1 1.4. Research questions 19 imprecise dating of transitions between condition states (Cherqui, Clemens-Meyer, Tscheikner-Gratl, et al., 2024), remains an under-explored research direction. Part III: Maintenance optimisation of multi-state components A follow-up research direction from the modelling of components with multi-state deterioration is how to use this information to support maintenance decisionmaking. In our case study with sewer mains, we find that their management has been approached in various ways, as detailed in Section III.3. These approaches include risk-based methods (Lee, C. Y. Park, Baek, et al., 2021), multi-objective optimisation problems (Elmasry, Zayed, and Hawari, 2019), Markov Decision Processes (Wirahadikusumah and Abraham, 2003), considering the network structure (Qasem and Jamil, 2021), and ML-based methods (Marc Ribalta and Rubión, 2023). Techniques such as RL, which are widely used for MPO applications (Ogunfowora and Najjaran, 2023; Marugán, 2023), remain largely underexplored in sewer asset management. Existing applications of RL in this field primarily address control problems (Yin, Leon, Sharifi, et al., n.d.) and the grouping of maintenance actions (Kerkkamp, Bukhsh, Y. Zhang, et al., 2022). A significant challenge in developing maintenance strategies for sewer pipes arises from the complex state space, defined by various system parameters, which often experiences “dimensionality explosion.” This complexity renders traditional exhaustive search techniques and exact solvers, such as STORM (Hensel, Junges, Katoen, et al., 2022), ine"cient. Consequently, there is a pressing need for methods capable of managing larger state spaces and delivering near-optimal policies by accounting for system characteristics. Our research addresses this gap by employing Deep Reinforcement Learning (DRL) for MPO in sewer systems. While it is well-known that DRL does not guarantee globally optimal policies, we explore the e!ectiveness of these techniques for the MPO of sewer mains. We consider within the problem formulation the stochastic nature of the Multi-State Deterioration of sewer mains and evaluate the impact of di!erent degradation models on policy e!ectiveness. 1.4 Research questions Each gap covered in this dissertation has a dedicated part with the following general research questions: - Part I: how to obtain e!cient and compact Fault Tree models from failure datasets in a robust and scalable manner? - Part II: how and to what extent is it possible to accurately model Multi-State Deterioration with applications in sewer mains?
RkJQdWJsaXNoZXIy MjY0ODMw