Mineral Processing
Complete Workflow Explained

Drill-core samples go through QEMSCAN or MLA for mineralogical analysis, drop-weight and SMC tests for hardness, liberation studies, and bench-scale separation tests. Whether the copper in a porphyry deposit sits in chalcopyrite or in secondary minerals like covellite and chalcocite reshapes the reagent regime. Whether gold in an ore occupies grain boundaries or is locked inside arsenopyrite determines whether gravity recovery works or whether oxidative pretreatment is necessary.

Orebodies have spatial variability that can shift flotation response over a few bench levels in the pit when a copper deposit transitions between supergene and hypogene zones. Geometallurgical block models tied into mine scheduling catch these transitions in advance. Without that link between geology and metallurgy, the plant ends up reactive, spending weeks adjusting reagent doses to compensate for a change that was visible in the drill data months earlier.

The Bond Work Index, a 1950s parameter designed for rod and ball mill circuits, shows up as the primary comminution input in nearly every feasibility study. For SAG or HPGR circuits it needs SPI, SMC, or JK Drop Weight data alongside it. Without those, mills get sized wrong. Comminution testwork is expensive and slow, and finance-driven project schedules cut it. On the geotechnical side of the same project, nobody would accept a dam designed from three soil samples. Mill sizing from a handful of composited Bond tests passes review routinely.

A composite blending hard siliceous rock with soft clay-rich rock produces test results that describe neither. The plant then alternates between overgrinding the soft material and undergrinding the hard. Variability testwork on individual geometallurgical domains costs three to five times more and gets cut from the budget at many projects. That cost saving during the study phase has destroyed more cumulative project value than any single metallurgical variable across the industry.

These two stages belong together in any discussion that tries to describe how a grinding circuit behaves, because the classifier output determines what comes back to the mill and the mill product determines what the classifier receives. Treating them as separate topics, which is what most articles do, hides the feedback loop that governs both.

Fifty to seventy percent of a plant's energy consumption and a dominant share of its capital cost sit in comminution. Primary crushers take run-of-mine rock down to 150 to 200 mm. Cone crushers in secondary and tertiary stages bring it to 10 to 15 mm with screens between each stage controlling product size. A blinded or worn screen propagates problems through the whole plant in ways that can take weeks to trace.

SAG mills are difficult equipment. Ore hardness, feed size distribution, ball charge percentage, mill speed as a fraction of critical speed, feed moisture, grate open area, pulp lifter geometry, trommel screen condition: move any one of these and the others respond. Critical size material, particles caught between being useful as grinding media and being small enough to discharge, accumulates inside the mill and chokes throughput. Pebble ports and recycle pebble crushers and ball charge management deal with it, and how well a plant manages critical size is one of the clearest markers of grinding circuit competence.

A SAG mill reaches peak throughput near its volumetric limits, which means occasional high-fill events. At operations where mill trips create incident reports, operators learn to hold the mill well below capacity. Fifteen to twenty percent below, sometimes more. This becomes the baseline over a period of months. Production targets get set against the conservative operation. New operators learn the conservative settings. The people who originally made the decision to hold back leave. Within a few years, the institutional memory that the mill was being constrained is gone, and the plant runs below potential without anyone recognizing the gap. This is one of those problems that cannot be solved by better instrumentation or control algorithms because it originates in management culture, not in process engineering.

HPGR compresses material between counter-rotating rolls, generates grain-boundary micro-fractures, and cuts comminution energy by 15 to 30 percent relative to SABC circuits. Operations in South America, Southern Africa, and Australia run it at full scale. It pairs with dry screening, which opens up low-water comminution circuits for arid regions.

Stirred mills, IsaMill and Vertimill being the two dominant machines, use small media and deliver roughly half the energy per ton of new surface area below 40 to 50 micrometers compared to ball mills. Any circuit requiring fine grinding should be using them.

The cyclone feed pump controls more of the grinding circuit's metallurgical performance than the organizational structure of most plants can detect. When the impeller wears, cyclone inlet pressure drops gradually over weeks and months. The cut point drifts coarser. More poorly liberated material reaches flotation. Recovery falls. The metallurgist adjusts flotation reagents. The maintenance planner tracks the pump on a separate schedule. These two responses happen in parallel, in different departments, reported in different documents, discussed in different meetings. The connection between pump wear and flotation recovery lives in the space between those departments, and at plants organized in conventional silos, nobody occupies that space. A single team with authority to track mass balance across the whole circuit and investigate across departmental boundaries would catch this in a week. That kind of cross-functional metallurgical accounting team exists at some operations, particularly in Southern Africa where the tradition of plant metallurgist oversight is stronger, and the difference in diagnostic speed is dramatic.

