| Popis: |
Highway bridges are vital links within the highway network, representing a major long term infrastructure investment. A well managed bridge stock is therefore fundamental to the safety and availability of the highway network as a whole. In carrying out bridge management functions, Highway Authorities face growing pressures arising from inadequate budgets and greater accountability, when many of the existing bridges have reached the upper limits of their design life spans while being subjected to increasing and unprecedented traffic loading. There are many factors that influence the decision making process in bridge management, including funding and prioritisation decisions, and hence a MultiCriteria Decision Making (MCDM) approach is vital to ensure optimisation and a satisfactory trade-off between conflicting factors. The two key outputs of this thesis are the development of a bridge management framework for implementing best practices in Highway Authorities, and the development and testing of Artificial Intelligence (AI) models to aid multi-criteria decision making in bridge management. The case study approach was adopted, based on Manchester City Council's Bridges Practice, for mapping accepted good asset management practices in developing the bridge management framework; and for providing the vital bridge maintenance scheme data used in designing the AI decision support models. The most significant factors influencing decision making in bridge management were established through a nationwide questionnaire survey undertaken within the UK Highway Authorities' practicing bridge managers. Highway authority interviews were also conducted within the different types of Highway Authorities nationally and within the case study to verify existing bridge management practices and to inform the development of the bridge management framework. Several decision support models were developed using three different AI techniques, namely Artificial Neural Networks, Fuzzy Logic and Genetic Algorithms, as standalone models, within hybrid systems or both. The developed AI models and hybrid systems were validated using bridge maintenance schemes not used in the development phases, and found to be effective, to varying performance levels, in predicting the output Prioritisation Score for each bridge maintenance scheme within the test data set. The output bridge management framework was validated through an interview with Manchester's Roads and Bridges Manager, as a target end-user, and the framework was judged as a valuable tool capable of delivering the necessary upgrading in highway authority bridge management practices. |
