40

Length scales evolution and localization phenomenon in sand

Mustafa I. Alsaleha,*, George Voyiadjisb, Khalid A. Alshiblib
a
Machine Research, Caterpillar Inc., Mossville, IL 61552, USA
b
Department of Civil & Environmental Engineering, Louisiana State University, Baton Rouge, LA 70803, USA

Abstract

A numerical model in the Cosserat continuum for strain localization phenomena in granular materials was developed
by the authors. The mathematical formulations used in the present numerical model are equipped with evolution
equations for the length scales through micropolar theory. The evolution equations of the internal length scales describe
any possible change in the contact surface between the particles, damage to the particles if it exists and/or any change in
the local void ratio within the domain. The solution for the shear bands thickness shows more accurate correlation with
the experimental results and less dependency on the mesh size when such evolution equations are used.

Keywords: Strain localization; Granular materials; Evolution of length scales; Micropolar theory

1. Introduction The present model assumes two different length scales

embedded in the Cosserat-based formulations. These
Granular materials exhibit a complex mechanical length scales are adjusted using the effect of the shape
behavior when subjected to high plastic deformation. indices and the surface roughness.
The microstructure of the particles and how they affect Evolution equations for the proposed length scales
the material behavior is essentially justified by the non- and their effects on the plastic strain localization are
uniformity in the shape and surface roughness at the introduced in this paper. Such equations are then
micro or even at the nano level. During the hardening implemented into the numerical model to study their
regime, granular materials behave as a continuum until effect on the plastic strain localization in granular
the failure or instability point, where deformations begin materials.
to localize into a small but finite shear zone, called the
shear band. Thicker shear bands indicate less shear
strength; the angle of inclination of such bands with
respect to the minimum principal axis gives an indica- 2. Proposed evolution equations for the length scales
tion of the stability of the soil mass. Since the shear band
properties are essentially dependent on some micro- The length scales are affected by the shape and the
length scale it is important to carefully incorporate this surface roughness of the particles; such effects are
parameter. accounted for through the following equations:
The Cosserat theory can successfully separate the ISPH
grain rotation from its translation adding three addi- ls ¼ lave ð1Þ
IR
tional degrees of freedom to any point in the 3-D
continuum. Granular materials undergo high rotational where ls, ISPH, IR and lave are the length of the surface of
and translational deformations at failure. The classical contact, sphericity index, roundness index and the mean
strain tensor fails to capture the real kinematics such as particle size (equivalent to d50), respectively.
micro-rotation in granular materials and other alter- This equation is proposed to account for the spheri-
native tensors need to be introduced instead [1,2]. city and roundness effect on the length of the contact
surface. All researchers assume that the length of the
contact surface is the mean particle size; however, it is
* Corresponding author. Tel.: + 1 309 578 3245; Fax: + 1 309 believed that this quantity will be affected by the shape
578 4277; E-mail: Alsaleh_Mustafa_I@cat.com of the particles. The length of the contact surface

# 2005 Elsevier Ltd. All rights reserved.

Computational Fluid and Solid Mechanics 2005
K.J. Bathe (Editor)
 M.I. Alsaleh et al. / Third MIT Conference on Computational Fluid and Solid Mechanics 41

increases as the sphericity index increases and decreases where k1 is a constant coefficient taken here as unity and
as the roundness index increases. Based on this argu- lo is the initial length scale, which is assumed to be as
ment Eq. (1) was proposed. follows:
The length of the arm of rotation is assumed to have ISPH IR
the following expression: IR d50 þ ISPH d50 þ R
lo ¼ ð8Þ
2
IR
la ¼ lave þ Ra ð2Þ It is important to mention here that Eq. (7) has no
ISPH
physical basis especially for granular materials since it
where la and Ra are the length of the arm of rotation and assumes an exponential decay for the initial length scale.
the mean surface roughness respectively. Similar to the The chosen length scales can decrease or increase based
length of the contact surface, the length of the arm of on the shape of particles and the final deformation in
rotation is assumed to be affected by the shape of the each loading step.
particles. Equation (2) shows that this quantity increases In this paper we propose a new evolution equation for
as the roundness index and surface roughness increase the length of contact and the arm of rotation as a
and decreases as the sphericity index increases. function of the shape indices and the Cosserat rotation.
The above shape indices are defined as follows: It is assumed here that such length scales are subjected
  to significant changes during plastic deformation
Dequ Dequ  attributed to the rotation of the particles. As the particle
ISPH ¼ 
 ð3Þ
ds dL  rotates and translates it causes the length scales to
Pact change because of the high non-uniformity in their
IR ¼ ds þdL  ð4Þ shapes. Based on this argument the following equations
2 are proposed and used in this paper:
in which Pact is the actual perimeter of the particle, Dequ ISPH
is the sphere-equivalent diameter, and ds and dL are the ls ¼ d50 sinð!c Þ ð9Þ
IR
shortest and longest dimensions, respectively [3].
IR
In Eqs. (1) and (2) it is assumed that for a given ele- la ¼ d50 cosð!c Þ þ Ra ð10Þ
ISPH
ment, which will involve few particles, all the parameters
are held constant throughout the simulation and only The above equations are proposed based on a thor-
the mean particle size lave will evolve based on the ough understanding of the behavior of granular
cumulative effective plastic strain and the point or materials. As mentioned earlier, the particles will exhibit
Cosserat rotation. It is worth noting here that the par- large translational and rotational deformations due to
ticles are treated as rigid bodies (no particle damage); the discreteness of such materials. The rotation of the
however, the evolution equations are used here to particles will cause the assigned length scales to change
account for the changes in the suggested length scales due to the high irregularity in the particle shape.
due to the translational and rotational deformations of
the nonuniform-shaped particles. Following Garcia [4]
the following evolution equation is proposed to update
the average length scale: 3. Results and discussion

