Advancing Carbon Reduction with an Intelligent Scrap Steel Robotic Grading System

Background

The steel industry, as a high-energy-consuming sector, consistently ranks among the top emitters of carbon dioxide. Statistics indicate that using one ton of recycled steel can reduce carbon emissions by approximately 1.6 tons. Enhancing the comprehensive utilization of scrap steel is a crucial pathway for the steel industry to achieve significant carbon reductions. Traditional sorting methods often rely on manual labor or mechanical separation systems that may not sort materials accurately. The customer's solution, Smart Recycled Steel Grading 2.0, provides objective, fair, and traceable grading results by integrating AI and robotic arms. This innovation represents a pivotal shift from exploratory deployment to industry-scale replication, delivering more accurate grading, broader application scenarios, and faster deployment for recycled steel grading processes.

Project Challenges

Developing an AI-powered machine vision and robotic sorting system presented several challenges: 

Reliability and Durability: The grading solution needed to be deployed on-site, requiring stable and reliable hardware capable of withstanding harsh industrial environments. 

Cost Efficiency: The existing proof-of-concept setup used expensive rackmount servers. A more cost-effective solution was necessary for on-site AI inferencing, without sacrificing performance. 

Flexible Configuration: Customizable hardware configurations were essential to adapt to various deployment scenarios.

Key Requirements

To address these challenges, the ideal robotic sorting system needed to offer: 

Comprehensive Certification: A fanless design paired with a robust certification system to ensure compliance and reliability in demanding environments. 

AI High Performance and Expandability: Support for NVIDIA GPUs, desktop-grade multi-core CPUs, up to 64GB of memory, and RAID data redundancy to ensure optimal AI performance. 

Durability: Features such as anti-vibration, shock resistance, wide voltage, and wide temperature range capabilities to ensure a long product lifecycle. 

Flexibility and Scalability: Modular and customizable options for seamless integration into diverse deployment scenarios.

Solutions

The customer adopted Advantech's ARK-3534B Edge Computer, equipped with an NVIDIA Tesla GPU card, as a cost-effective alternative to expensive rackmount servers. The solution provided: 

Compact Yet Powerful Design: A smaller footprint, ideal for on-site deployment in diverse environments. 

Expandability: The system can integrate a sophisticated suite of IR sensors, including steel detectors, motors and cameras, to ensure precise classification. 

Flexibility: Configurable to meet specific on-site requirements, with support for multi-device clustering and hot standby for enhanced reliability. 

Cost Efficiency: Reduced overall costs while maintaining or exceeding the performance of the previous server setup.

Outcomes

Implementing the ARK-3534B Edge Computer significantly improved the customer's scrap steel grading operations by: 

• Ensuring consistent and accurate sorting. 
• Lowering hardware costs without compromising performance. 
• Delivering a robust, scalable solution that maintains system stability, even in demanding industrial environments. 
• Accelerating the deployment and replication of the Smart Recycled Steel Grading 2.0 system, aiding the steel industry’s transition to greener practices.