This study focuses on aligning the business functions of maintainers and operators in the manufacturing industry for better costs/profits. In practice, maintainers and operators undertake maintenance operations management and production operations management respectively; such two types of operations management are closely intertwined with each other and hence should not be considered independently. In the maintenance planning literature, however, the balance between machine utilisation and increased risk of failure is rarely thoroughly discussed. Motivated by the maintenance planning problem in a large scale British coal-fired power plant, this study aims at developing a theoretical framework which facilitates aligning the business functions of maintainers and operators for better costs/profits in a relatively generic manufacturing industrial setting. Initially we consider a generic multi-asset production system. In such system we investigate the fundamental trade-offs between the decision making of maintainers and operators, and we further analyse how such trade-offs are underpinned by the existence of contracted period for sales and the associated potentially high penalty cost; based on such problem structure elicitation, we then develop a maintenance approach which integrates the operators' decision making as part of the maintainers' decision making in a conceptual framework and then further mathematically formulates such integrated maintenance planning problem as a Markov decision process (MDP). Such maintenance approach not only facilitates maintenance planning optimisation given existing machine utilisation behaviours, but also facilitates machine utilisation behaviour improvement as researchers/practitioners can conduct what-if analysis by changing the integrated utilisation behaviours in the MDP model. Next we consider a multi-level hierarchical physical structure which involves multiple aforementioned multi-asset production systems; such hierarchical structure is shared by many industrial cases. We scale up the MDP model to capture such complex hierarchical structure, in the context of the coal-fired power plant case. The resulted mathematical problem is too complex to be solved by exact methods, and we therefore develop a set of heuristics to solve the problem: we select a simulation-based computation heuristic and a value function approximation method from literature, and we further combine them with our own designed decomposition method and parameters number bounding method. We further discuss how the scaled-up mathematical model and heuristics can be applied to other industrial cases of interest. Finally, we conduct numerical tests to demonstrate the practical value of our maintenance approach, mathematical model and heuristics.
|Date of Award||31 May 2019|
- University Of Strathclyde
|Supervisor||Tim Bedford (Supervisor) & Kerem Akartunali (Supervisor)|