UBC Theses and Dissertations
Response of pile foundations to simulated earthquake loading : experimental and analytical results volume I Gohl, W. Blair
The analysis of the dynamic response of pile foundations to earthquake shaking is a complex problem and has been treated using concepts developed from the theory of elasticity, applicable to low level shaking, and to models incorporating non-linear soil response appropriate for stronger shaking intensities. A review of available field reports indicates that due to the lack of complete instrumental recordings describing the response of full scale pile foundations to earthquake loading, the above analysis techniques are in large measure unchecked. To provide a reliable data base suitable for checking various models of dynamic pile foundation response, a series of small scale model tests on single piles and pile groups embedded in dry sand foundations were carried out on shaking tables at the University of British Columbia. A similar series of tests were carried out using a geotechnical centrifuge equipped with a base motion actuator located at the California Institute of Technology. Under the centrifugal forces acting on the model, full scale stress conditions are simulated in the sand foundation. Since soil behaviour is stress level dependent, the centrifuge tests are considered to provide a more realistic simulation of full scale pile foundation behaviour. Both the shake table and centrifuge single pile tests were carried out using both sinusoidal and random earthquake input motions over a range of shaking intensities. From the data, details of soil-pile interaction were elucidated. This provided a basis for improvement in methods of estimating required input parameters used in the dynamic analysis of pile foundations. Prior to each test, shear wave velocity measurements were made throughout the prepared sand foundations using piezoceramic bender elements. This technique has proved particularly useful in the centrifuge environment since the bender element source and receivers could be triggered remotely from off the centrifuge arm while the model was in flight. The shear wave velocity data were used to compute small strain, elastic shear moduli in the soil which have been found to be in close agreement with predictions made using an equation proposed by Hardin and Black (1968). Elastic compression wave velocities were also identified from the bender element responses recorded during the shake table tests. The single pile tests demonstrated that significant non-linearity and strain softening occurs in near field soil response, which is responsible for reductions in fundamental vibration frequency and pile head stiffness parameters with increasing amplitudes of lateral pile vibration. An analysis technique developed to estimate average effective strains around a single pile leads to predictions of large modulus reduction around the pile, depending on the amplitude of pile vibration. Soil reaction pressures (p) due to relative horizontal movement between the soil and the pile (y) were deduced from the test data for various cycles of shaking, or so-called p-y curves. The cyclic p-y curves developed show clearly the non-linear, hysteretic near field response near the pile head. Approximately linear elastic p-y response occurs at greater depth. Backbone p-y curves computed using procedures recommended by the American Petroleum Institute (API) are in poor agreement with the experimental shake table and centrifuge measurements. Material damping inferred from the area within the p-y hysteresis loops increases, in general, with increasing pile deflection level. The experimental p-y hysteresis loops were reliably simulated using a Ramberg-Osgood backbone curve and the Masing criterion to model unload-reload response. Comparing the flexural response observed on single piles during the shake table and centrifuge tests, the depth of maximum bending moment relative to the pile diameter has been observed to be greater in the shake table tests. This can be anticipated from the laws of model similitude. Cyclic p-y curves developed from the shake table and centrifuge tests also show substantial differences, with the shake table p-y curves being stiffer than predicted using the API procedures, while the opposite behaviour was found in the high stress, centrifuge environment. Damping in the low stress level environment of the shake table has been found to be greater than under full scale stress conditions in the centrifuge. Two-pile tests, where the piles have been oriented inline, offline or at 45 degrees to the direction of shaking, indicate that pile to pile interaction is very strong for inline and 45 degree shaking, and is relatively minor for offline shaking. Interaction effects observed under low and high intensities of shaking die off with increasing pile separation distance at a quicker rate than predicted using elastic interaction theory. Interaction effects for inline and offline cyclic loading may be neglected for centre to centre pile spacings of about six and three pile diameters, respectively. For close pile separations during inline shaking, elastic theory underpredicts the extent of interaction. Similar conclusions were reached from the shake table and centrifuge tests conducted. Based on the experimental data and data available from the literature, modifications to elastic pile interaction coefficients have been suggested. Predictions of single pile response to earthquake shaking have been made using an uncoupled, sub- structure approach incorporating non-linear pile head springs and equivalent viscous dashpots (foundation compliances) derived from the test data. The foundation compliances account for the deflection level dependent stiffness and damping characteristics of the below ground soil-pile system. The measured free field surface motions have been used as the input excitation. Agreement between computed and measured pile responses was found to be excellent. A fully coupled analysis using the commercially available program SPASM8, where the below ground portions of the pile are directly considered in the numerical discretization of the problem has also been used. Interaction between the soil and vibrating ground is accounted for using a Kelvin-Voight model which includes non-linear Winkler springs and equivalent viscous dashpots to simulate radiation damping. Free field ground motions deduced from an independent free field response analysis using the computer program SHAKE are applied to the free field end of the soil-pile interaction elements. Using this full coupled model, the possible effects of kinematic interaction are accounted for. Results from the analysis show that SPASM8 underpredicts pile flexural response. A key difficulty in using an analysis of this kind is the accurate determination of free field input motions to be used along the embedded length of the pile. A computer program, PGDYNA, has been developed to analyse the uncoupled response of a superstructure supported by a group of foundation piles, taking into account non-linearity of the pile head compliances and the effects of pile group interaction. Interaction factors developed from the experimental test program were used to calculate deflection level dependent pile head stiffnesses. Preliminary testing of the program indicates that use of the free field surface motions as input, neglecting the effects of kinematic interaction, leads to an overestimate of pile group response.
Item Citations and Data