You are on page 1of 9

CHAPTER 9

9.1 Scope of Research

SUMMARY AND CONCLUSIONS

A significant number of cases of damage to piles and pile-supported structures during earthquakes have been observed, but few instrumented records of the response and performance of such structures during earthquakes have been obtained. Much of the observed damage to piles during earthquakes has been due to the effects of liquefaction and lateral spreading, though some important cases of seismically-induced pile failures in clay have been observed in the Mexico City and Loma Prieta earthqukes. A thorough review of field and laboratory experimental programs designed to investigate SSPSI has revealed that most efforts have focused on liquefaction problems, leaving a gap in our understanding of SSPSI in cohesive soils. The state-of-the-practice and current building code recommendations reflect the lack of consensus on how to evaluate SSPSI effects. Currently available analytical methods for SSPSI problems range from simple static analyses to derive pile head secant stiffnesses for input to dynamic structural models, to complete dynamic pile group interaction analyses. Many analytical tools consider visco-elastic response, and others model soil-pile nonlinearity with p-y curves, but nearly all have the common feature of uncoupling the site response, soil-pile interaction, and superstructure response components of SSPSI. To expand the database of pile performance during strong shaking, a series of scale model shaking table tests of model piles in soft clay was performed. This research effort had the goals of providing insight into a variety of SSPSI topics, and generating a data set

411

with which to calibrate an advanced SSPSI analysis tool being developed at U.C. Berkeley (Lok, 1999), as well as data suitable for evaluation of other analytical tools and methods.

9.2 Research Findings and Recommendations Principles of scale model similitude were used to derive a set of model scaling relationships that were used in a method of implied prototypes to relate model and prototype behavior. Consideration was given to the dynamic and nonlinear nature of soilpile interaction in developing the model soil and model piles for the testing program. A specialized flexible wall test container was designed to allow the soil to respond in the same fashion as the free-field, unencumbered by boundary effects. The performance of the shaking table was generally seen to be good, with reasonable reproduction of 1-D and 2-D input motions. Unwanted twist, pitch, and roll motions were present, however, and exerted unquantifiable influence on the model response. These spurious motions were, however, principally of high frequency content, and did not appear to significantly affect soil-pile-superstructure performance. The trends of model site response were consistent with free-field behavior; the motions amplified from base to surface and were coherent across the site. Vertical accelerations observed at the soil surface may have been due to a combination of surface wave reflections, table pitch and roll motions, and a soil column bending mode. Qualitative analysis of the single pile and pile group response yielded a number of interesting results. The single piles were seen to respond with components of inertial and kinematic interaction, though the inertial components produced upper bound bending moments. This result may suggest that developing pile demands from consideration of

412

inertial loading only may be acceptable for cases where site stiffness contrasts or ground failure (lateral spreading) do not exert significant soil loads or deformations on the piles. The response of pile groups was highly frequency dependent, which calls into question the applicability of applying pseudo-static analyses to such problems. Pile cap and free field motion variations illustrated wave scattering effects and the necessity of developing modified foundation input motions for substructuring analyses. Moderate effects of pile cap embedment were observed, particularly in contributing to pile group rocking stiffness, though further study is warranted in this area. A group that had been first subjected to large deformation static lateral loading was seen to have lesser seismic response than an identical group that had not been pre-loaded, suggesting that pre-loading remolded the near-field soils and base-isolated the group. A 5x3 pile group and single pile with the same average load per pile were subjected to 2-D shaking, and the single pile was seen to have greater bending moment demands than the group piles. This was attributable to the long period motion and resulting lack of resonance of the pile group superstructure, which did not impart large inertial loads to the foundation. Finally, tests evaluating pile raft performance and the effects of impounded water scouring the soil-pile gap and degrading resistance were somewhat inconclusive, but can be revisited with improved test designs. Site characterization included laboratory and in-situ testing to establish the undrained shear strength and shear wave velocity profiles. T-bars tests were adapted from centrifuge testing, and produced continuous profiles of strength versus depth. The

strength determined in the T-bar tests was found to include a shearing velocity rate effect, which was quantified in relation to vane shear testing. Hammer blow tests were

performed from the base and surface of the model container to determine the site shear

413

wave velocity profile. The latter method was found to yield good results except near the surface, where surface wave interference obscured the shear waves, and at depth, where the wave energy attenuated and reflected. Piles in very close proximity to the

accelerometer arrays were found to artificially increase the apparent wave velocity by proving alternate travel paths for the wave energy. Model-soil-specific dynamic modulus degradation and damping curves as a function of strain level were derived from testing with an advanced cyclic triaxial testing apparatus with internal strain and shear wave velocity measurements. SHAKE91 was successfully used to simulate the model free-field response, indicating that the model soil-container system adequately reproduced free-field site conditions. The small errors between the observed and predicted behavior may be

acceptable for pure site response analyses, but the propagation of these errors into the SSPSI analysis requires further study. The suite of pile head loading tests provided estimates of pile head lateral and axial stiffness. COM624P achieved very good predictions of the static lateral pile load tests. Increased deflections under axial load cycling were observed to agree with data published by Bea et al. (1980). T-z analysis demonstrated that the model piles were transferring load almost exclusively through end-bearing, as was intended. Pile group effects were not thoroughly analyzed, though a group efficiency factor of 0.8 was calculated from the static lateral group test. This value may be unconservative considering the group yield behavior at higher load levels. The end-bearing group was observed to engage in a near rigid body rotation, which may constitute the seismic

