Hydrocyclones and Classification: Sizing, Control, and Troubleshooting

Classification is one of the most pervasive unit operations in a mineral processing plant, yet it is often treated as an afterthought behind the flashier business of grinding or flotation. In reality, how well you classify determines how efficiently your entire circuit runs. A poorly performing classifier sends oversize material to the downstream process and fine material back to the mill, wasting energy, inflating circulating loads, and undermining downstream metallurgy. At the center of modern classification practice sits the hydrocyclone — a device of elegant simplicity that exploits centrifugal force to separate particles by size and density at throughputs no mechanical classifier could match. This guide covers classification from first principles through to hydrocyclone design, performance analysis, spiral classifier operation, circuit sizing, and the latest advances in the field. Whether you are designing a new comminution circuit, troubleshooting an existing one, or simply building your knowledge base, you will find rigorous, practical, and up-to-date treatment of every major topic in hydrocyclone mineral processing.

Table of contents

• What is Classification in Mineral Processing?
• Hydrocyclones: Operating Principles and Flow Patterns
• Hydrocyclone Geometry: Design Variables and Their Effects
• Hydrocyclone Performance: Cut Size, Efficiency Curves, and the Fish-Hook Effect
• Spiral Classifiers: Design, Operation, and When to Use Them
• Sizing and Selecting Hydrocyclone Circuits
• Classification in Closed-Circuit Grinding
• Troubleshooting Common Classification Problems
• Advances in Classification Technology
• Hydrocyclone selection and sizing framework
• Hydrocyclone troubleshooting matrix

What is Classification in Mineral Processing?

Classification, as defined by Heiskanen (1993), is a method of separating mixtures of minerals into two or more products based on the velocity with which particles fall through a fluid medium. In almost all industrial mineral processing applications that fluid is water, making wet classification the dominant form. The process is specifically applied to particles that are too fine — generally below 200 μm — to be screened effectively and economically.

The fundamental purpose of classification in a processing plant is twofold. First, classifiers are used to close grinding circuits, removing finished-size product and returning oversize material to the mill. This prevents overgrinding, reduces unnecessary energy consumption, and produces a product with a narrower particle-size distribution than would result from open-circuit grinding. Second, classifiers perform preparatory duties for downstream separation processes, conditioning feed streams for spirals, shaking tables, flotation cells, and other unit operations that require a controlled size or density distribution.

The Physics of Particle Settling

To understand why classifiers work, you must understand how particles behave in a fluid. When a particle falls through water it experiences three forces: the downward pull of gravity, an upward buoyancy force equal to the weight of displaced fluid, and an upward drag force that increases with velocity. As velocity builds, drag grows until it exactly balances the net downward force, at which point the particle travels at its terminal velocity.

For small particles (Reynolds number below 1) settling is governed by Stokes’ law:

v = g d² (ρ_s − ρ_f) / 18η

where v is terminal velocity, d is particle diameter, ρ_s and ρ_f are the densities of the solid and fluid respectively, and η is the fluid viscosity. The key implication is that terminal velocity depends on both size and density. A small dense particle — say fine galena — can therefore settle at the same rate as a large light particle of quartz. This equivalence is quantified by the free-settling ratio, the ratio of diameters of particles of two different densities that fall at the same rate. For quartz (s.g. 2.65) and galena (s.g. 7.5) in water, this ratio is about 3.6 under Stokes’ conditions and rises toward 4.0 in the turbulent regime described by Newton’s law.

Free Settling Versus Hindered Settling

Free settling dominates when solids content is below roughly 15% by weight, so that particle-particle interactions are negligible. Above this concentration the pulp density rises, effective viscosity increases, and the system transitions to hindered settling. Under hindered conditions, settling is better described by a modified Newton’s law in which the driving term uses the difference between particle density and pulp density rather than fluid density alone. The practical consequence is that hindered settling amplifies the effect of density on classification while reducing the effect of size. For the quartz–galena pair, the hindered settling ratio in a pulp of density 1.5 rises to about 5.2, which is higher than the free-settling ratio. Hindered-settling classifiers exploit this to sharpen density-based separations, whereas free-settling classifiers use dilute conditions to maximize the size-selectivity of separation.

