Automated material flow characterization of WEEE in sorting plants using deep learning and regression models on RGB...

Area of Science:

• Recycling and Waste Management
• Sensor Technology
• Artificial Intelligence

Background:

• Waste from electrical and electronic equipment (WEEE) is a growing global challenge.
• Recycling WEEE is crucial for recovering valuable and critical raw materials.
• Current manual sampling and sorting for WEEE analysis are labor- and cost-intensive.

Purpose of the Study:

• To develop an automated sensor-based material flow characterization (SBMC) method for shredded WEEE.
• To optimize WEEE recycling processes through accurate material flow composition analysis.
• To overcome limitations of existing SBMC methods not yet applied to shredded WEEE.

Main Methods:

• Developed a three-step SBMC method using RGB cameras for shredded WEEE.
• Utilized YOLO v11 deep learning for material identification (metals, plastics, PCBs).
• Employed k-nearest neighbors regression for particle mass prediction and material flow composition analysis.

Main Results:

• YOLO v11 achieved an mAP@0.5 of 0.990 for material identification.
• K-nearest neighbors regression predicted particle masses with a mean relative error below 5%.
• The combined YOLO and k-NN approach achieved a 4.94% error on a validation dataset.

Conclusions:

• The developed SBMC method provides an accurate and automated solution for analyzing shredded WEEE.
• This technology can enhance monitoring and control in WEEE sorting plants.
• The method facilitates efficient recovery of valuable raw materials from electronic waste.