1 s2.0 S0038080618302099 Main

HOSTED BY Available online at www.sciencedirect.
com
ScienceDirect
Soils and Foundations 59 (2019) 208–227
www.elsevier.com/locate/sandf
On the influence of grain shape on the cumulative deformations in

sand under drained high-cyclic loading
T. Wichtmann a,⇑, Th. Triantafyllidis b, L. Späth c
a
Chair for Geotechnical Engineering, Bauhaus University Weimar, Germany
b
Institute of Soil Mechanics and Rock Mechanics (IBF), Karlsruhe Institute of Technology (KIT), Germany
c
Ingenieurgruppe Geotechnik, Kirchzarten, Germany
Received 18 February 2017; received in revised form 17 September 2018; accepted 1 November 2018
Available online 14 February 2019
Abstract
The cumulative response of three granular materials with significantly different grain shape and surface characteristics (glass beads,
natural sand with subrounded grains and crushed sand with very angular particles) but identical grain size distribution curve has been
studied in drained cyclic triaxial tests. For each material, several tests with 100,000 cycles and different amplitudes, densities, average
mean pressures and average stress ratios have been performed. In case of glass beads and natural sand, an approximately square rela-
2
tionship between the residual strain accumulation rates and stress or strain amplitude was found (_eacc ðeampl Þ ), while an almost pro-
portional dependence was measured for the crushed sand (_e e ). The largest differences in the cumulative response of the three
acc ampl
tested materials were observed regarding the pressure-dependence of e_ acc . For glass beads and (less pronounced) for natural sand, the
residual strain accumulation rates decreased with average mean pressure, while the opposite tendency was obtained for the crushed sand.
At small pressures, the residual strains were much larger for the glass beads than for the natural sand and particularly the crushed sand,
while these differences in the accumulated strains almost diminished at larger pressures. Independent of the shape and the surface char-
acteristics of the particles, it was confirmed that the average stress ratio is the governing parameter of the cyclic flow rule. Finally, the
parameters of the high-cycle accumulation (HCA) model proposed by Niemunis et al. (2005) were analyzed considering the grain shape
parameters (aspect ratio, circularity) obtained from an automated grain shape analysis.
Ó 2019 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.
This is an open access article under CC BY-NC-ND license. (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Keywords: High-cyclic loading; Grain shape; Glass beads; Natural sand; Crushed sand; Drained cyclic triaxial tests
1. Introduction struction processes (e.g. vibration of sheet piles) or