Why Classification Matters for Grinding Circuit Performance

In a closed-circuit ball mill, the classifier is not a peripheral device — it is the control element. By removing product-sized particles before they can be reground, it improves circuit energy efficiency and increases throughput capacity. It also narrows the particle-size distribution of the overflow, which benefits flotation by reducing the proportion of both ultra-fine slimes and oversize particles. The presence of fine high-density minerals such as galena or gold in the cyclone feed introduces a density effect: these minerals preferentially report to the underflow relative to their size, increasing the circulating load on those minerals and leading to selective overgrinding. Understanding and managing this effect is one of the critical metallurgical considerations in circuit design.

Hydrocyclones: Operating Principles and Flow Patterns

The hydrocyclone — commonly abbreviated to just cyclone in plant language — is a continuously operating centrifugal classifier that uses the pressure energy of the feed stream to generate a rotating flow field, accelerating the settling rates of particles by orders of magnitude compared with gravitational settling alone. It has no moving parts, occupies a small footprint, handles high throughputs relative to its size, and can be manufactured in materials ranging from natural rubber and polyurethane to high-chrome white iron — all of which make it by far the most widely used classifier in mineral processing.

Construction and Basic Flow

A standard hydrocyclone consists of a cylindrical section at the top, which transitions into a conical section that narrows progressively to an open apex (also called the spigot or underflow discharge). Feed slurry enters the cylindrical section through a tangential inlet, imparting a swirling, vortex motion to the contents. The top of the cylindrical section is closed except for an axially mounted overflow pipe that extends downward into the body of the cyclone as the vortex finder. This critical component forces the incoming feed to travel downward before it can exit via the overflow, preventing direct short-circuit of coarse unclassified material straight up the axis and out.

The centrifugal acceleration generated by the rotating flow can reach hundreds of times the acceleration of gravity in small, high-pressure cyclones, far exceeding anything achievable in a gravitational classifier. This is why hydrocyclones can make fine separations — below 20 μm in small units — at high throughput rates.

Internal Flow Structure

The flow pattern inside a hydrocyclone is more complex than a simple inward spiral. There is an outer helical vortex that spirals downward along the cyclone wall, carrying coarse particles to the apex, and an inner helical vortex that reverses direction and spirals upward through the core and out the vortex finder as the overflow. Along the axis of the cyclone, low pressure causes dissolved air to come out of solution and form a continuous air core that extends from the apex to the vortex finder. This air core is not a design flaw — it is an essential feature of correct operation, allowing coarse particles to discharge freely at the apex and maintaining the dual-vortex structure.

Experimental work by Renner and Cohen (1978) identified four distinct zones within the body of a hydrocyclone, each with a characteristic particle-size distribution. A narrow annular zone adjacent to the cylindrical wall contains essentially unclassified feed material. A large conical zone contains fully classified coarse material resembling the underflow product. A narrow central zone surrounding the vortex finder contains fully classified fine material resembling the overflow. Classification actually takes place in a relatively narrow toroid-shaped region between these zones, where particles are radially stratified by size. This finding has important implications: only a portion of the cyclone body is doing useful classification work at any time.

The Centrifugal-Drag Force Balance

Particles within the rotating flow field experience two opposing radial forces. The outward centrifugal force tends to drive particles toward the cyclone wall and into the downward outer vortex, directing them to the underflow. The inward drag force of the fluid — which moves generally inward as it approaches the inner vortex — tends to carry particles toward the axis and up through the vortex finder. Particles reach a surface of zero radial velocity, called the envelope of zero velocity, where these forces are equal. Particles lying on this surface have an equal probability of reporting to either the overflow or the underflow. This is the conceptual basis of the d₅₀ cut size — the particle size at which half of the mass reports to each product stream.

Hydrocyclone Geometry: Design Variables and Their Effects

The performance of a hydrocyclone — in terms of cut size, throughput, and classification efficiency — is controlled by a set of interacting geometric dimensions. Understanding these relationships allows engineers to select and configure cyclones for a target separation or to diagnose performance problems in operating circuits.

