Inside the cyclone: how mines tune grind size in real time

Walk into any combined SAG/ball mill (semi‑autogenous grinding followed by ball milling) circuit and you’ll find the classification workhorse: the hydrocyclone. Slurry is blasted tangentially into a cone, a vortex forms, and centrifugal forces split fines from coarse. The mechanics are simple, the stakes are not.

Get the “cut size” wrong and you either starve downstream flotation/leaching or burn power pulverizing particles that were already fine. The fix: tight control of cyclone pressure, geometry, and solids — and a continuous read on product size.

As 911Metallurgist puts it, bigger particles are driven to the wall and leave via the apex (underflow), while fines ride the inner core out the top (overflow). That split is governed by the cyclone’s vortex finder and apex, the feed conditions, and the pump pressure — collectively setting the separation cut (P50/d50 or P80; P80 is the size at which 80% of the material is finer).

Hydrocyclone classification basics

Hydrocyclones control which fraction advances and which recirculates. Coarse particles exit the apex and return to the mill; fine particles exit the vortex finder and go to downstream processes (911Metallurgist; OneMine). Geometry and inlet conditions set the cut size. For example, a target P80 of ~150 µm means 80% of overflow solids should be finer than 150 µm (Scribd).

Operating envelopes are well known: hydrocyclone feed around 2.0 bar (180–250 kPa) at ~35–50% solids can yield classification efficiencies ~40–45% at a ~45 µm cut, with recirculating loads typically 200–500% (Australian patent AU2021473504A1). In practice, mills often operate in the 150–600% circulating‑load range by adjusting cyclone split fractions (opening/closing cyclones) (AU2021473504A1). The intent is clear: keep the underflow “dense” with oversize and the overflow relatively dilute with fines.

Operating parameters: pressure, solids, geometry

Pressure and flow. Raising feed flow/pressure strengthens the centrifugal field, boosting throughput and driving more fines to overflow — but it shortens residence time and can reduce separation sharpness. In a numerical study, increasing inlet velocity from 1 m/s to 5 m/s drove internal pressure from ~5.5 kPa to ~182.8 kPa and tangential velocity from ~2.0 m/s to ~11.2 m/s, with the authors noting that the higher feed rate “result[ed] in insufficient particle separation and reduced separation accuracy” (MDPI Separations). Operators therefore hold cyclone pressure constant (via the feed pump), since “a constant pressure ensures positive control over the classification” (911Metallurgist).

Solids concentration. Higher feed solids raise slurry viscosity and degrade separation: “the higher the feed concentration… the greater the slurry viscosity, which reduces the separation efficiency of the hydrocyclone” (MDPI Minerals). Rope‑like underflow (very dense flow) can capture fines and spike circulating load. Circuits temper this by controlling dilution or adding wash water to hold ~35–50% solids in the cyclone feed (AU2021473504A1).

Geometry. Apex (spigot) and vortex finder sizes have first‑order effects. A smaller apex reduces mass flow to underflow, forces some coarse to overflow (finer cut), and raises circulating load; enlarging the apex lets more slurry — including fines — leave as underflow (coarser cut). A recent “W‑shaped” spigot apex reduced fine (<74 µm) carry‑over into the underflow by ~1.5 percentage points in an iron ore plant (MDPI Water). Plants also add or remove cyclones in a cluster to match mill load. If pressure is too low or the apex too large, “all of the feed [can] be discharged out of the underflow” (loss of fine capture); excessively high pressure or a too‑small apex can collapse the air core and cause overload (911Metallurgist).

Cut size control and split ratio

Changing pressure or aperture shifts both cut size and split ratio. Increasing pressure or reducing apex tends to decrease the mass split to overflow (more slurry forced to underflow), concentrating solids in underflow; Zhang et al. observed that higher feed rates reduced the cyclone split ratio (overflow fraction) (MDPI Separations). Practically, overflow must stay dilute enough for flotation — often <30% solids — which caps how far operators can push split changes.

