Liquefaction of sandy soil has been observed to cause significant damage to infrastructure during major earthquakes. Historical cases of liquefaction have typically occurred in sands containing some portion of fines particles, which are defined as 75μm or smaller in diameter. The effects of fines on the undrained behaviour of sand are not however fully understood, and this study therefore attempts to quantify these effects through the undrained testing of sand mixed with non-plastic fines sourced from Christchurch, New Zealand. The experimental program carried out during this study consisted of undrained monotonic and cyclic triaxial tests performed on three different mixtures of sand and fines: the Fitzgerald Bridge mixture (FBM), and two Pinnacles Sand mixtures (PSM1 and PSM2). The fines content of each host sand was systematically varied up to a maximum of 30%, with all test specimens being reconstituted using moist tamping deposition. The undrained test results from the FBM soils were interpreted using a range of different measures of initial state. When using void ratio and relative density, the addition of fines to the FBM sand caused more contractive behaviour for both monotonic and cyclic loadings. This resulted in lower strengths at the steady state of deformation, and lower liquefaction resistances. When the intergranular void ratio was used for the interpretation, the effect of additional fines was to cause less contractive response in the sand. The state parameter and state index were also used to interpret the undrained cyclic test results – these measures suggested that additional fines caused less contractive sand behaviour, the opposite to that observed when using the void ratio. This highlighted the dependency on the parameter chosen as a basis for the response comparison when determining the effects of fines, and pointed out a need to identify a measure that normalizes such effects. Based on the FBM undrained test results and interpretations, the equivalent granular void ratio, e*, was identified from the literature as a measure of initial state that normalizes the effects of fines on the undrained behaviour of sand up to a fines content of 30%. This is done through a parameter within the e* definition termed the fines influence factor, b, which quantifies the effects of fines from a value of zero (no effect) to one (same effect as sand particles). The value of b was also determined to be different when interpreting the steady state lines (bSSL) and cyclic resistance curves (bCR) respectively for a given mixture of sand and fines. The steady state lines and cyclic resistance curves of the FBM soils and a number of other sand-fines mixtures sourced from the literature were subsequently interpreted using the equivalent granular void ratio concept, with bSSL and bCR values being back-calculated from the respective test data sets. Based on these interpretations, it was concluded that e* was conceptually a useful parameter for characterizing and quantifying the effects of fines on the undrained behaviour of sand, assuming the fines influence factor value could be derived. To allow prediction of the fines influence factor values, bSSL and bCR were correlated with material and depositional properties of the presented sand-fines mixtures. It was found that as the size of the fines particles relative to the sand particles became smaller, the values of bSSL and bCR reduced, indicating lower effect of fines. The same trend was also observed as the angularity of the sand particles increased. The depositional method was found to influence the value of bCR, due to the sensitivity of cyclic loading to initial soil fabric. This led to bSSL being used as a reference for the effect of fines, with specimens prepared by moist tamping having bCR > bSSL, and specimens prepared by slurry deposition having bCR < bSSL. Finally the correlations of the fines influence factor values with material and depositional properties were used to define the simplified estimation method – a procedure capable of predicting the approximate steady state lines and cyclic resistance curves of a sand as the non-plastic fines content is increased up to 30%. The method was critically reviewed based on the undrained test results of the PSM1 and PSM2 soils. This review suggested the method could accurately predict undrained response curves as the fines content was raised, based on the PSM1 test results. It also however identified some key issues with the method, such as the inability to accurately predict the responses of highly non-uniform soils, a lack of consideration for the entire particle size distribution of a soil, and the fact the errors in the prediction of bSSL carry through into the prediction of bCR. Lastly some areas of further investigation relating to the method were highlighted, including the need to verify the method through testing of sandy soils sourced from outside the Christchurch area, and the need to correlate the value of bCR with additional soil fabrics / depositional methods.
