En un momento en que la economía circular exige materiales más puros y procesos más inteligentes, TOMRA Recycling amplía su plataforma de inteligencia artificial GAINnext con tres nuevas aplicaciones de aprendizaje profundo y presenta FINDER, un sistema modular de recuperación de metales. La empresa apuesta por que la visión artificial y los sensores integrados puedan resolver lo que la clasificación tradicional nunca logró del todo: distinguir lo valioso de lo descartable con precisión industrial. Es un paso más en la larga historia humana de aprender a ver valor donde otros solo ven residuo.
TOMRA expands AI-powered recycling with GAINnext deep learning apps and new FINDER platform
Sorting purity above 95 percent transforms recovery from a technical headache into viable business.
Mark
Why does sorting PET trays matter so much that it needed its own deep learning application?
Mimi
Because for years, recyclers couldn't reliably tell the difference between food-grade trays and medical-grade ones just by looking at them. The shape is similar, the material is the same. Without that distinction, you contaminate the feedstock, and suddenly your recovered material is worth less or unusable. Once you hit 95 percent purity, you've solved the contamination problem and opened a real market.
Mark
The aluminum can application processes material 33 times faster than manual sorting. That's a staggering number. What does that actually mean for a recycling facility?
Mimi
It means one machine can do the work of dozens of people, but more importantly, it means the material moves through the line at industrial speed without bottlenecks. Manual sorting is slow and inconsistent. A machine running at that capacity keeps the whole facility moving and maintains consistent purity—98 percent or better. That consistency is what makes the material valuable to mills.
Mark
FINDER is modular, which seems like a business decision as much as a technical one. Why design it that way?
Mimi
Because recycling facilities don't all process the same waste streams. One facility might focus on e-waste, another on end-of-life vehicles. A modular system lets each customer buy what they need and upgrade later without scrapping the whole machine. It also means TOMRA can add new sensor technologies as they develop them without forcing customers to replace working equipment.
Mark
The system is designed to be maintained more easily. Is that a response to something specific?
Mimi
Downtime is expensive. If a machine is hard to access internally, maintenance takes longer, and the facility loses sorting capacity while it's being serviced. Easier access means faster repairs and less lost productivity. It's a small design choice that compounds into real operational savings.
Mark
You mentioned that all sensors and software are developed in-house. Why is that important to say?
Mimi
Because it signals control and accountability. If you're buying sensors from multiple vendors and integrating them yourself, you're managing multiple relationships and troubleshooting becomes complicated. When one company owns the whole stack, they can guarantee performance and stand behind it. That matters when you're investing in industrial equipment.
Der Puls
- Las líneas de clasificación tradicionales han dejado durante años materiales valiosos sin recuperar, limitando tanto la rentabilidad del reciclaje como sus beneficios climáticos.
- La presión de la industria siderúrgica por chatarra más limpia y la demanda de aluminio en circuito cerrado han convertido la pureza del material en una urgencia económica y medioambiental.
- GAINnext ahora distingue bandejas de PET por forma y uso, identifica compuestos de cobre en flujos de acero, y procesa latas de aluminio 33 veces más rápido que la clasificación manual con una pureza superior al 98%.
- FINDER integra sensores electromagnéticos, NIR e IA en una arquitectura modular que permite actualizar componentes sin reemplazar el sistema completo, reduciendo costes de instalación y mantenimiento.
- La plataforma en la nube TOMRA Insight añade visibilidad continua del proceso, posicionando el sistema para la optimización constante que exige el reciclaje industrial moderno.
En un momento en que la economía circular exige materiales más puros y procesos más inteligentes, TOMRA Recycling amplía su plataforma de inteligencia artificial GAINnext con tres nuevas aplicaciones de aprendizaje profundo y presenta FINDER, un sistema modular de recuperación de metales. La empresa apuesta por que la visión artificial y los sensores integrados puedan resolver lo que la clasificación tradicional nunca logró del todo: distinguir lo valioso de lo descartable con precisión industrial. Es un paso más en la larga historia humana de aprender a ver valor donde otros solo ven residuo.
TOMRA Recycling ha anunciado tres nuevas aplicaciones de aprendizaje profundo para su plataforma GAINnext y ha presentado FINDER, un sistema rediseñado de recuperación de metales. Juntos, representan un intento de resolver limitaciones que la clasificación por sensores convencionales no ha podido superar.