mechanical compaction (e.g. vibratory compaction). Per-
A high-cyclic loading, i.e. a loading with a large number manent deformations caused by such high-cyclic loading
of cycles (N > 103) and relative small strain amplitudes may endanger the serviceability of foundations and thus
(eampl < 103) may be caused by traffic (high-speed trains, must be accurately predicted in the design stage.
magnetic levitation trains), industrial sources (crane rails, The high-cycle accumulation (HCA) model of Niemunis
machine foundations), wind and waves (on-shore and off- et al. (2005) may be used for that purpose. It is based on
shore wind power plants, coastal structures), repeated fill- numerous drained cyclic tests performed on quartz sands
ing and emptying processes (locks, tanks and silos), con- with subrounded particles (Wichtmann, 2005; Wichtmann
et al. (2005, 2006, 2007a, 2007b)). The equations of the
Peer review under responsibility of The Japanese Geotechnical Society.
HCA model are summarized in the Appendix A. Applica-
⇑ Corresponding author. tions of the HCA model, amongst others to offshore wind
E-mail address: [email protected] (T. Wichtmann).
https://doi.org/10.1016/j.sandf.2018.11.001
0038-0806/Ó 2019 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.
This is an open access article under CC BY-NC-ND license. (http://creativecommons.org/licenses/by-nc-nd/4.0/)
T. Wichtmann et al. / Soils and Foundations 59 (2019) 208–227 209
power plant foundations, have been well-documented are affected by both the general shape and the surface pro-
(Galindo et al., 2014; Osinov, 2013; Wichtmann, 2016; file of the particles (schemes on the left-hand side and in the
Wichtmann et al., 2010c; Zachert, 2015; Zachert et al., middle of Fig. 1). A quantitative description of the surface
2014, 2015, 2016). For a simplified calibration of the roughness (right-hand side of Fig. 1) is a matter of ongoing
parameters of the HCA model, correlations have been pro- research at the IBF and not a topic of the current paper.
posed considering the grain size distribution curve (mean However, the shape of the sand particles and its surface
grain size d50, uniformity coefficient Cu) and index proper- roughness are often related to each other; i.e. angular par-
ties (minimum void ratio emin) from simple laboratory tests ticles tend to have a rough surface while rounded grains are
(Wichtmann et al., 2009, 2014, 2015). However, an estima- smooth (Jänke, 1968).
tion of all parameters from the correlations is only recom-
mended for making rough estimations because the 2. Tested materials
equations are valid only for the subrounded quartz sand
used in the experimental study; i.e. the influences of grain The particles of all three materials are composed of
shape or surface characteristics were not taken into consid- quartz. The natural sand is a fluvially deposited quartz
eration. A combined procedure was proposed as the mini- sand obtained from a sand pit near Dorsten, Germany.
mum standard for the HCA model calibration, where some The crushed quartz sand originates from the crushing of
of the parameters are estimated from the correlations and Rhine gravel. It was obtained from a gravel pit near Karl-
some others are determined from a single cyclic triaxial test sruhe, Germany. The glass beads used in this study are
(Wichtmann et al., 2015). commercially distributed by the company Worf Glaskugeln
The experimental study documented in this paper repre- in Mainz, Germany.
sents a first step to extend the simplified procedure for the Pictures of the grains of the three different materials
HCA model calibration considering the influence of grain taken with an electron scanning microscope are provided
shape. Since almost no test data on this point can be found in Fig. 2. In this test series, the characteristics of the parti-
in the literature, the present study was dedicated to a fun- cles on all three scales shown in Fig. 1 have been varied
damental examination of the influence of grain shape on simultaneously. From the crushed sand over the natural
the cumulative response under drained high-cyclic loading. sand to the glass beads, the shape becomes more compact
Three materials with significantly different grain shape and and more circular, the surface becomes less profiled and
surface roughness were chosen for this investigation: glass the surface roughness decreases.
beads, a natural sand with subrounded grains and a First, each of the three different raw materials glass
crushed sand with very angular particles. The strong influ- beads, natural sand and crushed sand was decomposed into
ence of the grain size distribution curve (Wichtmann et al., several grain sizes by sieving. Next, mixtures with the grain
2009, 2014, 2015) was eliminated by producing special mix- size distribution curve shown in Fig. 3 (d50 = 0.6 mm,
tures of all three test materials with the same grain size dis- Cu = 1.5) were produced for each material. The minimum
tribution curve. Beside the drained cyclic tests, some and maximum void ratios derived from standard tests
reference tests with monotonic loading were also according to German standard code DIN 18126 (loose
performed. placement with a funnel for emax, layerwise compaction
The characteristics of the particles, i.e. shape, surface under water for emin) as well as the grain densities qs are
profile and roughness, can be examined at different scales, summarized in Table 1. Larger void ratios emax and emin
see Fig. 1. Because the transition is smooth, it is difficult to for granular materials with more angular grains as evident
define clear boundaries between these scales. The parame- from Table 1 were reported by a number of researchers
ters applied for a quantitative description of the particle (Cho et al., 2006; Koerner, 1970; Rousé, 2008; Shahu
shape in this study (an explanation is given in Section 2) and Yudhbir, 1998; Shin and Santamarina, 2013;
Sukumaran and Ashmawy, 2001; Vaid et al., 1985). They
are a result of the asperities of the more angular particles
General form Profile of Surface which prevent a denser packing. The increase in the range
(elongated or surface roughness emax emin with increasing angularity is in accordance with
compact?) (Cho et al., 2006; Miura et al., 1997; Rousé et al., 2008).
In order to derive quantitative measures for the grain
Detail
shape, an analysis using the software ImageJ in combina-
tion with the Plugin Particles 8 was performed in analogy
to the procedure described by Cox and Budhu (2008)
(applied also by Ham et al. (2012) and Liu and Lehane
(2012)). Such automated analysis gives objective values,
e.g. Aspect ratio, e.g. „true whereas a manual grain shape analysis using graphical
Circularity roundness“
charts does not (e.g. Cho et al., 2006; Krumbein, 1941;
Fig. 1. Different scales of particle shape or surface roughness. (adapted Krumbein and Sloss, 1963; Powers, 1953; Rittenhouse,
from Santamarina (2012)) 1943).
210 T. Wichtmann et al. / Soils and Foundations 59 (2019) 208–227
a) b) c)
Fig. 2. Pictures of grains of (a) glass beads, (b) natural sand and (c) crushed sand taken with an electron scanning microscope (Flemming, 2015).
Silt Sand Gravel AGrain

