Geometry establishes the pressure zone
High-pressure grinding rolls (HPGR) apply compression across a fed bed between two counterrotating rolls. Geometry couples capacity and energy to breakage. Gap and material density influence solids flux, pressure and force influence power, and axial confinement influences the product distribution. Single-particle breakage can occur ahead of the compressed bed. Near the roll ends, loss of confinement can leave coarser material. The product from a press is thus assembled from regions of different loading rather than calculated at a single representative stress.
The model boundary determines the material streams available for plant calculations. A single-output model can calculate centre and edge PSDs for review and recombine them into one mass stream. A model with separately collected edge and centre streams can send edge product into an external recycle. Match the stream definitions when comparing cases. The central product of a press with side recycle can be finer than its total roll discharge even when grinding behaviour is unchanged.
Capacity and power under the Torres–Casali assumptions
The Torres–Casali HPGR representation starts from specified roll geometry and operating conditions. Let D be roll diameter, L roll length, g working gap and U peripheral velocity, in metres and metres per second as appropriate. Let ρₐ be feed bulk density and ρₑ extrusion density, in t/m³. The implementation evaluates a nip angle α from its cosine.
The geometry is physically meaningful only when the radicand is nonnegative and the cosine lies in the admissible range. A clipped implementation can keep a numerical calculation running beyond that range. Check the reported geometry and density for physical consistency. Critical size is x_c = g + D(1 − cos α). In this variant coarse feed above x_c is redistributed first by a single-particle breakage matrix and then joins the finer feed in the compressed bed.
Q_cap is capacity per roll pair in t/h, p is hydraulic pressure in bar, F is compression force in kN and P is power in kW. The coefficient 100 converts bar to kPa, which is kN/m². The capacity formula uses the extrusion density and a filled cross-section g × L moving at U. For multiple identical units, multiply per-unit capacity by their installed count on the same assumption. Power divided by predicted capacity is a modelled kWh/t duty. Measured power divided by actual dry solids throughput supplies a distinctly defined operating duty. When the model warns that incoming feed exceeds its predicted capacity, a PSD that balances the incoming feed does not establish a feasible operating condition.
The bed height in this geometry is z = (D/2) sin α and plug-flow exposure time is z/U in seconds. The code splits the roll in equal-width axial blocks. A normalized profile proportional to L² − 4yₖ² allocates total power at block centres yₖ measured from the roll middle. Selection rate by size and block follows the product of specific selection and the local power per solids inventory, with an hour-to-second conversion if the balance uses s⁻¹. The outlet size fractions are then the equal-width block average. Centre and edge diagnostics summarize subsets of the same axial calculation.
Three-zone and speed-solved variants
The calibrated Morrell–Tondo–Shi HPGR uses explicit pre-crushing, edge crushing and compressed-bed mechanisms. It converts the feed into a pre-crushed population, splits that population at a fitted edge fraction and processes the edge and bed streams with separate calibrated appearances before recombining them. It calculates a critical gap using compacted feed and flake density, and estimates edge mass fraction by γg/L for fitted coefficient γ. Capacity is a slip-corrected form of the extruded mass-flux expression, Q_cap = 3600 ρ_flake g L U c, with dimensionless multiplier c. Enter gross specific energy at the calibrated duty to estimate both feed and capacity power. The model can show how edge redistribution changes product sizing even at a fixed total duty.
The Campos HPGR solves for peripheral velocity given connected dry solids feed, a maximum allowable speed, effective extrusion geometry and an ejection correction. Stud penetration and the fraction of covered surface can modify the effective working gap. The model solves the throughput equation on its increasing physical branch.
Here fₑ is a fitted ejection fraction, φ is entered in percent, ν and t are dimensionless calibration parameters, and x_c and g are in metres. Check both the branch and speed cap because a mathematical trial near fₑ = 1 can inflate throughput outside the physically admissible region. In this implementation fₑ affects the hydraulic solve while the complete connected dry solids mass reports to the centre and edge products.
