant by using devices on the network “edge” to gather and process data locally and much closer to the source. The origins of the technology go back some 25 years, when Akamai Technologies Inc. introduced its first content delivery network to alleviate Internet bottlenecks. This network featured nodes placed at locations geographically closer to the end user that stored cached static content such as images and videos.

Edge computing replaces the concept of static nodes with edge devices that are also able to perform basic computational tasks. In 1997, computer scientist Brian Noble demonstrated how mobile technology could use edge computing for speech recognition. Two years later, this method was used to extend the battery life of mobile phones. Today, manufacturers and other businesses are increasingly looking seriously into how they can benefit from the technology.

“In terms of manufacturing applications, edge computing has only started being deployed for about five years or so,” says Craig Resnick, a vice president at the ARC Advisory Group. “However, its usage is steadily increasing for automation equipment by manufacturers in many industries. The implementation of edge computing is evolutionary rather than revolutionary, but you will begin to see applications in most plants in just a few years.”

Therefore, edge computing is a key emerging component of the Industrial Internet of Things (IIoT), according to Resnick. As the IIoT extends the network edge to industrial devices, machines, controllers and sensors, data will be generated faster and in greater volume than ever before.

Until a few years ago, manufacturers only considered using edge computing when their legacy servers died. Such thinking is slowly being replaced with the awareness that edge computing and the IIoT go hand in hand, and that, when integrated properly, both can optimize plant production. This fact is clear to the research firm MarketsandMarkets, which expects the global edge computing market to reach $6.72 billion by 2022.


Consider the Benefits

For plant managers, edge computing offers several real-world benefits. Number one is reduced network latency, which is the amount of time between when a sensor starts sending data and when an action is taken on the data.

Factors that impact this latency are numerous. They include the propagation delay through the physical media of the network; the time it takes to route data through the networking equipment (switches, routers, servers, etc.); and the amount of time it takes to process the data.

With edge computing, managers no longer need to wait several days for a cloud service to receive and analyze key data, and send back a recommended function. Instead, data generated by edge devices, such as sensors and actuators, can be used to create a machine learning function that can be deployed in a matter of hours.

A related benefit is the elimination of costs associated with storing, computing and transmitting data to the cloud. These can be substantial, depending on the velocity and volume of data that originates on the network edge. Less reliance on the cloud also results in less network jitter, which can lengthen latency to the point that it prevents the system from acting within the required time frame.

Equally beneficial is edge computing’s capability to achieve predictive maintenance on an assembly line or throughout a plant. Installed sensors constantly monitor machine health and identify signs of time-sensitive maintenance issues in real time. This data, in turn, is analyzed on the assembly line, enabling managers and workers to perform corrective actions on machines long before they stop working and halt production. Besides preventing costly plant shutdowns, preventive maintenance helps companies prolong the useful life of machines.

“Quick acquisition and processing of data, as well as secure data storage, are definite advantages of edge computing over cloud computing in select applications,” notes Resnick. “In addition, edge computing provides the real-time intelligence that gives managers more flexibility when they need to make important decisions in times of crisis, such as the recent coronavirus outbreak.”

Scott McClelland, vice president of product management and engineering at Harting Inc. of North America, says that edge computing can improve both process and product quality. It also offers effective asset management and tracking.

“The manufacturing process usually involves multiple machines,” explains McClelland. “Data provided from all of the machines, in the aggregate, can help a manufacturer improve how the process is performed at each step, and, in turn, the quality of the final product.”

As for asset management, he says that high-value parts, raw materials and tooling or molds can be equipped with location-type sensors or RFID tags (that act as sensors) so these items are easy to locate in the plant. The same sensors can even be placed on parts for immediate location status as they move through the production process.

Edge computing systems are flexible enough to be integrated with cloud environments to create a hybrid edge-cloud infrastructure. This is also known as IT (information technology)-OT (operational technology) convergence. Data, log records and application info generated at the edge can and should be linked back to the cloud, whether private or public. Likewise, cloud-based resources should be tied back to the edge, to ensure that production continues, even if the cloud temporarily disappears for some reason.

A hybrid edge-cloud setup can also increase data security and protect data from cyber-attacks when manufacturers transfer data among factories, across states or between countries. For companies that operate multiple plants, virtualized edge computing resources can strengthen data recovery by replicating and mirroring it between each facility over a private, secure network.

What provides security at the edge also works in the cloud, and vice versa. This lets manufacturers create a strong framework for full enterprise security, regulatory compliance and audits.


Learn the Language

Manufacturers looking to use edge computing need to first learn and understand its terminology, which correlates to network architectural tiers. At the bottom are edge devices. Above them are gateways. At the top is the actual edge computing platform.

IIoT experts have different views about what constitutes an edge device. Some consider sensors and actuators to be edge devices, whereas others believe they are simply nodes or data-generation points. There is consensus, however, that PLCs and programmable automation controllers are definitely edge devices.

“Life on the edge of the IIoT is where creativity knows no bounds,” says Jeff Miller, applications engineering manager at Mentor Graphics. “But, this sensor-driven design environment is quite complex. Each sensor signal is sent to an analog signal processing circuit in the form of an amplifier or a low-pass filter. The output connects to an analog-to-digital converter to digitize the signal, which is sent to a microprocessor to process and analyze the data.”

Actuators, like sensors, are transducers. But, whereas sensors sense and send, actuators act and activate. Or, as Miller points out, “they cause something to happen in the real world.”

