Numerical Simulation of Dynamic Soil-Pile-Structure Interactive Behavior Observed in Centrifuge Tests

Abstract

Being able to predict the dynamic behavior of the pile foundations used in many civil structures is a very important matter because it widely affects the performance of the entire structure. Several methods are used to evaluate the seismic behavior of such piles, including site investigation, physical model tests, and numerical modeling. The effectiveness of numerical modeling has recently become more important in light of the time-consuming, procedurally complex, and expensive nature of physical model tests. Most previous research which simulated the seismic behavior of piles using the numerical modeling method adopted a simplified approach which is relatively easy to apply but can lead to inaccurate results due to the many assumptions such an approach requires. On the other hand, three-dimensional continuum modeling is a computationally complex and time consuming process but is the most direct approach and able to obtain reliable and accurate results. Therefore, in this study, three-dimensional continuum modeling is used to evaluate dynamic soil-pile structure interaction under earthquake loading. First, three-dimensional dynamic numerical modeling based on the finite difference method was carried out to simulate the dynamic behavior of a soilpile-structure system in dry sand observed in dynamic centrifuge model tests performed by Yoo (2013). The three-dimensional numerical model was formulated in a time domain to effectively simulate the nonlinear behavior of soil using the commercial finite difference code, FLAC3D. As a modeling methodology, the Mohr-Coulomb model was adopted as a soil constitutive model. Soil nonlinearity was considered by adopting a hysteretic damping which can simulate the nonlinear reduction of soil shear modulus according to shear strain. An interface model which can simulate slip and separation between soil and piles was used and the stiffness of the interface material was determined with consideration given to the nonlinear relation of soil by means of a user-defined FISH function. A simplified continuum modeling method was adopted as a boundary condition. In this method, the soil medium was divided into a near field and a far field, the latter of which is not affected by soil-pile-structure interaction (SPSI). The mesh was created only for the near field to reduce the computing time. A far-field response was applied as a boundary condition at the boundary of the near field. Calibration of the proposed modeling method was carried out to minimize error and increase the reliability of the results generated by the numerical simulation. The results of a centrifuge test were used to demonstrate the capability of the model to reliably analyze a pile foundation under earthquake loading. The peak bending moment and maximum lateral pile displacement along the depth obtained from centrifuge tests and the numerical model showed good agreement for various input conditions. Validation of the proposed modeling method was performed using the test results from other cases to verify the applicability of the numerical model. The dynamic pile responses produced by the numerical simulation also showed good agreement with the test results in various conditions. Based on the calibration and validation process, the applicability of the proposed modeling method for various conditions was verified. A parametric study for various conditions was performed to investigate the dynamic behavior of a pile foundation by varying the weight of the superstructure, relative density, pile length, and pile head fixity. Parametric studies demonstrated which parameters have a considerable affect on the dynamic behavior of a pile foundation in dry soil deposits. A three-dimensional numerical simulation of dynamic SPSI in a liquefiable condition was also carried out. In order to simulate the development of pore water pressure according to shear deformation of the soil for the liquefied state, the Finn model was adopted. The Finn model is one of the models that adopt an effective stress method so that the liquefaction at each location can be directly recorded in real time. The Finn model was combined with the standard Mohr-Coulomb criteria used in the dry condition. The calibration of the proposed modeling method was carried out by comparing the results with those of the centrifuge test performed by Wilson (1998). The excess pore pressure ratio, pile bending moment, and pile head displacement-time histories calculated by the numerical model agreed reasonably well with the test results. The validation of the proposed method was later performed using the test results from another case. The calculated dynamic pile responses also show generally good agreement with measured pile responses of another case. Based on the calibration and validation procedure, it was confirmed that the proposed modeling method can properly simulate the dynamic behavior of a soil-pile system in liquefiable soil deposits. A parametric study was carried out to provide better insight into the dynamic behavior of pile foundations embedded in liquefiable sand under varying frequencies of input excitation, thicknesses of the liquefiable layer, relative densities, and weights of the superstructure. Parametric studies demonstrated which parameters have a considerable effect on the dynamic behavior of a pile foundation in liquefiable soil deposits. Moreover, the results obtained from parametric studies confirmed availability and applicability of the proposed numerical model comparing with the results from previous researches.