coarse fine medium coarse fine medium B
100 PGrain
ACircle
Finer by weight [%]
80
60
PCircle = PGrain
F
40 d50 = 0.6 mm Fig. 4. Definition of grain shape parameters Aspect Ratio = F/B and
Cu = 1.5 Circularity = AGrain/ACircle.
20
0 largest dimension in the orthogonal direction (see schemes

0.02 0.06 0.2 0.6 2 6 20 in Fig. 4). Circularity is obtained as the ratio of the cross-
Grain size [mm] sectional area of the grain AGrain divided by the area ACircle
Fig. 3. Tested grain size distribution curve (identical for all three test of a circle having the same perimeter as the original grain
materials). (Fig. 4). Circularity is thus a measure of how much the
shape of a grain resembles a circle. It should be kept in
mind, however, that this analysis is restricted to two-
For each fraction involved in the grain size distribution dimensional images of the grains, and that their third
curve shown in Fig. 3 (e.g. 0.1–0.125 mm, 0.125–0.16 mm, dimension was not taken into consideration.
etc.), images of several hundreds of grains were taken with For each fraction of a certain material, several hundreds
a flat bed scanner (resolution 9600 dpi). These images were of grains were analyzed in this way. Mean values of the
converted to a black-and-white format before analysis. shape parameters of all grains within a fraction were con-
Only particles with a size being in accordance with the sidered further. These mean values were weighted by the
grain fraction under consideration (no dust particles or mass fraction of that grain size in order to obtain a single
suchlike should be analyzed) and lying separately (i.e. with- value of the shape parameter for each material. These mean
out contacting neighbouring grains, in order to prevent the values of Aspect Ratio and Circularity are summarized in
analysis of conglomerates of several grains) were identified Table 1. Evidently, Circularity increases from the crushed
and analyzed by the software. The cross-sectional area, sand over the natural sand to the glass beads, while Aspect
perimeter, and geometric centre of gravity, etc. of the indi- Ratio simultaneously decreases.
vidual grains measured in pixels were determined. Based on
this information, several geometric parameters describing
the shape of the particles were calculated. Two shape 3. Monotonic tests
parameters are used further in this paper: the Aspect Ratio
and Circularity. Aspect Ratio is defined as the ratio of the Several drained monotonic triaxial tests were performed
length F of the longest axis divided by the length B of the on samples with different initial densities ID0 = (emax e0)/
Table 1
Mean grain size d50, uniformity coefficient Cu = d60/d10, grain density qs, minimum and maximum void ratios emin, emax and grain shape parameters
Aspect Ratio and Circularity for the three tested materials.
Material d50 [mm] Cu [–] qs [g/cm3] emin [–] emax [–] Aspect ratio Circularity
Glass beads 0.60 1.5 2.52 0.530 0.700 1.07 0.89
Natural sand 0.60 1.5 2.65 0.571 0.891 1.23 0.79
Crushed sand 0.60 1.5 2.66 0.763 1.149 1.31 0.73
(emax emin), with e0 being the void ratio prior to shearing, of the glass beads, the contractive phase at the beginning of
and an isotropic initial stress with an effective mean pres- shearing was extremely small.
sure p0 = 100 kPa (p = (r01 + 2 r03 )/3). A standard triaxial The increase of the shear strength with increasing parti-
cell for monotonic testing (not shown herein) was used. cle angularity is also confirmed by the diagram in Fig. 5c,
The samples measured 10 cm in diameter and 10 cm in where the peak friction angle uP is given as a function of
height (see a discussion of the influence of sample geometry the initial relative density. This agrees well with the litera-
in Section 6), and were prepared by dry air pluviation using ture (Andersen and Schjetne, 2013; Cavarretta et al., 2010;
a funnel. Different densities were achieved by changing the Cho et al., 2006; Gori and Mari, 2001; Guo and Su, 2007;
outlet diameter of the funnel. The fall height between the Jänke, 1968; Koerner, 1970; Liu and Lehane, 2012; Miura
outlet and the actual sample surface was kept constant et al. 1997; Miura et al., 1998; Oda, 1981; Rousé et al.,
(at about 2 cm) during pluviation. After preparation, the 2008; Sadrekarimi and Olson, 2011; Sezer et al., 2011;
samples were saturated with de-aired water. A back pres- Shahu and Yudhbir, 1998; Shin and Santamarina, 2013;
sure of 500 kPa was applied in all cases. The quality of sat- Sukumaran and Ashmawy, 2001; Tatsuoka, 2001; Xu
uration was checked by Skempton’s B-value. B-values and Sun, 2005; Zhuang et al., 2014). A higher angularity
larger than 0.99 were achieved in all cases. and surface roughness of the grains leads to higher inter-
Fig. 5a and b show the deviatoric stress q = r1 r3 or particle friction and interlocking between the particles, pro-
volumetric strain ev = e1 + 2 e3 versus axial strain e1 for viding restraint to particle sliding and rotation during
loose and dense samples. The fluttering q(e1) curves of deformation. The increased interparticle friction and the
the glass beads are in good agreement with test data restricted particle mobility allows the application of higher
reported earlier (Tatsuoka, 2001; Zhuang et al, 2014). They shear stresses to assemblies of more angular particles (Cho
are a result of the stick-slip response of the packing of et al., 2006; Guo and Su, 2007; Mair et al., 2002; Miura
round particles that favour rolling, thus forming an unsta- et al., 1997; Sukumaran and Ashmawy, 2001).
ble microstructure (Wei and Yang, 2014). Beside the larger As a measure of the initial stiffness, Young’s modulus
initial stiffness of the glass beads, it is clear from Fig. 5a E50 is presented as a function of ID0 in Fig. 5d. E50 repre-
that the shear strength increased with increasing angularity sents the secant stiffness between q = 0 and q = qmax/2.
of the granular material. The volumetric response for both The data in Fig. 5d corroborates the reduction in the initial
the initial contraction and the subsequent dilatancy was stiffness with increasing particle angularity. Similar
larger for the more angular materials (Fig. 5b). In the case tendencies have been reported earlier (Cho et al., 2006;
Legend
Material / ID0 loose / ID0 dense =
for
Glass beads / 0.30 / 0.91
a)+b):
Natural sand / 0.22 / 0.87
Crushed sand / 0.34 / 0.96
a) 600 c) e) 30
50
Peak friction angle ϕP [°]
Deviatoric stress q [kPa]
dense
Dilatancy angle ψ [°]
500 45 25
loose
400 40 20
300 35 15
200 30 10
100 25 5
0 20 0
0 5 10 15 20 25 30 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Axial strain ε1 [%] Initial relative density ID0 Initial relative density ID0
b) 2 loose d) 100 f) 2.0
Volumetric strain εv [%]
0 Glass beads Legend

