You are on page 1of 10

Three dimensional slope stability analysis using smoothed particle hydrodynamics

S. N. Manukulasuriya
Department of Civil Engineering, Monash University, Australia

Dr. H. H. Bui
Department of Civil Engineering, Monash University, Australia

Slope stability analysis using numerical techniques is important in assessing a soil slope without the need for costly on site field testing. In the past, methods of slope stability analysis include limit equilibrium methods (LEMs), the finite element (FEM) discrete element method (DEM). These methods of analysis have their flaws. LEMs and FEM are unable to model the discontinuous nature of a soil slope failure, as well as large deformation and post failure. DEM on the other hand require parameters which are not easily obtained. The limitations of these methods of slope stability analysis have made way for three dimensional slope stability analysis using smoothed particle hydrodynamics (SPH), a numerical method which uses Lagrangian framework. The stability of a particular slope configuration was assessed using SPH with different variations and boundary conditions. The soil slopes factor of safety (FOS) was gained using the shear reduction technique, where the shear strength reduction factor was plotted against the maximum displacement. The results of the analysis were validated against an FEM analysis conducted on the same soil slope. Furthermore, the critical failure mechanism was observed and compared with the FEM analysis conducted on the same configuration. The results obtained from the SPH analysis showed a similar or slightly higher factor of safety than the one attained using FEM. This is because the FEM analysis calculates their factors of safety at the point when the strength reduction factor cannot be increased any further, as the displacements which occur afterwards are too great for FEM to handle. The power of SPH is illustrated as the full failure of the slope is observed due to SPHs ability to handle large deformation. 1 INTRODUCTION Soil slopes are commonly formed in the ground whether it is natural, due to excavation, or fill. Consequences of soil slope failure include extra expenses, injury, and loss of life. For this reason it is imperative that a reliable and accurate form of soil slope stability analysis is available for engineers. Field work can be undertaken in order to analyse slope stability, however it can be very costly Computation methods of analysis are a cost effective alternative. Currently limit equilibrium methods (LEMs), the finite element method (FEM) is used for slope stability analyses, as well as the discrete element method (DEM), however these methods have limitations. The LEM method has been used in industry due to its simplicity, however it does not consider soil deformation and requires many assumptions. The inter-slide forces are ignored, and the critical failure surface must be initially speculated. FEM is currently a popular method of analysis and uses a grid based mesh. The use of a mesh however can lead to mesh tangling when simulating large deformation and post-failure behaviour of soil. The mesh used requires careful thought, and its creation should be undertaken by someone with experience. The DEM is a mesh-free method, and there does not have the issue of mesh tangling. Conversely, DEM does not use the motion of a continuum, and parameters for this method can be difficult to determine. Moreover, DEM processes use a great deal of computational power, making large scale deformation impossible to model. Smoothed particle hydrodynamics (SPH) is a mesh free Lagrangian alternative which was originally developed by Gingold and Monaghan (1977) and Lucy (1977) for astrophysical problems, and then commonly utilised for fluid simulation. SPH has also been recently developed by combining it with the DruckerPrager model to simulate slope failure and slope stability analysis, including large-scale deformation and the post failure of geo-materials (Bui et al., 2008). The accuracy and performance of SPH for geo-materials was validated against the finite element analysis of a slope commonly used for validation (H. N. Gharti et al., 2011). The results of this SPH analysis which include the factor of safety as well as the slope failure model (including failure surface) will be used to

