From Scrap Waste to Self-Correction: Performance Revolution in Production with Edge AI

[Erkan Teskancan]

Increasing globalization, labor shortages, supply chain issues, and sustainability pressures create a complex web of challenges for manufacturers. Under these conditions, manufacturers must improve quality, reduce scrap and emissions, and maintain productivity, but often lack the necessary personnel and expertise.

The solution lies in turning to digital tools. Artificial intelligence (AI) and machine learning (ML) technologies can bridge the gaps in achieving operational excellence. However, many AI tools are cloud-based for high processing power and unlimited scalability. Yet, in industrial environments, cloud usage is often impractical due to latency, poor connectivity, security constraints, and cost issues.

Production lines require real-time feedback loops; cloud connectivity cannot provide the necessary performance. Therefore, manufacturers are turning to new methods to use AI locally on-site without relying on the cloud.

### Edge AI and Real-Time Correction in Industry

• Industrial AI systems support physical AI applications in factories with real-time independence.
• AI instantly detects deviations and makes automatic machine corrections, virtually eliminating scrap.
• For example, if there are cameras on the production line taking 30-100 images per second, even millisecond delays can affect production.
• Cloud connectivity issues can halt production, especially in remote and challenging industrial areas.
• Additionally, cloud technology poses risks in terms of security and cost.

### Emerson's Industrial AI Solution

• Emerson uses industrial PCs (IPCs) equipped with specialized AI acceleration hardware.
• These systems provide low-latency, closed-loop feedback and optimize production without the need for the cloud.
• These IPCs have fanless, low-power designs that are resistant to harsh environmental conditions.

### Specialized Processors for Edge AI

• Instead of GPT, specialized AI processors are optimized for real-time inference.
• The SiMa.ai MLSoC platform is integrated into Emerson's next-generation IPCs, enabling high-speed image processing and language model inference.
• These devices are superior to GPU-based systems in terms of energy efficiency.

### Emerson PACSystems IPC Features

• Hardened casing, fanless cooling system
• Long-life CPU and memory components
• Durable construction for continuous operation
• Predictable performance suitable for factory environments

### Edge AI Use Case in Industry

A thermoplastic pipe manufacturer monitors the pipe wrapping process with an AI-powered visual inspection system. Human inspection was slow and inconsistent, and there was a shortage of equipment. The AI system instantly detected wrapping errors and made automatic corrections, reducing scrap to almost zero. Additionally, video recordings were provided for quality and compliance.

### Conclusion

• AI applications in industry are supported by local IPCs and edge AI processors for low latency and high security.
• These systems provide continuous, automatic correction and improvement.
• Application areas can extend to many sectors such as laser welding, automotive adhesives, life sciences, oil and gas, mining, and recycling.

This new approach allows for increased efficiency in production while reducing energy consumption and costs.