Real bulk-materials stockyards use chevron stacking to fight short-term grade swings in the ore feed (a "batch" arriving from different faces of a mine or different rail cars). The stacker makes repeated full-length passes over the pile, laying the material down in horizontal layers — each layer captures the feed's grade at a different moment in time. Longitudinal position along the pile is decoupled from the moment of deposit.
A bridge reclaimer then cuts a full vertical slice through every layer at once and blends them, so each tonne it draws is the average of many different moments of the original feed:
grade_out(x) = (1/N) · Σ grade_in(layer i at column x), i = 1..N
If the layer samples at a given column are reasonably independent, averaging N of them reduces the standard deviation by roughly √N (central-limit effect) — so a feed with high CV (coefficient of variation = std/mean) comes out the reclaim end far steadier. This simulation samples a genuinely noisy/drifting feed signal while stacking, stores every sample it wrote, and computes both CV numbers directly from that real data — nothing here is a canned "improvement" number.
- Feed grade variability — amplitude of the noise + slow drift injected into the incoming ore stream.
- Stacking layers — how many chevron passes build the pile (this is the blending ratio N).
- Reclaimer speed — how fast the bridge reclaimer sweeps across the finished pile, slicing and averaging.