Cyclone Diameter

The cylindrical diameter D_c is the primary design variable because it sets the scale of the device and its capacity. A fundamental principle is that larger diameter means coarser cut size. The physics underlying this are clear: for a given tangential velocity, the centrifugal acceleration (v²/r) decreases as the radius increases, so larger cyclones generate less centrifugal force per unit of feed pressure, resulting in coarser separations. Industrial cyclones range from 2.5 m in diameter — used for coarse, high-throughput applications — down to 10 mm micro-cyclones capable of cut sizes below 2 μm, though requiring very large numbers of units in parallel to handle plant-scale flows. The practical range for grinding circuit classification spans roughly 100 mm to 840 mm.

Vortex Finder Diameter

The vortex finder diameter D_o is arguably the most sensitive and easily adjusted design parameter in the field. Increasing the vortex finder diameter at constant pressure creates a larger low-resistance pathway for the inner vortex, which does two things simultaneously: it increases the volumetric capacity of the cyclone, and it drives the cut size coarser by reducing the centrifugal field strength in the classifying zone. Decreasing the vortex finder diameter does the opposite — finer cut size, lower capacity. Most cyclone designs offer interchangeable vortex finders as the primary field-adjustment tool for cut size control.

Apex Diameter

The spigot or apex diameter D_u controls the density and mass flow of the underflow discharge. It must be large enough to maintain the air core — the diameter of the air core is roughly equal to the apex opening — and to discharge the accumulated coarse solids without plugging. Under correct operation, the apex discharges a hollow umbrella-shaped spray at a 20–30° included angle, indicating that the air core is intact and coarse material is discharging freely. Too small an apex leads to roping, in which the air core is lost, the apex discharges a rope-like stream of extremely dense slurry, and coarse particles overflow through the vortex finder, severely reducing classification efficiency. Too large an apex results in an excessively dilute underflow with excessive water reporting to that stream, carrying unclassified fine solids with it and increasing the bypass fraction.

Cone Angle and Cylindrical Length

The cone angle and the ratio of cylindrical to total length influence the residence time of particles in the classifying zone and the intensity of the swirling flow. Shallower cone angles (around 10°) extend residence time, favoring finer, more efficient classification. Steeper angles (20° is standard in many older designs) give shorter residence time and are more appropriate for coarse separation at high throughput. The inlet geometry — whether tangential, as in most conventional designs, or involute — affects the turbulence level at entry and therefore wear rates and classification efficiency. Proprietary designs such as Weir’s CAVEX and Metso’s gMAX cyclones use modified inlet and internal geometries to reduce turbulence and improve classification performance.

Feed Inlet Area

The cross-sectional area of the feed inlet governs the entry velocity and therefore influences both throughput and the initial momentum of the rotating flow. Most manufacturers design inlets at approximately 7% of the cyclone cross-sectional area. An undersized inlet creates excessive entry turbulence and wear; an oversized inlet reduces swirl intensity and coarsens the separation. For non-circular inlets, the equivalent diameter is calculated from the inlet cross-sectional area for use in design equations.

Hydrocyclone Performance: Cut Size, Efficiency Curves, and the Fish-Hook Effect

Characterizing hydrocyclone performance requires more than a single number. The partition curve — also called the performance, efficiency, or selectivity curve — provides a complete picture of how each particle size fraction is distributed between the overflow and underflow products.

The Partition Curve and d₅₀

The partition curve plots the fraction of each size class in the feed that reports to the underflow as a function of particle size. It has a characteristic S-shape: approaching 100% recovery for the coarsest size fractions (which always report to the underflow), passing through 50% at the cut size, and approaching some non-zero value at the finest sizes. The d₅₀ — the particle size at which exactly 50% of feed mass reports to each product — is the most widely used measure of the classification cut point.

However, the raw partition curve does not fully describe the classification process, because a fraction of the feed water reports directly to the underflow, carrying with it a proportional amount of all size fractions regardless of their settling behavior. This is the bypass fraction, denoted R_w/u. Kelsall (1953) proposed correcting the raw partition curve by subtracting out this bypass contribution using:

C = (S − R_w/u) / (1 − R_w/u)

where C is the corrected fractional recovery to underflow and S is the actual fractional recovery. The resulting corrected curve passes through zero at fine sizes and defines a corrected cut size d_50c. In modern practice, cyclone manufacturers and simulators typically quote the corrected d₅₀ (dropping the subscript “c”), so care is needed when reading vendor literature.

