Efficient methodology and algorithms for resilience analysis of critical infrastructure systems subjected to natural hazards

• Main Author: Hao, Changyu
• Other Authors: Law Wing-Keung, Adrian
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Computer science and engineering::Mathematics of computing::Probability and statistics; Engineering::Computer science and engineering::Computing methodologies::Simulation and modeling
• Online Access: https://hdl.handle.net/10356/155433
• Institution: Nanyang Technological University

Critical infrastructure systems (CIS) are highly connected to human life and industrial productions. Well protected and resilient critical infrastructure systems are the key components of modern smart cities. In this research, we answer the following key questions on resilience of critical infrastructure systems (CIS): (1) how to characterize the multi-criteria system resilience behaviour under uncertainties? (2) how to efficiently simulate the system performance under a large number of uncertain scenarios? (3) how can we optimize the system design for better resilience performance under uncertainties? (4) how do we systematically improve and update our knowledge about the system? A multiple Complementary Cumulative Distribution Function (CCDF)-based stochastic resilience assessment method is proposed. Critical infrastructure systems can be subjected to stochastic environmental interruptions, and given an interruptive event, the system's behaviours are often uncertain. To take such uncertainties into consideration, the proposed resilience assessment method quantifies the system resilience using multiple metrics, each having explicit physical meanings and expressed as a complimentary cumulative distribution function (CCDF). The proposed method provides explicit information of the system behaviour when the system is subjected to a broad spectrum of stochastic events with different degrees of rarity and extremeness. The proposed method can provide decision makers a bird's eye view of the system resilience performance under the attacks of future uncertain events and guide the allocation of recovery resources as well as system enhancement strategies. Resilience assessment with uncertainties being considered will inevitably bring about higher computation requirements if a simulation-based method is adopted to estimate the CCDF curves, especially when the tail parts of the CCDF curves are of interest. It is a non-trivial or even computationally prohibitive task if brute-force Monte Carlo simulation is adopted. In this research, the state-of-the-art stochastic simulation methods are adapted to facilitate the analysis of the system resilience considering the uncertainty involved. A new scheme of how the samples are used is proposed to improve the estimation of CCDF curves. In our illustrative example, about 10 times of computational savings are achieve with the improved method compared with the brute-force Monte Carlo simulation. Resilience optimization in the framework of multi-objective optimization is studied. We propose a framework optimizing the exceedance probabilities directly, where the objectives are probability integrals. We solve the computational burden by an off-line strategy: preparing approximations to the objectives before the multi-objective optimization starts. These approximations are efficiently obtained by adapting and improving a density estimation based approximation method. A realistic urban drainage system (UDS) is shown as an illustration of the multi-objective optimization problem with resilience as objectives. We proposed a novel efficient Bayesian model updating method for systematically improving and updating our knowledge about the system. We observed significant computational saving with our proposed Bayesian model updating method, and the method is also more robust to smaller sample size. With this method, stakeholders can quickly obtain the state-of-the-art parameter of the system and be more confident in making decisions. To summarize, this thesis provides a series of tools for (1) resilience assessment under a whole range of uncertain events, (2) efficient simulation method for resilience analysis, (3) design optimization with resilience as the objectives, (4) system state updating and improving our knowledge about the system.