Edge Computing's Critical Role in Industrial AI

By Austin Locke and Kevin Olikara, Kalypso: A Rockwell Automation business

Artificial intelligence (AI) is transforming industrial systems, offering organizations the opportunity to unlock value, efficiency and greater autonomy. It has become a top priority for operations, IT and engineering leaders driven by labor shortages, plants operating at maximum capacity, and increasing demands for product quality.

The industrial sector is still in the early stages of AI implementation, but with companies allocating nearly 20% of their IT budgets to industrial data analytics and AI, analysts expect rapid growth and widespread adoption.

One area where AI is making a significant impact is in quality inspection, where Vision AI systems automate processes, establishing consistent quality across shifts, lines and plants. This also improves use of process specialists, allowing them to focus on exceptional cases.

AI is also being integrated into industrial control systems, optimizing process yields and reducing the need for constant operator surveillance.

Edge Computing’s Role

Edge computing plays a critical role as AI use cases become increasingly prevalent. Edge is an enabling capability, supporting applications that necessitate low latency, cost-effective processing.

Several manufacturers on the forefront of industrial AI are building “edge in” architectures. They’re combining AI models, edge processors and the Industrial Internet of Things (IIoT) to bring intelligence to the machine level. This architecture facilitates what analysts believe to be foundational to a future wave of "hyperautomation.”

Many manufacturers are in the early stages of defining an edge strategy. The most successful edge strategies are use-case driven, meaning the strategy is defined by the requirements of the application or problem being solved.

Two manufacturers have recently deployed industrial AI solutions that use an underlying edge capability. These examples illustrate real steps towards autonomous operations with AI, facilitated by edge computing.

Real World Case Studies

Quality Inspection and Closed-Loop Quality Control. A major food and beverage manufacturer has implemented Vision AI at the edge for quality inspection and closed-loop quality control. Facing significant workforce constraints and increased quality standards, its objective was to move from human-based defect detection to an automated system using vision-sensing systems.

An AI-enabled vision system was deployed at the edge to continuously monitor product variances and recommend changes to equipment settings.

Manufacturers that have successfully implemented this use case are seeing the quality inspection cycle time improved by 50-75% while enhancing the accuracy of inspection and labor productivity.

Advanced Process Control. A leading tire manufacturer implemented an AI-driven adaptive control system leveraging edge computing. The goal was to modify the automation controller behavior in response to changes in the dynamics of the manufacturing process to maintain system performance at optimal levels.

AI was used to identify causal relationships, predict process results and prescribe an optimal action to the automation controller. Closed loop feedback from an AI model required low latency and high reliability, which was only possible by having the AI run time operate side-by-side with the process.

The result of this application reduced out-of-tolerance events by 50%, ultimately improving plant capacity to meet increasing demand.

Scaling AI at the Edge

In the rapidly evolving era of AI, manufacturers that successfully scale their AI investments will benefit from a competitive advantage. Many organizations have pockets of excellence — minimum viable products or proof of concepts that show the value of industrial AI — but far fewer have the strategy, resourcing or tools to scale these successes.

One of our clients leads the digital transformation team for one of the world’s most successful consumer brands. He says their backlog of AI use cases for manufacturing control is so deep, that it could consume the entire digital transformation agenda on its own.

The challenge for this business isn’t demonstrating the value of AI, but building the strategy to scale it at a pace that will keep them ahead of their competition.

Organizations need to consider edge management and orchestration (EMO), cloud and industrial asset connectivity, and edge security tools and policies. EMO automates routine tasks, improves efficiency and provides real-time visibility of edge device status.

Cloud and industrial asset connectivity support seamless workflows. Edge security tools and processes mitigate cybersecurity risks and support a secure and scalable system.

Edge Management and Orchestration. At scale, effective fleet management is crucial for secure and efficient operations across the enterprise. Manual management is costly, prone to errors and poses security risks.

Fleet management automates tasks such as setting access rules, deploying applications and conducting software updates for both the operating system and applications. Real-time visibility, centralized control and deployment agility are especially important for AI applications, which evolve quickly and are subject to model learning and retraining.

As the quantity and variety of edge applications expands, an edge-management platform can apply logic to determine which updates should be deployed to specific edge devices and when.

Cloud and Industrial Asset Connectivity. As IT/OT convergence continues, cloud and edge must work together to enable industrial AI at scale. The cloud provides the necessary calculation for analyzing large data sets and building and training models. It can also provide tools for centralized management of edge content (e.g., EMO) as well as control over the machine learning (ML) model life cycle (Model Ops).

The industrial edge must aggregate data in a heterogenous environment, operationalize models and operate in real-time. Flexibility, resilience and adaptability are paramount. Building a coherent plan that harmonizes activity between the cloud and edge is essential for scaling AI in manufacturing.

Edge Security Tools and Processes. Unmanaged computers present serious cybersecurity risks. Having visibility to edge device telemetry, with complete granularity of every application instance on each edge device, is critical to identifying potential risks and prioritizing mitigation actions like software updates.