Analysis of soil failure modes using flume tests

Size: px

Start display at page:

Download "Analysis of soil failure modes using flume tests"

Barry West
5 years ago
Views:

1 Analysis of soil failure modes using flume tests A. Spickermann & J.-P. Malet Institute of Earth Physics, CNRS UMR 751, University of Strasbourg, Strasbourg, France Th.W.J. van Asch, M.C.G. van Maarseveen, D.B. van Dam & H. Markies Faculty of Geosciences, Utrecht University, Utrecht, Netherlands ABSTRACT: A basic requirement in landslide forecasting and in the design of early warning systems is the identification, qualification and the modelling of failure modes and triggering mechanisms. Besides numerical tools, experiments like flume tests can give a better understanding of the processes governing landslide development. This study focuses on the investigation of the hydrological triggering and the mode of failure of landslides by performing flume tests in the laboratory. To trigger slope failure a groundwater table is created by means of water supply from the bottom (upward infiltration) of the slope. The failure mode and the mechanisms leading to failure depend on many features as geometry, material, discontinuities and joints. Many flume experiments are done on sandy or non-cohesive materials. Triggering small-scale landslides on cohesive or more clayey materials by hydrological processes experiments is rather rare. This work considers both flume tests on sandy and cohesive materials. On the basis of these flume experiments the main differences in failure modes with respect to the material are discussed. To characterize failure modes four criteria are taken into account, (1) the time displacement behaviour, () the total time of the failure, (3) the description of the kind of failure and () possible mechanisms leading to the observed failure. Whereas failure in the performed flume experiment on sand occur relative fast and retrogressively in multiple slumps, the formation of a fully developed shear surface in the test on a clayey material is initialized over a long time period and takes place progressively. The differences in failure modes caused by certain failure mechanisms have to be taken into account in the design of early warning systems. Monitoring techniques must agree with the development of failure because an appropriate monitoring is the requirement for the development of early warning systems. 1 INTRODUCTION The identification, qualification and the modelling of triggering mechanisms and of failure modes is essential in landslide forecasting and in the design of early warning systems. Within a number of several triggering mechanisms the hydrological trigger is assumed to be the reason for the majority of landslides. A landslide that is triggered by a hydrological process can show a variety of different failure modes which depend on the given conditions of the slope as geometry, material, joints and discontinuities. The mechanisms leading to a certain failure mode can be studied by triggering small-scale landslides e.g. in terms of flume tests. In this study we use four criteria to describe the mode of failure, (1) distribution of displacements in time of single points at the slope surface, or velocity of single points respectively, () total time span of the failure process, (3) kind of failure, e.g. retrogressive or progressive failure and () the mechanism of failure. A detailed analysis of the distribution of displacements in time in rotational and translational landslides is done by Petley et al. (). It is demonstrated that one of two movement styles is evident during accelerating phases for the landslides observed. In a plot of the reciprocal velocity 1 / ν against time the first style has a linear and the second style has an asymptotic form. It is assumed that the first style is related to landslides in which crack propagation is the dominant process, i.e. progressive failure (Bjerrum, 197), whereas the second style is associated with movement occurring as a result of ductile deformation in existing shear zones. Continuous monitoring of landslides allows statements about the total time span of failure. Petley et al. () use the time unit days in their inverse velocity time plots of different landslides. Especially in landslides prepared over very long time periods it is very difficult to estimate the starting

