High-throughput-high-accuracy materials identification and sorting for recycling

Important results from the project

The project developed and integrated optics, electronics and AI for real-time identification of plastics on moving conveyor belts. For selected plastic types, 99+% identification accuracy was achieved, exceeding the target of at least 95%. The AI model runs directly on the demonstrator´s FPGA without an external computer. Full lateral fast scanning and industrial capacity validation remain as the next development steps.

Expected long term effects

The results could eventually enable cleaner plastic fractions, more material recycling and less incineration, landfilling and use of new fossil raw materials. A compact module that can be integrated into existing sorting lines could lower the threshold for industrial implementation. The technology could also strengthen material supply and resilience by making recycled plastic a more reliable source of raw materials.

Approach and implementation

The project was implemented in two interconnected work packages: AI programming and detection optics and electronics. Sensor data was collected, the AI model was trained and tested in real time on a moving conveyor belt. In parallel, optics and electronics were designed, purchased and installed and integrated with the conveyor belt. The setup worked well and the core milestones were achieved.

External links

Identinano webbsidan