SBIR Phase I: Multimodal Acoustic Sorting for Cost-Effective Plastic Recycling

The broader impact of this Small Business Innovation Research Phase I project is to improve the effectiveness, affordability and efficiency of plastic recycling in the United States. Current sorting technologies struggle to accurately classify plastic resins, especially materials that are black, shredded, or contaminated, leading to high costs, material loss, and increased landfill use. This project investigates a new approach that uses sound to identify plastic types, potentially enabling accurate sorting across a wider range of polymers at lower cost. By making advanced sorting technology accessible to smaller and rural recycling facilities, this work aims to increase reclamation rates, reduce waste, and expand the domestic supply of high-quality recycled materials. These improvements will support U.S. manufacturing by reducing reliance on imported recycled plastics and strengthen local supply chains. The project also has the potential to reduce environmental impacts associated with plastic waste and contribute to national recycling and sustainability goals. This project will investigate an unexplored and versatile modality for plastic classification based on acoustic signatures generated by material impacts. Unlike conventional near infrared, visual, or density-based methods, which are limited by color, form, and contamination, this approach leverages the intrinsic vibrational response of materials to enable classification independent of appearance or optical properties. This approach may also be fused with complementary sensing methods, such as computer vision, to improve performance across diverse material streams and form factors. The proposed work will evaluate whether acoustic data, collected using a low-cost sensor set, can achieve the precision and robustness required for industrial plastic sorting applications. The scope includes three core objectives: (1) development of a scalable dataset of synchronized audio and visual data for high-priority polymers and form factors, including municipal waste and shredded electronic plastics, (2) conducting audio feature sensitivity studies and machine learning model benchmarking to optimize classification performance, and (3) an investigation of relationships between acoustic signatures and underlying material and chemical properties through comparison with laboratory-based characterization. Key technical risks include sensitivity to material additives, environmental noise, and variability in real-world waste streams. If successful, this research will demonstrate a new class of low-cost, non-destructive material identification technologies capable of improving sorting accuracy, reducing system costs, and enabling broader deployment across diverse recycling and manufacturing applications. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.