2 point of failure. Also it is difficult to define the total time span of failure when it occurs within several events that take place after periods of almost no movement. With respect to the third criteria Wang & Sassa (3) have summarized failure modes obtained from flume tests on sandy materials. The observed failure phenomena vary between complete flow slide and retrogressive failure. It is shown that the failure mode depends greatly on the grain size. Another kind of failure is the so-called progressive failure, defined as the formation and growth of a shear surface in cohesive materials undergoing a transition between peak and residual strength (Bjerrum, 197). By now almost all of the experimental studies have been performed on clean sands, i.e. little work has been performed on silt and silt-clay mixture (Wang & Sassa, 3). But despite numerous studies which have been done there is still uncertainty in the explanation of the processes determining the failure, that are discussed in the forth criteria. Iverson et al. (1997), Okura et al. (), Wang & Sassa (3), van Asch et al. () investigate fluidization processes that are often the reason for landslides in sandy materials. Also the development of progressive failure in cohesive slopes is a poorly understood process. Petley et al. (5) present a model for the growth of landslide shear surfaces, and hence for progressive failure, based upon laboratory testing of materials under conditions that simulate those within a hydrologically triggered landslide, i.e. conventional isotropic consolidation and undrained shear tests. A better understanding of the different behaviour of slopes is required for the prediction of potential catastrophic failure. This study is an attempt to get a qualitative overview of the varying failure modes that take place in slopes in relation to different materials based on flume experiments. In the longer term the results obtained from these tests can be used in the design of early warning that will depend on the material. FLUME TEST APPARATUS AND TEST PROCEDURE The test apparatus is depicted in Figure 1. Figure 1. Scheme of flume test apparatus. The inner dimensions of the flume are 5 cm in length, cm in height and cm in width. The flume angle α is made changeable for different test requirements. The test apparatus consists of one transparent side. Images are taken in defined intervals from this side and from the front. The pore pressure is monitored by tensiometers. To investigate the hydrological trigger water infiltrates from the back bottom of the slope using a pump and a defined top water level that is kept constant (upward constant head infiltration). 3 FLUME TESTS To describe typical failure modes depending on the material using the four criteria given in the introduction two flume tests are analysed in sandy as well as in clayey material. The first test considers sandy material and is performed on loosely packed River sand. A test on clay from Zoelen, the Netherlands, is the representative for cohesive materials. Percent finer by weight [% 1 Limburg sand River white sand,1,1,1 1 1 Grain size [mm] Figure. Grain size distributions. 3.1 Test on River sand The first experiment is carried out on loosely packed River sand. The grain size distribution is given in Figure. Slope dimensions are 5 cm in height and 11 cm in length. The slope angle of about 13 leads together with the flume inclination of 5 to a total slope angle of about 3 (Fig. 3). (1) To analyse the distribution of displacements in time the movement of four points at the slope surface is measured (Figs. 3, ). After about 15 min of water supply failure starts at the toe and it continues in direction of the crest. The ending points of the graphs are limited because of observation reasons. It does not mean that the soil mass stops to move. In Figure 5 the geometry and sequence of slumps is displayed and Figure gives the development of slumps in time. The time displacement curves can be

3 described as almost linearly, i.e. the observed points move with a nearly constant velocity. confirmed by experiments from the literature (Wang & Sassa, 3). slump number Figure 3. Initial test conditions and observed points of test on River sand. 1:3:3 1:: 1::3 1:5: time [h:min:sec] Figure. Development of slumps in time. 1:5:3 1:: Point C Point D 1:3:3 1:: 1::3 1:5: time [h:min:sec] 1:5:3 1:: Figure. Time displacement curves of four points at the slope surface of the flume test on River sand. Figure 5. Geometry and sequence of slumps during retrogressive failure. () The total time span of failure inclusive subsequent movements is about 5 min. Figure 7 illustrates the last phase of retrogressive failure at time 1:5:9 [h:min:sec] containing the positions of the observed points B, C and D. (3) Failure occurs in sudden multiple sliding and can be characterized as retrogressive slumping combined with liquefaction. The observations can be Figure 7. Final phase of retrogressive failure (1:5:9) inclusive observed points B, C, D. () After a number of tests on sandy materials still it is not clear if liquefaction of the soil causes the development of a sliding surface or if shear failure causes liquefaction. In this experiment it is observed that failure starts at the toe with liquefaction. It is assumed that due to liquefaction of the soil at the toe a first slip surface (slump 1) is formed. Other tests on e.g. Limburg sand (Fig. ) have shown that failure starts at the toe of the slope with shallow retrogressive rotational failures that end up in liquefaction. There are different theories that are assumed to be reasons for liquefaction, e.g. summarized in van Asch et al. () and van Asch & Malet (submitted). The following mechanisms are considered to be possible reasons for the liquefaction in the performed flume experiment, a) seepage flow, water flow gradient that is approximately opposite to the direction of gravity force causes decrease of effective stress, b) natural porosity is near liquid limit and c) compression at the toe leads to excess pore pressures (decrease of effective stress). An exact analysis and interpretation of pore pressure measurements in the flume tests may help to get a better understanding. Once failure has started with liquefaction at the toe the slope turns in an unstable state that causes the development of a shear surface, i.e. the

