Abstract

Given a block-model of an open cut mine, a production scheduling policy defines which blocks should be extracted, when they should be extracted, and what should be done with them once extracted (i.e., sent to a mill, waste-dump, stockpile, etc.). The conventional three-step heuristic for constructing such a scheduling is as follows: First, compute an ultimate-pit. Second, subdivide the ultimate-pit into phases. Third, schedule the blocks in each bench-phase, taking into account the mining, milling and market/refining capacities. Though each of these steps is in itself an optimization problem, the three steps, when put together, constitute a piece-meal approach to the full problem. Recent developments make it possible to implement a direct optimization methodology using integer programming (IP) that can tackle real-sized problems. In this article we first compare such a direct approach to a commercial implementation of the traditional methodology (Gemcom Whittle) and find that the direct approach yields solutions with significantly higher value. In an attempt to explain this difference, we modify the IP approach so as to generate solutions more similar to those obtained by the traditional approach. We find that this modified approach still outperforms the commercial solver. We conclude that the direct approach attains higher value primarily because the commercial solver’s heuristics are not effective in the problems considered. In fact, we find that the traditional methodology, if executed exactly, yields near-optimal solutions on all our benchmark problems. We do not consider blending, stockpiles, or operational spatial constraints (e.g., only “best-case” solutions).