open access

https://doi.org/10.14943/doctoral.k16537

https://hdl.handle.net/2115/95705

2007年新潟県中越沖地震による軟弱粘土地盤の長期沈下予測の確実性と不確実性に関する研究

英語

2025-06-30

Hokkaido University

vii, 110p.

During the 2007 Niigataken Chuetsu-oki Earthquake, ground liquefaction was prominent at the foot of sand dunes and old river channels, whereas clayey grounds exhibited no significant immediate failures. However, long-term settlement was observed at Shinbashi in Kashiwazaki City, where cumulative ground subsidence of 71 mm was recorded over 14 years from the surface to a depth of 23m. To investigate the mechanisms of both seismic deformation and long-term post-earthquake settlement, extensive ground investigations, boring surveys, and laboratory tests were conducted. Results revealed that the clay at the site was highly structured, strongly compressible, and extremely soft. Finite element (FE) simulations were performed using the Transformation Stress-Cyclic Mobility (TS-CM) constitutive model, incorporating subloading and superloading concepts. Model parameters were calibrated based on laboratory tests, and soil-water coupling elasto-plastic FE analysis was used to simulate ground subsidence. The simulations closely matched field observations, enabling forward predictions of post-earthquake subsidence. Findings indicated that the highly structured clay exhibited greater long-term consolidation potential than lower-structured clay due to significant excess pore water pressure generation during seismic loading and subsequent consolidation. This conclusion was further validated through laboratory consolidation tests. Despite the agreement between deterministic FE simulations and field measurements, uncertainty in geotechnical parameters necessitates a stochastic analysis framework for more reliable long-term settlement predictions. To address this, the stochastic finite element analysis (SFEA) software JRiveruncertainty was developed. The software supports parallel computing, integrates predefined finite element solvers with custom computational models, and incorporates multiple sampling methods (e.g., direct Monte Carlo, Latin Hypercube Sampling, Sobol sequence). It also accommodates various probability distributions (e.g., normal, lognormal, exponential, Poisson) and features sensitivity analysis, spatial variability analysis (random field generation), and the stochastic response surface method (SRSM), such as Polynomial Chaos Expansion (PCE). Stochastic analysis was applied to quantify the influence of uncertain parameters on subsidence behavior. Sobol global sensitivity analysis identified permeability coefficient, initial soil structure, and critical state stress ratio as the three most influential parameters affecting soil deformation and longterm subsidence. Monte Carlo simulations with varying sample sizes and distributions assessed settlement reliability, revealing that: - For an allowable settlement of 100 mm, the Reliability of the subsidence no more than 100 mm in 20 years is 79% (corresponding to a failure probability of 21%) and the Reliability of the subsidence no more than 100 mm in 25 years is 44% (corresponding to a failure probability of 56%), highlighting long-term risks in soft clay consolidation. - The model factor for settlements remained stable across different sampling techniques and sample sizes, simultaneously with the uncertainty propagation analysis both demonstrating robustness in numerical predictions. - Latin Hypercube Sampling (LHS) significantly reduced computational costs while maintaining accuracy, with LHS using 1000 samples achieving similar precision to the Sobol sequence with 5120 samples. - As for Stochastic Response Surface Method (SRSM), it allows us to obtain reliable uncertainty quantification results with significantly reduced computational cost with proper sample selection and regression modelling. These findings underscore the importance of incorporating uncertainty quantification in geotechnical engineering, particularly in seismic risk assessment and long-term resilience planning. By integrating stochastic analysis with traditional deterministic models, this research provides a more reliable and computationally efficient approach for evaluating subsidence risks in soft clay, aiding informed engineering design and decision-making in earthquake-prone regions.

2025-06-30

甲第16537号

博士(工学)

(主査) 准教授 磯部 公一, 教授 渡部 要一, 教授 西村 聡, 教授 宮森 保紀

工学院（環境フィールド工学専攻）

博士論文

Version of Record