414

response mode in soft soils. Although group effects were not analyzed in great detail, the data developed here should provide a good basis for such analyses in subsequent studies. The influences of 2-D shaking were seen to be minimal, as structural inertial forces tended to resolve the motion to a strong axis for the simple single degree of freedom models tested. For single piles, perimeter soil resistance was not engaged, as the piles preferentially followed gaps developed in previous cycles. These findings support the notion of using 1-D SSPSI analyses for simple and regular structures, preferably with a multi-component combination rule to scale input motions. The derivation of p-y curves from the static and seismic test data was quite successful, and the experimental curves compared very well to those recommended by API. P-y curve initial stiffness and ultimate strength trends with depth were in good agreement with API static and cyclic curves. Degrading behavior due to hysteresis and gapping was observed, softening the near-surface response below API stiffness values, indicating that gapping is an important feature to analytically model. These findings also serve as a back-analysis validation of using p-y curves for SSPSI problems. The application of system identification techniques yielded estimates of single pile and pile group flexible base frequencies and damping factors. Single pile and pile group frequencies were consistent with sine sweep and seismic response. The pile group flexible base frequencies differed significantly from the fixed base case, and damping ratios from 10 20 % were observed. Damping for the single piles and groups was computed to be a function of load level, and single pile values ranged from 10 20 %. Estimates of pile head lateral stiffness and experimentally-derived values differed over a wide range, from 100 700 lb/in, and were a function of loading level and

415

consequent soil-pile nonlinearity. The methods examined for computing dynamic stiffness from elastic theory provided unrealistically high estimates of stiffness for the model tests. Appropriately selected secant stiffness values from the static lateral load tests provided more realistic descriptions of the observed soil-pile dynamic response for moderate levels of shaking. ATC-32 chart solutions provided marginally acceptable lower bound pile head stiffness estimates for very strong shaking events.

9.3 Recommendations for Future Research The recommendations for future research are threefold as they relate to: (1) further analysis of the existing data set, (2) modified test procedures for shaking table testing with the flexible container, and (3) other possible future research .

9.3.1 Data Mining The data set generated in this research effort provides a rich information base for extensive study of SSPSI. Much work remains to be performed in the analysis of group interaction in both the static and seismic tests. The rocking response of the pile raft foundation and cap contributions to dynamic group response may also be investigated. The procedures developed for the derivation of experimental p-y curves can be applied to the many single pile and group pile cases. The implications of response prediction with the range of pile head stiffness values observed can be studied. Nonlinear site response analyses may provide improved estimates of the model soil site amplification. Most importantly, the single pile and pile group seismic response can be simulated with the wide range of existing and developing analytical tools.

416

9.3.2 Improved Test Procedures First and foremost, the shaking table performance should be further improved, particularly by eliminating twist, pitch, and roll motions, which exert unknown effects on the model performance. If possible, tests with other scaling factors should be conducted in a modeling of models approach to further validate the scaling relationships. A stiff cohesive soil base layer impeding drainage and arresting consolidation is recommended, to ensure more stable model conditions. Model pile strain gaging quality control can be slightly improved, though the model soil fly ash provides a harsh (chemical) environment and some gage failures are inevitable. The positioning of soil accelerometers should be such that in a given test, one array is distant from the model structures; one or two dense vertical arrays is recommended. Multiple surface vertical accelerometers are useful for tracking the soil column deformation modes. Surface hammer blow shear wave velocity tests are promising for determining Vs profiles, and may be improved by using an accelerometer trigger on the hammer and stacking records as practiced in geophysics. Pile group effects may be better detected by constructing a pile cap or load frame with load measurement devices at each pile head, or in each pile row. Forced vibration tests should be implemented for both single piles and pile groups with equipment capable of delivering higher loadings and higher frequencies. From the testing standpoint, the model response to a suite of ground motions with different frequency characteristics would provide a broader understanding of SSPSI. A given model may be excited with several different ground motions at low amplitude, but only one large amplitude shake can be performed before site degradation occurs. An

417

alternative procedure could consist of exciting the model with progressively increasing intensisties of the same motion, rather than just low and high amplitude shakes. It may also be possible to shake on one axis, then the other, so that soil-pile nonlinearities would be minimized from one test to the next. Other strategies may include changing model frequency response by changing pile head masses from test to test. The shaking table capabilities for 2-D and 3-D shaking should be utilized, as this type of data provides unique and realistic performance assessments. In addition, it may be useful to vary the orientation of the model structures to the principal axes of shaking.

9.3.3 Future Shaking Table Research Topics The capability that has been developed at U.C. Berkeley to study SSPSI with scale model shaking table testing is a unique and valuable resource, and demands to be applied to further investigations. Some possible topics for future study include the following: battered pile supported structures, attracting large head loads, large diameter drilled shafts, possibly constructed of scale model concrete, friction pile axial, t-z, and rocking response, pile raft foundations, including seismically induced settlements, multi-mode structures and/or structures supported by multiple pile groups, ductile columns, designed to yield and thereby reduce foundation demands, layered site profiles, including impedance contrast and surficial fill cases, variable pile group spacing, and influence of pile cap base contact on dynamic lateral and rocking stiffnesses.

418

Continued study of these and other SSPSI topics can contribute to advancing the state-of-knowledge, which then can hopefully be translated into advancing the state-ofpractice.

419

You might also like