The society is increasingly demanding scientific accountability behind risk management of lifeline networks such as hazard mitigation planning, infrastructure maintenance and post-disaster responses. For rapid and condition-based risk management of lifeline networks, it is essential to have system reliability analysis (SRA) methods that can integrate analyses across physical scales, and can interface models and data from multiple fields of science and engineering smoothly for quantifying system-level risk. The proposed project will develop multi-scale SRA methods employing advanced network clustering algorithms for efficient, accurate and collaborative risk assessment of large-size networks. The project will also create a near-real-time network risk alert system through integration of a rapid SRA method with a hazard alert system to facilitate rapid decision support on hazard responses. In addition, an efficient time-varying network SRA method will be developed in which network reliability is continuously updated based on inspection results of network component deterioration in order to sustain the network reliability with optimal use of limited resources. The analysis methods and numerical tools developed in this project will help practitioners understand the hierarchical structure of lifeline networks and its impacts on network risk and decision making, develop network risk alert systems customized for actual risk management practice, and perform condition-based maintenance considering actual deterioration progress and its impacts on network-level risk. Education-focused research tasks include the development of interactive cyber-environment on network theory, virtual experiment on network downtime, interactive computer software simulating network flow and connectivity, and mobile phone applications to demonstrate IT-based risk management. The research results will be incorporated into the graduate level courses on risk and reliability of complex infrastructure systems. Active efforts will be made to recruit students from the groups that are underrepresented in science and technology fields using the well-established institutional fellowship programs.
The project developed new system reliability analysis (SRA) methods including selective recursive decomposition algorithm (S-RDA) and clustering-based network reliability analysis to promote risk-informed management and post-disaster operations of lifeline networks such as water, energy, gas, and transportation networks. The specific methods developed were designed to effectively identify the hierarchical nature of the network, and rapidly calculate the reliability of the network based on the current and future conditions of various components of the network. Because of the improvements of the speed of performing SRA, it is now possible to estimate the failure risk of large lifeline networks, and continuously update the risk based on infrastructure deterioration models as well as inspection data collected in the field. To further improve the speed of performing SRA on large lifeline networks, a new surrogate modeling approach was developed based on machine learning theories. The developed surrogate model can accurately predict the network failure based on the failures of individual components with great improvement in computational efficiency. Using the developed approach, one can now quickly approximate the probability of a network failure given the probability of individual components failing. Numerical experiments on the California gas distribution system show improvements in computational time by two orders of magnitude. Additionally, new techniques were developed to link data and information on seismic hazard and bridge fragility with traffic monitoring data and estimation algorithms to improve post-disaster traffic monitoring. Numerical experiments performed on road networks near the New Madrid Seismic Zone in New Madrid, Missouri demonstrated merits of such integration and a great potential of applications to other hazards and monitoring systems.
