AI ore sorting: TOMRA OBTAIN vs HPY Classic XRT vs MineSense

Choose where to separate value from waste

A mine can improve feed quality at several points between excavation and the mill. TOMRA OBTAIN, HPY Classic XRT, and MineSense ShovelSense use AI with physical sensors, but act at different scales. TOMRA and HPY address individual particles in a sorting circuit; MineSense measures material during digging to guide truck destinations. This comparison helps mineral-processing and grade-control teams choose the intervention that fits their operation.

TOMRA OBTAIN

TOMRA OBTAIN adds deep-learning recognition to mineral sorting on appropriate TOMRA equipment. Its neural network identifies individual particles, including material that appears clustered, so a sorter can make more precise detection and ejection decisions at high throughput. A plant can use that capability to pursue more capacity at a similar separation target or improve sorting precision at its existing flow rate. TOMRA positions OBTAIN for operating mines and material that previously appeared uneconomic, including low-grade stockpiles and waste dumps. Its wider mining portfolio supplies XRT and other sensor configurations for different minerals and recovery tasks. This makes OBTAIN relevant when the main bottleneck is recognizing and separating the right particles within a prepared feed. It is software within a physical sorting application, so compatibility, feed preparation, and the selected mineral matter. The product does not turn every TOMRA machine into the same AI system, and it is separate from TOMRA’s GAINnext recycling offering.

HPY Classic XRT

HPY Classic XRT combines single- or dual-energy X-ray detection with AI interpretation and high-speed air jets. X-rays produce a grayscale representation of each piece of material; the algorithm uses that information to distinguish target minerals from waste rock, and the jets perform the physical separation. The Classic product page gives a concrete model example with a 10–60 mm particle range and 30–60 tonnes per hour capacity. Those are model specifications, not a promise for every orebody. HPY’s wider technology description discusses customized sensor combinations and adaptive intelligence for changing grade and texture. Classic therefore suits a plant considering a defined XRT pre-concentration stage before grinding and flotation. Its appeal is the integrated recognition-and-separation machine rather than a standalone analytics dashboard. Buyers should match the selected configuration to actual mineral contrast and feed size; broader HPY portfolio features should not automatically be treated as included in a Classic installation.

MineSense ShovelSense

MineSense ShovelSense measures ore quality at the extraction face. High-speed X-ray fluorescence sensors mounted on shovels or loaders scan material while a bucket fills. Machine-learning algorithms developed for the orebody turn the spectrum into grade estimates and material classifications. Fleet-management integration can then direct a truck toward the mill, an ore stockpile, or waste, bringing the decision forward to the point where hauling begins. The bucket-level data also supports geological visibility, stockpile management, planning, and reconciliation. ShovelSense suits mines trying to reduce dilution or recover valuable material that a block-level designation would send to waste. It does not eject individual stones, so it addresses a different scale of sorting from TOMRA and HPY. MineSense’s separate BeltSense product can add conveyor measurements and downstream process-control data. That connected option expands the mine-to-mill view, but should not be assumed to be included with every ShovelSense deployment.

Particle separation versus truck routing

TOMRA OBTAIN and HPY Classic XRT are closest alternatives when a plant needs particle-level rejection from a prepared mineral stream. TOMRA emphasizes the recognition software’s precision/throughput tradeoff; HPY describes an integrated XRT machine with AI classification and air-jet sorting. MineSense instead changes where a load goes before the material reaches a sorting circuit. A mine with substantial grade variation inside mining blocks may benefit from ShovelSense’s earlier routing decision. A plant handling mixed particles after excavation needs the physical separation offered by TOMRA or HPY. These products can address successive stages, so choosing ShovelSense does not eliminate every reason for a downstream sorter.

The sensor determines the useful decision

HPY Classic XRT interprets X-ray transmission images, while ShovelSense uses X-ray fluorescence to estimate grade and relevant elemental characteristics. OBTAIN enhances recognition within a configured TOMRA sorting application. These are different information paths: particle contrast, bucket composition, and neural-network particle interpretation should not be collapsed into one generic AI accuracy score. HPY’s published model range gives an equipment-sizing starting point; MineSense’s ore-specific algorithms connect geology to routing; TOMRA’s application selection connects mineral properties to the sorter. For a processing engineer, the meaningful comparison is what each system can observe and where it can act on that observation.

Compare complete project scope

Numeric prices and billing intervals remain unconfirmed for all three offerings. TOMRA’s software and sorter configuration, HPY’s equipment and compressed-air requirements, and MineSense’s fleet integration create different project scopes. HPY’s published operating-cost or savings figures do not establish the purchase price of Classic XRT. Likewise, a TOMRA recovery example or MineSense downstream benefit is not a licensing rate. Compare the equipment, application software, installation, calibration, and service included in a proposal. A price for a recognition upgrade would not represent the same work as installing a new sorting machine or instrumenting the loading fleet.

Which should you choose?

Consider TOMRA OBTAIN for deep-learning particle recognition in a suitable sorting circuit, HPY Classic XRT for an integrated XRT pre-concentration machine, and MineSense ShovelSense for grade-based decisions at the extraction face. Start with the stage where waste enters your processing cost.