Jordan Journal of Civil Engineering, Volume 4, No.

3, 2010

A Numerical Study for Geomaterials Shear Strength Components Using

Discrete Element Models

Mustafa Alsaleh 1) and Keven Hofstetter 2)

1)
Engineering Specialist, Research and Development Engineer; Virtual Product Development Technology;
Caterpillar Inc., Mossville, IL 61616, Alsaleh_Mustafa_I@CAT.com ( Corresponding Author)
2)
Program Manager; Virtual Product Development Technology; Caterpillar Inc., Mossville, IL 61616
HOFSTETTER_KEVEN_W@CAT.COM

ABSTRACT
Geomaterials (ranging from clay to gravel) are usually composed of individual particles that have specific
engineering properties. Those particles once packed to a certain density, exhibit a distinguished
macromechanical behavior, which is a result of their micromechanical interactions at the contact levels. Soil
masses are usually subjected to direct normal and indirect shear stresses; yet, they normally show shear type
of failure as indicated by many researchers using experimental and numerical evidences. The shear strength
concept of friction and cohesion is discussed in this paper. A Discrete Element Code (developed and owned
by Caterpillar, Inc.) was used in this study to show that it is possible to drop the apparent cohesion portion
and compensate for that with additional frictional resistance. Apparent cohesive bonds usually fail before
mobilizing the fictional resistance and, therefore, we may not account on it to resist future stresses. The
numerical simulations results for triaxial tests and excavation operations showed consistency regarding the
proposed shear strength components. Triaxial simulations for fine-grained materials showed that it is possible
for a numerical model to capture the stress–strain behavior if the cohesion component is dropped and, instead,
additional frictional component is added to account for the dilation that many classical soil mechanics laws
usually ignore. Likewise, excavation operations showed similar results using the same proposed theory. Some
important observations regarding the apparent cohesion concept are discussed and shown in this paper.

KEYWORDS: Geomaterials, Apparent cohesion, Friction, Micromechanics and continuum

mechanics, Virtual triaxial tests, Excavation and Non- cohesion theory.

INTRODUCTION and stability. Mohr-Coulomb shear strength theory

assumes that there are two components that give the soil
The shear resistance components in geomaterials give its shear strength: internal angle of friction and apparent
the materials their shear strength; the name here (shear cohesion (Alsaleh, 2004). These are two
strength) indicates that those components should resist phenomenological parameters than can be obtained for a
shear stresses. Coulomb’s equation for the shear strength given soil utilizing simple laboratory tests and using
was used by many researchers to describe the constitutive linear regression analysis. During soil deformation, the
behavior of soils. Otto Mohr later came up with his particles tend to roll and slide over each other, while
known envelope that represents a limit for shear strength surface friction, angular friction and particle interlocking
act to prevent such kinematics from taking place (Alsaleh
Accepted for Publication on 15/7/2010. et al., 2004). Once those resistances are fully mobilized,

- 183 - © 2010 JUST. All Rights Reserved.

