Professional Documents
Culture Documents
GEOTECNIA
UNIVERSIDAD DE LA GUAJIRA
INGENIERIA CIVIL
RIOHACHA-2020I
1) INTRODUCTION
In this article, a physics-based 3D model is developed and tested. slope deformation models that
couples solid deformation with fluid flow in variable soils. The potentiality of the coupled model to
realistically quantify stresses and deformations necessary to cause the slope to break. Continuous
nature of modeling is highlighted approach that does not quantify the factor of safety of the slope
per se, unlike methods based on limit equilibrium concepts. Instead, our model predicts spatial and
temporal variations in internal stresses, pore water pressure, degree of saturation and deformation
within the slope, such as rainfall functions that vary spatially and temporally intensity. So instead of
having a scalar quantity called factor of safety, we evaluate the potential for failure of a given slope
based on the stresses and strains that arise from a function of prescribed force (i.e. rain intensity).
Strain band based stability criteria are used in this paper to assess the potential for slope failure.
Assess whether the proposed model can Integrate important variables into a physics-based
characterization of field conditions and still obtain a description of the slope failure, we tested the
approach with complete and comprehensive data from Coos Bay experimental catchment (CB1).
The highly instrumented site was originally chosen as a hillslope-scale ‘‘laboratory’’ for conducting
sprinkling experiments aimed at developing and testing hydrologic response models. Experimental
and field data generated from the site are plentiful, albeit most of them pertain to hydrologic
response data. However, the highly instrumented CB1 slope failed as a large debris flow in
November 1996, raising some interesting questions related to the geotechnical aspects of the site.
Given what we know about the topography, hydrologic constraints, and the geotechnical outcome
at CB1, this case study seems ideal to test the proposed continuum slope model. However, it must
be noted that despite much effort to constrain the site with extensive field instrumentation and
investigation, much uncertainty remains particularly with respect to the geotechnical and hydrologic
boundary conditions appropriate for the site. Therefore, to underscore what we still do not know
about CB1, we emphasize that the analysis reported in this paper pertains only to a slope similar to
CB1, and not to the slope at CB1 itself.
The slope at CB1 is steep, on the order of 43 degrees. Any critical state soil mechanics model would
predict an initial stress condition within the slope that lies on the ‘‘dilatant side’’ of the critical state
line because of the high shear stress to effective-mean-normal stress ratio generated by the steep
slope condition. It would thus be reasonable to characterize the constitutive response of the soil
skeleton with a Mohr–Coulomb plasticity or similar models, since the compression cap is unlikely to
be activated particularly when the effective-mean-normal stress decreases further as a result of the
loss of suction with increased saturation. A simpler plasticity model such as the Mohr–Coulomb
model also requires fewer material parameters than any of the more sophisticated constitutive
models available in the literature. For the soil at CB1, which is mostly colluvium,
laboratorydetermined values of friction angle are available.
2) CONTINUUM SLOPE MODEL
The degree of saturation, K and Ks are the elastic bulk moduli for the solid matrix and solid
constituent, respectively, pw and pa are the intrinsic pore water and pore air pressures,
respectively, and 1 is the Kronecker delta tensor. Borja demonstrated this expression for the
constitutive stress tensor to be energy-conjugate to the solid rate of deformation. Under special
cases, r0 reduces to Bishop’s stress when K/Ks = 0 (incompressible solid grains) and ww = v; to the
Skempton [60] and Nur-Byerlee [45] stress when ww = 1; and to the Terzaghi [62] stress when K/Ks
= 0 and ww = 1. The Bishop parameter v can be determined experimentally [6, 40], and the
substitution of degree of saturation ww in lieu of this parameter is a simplified approximation
derived from the volume averaging over a so-called representative elementary volume, REV. For
purposes of constitutive modeling of the soil layer, we will use the Bishop stress (K/Ks = 0), with v =
ww.
for the near-surface condition, the equilibrium of the momentum at the subsurface is given by the
equation:
where /s and /w are the solid and water volume fractions in the REV, respectively; qs and qw are the
intrinsic solid and water mass densities (i.e., mass of constituent a per unit volume of constituent a),
respectively; q is the total mass density of the mixture (ignoring the mass of air phase); and g is the
gravity acceleration vector. Assuming that water is incompressible, balance of water mass in the
sub-surface takes the form:
where v is the velocity of the solid matrix, qw ¼/wðvw vÞ is the relative discharge velocity, hb is the
specified rate source/sink, he is the rate of water exchange with the surface continuum, and d(.)/dt
denotes a material time derivative followingthesolidmotion.
where FINTðdÞ is the internal nodal force vector arising from the effective stress r0, G and GT are
the discrete gradient and discrete divergence operators, respectively, which depend on p through
the degree of saturation Sr in the unsaturated regime, M is a coupling matrix that vanishes at full
saturation, U is an effective conductivity matrix that also depends on the degree of saturation, and
FEXT and GEXT are prescribed vectors of momentum and fluid supplies.