Stiffness E50 [MPa]
80 Natural sand for c)-f) 1.6

Stress ratio ηc-d
-2
Crushed sand
-4 60 1.2
-6
-8 40 0.8
-10
dense 20 0.4
-12
-14 0 0
0 5 10 15 20 25 30 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Axial strain ε1 [%] Initial relative density ID0 Initial relative density ID0
Fig. 5. (a) Stress-strain curves q(e1), (b) volumetric strain ev(e1), (c) density-dependent peak friction angle uP, (d) Young’s modulus E50, (e) dilatancy angle
w and (f) stress ratio gc-d at the onset of dilatancy obtained from drained monotonic tests on the three test materials with different grain shape.
Miura et al., 1998; Shin and Santamarina, 2013), where the cell pressure 3
results were partially based on oedometric tests. The lower water in the cell
stiffness for materials with more angular particles can be plexiglas cylinder
explained by the fact that those materials have a higher
void ratio at the same relative density due to the larger emax sand sample
(h = d = 10 cm)
and emin values, which were addressed above. The looser
packing and the accompanying lower coordination number load cell
results in a softer matrix and thus higher compressibility V metal bellow
(Cho et al., 2006; Miura et al., 1998). pore pressure transd.
While shearing more angular granular materials, larger sealing
values of shear stress and strain are necessary to break
ball bearing
interlocking and to allow dilatancy. This is reflected by
the later onset of dilatancy in the tests on the crushed sand displacement transd.
(Fig. 5b). The lower dilatancy angles of the crushed sand load piston
(Fig. 5e) correspond to the smaller maximum inclination
pneumatic cylinder
of the ev(e1) curves (Fig. 5b), cf. (Liu and Lehane, 2012;
Tsomokos and Georgiannou, 2010; Zhuang et al., 2014), support frame
although the overall dilatancy (final value of volumetric
strain ev, see Fig. 5b) becomes more pronounced with
increasing angularity. The higher shear strength of the
materials with more angular particles is coincident with a Fig. 6. Scheme of the cyclic triaxial device used for this study.
higher stress ratio gc-d at the onset of dilatancy (Fig. 5f,
with g = q/p). In contrast to Guo and Su (2007), it was
found that gc-d was almost unaffected by density for all In the cyclic triaxial device shown in Fig. 6 the axial
three tested grain shapes. force is measured at a load cell being located directly below
As further addressed in Section 5, the particle breakage the bottom end plate of the sample, i.e. inside the pressure
effects can be regarded as negligible in the present test cell. The axial deformation is obtained from a displacement
series. transducer attached to the load piston. The system compli-
ance was determined in preliminary tests on a steel dummy
4. Cyclic tests and subtracted from the measured values. Volume changes
are determined via the squeezed out pore water using the
4.1. Test device, testing procedure and analysis of results burette system and a differential pressure transducer (not
shown in Fig. 6). Two pressure transducers are used for
A scheme of the cyclic triaxial device used for this study monitoring cell pressure and back pressure.
is given in Fig. 6. This device was designed and manufac- In the tests of the present series the lateral effective stress
tured with the aim of enabling tests with very large num- has been kept constant (r03 = constant), while the vertical
0
bers of cycles. The sample is encompassed in a pressure effective stress was varied with an amplitude r1ampl around
0
cell capable of dealing with pressures up to 1000 kPa. A an average value r1av (Fig. 8a). The effective stress path is
pneumatic loading system was used to apply the axial load- shown schematically in a p-q diagram in Fig. 9. After sam-
ing. The pneumatic cylinder is located at the bottom of the ple preparation by air pluviation, water saturation and a
