University of Tasmania
Browse

Smart design choices to reduce the vulnerability of naval vessels

Download (4.08 MB)
thesis
posted on 2023-05-28, 12:45 authored by Friebe, M
Survivability is the ability of a naval vessel to survive a combat incident by avoiding (susceptibility), withstanding (vulnerability) or recovering (recoverability). Vulnerability assessment is often divided into the structural and system vulnerability assessment. System vulnerability assessments are traditionally performed using manually built fault and success trees that model a simplified version of the functional failure relationships. This traditional approach has been very limiting, but more accurate and realistic methods were too computationally expensive to use. Furthermore, traditional vulnerability assessments also assume that the onboard systems are perfectly reliable and fully functional, which is a further simplification that may have significant consequences on vulnerability assessments. System reliability has never been included in traditional vulnerability assessment methods mainly due to the limitation in available computational power. However, with increasing readily available computing power, such enhancements are now realisable. For an accurate vulnerability assessment of a naval vessel, it is important to know the functional failure relationships between the systems of that vessel is not prone and subject to human erroneous input. Furthermore, to include system reliability into the vulnerability assessment helps to understand the actual vulnerability performance of a vessel better and to support naval architects to make design decisions with regards of longevity vulnerability enhancement at minimal cost. The objective of this research is to demonstrate and to develop a framework that can automatically generate, via machine learning, the functional failure relationships from an actual design of a naval vessel and that then identifies critical and sensitive components that negatively contribute to the vulnerability performance of the vessel. Once these failure relationships are derived, they are then used to model the system reliability with the help of Bayesian Network operations. In order to derive the machine learned failure relationships of an actual naval vessel and to determine the reliability effect of the naval vessel's equipment, the research is divided into three major methodological chapters. The first part investigates contemporary and state of the art vulnerability assessment techniques and uses a selected tool to perform an actual survivability assessment of a chosen system. This study also served as a basis to become familiar with the nature of the research domain. The second part extends the model of the naval vessel and performs a vulnerability assessment with further naval systems modelled to complete a holistic and comprehensive naval model. The results of this model are then analysed with a Bayesian machine learning algorithm and built into various Bayesian Network models. These Bayesian Network models are then used for a sensitivity analysis to identify critical systems and single point of failures. The third part uses the derived Bayesian Network from the previous part and utilizes the learned failure relationships of an actual vessel to include the reliability effect of the naval vessel's equipment into the survivability assessment. The results of this methodology are of diverse nature. The first study performing a state-of-the-art vulnerability assessment for various firemain layouts with different automation levels resulted in an overview comparing different firemain systems across various levels of automation and their according vulnerability performance. The second part of this study resulted in the development of a complete naval vessel and a framework that has the ability to analyse output results from a vulnerability assessment of that vessel. The framework automatically derives probabilistic failure relationships between the naval vessel's systems and to identify critical systems and single point of failures of that design. The third part of the methodology resulted in a study that demonstrates the proof of concept to include the naval vessel's systems reliability and to predict the naval vessel's vulnerability performance with respect to service time, resulting in a demonstration of the significance of system reliability in vulnerability assessments. The research has demonstrated the feasibility to use Bayesian Networks as a tool to analyse naval vessels and to improve their vulnerability performance. The developed framework uses Bayesian Networks to identify single point of failures, which when eliminated from the design, lead to an improved design. As the inputs from the survivability assessment are readily available, these inputs just have to be entered into the vulnerability analysing framework and their analysis is automated. This framework enables shipbuilders to quickly analyse and assess naval ships, which can then be done in less time and with fewer resources. Furthermore, the developed framework produces probabilistic functional failure relationships that, when supplied with the information about the naval vessel's system reliability, can estimate the degraded vulnerability performance of the naval vessel after a certain amount of service years. Thus, the results and outcome of this research can benefit the vulnerability assessment process as it allows for the quick identification of single points of failure and the ability to model the naval vessel's future service behaviour.

History

Publication status

  • Unpublished

Rights statement

Copyright 2020 the author Chapter 4 appears to be the equivalent of a pre-print version of an article published as: Friebe, M., Skahen, D., Aksu, S., 2019. A framework to improve the naval survivability design process based on the vulnerability of a platform's systems, Journal of ocean engineering, 173, 677-686

Repository Status

  • Open

Usage metrics

    Thesis collection

    Categories

    No categories selected

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC