Centrifuge Modeling of the Seismic Performance of Pile-Reinforced Slopes

An extensive program of centrifuge testing was conducted to quantify the improvements to seismic slope performance that can be achieved by installing a row of discretely spaced vertical precast RC piles. A key novelty of the work presented is the use of recently developed microreinforced concrete to produce realistically damageable model piles. Pile-reinforced slopes are a good example of a problem in which relative soil-pile strength is important in determining whether the soil or pile yields first, and in which the performance of a slope with structurally damaged piles may be of interest. The new model RC allows these factors to be properly accounted for in a reduced-scale physical model for the first time. Two different reinforcement layouts were considered, representing (1) a section specifically detailed to carry the bending moments induced by the slipping soil mass, and (2) a nominally reinforced section with low moment capacity. These were supported by further tests on conventional elastic piles that were instrumented to measure seismic soil-pile interaction. It was demonstrated that dynamic ground motions at the crest can be significantly reduced in amplitude by up to 20% with elastic piles spaced at $s/B=3.5$, both in terms of the peak acceleration and across the full response spectrum. Permanent deformations at the slope crest (e.g., settlement) were also reduced by up to 35% at $s/B=3.5$. These findings are consistent with previous suggestions made for the optimal $s/B$ ratio for encouraging soil arching between piles at maximum spacing both under monotonic conditions and from numerical investigations of the seismic problem. The improvements to slope performance were reduced slightly using the designed microreinforced concrete piles owing to deterioration in the bending properties arising from fatigue under the cyclic kinematic loading. This suggests that idealization of RC piles as elastic elements will likely be only an approximation of their true behavior. The importance of reinforcement detailing was also highlighted, with the nominally reinforced section yielding early in the earthquake, and the damaged piles subsequently only offering a small (although measureable) reduction in seismic slope performance compared with the unreinforced case. In addition to these findings, the data presented will be useful in validating future improved numerical models for predicting the performance of piled slopes and for the aseismic design of pile-reinforcing schemes.