The coupled formulation presented above requires mixed finite elements with nodal pressure and
displacement degrees of freedom. Low-order mixed finite elements, with equal order interpolation
for displacement and pressure, would be ideal particularly in 3D when the number of equations can
increase dramatically. Unfortunately, loworder mixed finite elements create a numerical difficulty
in that in the limit of full saturation and either low permeability or fast loading rates, the pore fluid
can impose near or exact incompressibility on the deformation of the solid matrix and create
spurious pore-pressure oscillation. To address this difficulty, we use a stabilization scheme proposed
by White and Borja [73] based on the concept of polynomial pressure projections (PPP), which was
originally proposed in [8, 9, 20] to stabilize the Stokes problem. The primary motivation for using
stabilization is computational efficiency: with this technique, it would now be possible to use
loworder finite elements, such as those shown in Fig. 1, without the undesirable pore-pressure
oscillation that otherwise afflicts the solution without stabilization.
The sediment in CB1 is a surface sediment that is derived from dead or fresh rock and soil. The
bedrock of colluvium is a turbidite sandstone. The soils are well mixed, they are not plastic and gravel
sandstones, the thickness is well defined from the soil borings. Saturated hydraulic conductivity was
determined from slug tests, soil-water content, and porosity from TDR measurements, and
hysterical capillary pressure relationships were established from experiments. Discharge data
suggests that runoff generation occurs from water stored in small, poorly connected pores and
bedrock fractures, and saprolite connecting with larger macropores during storms. Tracer data
suggest that the two most important flow pathways in CB1 are saturated fast flow through shallow,
fractured bedrock and vertical percolation in the vadose zone. Piezometer logs show that storm flow
in the shallow, fractured rock zone exerts the most prominent control over the development of pore
pressure in CB1. The tensiometer data say that the flow of water through the unsaturated zone
provides control over the development of pore pressure and cannot be ignored in slope stability
evaluation models.
Low confining stress triaxial shear tests demonstrate that the colluvium in CB1 has no cohesion.
Cohesion in logged forests in the Oregon Coast Range was estimated to be less than 10 kPa and may
have explained how such a steep slope at CB1 could be sustained by it. Low confining stress triaxial
tests gave an internal friction angle of 40 and no cohesion. This is in agreement with other
measurements near the site that gave internal friction angles ranging from 35-44 and no cohesion.
The stress-strain behavior, performed in low confining stress triaxial strength tests, was
approximately linear for low applied stresses with a clear transition to non-linear material behavior
in axial strains greater than 1-2%. The saturated density of the soil is 1,600 kg / cu.m. and the density
in the soil water content of 20-30% is around 1,200 kg / cu.m.
Fig. 1: Stabilized low-order mixed finite elements for solid deformation / coupled fluid diffusion:
quadrilateral with four nodes for 2D (left), and hexahedron with eight nodes for 3D (right).
Fig. 2 Location map of experimental CB1 basin near Coos Bay in the Oregon Coast Range, USA
Reproduced from Ebel et al.
Fig. 3 Topographic map of experimental basin CB1. The two-dimensional planar strain condition is
assumed along section A-A for the coupled solid strain flow analysis. The dashed curve B delineates
the extent of the debris flow zone in CB1 since November 1996. The largest region C defines the
boundary of the FE hydrologic model Ebel et al. [22] 3D. The color bar is elevation in meters.
4) Hydromechanical model
The problem of a hillside scale numerical simulation is to determine the extent of spatial and
temporal discretizations, and what initial and boundary conditions. Ebel et al, performed 3D fluid
flow simulations based on the Richards Equation of variable saturated flow in the subsurface using
a complete FE InHM code. Which simulated a part of the sediment and fractured rock in which they
believed they could reasonably impose a boundary condition of no flow. The spatial extent of your
domain description is shown in Fig. 3.
