Skip to main content

Opening the black box: the best of sensor-based sorting

Recyclers are eager to process every last flake, component and millimetre of waste. In order to succeed, it’s vital to understand what exactly is coming into the yard. Advanced sorting systems have proved to be a game-changer, although there is a lot of ground left to cover.

Read the full story here >>

The global waste sorting technology market is thriving, with new players stepping into the arena every year. Market analysts report this positive momentum will continue to build: already worth US$ 2.2 billion, the sector’s value is projected to reach almost US$ 10.5 billion by 2030.

Robotics booming

The robotic sorting segment alone is expected to witness 19% growth in the next four years. ‘We have not seen the true beginning of the impact of artificial intelligence (AI) in the recycling sector,’ comments AMP Robotics’ founder and ceo Matanya Horowitz. ‘Adoption rates are at roughly 2% if you factor in recyclers everywhere. Imagine the potential!’

His firm has brought smart sorting systems to 200 clients worldwide since opening its doors in 2014. ‘It’s been an exciting journey,’ he says. ‘Our first milestone was increasing the pick rate to 80 picks per minute, doubling the initial 40 picks by a manual worker. We’re working now to push it to 120 picks to triple the number.’

‘I believe the next five to ten years will see a massive uptake in installations; I’m not just talking about our business, but the entire industry,’ Horowitz says with confidence. ‘Systems are only getting better, faster and smarter. Besides, they could make recycling a lot safer too.’

AI predicts particle size

Naturally, today’s sorting industry has been heavily inspired by state-of-the-art artificial intelligence-based systems. Lisa Kandlbauer of Leoben University in Austria has looked at how the technology can help with particle size prediction in solid waste. She used a bi-layered neural network to set apart five particle sizes: 10-20mm, 20-40mm, 40-60mm, 60-80mm and fractions larger than 80mm.

‘In the first experiment, we achieved a classification accuracy of 75%, and then we reached 82% for virtually all classes,’ she reports. Using another method, the RUSboosted Trees model, the classification was about 65% accurate. The best score was the 97% achieved by the neural network method.

‘Especially for size classes of 20-60mm of paper and cardboard, the new AI-powered models can be seen as a suitable prediction method,’ she declares, noting that, on average, the level of falsely-classified particles was far below 20%.

‘To understand and explain abnormalities in the results, relations between the individual descriptors and images need to be investigated,’ Kandlbauer states. ‘Nevertheless, the presented AI method shows a promising way to predict and measure particle size in real-time when treating commercial waste.’

Transparency needed

Researcher Steffen Rüger of the Fraunhofer Institute believes AI can also be used to enhance aluminium recovery. Combining dual X-ray transmission technology and deep learning to sort a database of almost 75 000 metal fractions into pure aluminium has yielded output fractions with a 91.7% purity and similarly high recovery levels.

In Rüger’s view, the final deployment of innovative deep learning systems depends mostly on whether or not they are accepted by society and industry. ‘That’s an easier-said-than-done scenario,’ says the researcher. ‘With new technology, people don’t understand how it works – if they can believe the hype. That’s why transparency is key.’

And he concludes: ‘We employ local explainable AI solutions to hopefully open these black box models. To provide insights and foster trust amongst the people in the value chain. That’s step one.’

Rüger is confident the recycling industry has far more to learn from smart sorting systems, inspired by a previous study that successfully automated detection of bone splinters in different types of tissue with 90-99% accuracy.

Novel PCB approach

The recycling of printed circuit boards (PCBs) could benefit from a deep-learning, assisted X-ray imaging approach, reports Markus Firsching of the Fraunhofer Institute in Germany. ‘At the moment, no system is available that can estimate the monetary value of used circuit boards. Our team is working on an intelligent and robust automated method that sorts these based on features in the DE-XRT images.’

The solution is said to work even in dirty environments, making it a promising tool for recyclers. ‘We first created a sample of 104 individual circuit boards,’ Firsching says. ‘Our approach relies on the detection of components that contain valuable materials, like integrated circuits. Think of ball grids, pin grid arrays, tantalum capacitators and connectors.’

Fraunhofer’s research achieved accuracies of 87.8%, 82.5%, 79.2% and 88.8% in four different classes, namely integrated circuits, tantalum capacitators, connectors and ball/ pin grids. Their monetary values were very close to the expected values.

The power of LIBS

Louise Hagesjö of the Swerim research lab in Sweden’s capital Stockholm described laser-induced breakdown spectroscopy (LIBS) as a ‘fast, contactless and flexible technique’ for the chemical analysis of various materials. Her team created a testbed known as SenSoRe to track the innovative sorting method (using robotic sorting and artificial intelligence too) while also building a platform to share know-how with other research institutes, recyclers and sensor manufacturers.

The five-year R&D project was launched in 2019 with funding from the Swedish government. According to Hagesjö, the first pilot was meant as a quality control step for non-ferrous metal scrap. ‘Identification of light elements such as aluminium and magnesium makes LIBS an interesting addition to existing X-ray fluorescence solutions by allowing continuous monitoring of shredded fractions.’

The best results were achieved by adjusting the angle of the camera scanning the flow of the material, placed both horizontally and at an incline. ‘Our prototype initially detected around 10% of the samples because analysis is typically done by focussing on one spot in the middle of the belt.’

This was raised to virtually 100% for non-ferrous fractions during the first week of testing, measuring 39% copper content, 39% stainless steel, 13% brass, 8% zinc and 1% lead – ‘exactly what we expected from this batch,’ Hagesjö remarks.

A second experiment sorted plastics with an accuracy of 80%. She believes that training the algorithm and using a higher sampling frequency (a pulsing laser with a repetition rate of over 10Hz) could boost analysis of black plastics.

‘The next step is to do more field tests at industrial recycling facilities,’ Hagesjö adds. ‘I’d love to hear from recyclers and investors wanting to partner with us to realise this.’

Read the full article in Recycling Technology >>

Don't hesitate to contact us to share your input and ideas. Subscribe to the magazine or (free) newsletter.

You might find this interesting too

Helping lithium-ion battery recyclers move forward  
Uni researchers advance PV panel recycling
Bringing together ‘deep skills’

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Subscribe now and get a full year for just €169 (normal rate is €225) Subscribe