oreblocks: License-Free Synthetic Ore-Body Block Models with a Stamped Exact Ultimate-Pit Optimum

Abstract

Benchmarking an open-pit optimiser, a mine-dispatch simulator, or a teaching tool needs realistic ore-body block models with a known answer, but the standard public instances (the MineLib library) are granted for academic download only and cannot be redistributed, and real mine models are proprietary. We present oreblocks, a small pure-Python package that generates synthetic three-dimensional ore-body block models of the same nature as the public instances (seeded deposit archetypes with per-block grades, bench structure, slope precedence and net-value economics) and stamps each one with its exact ultimate-pit optimum. The optimum is computed by the classical reduction of the ultimate-pit problem to a maximum closure and thence to a minimum s-t cut (Picard's theorem), solved by a Dinic max-flow with self-verifying closure and value-identity checks, and every instance is emitted in exact MineLib .blocks/.prec/.upit format with a sidecar carrying the stamped optimum, so any solver can be validated against a known-by-construction answer without touching licensed data. We validate the solver independently: on all four deposit archetypes its pit value agrees with a totally-unimodular linear program solved by a separate method (SciPy HiGHS) to relative 1e-15, i.e. to machine precision, and the exact solve runs in under a second up to 32,000 blocks. The package is deterministic, MIT-licensed, and consumed by downstream pit, dispatch and haulage tools; every reported number regenerates from the archived validation run. Source code and reproducible artifacts (MIT): https://github.com/fsantibanezleal/CAOS_OreBlocks .