Sharpness of Separation

The steepness of the partition curve around the cut point is a measure of classification efficiency known as the sharpness of separation. One widely used index is the imperfection I, defined as:

I = (d₇₅ − d₂₅) / (2 d₅₀)

where d₇₅ and d₂₅ are the sizes at which 75% and 25% of feed reports to the underflow, respectively. A perfect step-function classifier would have I = 0; real hydrocyclones typically yield values in the range 0.2–0.5 for the total solids partition curve. The partition curve can also be fitted to the exponential model of Plitt (1976), in which a parameter m describes sharpness — higher m indicates a steeper, more efficient separation. In practice m values rarely exceed 3, with values near 1.5–2.5 being typical for industrial cyclones. Importantly, calculating m separately for individual minerals in a multi-density feed often reveals that each mineral is classified more sharply than the aggregate solids curve suggests, because the aggregate is a weighted average of several overlapping mineral curves.

The Fish-Hook Effect

A notable anomaly sometimes observed in partition curves is a reversal at the fine end, where recovery to the underflow increases as particle size decreases below a certain threshold — producing the characteristic shape of a fish hook. This has been attributed to the idea that very fine particles follow the water split rather than any classification force, so the finest fractions split in proportion to the water reporting to the underflow. Whether the fish-hook represents a genuine physical phenomenon or is an artifact of measurement methodology and multi-density feeds has been debated extensively in the literature (Bourgeois and Majumder, 2013; Nageswararao, 2014). Computational fluid dynamics modeling has been used to investigate this question without providing a definitive resolution. For practical circuit design, the fish-hook is generally accounted for through the Kelsall bypass correction, and its impact on downstream processes can be managed by minimizing the water split to the underflow.

Effect of Density on Cyclone Performance

In circuits treating ores with significant density variation — for example, a lead-zinc ore containing galena (s.g. 7.5), sphalerite (s.g. 4.0), pyrite (s.g. 5.0), and quartz gangue (s.g. 2.65) — each mineral has a different effective d_50c. Galena reports preferentially to the underflow at sizes where quartz would overflow, leading to selective enrichment of the coarsest-density minerals in the circulating load. This density-driven behavior is both a challenge — since it causes overgrinding of dense valuable minerals — and an opportunity, since it creates a circulating load enriched in high-value minerals that can potentially be recovered using gravity concentration devices (jigs, Knelson concentrators) or flash flotation cells installed within the circuit.

Spiral Classifiers: Design, Operation, and When to Use Them

Before hydrocyclones became dominant in the 1960s and 1970s, spiral classifiers (and their close relatives, rake classifiers) were the standard tool for closing grinding circuits. Today they are found mainly in older plants, in operations requiring coarse classification (above about 150 μm), and in washing and ore-preparation plants. Understanding spiral classifiers remains important because they continue to be specified in certain applications where their characteristics offer advantages over cyclones.

How a Spiral Classifier Works

A spiral classifier consists of an inclined trough partially filled with water, within which a continuously rotating helical screw (the spiral) rakes settled solids upward against the flow of liquid. Feed slurry enters the trough and forms a settling pool. Coarse particles settle quickly to the bottom of the trough and are transported upward by the spiral to the sand discharge at the elevated end of the machine. Fine particles remain suspended and are carried by the horizontal flow of water to the overflow weir at the lower end, discharging as the fine product.

The separation mechanism is predominantly free settling: the horizontal velocity of water toward the overflow weir must be high enough to carry fine particles but low enough that coarse particles can settle against it. This sets the cut size. The depth of the settling pool, controlled by the height of the overflow weir, determines the residence time available for settling — higher weirs produce finer separations. Dilution of the feed is the most critical operating variable: increasing water addition lowers pulp density, increases free settling, and — up to a critical dilution of about 10% solids — produces finer overflow products. Above this point, increasing dilution actually coarsens the product because the rising velocity of the overflow current becomes the dominant factor.

Spiral Versus Rake Classifiers

Rake classifiers use reciprocating rake mechanisms rather than a continuous spiral. They can handle coarser and denser sands, but the reciprocating action means that rakes are withdrawn from the pulp periodically, during which some settled material may slip back. Spiral classifiers can operate at steeper angles than rake classifiers, which improves drainage and produces drier, cleaner sand products. For this reason spiral classifiers have largely displaced rake classifiers in applications where both can be used.

