Edge computing and hybrid cloud: scaling AI within manufacturing

By | 2 minute read | August 3, 2021

Manufacturing plant

Industry 4.0, or the fourth Industrial Revolution, promises higher production volumes, lower costs and better quality products. Most manufacturers pursue such gains through digitization with big data, AI and the Internet of Things. Often, however, they build solutions in isolation—AI deployments here and there without an enterprise strategy.

We at IBM Systems Manufacturing have a different strategy. By combining the hybrid cloud with edge computing, we can scale the value of AI across the global manufacturing enterprise.

Consider, for example, our first-of-a-kind AI visual inspection system. It’s deployed on assembly lines at the network edge in IBM plants in Canada, Hungary, Mexico and the U.S., and we’re managing it using IBM Hybrid Cloud and edge computing technology.

AI visual inspection at the network edge

The payback was immediate. Compared to a human inspector, AI automation reduces inspection times from 10 minutes to one minute in one use case and increases inspection accuracy. In fact, our teammates who inspect components are relieved to have a computer doing this tedious work.

We manage the AI models and edge devices from a central location via the cloud, an automated process that reduces software maintenance costs by 20%. All told, we expect this “automation of our AI automation” to save millions of dollars annually.

It’s clear that edge computing can drive a profound transformation of how manufacturers use big data, AI and the IoT.

Instead of sensors just sharing information with a central hub, today’s endpoint devices are becoming smarter with their own compute and memory resources. Many are full-fledged computers that require management, software updates and security patches. Real transformation is happening using AI in such devices at the network edge.

Edge computing delivers real-time insights

In the automated inspection example, AI models are deployed to the edge devices for inferencing. The images are big-data class and the workloads are compute-intensive. That’s why we process them at the edge, where the data is created, so we can detect anomalies and act on them in real time. Edge computing eliminates problems of bandwidth and latency from running inferencing in a data center.

Now, as much as we’d like the manufacturing line to be static, we constantly make small changes to improve quality. Each change requires a new AI model that must be retrained and then deployed to the edge devices.

Manual deployment is possible for a few stations. But what if a manufacturer has thousands or even hundreds of thousands of edge inspection stations around the globe, all requiring AI model updates and device management? Automation from the hybrid cloud and edge computing accelerates these tasks.

The power of edge application management

Powered by IBM Edge Application Manager, our solution leverages hybrid cloud tools like the Red Hat OpenShift Container Platform to let Model Engineers automatically scale AI model deployments to edge devices across the enterprise. This relieves the burden of supporting the solution without slowing production. It’s truly the best of both worlds, delivering AI insights at the edge along with centralized IT management.

The National Association of Manufacturers’ Manufacturing Leadership Council recently recognized our edge inspection solution for AI and Advanced Analytics Leadership. We at IBM Systems Manufacturing believe it’s the next logical step in manufacturing.

Our journey started using big data, AI and IoT technology, which are not necessarily scalable. By applying the hybrid cloud and edge computing, we’ve added a revolutionary dimension that brings us closer to Industry 4.0.

 Hear Christine Ouyang discuss manufacturing automation with AI and edge computing: