Artificial intelligence applied to metal recycling It has gone from being an almost futuristic concept to becoming a very real and, above all, profitable tool for scrap yards, treatment plants, and large steel groups. In just a few years, AI, robotics, and sensors have infiltrated sorting lines to improve performance, reduce risks, and get the most out of every kilogram of scrap metal.
Far from being a luxury reserved for large companies, AI-based solutions are now within reach of SMEs and medium-sized plantsThanks to leasing models, pay-per-use options, and agreements with technology providers who share the investment risk, the sector is undergoing a true silent revolution. Simultaneously, R&D projects like CIRIAMET and cutting-edge developments from companies such as TOMRA, AMP Robotics, and machine vision solutions for recycling plants demonstrate that the sector is experiencing a genuine, albeit quiet, revolution.
AI is now fully integrated into metal waste management
At the BIR World Recycling Convention held in Bangkok, the Ferrous Metals Division delivered a clear message: Artificial intelligence is transforming the management of scrap metal And it's no longer the preserve of a select few. According to its president, steel recyclers are characterized by being ahead of the curve, innovating, and adapting, and AI fits perfectly into that mindset.
During the session it was emphasized that these technologies allow detect hidden risks such as lithium-ion batteries Before they cause fires or explosions, improve plant safety, refine internal logistics, and strengthen the circular economy of steel and other metals. In an environment with volatile prices and unpredictable incoming flows, the ability to "see" and better understand incoming material has become critical.
The speakers emphasized that AI is becoming established as a strategic tool in all types of plantsFrom large, highly automated facilities to family businesses managing modest volumes but seeking greater accuracy and traceability, the overall message was clear: those who don't get on board will be left behind.
One of its great advantages is its cross-cutting applicability: From advanced sorting of ferrous and non-ferrous scrap metal, to collection planning and energy optimizationIn addition, there is the possibility of integrating AI with X-ray systems, hyperspectral vision, spectroscopy or blockchain, configuring authentic “smart factories” for metal recovery.
Furthermore, experts from major recycling companies highlighted how AI contributes to to improve the final quality of the metallic fractionsopening doors to more demanding markets and reducing "downcycling", that is, the loss of value of recycled raw materials compared to virgin ones.
From materials intelligence to advanced automatic sorting
One of the most powerful technological lines is the so-called “materials intelligence”This technology combines AI with X-ray sensors or other analytical systems to understand the composition of scrap metal in real time. Companies like the US-based Visia have developed equipment that scans incoming streams to identify everything from alloys to hazardous components.
One of the biggest headaches for the sector is that The flow of material arriving at the plants is extremely variableEach truck can carry very different mixtures, with hidden batteries, parts with complex coatings, or remnants of electronic components. This volatility makes planning difficult, affects quality, and multiplies risks.
Thanks to sensors and AI algorithms, these systems are able to detect up to dozens of different types of batteriesestimate their chemical composition and risk level, and automatically separate them from the main flow. In some cases, detection accuracy approaches 97%, significantly reducing incidents and safety shutdowns.
In Europe, projects like CIRIAMET go a step further by applying computer vision, spectroscopic analysis, and machine learning and deep learning models to classify complex metal scrap, mainly from end-of-life vehicles and lithium-ion batteries. The aim is to obtain high-purity concentrates of critical metals that can be reintegrated into high value-added applications.
This type of research lays the foundation for more efficient and circular metallurgical processesThey are capable of working with high-quality secondary raw materials instead of poorly controlled mixtures that only allow for lower-value uses. In practice, this means transforming complex waste into resources almost tailored to industry's needs.
Leading projects: the CIRIAMET case and the new scrap metal of electric vehicles
The CIRIAMET project, funded under the Basque Government's ELKARTEK program, focuses on an increasingly urgent challenge: How to identify, classify and recover valuable metals in the recycling of hybrid and electric vehiclesThe vehicle fleet is changing, and with it, so is the scrap metal arriving at the plants.
The metallic fractions resulting from the treatment of these vehicles They are not homogeneous from a chemical point of viewThey are made by mixing different aluminum alloys, special steels, and components with complex coatings or treatments. If melted as is, the result is usually a metal that does not meet the quality specifications required by the industry for advanced applications.
To solve this, CIRIAMET proposes a two-stage flow: first, the Automatic sorting of complex scrap metal using machine vision, spectroscopy and AIThen, it undergoes automated separation to obtain high-purity streams specific to each type of metal or alloy. This generates new value-added streams that can be used in cutting-edge sectors.
The GAIKER technology center plays a key role in the project, with lines of work that include the Automatic classification of aluminum alloys into non-ferrous fragments, the detection of unsuitable materials on conveyor belts as a preliminary step to their robotic removal and the intelligent location of joining elements in lithium-ion batteries to facilitate automated disassembly.