When to Choose a Spiral Classifier Over a Hydrocyclone

Spiral classifiers offer several practical advantages in specific situations. They operate at low pressure with no pump requirement, which saves capital and operating cost in coarse-classification duties. They produce a physically dewatered sand product that is convenient to handle, in contrast to the wet underflow slurry from a cyclone. They are also tolerant of variations in feed rate and density, making them more forgiving in small, low-automation plants. However, their fundamental limitation is inability to make fine separations economically: producing an overflow at 75 μm or finer requires dilution to pulp densities so low that the overflow product must be rethickened before downstream processing. For any application requiring separations below about 150 μm, or where high throughput is needed in a small footprint, hydrocyclones are unambiguously superior.

Hydraulic Classifiers

Hydraulic classifiers — also called teeter-bed or hindered-settling classifiers — operate on a different principle. Water is introduced upward through the classifier body, creating an upward current that counters particle settling. By using a series of sorting columns with progressively decreasing upward velocities, the feed can be sorted into a series of sized or density-separated products. The hindered-settling version exploits the enhanced density effect at high pulp densities to produce spigot products enriched in dense minerals. These devices remain relevant as feed preparation equipment for shaking tables and fine-particle gravity circuits, where narrow size distributions are required.

Sizing and Selecting Hydrocyclone Circuits

Selecting the right hydrocyclone for a given application is a multi-step engineering exercise that combines empirical models, manufacturer data, and simulation tools. Getting it right at the design stage — or at least defining a range of acceptable designs — prevents the costly underperformance problems that arise from mismatched cyclone circuits.

Empirical Models: Plitt and Nageswararao

The most widely applied empirical models for hydrocyclone design are those of Plitt (1976) in its modified form (Flintoff et al., 1987) and Nageswararao (1995). Both models express the corrected cut size d_50c as a function of cyclone geometry (D_c, D_i, D_o, D_u, cone length h), operating conditions (feed flow rate Q_f, feed pressure P, feed solids concentration C_v), and material properties (solids density ρ_s, fluid viscosity η). Both require feed-specific calibration constants, emphasizing that hydrocyclone performance is sensitive to the characteristics of the actual ore and slurry being treated — generic default values provide a starting point, but should not be trusted for final design without testwork.

A key practical relationship from Plitt’s model is that d_50c increases with cyclone diameter, increases with feed solids concentration, decreases with feed flow rate (and therefore with feed pressure), and increases with apex diameter or decreasing vortex finder diameter. These trends give the operating engineer the levers needed to adjust cut size in the field. For a quick preliminary estimate, the Arterburn (1982) technique uses a performance chart for a “base” cyclone size and applies correction factors for operating conditions to derive the cut size for a given cyclone diameter.

Sizing for Capacity: Parallel Clusters

Because achieving a fine cut size requires small cyclone diameter, and small cyclones have correspondingly low individual capacities, large-scale fine classification requires many cyclones operating in parallel — arranged in clusters, nests, or batteries. For example, a de-sliming plant may use thousands of small (51 mm) cyclones. Reliable equal distribution of feed to all cyclones in a cluster is critical: unequal feed distribution causes individual cyclones to operate at pressures and flow rates different from the design point, degrading both cut size and efficiency. Manifold systems must be carefully designed to ensure equal pressure at every cyclone inlet, and regular inspection is needed to identify and repair blocked cyclones, which are a common problem in fine-classification service.

Scale-Up Relationships

When scaling from a tested cyclone size D_c1 to a new size D_c2, the designer must consider that diameter, flow rate, and pressure are interdependent. Keeping the same pressure while increasing diameter raises both flow rate and cut size. Keeping the same cut size while increasing diameter requires a change in either pressure or flow rate. The empirical scale-up exponents published by various authors differ somewhat, but a practical set of values widely used in the industry gives: cut size scales with (D_c2/D_c1)^1.54 at constant flow rate, and flow rate scales with (D_c2/D_c1)^2.0 at constant pressure. These relationships must always be checked against the model equations rather than applied in isolation.

Using Simulation Software

Modern circuit design for classification relies heavily on simulation packages such as JKSimMet, MODSIM, and Limn. These incorporate calibrated versions of the Plitt and Nageswararao models and enable engineers to evaluate dozens of circuit configurations quickly, testing the sensitivity of product size and circulating load to changes in cyclone geometry, number of units, feed conditions, and grinding mill behavior. Simulation is particularly valuable for optimization studies on operating plants, where field tests at multiple operating points are impractical.

