Underground production scheduling optimization with ventilation constraints

Underground production scheduling has traditionally been a time-consuming, manual task in which the goal is to achieve corporateor operation-defined production targets. This dissertation presents a generalized, mathematical formulation that results in a largescale integer optimization model. The model maximizes discounted gold ounces mined and determines the optimal or near-optimal sequence of activities related to the development, extraction and backfilling of an underground mine. Constraints include physical precedence and resource capacities. The research uses data from an existing underground mine; however, the model formulation has the ability to serve other underground mines with similarly structured data and provides the ability to customize constraints. Additionally, the model includes a constraint that treats available mine ventilation as a consumable resource. Diesel particulate matter, DPM, is a primary contaminant found in underground mining. Ventilation is used to dilute DPM below regulatory levels; however, mine ventilation is a limited resource, meaning that airflow through a mine cannot easily be increased once a ventilation system has been implemented. Large integer optimization models are traditionally solved using the branch-and-bound algorithm; however, results show the benefit of using a solver called OMP, originally designed for open pit mining applications, which integrates a specialized linear programming algorithm and a heuristic to induce intergrality. The author evaluates two solution methods, the CPLEX optimization package and the OMP Solver, and compares solution time. The OMP Solver schedules are compared to manual production schedules, with and without ventilation constraints. These comparisons show that with the OMP solver, it is possible to produce higher quality, implementable schedules in much less time relative to the manually generated schedules. Furthermore, the OMP solver outperforms CPLEX at this task.