bedblend

An engine for TRUCK-BUILT run-of-mine stockpiles: haul trucks, a dozer, a pile that constrains its own construction, Gray-Thornton kinetic size segregation down the dumped face, a per-cell lot ledger with provenance, and reclaim by a loader working a face.

pip install bedblend

What it models, and what it deliberately does not

A heaped-fill, truck-dumped, pre-crusher stockpile: the kind built on a prepared pad by haul trucks and a dozer, lift by lift, and reclaimed by a loader. It is the inverse of an open pit. A pit cuts benches downward; a stockpile adds lifts upward, with the same primitives, a working level, a face at the angle of repose, a berm, and a ramp to the next level. Reclaim inverts it again.

It does not model chevron, windrow or cone-shell beds. Those are built by CONVEYOR STACKERS. Of the five pre-crusher stockpile types only blended-in-blended-out is a chevron, and the software that builds those beds is written for conveyor systems (Young and Rogers, Minerals 2021, 11, 636, figure 1). Trucks do not build a chevron bed, and offering those geometries alongside trucks is a category error that an earlier version of this library made.

import bedblend as bb

terrain = bb.Terrain.flat(60, 60, 2.5)
plan = bb.rectangular_yard(n_areas=1, area_width_m=90.0, area_length_m=90.0,
                           bench_height_m=6.0, n_benches=2, margin_m=30.0)
fleet = bb.Fleet.of(4, bb.TruckSpec(), (15.0, 75.0), repose_deg=37.0)
loads = bb.payloads_from(bb.dig_sequence(n_loads=600, seed=7), seed=7)

built = bb.build(terrain, plan, fleet, loads, repose_deg=37.0, seed=20260801)
print(f"{len(built.placed)} placed, {built.refusal_rate:.1%} refused")

face = bb.ReclaimFace(method=bb.ReclaimMethod.FULL_HEIGHT, position_m=30.0,
                      depth_m=10.0, width_m=90.0, loader=bb.LoaderSpec())
cuts = bb.campaign(built.terrain, built.model, face,
                   cut_tonnes=3000.0, n_cuts=24, repose_deg=37.0)

var_in = bb.tonnage_weighted_variance([l.grade for l in loads], [l.tonnes for l in loads])
var_out = bb.tonnage_weighted_variance([c.grade for c in cuts], [c.tonnes for c in cuts])
print(f"{len(cuts)} cuts, VRR {bb.vrr(var_in, var_out):.3f}")

The ratio is variance OUT over variance IN, on a tonnage base, so lower is better and 1.0 means the pile did nothing. Report it against the 1/N bound: the gap between them is what the pile fails to recover, and it is driven by the autocorrelation of the incoming stream. Layers only average if they are independent.

What is in it

What it does NOT claim

• The angle of repose is imposed, not emergent. A continuum height-field model, not DEM. It reproduces the geometry a given repose angle produces; it does not predict that angle from particle properties.
• Two segregation coefficients are anchored, not measured. PERCOLATION_COEFFICIENT and PECLET_DEFAULT are taken from the literature rather than fitted to a material. The size distribution down the face is SOLVED from the conservation law; where the mass lands and how much overruns the toe remain published operational observations, and neither touches the size split.
• Trajectory segregation is not modelled. Only kinetic sieving is. What rolls beyond the toe is reported as an overrun magnitude with a solver-derived composition.
• No fleet scheduling. One truck is routed per load and per cut; there is no queue, no cycle time and no spot time.
• It is not a blending optimizer. It evaluates a pile; it does not choose one.
• It is not plant metal accounting. It stops at the reclaimed stream.

Determinism

A run is a pure function of (parameters, seed). Everything stochastic comes from one 32-bit xorshift written out explicitly rather than using random or numpy, so the same stream can be reproduced bit for bit by an implementation in another language. That is what makes an in-browser mirror of this engine checkable against it. The core is dependency-free for the same reason.

References

• Gray, J.M.N.T. and Thornton, A.R. (2005), A theory for particle size segregation in shallow granular free-surface flows, Proc. R. Soc. A 461, 1447-1473. doi:10.1098/rspa.2004.1420
• Gray, J.M.N.T. and Chugunov, V.A. (2006), Particle-size segregation and diffusive remixing in shallow granular avalanches, J. Fluid Mech. 569, 365-398. doi:10.1017/S0022112006002977
• Young, A. and Rogers, W.P. (2021), Modelling of pre-crusher stockpiles, Minerals 11(6), 636. doi:10.3390/min11060636
• Young, A. and Rogers, W.P. (2022), Dump geometry from 28 UAV-surveyed dumps, Mining 2(1), 92-114. doi:10.3390/mining2010006
• Bak, P., Tang, C. and Wiesenfeld, K. (1987), Self-organized criticality: an explanation of 1/f noise, Phys. Rev. Lett. 59, 381. doi:10.1103/PhysRevLett.59.381

Licence

MIT. See LICENSE.