Professional Documents
Culture Documents
3198
Static and dynamic soil pressures over a primary crusher station at a copper
mine in Chile
Bogart Mendez1, Diego Rivera2 and Erdem O Tastan3
1
Rizzo Associates Chile, S.A., Pte. Errazuriz 3724, Las Condes, Santiago, Chile
Rizzo Associates Inc., Penn Center Boulevard, Suite 100, Pittsburgh, PA 15235
3
Rizzo Associates Inc., Penn Center Boulevard, Suite 100, Pittsburgh, PA 15235
2
ABSTRACT: A 36-m high crusher station will support the ROM Pad fill along its
lateral and back perimetral walls. This paper describes the analysis performed to
determine soil-induced lateral stresses on the crusher stations back wall under both
static and dynamic loading. A 3D numerical model is used for the analysis. Static and
dynamic material properties are obtained from geotechnical and geophysical
explorations conducted on site. The seismic environment is characterized considering
Chilean regulations. Soil-structure interaction is considered using interface elements.
Dynamic soil stresses obtained from the numerical model are compared to the
analytical solutions available in the literature. Soil-induced stresses as obtained from
the numerical model are between active and at-rest seismic stresses. Results reveal
that for certain applications, like the one described in this paper, selecting the
appropriate analytical solution, e.g., yielding vs non-yielding conditions, is not
straightforward and the use of numerical models is recommended.
INTRODUCTION
The construction of a new primary crusher station is proposed as part of a mine
expansion plan. The crusher station is about 36 m high and will serve as a retaining
element on three of its four faces to support the surrounding ROM Pad fill.
The dynamic behavior of the crusher station is of great interest due to the (1) high
seismicity in the area, (2) high frequency content and significant vertical acceleration
of Chilean earthquakes and, (3) the complex interaction between the ROM Pad fill
and the crusher station due to the large mass of the ROM Pad behind the building.
To address the unique difficulties, a detailed 3D numerical model is developed with
a particular interest to investigate the type of soil-induced lateral stresses developed
under both static and dynamic loading conditions. The model captures 3D stress
effects, and considers stress-dependency, spatial variation of material properties and
soil-structure interaction.
Soil-induced lateral stresses are also computed with simplified conventional
methods commonly used in engineering practice for yielding and non-yielding
conditions.
Page 1
3199
MODEL GEOMETRY
The dimensions of the model are 280 m 350 m in plan view with variable depth
down to 100 m, as seen in Fig. 1.
The ROM Pad fill will be built in two stages: the first stage consists of the ROM
Pad marked as zone A in Fig. 1(a). The second stage or final fill corresponds to the
backfill between the first stage of the ROM Pad, the crusher station and MSE walls
[zone C in Fig. 1(a)]. Zone B is the excavation for the conveyor system that will
transport crushed material.
MSE walls will be used as the front face (boundary) of the ROM Pad. Fig. 1(a)
depicts the plan view of the FLAC3D model, and Fig. 1(b) depicts an isometric view
of the geometry. Dimensions shown in Fig. 1(a) are as follows: width of mat
foundation, B = 23 m, length of final fill behind the building, d = 60 m.
Cross sections A and B are shown in Fig. 2. Fig. 2(a) shows that the depth from the
building foundation to the rock stratum varies along the foundation length, L (42 m),
i.e. the rock stratum dips.
Tetrahedral elements constitute the model mesh. Since FLAC3D uses low order
constant stress and strain elements, the use of Nodal Mixed Discretization (NMD) is
considered for an accurate modeling of plasticity material response (Itasca, 2009).
The mesh geometry is optimized in terms of computational time vs. mesh
refinement. The same mesh is used for both static and dynamic analyses. The
optimization is based on static stress comparison, free-field seismic site response and
computing time with/without NMD. The use of a single model for static and dynamic
computations is a key feature for the project analysis stage, as this allows for a
reasonable run time.
To adequately address the high frequency content, the mesh element size follows
the recommendations presented by Kuhlmeyer and Lysmer (1973).
The interaction of the ROM Pad fill and the crushing station is addressed using
interface elements in the model.