The gross specific energy comes from force, speed and a calibrated compression angle. An energy utilisation factor can reduce effective grinding energy as saturation increases. The axial power distribution uses a normalized Fourier profile, and the energy-driven population balance in each axial section resolves selection and appearance. Centre and edge collection are assigned by overlap with the appropriate width fractions. A single axial section necessarily has the same PSD shape in the two streams. Increasing resolution can reveal shape differences if the calibrated axial profile has significant variation.
| Representation | Inputs carrying key operating information | Product interpretation |
|---|---|---|
| Torres–Casali | Specified speed, gap, pressure and density inputs | Total outlet with axial centre and edge diagnostics |
| Morrell–Tondo–Shi | Calibrated pre-crush and separate zone appearances, edge and slip factors | Combined outlet after independent edge and bed calculations |
| Campos | Connected feed, speed limit, ejection law and energy utilisation | Distinct centre and edge material streams |
Worked duty and collection
Consider the Torres–Casali geometry D = 1.2 m, L = 1.2 m, g = 0.010 m, U = 1.5 m/s, ρₐ = 1.8 t/m³, ρₑ = 2.4 t/m³ and p = 55 bar. The nip expression gives α = 4.296°. Substitution gives these capacity, force and power estimates.
The mean calculated energy at capacity is 2.86 kWh/t. Bed height is approximately 0.04494 m and critical size 0.01337 m. This calculation establishes the operating variables. A calibrated selection function and breakage distribution are still required to predict a PSD.
For a separate two-product balance, suppose 200 t/h dry solids leave the rolls, 25% in the collected edge and 75% in the centre. The passing at 1 mm is 42% in the edge and 70% in the centre. Total passing at 1 mm is 0.25 × 0.42 + 0.75 × 0.70 = 0.63. Returning the edge to the same press changes the incoming rate and product interpretation. At steady state with no other external solids loss, the centre rate equals fresh feed. For a fixed fresh feed of 150 t/h and edge recycle fraction 0.25 of roll throughput, total feed is 150/(1−0.25) = 200 t/h. This exceeds the 155.52 t/h geometry in the first calculation, so combining those design assumptions is infeasible.
Fitting evidence and testing validity
Start with measured pressure, speed, gap, dry solids feed, extrusion density and power. Fit flow and power on that boundary before tuning the PSD. Use single-particle tests, bed breakage measurements and separate edge samples to distinguish pre-crushing, axial profile and appearance. A total product PSD may constrain their combined outcome while leaving individual zone parameters uncertain. If the task concerns side recycle, gather physical edge and centre samples. Weight those samples by their measured rates and compare with the whole discharge before fitting the detailed model.
A case can have a plausible power draw and produce the wrong distribution if the wrong ore is calibrated, the scope includes a return stream already counted in feed, or the sizing data use dispersed or compacted flakes inconsistently. Check capacity, residual solids by size, and whether PSD evidence covers the intended operating range. A well-fitted nominal case warrants additional comparison at different gap, speed or ore before scale-up.
Exercises and solutions
- At fixed geometry and extrusion density in the Torres–Casali throughput formula, what is the capacity change from increasing roll speed from 1.5 to 1.8 m/s?
- Total HPGR product has 68% passing 1 mm. Edge mass is 20% of outlet and has 48% passing 1 mm. Find centre passing.
- A 25% edge fraction is fully recycled. Find the maximum fresh solids feed at 155.52 t/h dry capacity.
Solutions. (1) Capacity is proportional to speed, so it rises by 20% to 186.624 t/h on the stated conditions. (2) Centre passing is (0.68 − 0.20 × 0.48)/0.80 = 0.73. (3) Fresh feed is 155.52 × 0.75 = 116.64 t/h when the press runs at capacity and the entire edge stream returns.
Selected references. Morrell, Tondo and Shi (1997), modelling and scale-up of high-pressure grinding rolls. Daniel and Morrell (2004), model verification and scale-up. Torres and Casali (2009), pressure-profile population balance. Campos et al. (2021), pressed-bed and scale-related refinements. Rodriguez et al. (2023), hybrid modelling of roll pressing.
