An investigation into the seismic performance and progressive failure mechanism of model geosynthetic reinforced soil walls

Abstract

Reinforced soil is a construction method in which layers of material based on plastics or metals are placed in between layers of soil. This sandwich-type construction is much stronger than just the soil and as a result, steep slopes and high vertical retaining walls may be made from the combined materials. The facing of reinforced wall can be left open so that grass and other foliage can grow out of it (called “soft facing”) or panels may be placed in front which are either in segments (“segmental”) or as a single stiff unit (“full height rigid”). Generally, the full height panels provide additional support by tying all the layers together, so these walls are considered to be particularly good for areas in which earthquakes can occur.

This study considered the resistance to earthquakes provided by vertical model reinforced soil walls which had a full height rigid facing attached and five layers of geosynthetic (i.e. flexible plastic grid) reinforcement (GRS walls). The model walls were constructed to be 0.9m high and were placed on a shaking table which allowed simple horizontal sinusoidal shaking to be applied to them. The shaking frequency and length were scaled so that the wall modelled a prototype that was similar to a 4.5m high wall exposed to shaking induced by moderate to strong earthquakes.

For twelve different GRS model tests, variations were made to the length of the reinforcement (0.6m to 0.9m with some cases including longer lengths just at the top) and the density of the soil, and for some tests an additional load was applied to the soil surface. Sensors were placed at different points in the wall’s soil to determine the acceleration and displacement during shaking, and were also placed into the reinforcement to measure the reinforcement load near the wall. High-speed cameras were placed at the side of the model to record the deformation during testing. Processing and detailed analysis of the images enabled movement of the soil that the naked eye couldn’t see to be detected and this image data was able to be linked to the sensor data within the wall to build up a complete picture of loads and movement.

Results showed that walls made with more compact soil, or with longer top layers of reinforcement, required greater shaking to deform and to fail. They also showed how the shaking is amplified within the soil towards the surface, how the soil deforms and fails mostly by rotation, and how increasing the top layer changes the way the soil responds during loading. Results also showed that adding a surface load initially made the wall more stable when it was shaken at low amplitude but as the shaking increased it destabilised the wall. The greatest load was found to occur in the bottom layers of reinforcement.

The results will be used to improve the design of reinforced soil for seismic loading by comparing the data against what is predicted in design. In particular, the study gives new information on reinforcement loads and how these link to deformation of the wall. This can help with aspects of detailed design – such as connections between the reinforcement and the wall panel – and will support the development of the seismic guidelines for GRS walls in New Zealand.

---

Technical Abstract

Geosynthetic reinforced soil (GRS) walls involve the use of geosynthetic reinforcement (polymer material) within the retained backfill, forming a reinforced soil block where transmission of overturning and sliding forces on the wall to the backfill occurs. Key advantages of GRS systems include the reduced need for large foundations, cost reduction (up to 50%), lower environmental costs, faster construction and significantly improved seismic performance as observed in previous earthquakes. Design methods in New Zealand have not been well established and as a result, GRS structures do not have a uniform level of seismic and static resistance; hence involve different risks of failure. Further research is required to better understand the seismic behaviour of GRS structures to advance design practices.

The experimental study of this research involved a series of twelve 1-g shake table tests on reduced-scale (1:5) GRS wall models using the University of Canterbury shake-table. The seismic excitation of the models was unidirectional sinusoidal input motion with a predominant frequency of 5Hz and 10s duration. Seismic excitation of the model commenced at an acceleration amplitude level of 0.1g and was incrementally increased by 0.1g in subsequent excitation levels up to failure (excessive displacement of the wall panel). The wall models were 900mm high with a full-height rigid facing panel and five layers of Microgird reinforcement (reinforcement spacing of 150mm). The wall panel toe was founded on a rigid foundation and was free to slide. The backfill deposit was constructed from dry Albany sand to a backfill relative density, Dr = 85% or 50% through model vibration.

The influence of GRS wall parameters such as reinforcement length and layout, backfill density and application of a 3kPa surcharge on the backfill surface was investigated in the testing sequence. Through extensive instrumentation of the wall models, the wall facing displacements, backfill accelerations, earth pressures and reinforcement loads were recorded at the varying levels of model excitation. Additionally, backfill deformation was also measured through high-speed imaging and Geotechnical Particle Image Velocimetry (GeoPIV) analysis. The GeoPIV analysis enabled the identification of the evolution of shear strains and volumetric strains within the backfill at low strain levels before failure of the wall thus allowing interpretations to be made regarding the strain development and shear band progression within the retained backfill.

Rotation about the wall toe was the predominant failure mechanism in all excitation level with sliding only significant in the last two excitation levels, resulting in a bi-linear displacement acceleration curve. An increase in acceleration amplification with increasing excitation was observed with amplification factors of up to 1.5 recorded. Maximum seismic and static horizontal earth pressures were recorded at failure and were recorded at the wall toe. The highest reinforcement load was recorded at the lowest (deepest in the backfill) reinforcement layer with a decrease in peak load observed at failure, possibly due to pullout failure of the reinforcement layer. Conversely, peak reinforcement load was recorded at failure for the top reinforcement layer.

The staggered reinforcement models exhibited greater wall stability than the uniform reinforcement models of L/H=0.75. However, similar critical accelerations were determined for the two wall models due to the coarseness of excitation level increments of 0.1g. The extended top reinforcements were found to restrict the rotational component of displacement and prevented the development of a preliminary shear band at the middle reinforcement layer, contributing positively to wall stability. Lower acceleration amplification factors were determined for the longer uniform reinforcement length models due to reduced model deformation. A greater distribution of reinforcement load towards the top two extended reinforcement layers was also observed in the staggered wall models.

An increase in model backfill density was observed to result in greater wall stability than an increase in uniform reinforcement length. Greater acceleration amplification was observed in looser backfill models due to their lower model stiffness. Due to greater confinement of the reinforcement layers, greater reinforcement loads were developed in higher density wall models with less wall movement required to engage the reinforcement layers and mobilise their resistance.

The application of surcharge on the backfill was observed to initially increase the wall stability due to greater normal stresses within the backfill but at greater excitation levels, the surcharge contribution to wall destabilising inertial forces outweighs its contribution to wall stability. As a result, no clear influence of surcharge on the critical acceleration of the wall models was observed. Lower acceleration amplification factors were observed for the surcharged models as the surcharge acts as a damper during excitation. The application of the surcharge also increases the magnitude of reinforcement load developed due to greater confinement and increased wall destabilising forces.

The rotation of the wall panel resulted in the progressive development of shears surface with depth that extended from the backfill surface to the ends of the reinforcement (edge of the reinforced soil block). The resultant failure plane would have extended from the backfill surface to the lowest reinforcement layer before developing at the toe of the wall, forming a two-wedge failure mechanism. This is confirmed by development of failure planes at the lowest reinforcement layer (deepest with the backfill) and at the wall toe observed at the critical acceleration level. Key observations of the effect of different wall parameters from the GeoPIV results are found to be in good agreement with conclusions developed from the other forms of instrumentation.

Further research is required to achieve the goal of developing seismic guidelines for GRS walls in geotechnical structures in New Zealand. This includes developing and testing wall models with a different facing type (segmental or wrap-around facing), load cell instrumentation of all reinforcement layers, dynamic loading on the wall panel and the use of local soils as the backfill material. Lastly, the limitations of the experimental procedure and wall models should be understood.