Envisan-GEM is developing an AI-powered sorting line for construction and demolition waste (CDW) together with the Faculty of Civil Engineering at Czech Technical University in Prague (CTU) and AI firm BigHub. The three-year project (SQ01020186) is funded by the Technology Agency of the Czech Republic.
"This project is a prime example of the ideal synergy between academia and practice," said Václav Nežerka, Associate Professor in the Department of Physics at CTU's Faculty of Civil Engineering.
The problem is recyclate quality. Construction generates up to 40% of all waste, Europe produces more than 10 billion tonnes of CDW a year and the Czech Republic about 25 million tonnes. Official recycling rates for CDW reach up to 90%, but most of that material is downcycled into backfill or subbase rather than reused as high-value aggregate, because it is contaminated or poorly sorted. Replacing part of the natural aggregate in concrete with recyclate can cut material costs by up to 20%, but only if the recyclate meets standards such as EN 12620.
The line uses a high-resolution RGB camera to capture the structure, colour and surface of fragments on a conveyor in real time; a second camera with a wavelength filter can be added if recognition accuracy proves insufficient. Machine-learning models classify each piece as concrete, brick, ceramics, wood or another component, and a pneumatic system then fires a jet of compressed air to divert it into the correct material stream.
BigHub is designing the line's intelligence layer across three areas: computer vision and material classification; MLOps and cloud infrastructure on Microsoft Azure, covering model versioning, performance monitoring and continuous retraining; and reports and dashboards to track sorting performance and recyclate quality. According to Nežerka, CTU contributes scientific expertise, knowledge of materials and the theoretical framework, while BigHub brings the AI know-how.
The data-collection prototype and evaluation software are complete. No sorting accuracy or throughput figures have been published yet; further development with industrial deployment is expected by 2027, with commercialisation as the stated goal and potential for export to other EU markets.
Details
- City
- Hůry
- Organization
- Envisan-GEM