The deformation in the bedrock would be too small to be of concern in the solid in the deformation
/ flow analysis. However, there are two important factors that make the present simulations more
demanding than the fluid flow simulations performed by Ebel et al .:
a) Each node of the FE mesh is composed of 3 degrees of freedom in 2D, and 4 DOFs in 3D; so
the number of equations increases very fast.
b) The water retention curve has a steep slope near the wetting front, which requires a high
resolution mesh throughout in the unsaturated zone.
Knowing how difficult it is to model the exact conditions of CB1 due to doubts in defining what
constitutes a 3D representation, we think it would be easier to consider a much simpler 2D planar
deformation representation, where the boundary conditions can be restrict more easily. To do this,
we selected a cross section that passes through the steepest part of basin CB1, shown as section AA
in Fig. 3, building the mesh shown in Fig. 4. This 2D representation of the slope in CB1 is conservative
in the sense that the out-of-plane forces of the 3D effect emanating from lateral root cohesion,
friction, etc. The mesh shown in Fig. 4 has 20,000 nodes and 18,981 mixed quadrilateral finite
elements, resulting in a total of 60,000 DOFs.
Figures 5 and 6 show the suction / saturation and saturation / relative permeability relationships
that are used in the simulations. The degree of saturation (Sr) is determined from the Van Genuchten
model as:
The model contains some parameters: w1 is the residual water saturation, w2 is the maximum water
saturation, (sa) is a scale pressure, and n and m are constants that define the shape of the saturation
curve. The constants n and m are not independent, but are related to each other as:
The values of the parameters used are in Table 1. To calibrate the model, the in situ retention curves
measured by Torres et al. For the CB1 site, hysterical effects were neglected and only wetting
measurements were used for calibration. In the hydrological model, except for a slight adjustment
in the scale pressure. The steep geometry of the saturation and relative permeability relationships
proved to be numerically challenging in the coarse mesh, and a slight flattening of the curve
improved the robustness of the model. The elastic parameters K and m are typical of the sand /
gravel mixture subjected to a confining pressure comparable to that prevailing in the CB1 sediment.
The elastic parameters influence the displacements, but have little effect on the failure mechanism,
which is largely determined by the plasticity model. The parameters of the Mohr-Coulomb plasticity
model are cohesion c, angle of friction /, and angle of dilation w, and range of values is also
summarized in Table 1.
Fig. 4 Finite element mesh for simulations of problems similar to CB1 (cf. section A-A of Fig. 3)
Fig. 5 Soil-water retention curve data and van Genuchten curves used by Ebel et al. for hydrological
simulations. We used (approximately) the wetting curve of Ebel et al. for the saturation-suction
relationship to capture the mechanical response during the wetting phase of the numerical
simulations.
Fig. 6 Variation of relative permeability with degree of saturation based on the wetting curve.
5) Results
In general, the failure mechanisms generated by the 2D slope model are complex and highly
dependent on the imposed flow boundary conditions. In the first part, we show the impact of
boundary flow conditions on the resulting deformation and failure patterns. In the second part, we
carry out parametric studies to investigate the sensitivity of calculated deformations and failures to
variation in material parameters.
We consider three flow boundary conditions - Case A: a zero fluid pressure is prescribed at the slope
surface; Case B: rain infiltration is prescribed on the slope surface in the form of fluid flow; and case
C: infiltration of fluids into the sediment at the interface between the sediment and the fractured
bedrock is prescribed. The material parameters are summarized in Table 1. For the current
simulations, baseline values of c = 4 kPa for cohesion / = 40 for friction angle and w = 25 for dilation
angle were used.
The gravitational load was turned on along with this initial pore water pressure to obtain the initial
effective stresses, after which the nodal displacements were reset to zero. A standard 2 9 2 Gauss
rule of integration was used for stabilized quadrilateral mixed elements, and an implicit backward
scheme was used for time integration.
For case A, we assume that the pressure of the porous water at the slope surface to rise linearly
from -1.5 kPa at = 0, to 0 kPa at = 3 h (Dirichlet limit condition), with all other flow limits of fluids
that are assumed to be impermeable (Neumann boundary condition). This represents heavy rainfall,
enough to saturate the slope surface within 3 h and keep it saturated thereafter. Figure 7
summarizes the hydromechanical responses calculated at the beginning of the primary and pending
secondary failure mechanisms. The general mechanism is as follows. As the sediment becomes
saturated, local plastic zones develop at the base and stretch the slope segments unevenly.
