Automated Scheduling of Mining Operation Tasks

Abstract: The task of scheduling mining operations is a strikingly tough task yet it is still largely done manually by hand or with the help of simple gantt planning tools. This thesis aim is to explore the feasibility of an automatic scheduling solution that can incorporate the constraints specific to mining operations. A constraint programming based solution is presented and evaluated based on its correctness, viability and performance. With its rich set of operators, the constraint programming library OR-Tools is able to capture most of the mining specific constraints and two different objective functions are developed to suit different use cases. One is the well established makespan objective which purpose is to minimize the completion time of the last task. The second objective function, named the sub goal deviation objective, minimizes the deviation from the overall production goal divided into sub goals.  The underlying scheduling problem is notoriously hard to solve optimally for large instances. This is supported by several related studies and also by experimental results. To mitigate the performance degradation for large scheduling instances, an iterative solver strategy is presented. With this strategy the scheduler is able to solve much larger instances and initial tests resulted in the same objective values as the optimal strategy. A rescheduling procedure is presented to support schedule maintenance due to unforeseen circumstances such as delays or machine breakdowns. It is concluded that automatic scheduling and rescheduling is feasible but that it first needs to be evaluated by experienced schedulers in the field before being applied in a production environment.