XRT technology differentiates particles by atomic density, enabling precise separation of mineralised and host rock material at the individual particle level

Decision Lens

The core tension here is familiar: variable feed grade and mixed material sources erode plant efficiency unless waste is removed before it consumes energy in crushing and grinding. Soma Gold’s move to deploy Tomra Mining’s COM Tertiary XRT 1200 sorter at El Bagre is a direct response to that problem. The confirmed mechanism — dual-energy XRT sensing combined with deep learning classification — performs particle-level sorting even at high belt occupancy and across varying size fractions. The TS100 ejection modules add a second cost lever beyond throughput, reducing compressed air consumption by up to 70% compared to conventional systems. For operations directors evaluating pre-concentration options, this case provides a vendor-backed dataset from test-centre conditions, though in-situ production verification at scale remains pending.

90-Second Brief

In recent days, soma Gold Corp is installing a COM Tertiary XRT 1200 sorter at its El Bagre gold operation in Antioquia, Colombia, following benchmark test work conducted at Tomra’s Test Centre in Germany. The system uses dual-energy X-Ray Transmission sensing and AI algorithms, branded Obtain and Contain, to classify and eject non-valuable particles before they enter energy-intensive downstream stages. Test results confirmed consistent upgrade performance across multiple material types and size fractions. Implementation is being executed in partnership with Dismet, Tomra’s local integrator in Colombia.

What’s Actually Happening

El Bagre processes gold-bearing material from both Soma Gold’s own mines and third-party sources — a dual-feed model that creates feed variability as a structural condition, not an occasional problem. Conventional processing handles that variability by over-sizing throughput capacity; the sorting-first approach inverts that logic by narrowing what enters the circuit.

XRT technology differentiates particles by atomic density, enabling precise separation of mineralised and host rock material at the individual particle level. Tomra’s Obtain module applies deep learning to classify particles accurately even at high belt occupancy — a critical parameter for continuous, high-volume plant integration. Contain adds neural network detection for mineralisation embedded within host rock, targeting the classification errors that simpler optical or density methods miss.

The TS100 ejection modules reduce compressed air consumption by up to 70% compared to conventional ejection systems, directly lowering one of the main operating cost components in sorting circuits. Implementation is a joint project with Dismet, Tomra’s Colombian integrator, handling site-specific flowsheet integration at El Bagre.

Why It Matters for Mining Operations Directors?

Pre-concentration before crushing and grinding is not a new concept, but adoption at operating gold plants has lagged behind the technology’s maturity — largely because variable feed mineralogy has historically undermined sensor reliability. The El Bagre deployment addresses that directly: test work was conducted on representative Colombian material across size fractions and feed types, including challenging small-particle scenarios and dense material flows.

For operations directors managing similar constraints — mixed ore sources, variable head grades, energy-intensive comminution circuits — the operational logic is transferable. Removing waste before the mill reduces per-tonne energy consumption, extends liner and media life, and stabilises feed grade to flotation or leach circuits. Better feed consistency also reduces reagent overconsumption and improves recovery predictability.

The compressed air reduction is not peripheral. Compressed air is a significant fixed-cost infrastructure element at many underground and remote operations, and a 70% reduction in ejection air demand lowers both operating cost and maintenance burden on compressor capacity — relevant wherever compressed air is already a constraint.

The Forward View

Sensor-based sorting is moving from pilot-scale curiosity to production-flowsheet integration at smaller and mid-tier gold operations. If Soma Gold’s El Bagre installation delivers on the test-work upgrade ratios under continuous production conditions, it will add a datapoint that accelerates consideration at larger operations processing marginal or transitional ore.

The broader signal is directional: AI-enhanced XRT sorting is now being paired with local integration partners in jurisdictions like Colombia — a market historically underserved by advanced processing technology vendors. That commercial maturation matters for operations directors in Latin America and similar jurisdictions where technology access and local technical support have previously been barriers.

For processing plant managers evaluating comminution circuit bottlenecks, the next decision window is likely tied to crusher or mill relining shutdowns — when reconfiguring feed preparation is least disruptive. Operations with mixed ore sources or third-party feed agreements should evaluate whether pre-concentration addresses a structural feed variability problem, not merely a throughput one.

What We’re Uncertain About?

  • Production-scale performance versus test-centre results. The upgrade ratios and sorting consistency confirmed at Tomra’s facility used representative samples under modelled conditions. Whether those results translate without degradation into continuous plant operation at El Bagre — with real-time feed variability — has not yet been reported. Resolution requires operational production data post-commissioning.

  • Quantified upgrade ratio and recovery trade-off. The source confirms improvement in mineralised material upgrading but does not disclose specific upgrade factors, mass yield to concentrate, or gold recovery in the reject stream. Without those numbers, comparative assessment against alternative pre-concentration methods is not possible from available evidence.

  • Economic threshold for applicability. The cost-benefit case depends on site-specific energy costs, ore variability, and capital cost of sorter integration. The El Bagre case does not provide a payback horizon or AISC impact figure. Operations directors would need those inputs to assess applicability to their own asset.

  • Commissioning timeline. The sorter order has been placed and the project is in implementation phase, but no commissioning date or ramp-up schedule is confirmed in available reporting.

One Question to Bring to Your Team

If you removed the bottom quartile of your feed — by atomic density — before it reached the crusher, what would that do to your energy cost per tonne processed, your mill availability, and your reagent consumption in the back end of the circuit?

Sources

  • Engineerlive — Tomra Mining’s sensor-based sorting used to advance mineralised material upgrading in Colombia (Link)

Tags