Sensor-based ore sorting (SBS) has moved from a niche curiosity to a genuinely disruptive force in modern mineral processing. By individually examining each rock particle with a contactless sensor and physically ejecting waste before it enters the mill, sorting technology allows mines to concentrate only on material that carries value. The result is lower energy consumption, reduced water use, smaller tailings volumes, and often a step-change improvement in head grade to the comminution circuit. This guide explains every aspect of sensor-based ore sorting: the physics and engineering behind each sensor type, how a commercial installation is configured, the economic case for adoption, which ore types benefit most, and the frontiers that are likely to shape the next decade of development. Whether you are a metallurgist evaluating a new project, an engineer optimising an existing flowsheet, or a student building foundational knowledge, this complete reference will take you from first principles to industrial practice.
What Is Ore Sorting?
Ore sorting is the automated process of examining individual rock particles, making a real-time classification decision based on a measured physical or chemical property, and then physically separating the particles into at least two streams: a valuable fraction that proceeds to downstream processing and a rejected waste fraction that is discarded. In its broadest definition the term has existed since Georgius Agricola documented hand sorting in De Re Metallica in 1556, and the economic logic has barely changed: why crush, grind, float, and leach rock that carries no value?
Modern sensor-based sorting extends that ancient practice by replacing the human eye and hand with electronic sensors, high-speed data processors, and pneumatic or mechanical ejection systems. Wotruba and Harbeck (2012) define SBS as any application where particles are singularly detected by a sensor technique and ejected by an amplified mechanical, hydraulic, or pneumatic process. The key qualifier is “singularly”: unlike bulk-stream techniques such as dense-media separation or flotation, a sorter measures properties particle by particle, which is both its great strength and its throughput limitation.
Within a mineral processing flowsheet, SBS can be inserted at several logical positions. Pre-concentration removes a fraction of high-grade coarse particles from run-of-mine (ROM) material before fine grinding, typically treating particles coarser than about 5 mm. Waste rejection removes clearly barren rock before any further beneficiation. Concentration produces a directly saleable product, common in industrial minerals and diamonds. Ore-type diversion routes different lithological domains to different plant lines for specialised treatment. Each position offers a different economic return, and the most compelling case is almost always pre-concentration, where eliminating even 20–30% of the ROM feed as barren waste can dramatically reduce the unit cost of production downstream (Wills and Finch, 2016).
Sensor Technologies: XRT, Optical, Laser, Near-Infrared, and Electromagnetic
The choice of sensor is the single most important technical decision in designing an ore-sorting installation. Each sensor interacts with a different physical or chemical property of the rock, and the suitability of each technology is entirely determined by whether a detectable contrast exists between valuable mineral and gangue in that specific property.
X-Ray Transmission (XRT)
X-ray transmission sorters pass a fan-shaped X-ray beam through the falling or belt-borne rock and measure the attenuation of that beam on the far side. Attenuation is dominated by the atomic density of the material: heavy elements such as lead, zinc, copper, and tin attenuate hard X-rays far more strongly than light elements such as silicon, aluminium, and calcium. XRT therefore provides a signal that is tightly correlated with elemental composition without requiring contact with the rock surface. High-resolution linear detector arrays can produce a two-dimensional atomic-density map of each particle, allowing the processing algorithm to reject or accept based on a threshold density or a spatial pattern.
XRT has become arguably the most widely deployed sorting technology in the mining industry. Its strength is that it sees through surface coatings, dust, and moisture, all of which blind optical sensors. At the Mittersill scheelite mine in Austria, XRT sorters reject approximately 25% of the 130 t/h ROM stream as barren waste before milling, with feed grades as low as 0.03% WO3 (Mosser and Gruber, 2010, as cited in Wills and Finch, 2016). In diamond recovery, XRT is now the dominant final-recovery technology at many kimberlite operations, delivering single-stage recoveries of 96–98% of diamonds from pre-concentrated DMS product (TOMRA, 2020).
