Development of a maintenance analytics framework for the diagnosis and prognosis of marine systems

Prognostics and Health Management (PHM) approaches are gaining popularity within recent years due to the growing need for enhancements in ship automation and intelligence. Although PHM technologies have been widely investigated and achieved a certain level of maturity in industries such as aerospace, manufacturing, and railway, it is an undeniable fact that the shipping sector is still in its infancy in this regard and further research is required on the matter. For this reason, the main aim of this thesis is to enable Smart Maintenance within the shipping sector. Accordingly, this thesis presents a Maintenance Analytics (MA) framework for marine systems. This framework is primarily constituted by three modules: 1) data pre-processing, which ensures the data quality and integrity required in the subsequent modules, 2) diagnostic analytics, which determines the current state of marine systems, and 3) predictive analytics, which aims to predict the Remaining Useful Life (RUL), thus establishing the future health state of marine systems. In total, eight novelties have been ascertained to contribute towards the analysis and formalisation of Machine Learning and Deep Learning approaches to ensure the applicability of Artificial Intelligence within the shipping sector, thus facilitating implementation of better maintenance strategies. To analyse the performance of such a novel MA framework, a total of eight case studies of distinct marine systems are introduced. Results demonstrate the importance of exploring, analysing, and formalising novel holistic MA frameworks to assist with decision-making processes related to maintenance strategies to guarantee the robustness and flexibility of O&M activities whilst facilitating both reliability and availability of marine systems, thus reducing downtime and operational cost and enhancing ship/company profitability. Through the implementation of the eight distinct case studies, the developed MA has demonstrated its high accuracy in detecting and identifying faults for determining the diagnosis and in predicting the RUL for the prognosis of marine machinery. For instance, the fault detection phase and fault identification phase of the diagnostic analytics module have achieved an average accuracy of 92.5% and 95%, respectively.Prognostics and Health Management (PHM) approaches are gaining popularity within recent years due to the growing need for enhancements in ship automation and intelligence. Although PHM technologies have been widely investigated and achieved a certain level of maturity in industries such as aerospace, manufacturing, and railway, it is an undeniable fact that the shipping sector is still in its infancy in this regard and further research is required on the matter. For this reason, the main aim of this thesis is to enable Smart Maintenance within the shipping sector. Accordingly, this thesis presents a Maintenance Analytics (MA) framework for marine systems. This framework is primarily constituted by three modules: 1) data pre-processing, which ensures the data quality and integrity required in the subsequent modules, 2) diagnostic analytics, which determines the current state of marine systems, and 3) predictive analytics, which aims to predict the Remaining Useful Life (RUL), thus establishing the future health state of marine systems. In total, eight novelties have been ascertained to contribute towards the analysis and formalisation of Machine Learning and Deep Learning approaches to ensure the applicability of Artificial Intelligence within the shipping sector, thus facilitating implementation of better maintenance strategies. To analyse the performance of such a novel MA framework, a total of eight case studies of distinct marine systems are introduced. Results demonstrate the importance of exploring, analysing, and formalising novel holistic MA frameworks to assist with decision-making processes related to maintenance strategies to guarantee the robustness and flexibility of O&M activities whilst facilitating both reliability and availability of marine systems, thus reducing downtime and operational cost and enhancing ship/company profitability. Through the implementation of the eight distinct case studies, the developed MA has demonstrated its high accuracy in detecting and identifying faults for determining the diagnosis and in predicting the RUL for the prognosis of marine machinery. For instance, the fault detection phase and fault identification phase of the diagnostic analytics module have achieved an average accuracy of 92.5% and 95%, respectively