compare with the already established finite element method. 2 REVIEW AND COMPARISON OF PAST SLOPE STABILITY NUMERICAL ANALYSES 2.1 2-D slope stability analysis In the part it was very common for practitioners to utilise 2-D methods of slope stability analysis. This is due to the simplicity of modeling a slope stability problem in only two dimensions. Furthermore, 2-D slope stability methods of analysis yield more conservative factors of safety than their 3-D counterparts. For this reason, practitioners have felt that there is no need to develop 3-D slope stability methods of analysis. 2-D slope stability analyses such as the limit equilibrium method (LEM), which include Bishops simplified method (Bishop, A. W., 1995), Felleniuss method (Fellenius, 1936), Morgenstern and Prices method (Morgenstern, N.R & Price, V.E., 1965), Spencers method (Spencer, 1967) and Janbus method (Janbu, 1968), have been popular as they do not require the stress strain relation of the soil. On the other hand these methods of analysis require information on the initial geometry of the slope, including the location of the critical slip surface. The analysis involves dividing the soil slope into slices, which assume that the inter-slice forces are negligible. Consequently, the slices are not truly in equilibrium, and ultimately lead to less accurate calculations (Griffiths & Marquez, 2007). 2.2 Finite element method Slope stability analysis using elasto plastic finite elements, name the finite element method (FEM), has been traditionally used in the past as a 3-D alternative. FEM is able to deal with complicated geometries,
boundary conditions and property variability. The major advantage of the finite element method is that it considers soil deformation properties. Moreover, the shape and location of the failure surface does not need to be initially speculated, as the algorithms allow failure to occur naturally through the zones where shear stress has been induced by gravity, and overcomes the soil shear strength. This method of analysis does not utilise slices or columns, and therefore there are no assumptions made about the side forces of each slice. This in turn means that the overall equilibrium of the slope is not affected by unjustified assumptions (Griffiths & Marquez, 2007). FEM is however time consuming. Slopes are modeled using an appropriate mesh and therefore the accuracy of the results obtained is determined by users experience. The mesh should have enough elements to model the slope stability problem accurately, but on the other hand too many elements can cause the mesh to collapse. Many

elements also accentuate rounding off errors. The main disadvantage of FEM is the use of the mesh grid. Mesh tangling does not allow FEM to model large deformation and post failure behaviour. Remeshing can help tangling issues, but it is a complicated process (Bui H. H et al., 2011).

2.3 Spectral element method The spectral element method (SEM) is a higher order FEM which allows for simpler commutation of calculations involving the stiffness matrix, strain and stress (Hom Nath Gharti, D. K., 2012). Consequently, elements of higher orders are easier to implement using SEM than with FEM. This type of meshing is quite challenging and there are only a few hexahedral meshing tools available. 2.4 Mesh free solutions Discrete-element method (DEM) is a mesh free method of slope stability analysis which can model large deformation and post failure behaviour. It is relatively simple however requires parameters which are difficult to determine. As a result this method is not accurate. Smoothed particle hydro dynamics (SPH) is a mesh free slope stability analysis which is able to simulate slope stability failure in the field accurately. SPH is able to cater for large scale exaction work, and therefore can yield the closest results to actual field results. SPH is able to model discontinuous failure of soil and evaluate the stability of a slope. It is able to simulate the post-failure behaviours of soil which is not possible using the finite element method. The factor of safety is estimated using a modified failure criterion with shear strength reduction techniques. 3 SIMULATION APPROACHES 3.1 Simulation framework The approach undertaken when simulating problems using SPH utilises a set of discrete particles. The following sections detail how the equations governing continuous motion can be discretised in order for particles to behave appropriately within the SPH framework. The co-ordinate system used is a Lagrangian framework, as it is necessary to monitor the particles as the move. The particles contain field variables such as the stress, density, temperature and displacement, as well as holding a mass. The particles behavior is dictated by the kernel function (smoothing function) which determines range of interparticle interaction (Bui et al., 2011).

3.2 Movement of soil particles The movement of soil particles is simulated within the SPH frame wake by using the following equation: (1)

these relations are detailed in Bui et al., 2011. The yield criterion which is chosen for this model is the Drucker-Prager yield criterion. When applying this condition, the stress tensor can be ultimately calculated in terms of the strain tensor with equation 5:

The parameters used here are the displacement (u), where the raised dot indicates the time derivative, density , gradient total stress tensor , acceleration caused by gravity (g) as well as the Cartesian components x, y, z which are represented by and . If the pore water pressure is taken as zero, then the following equation can be used: (2)

(5)

This soil constitute model has been validated within the SPH framework (Bui et al., 2008) and uses six soil parameters. These are the cohesion coefficient (c), friction angle (), dilatency angle (), Youngs modulus (E), Poissons ration (v) and soil density (). A detailed explanation of the mathematics involved in this approach is found in Bui et al., 2008. 4 METHODOLOGY The SPH program was written using FORTRAN, and utilised the framework outlined in the previous section. Slope stability analysis was conducted by assessing the factor of safety (FOS) for the slope under consideration. The shear strength reduction method is used in the SPH analysis as outlined in Bui et al., 2011. The shear strength reduction method reduces the real shear strength parameters of a soil slope by a factor (shear strength reduction factor, SRF) to cause the slope to fail. The value of SRF at which the slope fails is then defined as the FOS for the slope. Equations 5 and 6 illustrate this concept: (6) (7) Where c and are the real shear strength parameters, ct and t are the reduced shear strength parameters, and SRF is the shear reduction factor. The analysis is performed by initially setting the SRF to 1, which should be stable under gravitational loading, (otherwise the soil slope under consideration is unstable). The SRF is subsequently increased by small increments until the slope fails. The FOS is then defined at the SRF that failure occurs.

