ARTIFICIAL INTELLIGENCE

Software-defined manufacturing (SDM) is a product of Industry 4.0 concepts that have been talked about for more than a decade. The idea is to make factories more like data centers, allowing operational technology (OT) operations to be orchestrated much like modern cloud workloads that can be routed to, and executed on, any node with the resources to perform the task at hand.

SDM was especially top-of-mind during the embedded world 2024 Exhibition & Conference, which kicked off with a Day 0 networking event featuring a conversation about AI and manufacturing with Christine Boles, Vice President of the Network & Edge Group at Intel and General Manager for Federal & Industrial Solutions at Intel. She reflected on the past couple of years, from when software-defined manufacturing was in its infancy to its current evolution, where it is now more of a reality thanks to ecosystem partners and industry groups like the Open Process Automation Forum (OPAF) and the PCI Industrial Computer Manufacturers Group (PICMG), all of which are working to lay the foundation and future for SDM—enabled by workload consolidation, virtual programmable logic controllers (vPLCs), time-sensitive networking (TSN), and “AI Everywhere.”

“AI is driving momentum and it’s helping enable software-defined manufacturing,” Boles said. “Because of how fast the technology moves, it’s important that we, as an industry, work to make adoption, deployment, and management as easy as possible. That way, more manufacturers will be able to take the first step.”

Hybrid AI Lays the Groundwork for Software-Defined Manufacturing

From computer vision to real-time networking, there was a particular focus on enabling the intelligent edge enroute to an SDM paradigm, specifically around the use of edge AI in manufacturing, with exhibitors demonstrating how industrial operators can leverage it to:

• Optimize productivity through real-time adjustments to production lines.
• Control costs and increase efficiencies with predictive maintenance that inspects products for defects and performs root cause analysis.
• Secure increasingly connected automation infrastructure by providing extra layers of protection in platforms like intrusion detection and prevention systems.

Edge AI is particularly well suited to these tasks because the proximity to data sources permits faster decision-making and reduces the networking costs of sending information to the cloud. But for the most part, industrial edge AI will demand revamped compute and networking infrastructure, Boles explained.

She shared that she and her team at Intel have been working with manufacturers to better understand their requirements. What they learned is “legacy infrastructure that is still very fixed function, with very little flexibility” and is “a sticking point for new workers and a barrier to implementing technologies like AI.”

As a result, “a hybrid AI approach is necessary where manufacturing is concerned—one that combines cloud software with edge AI,” she said.

Hybrid AI is a distributed method of operating on AI workloads that simultaneously delivers the real-time advantages of edge AI with the deep performance and comprehensive insights of the cloud. It addresses current obstacles to industrial edge AI by allowing operators to make use of cloud-based AI services today while their operational technology infrastructure is incrementally upgraded with AI-enabled platforms. Once fully deployed, these architectures enable the type of layered intelligence required for true SDM.

And “upgraded” is a relative term. “There’s a general assumption that if you use AI, you need a GPU, but you don’t,” Boles said. “What it really comes down to is knowing what kind of function you’re trying to do, and recognizing that when you start getting into heavier vision-based data streams, you’ll need higher-performance compute and potentially a different type of acceleration.”

Partner Collaboration Drives Innovation in a Software-Defined Industry

Boles concluded that ultimately partnerships enable technological advancements and innovation.

“We continue to work with our ecosystem of partners to bring solutions that really help change and transform the industry,” she explained.

For instance, Boles highlighted NEXCOM and the NexAIoT team, providers of industrial and manufacturing solutions, which deployed an autonomous mobile robot built on Intel^® Core^™ processors, Intel^® RealSense^™ cameras, and OpenVINO^™. Solomon Technology Corporation, which specializes in 3D vision systems, built an automation defect classification solution leveraging OpenVINO and the Intel^® Edge AI Box. And Chieftek Precision, a manufacturer of high-quality linear motion robotics and robotic controllers, which developed an AI-powered miniature robotic arm tailored to high-precision manufacturing applications, thanks to Intel processors, Intel^® Edge Controls, OpenVINO, and Intel vPro^® technologies.

Moving forward, Intel will continue to provide more opportunities to collaborate with ecosystem partners, accelerate progress within the manufacturing community, and support end users in this space. With programs like the recently announced Intel^® Industry Solution Builders, Intel empowers industry specific communities for both Intel^® Partner Alliance partners as well as end users in their adoption journey, providing access to resources, training, and engagement opportunities across industries.

“Intel has always worked with our ecosystem partners to bring solutions and solve industry challenges,” Boles added. “For Intel, our partner ecosystem is both our main go-to-market and our primary path to innovation. That’s as true in manufacturing as in any other sector.”
 

This article was edited by Christina Cardoza, Editorial Director for insight.tech.