Fig. 7 Multiple sliding block mechanism on a steep slope subjected to an increase in porous water
pressure limit condition from -1.5 kPa to 0 kPa in 3 h on sloping surface
For case B, we specify a seepage infiltration of 15 mm / h on the slope surface (Neumann boundary
condition) maintaining the same flow boundary conditions on the other faces as in case A (an initial
set of conditions fully human flow contour). This ultimately caused the lower end of the slope to
saturate and the pore water pressure to rise indefinitely. Since the pressure of porous water cannot
exceed zero at the slope surface, the boundary condition changed from a seepage type (Neumann)
to a pressure type (Dirichlet) at the slope surface where the porous water pressure reached zero.
Figure 8 shows the calculated hydromechanical responses. The failure mechanism is similar to a
multiple block type as in case A, but note that the new boundary conditions have mobilized the
failure within a different segment of the slope (at x = 16 and 25 m). The location function also shows
a propensity of the sediment to develop the location of the shear strain where the plastic strain is
concentrated.
Fig. 8 Multiple sliding block mechanism on a steep slope subject to 15 mm / h rain infiltration on the
slope surface.
For Case C, we also assume an initial all-human boundary condition with zero flow at all surface
boundaries except over a finite segment between x = 12 and 18 m at the sediment-rock interface.
Therefore, this segment of the fractured bedrock is assumed to be a source of fluid. There has been
much speculation about the flow of fluids from the fractured bedrock at CB1. Although its
importance in hydrological aspects has been elucidated from previous fluid flow simulations [22], its
impact on the mechanical responses of sediments is poorly understood. To isolate the effect of fluid
infiltration from the fractured bedrock, we assume that the upper surface is a no-flow boundary
unless the pressure plume reaches it, at which point it changes to a face p = 0. Figure 9 suggests that
With just the influx of fluids from the fractured rock, a great deal of seepage is needed to build up
enough pressure to fail the slope.
Fig. 9 Failure mechanism on a steep slope subjected to upward seepage at the base of the slope.
5.2 Impact of the variation of the resistance parameters on the failure mechanism
Figure 10 shows the impact of uncertain cohesion on the failure mechanism of the 2D slope model.
The friction and expansion angles were set at 40 and 25, respectively, and the loading condition was
defined by case B (prescribed leakage on the slope face). We recall that at the CB1 site, the
parameter c was derived mainly from the cohesion of the lateral root, which could vary in space and
time. For example, as the sediment stretches, the root force could be mobilized, leading to an
increase in c, but could also subsequently decrease as the roots are withdrawn from the host
sediment. The figure shows the failure mechanisms and the cohesion failure time of 2, 4 and 8 kPa.
At 2 kPa the slope failed for the initial setting, just below its own weight. The trends observed are
what one would expect: greater cohesion delays the moment of failure. However, at c = 8 kPa, only
a single zone of plastic emerged on the slope face, not multiple.
Fig. 11 Variation of the failure mechanism and timing of the failure with the friction angle.
Finally, we consider the effect of the dilation angle on the predicted failure kinematics. Figure 12
compares the distribution of plastic deformation on the slope using two dilation angles, 10 and 25,
while the friction angle remains fixed at the base value of 40. In both cases, a failure surface is
observed that is extends to the slope surface approximately x = 25 m, suggesting that this failure
mode is insensitive to dilation. In the low dilation angle simulation, however, the secondary failure
in the region x = 16 m is suppressed. There is some indication that with additional strain, a secondary
failure may initiate at this point, but the moment of the two would not be as closely linked as in the
case of the high angle of expansion.
Fig. 12 Variation of the failure mechanism and failure timing with angle of dilation.
This assessment is particularly true since completely three-dimensional analyzes are considered in
complex domains. However, there are at least two non-trivial issues that must be addressed to apply
this algorithm to coupled fluid flow / solid deformation problems. The first is that a suitable
refinement / coarsening criterion must be developed that takes into account the coupled nature of
the fluid-solid problem, see [23, 56]. The second consideration is that the application of adaptive
mesh refinement to a plastic medium introduces additional questions about how to handle the
treatment of internal plastic variables.