In edge computing, an actuator operates in the reverse direction of a sensor. It takes an electrical input and turns it into physical action. For instance, an electric motor, a hydraulic system, and a pneumatic system are all different types of actuators.

 “Because a sensor simply generates data and sends it out, it is right to consider it a node or a data-generation point,” opines McClelland. “But, an edge device is something that collects and analyzes data from the nearby machine it needs to evaluate. This device is also robust and has a high IP rating.”

Any and all data obtained by an edge device gets sent to an IIoT gateway, which may process it and only send the most relevant data back through the cloud, reducing bandwidth needs. Or, the gateway can send the data back to the edge device in the case of real-time application needs.

In simple terms, the gateway is a bridge between the device level and the level where the device-obtained data gets fully leveraged. It is Linux-based, and can be hardware, software or a combination of both.

The gateway’s typical functions include connectivity aggregation, and data encryption and decryption for security. Often times, it pre-processes the data from sensors and other data points.

“Industrial gateways provide an effective link between non-IP-based automation networks and the enterprise, i.e., the hardware and software needed by a large company,” notes Chantel Polsonetti, a vice president at ARC Advisory. “The gateway may [also] be used to interface automation-specific machine networks to Ethernet ‘backhauls’ (the connections from a wireless cell tower to the Internet).”

Polsonetti adds that gateways make it easy to integrate legacy automation protocols through their protocol-conversion capabilities. Plus, they help insulate OT assets from the IT environment, thereby addressing security concerns about device or machine access.

The top layer of the edge computing architecture is the platform, which easily connects to edge devices and cloud services using IIoT industrial protocols. Other platform functions include developing and managing IIoT edge computing applications, and visually composing flows to manage, analyze and route all data.

At the 2019 ASSEMBLY Show, Kinexon Inc. showcased its open-edge computing platform KinexonRIoT (Real-time Internet of Things). The platform is capable of processing up to 500,000 data points per second with a latency of less than 50 milliseconds. Internal intelligence optimizes and automates production and logistics processes in real time. Its Event Recognition feature monitors coupling, geofence and collision events. Application examples for the latter include accident prevention in intra-logistics traffic, and process reliability for screwdriving processes.


The Ways of Implementation

Before plant management can implement an edge computing system, it needs to be sure of the project’s ROI for the company. Verifying ROI should begin with conducting a small-scale initial experiment on legacy machines to show the value potential that a full-blown edge computing system will add to a specific plant or the organization as a whole.

This setup should be simple, use a specific data set for measurement and specifically define success metrics so ownership can be assured of a positive ROI within a concrete timeframe. Simplicity is key, because the more complex the initial setup, the longer it will take to fully deploy edge computing and the lower the likelihood that it will be successful.

By seeing the success of just a few retrofitted machines, a manufacturer can then decide when it wishes to scale up the installation. This allows costs to be spread over a longer period and thorough planning to be undertaken. Equally important, this approach lets a company focus on what it specifically wants to achieve from digitalization, thereby helping it upgrade one or more facilities and profit from the benefits of improved performance.

“Companies always want a quick and reasonable ROI when using edge computing,” points out Resnick. “If a feasibility study can calculate even a 1 percent improvement in overall equipment effectiveness, then that’s often a worthwhile ROI, even though it may take two months or two years to achieve. Plant management understands that sometimes the main benefit of edge computing is lessening the amount being wasted, rather than increasing the amount of production.

“Eliminating unscheduled downtime, alone, may be a worthwhile ROI for edge computing,” Resnick continues. “Let’s say, for example, that a 1-hour downtime costs the company $100,000 in lost production. Eliminating five such events means a saving of $500,000, which is a quite nice ROI, so long as implementation cost is less.”

Having established ROI, plant managers next must decide on project partners. A manufacturer can attempt to go it alone, provided its in-house IT and OT teams are knowledgeable enough to handle all aspects of implementation. But, this scenario is not the norm.

Usually, according to Resnick, plant managers either work closely with their automation suppliers, system integrators or machine builders to install an edge computing platform that all of the parties agree upon. Or, plant managers, in agreement with these other parties, first seek technology input from an edge computing specialist, like ARC Advisory, before acquiring and installing the platform.

If possible, it’s a good idea to perform trials with multiple technology platforms. The goal is to determine the level of compatibility and interoperability of each platform in different parts of the plant.

A fourth approach to implementation is for management to work solely with an edge computing platform vendor, such as Harting. This approach may require several meetings, as well as customization of the platform to meet the management’s specific goals for the plant.

“We specifically determine the plant’s hardware, software, sensor and platform needs, and we provide all of it,” explains McClelland. “If older equipment is to be used, we need to know which protocols they use and which machines have or need network connectivity, in order to pick the right sensors. For plants with newer machines that use many protocols, we’d do a digital retrofit to translate the many protocols into one. This saves time and reduces human error when collecting information.”

In 2015, Harting introduced its MICA (modular industry computer architecture) edge computing platform. MICA features modular hardware and Linux-based open source software that allows for rapid development and quick deployment using Debian-based containers. The platform easily connects to sensors, machine tools, RFID readers and antennas, smart infrastructure and more. Capabilities include machine health monitoring, condition-based maintenance, process monitoring and asset tracking.

“The hardware and software we tweak for each application,” notes McClelland. “If a machine needs to be modified during implementation, we’ll work with the vendors and integrators as needed. As for sensors, sometimes we use what’s already installed, but other times we install new ones. These may vary from custom ones we create, to those in the Bosch CISS kit, or standard or IO-Link models from other third parties.”