A UK-based client needed to understand the feasibility of using robotic sorting to pick aluminium.

They were most interested to separate cans from other aluminium items and wanted to compare the strength of AI  against other sorting technologies in identifying target items.

The Challenge

The company approached Recycleye to understand if AI could be used for two different applications for aluminium.

The first application was to more accurately identify foils, aerosols and other aluminium contaminants going through their eddy current from the cans, to ensure maximum purity could be achieved by the robot. This is referred to as negative picking of the aluminium QC stream.  The second application was to pick aluminium cans from a reject line, to ensure maximum value is extracted. This is known as positive picking.

Revenue was a key consideration, with a desire to effectively pick and optimize aluminium recyclates, given the comparatively high market value of the material.  Safety was also an important consideration, with fires and explosions of gas cannisters a major concern, and other identification technologies can struggle to identify and differentiate them from cans.

The client asked us to focus on the first application, for which the current stream input was 80% pure.


Purity increased +8%

Potential value +20%

Improved fire safety

Multiple class detection

The model performed well and was able to improve purity of the output on the incoming line.  In this line example, where input purity was 80%, our model increased aluminium can output purity to 93%, compared to the current sorting technology, which was producing 85%.

This result would have a significant impact on the value generated by this black bag MRF.  Based on achieving an average of around £700 per tonne at 15% contamination, compared to £850 for 5% output contamination, this result would mean an increase in output value of +20%.

With regards to safety, the AI detected cannisters, ensuring they could be removed before the baler. Given the importance of this feature, we’re developing automated alerts for gas cannisters and other dangerous objects such as batteries.

In addition, the model also accurately identified PET as a different material class, enabling the client to ensure they can extract the most value from other items in the waste line.

Identification and picking of aluminium cans

“Managing the value generated by a MRF must be a dynamic process. While the price of aluminium has continued to raise over the past few years, purity and safety are key considerations in separating cans from aerosols for many of our clients”.

This work demonstrated the ability of Recycleye to correctly identify aluminium cans to maximise recovery.”

– Zoe Cook, Technical Sales Manager

Knowing the volume of other valuable material classes in a waste stream gives the MRF manager the ability to respond dynamically to market conditions, using negative or positive picking, to extract maximum value from different categories in the sorted waste.

This work demonstrated the ability of Recycleye to correctly identify aluminium cans to maximise recovery.

  • 8% improvement in purity of offtake
  • Maximised recovery of high value recyclates (estimated at +20% for aluminium cans)
  • Reduced risk of fires
  • Supporting plant profitability