Deformational properties of soil, in terms of modulus and damping, exert a great influence on seismic response of soil sites. However, these properties for sands containing some portion of fines particles have not been systematically addressed. In addition, simultaneous modelling of the modulus and damping behaviour of soils during cyclic loading is desirable. This study presents an experimental and computational investigation into the deformational properties of sands containing fines content in the context of site response analysis. The experimental investigation is carried on sandy soils sourced from Christchurch, New Zealand using a dynamic triaxial apparatus while the computational aspect is based on the framework of total-stress one-dimensional (1D) cyclic behaviour of soil. The experimental investigation focused on a systematic study on the deformational behaviour of sand with different amounts of fines content (particle diameter ≤ 75µm) under drained conditions. The silty sands were prepared by mixing clean sand with three different percentages of fines content. A series of bender element tests at small-strain range and stress-controlled dynamic triaxial tests at medium to high-strain ranges were conducted on samples of clean sand and silty sand. This allowed measurements of linear and nonlinear deformational properties of the same specimen for a wide strain range. The testing program was designed to quantify the effects of void ratio and fines content on the low-strain stiffness of the silty sand as well as on the nonlinear stress-strain relationship and corresponding shear modulus and damping properties as a function of cyclic shear strains. Shear wave velocity, Vs, and maximum shear modulus, Gmax, of silty sand was shown to be significantly smaller than the respective values for clean sands measured at the same void ratio, e, or same relative density, Dr. However, the test results showed that the difference in the level of nonlinearity between clean sand and silty sands was small. For loose samples prepared at an identical relative density, the behaviour of clean sand was slightly less nonlinear as compared to sandy soils with higher fines content. This difference in the nonlinear behaviour of clean sand and sandy soils was negligible for dense soils. Furthermore, no systematic influence of fines content on the material damping curve was observed for sands with fines content FC = 0 to 30%. In order to normalize the effects of fines on moduli of sands, equivalent granular void ratio, e*, was employed. This was done through quantifying the participation of fines content in the force transfer chain of the sand matrix. As such, a unified framework for modelling of the variability of shear wave velocity, Vs, (or shear modulus, Gmax) with void ratio was achieved for clean sands and sands with fines, irrespective of their fines content. Furthermore, modelling of the cyclic stress-strain behaviour based on this experimental program was investigated. The modelling effort focused on developing a simple constitutive model which simultaneously models the soil modulus and damping relationships with shear strains observed in laboratory tests. The backbone curve of the cyclic model was adopted based on a modified version of Kondner and Zelasko (MKZ) hyperbolic function, with a curvature coefficient, a. In order to simulate the hysteretic cycles, the conventional Masing rules (Pyke 1979) were revised. The parameter n, in the Masing’s criteria was assumed to be a function of material damping, h, measured in the laboratory. As such the modulus and damping produced by the numerical model could match the stress-strain behaviour observed in the laboratory over the course of this study. It was shown that the Masing parameter n, is strain-dependent and generally takes values of n ≤ 2. The model was then verified through element test simulations under different cyclic loadings. It was shown that the model could accurately simulate the modulus and the damping simultaneously. The model was then incorporated within the OpenSees computational platform and was used to scrutinize the effects of damping on one-dimensional seismic site response analysis. For this purpose, several strong motion stations which recorded the Canterbury earthquake sequence were selected. The soil profiles were modelled as semi-infinite horizontally layered deposits overlying a uniform half-space subjected to vertically propagating shear waves. The advantages and limitations of the nonlinear model in terms of simulating soil nonlinearity and associated material damping were further scrutinized. It was shown that generally, the conventional Masing criteria unconservatively may underestimate some response parameters such as spectral accelerations. This was shown to be due to larger hysteretic damping modelled by using conventional Masing criteria. In addition, maximum shear strains within the soil profiles were also computed smaller in comparison to the values calculated by the proposed model. Further analyses were performed to study the simulation of backbone curve beyond the strain ranges addressed in the experimental phase of this study. A key issue that was identified was that relying only on the modulus reduction curves to simulate the stress-strain behaviour of soil may not capture the actual soil strength at larger strains. Hence, strength properties of the soil layer should also be incorporated to accurately simulate the backbone curve.