Equation 2 describes continuous motion, and therefore must be discretised in order to be implemented onto individual soil particles within the SPH framework. The SPH approach described in (Bui et al. 2008) is applied to equation to in order to find the movement of the discrete particles. This leads to the following equation being implemented: ( ) (3)

where i represents the particle which is under consideration; i represents the density of particle i and j represents the density of particle j respectively; N is the number of adjacent particles (adjacent particles are those within the support domain of particle i defined by the kernel function); mj is the mass of particle j; is a stabilisation term applied to counteract fluctuation of the stress and tensile instability (Bui et al., 2008); is the kernel function (Monaghan and Lattanzio, 1985) which provides smoothing and defines the range of interparticle interaction.

3.3 Stress strain relationship For the stress-strain relationship to be established, a constitutive model is used. Rate of strain of a particular elasto-plastic material ( can be divided into two main components. These are the elastic strain rate tensor ( and the plastic strain rate tensor ( . This is shown in equation 4: (4)

First the elastic strain rate tensor is looked at. This can be defined using Hookes. The plastic strain-rate tensor is found y using the plastic flow rule. Both

5 SPH APPLICATIONS AND VALIDATIONS Several slope stability analyses were conducted on the slope configuration presented in figure 1 and figure 2. The results of these analyses were ultimately validated against the FEM and SEM results determined in H. N Gharti et al. (2011) using an identical slope. This slope was initially used in 2D by Fredlund and Krahn, and has also been used by many authors to validate their methods of slope stability analysis. Three different cases are explored, which includes implementing a weak layer and water table.

sons ratio are taken as 105 kN/m2 and 0.3 respectively for all layers of soil. The dilation angle used in H. N Gharti et al.,2011 is zero, however for SPH analyses, it is recommended to use a dilation angle which is half of the cohesion, as discussed in Bui et al., 2008. The boundary conditions for all three cases are full-fixity (rough) at the base, as well as one side of the slope (nearest side in figure 1), with the other being a free-roller (smooth condition). H. N Gharti et al. (2011) contains results of several analyses conducted using three different slope stability analysis methods; Finite element method (FEM) adapted from Smith and Griffiths, serial SEM and parallel SEM. The FEM method used utilises 20-node hexahedral elements with four Gauss points. The analysis in H. N Gharti et al. (2011) using this slope configuration has been conducted on only half of the slope under consideration. The analysis conducted using SPH uses the full configuration, and consequently the simulation will be more two dimensional, and therefore the results obtained in this analysis will tend to be more conservative.
Table 1: Comparison of the factor of safety obtained using different methods

Figure 1: Slope configuration used for analysis and validation. Dry region (light grey), wet region (dark grey) and 0.5 thick weak layer (black)

Case 1 Xing Bishops modified method Janbus simplified method CLARA GLE Chen et al. Chen et al. FEM (Griffiths and Marquez FEM SEM SPH 2.18 2.18 2.4 2.262 2.187 2.17 2.122 -

Case 2 1.553 1.607 1.558 1.62 1.603 1.717 1.603 1.58 1.57 1.57 1.5 - 2

Case 3 1.441 1.511 1.481 1.54 1.508 1.49 1.5 - 2

Figure 2: 2D view of slope configuration (Zhang 1988)

Case 1 is a homogeneous dry slope which have the following material properties; soil unit weight ( = 18.8 kN/m3, friction angle () = 20o and cohesion (c) = 29 kN/m2 . The FEM mesh created consisted of 1670 elements with an average mesh size of 2. Case 2 is a non-homogeneous dry slope which contains a thin 0.5 metre weak layer situated 5 metres from the bottom of the configuration, as can be seen in figure 1 and figure 2. The weak layer has the following properties; ( = 18.8 kN/m3, friction angle () = 0o and cohesion (c) = 10 kN/m2. This layer can be thought of as a geo-synthetic, and is placed within the soil which has the properties which were outlined in case 1. Case 3 introduces a water table, as well as containing the weak layer. The level at which the soil is saturated is represented in figure 2 by the dark grey. The Youngs modulus and pois-