A Numerical Study… Mustafa Alsaleh and Keven Hofstetter

translational and rotational motions are not restricted behavior of soil masses.
anymore. As a result, tremendous deformations occur and It is generally agreed that the shear strength of the
the soil mass enters an instability phase that eventually normally consolidated clays, sands and gravels is highly
leads to failure and massive fragmentations. The ultimate dependent on the microproperties of the solid particles
resistance stress at this point is called the shear strength (size, local void ratio, shape and surface roughness)
of the soil. The shear failure surface of a geomaterial is a (Alsaleh, 2004), density of the soil mass, existence of
non–linear thin surface that experiences tremendous shear water and the level of the effective stress (Alsaleh et al.,
and volumetric plastic strains (localized strains) (Alsaleh, 2004; Terzaghi and Peck, 1948). Such types of
2004). Apparent cohesion has been defined by the geomaterials are known as φ -soils, described only by
geotechnical researchers as the tensile resistance that the internal angle of friction (macrolevel). The classical
essentially builds up between two adjacent particles due definition is the summation of: (i) the repose angle,
to the suction at the contact surface. Such a strength which depends on the grain microproperties, and (ii) the
component can simply collapse under very small strains. dilation component, which depends on the density and
Using the above-mentioned argument, it is clear that the the level of the effective stress.
apparent cohesion is not sustained once the soil mass is In slightly overconsolidated clayey to silty soils, the
sheared; therefore, the Mohr-Coulomb shear strength shear resistance mechanism is known to be a
theory needs to be revisited and discussed carefully. combination of cohesive and frictional components
Using an assembly of discrete particles (rigid rods), Rowe often represented by a Mohr – Coulomb (M–C) failure
(1962) has proven and showed that this theory has some envelope. Geotechnical engineers call this type of soil a
limitations, especially when it is applied to granular c-φ soil; the cohesion and the friction angle are used
materials due to the fact that it does not account for soil here to compute the shear strength of the soil. The
dilatancy. Traditionally, geotechnical engineers have authors see a danger in combining these two quantities
been classifying the soil into two types: cohesive and together due to the physical fact that the cohesive bonds
cohesionless soils; however, these terminologies could be (if they exist) would break down prior to the
misleading in many occasions. Instead, fine and coarse- mobilization of the frictional resistance components.
grained soils should replace these terms. This is an Considering highly overconsolidated clays, the
important classification that would depend on the level of geotechnical community has been using very high
the hydraulic conductivity and the specific surface area. cohesion quantities and very low or zero friction angles
In fine-grained soils, the structure is able to hold water according to the classical M–C failure criterion. It is
molecules for longer times, creating negative pore water believed here that this approach might be a serious error
pressure that produces suction at the contact area between and its applicability to such types of soil should be
two adjacent particles (Santamarina, 1997), resulting in questioned. The clay sheets are compacted closer to
apparent cohesion. In coarse sand, this phenomenon does higher densities due to the high maximum past pressure
not exist as fluid seeps out the voids quickly, leaving no and the micromechanical interactions are more efficient
chance for the negative pore water pressure to build up. to give higher macro shear strength. Therefore, the
As indicated by Alsaleh et al. (2006), there are authors are supporting the φ -soil concept to be applied
several microproperties that control the behavior and the to such types of soils.
strength of a discrete particle system (soils). Particle
size distribution, local void ratio, particle shape and MODELING PARAMETERS
surface roughness are good examples of such properties. As previously mentioned, the internal angle of
Micromechanical based material parameters and laws friction is a continuum–based parameter that accounts
are needed to better describe the micro and macro for the particle rotational and sliding resistance. This

- 184 -
Jordan Journal of Civil Engineering, Volume 4, No. 3, 2010

400

350

300

250
Shear Stress (kPa)

φ
200

150

100

50

c
0
0 50 100 150 200 250 300 350 400
Normal Stress (kPa)
Conf Press = 20.0 kPa
Conf Press = 40.0 kPa
Conf Press = 60.0 kPa
M-C Envelop

Figure 1: Classical M-C Failure Envelope Using Lab Triaxial Test on Silty Clay

parameter is meant to account for the particle–to– the critical state and the angle of friction in this case
particle frictional resistance; however, the difficulty is dependent only on the particle size, shape and
in measuring such microvalues enforces the surface roughness. Therefore, the critical state or the
continuum–based quantity. The critical state soil constant volume friction angle is a unique value for a
mechanics approach separates the peak friction angle certain soil type. On the other hand, soil particles that
into two components; constant volume or critical are confined under high stresses show lower peak
state angle of friction and dilatancy angle (Wood, friction angles than those which are confined under
1990). This leads to the conclusion that one should low stresses because dilation is reduced at high
consider the level of effective stress before providing stresses. A preliminary conclusion can be made here;
a value for the peak angle of friction. If the soil does the unique linear M–C failure envelope (Figure 1) for
not undergo any volumetric changes, then it is called

