Özel sıvılaşma düzeneği içinde çevrimsel basit kesme deneyleri altında kum davranışının sayısal modellenmesi

Liquefaction is one of the most catastrophic effects of earthquakes to the built environment. Liquefaction causes decrement of shear strength in fully saturated loose sands due to excess pore water pressure increment during a repeated loading or dynamic excitation, such as an earthquake. The devastating results of liquefaction can be listed as bearing capacity failures, lateral spreading, differential settlements, etc. In order to reduce liquefaction related failures, some mitigation techniques are implemented in practice. Induced-Partial Saturation (IPS) which has been recently proposed by Yegian et al.(2007) and Eseller-Bayat (2009) is a technique to mitigate liquefaction by generating air/gas in fully saturated liquefiable sand sites. An experimental study was performed to investigate excess pore water generations in partially saturated sands, that were tested in cyclic simple shear liquefaction box (CSSLB) by Eseller-Bayat (2009). This thesis study includes the initial tasks of a TUBİTAK funded project (No: 213M367) with a title ''Dynamic Response of Sands Mitigated by IPS (Induced Partial Saturation) under New and Existing Structures''. The primary goal of this project is to numerically model partially saturated sands and determine how much partial saturation should be induced in liquefiable areas and how accurate the RuPSS (excess pore water pressure ratio in partially saturated sands) empirical model predicts the liquefaction response of remediated sites by IPS. First, the shaking table tests and the CSSLB box setup are numerically modeled in FLAC3D. Fully saturated sand specimens are tested under cyclic simple shear conditions at a range of relative densities (Dr) and shear strain amplitudes (γ) and compared with the shaking table test results from excess pore water pressure ratio (ru) and number of cycles to liquefaction (NL) point of view. More numerical data are obtained for high effective stresses that could not be achieved under shaking table tests due to the limitations of the CSSLB setup. A formulation for NL is developed as a function of relative density (Dr), simple shear strain amplitude (γ) and initial effective stress (σ'v0). Then, partially saturated specimens are modeled in FLAC3D, by changing the bulk modulus of air-water mixture based on Koning (1963) equation and are compared with the shaking table test results in terms of maximum excess pore water pressure ratio (rumax) and number of cycles to rumax (Nmax). It is concluded that the rumax values obtained in numerical model analysis are lower for high degree s of saturation (S) and higher for lower degrees of saturation than experimental results. Also, Nmax values are achieved higher in numerical analysis results. Future research is needed for modelling the bulk modulus of air-water mixture for numerical modeling. A procedure is suggested in this study for determining the initiation of liquefaction by using NL formulation developed and equivalent number of earthquake cycles Nγ . If Nγ / NL ≥1 liquefaction initiates in that sand layer. The procedure is confirmed with the results of a free-field 15 m sand layer numerical model tested under an earthquake record with magnitude M. RuPSS model prediction is also compared with the free-field 15 m partially sand layer results. The excess pore water pressure ratio ru reached rumax in both results, however the RuPSS model predicted the rumax values higher than the numerical model results. Finally, important issues in numerical modeling of experimental and free-field fully and partially saturated sands for dynamic and liquefaction conditions in FLAC3D are presented in this study.