5.1 Homogeneous dry slope Case 1 of the slope configuration shown in figure 1 was analysed. The slope was modelled using 26400 SPH particles. As it is difficult to model a slope which has dimensions which are not whole numbers (the distance between particles would otherwise have to be reduced to extremely tiny increments in order to model exact dimensions, which require very high computational power), the dimensions of the slope were rounded off to the nearest whole number. This assumption would have an insignificant impact on the final results. Moreover, the analysis which was conducted in H. N Gharti et al. (2011) used grids which had an average mesh size of 2 metres, making the dimensions of that analysis similar to this SPH analysis.

Displacement vs. SRF


0.3 Maximum displacement 0.25 0.2 0.15 0.1 0.05 0 0 0.5 1 1.5 SRF 2 2.5 3

Figure 3: Maximum particle displacement vs. SRF (Case 1)

This displays the contour plot for the particles displacement from their original positions after 5000 iterations. The 3D SPH model predicts a very clear slip failure surface, which elicits the power of SPH. The illustration of the soil slope failure mechanism can be compared with that of the slope failure achieved using FEM in H. N. Gharti et al. (2011) in figure 5. The failure mechanism is not as easy to determine using the FEM analysis. Furthermore, the model is deformed, rather than displaying a slope collapse which soil particles would undergo. The disadvantages of FEM are brought to light here, as the discontinuous failure along the slip surface cannot be modelled. Also, it is impossible to predict the post failure behaviour of the soil using the FEM model.

The initial stress conditions were applied by a gravity force onto all the particles. The strength reduction factor was firstly set at 1, and the maximum displacement was recorded after 5000 time steps were ran. This SRF was subsequently increased gradually, and the results were plotted as shown in figure 3. The factor of safety is taken at the point when there is a sudden large increment in the maximum displacement observed. Case 1 displayed a clear large jump in maximum displacement over a small SRF increment. The increase in maximum displacement results in an almost vertical rise, once the factor of safety reached approximately 2.4. The factor of safety in this case is therefore taken as 2.4. The rapid increase in displacement, once the factor of safety reached 2.4, also illustrated the fast collapse of the homogeneous dry slope. For a more accurate factor of safety, the SRF increments would need to be decreased. This would consequently require greater computational power and time, and can be possibly explored further in future studies with increasing technology. The corresponding factor of safety found in H. N Gharti et al. (2011) for this case is 1.8, which is lower than the factor of safety obtained in the SPH analysis. This is attributed to the finite element method, as well as the spectral element methods inability to cope with large deformation due to mesh tangling. The program simply cannot run any further after the slope undergoes large displacement, and therefore cuts off at a lower factor of safety. Because the factor of safety determined using SPH includes information on high displacements, this can be thought of as a more realistic estimation of the actual factor of safety. Additionally, the factor of safety obtained in this analysis will be more conservative, as it tends to represent a more two dimensional analysis as the width of the slope in the SPH analysis is longer than the width used in H. N. Gharti et al. (2011). The critical failure mechanism for the case 1 homogeneous dry slope is seen in figure 4.

Figure 4: Homogenous dry soil slope critical failure mechanism (Case 1) using SPH

Figure 5: Failed slope for case 1 using FEM H. N Gharti et al. (2011)

5.2 Introduction of a weak layer Case 2 introduces a thin weak soil layer located 5 metres from the bottom of the slope. This weak layer can be intuitively seen to reduce the overall strength of the slope, resulting in a lower factor of safety. Figure 6 displays the results of the maximum displacement vs. SRF plot. Unlike case 1, this plot does not give a clear FOS. The change in maximum displacement determined does not show major jumps as

the SRF is increased. As a result, the graph does not have a decisive vertical section. This indicates that the introduction of a weak layer has resulted in the soil slope only gradually failing, as opposed to a rapid collapse. Even though a clear FOS cannot be obtained from figure 6, it can be seen that the maximum displacement of the soil particles shows a bigger increase with every SRF increment which occurs when the SRF is around 1.5. This indicates that a FOS could be defined within this range. However, for 3-D SPH analyses of soil slopes which only show gradual failure such as this case, a FOS may not be appropriate for assessing its stability. Quantifying a slopes stability using 3D SPH which shows gradual failure, other than a FOS, is a potential area for future research. Even though the FOS determined for the case 1 slope is higher than the FOS of case 2, it can be deduced from the graphs that the case 1 soil slope collapses rapidly. The overall strength of the case 1 soil slope is higher than that of cases 2, as case 2 introduces a weak layer. However, the case 2 soil slope fails gradually, and a significant collapse is not observed.
Displacement vs. SRF (Case 2)
1.4 Maximum displacement (m) 1.2 1 0.8 0.6 0.4 0.2 0 0 1 SRF 2 3

