Advanced deep learning for automated ferrous sorting in steel recycling
An innovative deep learning system automates the sorting of ferrous materials from scrap in steel recycling. This solution enhances accuracy, efficiency, and sustainability, reducing manual labor and errors while ensuring reliable classification and transparency through explainable AI.
Overview
This advanced deep learning system is designed to automate the identification and sorting of ferrous materials from scrap, enhancing the accuracy and efficiency of steel recycling processes. By integrating this system, industries can significantly reduce manual labor and errors, ensuring a scalable, high-performance solution that minimizes disruptions to existing operations. The technology not only optimizes resource utilization but also contributes to more sustainable and efficient recycling processes.
Technical specifications
The system leverages the SwinV2+regression model to automate yield estimation and improve sorting accuracy. It operates through two main sections:
- Automated Yield Estimation: Utilizes SwinV2 with regression to predict yield, extracting hierarchical features from scrap images and mapping them to obtain ferrous yield.
- Ferrous/Non-Ferrous Classification: Enhances sorting after magnetic separation using SwinV2, which extracts both global and local features. Conformal prediction ensures reliable classification, while Explainable AI (XAI) enhances transparency by identifying decision-relevant regions.
- Benefits: The technology promises automated yield estimation, increased classification accuracy, improved reliability with conformal prediction, and enhanced transparency through XAI.
Technology readiness level
This system is currently at Technology Readiness Level 5, indicating that it has been validated in relevant environments and is a step closer to commercial deployment. Future developments will focus on refining the system for broader industrial application.
About APJ Abdul Kalam Technological University
APJ Abdul Kalam Technological University is a state technological university headquartered in Thiruvananthapuram, Kerala, with a broad engineering-focused academic and research mandate. Its statewide affiliated-college network extends access to faculty expertise, postgraduate education, and doctoral supervision beyond the central campus. University schools provide an industry-oriented platform for postgraduate training, innovation, entrepreneurship mentoring, incubation support, and collaboration with research networks. E-governance and affiliation systems give companies a structured route into Kerala’s technical-education ecosystem, while part-time PhD pathways can accommodate professionals working in industry and other organizations. Research is supported through competitive funding from Indian national agencies and government programs.
delivered to your inbox.