f d50 The numerical model used in this paper is a Cosserat

lAve ¼ 1 ð5Þ
þ d50 pm continuum-based model, which enables us to incorpo-
rate the discussed length scales. The model was used to
where d50 is the mean particle size, m and f are material predict the behavior of the very dense F-75 Ottawa sand.
constants, is a constant taken as mean surface It was found that the numerical solution has some minor
roughness of particles, and p is the effective plastic strain mesh dependency if one chooses to use a constant length
as follows: scale, while it is not mesh-dependent if an evolution
equation is used for the internal length scale. Figure 1
Z t qffiffiffiffiffiffiffiffiffiffi shows the difference between the predicted shear bands
p¼ _ij00 _ij00 ð6Þ using a constant and evolving length scale (Eq. 10) for
0 two different mesh sizes. In Fig. 1(a) the predicted shear
band using a constant length scale was found to have
A simpler evolution equation was used in this paper as some mesh dependency, while in Fig. 1(b) with a vari-
able length scale it was found that mesh dependency
lave ¼ lo e
k1 p ð7Þ decreases and consistent shear band thickness was
42 M.I. Alsaleh et al. / Third MIT Conference on Computational Fluid and Solid Mechanics

Fig. 2. Comparison between measured and predicted shear

band thickness and inclination angle for the very dense F-75
Ottawa sand under confining pressure of 100.0 kPa. (a) Model
prediction. (b) Experimental results [5].

Alshibli [5] under plane strain conditions for the very

dense F-75 Ottawa sand. The numerical results compare
well with the experimental observations.
Fig. 1. Effect of the length scale and mesh size on shear band:
(a) constant internal length scale; (b) variable internal length
scale. 4. Conclusions

The present numerical model was developed in order

obtained using a different mesh size. Three different to accommodate any evolution expression for the
evolution Eqs. (6) through (10) were used and all gave internal length scales. Three different evolution equa-
similar solutions for the shear band thickness and tions were proposed and used in this paper; using these
inclination. However, comparing the predictions with evolution equations was found to improve the numerical
experimental results in Alshibli [5], it was found that Eq. analysis for the strain localization in granular materials.
(10) gave closer predictions that compare better with the The degree of mesh dependency was found to appreci-
experiments for both the thickness and the inclination ably decrease once such evolution equations are used
(see Tables 1 and 2). Figure 2 shows comparisons and the shear band thickness evolves similarly to the
between model predictions and experimental results in internal length scale during the deformation process.

Table 1
Comparison between measured and predicted shear band thickness

Material Confining pressure Initial void Mean grain size Measured shear Predicted shear Predicted shear Predicted shear
(KPA) ratio (mm) band thickness band thickness band thickness band thickness
(mm) (mm)a (mm)b (mm)c
*F-sand 15.00 0.629 0.22 2.97 3.07 3.00 2.96
*F-sand 100.0 0.629 0.22 2.91 2.80 2.86 2.88
*F-sand 15.00 0.495 0.22 3.00 3.10 3.12 3.03
*F-sand 100.0 0.495 0.22 3.05 2.81 2.92 2.93
+
C-sand 15.00 0.767 1.60 17.33 17.96 17.90 17.52
+
C-Sand 100.0 0.767 1.60 17.00 16.61 16.74 16.92
* F-75 Ottawa sand, + Coarse silica sand
a
Using Eq. (5)
b
Using Eq. (7)
c
Using Eqs. (9), (10)
 M.I. Alsaleh et al. / Third MIT Conference on Computational Fluid and Solid Mechanics 43

Table 2
Comparison between measured and predicted shear band inclination angle

Material Confining pressure Initial void Mean grain Measured shear Predicted shear Predicted shear Predicted shear
(KPa) ratio size (mm) band inclination band inclination band inclination band inclination
(O) (O)a (O)b (O)c
*F-sand 15.00 0.629 0.22 51.6 52.00 52.10 51.20
*F-sand 100.0 0.629 0.22 53.7 54.30 54.00 53.80
*F-sand 15.00 0.495 0.22 58.0 57.20 57.60 58.10
*F-sand 100.0 0.495 0.22 59.2 58.10 59.30 58.90
+
C-sand 15.00 0.767 1.60 51.4 50.60 50.90 51.50
+
C-sand 100.0 0.767 1.60 53.2 48.90 49.60 51.70
* F-75 Ottawa sand, + Coarse silica sand
a
Using Eq. (5)
b
Using Eq. (7)
c
Using Eqs. (9), (10)

Acknowledgment [2] Oda M, Iwashita I. Mechanics of Granular Materials: An

Introduction. A.A. Balkema, 1999.
The authors gratefully acknowledge the financial [3] Alshibli KA, Alsaleh MI. Characterizing surface rough-
support provided by NASA/George C. Marshall Space ness and shape of sands using digital microscopy. ASCE J
Comput Civil Eng 2004;18(1):36–45.
Flight Center under contract NAS8-01042.
[4] Garcia JJ. The double effect of grain size on the work-
hardening behavior of polycrystalline copper. Scripta
Metall Mater 1994;31:487–489.
References [5] Alshibli KA. Localized Deformations in Granular Mate-
rials. PhD Dissertation, University of Colorado at
[1] Vardoulakis I, Sulem J. Bifurcation Analysis in Geo- Boulder, CO, 1995.
mechanics. London: Blackie Academic and Professionals,
1995.