Figure 7: Soil slope at failure for the case 2 using SPH analysis

Figure 8: Case 2 soil slope failure surface H. N Gharti et al. (2011)

Figure 6: Maximum particle displacement vs. SRF (Case 2)

The FEM analysis conducted in H. N Gharti et al. (2011) gave a FOS of 1.57 for case 2. When conducting a FEM analysis, the soil slope starts to experience larger deformation as the SRF gets higher and so the program is unable to simulate any further. The FOS is taken at this point for the FEM analysis, and therefore does not necessarily reflect the true stability of the slope. The advantages of SPH are highlighted in this situation. SPH is able to handle large deformation and can resultantly continue the deformation process. The full extent of the soil slope failure can then be analysed, which allows for vital information to be obtained. For example, it would be difficult to determine the rate at which a slope would collapse with an FEM analysis as the full extent of the failure is unable to be modeled.

The soil slope failure mechanism for case 2 is shown in figure 7. Failure is observed in this analysis along the weak soil layer, as opposed to the circular failure seen in case 1. Displacement is also observed below the weak layer. The effects of the introducing a weak soil are highlighted here. The failure surface witnessed using SPH can be compared to the FEM analysis in H. N Gharti et al. (2011) in figure 8. Failure in the FEM analysis also occurs along the weak layer, giving further confirmation that SPH can accurately model a non-homogeneous soil slope. However the FEM analysis does not have any node displacement below weak layer. The weak layer is modeled in the SPH program using a single plane. Further research should be undertaken in the future to determine the accuracy of this failure surface. A thicker weak layer should be used, and possibility larger particle sizes. 5.3 Partially wet slope with a weak layer Extending on the model used in case 2, a water table is introduced, which is represented in the dark grey shade in figure 1. Figure 9 illustrates the pore water pressure distribution throughout the model. Like

case 1 and 2, 5000 time steps were undertaken. The displacement vs. SRF plot attained from the simulation is shown in figure 12. It can be seen that the relation is similar to that of case 2. Larger jumps of maximum displacement start to take place at a SRF of approximately 1.5 to 2. Similarly to case 2, a FOS is difficult to determine from this plot, as there is no decisive SRF increment which shows big changes in maximum displacement. For a FOS to be observed from a displacement vs. SRF plot, very small SRF increments should yield large changes in maximum displacement, which would result in the curve becoming almost vertical.

was not performed for case 3) in H. N Gharti et al. (2011) was 1.49. As earlier mentioned, the FOS of the SPH would most likely be around 1.5 to 2, as this is the SRF where larger displacements are observed. Therefore, the results of the SPH analysis using a water table coincide with that of the results obtained in H. N Gharti et al. (2011). Figure 10 illustrates the failure mechanism of the slope using SPH. Similarly to case 2, the failure can be seen be have occurred along the weak layer and also below. The failure mechanism of the SEM analysis shown in figure 12 is similar to the SPH analysis, as failure also occurs through the weak layer of soil. Nevertheless, like case 2, further analysis should be undertaken to validate this failure surface using a thicker weak soil layer and bigger particles.
Displacement vs. SRF (Case 3)
2.5 Maximum displacement (m) 2 1.5 1 0.5 0 0 1 2 3

Figure 1: Pore water pressure incurred after the introduction of a water table

SRF
Figure 12: Maximum particle displacement vs. SRF (Case 3)

5.4 Comparison of SPH results with H. N. Gharti et al. (2011)


Figure 2: Soil slope failure with water table and weak layer (Case 3) using SPH

Figure 11: Case 3 soil slope failure surface using SEM H. N Gharti et al. (2011)

When the displacement using the SPH analysis for all three cases were plotted on separate axes, the factor of safety for cases 2 and 3 were not clearly defined due to the curve never reaching a decisive vertical relation. This has been attributed to SPHs ability to simulate large deformations, and so displayed the full behavior of the soil slope failure. Due to this, validation of the SPH results against the FEM and SEM analysis conducted in H. N. Gharti et al. (2011) becomes more difficult. Subsequently, the SPH results for all three cases were limited to a displacement of 0.3m. This allows the shapes of the curves for the FEM and SEM analysis to be compared with the SPH analysis.