- 185 -
A Numerical Study… Mustafa Alsaleh and Keven Hofstetter

400

φnew1

φnew2
300
φnew3
Shear Stress (kPa)

200

100

0
0 50 100 150 200 250 300 350 400
Normal Stress (kPa)
Conf Press = 20.0 kPa
Conf Press = 40.0 kPa
Conf Press = 60.0 kPa
Failure Envelope 20kPa
Failure Envelope 40kPa
Failure Envelope 60kPa

Figure 2: Confining Stress-Dependent Failure Envelope and Friction Angles for Silty Clay Using
the Same Triaxial Tests

multiple levels of confining stress is incorrect and, the Mohr circle that is associated with a particular
instead, there exists a linear failure envelope for each confining pressure. This yields a higher peak angle of
level (Figure 2). As shown in Figure 1, where the friction that would compensate for the assumed zero
classical Mohr–Coulomb theory is used to obtain the cohesion resistance.
peak friction angle, it is obvious that this failure As clarified above, the apparent cohesion concept
envelope could show a non-zero cohesion value. If we was used first to fit the experimental results into a linear
do not accept the concept of cohesion in soil, then the equation (Mohr–Coulomb shear strength theory, see
failure envelope is forced through a zero intercept on Figure 1). The following equation describes the failure
the shear stress axis and at the same time is tangent to envelope:

- 186 -
Jordan Journal of Civil Engineering, Volume 4, No. 3, 2010

τ = C + σ tan (φ ) (1) C is the apparent cohesion,

where, σ is the normal stress and
τ is the shear stress, φ is the internal angle of friction
500

400
Deviatoric Stress (kPa)

300

200

100

0
0 5 10 15 20 25
Axial Strain(%)
20 kPa
40 kPa
60 kPa

15
(a)

10

5
Volumetric Strain (%)

0 5 10 15 20 25

10

15

Axial Strain (%)

20kPa
40 kPa
60 kPa

(b)
Figure 3: Effect of the Confining Stress on (a) Deviatoric Stress and
(b) Volumetric Strain Using DEM Simulations

- 187 -
A Numerical Study… Mustafa Alsaleh and Keven Hofstetter

700

600 90 kPa

500
Deviatoric Stress (kPa)

400

60 kPa

300

30 kPa
200

Virtual Triaxial Test

100

Laboratory Triaxial Test

0
0 2 4 6 8 10 12 14
A x i al St ra in( % )

Figure 4: Comparison between Predicted and Measured Deviatroic Stresses for Crushed Limestone

The apparent cohesion is a continuum-based Figure 2.

parameter that is trying, to some extent, to account for Considering the example of a soil pile, it may be
the small tensile forces that might exist in a partial- described by the shear strength resistance within the soil
saturated fine- grained soil mass. The resistance that this mass and by other parameters (boundary conditions,
parameter is providing within the numerical model is moisture content,… etc.). The pile can stand stable or in
acceptable and has a physical meaning only if the soil equilibrium without any external support at a certain
mass is subjected to very small deformations, when the slope; the angle of this slope is usually called the repose
frictional resistance is not yet mobilized and the suction angle. This angle can be measured using different
bond between two adjacent particles is not yet broken. techniques (Santamarina and Cho, 2001), dependent on
However, in most engineering applications, the soil the microfabric of the material (Alsaleh et al., 2006).
masses undergo large deformations that pass this limit. The angle increases with the decreasing particle size,
At that point, the soil can use only the frictional decreasing the coefficient of uniformity and increasing
component to resist any further shearing. This fact the grain angularity and surface roughness.
indicates that the cohesive resistance term can be Given our physical understanding for the discrete
dropped from Eq. 1 and additional frictional resistance materials behavior of soil, we can say that deceasing the
added so that the soil provides the same shear particle or the grain size will significantly increase the
resistance. Using this argument, Eq. 1 is rewritten as: frictional resistance due to the increase in the total
specific area. Therefore, the finer the soil fabric, the
τ = σ tan (φ new ) higher the actual internal microfrictional components,
leaving no need to consider the apparent cohesive
where, φ new is the internal angle of friction shown in resistance.

