What has to be identified
A crusher relates the connected feed to a new population of particles. Its minimum duty is to preserve solids by component, account for water, and redistribute size. Crusher power can be a separate model or a consequence of the chosen breakage representation. A useful distinction is how a model predicts the product. A prescribed-product crusher assigns a measured product curve. An empirical primary crusher generates a curve from a machine opening and fitted product-shape parameters. A matrix crusher predicts how individual portions of the feed enter breakage and how their daughter fragments escape. Each representation answers a different calibration question.
In a closed circuit with a screen, the single-pass crusher product is the discharge directly before screening. The external oversize return is a separate stream. Internal passages in a breakage zone are handled inside the crusher model. In the circuit balance, the external recycle contributes directly to feed. Double-counting returned oversize in both places exaggerates exposure to breakage.
Classification, appearance and internal recirculation
Let f be a column of feed mass flow rates by size interval for one solid component. Let C be a diagonal matrix of fractions selected for breakage during an internal pass. Let B be an appearance matrix with each column describing the fractions formed when its parent breaks. Index from coarse to fine. Column j has a same-size fraction Bⱼⱼ and can have entries in finer rows i > j. For a strictly size-reducing process, entries in coarser rows are zero. Use a consistent basis for all components and check that every column sums to one.
Let u denote the total internal flow presented for selection, including the products of internal breakage events. Let p be the measured discharge corresponding to particles escaping selection.
A general solver may compute p = (I − C)(I − BC)⁻¹ f. A triangular, class-resolved calculation is clearer for this physical structure. For each class i, form its available mass Uᵢ, including flow from all already processed coarser classes. Mass leaves a class by bypassing breakage or transferring to a finer interval.
Here Lᵢ is the fraction of a pass that removes material from class i, pᵢ is direct discharged mass and fᵢ→ₖ is the mass transferred to finer class k. The outgoing stream and all daughter transfers from that class sum to Uᵢ because its bypass probability plus the probabilities of daughters finer than i equals Lᵢ. A populated class with Cᵢ = 1 and Bᵢᵢ = 1 is trapped. The correct response is to expose an infeasible breakage or selection calibration. Adding numerical regularization can hide the physical problem.
A common classification function uses thresholds K1 and K2 and positive exponent K3. It sets C = 0 at or below K1 and C = 1 at or above K2. The transition follows a power law.
Use the representative class size convention of the calibrated model. Replacing a lower aperture with the geometric mean can shift selection substantially on a wide mesh. In the Modified Whiten implementation the thresholds are correlated to the closed-side setting by K1 = α CSS and K2 = β CSS + γ. Setting changes move the selection range. A fitted γ is a length and must use the same units as CSS. The Whiten Crusher also couples its classification settings to its operating correlation, so opening and rate requirements belong in the calibration. Generic Matrix Crusher uses the same conservation architecture with a Broadbent–Callcott type appearance function, making it useful when detailed component breakage tests are unavailable.
Worked calculation
A machine accepts 90 t/h coarse and 10 t/h middle class, with no fine feed. In the coarse class, selection is 0.8 and the appearance fractions to coarse, middle and fine are 0.25, 0.50 and 0.25. In the middle class, selection is 0.4 and the appearance fractions to middle and fine are 0.25 and 0.75. Fines bypass completely.
For coarse material, L₁ = 1 − 0.8 × 0.25 = 0.8. Coarse discharge is 90 × 0.2/0.8 = 22.5 t/h. Middle gains 90 × 0.8 × 0.50/0.8 = 45 t/h and fines gain 22.5 t/h. In the middle class, U₂ = 10 + 45 = 55 t/h and L₂ = 1 − 0.4 × 0.25 = 0.9. Middle-class discharge is 55 × 0.6/0.9 = 36.667 t/h, and its additional fine transfer is 55 × 0.4 × 0.75/0.9 = 18.333 t/h. The total product follows by adding each contribution to its destination class.
| Size interval | Product solids (t/h) |
|---|---|
| Coarse | 22.500 |
| Middle | 36.667 |
| Fine | 40.833 |
| Total | 100.000 |
If feed water is 12 t/h and the crusher has no added or removed water, 12 t/h reports to product. If there are two solid components with different matrices, repeat the calculation for each. Summing component products gives the bulk PSD while their local fractions define composition by size.
Impact calibration, empirical curves and measured products
A calibrated Kojovic impact representation links impact energy to component appearance. With a rotor at n rpm and effective radius r in metres, tip speed is 2πrn/60. If the local model uses the tip kinetic energy as an impact-energy proxy, E = v²/7200 kWh/t dry solids. A calibrated material relation connects impact energy to t₁₀.
A is entered as a percentage, b has units of t/kWh, and t₁₀ is the mass fraction passing one tenth of the parent size after breakage. Measured t-family rows at relative apertures can provide interpolation about that fraction. Differences between adjacent cumulative fractions supply individual daughter masses. A row must preserve monotonic passing from its smallest to largest relative aperture. Outside the measured t₁₀ range an extrapolation may remain mathematically bounded while failing to represent the ore. A fit to v describes the energy input assumption, while A, b and the appearance table jointly describe the material response.
King Primary Crusher uses an independent empirical product curve for each component. Its normalized aperture is x/OSS, with x and the open-side setting in the same length unit. It has two size regions joined at x/OSS = 0.5 and sets their shape from a component product type index. An optional feed-fines treatment represents fines entering along with the empirical product. This family requires a reliable opening and measured product evidence to fit the shape. A Vendor Known Product Crusher instead fixes a known product specification. It suits cases where the relevant PSD is available and the survey does not identify internal breakage.
What a crusher survey can identify
One feed and product survey can support a product curve or constrain a combined selection/appearance response. It usually cannot identify all thresholds, an unrestricted appearance matrix and power independently. If crusher discharge is coarser than expected, lower selection and less intense appearance can produce similar results. Fit the appearance matrix first with single-particle or drop-weight evidence where possible. Fit selection settings against surveys at more than one opening or load. A normalized residual measures the error conditional on available data. Compare alternate parameter sets to test for nonuniqueness.
Use a reference survey to reconcile overall flows and compare coarse bypass, shoulder shape and fine fraction. Add data near a suspected setting transition. Record whether streams were sampled at steady conditions and on the same measurement boundary. Product PSDs are more informative about retained size masses than power if no calibrated energy expression is connected to the breakage model.
Exercises and solutions
- For one occupied class C = 0.9 and Bᵢᵢ = 0.4. Find the fraction of available mass discharged directly from that class after internal recirculation.
- In the worked example set middle selection to zero, leaving other values fixed. Find all product class flows.
- CSS is 20 mm, α = 0.5, β = 1.3 and γ = 4 mm. Compute K1, K2 and C for 20 mm at K3 = 2.
Solutions. (1) L = 1 − 0.36 = 0.64 and the direct fraction is 0.10/0.64 = 0.15625. The remaining 0.84375 transfers finer. (2) Coarse remains 22.5 t/h. Middle discharges all 55 t/h that entered it, and fines receive 22.5 t/h from coarse breakage alone. (3) K1 = 10 mm, K2 = 30 mm, and C = 1 − (10/20)² = 0.75.
Selected references. Broadbent and Callcott (1956), size distributions from breakage. Whiten, Walter and White (1979), breakage functions for crusher models. Kojovic, Napier-Munn and Andersen (1997), laboratory-determined energy and breakage functions.