test device, so that the axial loading is applied in the successful B value test, the sample was consolidated for
upwards direction. This arrangement enables sample one hour at the average effective stress of the test. This
0
preparation inside the test device, i.e. the sample can be average stress is described by the vertical component r1av
prepared directly on the load piston (see the photos in 0
and the horizontal component r3av or by the average values
Fig. 7). The top end plate of the sample is rigidly connected 0 0
of effective mean pressure p = (r1av + 2 r3av )/3 and devia-
av
with the upper plate of the loading frame, which is decou- 0 0
pled from the lid of the pressure cell. The bottom and the toric stress qav = r1av r3av , with the average stress ratio
top end plate are both equipped with a central porous defined as gav = qav/pav. Afterwards the cyclic loading in
0
stone 15 mm in diameter. The drainage lines from these the axial direction with the amplitude r1ampl was superposed
porous stones are connected to a burette used for the mea- to this average stress. Since the lateral stress was constant
0 0
surement of volume changes. As in the monotonic tests, during the cycles (r3ampl = 0, r03 = r3av = constant) and
samples with a diameter of 100 mm and a height of the test was performed with open drainage, the resulting
100 mm were used in the present test series. The influence effective stress path during the cycles is inclined by 1:3 in
of the sample geometry in the cyclic tests is further dis- the p-q diagram (Fig. 9). The amplitudes of deviatoric
0
cussed in Section 6. In order to eliminate end friction and vertical stress are identical, i.e. qampl = r1ampl , while
effects, smeared end plates composed of a layer of grease 0
that of mean pressure is pampl = r1ampl /3.
and a rubber disk were used.
Fig. 7. Photos of sample preparation by air pluviation.
Fig. 8. (a) Axial effective stress r01 and (b) axial strain e1 versus time in a drained cyclic triaxial test.
0 0
Due to larger deformations the first irregular cycle resulting in a value r3av /r1av = 0.5 being typical for many
(Fig. 8) was applied with a low loading frequency of in situ conditions). The stress amplitude in the reference
0
0.01 Hz while a frequency of 0.2 Hz was chosen for the test has been selected as qampl = r1ampl = 60 kPa, which
subsequent 105 regular cycles. The data were continuously leads to strain amplitudes e ampl
4104 being typical for
recorded during the whole test. From this huge amount of a high-cyclic loading in real problems. Compared to the
data, the data recorded during the first 24 cycles and during reference test, lower or higher deviatoric stress amplitudes
five cycles at N = 50, 100, 200, 500, . . ., 5 104 and 105 were qampl between 20 and 80 kPa were applied in the first series
extracted for a further analysis. of tests (Fig. 10a). All other boundary conditions (density,
For each of the three materials four series of drained average stress) were chosen as in the reference test. In the
cyclic triaxial tests were performed. The stress paths are second series, the initial relative density ID0 was varied
shown schematically in Fig. 10. As reference a test with between medium dense and dense, keeping the average
medium relative density, an average mean pressure and cyclic stresses identical to the reference test
pav = 200 kPa and an average stress ratio gav = 0.75 has (Fig. 10b). For each material three or four different densi-
0 0
been chosen (i.e. r1av = 300 kPa, r3av = 150 kPa, thus ties were tested. Four or five different average mean pres-
sures in the range 50 kPa pav 300 kPa were examined
in the third series (Fig. 10c), while the density, the average
q stress ratio gav = 0.75 and the amplitude-pressure ratio
L
CS
f = qampl/pav = 0.3 were the same as in the reference test.