X-Ray Fluorescence (XRF)
XRF excites elements within the rock surface with a primary X-ray beam; the elements respond by emitting characteristic fluorescent X-rays whose energies identify the elements present. Unlike XRT, which measures bulk attenuation, XRF provides direct elemental analysis of the particle surface, detecting elements with an atomic number above about 20. The RADOS XRF chute-type sorter has operated at more than 49 sites across over 20 commodities, primarily in Russia, separating copper-zinc ores and other polymetallic deposits where surface elemental content is a reliable proxy for particle value (Wills and Finch, 2016). The system requires feed grades above about 0.1 wt% for effective detection of most target elements.
X-Ray Luminescence (XRL)
Diamonds luminesce when irradiated by X-rays, a property that is far more consistent and reliable than the natural hydrophobicity exploited by the older grease-belt recovery method. XRL sorters have been used in diamond operations since the 1960s and now constitute the standard final-recovery stage at most kimberlite and alluvial operations worldwide. Modern wet and dry XRL machines operate in multi-stage circuits to ensure very high diamond recoveries while rejecting waste efficiently.
Optical and Colour Sorting
Colour sorters measure the reflectance of the rock surface across the visible spectrum, or a selected portion of it, using line-scan cameras or laser-based photomultipliers. The technology is the oldest form of automated sorting and remains competitive wherever the valuable mineral or the gangue has a distinctive colour or surface texture. Applications include magnesite, limestone, talc, feldspar, phosphate, and several gold-bearing ore types where barren quartz appears distinctively lighter or darker than mineralised rock. The main limitation is sensitivity to surface contamination: fine clay coatings or moisture can mask the underlying colour signal, which is why particle washing before the sorting stage is sometimes required.
Near-Infrared (NIR) Spectrometry
NIR sensors measure the reflectance or absorption of radiation in the near-infrared wavelength range, producing a spectral fingerprint that is sensitive to the vibrational modes of specific molecular bonds, particularly hydroxyl groups in clay, carbonate, and silicate minerals. NIR has been successfully applied to waste elimination from boron minerals (colemanite and ulexite), talc-carbonate separation, and porphyry copper ore characterisation (Dalm et al., 2014, as cited in Wills and Finch, 2016). Its chief advantage is specificity: NIR can distinguish mineralogical varieties that look identical in colour.
Electromagnetic Induction
Electromagnetic (EM) sensors detect the conductivity and magnetic susceptibility of rock particles by measuring the phase shift and amplitude change induced in a tuned coil beneath the belt. Conductive sulfide minerals such as pyrite, chalcopyrite, and galena produce strong EM responses even at modest liberation. The RTZ Model 19 sorter, which processed 25–150 mm rocks at up to 120 t/h, demonstrated the commercial viability of this approach for sulfide-bearing base-metal ores (Wills and Finch, 2016). EM sensors can be combined with optical sensors to improve discrimination in mineralogically complex feeds.
Radiometric Sorting
Radioactive minerals, principally uranium-bearing ores, emit natural gamma radiation that can be detected by scintillation counters beneath or around the conveyor belt. Radiometric sorting has a long history of pre-concentrating uranium ore in Australia, Namibia, South Africa, and Canada, and was the basis for early industrial sorters in the 1940s and 1950s. The technology remains highly relevant for uranium operations seeking to upgrade ROM ore before leaching.
How an Ore Sorter Works: The Four Subsystems
Regardless of the specific sensor technology, every commercial ore sorter is built around four fundamental subsystems: particle presentation, sensing, electronic processing, and separation. Understanding each subsystem is essential for specifying, commissioning, and optimising a sorter.
Particle Presentation
Individual particle measurement requires that each rock be presented to the sensor in a reproducible, non-overlapping manner. Two engineering architectures are commercially dominant. In the belt-type system, a conveyor belt carries a monolayer of particles past the sensor head at a controlled speed, typically 1–3 m/s. The belt system allows the sensor to be mounted above or below the material and is well suited to belt-mounted detectors such as XRT linear arrays and NIR spectrometers. In the chute-type system, particles free-fall from a high-incline chute past the sensor zone; they are already travelling as individual projectiles, which simplifies ejection geometry. The chute system is widely used for XRF, XRL, and radiometric applications.