BOUNDARY CONDITIONS
For static analysis, roller boundary conditions were assigned to each model outer
face, i.e. the boundaries were restrained only in the direction normal to each face.
The distance from the area of analysis (location of the crusher station) to the model
boundaries is large enough so that boundary locations do not affect the solution. An
analysis is conducted to verify that at the boundaries of the model the stress increment
distribution due to foundation loading is less than 5 % of the maximum value.
For dynamic analysis, free-field boundaries are used for the lateral boundaries of the
model, while a quiet boundary is utilized for the base of the model. These boundary
conditions are fully incorporated into the dynamic option of FLAC3D. The free-field
boundaries at the sides of the model account for the free-field motion that would exist
in the absence of the structure. On the other hand, the quiet boundary at the base is
very effective in absorbing the body waves approaching the boundary, so that they are
not reflected back into the model.
Page 2
(a)
3200
(b)
FIG. 1. (a) Plan and (b) isometric view of the numerical model in FLAC3D.
(a)
(b)
FIG. 2. View of cross section A (a) and cross section B (b) in the FLAC3D model.
MATERIAL PROPERTIES AND CONSTITUTIVE MODELS
The model developed for the analysis of the crusher station involves excavations in
natural soils, backfilling and embankment construction.
A geotechnical exploration campaign was performed to characterize the site. The
campaign included geophysical testing (downhole, seismic refraction and refraction
microtremor method), laboratory testing, in-situ plate bearing testing, SPT,
exploration pits and drilling down to the rock stratum with core sampling. The
exploration campaign aimed to characterize both the geotechnical units and their
spatial distribution over the site. Four main formations are identified:
Page 3
3201
The spatial distribution of these strata was determined mainly from geophysical
exploration and drilling data. The variable thickness of all strata across the site can be
seen in Fig. 1 and Fig. 2. Only the first rock stratum is considered in the analysis, as
rock strains are assumed to be negligible in higher depths due to high stiffness.
The strength of geomaterials is based on triaxial testing of reconstituted samples,
the results of in-situ plate bearing tests and engineering judgment based on the
knowledge of in-situ soils.
The Mohr-Coulomb constitutive model is considered as a good approximation for
modeling the soil strength under static load conditions for Stratum I and Stratum II,
while a linear elastic material model is attributed to the rock Stratum III. In order to
obtain a better approximation of the soil stiffness variation with confining stresses for
Strata I and II, the Youngs modulus is calculated considering the variation of the
octahedral stress and the soil at-rest pressure coefficient. The variation is obtained
using all the available data gathered in the exploration campaign.
The properties for backfill materials are estimated based on available results from
laboratory testing and upon engineering judgment along with local experience with
the materials on site. It is considered that all backfill materials are granular and will
be compacted up to 95% of its maximum dry density according to regular
construction standards. Backfill materials are divided in two groups: F1 is the group
for the ROM Pad fill and the final fill material and F2 includes the fill used for the
unreinforced portion of the MSE walls and structural fill materials. The ROM Pad
material is modeled with a linear elastic constitutive model. This model is adopted
because the ROM Pad fill yielding analysis is not a priority for the project, as this
embankment is allowed to accommodate large strains without posing any risks to the
operation of the crusher station. Youngs modulus for backfill materials is also
considered to be stress-dependent.
Since the purpose of the analysis is not to evaluate the stability and behavior of the
MSE walls, they are accounted for in a simplified manner. The MSE walls are
modeled as high stiffness soil blocks, in accordance with the usual external global
stability analysis of such reinforced soil masses.
The vertical faces of the MSE walls pose a complex modeling issue if the soil
reinforcement is not explicitly accounted for, because of the material failure of such
vertical faces when considering a material failure criterion. Hence, to avoid such
modeling problems an elastic constitutive model is used for the MSE wall elements.
The static strength and deformability properties for in-situ soil and backfill
materials are shown in Table 1.
The mat foundation is modeled as a reinforced concrete using solid elements with
linear elastic behavior. The Crusher station building is modeled as a single solid
volume, using a linear-elastic constitutive model; equivalent material properties are
considered to adequately represent its stiffness; equivalent properties are computed
from the actual building lateral stiffness.