Typical moves are straightforward. If the overflow P80 coarsens above spec, raise pump pressure and/or tighten the underflow (smaller orifice) to push the cut finer. If overgrinding fines, lower the pressure or open the apex slightly to let more solids report to underflow and preserve throughput. Stable control hinges on maintaining the hydrocyclone feed pump and valves (911Metallurgist).

Online particle size analyzers

Hourly lab checks can’t keep up. Offline PSD (particle size distribution) via sieving or laser diffraction every 30–60 minutes is too slow for tight control; upsets slip through. Continuous online analyzers now feed real‑time grind size to operators and control systems.

Technology options span laser diffraction, acoustic sensing, and imaging. Laser diffraction probes (e.g., Malvern Insitec or Endress+Hauser analyzers) sample slurry — often via a bypass tube — and measure laser‑light scattering, covering ~0.1–2,500 µm and handling wet slurries (ATA Scientific). Acoustic systems such as CiDRA’s CycloneTrac insert a probe directly into the overflow through a 50 mm port and infer PSD from stress waves generated by particle impacts; they output a PSD “trend” every few seconds, require no sample handling or moving parts, and can track the target P80 (OneMine). Other approaches include optical imaging with machine learning and older centrifugal/radiometric sensors (e.g., “Hydrosizer” shear‑through‑inertia sorters with gamma detectors) (ResearchGate; 911Metallurgist).

Benefits, payback, and control integration

Continuous PSD “allows problems to be rectified before they can escalate,” delivering “much greater control of particle size,” “faster optimisation of plant operation,” “less energy consumption,” and “a reduction in waste” (ATA Scientific; ATA Scientific). Payback is frequently 6–12 months from higher throughput and recovery (ATA Scientific).

Plants wire the analyzer’s overflow P80 into feedback loops to adjust cyclone pump speed or splitting valves. Some run automatic pump/load control to hold a target overflow P80 or underflow density. Emerging AI (e.g., convolutional neural networks on slurry images) shows promise, though practical deployment is still developing (ResearchGate). At minimum, online PSD lets operators detect upsets immediately. As Poplawski notes, near‑line lab measurements were too infrequent “for reliable use in automatic control,” whereas field trials with an online acoustic probe showed stable, real‑time tracking, with the sensor “constantly in contact with many particles… orders of magnitude more [data] than traditional sample‑based” instruments (OneMine; OneMine).

Key numbers in practice

Typical setpoints: pump pressure ~1.8–2.5 bar, feed solids 30–50%, and d50 cut sizes in the tens of microns for base‑metal circuits (AU2021473504A1). One example: ≈200 kPa and ≈40% solids delivered ~45% separation efficiency at 45 μm d50 (AU2021473504A1). Doubling feed velocity in simulation increased internal pressure ~30× and velocities ~5–6× (MDPI Separations), while raising slurry density quickly degraded efficiency (MDPI Minerals).

In a well‑controlled circuit, only the fine fraction (<target P80) moves to flotation or leach, while 2–5× the feed tonnage recirculates as coarse back to the ball mill. If overflow P80 drifts coarse, raise pump speed or partially close a valve to lift pressure (finer cut); reverse to increase throughput. Inline particle size sensors (laser or acoustic) verify that adjustments hold target and contribute to “higher return on investment” due to better size control (ATA Scientific).

Sources and references

Authoritative studies and industry reports on hydrocyclones and PSD control: 911Metallurgist; MDPI Separations; MDPI Minerals; MDPI Water; Australian Patent AU2021473504A1; ATA Scientific; ATA Scientific; OneMine; OneMine. References include Zhang et al. (2022) Separations 9(11):349; Liu et al. (2024) Water 16(7):1070; Wang & Chen (2021) Minerals 11(3):307; Australian Patent AU2021473504A1 (2021); Poplawski (2019) CIM J.; ATA Scientific (2017); 911Metallurgist (2024). Underflow/overflow PSDs and efficiencies above are drawn from these sources (MDPI Separations; MDPI Minerals; MDPI Water; AU2021473504A1; ATA Scientific; OneMine).