- 188 -
Jordan Journal of Civil Engineering, Volume 4, No. 3, 2010

400

350

300

250
Deviatoric Stress (kPa)

200

150

100

50

0
0 5 10 15 20 25
Axial Strain (%)
Init Void Ratio = 0.75
Init Void Ratio = 0.60
Init Void Ratio = 0.50

(a) Deviatoric Stress

10

5
Volumteric Strain (%)

0 5 10 15 20 25

10

Axial Strain (%)

Init Void Ratio = 0.75
Init Void Ratio = 0.60
Init Void Ratio = 0.50

(b) Volumetric Strain

Figure 5: The Effect of the Initial Density (Void Ratio) on the (a): Deviatoric Stress and
(b): Volumetric Strain Using DEM Simulations

- 189 -
A Numerical Study… Mustafa Alsaleh and Keven Hofstetter

le 1
R2
2 R2
1
R1 R1 gap
2

R pillar

(a) Tensile Stress Pillar between Two Particles

Pillar Tensile Stress

Ecoh Ecoh damaged

ε1 ε2
Pillar strain

(b) Pillar Tensile Strength

Figure 6: Apparent Cohesion Model Implemented within the Current DEM Code

- 190 -
Jordan Journal of Civil Engineering, Volume 4, No. 3, 2010

350

300

250
Deviatoric Stress (kPa)

200

150

100

50

0
0 5 10 15 20 25 30 35
Axial Strain (%)
Measured @ 20 kPa
Measured @ 40 kPa
Measured @ 60 kPa
Predicted @ 20 kPa
Predicted @ 40 kPa
Predicted @ 60 kPa

Figure 7: Comparison between Predicted and Measured Deviatroic Stresses for Silty Clay Using
Friction and Cohesion Parameters

MODELING OF GEOMATERIALS that is mostly used to model the particle flow and
There are several numerical tools to model the stress–strain transfer for granulates. Granular material in
behavior of geomaterials. Discrete Element (DEM), geotechnical engineering is considered a cohesionless
Finite Element (FEM) and Mesh Free Methods (MFM) material, which is an acceptable assumption. Modeling
are the most common tools to model the stress transfer a granulate assembly normally requires microfrictional
and the particle flow mechanisms in soils. Each of the parameters, particle-to-particle stiffness and damping
above-mentioned models has its own applications, parameters to solve for the system dynamics. Such
advantages and limitations. In any of the above- parameters are, relatively speaking, obtainable using
mentioned tools, there is a need for stress–strain some simple engineering tests. However, once we try to
relations; of course such relations require material model fine–grained cohesive soil, two issues seem to be
properties (shear strength parameters, stiffness limiting the application of the DEM. The first issue is
modulus,… etc.). the length scale; we are trying to model very small
Let us consider a Discrete Element Model (DEM) particles (in the order of angstroms) using a size of

- 191 -
A Numerical Study… Mustafa Alsaleh and Keven Hofstetter

350

300

250
Deviatoric Stress (kPa)

200

150

100

50

0
0 5 10 15 20 25 30 35
Axial Strain (%)
Measured @ 20kPa
Measured @ 40kPa
Measured @ 60 kPa
Predicted @20kPa
Predicted @ 40kPa
Predicted @ 60kPa

Figure 8: Comparison between Predicted and Measured Deviatroic Stresses for Silty Clay Using
Zero–cohesion theory

multiple inches. The second issue is the cohesive grained soil.