Page 4
3202
For dynamic analyses, a linear-elastic model is used for all materials. Elastic moduli
for all strata are obtained based on geophysical data (Gmax). Strength and material
properties are shown in Table 2.
Table 1. Strength and elastic material properties for static analyses.
Material
K0
(kg/m3) ( ) Es (MPa)(1)
Stratum I
1910
38
56
0.38
Stratum II
2070
40
114
0.36
Stratum III (Rock)
2120
4272
ROM Pad (F1)
1910
130
Final fill (F1)
1910
36
130
0.41
Unreinforced MSE (F2)
1910
38
200
0.38
Structural Fill (F2)
1910
38
257
0.38
Reinforced MSE
1910
820
Crusher Station
1600
4000
Foundation Mat
2400
22000
(1)
Static
Elastic
Modulus
(large
strains),
values
at
center
of
strata
(2)
Mohr-Coulomb model
Model
MC(2)
MC
Linear Elastic
Linear Elastic
MC
MC
MC
Linear Elastic
Linear Elastic
Linear Elastic
Stratum I
0.33
330
553
Stratum II
0.30
550
1628
Stratum III (Rock)
0.30
1450
11589
ROM Pad and final fill (F1)
0.33
330
553
Un. MSE and structural fill (F2) 0.33
330
553
Reinforced MSE
0.33
402
820
Crusher Station
0.30
4000
Foundation Mat
0.30
22000
(1)
Dynamic Elastic Modulus (small strains), values at center of strata
Model
Linear Elastic
Linear Elastic
Linear Elastic
Linear Elastic
Linear Elastic
Linear Elastic
Linear Elastic
Linear Elastic
Page 5
3203
Page 6
1.20
1.20
1.00
0.80
Sa (g)
0.80
Sa (g)
1.00
3204
0.60
0.60
0.40
0.40
0.20
0.20
0.00
0.00
0.10
1.00
F (Hz)
10.00
100.00
0.10
1.00
F (Hz)
10.00
100.00
Sa (g)
(a)
(b)
FIG. 3. (a) Comparison of the synthetic motion spectrum to the target NCh433
spectrum, and (b) comparison of the surface spectrum obtained using the
deconvolved motion.
1.80
1.60
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00
0.10
1.00
F (Hz)
10.00
100.00
ANALYSIS RESULTS
Results are presented in terms of soil-induced lateral stresses for both static and
dynamic loading phases. Dynamic loading is applied after static equilibrium is
reached. Static analysis considers construction sequence and stress-dependant soil
modulus. Further details about static analysis are provided in Mndez and Rivera
(2013).
Static Soil Stresses
The sequence for static analysis is as follows;
1) initial equilibrium of the free field to achieve geostatic stress state,
2) initialization of model displacements and construction of the first stage of the
ROM Pad [zone A in Fig. 1(a)],
3) performing all the excavations required in the model [zone 3 of Fig. 1(a) and
excavation for the building mat foundation],
4) initialization of model displacements and beginning the simultaneous construction
of the building and the MSE walls on each side of the building,
5) construction of the final fill behind the building (second construction stage of the
ROM Pad). This sequence of analysis is based on the planned execution stages of the
Page 7
3205
project. The initialization of model displacements is done to address the effect of the
fills around the building on both induced settlements and soil-induced lateral stresses.
Computed FLAC3D static soil-induced stresses over the center of the back wall of
the crushing station building are presented in Fig. 5(a), along with theoretical values
for active and at rest stresses. Also, the values of total lateral force for each case are
presented in the same figure. The static soil-induced stress distribution, as well as the
total lateral force, is very similar to the at rest condition and not to the active stress
distribution, as commonly assumed in the design of conventional retaining walls. It is
reasonable to attribute the development of the at rest soil-induced stress distribution to
the influence of the ROM Pad fill behind the building. This is because the weight of
the ROM Pad causes settlements behind the building, thus inducing rotation of the
crushing station towards the backfill [see Fig. 5(a), contours correspond to horizontal
displacement], consequently developing a soil-induced stress state very similar,
although somewhat larger than, to the at rest condition. However, it is interesting to
note that the numerically-computed soil-induced stress distribution alternates between
the active and the at rest theoretical values. Also, note that near the top of the wall,
stresses are lower than the active stresses. These values are computed as a result of
the interaction between the fill and the building through the interface element, which
can account for full separation between both materials. It is likely that these lower
stresses result from a three-dimensional effect of the upper part of the wall, which is
less confined than the lower parts of the building wall. This complex behavior of the
soil-induced stresses cannot be obtained using simplified conventional methods.