At the same time, the consortium is assessing the impact that these technologies may have on circularity, environmental footprint and sustainability of the entire value chainFrom the collection of end-of-life vehicles (ELVs) to the manufacture of new products using recovered metals, the process is a comprehensive one. The use of hyperspectral imaging and laser-induced plasma spectroscopy (LIBS), combined with deep learning models, enables highly precise identification of each target material.
Smart recycling plants and the deep learning revolution
In many metal recycling plants, machines have long been performing tasks that were previously purely manual: classification, screening, cutting, pressing or mechanical separationWhat is changing now is the "brain" that coordinates all those teams, thanks to deep learning and real-time data analytics.
Deep learning is based on multi-layered artificial neural networks These systems are capable of identifying highly complex patterns in images, sensor signals, or process data. The more difficult the classification task, the more layers and data are needed, but also the greater the accuracy that can be achieved without changing the plant's hardware.
In practice, this technology allows machine vision systems Recognize metals by type, color, shape, size, or textureadapting over time to variations in input material. Instead of replacing equipment, many improvements are implemented through software updates that add new capabilities to existing lines.
Deep learning also drives the creation of new material flows that previously could not be separated with sufficient qualityThis opens up business opportunities for higher-value products and more demanding markets. This is especially relevant for strategic metals and light aluminum alloys, where a few extra points of purity can make a significant difference in price.
At the same time, the introduction of these technologies is generating a change in the labor structure of the sector: Fewer repetitive manual sorting tasks and more profiles linked to data analysis, advanced maintenance, and monitoring of automated processesFar from "eliminating" human labor, it redirects it towards higher value-added activities.
Robotics, advanced sensors, and high-precision separation
Robotics has moved beyond science fiction and become an everyday ally in metal recycling plants. The new systems combine High-precision robotic arms with AI-guided cameras and sensors, capable of identifying and collecting specific parts on very fast conveyor belts and demanding environments.
This automation not only increases productivity, but also It reduces the exposure of operators to hazardous materials, dust, loud noises, and potential explosions.In the case of lithium-ion batteries, for example, some plants use self-learning smart cameras that detect these batteries and expel them into containers with sand, minimizing the risk of fires.
Beyond robotics, sensor technology has made a huge leap in recent years: high-power magnets to capture ferrous metals, eddy current separators to recover aluminum and other non-ferrous metalsNIR sensors, LIBS spectroscopy and X-ray transmission (XRT) to differentiate alloys by their composition or atomic density.
In the case of aluminum, very high recovery rates have been achieved; some eddy current systems are capable of to recover up to around 95% of the aluminum present in certain waste fractionsThis significantly improves the plant's economic return and reduces the pressure on primary extraction.
These advances are integrated into authentic Smart recycling plants connected via the Internet of Things (IoT) and big data platformsSensors distributed throughout the line monitor in real time the performance, energy consumption, stops, purity of each flow and other key parameters, facilitating fine adjustments and tactical decisions almost instantly.
The TOMRA ecosystem: GAINnext, X-TRACT and AUTOSORT PULSE
One of the most advanced examples of AI application in metals is the TOMRA Recycling ecosystem, which has expanded its portfolio with a deep learning-based solution called GAINnext for refining aluminum scrap profiles. This is the first time this technology has been applied specifically to the metallurgical industry on this scale.
In practice, GAINnext takes care of Clean the aluminum profile fraction by removing the lightweight aluminum crankcaseThis component, which typically has a high silicon content, can compromise the quality of the final alloy if not removed. The result is a cast-ready aluminum profile with very high purity levels and reduced silicon content, which commands higher prices on the market.
The system is combined with X-TRACT, TOMRA's solution that uses XRT technology to separate aluminum from heavy metals by atomic densityFirst, the zorba (crushed mixture of non-ferrous metals) is processed to produce a high-purity aluminum scrap known as "twitch"; then, it is further refined by removing high-alloy crankcase aluminum and high-density profile aluminum.
Once the profile fraction is obtained, the GAINnext integration allows to detect and eject with great precision the remaining lightweight crankcase Thanks to the analysis of RGB images and neural networks trained over years, the system can process hundreds of thousands of images per millisecond and perform up to 2,000 ejections per minute, mimicking human vision but at a speed unattainable for a human operator.
For those who need to raise the bar even higher, the aluminum profile fraction can undergo an additional stage with AUTOSORT PULSE, which applies Dynamic LIBS to differentiate between alloy series such as 5xxx, 6xxx and othersand even perform intra-alloy sorting. This opens the door to avoiding downcycling and recovering scrap metal suitable for demanding applications such as automotive or construction.