characteristics that the soil mass might have as a result
of the negative pore water pressure within the tiny RESULTS AND DISCUSSION
voids. Many researchers tried to resolve the length scale The Discrete Element code developed by Caterpillar,
issue by representing group of very small particles as Inc. (Hofstetter, 2002) was used in this study to simulate
large solid grains (spheres in most cases) and triaxial experiments and excavation applications. The
compensate for the high micro frictional components by triaxial simulations have been performed to show that the
some false cohesion parameters. This approach requires current DEM code is capable of capturing the real
building some cohesive links or bonds at the particle behavior for both fine-grained and coarse-grained
level to keep the particles bonded to each other until a materials with an acceptable level of accuracy. Figure 3
certain strain-threshold value. Once this strain-threshold shows the effect of the confining pressure on the
is reached, failure criterion is met and the particles will deviatoric stress and volumetric strain predictions using
travel apart from each other. This use of cohesion does DEM triaxial simulations. Dilation increases as the
not represent the true soil behavior; yet, researchers and confining stress decreases and, as a result, the predicted
developers tend to use such technique to model fine– peak friction angle increases. Figure 4 shows a

- 192 -
Jordan Journal of Civil Engineering, Volume 4, No. 3, 2010

(a) With Cohesion

(b) Without Cohesion

Figure 9: Soil Piles Built at Steep Angles Using Caterpillar, Inc. DEM Code

- 193 -
A Numerical Study… Mustafa Alsaleh and Keven Hofstetter

5
1.6 .10

5
1.4 .10

5
1.2 .10

1 .10
5
Force in the Horizantal Direction

8 .10
4

6 .10
4

4 .10
4

2 .10
4

4
2 .10

4
4 .10
0 1 2 3 4 5 6 7 8 9 10 11 12 13
ExcavationTime
With Cohesion
Non - Cohesion

(a)
4
4 .10

2 .10
4

4
2 .10
Force in the Vertical Direction

4 .10
4

4
6 .10

4
8 .10

1 .10
5

5
1.2 .10
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Excavation Time
With Cohesion
Non - Cohesion

(b)

Figure 10: Comparison between Forces Acting on the Bucket during Excavation Using
Both (a) Cohesive and (b) Non-Cohesive Parameters

- 194 -
Jordan Journal of Civil Engineering, Volume 4, No. 3, 2010

comparison between predicted and measured deviatoric shear strength components for fine–grained soil piles. The
stresses for medium-dense crushed limestone. The modeled soil particles are usually required to be equipped
material parameters were obtained from simple with cohesive bonds in order for the pile to be stable at a
laboratory tests and, thereafter, were mapped into the given steep angle. In this case, the cohesive bonds
DEM code. The classical internal angle of friction was algorithm needs to be enabled, which will introduce
back-calculated using the virtual triaxial test results and intensive computational overhead. Instead, using the non–
the values agree with laboratory experimental cohesion theory with the additional frictional resistance
measurements. Likewise, the DEM simulations showed (φ new, which includes the effect of dilation and other
that the density effect agrees with the fundamentals of microproperties) can provide the adequate shear strength
soil mechanics; dense packed particles seem to dilate and significantly reduce the computational cost. Figure 9a
while loose packed particles tend to contract (Figure 5). shows a fine–grained soil built at a steep angle using
Triaxial test results for fine–grained soil (silty clay) cohesive bonds; this pile could be rebuilt using the non–
were used to calibrate for the DEM apparent cohesion cohesion theory at the same steep angle (Figure 9b).
model parameters. The apparent cohesion model Simulating excavation, both models retain almost the same
implemented in this DEM code is essentially a non- vertical and horizontal forces (see Figure 10) with much
linear tensile pillar that is described in Figure 6. Using lower computational cost for the non–cohesion case.
the classical soil shear strength laws (Figure 1), the
apparent cohesion and internal angle of friction were CONCLUSIONS
obtained and used to predict the stress-strain behavior A numerical study for the shear strength components
for this material. Figure 7 shows comparisons between (friction and cohesion) was performed for fine and
model predictions and laboratory measurements. In course–grained soils using Caterpillar DEM code. The
these predictions, the internal angle of friction was classical soil shear strength laws (mainly M–C theory)
assumed to be constant following the M–C theory; this assume a constant internal angle of friction and ignore
assumption of course, does not let the model respond to the dilation effect. The authors revisited the definition
the effect of the confining pressure. In other words, the for the apparent cohesion and pointed out the limitations
dilation effect is not being captured; on the other hand, of the concept. The apparent cohesion, if existing in
if the non–cohesion theory (Figure 2) is adopted, the partially saturated fine–grained soil, fails at very low
internal angle of friction becomes highly dependent on the strain levels and we may not account on it. Numerical
level of the confining stress. This dependency agrees with results using a micromechanical–based DEM model
the real behavior of a discrete system. The triaxial test showed that the cohesion part can be ignored and,
results for the silty clay were analyzed using the non– instead, additional frictional components which are
cohesion theory and the measured internal angles of function of dilation and microproperties can be used to
friction were used to predict the stress-strain behavior for compensate for that part and the model will still retain
the three different confining stresses. The comparisons in acceptable results with much lower computational cost.
Figure 8 show that the non–cohesion theory can predict the The proposed alternative was applied to triaxial
constitutive behavior of the fine–grained soils. compression tests and excavation operations. The
In many engineering practices, the design engineer is findings of the study are supported by the fact that the
required to replicate or build a geotechnical structure using shear strength of soil is essentially caused by the
numerical tools: soil piles, slopes, earthfill dams,… etc. particles contacts, overlapping and interlocking
Caterpillar machines deal with various types of regardless of the particle size distribution. Then, the
geotechnical structures. The authors chose to use shear strength is function of density, surface roughness
excavation applications as an example on modeling the of the particles, angularity, spherecity and size.

