Towards manufacturing sustainability: Automated sorting of composite building materials for improved recycling

Recycling composite building materials, which include elements such as concrete, gypsum and bricks, is essential to mitigating the environmental impact in the construction industry and benefit manufacturing sustainability. By reusing these materials, it not only reduces waste, but also contributes significantly to reducing CO2 emissions, as recycling requires less energy compared to producing new materials. The main process in recycling is sorting materials, which is still carried out manually and involves skilled workers. These methods can be labour-intensive and ineffective. Therefore, introducing automation and partial automation to this sorting process can make a big difference. This paper aims to design an automated sorting system using the image information from two camera systems and laser-induced breakdown spectroscopy (LIBS), which provides information on the chemical composition of the sample in a measurement range of about 100 µm. Due to the heterogeneity of building debris, real-time selection of suitable measuring points on the sample surface is necessary for the targeted use of LIBS. An algorithm is developed that automatically determines suitable measuring points based on the sample topology recorded by a distance sensor and the RGB data from a video camera. Therefore, clustering (k-means) is applied to the RGB-data to gather information on the heterogeneity. LIBS-measurable areas are extracted from the distance sensor's data. A tree of possible measurements is built, and the highest rated branch provides the measurement positions for LIBS. The results show that the algorithm finds suitable measurement positions in which the heterogeneity of the surface composition is mapped. Particularly in the case of limited measurement points the algorithm provides positions for high surface representation of the LIBS data. The validation of the developed system is carried out on synthetic data and testing on photogrammetrically recorded real samples. The refinement of the sample's material map (e.g by optimizing the clustering) can elevate the method towards industrial usability. This will contribute to the implementation of circular economy.