8 8.2. Markov Process-based Prognostics: Multi-state deterioration modelling 179 end-users like risk asset managers, where both expressiveness and interpretability are crucial. By implementing MOEAs, we successfully achieved this consistency. Furthermore, we observed improvements in convergence speed and scalability, which is critical for real-world engineering challenges involving numerous basic events. In hindsight, this proved to be an e!ective approach. However, applying MOEAs to larger datasets presents additional challenges. Evolutionary algorithms rely heavily on random mutations, and in large solution spaces, this randomness can require substantial computational power, making convergence di"cult. This limitation is a key drawback of using MOEAs for inferring FTs. For instance, most of our validation cases are derived from real-world engineering problems which, while representative, are relatively small, involving up to 24 Basic Events and 26 Minimal Cut Sets. Other real-world problems could be much larger, with incomplete and noisy datasets. Other type MOEAs, or advanced methods that rely less on randomness, combined with improvements in computational power and strategic implementations such as parallelisation, could enhance the e!ectiveness of MOEAs for inferring FTs from data. 8.2 Markov Process-based Prognostics: Multistate deterioration modelling Overview of the research problem in Part II Accounting for reliable deterioration models is key in Prognostics and Health Management (PHM), as they help predict failures and undesirable states, enabling timely optimal actions. Various types of deterioration models exist depending on the system/component/process of interest, and the literature is extensive on this subject. Part II of this dissertation focuses on deterioration modelling in sewer main systems, which are complex due to their scale, numerous components, and dependencies. We centre our attention on Multi-State Deterioration Models (MSDMs) using Markov chains, which aim to predict damage severity levels. These models have been addressed in the literature, but several aspects remain open for discussion, including model assumptions, and data availability. Recap on contributions in Part II In Part II, we used Markov chains to model Multi-State Deterioration (MSD) in sewer mains. Our contributions are three-fold:
RkJQdWJsaXNoZXIy MjY0ODMw