Ore Sorter Cutoff Lab - 2D Grade Scatter

Every rock is a point: true grade against scanner reading. Drag the cutoff and watch the ore you lose and the waste you hoist.

Switch to 3D

Presets

Playback

Ore and sorter

Same parameters as the 3D drift.

Economics

Sample

More rocks smooth the scatter.

Layers

Model checks

The science

Why sort underground

Every tonne of barren rock hoisted up a shaft costs money twice: once for hoisting and haulage, and again for crushing and grinding at the mill. A sensor-based sorter on a conveyor measures each rock (here a generic X-ray or near-infrared scanner), and an air-jet bank blows rocks whose measured grade is below a cutoff c into a waste bin, from where they go back underground as backfill. Only the accepted stream rides the skip to surface.

The model

True rock grades are log-normal, ln g ~ N(μ, s2) with mean G and coefficient of variation cv (s2 = ln(1 + cv2)). The scanner reads ln m = ln g + e with e ~ N(0, σ2). A rock is kept when m ≥ c. Counting statistics make the noise fall with dwell time, so σ2 = σ02 v / vref - a faster belt is a noisier scan. Closed forms: the mass kept is A = 1 − Φ((ln c − μ) / √(s2+σ2)) and the copper kept (the size-biased distribution of ln g is N(μ+s2, s2)) is R = 1 − Φ((ln c − μ − s2) / √(s2+σ2)). Net value per tonne of feed is V = R·G·P·r/100 − csort − A(choist + cmill) − (1−A)cbackfill (P = metal price, r = mill recovery). A perfect sorter would use the break-even grade (choist + cmill − cbackfill) / (P r / 100); with noise the best cutoff shifts, and the page finds it by golden-section search and confirms it with a brute-force scan.

Honest simplifications: every rock is treated as the same mass (real sorters count particles of different sizes); one grade and one metal; no liberation or surface-coating effects; the scanner noise law and every unit cost are typical illustrative values, not data from a mine; the animated rocks each stand for 10 kg and the skip is shown filling after 28 rocks so the cycle is visible. Percentages and USD figures are model outputs for learning, not an engineering study.

What to try
  • Drag the white cutoff line up and down on the plot: the four quadrants (ore kept, ore lost, waste hoisted, waste ejected) and the ROC point move together.
  • Switch the break-even line off and on: it marks the grade at which a rock just pays for hoisting and milling.
  • Compare the blue ROC curve with the dashed noise-free one: the gap is what belt speed costs you.
  • Press Draw a new sample and watch the tallies wobble around the closed-form values.
ROC curve of the sorter
Grade-recovery trade-off
Misclassified mass against belt speed
Made with 🟠 Blender