- 195 -
A Numerical Study… Mustafa Alsaleh and Keven Hofstetter

ACKNOWLEDGEMENTS Peoria and Tucson Proving Grounds for providing the

The authors would like to acknowledge Caterpillar soils used in the study.

REFERENCES Dynamic Interaction of a Dozer Blade Earthen

Material. 14th International Conference of the
Alsaleh, M.I. 2004. Numerical Modeling for Strain International Society for Terrain-Vehicle Systems.
Localization in Granular Materials Using Cosserat Vicksburg, MS, USA, Oct., 20-24.
Theory Enhanced with Microfabric Properties. Ph.D. Rowe, P.W. 1962. The Stress-dilatancy Relation for Static
Thesis, Dept. of Civil and Environmental Engineering, Equilibrium of an Assembly of Particles in Contact.
Louisiana State University. Proc. R. Soc., London, Ser. A, 269:500-527.
Alsaleh, M.I., Alshibli, K.A. and Voyiadjis, G.Z. 2004. On Santamarina, C. 1997. Cohesive Soil: A Dangerous
the Bridging of the Length Scale and the Behavior of Oxymoron, A Note Located at URL http:
Granular Materials, Geo Jordan 2004: Advances in //geosystems.gatech.edu/Faculty/Santamarina/General/
Geotechnical Engineering with Emphasis on Dams, Publications/Electronics /Dange_Oxy/Dangeoxi.zip
Higway Materials and Soil Improvement, ASCE Santamarina, J.C. and Cho, G.C. 2001. Determination of
Geotechnical Practical Publication No. 1,191-199, Critical State Parameters in Sandy Soils - Simple
Irbid. Jordan. Procedure, ASTM Geotechnical Testing Journal, 24
Alsaleh, M. I., Alshibli, K. A. and Voyiadjis, G. Z. 2006. (2): 185-192.
The Influence of Micro-Material Heterogeneity on Terzaghi, K. and Peck, R.B. 1948. Soil Mechanics in
Strain Localization in Granular Materials, ASCE Engineering Practice, Wiley and Sons.
International Journal of Geomechanics, 6 (4): 248-259. Wood, D.M. 1990. Soil Behavior and Critical State Soil
Hofstetter, K.W. 2002. Analytic Method to Predict the Mechanics. Cambridge University Press, New York.

- 196 -