Production Scheduling Optimisation Using Mathematical Programming and Machine Learning for Underground Mining Operations

Abstract

For decades, standalone mathematical models have dominated the optimisation of production schedules in underground mining. However, recurrent inconsistencies between schedule forecasts and actual production suggest the robustness of optimal solutions is impaired by inaccurate inputs. To address this, the study leverages emerging Machine Learning (ML) applications to generate ML-predicted dilution schedule inputs on a per-stope granularity, improving the schedule’s forecast precision by at least 3.6% and 1.6% for mined tonnes and Net Present Value, respectively.