Here’s how a layperson – me – explains what manufacturing is: It means taking raw materials and turning them into finished products.
If you want a more formal definition, here’s one from the U.S. Bureau of Labor Statistics: “The manufacturing sector comprises establishments engaged in the mechanical, physical, or chemical transformation of materials, substances, or components into new products.”
It sounds old-school and highly physical – and perhaps not exactly fertile terrain for computing innovation. Yet manufacturing, just like the overall industrial sector, is a natural fit for edge computing and related trends like IoT, AI, and machine learning.
Automation is a big deal in manufacturing and has been for eons. (The industry even has entire trade publications devoted to the subject.) When the broader business world talks about how people and machines – or people and code, in the case of technologies like RPA and AI/ML – will work alongside one another now and in the future, manufacturing CIOs smile and nod knowingly.
There are tons of machines, robotics, sensors, and other devices generating massive amounts of data. To maximize the value of that data, manufacturing companies need maximum flexibility in their IT infrastructure. For reasons similar to the industrial sector, edge computing architecture isn’t an unlikely choice in manufacturing settings – it’s a natural one.
[ Developing an edge strategy? Also read Edge computing: 4 pillars for CIOs and IT leaders. ]
“The manufacturing industry continues to push toward edge applications – from robots scurrying around the warehouse to acoustic calibrators to cameras spotting flaws on the manufacturing line – to advance factory automation and efficiency,” says Brian Sathianathan, CTO at Iterate.ai. “There’s no question that edge computing is, and will continue to be, hugely important to the industry. The challenge for CIOs in this industry, though, is how to put the power of edge systems into place while ensuring edge applications stay always-on and don’t wreak havoc on their networks.”
Edge computing gives manufacturing CIOs a model for making strategic decisions about what should run in, say, a warehouse or on an assembly line – and what should run in a centralized cloud or data center, and what will flow from cloud to edge and vice versa.
As Red Hat technology evangelist Gordon Haff told us recently, “The idea is that you often want to centralize if possible, but keep decentralized as needed.” And Haff’s technical evangelist colleague Ishu Verma points out that edge architecture also enables IT leaders to standardize their edge operations on the same practices and tools used in their centralized environment(s).
“This approach allows companies to extend the emerging technology best practices to the edge – microservices, GitOps, security, etc.,” Verma says. “This allows management and operations of edge systems using the same processes, tooling, and resources as with centralized sites or cloud.”
While potentially true for any industry, this is particularly important in a sector like manufacturing, in which an organization could very well have thousands of edge nodes (or more) running in highly diverse, tough settings.
5 examples of edge computing in manufacturing
With that in mind, here are five examples of manufacturing organizations that can use edge computing.
1. Quality control automation
Again, automation is typically a big deal in manufacturing, though how it manifests can vary considerably.
“Manufacturing facilities can have minimal automation all the way to a fully automated production line,” says Andrew Nelson, principal architect at Insight.
Edge/IoT implementations can become increasingly useful as an environment moves toward the “fully automated” end of the spectrum.
Quality control automation on a production line is a good example, according to Nelson, and is common in settings such as a canning line in the beverage industry or the packaging process in the food or agribusiness settings.
A mix of computer vision, sensors, and other instrumentation can detect anomalies or other issues; being able to act rapidly on that data requires keeping it as close to the process as possible.
2. Warehouse automation
A similar but separate automation use case is in the warehouse, where functions like inventory management are rich with data and opportunities for increased efficiency.
“Some manufacturers run warehousing next to the production lines,” Nelson says. “Computer vision can be used to manage inventory levels and help with product picking. RFID/BLE earlier can also be leveraged for item locations and quantity levels. Smart shelves can be instrumented with sensors as another data point.”
Sending all of that data back to a cloud or centralized data center isn’t likely the most effective option from a cost or performance standpoint. Edge deployments create the flexibility to make more optimal decisions about what to run locally in the warehouse, whether for latency, cost, security, or any other reason.
3. Production line diagnostics
We hear lots about “predictive analytics” these days, but it’s a broad term – its actual value depends on business- or industry-specific applications, and manufacturing has a big one: using machine data to more precisely monitor and predict when the vast number of moving parts and pieces in a manufacturing setting will break down or otherwise require maintenance.
“The [production] line itself can be instrumented to predict issues with bearings, belts, motors, etc.,” Nelsons says. “In many cases, a line going down for maintenance can cost a company a lot. If you can predict or triage the issues quickly, you can minimize the downtime” and potentially save significant ongoing costs.”
In that context, latency becomes expensive. Processing that data locally can produce a tangible financial ROI. And that ROI can be magnified by combining this type of predictive analytics with the quality control/quality assurance automation Nelson described above.
“This can be merged with the Q/A processes in one landscape with multiple benefits and larger ROI,” Nelson says.
4. Product logistics and tracking
This category extends the edge of the edge, enabling inventory tracking and other uses even as products move out of the manufacturing environment into other phases of the supply chain.
“RFID and Bluetooth low emission [technologies] can be used to track products as they move through the line and out of production into crates and pallets and even when moving to shipping containers,” Nelson says. “Trucks can be scanned on the way in and out of a warehouse to address both input and output product levels.”
It’s a reminder that, as edge servers and applications, the boundaries of “the edge” may continuously expand.
5. The "golden" use case: AI/ML applications
If reducing latency is the most common driver of edge computing strategies, then AI/ML workloads seem likely to become the golden use case, at least in manufacturing.
“The most powerful manufacturing edge deployments are dependent on the power of the AI fueling them, but getting smart machines working seamlessly at the edge requires a lot of data,” says Sathianathan, the Iterate.ai CEO.
The problem isn’t a lack of available data – all of the above use cases reflect the reality that manufacturing CIOs are awash in information. In fact, Sathianathan says manufacturing has an advantage over some other industries when it comes to AI/ML because so much of an organization’s data is machine-generated.
[ Related read: Edge infrastructure: 7 key facts CIOs should know about security. ]
“Unlike data in other sectors that include much more bias and noise, manufacturing system data is ‘golden data’ that is particularly relevant and valuable,” he says.
The challenges occur when trying to send all of that data back from the manufacturing site to the cloud or data center. As Sathianathan told us recently, there can be such a thing as “too much data” to pass from a factory or warehouse floor through the local network and to the cloud and back again.
“That’s no good, because, as manufacturing CIOs know, decisions must be made instantly to be effective,” Sathianathan says. “And while some downtime is usually acceptable in standard IT environments, that’s simply not the case in manufacturing. The costs of halting production lines because edge applications are faltering can be hundreds of thousands of dollars per minute – there just isn’t room for error.”
As edge computing and AI/ML technologies mature, both in terms of infrastructure and in terms of developing lighter-weight applications (via low-code and other tools), they become a match made in IT heaven.
“Advances in AI and edge servers with GPU-centric architectures are now becoming available and, for manufacturing CIOs, it’s a much better solution to start placing AI applications on the edge,” Sathianathan says.
[ Learn how leaders are embracing enterprise-wide IT automation: Taking the lead on IT Automation. ]
