TOMRA Mining Adds Deep Learning System to Sorting Line

TOMRA Mining has introduced CONTAIN™, a deep learning-based ore classification system designed to improve the sorting of inclusion-type ores. The company says the system uses real-time analysis of X-ray imagery to detect complex mineralizations such as tungsten, nickel and tin that may be missed by conventional sorting methods. CONTAIN™ is built to integrate with TOMRA’s existing platform, including its COM XRT and OBTAIN™ technologies, allowing operators to fine-tune sorting performance and recovery rates across a wide range of ore types.

The technology uses convolutional neural networks trained on tens of thousands of ore samples, according to TOMRA. This enables the system to assign probability scores to each rock based on the likelihood of subsurface mineral content, helping operators reduce losses and maintain consistent product quality. Trials at Wolfram Bergbau in Austria reportedly resulted in increased throughput, lower tails grade, and a 33% reduction in ore mineral losses.

“CONTAIN is exceptionally accurate in evaluating the value of a rock, making sorting thresholds for such relatively low-grade ores economically viable,” said Stefan Jürgensen, Software Team Lead at TOMRA Mining.

For more information