TOMRA Mining has introduced CONTAIN, a cutting-edge deep learning-based ore sorting solution designed specifically to tackle the challenges of sorting inclusion-type ores. Leveraging artificial intelligence, this new technology offers unprecedented accuracy and efficiency in classifying ores containing valuable minerals such as tungsten, nickel, and tin—even when inclusions are subtle and traditionally difficult to detect.
At the heart of CONTAIN is a sophisticated convolutional neural network (CNN) that enables real-time analysis of X-ray transmission (XRT) images. Unlike traditional sorting technologies that rely on basic particle data, CONTAIN recognises complex mineralogical patterns within rocks, assigning a probability score based on the likelihood of valuable ore inclusions. This advanced classification enables a more precise separation of ore from waste, significantly improving product grade while reducing dilution.
Intelligent Sorting for Complex Ores
Inclusion ores—where valuable minerals are embedded within host rocks—have traditionally posed a challenge for sensor-based sorting systems. However, TOMRA’s deep learning platform fundamentally changes the game. By training the system on tens of thousands of ore samples, TOMRA’s AI can detect visual cues imperceptible to conventional methods.
“Our system was built from the ground up to handle the complexity of inclusion-type ores,” explains Stefan Jürgensen, Software Team Lead at TOMRA Mining. “With CONTAIN, operators can dynamically adjust the grade-recovery threshold via an intuitive touchscreen interface, providing full control over yield and concentrate quality.”
Seamless Integration and Industry-Ready Design
Developed entirely in-house, CONTAIN integrates seamlessly with TOMRA’s existing mining sorters, making deployment simple and cost-effective for existing customers. The system is built to operate efficiently across a wide range of ore types and feed conditions—whether dealing with dense, high-speed input streams or large particle sizes of up to 65 mm.
Importantly, CONTAIN does not require specific throughput or spacing, a feature that makes it especially suited for the real-world demands of mining operations. Its ability to maintain high performance under challenging conditions means it can add value even when processing low-grade ores with complex mineral inclusion patterns.
Transforming Economic Viability
“Traditional technologies can be configured to detect low-grade material,” adds Jürgensen, “but often at the cost of product dilution, with large amounts of waste rock ending up in the concentrate stream. CONTAIN’s pinpoint accuracy makes it possible to economically sort ores that were previously uneconomical.”
This technology breakthrough is already making an impact. David Comtesse, Production Manager at Wolfram Bergbau- und Hütten AG, shared:
“We were absolutely overwhelmed by what CONTAIN could do. It identified inclusions we didn’t think were detectable and did so with incredible precision—even at grain sizes up to 65mm. It immediately changed the way we think about sorting and processing. This isn’t just an upgrade-it’s a completely new level of performance.”
Decades of Innovation
TOMRA has long been at the forefront of sensor-based sorting innovation. The company was the first to implement AI-driven image processing in mining back in 1993. In 2018, it established a dedicated deep learning team to accelerate the development of AI-enabled solutions. CONTAIN is the first commercial solution to emerge from this focused R&D effort.
Looking Ahead
With the introduction of CONTAIN, TOMRA Mining continues its mission to drive sustainability and efficiency in the mining industry. By enabling smarter, data-driven sorting, the technology helps reduce energy consumption, minimise water usage, and decrease waste—delivering both environmental and economic benefits.
Learn more about CONTAIN and TOMRA’s sensor-based sorting technologies at www.tomra.com/mining.