Consequently, it is recommended that the FOS be attained using a different method for slopes displaying a displacement vs. SRF relationship of this sort. The FOS obtained in the SEM analysis (FEM analysis

The resultant curves were plotted on a common axis, and are displayed in figure 13. These curves can be compared with figure 14, which contains the plots obtained from the FEM and SEM analysis. It can be seen from these plots that the shapes of the curves from the SPH analysis closely resemble the FEM and SEM relations. The SPH curves tend to extend more vertically at around the same SRF as that of the FEM and SEM analysis. The factor of safety in the SPH analysis is slightly higher than the FEM and SEM analyses, as the SPH results are more conservative due to the geometry used as highlighted earlier. 5.5 Change in boundary conditions Slope stability analyses were also conducted using SPH on the three cases while both boundaries were fixed (rough). Figure 18 shows the displacement vs. SRF relations for all cases with fixed boundaries. For case 1, significant change in displacement is observed at a SRF of around 2.4. The maximum displacement observed however at failure is around 0.09 m, which is nearly a third of the maximum displacement at failure observed when one of the boundaries was smooth. The FOS is harder to define for cases 2 and 3. Even though the implementation of two rough boundaries appear to have produced a more definitive relation than when one of the boundaries was smooth, using this relation to determine the FOS is still not appropriate. The failure surfaces can be seen for all three cases in figure 15, figure 16 and figure 17. The maximum displacement occurs throughout the middle of the slope, as the side boundaries are fixed. The slope failure observed with relation to time for case 3 using the fixed boundaries can be seen in figure 19.

Figure 3: Case 1 homogeneous soil slope failure (both sides fixed)

Figure 4: Case 2 non-homogeneous soil with weak layer slope failure (both sides fixed)

Figure 5: Case 3 non-homogeneous soil slope failure with water table (both sides fixed)

0.3 Maximum displacement (m) 0.25 0.2 0.15

SPH Displacement vs. SRF (All cases)


Case 3 Case 2

Case 1
0.1 0.05 0 0.5 0.75 1 1.25 1.5 SRF 1.75 2 2.25 2.5

Figure 13: SPH displacement vs. SRF plot for all cases, while only using data which is below a maximum displacement of 0.3m

Figure 14: Maximum displacement vs. SRF plot for all cases using FEM and SEM (H. N Gharti et al. (2011)

5.6 Defining the factor of safety Figure 18 also illustrates clearly the relative difference in maximum displacement between the three cases. It can be seen that case 1 (homogeneous dry slope) slope failure is more gradual, and has a much lower maximum displacement that when a weak layer is added. The addition of a water table has only slightly increased the maximum displacement of the soil particles, and yields a similar shaped curve to case 2 which does not have a water table. An FEM analysis of the three cases with fixed boundaries should be undertaken in order to determine the accuracy of these results. This is left for future research.
Displacement vs. SRF (all cases with both sides fixed)

0.8 Maxumum displacement (m)

0.6

Case 1 Case 2

0.4

Case 3

0.2

In the 3D SPH slope stability analysis conducted, a FOS was easily defined when rapid failure and collapse of the soil slope occurred. When the soil slope failed gradually, the resultant displacement vs. SRF plot did not show a definitive FOS. In Bui et al., 2011, the FOS of the SPH analysis was defined using a different approach. The maximum displacement of soil was plotted against the number of userspecified iterations for each SRF increment. The number of user-specified iterations was dependant on each problem. The FOS was determined by analysing the nature of the relation of the curves. If the curve displayed a convex shape within the specified number of iterations, then the solution was considered convergent. The SRF is then increased until the curve became concave. This concave relation represented an unconverged SPH solution, and taken as the FOS of the slope, as seen in figure 20. This is a more desirable method of defining the FOS for SPH as it displays a clear cut FOS without having to define a point at which failure occurs based on the amount of maximum displacement taken place. The method for defining the FOS in Bui et al., 2011 is recommended for future research using 3D SPH.

