2 Chapter 1. Introduction 1.2 Main concepts This section provides a concise overview of the main concepts and relevant literature for this dissertation, acknowledging that some details are beyond its scope. Prognostics and Health Management (PHM), as described in Section 1.2.1, serves as the overarching concept linking the research directions explored in this dissertation. Figure 1.1 illustrates the main concepts discussed and their relationships. Prognostics and Health Management (PHM) Maintenance Optimisation Markov process-based Prognostics (Section 1.2.1) (Section 1.2.3) (Section 1.2.4) Sewer Deterioration Modelling Markov Chains for Multi-State Deterioration Reliability Modelling Markov Decision Process Deep Reinforcement Learning Fault Tree Analysis Multi-Objective Evolutionary Algorithms (Section 1.2.2) Figure 1.1: Main theoretical concepts used in this dissertation. Section 1.2.2 introduces Reliability Modelling, focusing on Fault Tree Analysis and the use of Multi-Objective Evolutionary Algorithms to automatically infer the structure of Fault Tree models. Section 1.2.3 introduces Markov Process-based Prognostics, with emphasis on Markov chains for Multi-State Deterioration with applications on Sewer Deterioration Modelling. Finally, Section 1.2.4 introduces Maintenance Optimisation with applications to sewer mains, where we use Markov Decision Process to cast the optimisation problem and approximate near-optimal maintenance policies via Deep Reinforcement Learning. 1.2.1 Prognostics and Health Management Prognostics and Health Management (PHM) is a holistic multi-disciplinary computation-based engineering paradigm that employs measurements, models, and software for the e!cient and comprehensive management of system health (Sheppard, Kaufman, and Wilmering, 2008; Lee, F. Wu, Zhao, et al., 2014; W. Zhang, D. Yang, and H. Wang, 2019; Fink, Q. Wang, Svensén, et al., 2020; Hu, Miao, Y. Si, et al., 2022; L. Zhang, Lin, Liu, et al., 2019; Zio, 2022). Notice that in the literature, the terms Predictive Maintenance (PdM) and Condition-Based Maintenance (CBM) greatly overlap with PHM. However, the use of PdM and CBM may be ambiguous and, in some cases, interchangeable (Esteban, Zafra, and Ventura, 2022). For consistency, in this dissertation, we adopt PHM as a more general term.
RkJQdWJsaXNoZXIy MjY0ODMw