Classification in Closed-Circuit Grinding

The most consequential application of hydrocyclones is in closed-circuit grinding, where the cyclone overflow defines the plant product size while the underflow returns to the mill as circulating load. The performance of the grinding circuit — in terms of product size, energy consumption, and throughput — cannot be understood independently of the classifier behavior.

Circulating Load and Its Implications

The circulating load (CL) is the mass flow rate of material returning to the grinding mill divided by the mass flow rate of new feed. A high CL means the mill is processing a large quantity of already-ground material alongside fresh feed. In principle, a high CL improves grinding efficiency by ensuring that particles are never underground and always returned for further size reduction if they have not reached the target size. In practice, however, a CL that is too high creates excessive fines in the mill feed (from bypassing in the cyclone underflow), which cushion particle-particle impacts and reduce grinding rate. Typical CL values in ball mill circuits are 150–400%, with the optimal value depending on the ore hardness, target grind size, and cyclone efficiency.

The Density Effect in Grinding Circuits

In circuits treating multi-mineral ores, the density-dependent classification behavior of hydrocyclones creates mineral-specific grinding environments. Dense minerals such as galena or gold accumulate in the circulating load because they preferentially report to the cyclone underflow relative to their size. This selective overgrinding fine-grinds the dense valuable mineral to particle sizes well below the target P₈₀ for the bulk solids, which can reduce recovery in downstream flotation. The circuit designer has several options for managing this: using a coarser cyclone cut size, installing in-circuit gravity recovery devices on the cyclone underflow, or switching to screen classification, which is density-independent.

Screens Versus Cyclones in Grinding Circuits

High-frequency vibrating screens — such as the Derrick Stack Sizer — have gained traction as alternatives to hydrocyclones in grinding circuit closure because they classify on the basis of size alone, without any density bias, and produce a very sharp partition curve with essentially zero bypass. Case studies, such as the replacement of cyclones with screens at the El Brocal concentrator in Peru (Dündar et al., 2014), have shown improved grinding circuit capacity after the switch to screens, attributed to the elimination of fine material bypass that would otherwise return to the mill and cushion grinding. However, screens are significantly more expensive to install and maintain than cyclones, and some circuits have switched in the opposite direction — from screens back to cyclones — specifically to take advantage of the additional size reduction of high-density minerals that cyclones provide.

Flash Flotation and Gravity Recovery Within the Grinding Circuit

The enrichment of dense minerals in the cyclone underflow circulating load creates an opportunity to recover valuable liberated particles before they are overground. Installing a gravity concentrator — most commonly a Knelson or Falcon centrifugal concentrator — on the cyclone underflow, or a flash flotation cell on the mill discharge, allows early extraction of free gold, free sulfides, or other dense valuable minerals. This strategy can dramatically improve overall plant recovery for gold operations and reduces the mass of material that must be processed through the main concentrator. The Gravity Recoverable Gold (GRG) test, standardized by Laplante et al. (1995), quantifies the proportion of gold in an ore that is amenable to this approach and is now a standard component of gold project metallurgical testwork programs.

Troubleshooting Common Classification Problems

Even well-designed classification circuits encounter operational problems. Recognizing the symptoms and understanding their root causes allows problems to be resolved quickly before they cause metallurgical damage or circuit downtime.

Roping Underflow

Roping — in which the apex discharges a thick, rope-like stream rather than the correct hollow-cone spray — is the most serious and immediately damaging malfunction. It occurs when the apex is too small to discharge the accumulated coarse solids, the circulating load is excessively high, or the feed contains an unusually high proportion of coarse dense material. The consequence is loss of the air core, collapse of the inner vortex, and coarse oversize material reporting to the overflow. Downstream effects can include severe loss of classification efficiency, overloading of downstream processes with coarse particles, and damage to flotation cells if coarse sulfides or slimes suddenly appear in the feed. Solutions include increasing the apex diameter, reducing the circulating load, or redistributing feed across more cyclones in a cluster.

