AI-based Sorting : TOMRA introduces deep learning solution to enhance wrought aluminum scrap recycling

This development represents the first application of GAINnext™ - TOMRA's revolutionary deep learning based sorting technology - in the metals industry. This AI innovation enables aluminium scrap processors and recyclers to upgrade their wrought aluminium scrap and produce exceptionally high purity fractions by instantly recovering low alloy castings from the wrought fraction and reducing alloying elements such as silicon. GAINnext™ complements TOMRA's X-TRACT™, which separates based on atomic density. 

TOMRA's X-TRACT™ with X-Ray Transmission (XRT) technology is the industry standard for sorting aluminium from heavy metals based on atomic density. X-TRACT™ first sorts shredded mixed non-ferrous metals (Zorba) to produce high purity aluminium scrap (Twitch), then removes high alloy cast aluminium and high density wrought aluminium to further refine the Twitch. This process produces a high quality fraction containing both wrought aluminium and small amounts of low alloy cast aluminium, which other currently available sorting methods cannot fully separate. 

Now, by integrating GAINnext™ into the sorting process after X-TRACT™, a pure forged fraction can be achieved. GAINnext™ harnesses the power of deep learning and artificial neural networks. Using RGB cameras, it rapidly processes tens to hundreds of thousands of images per millisecond, 'sensing' and 'classifying' materials by shape, size and dimension with exceptional accuracy. This advanced system, trained over years by TOMRA's experts, mimics human vision but with superior speed, enabling high throughput sorting of up to 2,000 rejects per minute. In its first application in the metals industry, GAINnext™, by accurately detecting and removing low alloy castings from the wrought fraction produced by TOMRA's X-TRACT™, delivers an exceptionally high purity wrought product that commands premium market prices.

For TOMRA customers seeking even higher levels of granularity in aluminium sorting, the wrought fraction recovered by GAINnext™ can be further refined using TOMRA's AUTOSORT™ PULSE system, which uses dynamic laser induced plasma spectroscopy (Dynamic LIBS) to accurately identify and sort different alloy types. TOMRA's breakthrough innovation, AUTOSORT™ PULSE, which will be launched in 2023, uses precise elemental composition analysis and advanced dynamic laser detection to accurately distinguish between different alloy types, such as 5xxx and 6xxx aluminium. 

This combination of X-TRACT™, GAINnext™ and AUTOSORT™ PULSE represents the pinnacle of sorting innovation, delivering the highest level of material separation available on the market today. 

Tom Jansen, Head of the Metals Segment, Sales at TOMRA Recycling, comments: “We are excited to launch this new deep learning-based application to upgrade wrought aluminum. Deep learning is a powerful addition to our traditional sensor-based technology, and expanding our GAINnext™ ecosystem aligns with our broader AI strategy. Building on our proven success and the significant results our customers have achieved with GAINnext™ in the waste segment, we are confident in delivering similar value to our metals customers.

“This latest application enhances dry mechanical sorting of mixed metals, offering a cost-effective solution with a low cost per ton thanks to its high throughput and significantly reduced manual sorting requirements. Our customers will be able to consistently produce smelter-ready, premium-grade aluminum with exceptional purity levels. This is a significant enhancement to our metal recycling solutions portfolio, and we anticipate wide adoption by recyclers seeking to refine scrap and generate high-quality feedstock for circular manufacturing.”

TOMRA pioneered deep learning technology in 2019, and GAINnext™ has been proven in the field across various complex recycling streams, including waste, plastics, paper, metals and wood. In 2024, TOMRA introduced a groundbreaking GAINnextTM application to separate food-grade from non-food-grade plastics for PET, PP and HDPE. This latest expansion of the GAINnextTM ecosystem is part of TOMRA’s broader AI strategy. Further GAINnext™ applications are due to be announced in 2025.