0 0 1 SRF 2 3

Figure 6: Maximum displacement vs SRF plots for all three cases (using fixed boundaries)

Time = 0 seconds

Time = 1.05 seconds

Time = 2.10 seconds

Time = 3.15 seconds

Time = 4.20 seconds

Time = 5.25 seconds

Figure 19: Slope failure of case 3 (Non-homogenous with water table) using fixed boundaries with respect to time modelled with SPH

taining a factor of safety should be looked into for soil slopes showing gradual failure. 7 ACKNOWLEDGEMENTS Special acknowledgement is given to Dr. Ha H Bui for teaching the SPH program, as well as for the constant guidance and supervision given throughout the course of this project.
Figure 20: Factor of safety determined in Bui et al., 2011. Factor of safety taken as 1.4, the first concave relation

8 REFERENCES Bishop, A. W. (1995). The use of the slip circle in the stability analysis of slopes. Geotechnique 5, No .1, 7-17 Bui et al.,(2008).Lagrangian mesh-free particle method (SPH) for large deformation and failure flows of geo-material using elasticplastic soil constitutive model. International Journal for Numerical and Analytical Methods in Geomechanics, Vol.32, Issue 12, pp. 1537-1570. Bui HH, Fukagawa R, Sako K. (2008). Slope stability analysis and slope failure simulation by SPH. 2008. ICSMGE 2009; 2; 1578-1581. Bui H. H et al. (2011). Slope stability analysis and discontinuous slope failure simulation by elasto-plastic smoother particle hydrodynamics (SPH). Geotechnique 61, No.7, 565-574 Fellenius, W. (1936). Calculation of the stability of earth dams.Proc. 2nd Congr. Large Dams, Washington, DC 4, 445462. Gingold, R. A. & Monaghan, J. J. (1977). Smoothed particle hydrodynamics: theory and application to non spherical stars. Monthly Notices R. Astronom. Soc. 181, 375389. Griffiths, D. V. & Marquez, M., 2007. Three dimensional slope stability analysis by elasto-plastic finite elements. Geotechnique, 57(6), 537-546 Hom Nath Gharti, D. K. (2012). Application of an elastoplastic spectral element method to 3D slope stability analysis. International journal for numerical methods in engineering, 1-26 Janbu, N. (1968). Slope stability computation, Soil Mechanics and Foundation Engineering Report. Trondheim: Technical University of Norway Lucy, L. (1977). A numerical approach to testing the fission hypothesis. Astronom. J. 82, 10131024. Zhang, X. (1988). Three-dimensional stability analysis of concave slopes in plan view. J. Geotech. Engng, ASCE 114, No. 6, 658 671.

6 CONCLUSIONS The performance of the 3D SPH slope stability analysis method was critiqued and evaluated in this project. The method of soil slope analysis was conducted on various soil slope situations and configurations. The results of these analyses were validated against FEM and SEM analyses conducted on the same slopes in H. N Gharti et al. (2011). The problems associated with FEM were highlighted throughout the analysis. The factor of safety determined for the homogenous dry slope (case 1) was found to be approximately 2.4 using the 3D SPH analysis. The FEM and SEM analysis in H. N Gharti et al. (2011) yielded a lower factor of safety. This has been attributed to FEM and SEMs inability to cope with large deformation due to mesh tangling. The factor of safety obtained in the FEM and SEM analysis is simply the highest strength reduction factor that the program could handle. A more conservative FOS also expected when conducting the analysis, as the width of the slope used was greater than the width used in H. N Gharti et al. (2011), consequently representing a more two dimensional problem. The analysis on non-homogeneous and saturated soil slopes with SPH resulted in gradual failure, and therefore a decisive factor of safety was unable to be determined. Because the SPH program was able to model the larger displacements as the strength reduction factor was, observation of this behavior is made possible. It is concluded that larger deformations are able to be modelled with SPH. Furthermore, higher factors of safety are observed as the program is able to handle large deformation, or high strength reduction factors. When the maximum displacement vs. SRF relation using SPH was directly compared to the FEM relation for all three cases, it was seen that both analyses produced similar shaped graphs. This observation validates the SPH analysis undertaken. It is recommended that in future studies that a different approach is used to obtain the factor of safety when conducting a SPH slope stability analysis. Furthermore, different methods of conducting an assessment of slope stability other than ob-

You might also like