The feed preparation upstream of the sorter is critical. Material must be screened to a controlled size range; the ratio of maximum to minimum particle size (the size range coefficient) should typically not exceed three, because a wide size range degrades separation efficiency for all ejection technologies. The technically feasible upper size limit is around 350 mm, but economic liberation of barren waste is usually achieved below 100 mm. The lower economic size limit is approximately 10–20 mm because operating costs are inversely proportional to average particle size and weight: treating 10 mm particles requires roughly ten times more machine capacity per tonne of product than treating 100 mm particles (Wills and Finch, 2016).
Sensing
The sensor performs a real-time, contactless measurement of a physical or chemical property of each particle as it passes through the detection zone. Measurement speed is critical: a particle travelling at 2 m/s on a belt occupies the sensor field of view for only a few milliseconds, so the sensor and its associated electronics must acquire and transmit a measurement in that window. Modern line-scan cameras operating at several thousand lines per second and XRT detector arrays with microsecond response times are well matched to these demands.
Electronic Processing
The raw sensor signal is processed by algorithms that convert the measurement into a classification: accept or reject, and in more sophisticated systems, an estimated grade or probability of value. Site-specific algorithms are tuned during commissioning using representative ore samples. Machine learning approaches are increasingly used to handle multi-sensor data fusion, where two or more sensor channels are combined to improve discrimination accuracy. The location and velocity of each detected particle is tracked from the sensing zone to the ejection zone, which introduces a time delay that the control system must compensate for accurately.
Separation
Physical separation is achieved by selectively deflecting target particles from their natural trajectory. The dominant technology is a linear array of approximately 200 high-speed pneumatic air valves, each individually addressable, that fire a pulse of compressed air to deflect a particle into the reject bin; undeflected particles fall into the accept bin, or vice versa. Air valve arrays achieve ejection response times of less than 1 ms and allow selective treatment of individual particles even at high belt speeds. For low-throughput single-particle applications such as XRL diamond sorters, mechanical ejectors are preferred. Water jets have been evaluated but have not achieved commercial adoption in hard-rock mining applications (Wills and Finch, 2016).
Economics and Grade-Recovery Benefits
The economic case for sensor-based ore sorting rests on several interacting mechanisms. When barren waste is removed before milling, the mass flow to the comminution and concentration circuits decreases, allowing either a reduction in installed capacity or an increase in throughput. Every tonne of waste that is not ground, floated, or leached avoids the associated energy, water, reagent, and labour costs. Capital and operating costs of SBS installations are typically roughly half those of comparable dense-media separation circuits, partly because SBS operates dry and avoids the dense-media recovery and regeneration infrastructure (Wills and Finch, 2016).
The highest economic benefit accrues when the sorted waste fraction carries a grade close to zero and the sort is performed at coarse particle size, maximising both the mass rejection ratio and the value upgrade ratio. Economic modelling of combined prompt gamma neutron activation analysis (PGNAA) and XRF sorting has predicted a 6.5% increase in net smelter return at copper operations, and XRT sorting has been shown to reduce capital and operating costs by up to 20% in certain base-metal applications (Robben, 2014, as cited in Lessard et al., 2014; MDPI Minerals, 2019).
Pre-concentration also has a profound effect on mine-level economics. By rejecting waste from the ROM stream, a sorting circuit effectively lowers the cut-off grade, allowing previously uneconomic blocks or marginal waste dumps to be processed profitably. This can extend mine life significantly. Sorting is also the technology of choice for upgrading historic waste-rock dumps that were produced when processing technology was less efficient (von Ketelhodt, 2009, as cited in Wills and Finch, 2016). With declining average ore grades globally, the ability to valorise low-grade material at coarse particle size before incurring comminution costs is strategically important.
From a sustainability standpoint, by reducing the mass of material processed through wet circuits, sorting reduces specific water consumption, specific energy consumption, and the volume of tailings requiring disposal. These benefits align with increasing regulatory and investor pressure to reduce the environmental footprint of mining operations.
Applications by Ore Type
The suitability of ore sorting is governed by two conditions: sufficient liberation of valuable mineral and gangue at coarse particle size, and a detectable contrast in at least one physical or chemical property. Where both conditions are met, the technology can be highly effective.
