Remove chlorinated waste from refuse derived fuel with rapid recognition technology

Abstract

Refuse-derived fuel plays a crucial role in waste-to-energy applications, offering a sustainable solution to mitigate global warming and waste management challenges. However, chlorine contamination in RDF poses significant industrial challenges, including severe boiler corrosion, unplanned downtime, and toxic gas emissions, highlighting the urgent need for efficient chlorine detection and removal. This study proposes a methodology combining near-infrared spectroscopy with deep learning architectures, including ResNet and CNN. A fuzzy labeling approach was implemented to enhance the adaptability of sorting to chlorine levels compared to binary classification. A dataset with 35 typical industrial solid wastes including textile, plastics and artificial leathers containing chlorine from 0 % to 34 % was built. Under simulated industrial conditions, the ResNet-based model achieved a classification accuracy of 87.6 % for new RDF materials. This advancement provides a reliable, scalable solution for detecting chlorine in diverse RDF scenarios, marking a substantial step forward in waste-to-energy processing and offering practical benefits to the industry.