Abstract
Global trade relies heavily on seaborne transportation, highlighting the critical role of cargo ships in maintaining the global supply chain. However, the deterioration of ship hull structures over their lifecycle poses significant challenges for structural integrity, operational safety, and maintenance planning. Traditional preventive maintenance approaches, centred on periodic inspections, offer limited opportunities for timely, data-driven decision-making, particularly given the relatively infrequent surveys and the exponential progression of deterioration modes such as corrosion-induced thickness loss (CITL). Structural Health Monitoring (SHM) offers a promising alternative by using structural response data to provide continuous or event-triggered health-state information to support condition-based or predictive maintenance. In this thesis, a comprehensive strain-based SHM framework for CITL monitoring in ship hull structures is proposed and evaluated, with uncertainty quantificati ...
Global trade relies heavily on seaborne transportation, highlighting the critical role of cargo ships in maintaining the global supply chain. However, the deterioration of ship hull structures over their lifecycle poses significant challenges for structural integrity, operational safety, and maintenance planning. Traditional preventive maintenance approaches, centred on periodic inspections, offer limited opportunities for timely, data-driven decision-making, particularly given the relatively infrequent surveys and the exponential progression of deterioration modes such as corrosion-induced thickness loss (CITL). Structural Health Monitoring (SHM) offers a promising alternative by using structural response data to provide continuous or event-triggered health-state information to support condition-based or predictive maintenance. In this thesis, a comprehensive strain-based SHM framework for CITL monitoring in ship hull structures is proposed and evaluated, with uncertainty quantification (UQ) placed at its core. The framework encompasses tasks across the SHM hierarchy, from damage detection and quantification to prognosis, using both data-driven and model-based approaches. By integrating SHM outcomes into decision-making processes, the framework aims to enable a transition from traditional inspection-based maintenance to adaptive, data-centric strategies. Key contributions of this thesis include the demonstration of the feasibility of strain-based CITL monitoring for ship hull structures, the development of an optimal sensor placement (OSP) framework for SHM system design, and an evaluation of the value of information (VoI) to assess the benefits of investing in such systems. Computational challenges associated with model-based SHM for large-scale structures have also been addressed through surrogate modelling techniques, such as convolutional variational autoencoders (CVAEs), which enable uncertainty-informed structural response predictions with significantly reduced computational costs while maintaining accuracy. Although conducted primarily in a numerical environment—using a high-fidelity Finite Element (FE) model based on a realistic case study of a ship hull—this research incorporates experimental investigations on lab-scale specimens to address challenges not amenable to numerical investigations. These investigations explore the role of UQ in training data generation for data-driven SHM and examine model selection for model-based SHM tasks through Bayesian decision theory, evaluating its impact on SHM performance and downstream decision-making. The findings from these experimental campaigns extend beyond the specific context of ship hull SHM, contributing to the field of SHM more broadly. This thesis seeks to advance the integration of SHM into established hull maintenance practices, to ultimately bridge the gap between academic research and practical applications. By demonstrating the feasibility and value of SHM for ship hull structures, it provides a pathway for industry stakeholders to adopt more informed and adaptive maintenance strategies. Additionally, critical areas for future research have been identified, including extending the framework to other damage modes, refining decision-making tools, and addressing data scarcity through population-based SHM approaches.
show more