slump. We propose that the movement of the soil mass along the developed shearing zone is flow controlled, i.e. dependent on the down slope flow of the failed mass. 3.

Before the slope was built the material was saturated in buckets what leads to a strong compacted slope with a height of cm, a width of cm, a length of 57 cm and a slope angle of 35 with respect to a

4 slump. We propose that the movement of the soil mass along the developed shearing zone is flow controlled, i.e. dependent on the down slope flow of the failed mass. 3. Flume test on Zoelen clay To present a typical failure mode in cohesive material an experiment on clay from Zoelen, the Netherlands, is used. Before the slope was built the material was saturated in buckets what leads to a strong compacted slope with a height of cm, a width of cm, a length of 57 cm and a slope angle of 35 with respect to a horizontal flume surface (α = ). The flume was tilted back with a flume angle of -1 (-9-), that gives a total slope angle of γ = 1. After three days of drying water infiltration was started (9-9-). After another two days the flume was tilted to α = 11 (1-1-), which gives a total slope angle of γ =. Final Failure occurred after additional three weeks of supplying water (/5-1-). (1) The displacements in time are measured using two points at the slope surface (Fig. 9). Figure 1 presents the time displacement curves of the whole failure process (5 days). The curve can be divided into two parts, where the first part is characterized by a very slow movement that took place from 1-1- until -1-. Figure 1. Time displacement curves of two points at the slope surface of the flume test on Zoelen clay : : : time.1. :.1. 1:.1. 1:3 Figure 11. Time displacement curves of the last phase of failure in the test on Zoelen clay. Figure 1. Slope after failure, -1-, inclusive final positions of observed points A, B. Figure. Initial test conditions and observed points of test on Zoelen clay time In the second part the movement velocity increases and final failure occurs. In Figure 11 the second phase of the time displacement behaviour is illustrated in detail. Even if the movement along the slip surface in the second phase is faster than the movements of the first phase it can be described as rather moderate compared to the slumping velocities in sandy material. Also the diagram (Fig. 11) shows that final failure occurs in two stages, both on the 5-1-, the between : and 3: and the second between 5: and 7:. () The whole failure process that includes two parts of different movement velocities lasts about three weeks. Figure 1 shows the slope after failure on the -1- including the final positions of the observed points A and B. (3) The first part of the failure process is characterised by the development of fissures and cracks. Pore pressure measurements suggest that these cracks develop because of desiccation. This phase can be described by the initialization of failure, i.e. the formation and the growth of a completely developed shear surface. The second phase contains the movement of the soil mass along the developed shearing zone. Final failure occurs within one slump in two stages. The whole process is characterised as