Diamonds
Diamond processing is the ore-sorting application with the longest history and the highest commercial penetration. The unique X-ray luminescence of diamonds provides an unambiguous signal that sorters can detect at high reliability. XRL sorters handle virtually all diamond operations at the final recovery stage after DMS concentration, and XRT is increasingly used earlier in the circuit. TOMRA reports that a 2,492-carat stone, one of the largest rough diamonds in recorded history, was recovered at the Karowe Mine in Botswana using XRT technology (TOMRA, 2020).
Base Metals (Copper, Zinc, Lead)
Vein-type, brecciated, and layered base-metal mineralisation can show strong liberation at coarse sizes, making XRT, XRF, and EM sorting effective for waste rejection. The Svyatogor Cu/Zn mine in Russia uses RADOS XRF sorters to reject low-grade material, and numerous copper operations have trialled XRT pre-concentration. Studies of porphyry copper ores using NIR spectroscopy have also shown that alteration mineralogy is a useful proxy for copper grade at the particle level (Dalm et al., 2014).
Uranium
The natural radioactivity of uranium ores makes radiometric sorting the obvious choice. Pre-concentration of uranium ore by radiometric sorting has been practised at mines in South Africa, Namibia, Australia, and Canada since the 1950s. At the Rössing Mine in Namibia, gamma scintillation counters beneath the belt detected higher-grade ore pieces, enabling selective diversion to the leach circuit.
Industrial Minerals
Limestone, magnesite, talc, feldspar, phosphate, and other industrial minerals are frequently sorted by colour or NIR. In these applications, product quality rather than metal grade is the target specification: for example, a pure-white marble or a talc product free from carbonate inclusions may command a significant price premium. Colour sorting has been used commercially for these applications since the 1970s.
Scheelite and Tungsten
Scheelite exhibits strong ultraviolet fluorescence, which was the original basis for sorting at King Island Scheelite in Tasmania. XRT has since proven superior because it is not affected by surface condition, and the Mittersill Mine in Austria operates XRT sorters at 130 t/h to reject approximately 25% of ROM feed as barren aggregate (Mosser and Gruber, 2010).
Coal
Optical and photometric sorters have been applied to coal preparation, where the colour contrast between bright coal and pale shale or bone provides a useful sorting signal. XRT is also applicable in coal, where the density contrast between clean coal and mineral matter (ash) is exploited.
Integration into Mining Flowsheets
Integrating a sorter into an existing or new flowsheet requires attention to both the upstream preparation circuit and downstream process interactions. A typical SBS installation consists of a primary or secondary crusher, a screening circuit to define the feed size fraction, the sorter itself, and a compressed air supply system. For high-throughput applications, multiple sorter units are operated in parallel; cascading rougher-scavenger arrangements improve both separation efficiency and overall availability.
The ideal position in the flowsheet is immediately after sufficient liberation has been achieved but before any wet processing. Treating freshly blasted ROM material at primary crushed size (typically 50–150 mm) maximises economic benefit because particle sizes are large, throughput per unit is high, and the volume of material rejected is still coarse and can be easily disposed of as aggregate or waste rock fill. Secondary crushed size (20–50 mm) is used when primary liberation is insufficient.
For underground mining, the compact footprint of containerised sorter units makes near-to-face pre-concentration technically feasible. About 30% of new machine installations worldwide are containerised systems (Wills and Finch, 2016). Eliminating waste underground reduces hoisting costs, shaft wear, and surface infrastructure requirements, and several conceptual studies have demonstrated positive economics for underground sorting in narrow-vein and cut-and-fill stoping operations (Murphy et al., 2012).
The reject stream from the sorter must be assigned a disposition. Options include waste-rock dump, mine backfill aggregate, construction aggregate for sale, or heap leaching if the reject grade, though below cut-off for conventional processing, warrants a low-cost hydrometallurgical option. The accept stream may either proceed directly to a conventional mill or, if the sort has produced a high enough grade, to a concentration-only circuit such as leaching or gravity concentration.
Limitations and Ore Amenability Testing
Sensor-based ore sorting is not universally applicable. The technology requires a detectable property contrast between valuable and gangue particles at the target particle size. Where mineralisation is truly disseminated throughout the rock matrix at a scale below the particle size being tested, the sorting signal will be ambiguous and recovery losses will be high. Massive sulfide deposits, fine-grained disseminated gold, and fine-grained porphyry systems with pervasive alteration may offer poor sorting response unless liberation at coarse sizes can be demonstrated.