Therefore, for non-conventional cases, like the one presented herein, selection of the
adequate simplified method for computing lateral soil stresses can be challenging and,
in some cases, not possible. Accordingly, a detailed numerical evaluation of soilinduced lateral stresses and forces is recommended.
Additionally, Fig. 5(b) presents the soil-induced stress contours over the back wall
of the building (compression stresses are negative). A fairly regular stress distribution
can be observed in Fig. 5(b). Small stress concentrations result on the lower corners
of the building, where the building and the mat foundation join. However, these stress
concentrations are not expected to influence the overall stresses.
Dynamic Soil Stresses
After computing the static stress state, seismic loading is applied in the horizontal
(Y direction, see Fig. 5) and vertical directions. A linear elastic analysis is performed,
considering a combination of local and Rayleigh damping for the high frequencies.
Soil-induced stress distributions computed with analytical methods are also presented
for comparison purposes. It is worth to recall that the building has three of its faces
covered by backfill, and only the front face is free, as seen in Fig. 1.
Static and dynamic stresses (total stresses), are shown in Fig. 6(a). Results are
computed with both analytical methods and FLAC3D. Results for FLAC3D are
presented in terms of average stresses (average of stress time series for all nodes at
the center of the back wall along its height), average stresses plus two standard
deviations (Av+2SD) and as the collection of maximum stresses along the wall height
(Max).
Page 8
3206
(a)
(b)
FIG. 5. FLAC3D-computed dynamic soil stresses for one and two acceleration
components
The envelope of maximum stresses assumes that all nodal maximum stresses occur
at the same time step, a very conservative condition. However, it is observed that this
plot agrees well with results from Ostadan and Wood methods, which account for
non-yielding wall conditions.
On the other hand, the average FLAC3D results align well with the M-O method for
active conditions (using a horizontal seismic coefficient, kh, equal to Ao/2). A soilinduced stress distribution between both limits is computed as the average distribution
plus two standard deviations. Fig. 6(a) shows that this distribution is similar to the MO method using Whitman (1991) kh = A0 recommendation to approximate the nonyielding wall condition. The overall trend of inverted triangle distribution is fairly
reproduced in FLAC3D results, as shown in Fig. 6(b). It is noted from the stress
distribution that the lower part of the wall is subjected to high stresses. This is
attributed to the base shear developed at the bottom of the building; the mat
foundation is fully buried along its perimeter and total height (2 meters), while the
building has a free face on its front. As a result of this particular condition, it is
expected to observe larger lateral stresses at the bottom of the wall, as computed with
FLAC3D, and shown in Fig. 6.
Seismic lateral forces obtained from FLAC3D analysis (Av+2SD) fall somewhere
in between results from simplified analytical methods. The difference among these
methods is relatively high. Therefore, a detailed assessment of the nature of the
Page 9
3207
(a)
FIG. 6. (a) Static + dynamic soil stress, (b) dynamic stress
(b)
CONCLUSIONS
From the static and dynamic analyses carried out, it is possible to draw the
following conclusions:
- The weight of the ROM Pad backfill behind the building induces a tilt towards the
fill, thus developing a soil-induced lateral stress distribution close to the at-rest
condition. This condition is observed from the results of the analysis, and it was not a
preliminary assumption as required when using simplified analytical methods.
- Dynamic soil-induced lateral stresses are between the active and at-rest condition,
as observed from numerical results. However, numerical results align fairly well with
the M-O results for the case of considering a horizontal seismic coefficient equal to
the value of the peak ground acceleration. Despite this trend for lateral stresses,
numerical results for lateral seismic forces are about 60 % larger than those computed
with the latter criterion.
Page 10
3208
Page 11