Digitization, predictive maintenance and traceability
The introduction of AI is not limited to the physical classification of materials. It is becoming increasingly important in... global digitization of recycling plantswith systems that collect and analyze data to improve efficiency, plan maintenance and ensure traceability.
Platforms like Tomra Insight allow remotely monitor the performance of sorting machinesVisualizing key metrics on dashboards accessible from a computer or mobile device makes it easier to detect deviations in purity, flow rates, or energy consumption, and to anticipate minor issues before they become costly breakdowns.
Predictive maintenance relies on sensors, cameras, and algorithms that They learn to recognize wear patterns and abnormal conditionsEarly warnings help schedule planned shutdowns, extend the lifespan of equipment, and cut operating costs associated with unforeseen failures.
In parallel, many companies are making progress in digital traceability and comprehensive documentation of the waste flowThis responds both to the demand for transparency from industrial customers and to stricter regulatory requirements regarding waste management, recycled content and environmental footprint.
Blockchain applications are even being explored for create immutable records of the origin, treatment and final destination of recycled metalsreducing fraud in the supply chain and facilitating the verification of environmental and responsible origin certifications.
Economic accessibility, SMEs and business models
One of the recurring fears in the sector is that Investment in AI and robotics will put smaller companies at a disadvantageHowever, more and more technology providers are opting for flexible models, aware that SMEs represent an essential part of the recycling ecosystem.
In practice, formulas like the following are becoming more widespread: equipment leasing, payment per ton processed, or strategic alliances In these arrangements, the supplier assumes part of the risk and shares the benefit obtained through improved classification. This makes it easier for small and medium-sized plants to access cutting-edge technology without an unaffordable initial outlay.
Industry experts point out that SMEs have very interesting niches that they can exploit They combine their practical knowledge of the material with well-tailored AI solutions. The key is not to try to replicate the model of large plants, but to find specific applications where the technology makes a clear difference in quality, safety, or profit margin.
Furthermore, the return on investment is usually relatively quick: studies and real-world cases point to recovery periods of between 12 and 24 months thanks to greater automation, reduced risks, less need for manual sorting, and the creation of higher value products.
All of this reinforces the idea that AI is not just an “extra” or a technological embellishment, but a very powerful lever to improve competitiveness even in tight margin marketsespecially when combined with a good business strategy and rigorous operational management.
Environmental impact, energy and circular economy
Beyond the economic benefit, AI has an important role to play in reduce the environmental impact of metal recyclingBy improving the accuracy of sorting, recovery rates increase and the amount of waste that ends up in landfills or in energy recovery decreases.
Less diverted material and more high-quality scrap mean less need to extract virgin ore and lower energy consumption associated with primary productionSince recycled metals typically require much less energy than those produced from ore, every ton recovered correctly translates into a considerable reduction in greenhouse gas emissions.
AI also helps to optimize the energy consumption of the plants themselvesby automatically adjusting the operation of motors, conveyors, shredders, or sorting systems based on the actual load and process conditions. This reduces costs and the facility's carbon footprint.
At the urban and regional level, predictive analytics systems allow to better estimate the generation of metal waste in specific areashelping authorities and concessionary companies to size infrastructures, collection routes and more effective management policies.
All these innovations contribute to strengthening the circular economy of metals, closing the cycle from product design to its reincorporation as secondary raw materialAI acts as a common thread that connects data, processes, and decisions, giving coherence to the system.
Employment, professional profiles and social perception of AI
One issue that raises many questions is the effect of AI on employment within the metal recycling sector. What is observed in practice is a reconfiguration of functions rather than a massive destruction of jobsCertain repetitive manual sorting tasks tend to decrease, but others related to supervision, quality control, and data management increase.
Operators who previously only separated parts can now to leverage the information generated by AI systems to make better-informed decisions, identify optimization opportunities, adjust process parameters or detect deviations in the purity of the fractions.
On the other hand, the demand for Technical profiles focused on advanced equipment maintenance, data analysis, systems integration, and algorithm developmentThe challenge lies in facilitating training and professional development for the workforce so that they can take advantage of these new opportunities.
On a cultural level, it is evident that Science fiction has fueled certain fears about AIpresenting it as an almost autonomous force that could get out of control. However, in the context of industrial recycling, the reality is much simpler: these are tools designed to assist human teams, not to completely replace them.
The key is to maintain a responsible approach, with constant supervision and clear rules about which decisions are automated and which are reserved for people. As long as these principles are respected, AI can be seen more as an ally than a threat to the sector.
The combination of artificial intelligence, robotics, advanced sensors, and digitalization is redefining metal recycling, offering Safer, more efficient and sustainable plants, capable of extracting the maximum value from each metallic waste.The speed of change is high, but those who know how to integrate these tools judiciously will have a very solid position in the new circular economy scenario that is consolidating.