Clay minerals change slurry rheology when the ore shifts into clay-rich zones. Viscosity rises, the cyclone cut point moves coarser, fines bypass to the underflow, the mill receives material it cannot grind further, the overflow gets coarser. Adding water, the reflex response, can reduce mill residence time and worsen things. Proper response involves blending, rheology modifiers, cyclone geometry changes, and sometimes bypass circuits. At many operations the throughput drops and stays down until the geology changes again. The phrase "difficult ore" gets used as an explanation and as a signal that no further investigation is needed, which is convenient and expensive.

This section is longer than the others because flotation is where the largest number of interacting variables converge in the smallest physical space, and because the gap between how flotation is described in textbooks and how it behaves in an operating plant is wider than for any other unit operation in the flowsheet.

A flotation circuit runs on surface chemistry under turbulent three-phase flow conditions on mineral surfaces that are heterogeneous, partially oxidized, and contaminated by ions from the process water. Every reagent interacts with every other reagent and with the water chemistry and with the mineral surface condition, so adjusting any single input shifts selectivity across the whole circuit.

The hydrodynamic conditions at the point where collector is added change its adsorption behavior. Collector dosed into mill discharge at high turbulence produces different surface coverage than the same quantity added gently to a conditioning tank. Same molecule, same dose, same ore. Different metallurgy. This is not a minor effect. It is large enough to explain a significant fraction of the gap between laboratory flotation performance and plant flotation performance, because a 2.5-liter bench cell and a 300-cubic-meter mechanical cell create completely different hydrodynamic environments. No amount of reagent optimization in the lab can account for conditions that do not exist in the lab. The problem has no clean solution. Making a lab cell behave like a plant cell is physically impossible, and developing predictive models of collector adsorption kinetics as a function of hydrodynamic environment is an active research problem with limited practical progress so far.

The rougher circuit pulls out as much valuable mineral as possible and accepts gangue contamination. Cleaners take rougher concentrate and reject gangue through re-flotation under more selective conditions. Scavengers work on rougher tails to recover what the rougher missed. Cleaner tails recycle, creating internal loads that affect retention time and reagent consumption throughout the circuit.

Column cells use counter-current wash water to suppress entrainment and produce cleaner concentrate than mechanical cells. For zinc operations with iron penalties or copper operations with arsenic penalties, the economic gap between a well-configured and a poorly configured cleaner circuit runs to tens of millions of dollars per year in smelter settlements.

The regrind mill that treats rougher concentrate before cleaning connects to this directly. The target regrind size has to account for the cleaner circuit's ability to reject the finer gangue it will create: whether the cleaners are columns or mechanical cells, what froth depth they operate at, what wash water rate columns use. When the comminution team sizes the regrind and the flotation team designs the cleaners and they present their work in separate chapters of the feasibility study, the interaction between regrind size and cleaner entrainment falls in the gap between the two teams.

Froth depth determines how much entrained gangue drains back to the pulp before bubbles reach the cell lip. Deeper froth drains more gangue and produces cleaner concentrate at the cost of recovery, because weakly attached valuable particles also detach during the longer transit. Most plants control flotation through reagent dosage and air rate while froth depth is fixed by launder height. Adjustable launders or dart valves exist and would give operators an independent metallurgical lever. They add mechanical complexity and require training that most operator workforces have not received, which is why they remain uncommon despite being understood for decades.

Centrifugal gravity concentrators like Knelson and Falcon units recover fine free gold that spirals and jigs cannot catch. Gravity-recovered gold does not go through the smelter, so its net value per ounce exceeds flotation gold. A large number of older gold and gold-copper plants still lack a gravity circuit despite the economics of retrofitting one being favorable.

Gold is malleable. It flattens in the mill rather than breaking, which increases its surface area and decreases its settling velocity. The longer gold stays in the grinding circuit, the harder it gets to recover by gravity. Pulling a feed from the SAG mill discharge captures coarse gold early, before classification and ball milling have deformed it. Pulling from the cyclone underflow captures gold that the cyclone called coarse based on its density, which may mean particles that are physically quite small. The optimal bleed point depends on the gold size distribution and on how much plastic deformation the specific grinding circuit induces, which means it should be determined through testwork at each operation rather than defaulted to the cyclone underflow as a matter of convention.