Mc(ϕp)
average Finally, in the fourth test series (Fig. 10d) the average stress
stress 1
Mcc(ϕcc) ratio gav was varied between 0.0 and 1.25, keeping density,
σav ηav = qav / pav pav and qampl as in the reference test. Four or five tests with
1
1 different gav values were conducted for each material. Each
q ampl = σ'1ampl of these tests was performed on a fresh sample, i.e. no mul-
qav tistage testing was done.
Note, that the initial relative density ID0 = (emax e0)/
(emax emin) given herein is calculated with the void ratio
p ampl = σ'1ampl/ 3 e0 measured at the average stress (pav, qav) prior to the start
of the regular cycles. A value ID0 = 1.18 was obtained in
pav p
one of the tests on the crushed sand. In principle, relative
Fig. 9. Stress path in a cyclic triaxial test shown in the p–q diagram. densities larger than 1.0 are possible due to two different
Fig. 10. Stress paths in the four series of cyclic tests performed on each material shown in the p–q diagram.
reasons: First, it has to be considered that emin and emax are of accumulation m ¼ e_ acc =k_eacc k (unit tensor, see Appendix
determined from standardized testing procedures. By cer- A). In the triaxial case, the intensity of accumulation is cal-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
acc 2 2ffi
tain sample preparation techniques like air pluviation using
culated as e_ acc ¼ e_ 1 þ 2 e_ acc 3 with the rates of axial
a funnel with a very small outlet diameter one can reach
void ratios slightly below emin. Second, a void ratio e0 - and radial strain accumulation. The direction of accumula-
< emin can be reached during the increase of pressure to tion can be expressed by the ratio of the volumetric and
pav, in particular in case of a high compressibility of the deviatoric strain accumulation rates
pffiffiffiffiffiffiffiffi
grain skeleton as observed for the angular crushed sand. x ¼ 3=2trðmÞ=km k ¼ e_ v =_eq acc acc
with e_ v ¼ e_ 1 þ 2_eacc
acc acc
acc 3
Additionally, also cyclic loading may cause a compaction and e_ q ¼ 2=3 e_ 1 e_ 3 . In the following, the test results
acc acc
resulting in e < emin after a large number of cycles are first analysed regarding the direction of accumulation
(Wichtmann and Triantafyllidis (2005)). and then with respect to its intensity.
Fig. 11 presents data recorded during the first 24 regular
cycles of three tests on the natural sand with the reference 4.2. Direction of accumulation - high-cyclic flow rule m
stress conditions (pav = 200 kPa, gav = 0.75, qampl = 60
kPa) and different initial relative densities ID0. The dia- Fig. 12 presents the accumulated deviatoric strain eacc q as
grams show the relationships between deviatoric stress q
a function of the accumulated volumetric strain ev mea-acc
and axial strain e1 or volumetric strain ev, respectively,
sured in selected tests. Each data point in those diagrams
and the curves of void ratio change e e0 versus effective
refers to a certain number of cycles (N = 1, 2, 5, 10, . . .,
mean pressure p. The accumulation of axial and volumetric
105). For all three materials, the direction of the eacc acc
q –ev
strain and the associated reduction of void ratio are obvi-
ous in these representations. The lower the density, the lar- strain paths was found to be rather independent of stress
ger is the accumulated strain or void ratio change per cycle. amplitude, density and average mean pressure (see the dia-
Since the HCA model predicts the strain accumulation grams for crushed sand in Fig. 12a–c). For all three mate-
rates due to the regular cycles only, the first irregular cycle rials, the average stress ratio gav is the governing parameter
is not discussed in this paper. In the following, N = 1 thus for the direction of accumulation (Fig. 12d–f). The larger
refers to the end of the first regular cycle (Fig. 8). In the gav the larger is the deviatoric component of the strain
HCA model of Niemunis et al. (2005) the tensor of the rate accumulation rate, i.e. the lower is the ratio e_ acc
v =_
eacc
q . The
of strain accumulation is described by e_ acc ¼ e_ acc m, i.e. as dependence of the strain rate ratio on g can be well
av
the product of the scalar intensity of accumulation described by the following equation adopted from the flow
e_ acc ¼ k_eacc k (norm of strain rate tensor) and the direction rule of the Modified Cam clay model:
a) ID0 = b) c)
80 0.49 0.60 0.81 80
Void ratio change e - e0 [-]
0
40 40
q - qav [kPa]
q - qav [kPa]
0 0 -0,001
-40 -40
-0,002
-80 -80
0 0.1 0.2 0.3 0 0.05 0.10 0.15 -30 -20 -10 0 10 20 30
Axial strain ε1 [%] Volumetric strain εv [%] p - pav [kPa]
Fig. 11. Data during the first 24 cycles of three tests on the natural sand with the reference stress conditions (pav = 200 kPa, gav = 0.75, qampl = 60 kPa)
and different initial relative densities ID0: (a) Deviatoric stress q qav versus axial strain e1, (b) Deviatoric stress q qav versus volumetric strain ev, (c)
Void ratio change e e0 versus effective mean pressure p pav.
a) 0.8 Crushed sand b) c) 0.8

