Reliability assessment of structural dynamic systems subject to earthquake-induced tsunamis

• Main Author: Shao, Zhe
• Other Authors: Tiong Lee Kong, Robert
• Format: Theses and Dissertations
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering::Civil engineering::Structures and design
• Online Access: https://hdl.handle.net/10356/69064
• Institution: Nanyang Technological University

To date, it is neither practical nor possible to accurately predict such hazard and construct perfectly effective countermeasures against them. Therefore, it is of our great interest to quantify the structural risk caused by earthquake-induced tsunamis. Despite continuous advancement in numerical simulation of tsunami and wave-structure interaction, it still remains a computationally challenging task to evaluate the reliability of structural dynamic system, especially when uncertainties related to the system and its modelling are taken into account. In this study, attempts have been made from two perspectives. Firstly, the focus is concentrated on the physical numerical simulation of the tsunami wave-structure interaction. The uncertainty arises from both the stochastic process of earthquake source generation and the structural parameters. The stochastic earthquake source model is proposed to generate tsunami profile.Secondly, various approaches based on a novel integration of the Subset Simulation algorithm and two metamodels, i.e. modified moving least squares (MLS), as well as Gaussian processes (GP), are proposed to evaluate the reliability of the dynamic system. The metamodels are used to replace the computationally expensive physical numerical simulations. The results showcase the ability of the proposed stochastic earthquake source model and the numerical simulation tools in modelling the complete physical process. Furthermore, the novel integration of the Subset Simulation and metamodels demonstrates potential benefit in estimating failure probability of a highly nonlinear complex system. The effectiveness of the proposed algorithms is discussed by comparing the results with the results obtained from the original Subset Simulation without using the metamodels.