Tungsten West plc is deploying six TOMRA Mining XRT ore sorters equipped with OBTAIN™ Deep Learning at the Hemerdon tungsten mine in Devon, England, as part of a redesigned processing plant targeting full-scale production in the first quarter of 2027.

The new sorting house will process material from the primary and secondary crushing circuit in the 10–80 mm size range, using sensor-based sorting to separate mineralised material before it enters the downstream processing stages.

Three processing fractions

After secondary crushing, a wet screen separates the material into three fractions:

• Below 10 mm — sent directly to pressure jigging and dense media separation.
• 10–30 mm — processed by three COM Tertiary XRT machines.
• 30–80 mm — processed by three COM XRT 2.0 units.

The system is designed to send approximately 30% of the incoming material, representing the mineralised fraction, to the downstream concentrator for tungsten and tin recovery. The remaining approximately 70% is separated as a premium aggregate by-product for the construction industry.

OBTAIN™ Deep Learning

OBTAIN uses Deep Learning and a neural network to identify the properties of individual particles during high-throughput XRT sorting. One challenge in high-volume ore sorting is that particles can arrive in clusters. When this happens, sensors can interpret several particles as a single object, making precise ejection decisions more difficult.

OBTAIN is designed to break down these clusters algorithmically and make sorting decisions at the individual-particle level, even at high feed rates.

According to TOMRA Mining, the technology allows each sorter to process around 80% more material while maintaining the same quality and recovery rates.

“With OBTAIN, we can feed each sorter with about 80% more material and have the same quality and the same recovery rates as before. In short, we can simply run double the capacity.”

— Jens-Michael Bergmann, Global Segment Manager Industrial Minerals, TOMRA Mining

Less material through the full processing circuit

At Hemerdon, tungsten mineralisation occurs primarily as discrete inclusions within the host rock, making particle-level detection important for recovery and waste separation. By removing approximately 70% of the mined material before downstream crushing, grinding, milling, flotation and jigging, the system is designed to reduce the amount of material requiring further processing.

Installation of all six TOMRA Mining sorters is scheduled through Q4 2026, with Hemerdon targeting full production from the end of Q1 2027.

#Tungsten #OreSorting #XRT #TOMRA #DeepLearning #AI #MiningTechnology #CriticalMinerals #MineralProcessing #SortingTechnology #CircularEconomy #CircularEconomyNews