Honours Thesis by: Thomas J. Neaverson
Academic Supervisors: Dr Stefan Williams & A/Prof. Guodong Shi
Industry Supervisors: Mr Brad Waldon & Mr Marcus Valastro
The scheduling of maintenance work in heavy engineering industries is a complex managerial and logistical challenge. In cases where maintenance requires the shutdown of a plant, it is desirable to schedule tasks in a way that minimises the total shutdown period. However, other considerations, such as priority, resource capacity and precedence requirements may compete with this goal. This thesis investigates the optimisation of the maintenance scheduling process through the development of a computerised maintenance management system. A modular design structure is proposed for storing, scheduling and visualising maintenance work information. Particular focus is given to data driven decision-making processes for scheduling of tasks in large-scale maintenance operations. The study employs a multi-faceted approach, comparing and exploring exact and meta-heuristic techniques to enhance scheduling accuracy and efficiency. The findings highlight the efficacy of a simulated annealing algorithm for near-optimal schedule generation. This approach is generalised to cases where the duration of work is uncertain, which can cause conflicts that render a previously optimal schedule infeasible. The formulation for a Markov decision process which is capable of generating the optimal re-scheduling policy under the presence of uncertainty is proposed.