La primera aplicación aprende a distinguir bandejas de PET para alimentos —envases de comida para llevar, bandejas de supermercado, embalajes médicos— por forma y uso, alcanzando una pureza superior al 95% y convirtiendo su recuperación en un modelo de negocio viable. La segunda identifica automáticamente compuestos de cobre mezclados con acero, mejorando la calidad de la chatarra para las acerías que buscan descarbonizarse. La tercera, adaptada de Norteamérica al mercado europeo, procesa latas de aluminio usadas 33 veces más rápido que la clasificación manual, con una pureza del 98% o superior.
FINDER, la nueva plataforma modular de clasificación de metales, combina un sensor electromagnético de nueva generación con tecnología NIR e IA integrada. Puede configurarse según las necesidades actuales del reciclador y ampliarse con nuevos sensores sin sustituir el sistema completo. Su estructura mecánica rediseñada facilita el mantenimiento, y todos los sensores y software son de desarrollo propio, según el director de tecnología de TOMRA.
El sistema puede conectarse con GAINnext para añadir clasificación por aprendizaje profundo, y con TOMRA Insight, una plataforma en la nube que ofrece visibilidad completa del flujo de materiales. TOMRA presentará estas innovaciones en SRR 2026, que se celebra del 9 al 11 de junio en Madrid, señalando que la apuesta por la clasificación inteligente es una dirección estratégica sostenida, no un anuncio puntual.
TOMRA Recycling is pushing deeper into artificial intelligence as a tool for sorting waste, announcing three new deep learning applications for its GAINnext platform and unveiling a redesigned metal recovery system called FINDER. The moves represent an attempt to solve problems that traditional sensor-based sorting has struggled with for years—problems that, until now, have limited what recyclers could actually recover and how pure the recovered material could be.
The first of the three new applications targets PET food trays, a material that has become increasingly valuable as both a feedstock and a source of confusion in sorting lines. By training GAINnext on thousands of images, the system learned to distinguish between takeout containers, supermarket trays, and medical-grade packaging based on shape and intended use. The result is sorting purity above 95 percent—a threshold that transforms PET tray recovery from a technical headache into a viable business model.
The second application addresses metals, specifically the challenge of identifying copper-bearing compounds mixed with steel. This matters because steelmakers pursuing decarbonization need cleaner scrap, and complex materials like electric motors have historically been difficult to separate from oxidized or dirty material streams. GAINnext can now identify these compounds automatically, improving scrap quality and its value as furnace feedstock—a direct contribution to the steel industry's climate goals.
The third application focuses on aluminum recovery from used beverage cans, or UBC. After initial deployment in North America, TOMRA adapted the system for European markets. The numbers are striking: the application processes material 33 times faster than manual sorting while achieving purity levels of 98 percent or higher. By immediately detecting and ejecting non-UBC materials, the system enables closed-loop aluminum recycling that is both more efficient and more automated.
Beyond the software, TOMRA introduced FINDER, a modular platform for metal sorting that integrates AI capabilities alongside multiple sensor technologies. The system is built around a new electromagnetic sensor that maximizes metal-to-non-metal separation, but it can be configured with additional sensors—including near-infrared technology—depending on what a recycler needs to handle. FINDER addresses non-ferrous metals like copper and brass, recovers electrical cables, cleans stainless steel, separates metals from non-metals, and sorts circuit board fragments from both end-of-life vehicles and electronic waste.
The architecture is intentionally flexible. Recyclers can configure FINDER to their current needs and add or upgrade sensors over time without replacing the entire system. Plug-and-play integration reduces installation costs and downtime. A redesigned mechanical structure gives technicians easier access to internal components, cutting maintenance time and operational expense. All sensors and software are developed in-house, according to Ralph Uepping, senior vice president and chief technology officer of TOMRA Recycling, which he frames as a guarantee of reliability and performance.
The system can also be paired with GAINnext, layering deep learning classification on top of the sensor-based sorting. A digital interface allows operators to manage multiple machines from a single dashboard, and an optional cloud platform called TOMRA Insight provides full visibility into process flow and material movement. This connectivity positions the technology for the kind of continuous optimization that industrial recycling increasingly demands.
TOMRA plans to showcase both FINDER and the new GAINnext applications at SRR 2026, a recycling industry conference running June 9-11 in Madrid. The company's Iberian leadership indicated that visitors will see these innovations alongside other AI-driven solutions, signaling that the push toward intelligent sorting is not a one-time announcement but an ongoing strategic direction.
Bemerkenswerte Zitate
A single modular system offering multiple sensor technologies allows customers to configure the system according to their operational needs, with the ability to add or update sensor technologies over time.— Ralph Uepping, senior vice president and chief technology officer, TOMRA Recycling
Möchten Sie die ganze Geschichte?
Das Original lesen bei Residuos Profesional ↗
Technology