Excessive Bypass

High bypass fractions (R_w/u above about 25–30%) indicate that too much water is reporting to the underflow. This is usually caused by an oversized apex, low feed pressure, or high feed density. High bypass sends unclassified fine material — which should report to overflow — back to the grinding mill instead, increasing the circulating load and reducing grinding efficiency. The remedy is to reduce apex diameter, increase feed pressure, or reduce feed solids concentration.

Coarse Overflow

If the cyclone overflow is coarser than the target product size, the most common cause is that the cyclone is too large for the application — the d₅₀ is too coarse. Other causes include high feed density, a vortex finder that is too large, or an unusually high proportion of light-density gangue minerals that are difficult to classify. The systematic diagnostic approach is to measure the partition curve and compare the observed d_50c with the design value. If they agree, the cyclone geometry needs changing; if the observed d_50c is finer than expected but the overflow is coarser, the problem may lie with excessive bypass or poor slurry preparation.

Pump and Sump Issues

Hydrocyclone performance depends on a steady, controlled feed pressure, typically maintained by a centrifugal pump drawing from a sump. Fluctuations in sump level cause feed pressure to vary, which in turn varies the flow rate and cut size of the cyclones. Pump cavitation, worn impellers, and poorly sized sumps are common sources of pressure instability. Automatic level control on the sump, combined with variable-speed pump drives, provides the steadiest control of operating conditions and is standard practice in modern plants.

Blocked Cyclones in a Cluster

In a multi-cyclone cluster, individual cyclones can become blocked by tramp oversize material (pieces of grinding media, rubber liner, or rock fragments), causing the remaining cyclones to operate at higher flow rates than designed. Online monitoring systems using acoustic sensors or pressure measurement to detect blocked or malfunctioning cyclones are now commercially available and are particularly valuable in large clusters where manual inspection of each cyclone is impractical.

Advances in Classification Technology

Classification technology is not static. The last two decades have seen significant advances in both equipment design and process understanding, several of which are already in routine industrial use while others are still emerging.

Computational Fluid Dynamics Modeling

The internal flow field of a hydrocyclone — involving turbulent swirling flow, an air-water interface at the air core, and dispersed solid particles spanning a wide size range — was long regarded as beyond rigorous computational treatment. Advances in computational fluid dynamics (CFD), combined with turbulence models, volume-of-fluid methods for the air core, and discrete element method (DEM) particle tracking, have now made realistic simulation of hydrocyclone performance feasible for research purposes. CFD models have been validated against experimental data from techniques including particle image velocimetry, laser Doppler velocimetry, and positron emission particle tracking (PEPT), and have been used to study phenomena such as the fish-hook effect, short-circuit flows, and the influence of cone geometry on classification efficiency. While CFD models are still too computationally demanding for routine circuit design, they are increasingly being used for equipment development and to refine empirical model predictions for unusual geometries or operating conditions.

Improved Cyclone Designs

Proprietary cyclone designs incorporating involute feed entries, optimized cone profiles, and wear-resistant internal linings have improved both the efficiency and the service life of hydrocyclones, particularly in abrasive applications such as iron ore and copper sulfide circuits. The CAVEX design from Weir Minerals and the gMAX design from Metso Outotec are examples of commercially successful designs that incorporate these improvements. Semi-inverted hydrocyclone orientations, in which the cyclone operates inverted or at various angles from vertical, have been shown in pilot-scale studies to improve separation efficiency significantly in certain applications by modifying the gravitational contribution to particle settling.

Smart Monitoring and Control

Online monitoring of cyclone health and performance has evolved from simple pressure gauges to sophisticated sensor arrays. Acoustic sensors detect the characteristic sound signature of roping versus spray discharge; vibration sensors identify mechanical issues; electrical resistance tomography and microwave sensors can infer the internal distribution of solid and fluid phases without intruding into the process stream. Centralized control systems such as FLSmidth’s SmartCyclone integrate sensor data with process models to provide real-time indicators of cyclone health, detect blocked cyclones automatically, and adjust operating conditions to maintain target product size. These systems have demonstrated measurable reductions in product size variability and circuit downtime in commercial installations.

Screens as Complements to Cyclones

High-frequency fine screens are increasingly being considered not as alternatives to cyclones but as complements in integrated classification circuits. A two-stage circuit in which a cyclone makes a coarse preliminary classification and a screen makes the final fine cut can combine the high throughput and low cost of cyclones for the bulk of the separation duty with the density-independence and zero bypass of screens for the final product cut. Research comparing the performance and economics of these hybrid approaches is an active area of investigation.

