TY - BOOK

T1 - Computational modeling and uncertainty quantification of geological hazards under dynamic loading

AU - Wang, Ruohan

PY - 2024/8/29

Y1 - 2024/8/29

N2 - Earthquake-induced landslides, among the significant geological hazards, result in substantial property damage annually. At present, the investigation of landslide analysis and failure patterns through computational methodologies continues to be an active research area within geotechnical engineering. The entire evolution process of landslides under dynamic loading are widely recognized as large-deformation processes. However, when considering the large deformation and failure behavior of slopes, such as landslides and collapses, especially under dynamic loading, numerical analyses conducted using traditional computational modeling tools (e.g., finite element method) often encounter convergence difficulties. The underlying reason is that the distorted mesh under large or even discontinuous deformation will cause severe numerical singularity problems. Hence, there is an urgent need to develop precise computational modeling tools to address these problems. On the other hand, it is widely believed that landslide issues are primarily affected by uncertainty, which manifests at various stages of analysis, ranging from site characterization to material property evaluation, and from design to consequence assessment. Traditional geological hazard analysis methods struggle to effectively quantify the uncertainties induced by these complex factors. The probability distribution and statistical characteristics of geotechnical parameters are primarily estimated from exploration testing samples. However, due to technical and economic limitations, these samples are often scarce and unevenly distributed, failing to fully represent the spatial variability of parameters. Even efficient surrogate model methods may struggle due to the curse of dimensionality, limiting their effectiveness. Current research has yet to effectively tackle the high-dimensional problems arising from spatial discretization and the time-consuming nature of repetitive calculations in geotechnical numerical models. Therefore, there is an urgent need to develop efficient methods for probabilistic inversion of spatially variable soil parameters. Hence, this thesis primarily involves addressing two key issues: advancing appropriate numerical tools and systematic uncertainty quantification in landslide hazards. To solve the above two key issues, this thesis reports four contributions mainly focusing on a robust technology for landslide analysis that can accurately simulate the entire evolution process and quantify the impact from a series of uncertainty source, especially from a numerical perspective. As for the first key issue, my first contribution was proposing a computational approach for large-deformation analysis leveraging non-ordinary state-based peridynamics (NOSBPD) features, whereas an effective procedure had been proposed to couple the random field theory with the NOSBPD algorithm, which enabled an accurate assessment of the risk induced by the spatial variability on geotechnical structures. As the second development, to evaluate the probability of landslide risk and seismic reliability under stochastic pulse-like ground motions, a novel run-out distance assessment framework aligned with target spectrum specified by codes was established as an extension of the proposed computational modeling technology. For the second key issue, uncertainty quantification in geological hazards, the third contribution proposed a real-time model-updating-based adaptive approach. This approach incorporated random field theory and introduced the concept of dimensionality reduction using the Karhunen-Loève expansion method for geotechnical structures subjected to sequential seismic ground motions. Specifically, leveraging data from shaking table tests, the proposed approach accurately predicted the seismic responses of geotechnical structures and performed inverse analysis on the spatial distribution field of real-site soil parameters. Additionally, the fourth contribution conducted a series of investigations focusing on exploring the practical applications of random fields in geotechnical engineering. Specifically, for embankment dam structures, the inherent spatial variability of hydraulic conductivity was quantitatively assessed for its impact on the phenomenon of piping leading to geotechnical structure failure. In summary, this thesis offers a series of original contributions focusing on the computational modeling and uncertainty quantification of geotechnical engineering structures and systems. The results presented in this thesis indicate that the methods proposed are valuable tools to support various practical processes related to precise computational modeling and systematic uncertainty quantification in geotechnical engineering.

AB - Earthquake-induced landslides, among the significant geological hazards, result in substantial property damage annually. At present, the investigation of landslide analysis and failure patterns through computational methodologies continues to be an active research area within geotechnical engineering. The entire evolution process of landslides under dynamic loading are widely recognized as large-deformation processes. However, when considering the large deformation and failure behavior of slopes, such as landslides and collapses, especially under dynamic loading, numerical analyses conducted using traditional computational modeling tools (e.g., finite element method) often encounter convergence difficulties. The underlying reason is that the distorted mesh under large or even discontinuous deformation will cause severe numerical singularity problems. Hence, there is an urgent need to develop precise computational modeling tools to address these problems. On the other hand, it is widely believed that landslide issues are primarily affected by uncertainty, which manifests at various stages of analysis, ranging from site characterization to material property evaluation, and from design to consequence assessment. Traditional geological hazard analysis methods struggle to effectively quantify the uncertainties induced by these complex factors. The probability distribution and statistical characteristics of geotechnical parameters are primarily estimated from exploration testing samples. However, due to technical and economic limitations, these samples are often scarce and unevenly distributed, failing to fully represent the spatial variability of parameters. Even efficient surrogate model methods may struggle due to the curse of dimensionality, limiting their effectiveness. Current research has yet to effectively tackle the high-dimensional problems arising from spatial discretization and the time-consuming nature of repetitive calculations in geotechnical numerical models. Therefore, there is an urgent need to develop efficient methods for probabilistic inversion of spatially variable soil parameters. Hence, this thesis primarily involves addressing two key issues: advancing appropriate numerical tools and systematic uncertainty quantification in landslide hazards. To solve the above two key issues, this thesis reports four contributions mainly focusing on a robust technology for landslide analysis that can accurately simulate the entire evolution process and quantify the impact from a series of uncertainty source, especially from a numerical perspective. As for the first key issue, my first contribution was proposing a computational approach for large-deformation analysis leveraging non-ordinary state-based peridynamics (NOSBPD) features, whereas an effective procedure had been proposed to couple the random field theory with the NOSBPD algorithm, which enabled an accurate assessment of the risk induced by the spatial variability on geotechnical structures. As the second development, to evaluate the probability of landslide risk and seismic reliability under stochastic pulse-like ground motions, a novel run-out distance assessment framework aligned with target spectrum specified by codes was established as an extension of the proposed computational modeling technology. For the second key issue, uncertainty quantification in geological hazards, the third contribution proposed a real-time model-updating-based adaptive approach. This approach incorporated random field theory and introduced the concept of dimensionality reduction using the Karhunen-Loève expansion method for geotechnical structures subjected to sequential seismic ground motions. Specifically, leveraging data from shaking table tests, the proposed approach accurately predicted the seismic responses of geotechnical structures and performed inverse analysis on the spatial distribution field of real-site soil parameters. Additionally, the fourth contribution conducted a series of investigations focusing on exploring the practical applications of random fields in geotechnical engineering. Specifically, for embankment dam structures, the inherent spatial variability of hydraulic conductivity was quantitatively assessed for its impact on the phenomenon of piping leading to geotechnical structure failure. In summary, this thesis offers a series of original contributions focusing on the computational modeling and uncertainty quantification of geotechnical engineering structures and systems. The results presented in this thesis indicate that the methods proposed are valuable tools to support various practical processes related to precise computational modeling and systematic uncertainty quantification in geotechnical engineering.

U2 - 10.15488/17928

DO - 10.15488/17928

M3 - Doctoral thesis

CY - Hannover

ER -