Optimization-based procedures for underground mine planning

Abstract

There is need for greater collaboration across the disciplines of mining, geology, operations research, statistics, and computer science to improve underground mine planning. Active areas of research include, inter alia: (i) integrating design and scheduling -- including improved geotechnical modeling, and (ii) addressing the volatility of real-time operations through more robust schedules. We address (i) through examination of a strategic underground mine design and scheduling problem by considering an ore body partitioned into panels, each of which is extracted by a specific method. An integer programming model prescribes an optimal set of methods with which to extract each panel and the corresponding schedule to maximize the net present value. The solution we provide for a base-case industry data set results in a design and corresponding schedule with 44% scaled additional value, compared to the best industry-derived solution for this strategic planning model. Related specifically to (ii), we relax the assumption of perfect knowledge regarding value and duration of each activity in an underground mining operation and present a stochastic programming model whose tractability is questionable for realistic-sized instances, and demonstrate that by relaxing certain constraints and developing a heuristic that exploits the resulting mathematical structure, we can obtain good-quality solutions, feasible for practical time horizon lengths, even in the presence of the relaxed constraints, within several hours, at most. We further demonstrate empirically that the solution quality improves relative to solving a deterministic equivalent based on point estimates of value and duration data.