Hydrocyclone selection and sizing framework

Cyclone diameter alone does not define performance. Sizing must reconcile capacity, target partition, water split, pressure, feed density, geometry, and the number of operating units.

Design question	Evidence required	Design implication
What classification result does the downstream process need?	Target partition curve, product-size distribution, bypass, and downstream recovery response	Defines the required cut and sharpness rather than relying only on a nominal P80.
How variable are flow and solids loading?	Time-based plant data or variability scenarios for flow, density, PSD, and rheology	Determines cluster turndown, spare units, control range, and the need for staged operation.
Can the pump and sump maintain stable feed?	Pump curve, available head, sump volume, control response, and line losses	Prevents selecting cyclones that cannot operate at stable pressure and density.
Which geometry is appropriate?	Vendor testwork or calibrated model covering inlet, vortex finder, apex, cone angle, and diameter	Links geometry changes to capacity, water split, cut size, and roping risk.
Is the slurry prone to wear, blockage, or unusual rheology?	Mineralogy, abrasion data, particle shape, viscosity, and oversize contamination	Changes material selection, inspection interval, apex design, screening, and model confidence.
How will performance be measured?	Sampling locations, pressure and density instruments, online size measurement, and survey procedure	Ensures the circuit can be diagnosed and controlled after commissioning.

Hydrocyclone troubleshooting matrix

Interpret pressure, density, flow pattern, and product sizing together. A single pressure reading cannot distinguish every failure mode.

Observed symptom	Likely mechanisms to investigate	Priority checks
Roping or a solid underflow discharge	Apex restriction, excessive solids loading, high feed density, or coarse oversize	Apex inspection, underflow pattern, feed density and PSD, pressure, and oversize-screen performance.
Excessively dilute spray underflow	Apex too large, low solids loading, low feed density, or unsuitable geometry	Apex size, feed solids, pressure, water additions, and underflow density.
Overflow becomes coarser	Pressure or density change, worn vortex finder, fewer operating units, feed PSD shift, or pump limitation	Pressure and density trends, active cyclone count, geometry wear, feed PSD, pump and sump behavior.
Too many fines report to underflow	High water recovery to underflow, bypass behavior, excessive feed density, or geometry mismatch	Water split, partition curve, feed density, apex and vortex-finder dimensions.
Cyclones in one cluster perform differently	Unequal feed distribution, blocked ports, wear differences, air ingress, or valve condition	Per-cyclone pressure and product samples, manifold inspection, valve position, and component measurements.

Final selection should use vendor or pilot testwork and a calibrated model where possible, followed by plant surveys using correctly timed and representative samples.

Additional Resources and Industry Links

1. Wills, B.A. & Finch, J.A. (2016). Wills’ Mineral Processing Technology, 8th Edition. Butterworth-Heinemann/Elsevier.
2. Plitt, L.R. (1976). A mathematical model of the hydrocyclone classifier. CIM Bulletin, 69(December), 114–123.
3. Nageswararao, K., Wiseman, D.M. & Napier-Munn, T.J. (2004). Two empirical hydrocyclone models revisited. Minerals Engineering, 17(5), 671–687.
4. University of Alaska Fairbanks AMIT 145: Lesson 2 — Classifying Cyclones. Mining Mill Operator Training.
5. 911Metallurgist: Hydrocyclone Design and Sizing Parameters — Calculations.
6. McLanahan Corporation: 6 Factors That Affect Hydrocyclone Performance.
7. Dündar, H., et al. (2024). Comparing the Performance of Hydrocyclones and High-Frequency Screens in an Industrial Grinding Circuit: Part I — Size Separation Assessments. Minerals, 14(7), 707. MDPI.
8. Benefits of semi-inverted hydrocyclones in closed grinding circuit — a pilot scale study. Minerals Engineering, 2025. ScienceDirect.
9. In-depth characterisation of hydrocyclones: Ascertaining the effect of geometry and operating conditions on their performance. Advanced Powder Technology, 2023. ScienceDirect.
10. Napier-Munn, T.J., Morrell, S., Morrison, R.D. & Kojovic, T. (1996). Mineral Comminution Circuits: Their Operation and Optimisation. JKMRC, The University of Queensland.