Seismic Modelling of Rock-Fill Embankments on Deep Loose Saturated Sand Deposits

L M N Peiris, Cambridge University
Geotechnical Engineering Group

Abstract

There are many case histories of earth embankment and dam failures due to earthquakes, reported in the literature. These geotechnical structures built of sand and silt showed the occurrence of slope sliding, slumping and cracking leading to the embankment collapse as a result of their underlying foundations liquefying during earthquakes. After World War II, sand and gravel aggregate or crushed rock became a widely used material in constructing embankments, earth and rock-fill dams, artificial islands and coastal defences. Since many of these structures are found in seismically active regions throughout the world, it is possible that they could also experience failure as a result of weakening of their underlying foundations. Hence, there is a need to investigate the failure mechanisms of rock-fill embankments on loose liquefiable deposits subjected to earthquakes.

Physical model tests on gravel embankments on deep loose saturated sand deposits were carried out on board the Cambridge University Geotechnical Centrifuge, to study the behaviour of rock-fill embankments on deep loose deposits. The models were prepared in the Cambridge University Equivalent Shear Beam (ESB) box, modelling the dynamic behaviour of a semi-infinite prototype soil deposit. The earthquakes were introduced to the models using the bumpy road shaker and the newly developed Stored Angular Momentum (SAM) actuator. All the centrifuge tests except one, involved the use of the SAM actuator.

The physical model tests on gravel embankments showed that, the primary mechanism of embankment failure is gravel penetration into the underlying foundation during the earthquakes. Gravel penetration resulted in significant crest settlements of the gravel embankments during strong earthquakes. The crest settlements observed were less than the depths of gravel penetration. This may be attributed to the structure of the embankment gravel altering from a densely interlocked to one of expanded, upon passing through the embankment/foundation interface. The expansion of the gravel embankment in the foundation alone cannot account for the crest settlements being greater than the depths of gravel penetration. Once the gravel near the base of the embankment has penetrated, the deformation of the interlocked gravel above the embankment/foundation interface could form an arch, which would slow down the flow of gravel into the foundation. Lateral spreading of the embankment also contributed to the crest settlements observed, which was primarily due to gravel near the toe region penetrating and de-stabilizing the slopes. This was indicated in a gravel embankment test with a water saturated foundation, where neither gravel penetration nor lateral spreading were observed. Using water inhibits large positive pore pressure generation, which is a feature of oil saturated models where gravel penetration was observed.

Physical model tests on a sand and a rigid embankment were carried out to compare the dynamic behaviour with those of the gravel embankments. The failure of the sand embankment was largely due to lateral spreading during the earthquakes. The rigid embankment settled into the foundation by displacing fluidized material trapped beneath its base, without any embankment deformation.

The dynamic behaviour of the underlying foundation was responsible for the gravel embankment penetration observed during the earthquakes. Once the pore pressures rise, the increased hydraulic gradients in the vicinity of the gravel embankment/foundation interface would initiate gravel penetration. When the foundation reaches the "characteristic state", a potential exists for large strains to be mobilized. This allows the gravel to penetrate deep into the foundation. The arrival at the characteristic state is indicated by the pore pressures dropping as a result of the dilating tendency. The zone below the embankment where these dilation effects exist, is described as the "dilation zone". This dilation zone expands deeper into the foundation in stronger earthquakes or earthquakes that last longer. The structure of the embankment gravel alters from a densely interlocked to one of expanded upon passing through the embankment/foundation interface. Gravel penetration and the structure expansion mobilizes more shear strains, resulting in the foundation experiencing residual dilation effects, reflected in a residual drop of the pore pressures. The dilation zone setup in the foundation led to pore fluid migration into the region from the liquefied free field. However, since the free field was softer than the foundation below the embankment, there was a possibility of an increase of the total vertical stress below the embankment. This increase in total vertical stress and the pore fluid migration resulted in the pore pressures below the embankment rising after the effects of dilation are shown.

The stability of the embankments tested in the centrifuge were analysed using slope stability and bearing capacity evaluation procedures. The slope stability analysis predicted the gravel embankments to remain stable, despite the physical model tests showed lateral spreading. This is because the slope stability evaluation procedure cannot account for the slope instabilities caused by gravel penetration. The steeper slope of the sand embankment was predicted to have a slope failure, which experienced a significant lateral spreading during the earthquake. Dynamic bearing capacity solutions applied to the rigid and the sand embankment revealed that the bearing capacities are higher than the applied loads and a failure geometry cannot be accommodated within the bounds of the model.

The embankment stability calculations and the experimental observations show that the limit equilibrium slope stability evaluation procedures are strictly not applicable for rock-fill embankments on loose deposits. They fail to address the issue of rock-fill penetration, which is the primary failure mechanism. The dynamic bearing capacity solutions assumed no pore pressure generation during an earthquake. Hence, these methods are not applicable, unless the degradation of the bearing strength is accounted for.

The behaviour of the foundation below the sand and the rigid embankment was modelled using the finite element program SWANDYNE. The patterns of pore pressure results indicated the effects of dilation, pore fluid migration and total vertical stress re-distribution in the foundation below the embankment. Free field liquefaction was also seen the results. These effects are reflected in the acceleration attenuations seen in the foundation model. The results generally compared well with the centrifuge test data, which suggests that, SWANDYNE could make class ‘C’ predictions on the behaviour of foundations under embankment loading experiencing earthquakes.

The physical model tests conducted for this thesis, were the first ever research experiments to be carried out using the new SAM actuator. Since this research program began when the actuator was under testing, several configurations of the shaker were used resulting in earthquakes with different features. Hence, the tests done for this thesis form part of the testing program and future refinements that the shaker may require to enhance its performance. The boundary effects of the ESB box were further examined continuing from the work done by other researchers at Cambridge.