Magnetite iron ore goes through low-intensity magnetic separators. Weakly paramagnetic minerals need high-intensity or high-gradient units. Heavy mineral sand operations use electrostatic separation to sort conductors from non-conductors, and feed temperature and humidity control is critical because small deviations in either collapse separation efficiency. Operations in humid tropical climates deal with this as a constant, and it creates a performance gap relative to arid-climate operations running identical equipment that rarely gets discussed openly because both sets of operations are usually owned by different companies operating in different regulatory environments with different reporting standards.

Dense medium separation immerses feed in a ferrosilicon or magnetite suspension of controlled density to reject barren rock at coarse sizes. Sensor-based ore sorting, using X-ray transmission or optical or nuclear sensors, does particle-by-particle rejection at even coarser sizes and can discriminate on composition rather than density alone.

Treating pre-concentration as a metallurgical optimization understates its reach. It is a resource definition decision that changes the size of the orebody. Most feasibility studies evaluate pre-concentration as an option, as something to add to the flowsheet if the economics justify it. The framing should be inverted: the default should be to include pre-concentration and then evaluate whether there is a reason to omit it. The difference in framing changes which projects get built with it and which do not.

These four topics interact so heavily that separating them into individual sections would obscure the connections.

Thickeners settle solids with flocculant. Flocculant performance depends on water chemistry, particularly divalent cation concentrations of calcium and magnesium. A plant that increases water recycling finds its flocculant underperforming because the cation concentrations in the return water have shifted. The thickener underflow gets less dense. If the tailings strategy is paste or filtered dry-stack, the downstream dewatering equipment now receives feed outside its design envelope. The whole chain from flotation tails to tailings deposition is linked through water chemistry.

Filtration brings concentrate moisture to 8 to 12 percent for shipping. Every percentage point of excess moisture in a million-tonne-per-year operation means paying ocean freight to move water. The connection between filter performance and realized revenue per tonne of concentrate sold is direct and large enough to justify filtration being treated as a revenue-critical operation rather than a back-end utility.

Mount Polley, Samarco, and Brumadinho ended the era of treating tailings as an afterthought. Filtered dry-stack tailings, paste tailings, and co-disposal with waste rock are the current direction, each constraining plant design in specific ways: dewatering capacity, thickener specification, geotechnical requirements. In some jurisdictions the tailings facility now costs more than the processing plant, which inverts the traditional capital cost hierarchy and means that any flowsheet change reducing tailings volume carries a multiplied economic benefit.

The lead circuit metallurgist sees poor selectivity and adjusts depressant. The copper ions keep arriving through the water. Nothing improves until someone samples and analyzes the return water and traces the activation mechanism back to its source, which may be in a completely different part of the plant.

Lime, the highest-tonnage reagent in most sulfide plants, varies in calcium oxide content, reactivity, and impurity profile depending on the limestone source and calcination conditions. High-silica or under-calcined lime uses more per unit of pH change and introduces ions that interfere with flotation. Feasibility studies assume a generic specification. The gap between that specification and what is commercially available at the project's location can be substantial, and it has operational consequences that persist for the life of the mine.

Model predictive control stabilizes SAG mills well. Machine learning applied to flotation works within the ore domain it was trained on and degrades outside it. Combining mechanistic structure with data-driven calibration holds up better across domain transitions than either approach alone.

The constraint at most plants is instrument maintenance. A particle size analyzer drifting out of calibration, a density gauge with a fouled source, an XRF with a worn sample presentation system: each corrupts control system inputs. The control system then optimizes against corrupted data, which is worse than not optimizing at all because it actively drives the plant away from the correct operating point while everyone believes the system is working. The gap between plants with dedicated instrument technicians and enforced calibration schedules versus plants where instrument maintenance competes with mechanical maintenance for the same workforce and budget is a gap that shows up in metallurgical performance consistently and across commodities and geographies.

Ore hardness propagates through the SAG mill into the cyclone cluster into flotation into concentrate quality into smelter settlements into revenue into the mine plan. Every stage is coupled to every other through mass flow, water chemistry, energy balance, and economics. The flowsheet shows boxes and arrows. The plant runs on the interactions between those boxes, and the interactions are where performance is won or lost.