The economically feasible lower size limit of approximately 10–20 mm is a genuine constraint. Processing finer material dramatically increases machine throughput requirements and capital cost per tonne, often making dense-media separation or gravity concentration more cost-effective below this size. Similarly, a wide size range in the feed degrades separation efficiency because the ejection system is calibrated for a defined particle size range; a size range coefficient exceeding three is generally problematic (Wills and Finch, 2016).
Ore amenability testing is essential before any commercial commitment. Testing should include representative sampling from all ore domains expected during mine life, not just from high-grade zones. Key tests include: laboratory-scale sorter trials on crushed and screened samples to generate grade-recovery curves; liberation analysis at the target particle size by automated mineralogy (QEMSCAN or MLA); and assessment of property variability through the ore body. TOMRA and other suppliers operate dedicated test centres where pilot-scale trials can be performed under controlled conditions on split samples.
Complicating factors include moisture sensitivity of optical and NIR sensors, which may require washing or drying circuits; surface oxidation altering spectral properties in weathered ore zones; and high slimes content, which blinds sensors and reduces air ejector reliability. Careful feed preparation, particularly screening to remove fines below about 10 mm, is usually the most important operational intervention for maintaining sorting performance.
The Future of Ore Sorting
The trajectory of sensor-based ore sorting is strongly positive. Three convergent forces are driving accelerating adoption: falling ore grades and rising processing costs, increasing availability of low-cost sensing hardware originally developed for the food and recycling industries, and tightening environmental standards that reward reduction in water and energy use per unit of metal produced.
Multi-sensor fusion is the most technically significant near-term development. Individual sensors have complementary strengths and weaknesses; combining XRT atomic-density data with NIR spectral data, for example, can discriminate ore types that neither sensor can resolve alone. Research published in MDPI Minerals reviews sensor fusion as a route to handling the complex, multi-mineral feeds that single-sensor systems struggle with (MDPI Minerals, 2022). Machine learning algorithms are increasingly being applied to fuse multi-sensor signals in real time.
Process analytical technology (PAT) approaches, where ore-sorting data are integrated with plant-wide process control, offer a further step forward. If the sorter simultaneously characterises the grade and mineralogy of the ROM feed and communicates that information to the mill control system, downstream processes can be adjusted proactively rather than reactively. This closes the loop between mine production and mill operation in a way that was not previously possible.
Underground sorting remains an important frontier. The economic case for near-to-face waste rejection in deep underground mines is compelling: hoisting costs are a major component of operating cost, and eliminating barren tonnes before hoisting them to surface is economically attractive. Miniaturisation of sensor packages and the development of robust, dust-resistant electronics are making underground deployment progressively more practical.
Finally, the application of sorting to tailings and waste-rock reprocessing is gaining momentum. Historic waste dumps and tailings impoundments may contain material that, when assayed particle by particle, yields an economically sortable fraction. With global pressure to reduce new land disturbance and to remediate legacy mining sites, sensor-based sorting of mine waste is emerging as both an economic opportunity and an environmental responsibility (Canadian Mining Journal, 2023).
References and Further Reading
- Wills, B.A. & Finch, J.A. (2016). Wills’ Mineral Processing Technology, 8th Edition. Butterworth-Heinemann/Elsevier.
- Robben, C. & Wotruba, H. (2019). Sensor-Based Ore Sorting Technology in Mining—Past, Present and Future. Minerals, 9(9), 523. MDPI.
- Robben, M. & Wotruba, H. (2022). A Review of Sensor-Based Sorting in Mineral Processing: The Potential Benefits of Sensor Fusion. Minerals, 12(11), 1364. MDPI.
- TOMRA Mining. (2020). Sensor-Based Ore Sorting Solutions. TOMRA Sorting Solutions.
- Canadian Mining Journal. (2023). TOMRA Mining’s sensor-based sorting tech reshaping global mining economics.
- Mining Technology. (2023). How can sensor-based ore sorting turn waste into wealth?
- Lessard, J., et al. (2014). Developments of ore sorting and its impact on mineral processing economics. Minerals Engineering, 65, 88–97.