A total of 500 waste-to-energy plants are currently operational in the European union. ‘These generate approximately 19 million tonnes of bottom ash every year,’ observes Liesbeth Horckmans of Belgian research firm VITO Sustainable Materials. She presents the INSTAnT characterisation technology as a way to boost recovery in this niche market.
At the moment, around 80% of metals embedded in bottom ash are recovered. ‘This is a good start, but we can do better,’ Horckmans told delegates at the recent Sensor-Based Sorting Conference in Aachen, Germany.
‘We use a combination of dual X-ray transmission and 3D laser triangulation sensors to get in-depth knowledge of the particle,’ she explains. After alignment and segmentation, features including mass, density, atomic number and shape can be determined. Subsequently, the data is processed by a machine learning model to classify the particles into six distinct classes. There are; glass, ceramic, slag, ferrous metals, aluminium and copper.
A total of 500 particles were handpicked for each of the three size fractions; 4-8mm, 8-20mm, and 20-40mm. The material used was the output of Suez bottom ash treatmeant plant in Grimbergen, Belgium. The test samples were scanned at a belt speed of 20mm per second.
Feasibility experiments were conducted at the Tomra Sorting’s testing site in Müllheim-Kärlich in Germany, using both the Autosort Laser (NIR) machine, the Combisense colour line camera, and a prototype laser unit that sorts based on laser illumination.
‘Our method has an overall accuracy of at least 90%,’ Horckmans reports. She does lament that overlap in the training fractions complicates perfect particle classification. After applying the CombiSense Chute machine to separate glass from the mineral fraction, purity was notably increased to at least 98% for slag.
Thanks to optimisation trials, glass purity was increased from just 9.6% to 81.9% in one step, with a 94.3% recovery rate. The purity rate of the aggregate was pushed to 99%. ‘The main challenge standing in the way of industrial implementation is the intensive and long-term preparation of representative pure training fractions,’ Horckmans concludes.
Would you like to share any interesting developments or article ideas with us? Don't hesitate to contact us.