5 progressive failure that is typical for cohesive materials. () In general it is assumed that progressive failure is caused by the decrease of shear resistance, i.e. the reduction of shear parameters, as friction angle and cohesion. According to Petley et al. (5) the small displacements in the first part of the failure process are caused by microcrack formation. Even if the microcracks weakens the slope slightly it is still in a stable state. Shear surface formation starts when the microcrack density reaches the point at which interaction between them begins. Petley et al. (5) propose that during this early phase of shear surface development, a reduction of pore pressure can stop the deformation. One can conclude that the constant water supply in the flume experiments supports the development of microcracks and further the formation of a slip surface. The acceleration to failure in the time displacement curve starts if shear stress exceeds shear strength. From this point pore pressures become increasingly unimportant for the deformation. CONCLUSIONS Up to now flume experiments are mainly carried out on sandy materials. The results show that flume experiments are a good opportunity to investigate the hydrological trigger of both slopes consisting of non-cohesive (sandy) materials and slopes consisting of cohesive (clayey) materials. With respect to the four criteria to characterize failure modes in the conducted flume experiments given in the introduction following results are obtained. (1) The failure in sand occurs suddenly, without any visible precursor, associated with a high and constant velocity. In contrast to this the time displacement graphs of the cohesive slope are divided into two parts. The first part is characterized by a very slow and constant movement. In the second part the movement velocity increases and final failure occurs. () The whole failure process in River sand takes place within a few minutes whereas the failure in the test of the Zoelen material is related to a long time period and lasts about weeks. (3) The test on sand can be described as retrogressive failure that proceeds as a series of multiple slide events and is connected to fluidization processes. The experiment performed on cohesive material shows a failure mode completely different from the one of sand. Final failure is initialized over a long time period and occurs progressively. After a shearing surface is formed the soil slumps down along the surface in several stages, with different velocities, that could be related to a renewed building up of pore pressure in opening fissures. Failure takes place without any fluidization. () The retrogressive failure in the performed test on River sand starts with liquefaction at the toe. The mechanisms leading to liquefaction are not clearly indicated. Possible explanations could be the seepage forces caused by the infiltration of water or that the material is near the liquid limit. It is emphasised that other tests performed e.g. on Limburg sand does not start with liquefaction, but with the formation of a sliding surface. In this case the contraction during shearing may generate liquefaction van Asch & Malet (submitted). To explain the failure mechanism of the test on Zoelen clay a model for the growth of landslide shear surfaces, i.e. for progressive failure, according to Petley et al. (5) is used, where shear surface growth starts with microcracking. After point interaction and coalescence of microcracks is achieved further shear surface development leads to stress concentration and acceleration of sliding movements. The difference in failure initiation in cohesive and non-cohesive materials has to be taken into account in the development of early warning systems. Especially the long term failure initialization in cohesive slopes showing progressive failure can be used in forecasting potential catastrophic failure. Due to the sudden occurrence of landslides in sandy material the measurements of displacements as indicator of failure is assessed to be insufficiently. It has to be studied if pore pressure or alternatively precipitation measurements can be a useful tool to predict potential failure. REFERENCES Bjerrum, L Progressive failure in slopes of overconsolidated plastic clay and clay shales. Journal of the Soil Mechanics and Foundations Division of the American Society of Civil Engineers 93: 1-9. Iverson, R.M., Reid, M.E. & LaHusen, R.G Debris flow mobilization from landslides. Annual Review of Earth and Planetary Sciences 5: Okura, Y., Kitahara, H., Ochiai, H., Sammori, T. & Kawanami, A.. Landslide fluidization process by flume experiments. Engineering Geology : 5-7. Petley, D.N., Bulmer, M.H. & Murphy, W.. Patterns of movement in rotational and transational landslides. Geology 3(): Petley, D.N., Petley, D.J., Bulmer, M.H. & Carey, J. 5. Development of progressive landslide failure in cohesive materials. Geology 33(3): 1-. Wang, G. & Sassa, K. 3. Pore-pressure generation and movement of rainfall-induced landslides: effects of grain size and fine particle content. Engineering Geology 9: Van Asch, Th.W.J., Malet, J.P. & van Beek, L.P.H.. Influence of landslide geometry and kinematic deformation to describe the liquefaction of landslides: some theoretical considerations. Engineering Geology : Van Asch, Th.W.J. & Malet, J.P. Fluidization potential of soils as an important aspect in landslide hazard analyses. submitted to Natural Hazards and Earth Systems.

Analysis of soil failure modes using flume tests

Analysis of soil failure modes using flume tests A. Spickermann & J.-P. Malet CNRS UMR 7516, School and Observatory of Earth Sciences, University of Strasbourg, Strasbourg, France Th.W.J. van Asch, M.C.G.