N = 105 0.8 Crushed sand Crushed sand
q [%]
q [%]
q [%]
pav = 200 kPa pav = 200 kPa ηav = 0.75
0.6 ηav = 0.75 qampl/pav = 0.3
Acc. dev. strain εacc
ηav = 0.75

0.6

0.6
ID0 = 0.62 - 0.69 qampl = 60 kPa ID0 = 0.59 - 0.70
0.4 qampl [kPa] = 0.4 ID0 = 0.4 pav [kPa] =

80 1.18 300
60 0.83 200
0.2 0.2 0.68 0.2 100
40
20 0.55 50
0 0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Acc. vol. strain εacc
v [%] Acc. vol. strain εacc
v [%]
d) e) f)
1.0 Glass beads 1.6 Natural 1.2 Crushed sand
pav = 200 kPa
ηav ηav =
q [%]
q [%]
q [%]
=
1.2 qampl = 60 kPa sand 1.0
0.8 0.875 1.25
ηav =

ID0 = 0.58
0.75 0.8 1.00
- 0.65 1.25
0.6 0.50 0.8 0.75
0 1.00 0.6 0
0.75
0.4 pav = 200 kPa 0.4 pav = 200 kPa
0.50 0.4
qampl = 60 kPa 0 qampl = 60 kPa
0.2 ID0 = 0.56 - 0.62 0 0.2 ID0 = 0.67 - 0.75
0 -0.4 0
-0.2 0 0.2 0.4 0.6 0.8 1.0 -0.4 0 0.4 0.8 1.2 1.6 0 0.4 0.8 1.2
Acc. vol. strain εacc
v [%]
Fig. 12. eacc acc

q –ev strain paths measured (a)–(c) for crushed sand in tests with different stress amplitudes, densities and average mean pressures and (d)–(f)
for glass beads, natural sand and crushed sand in tests with different average stress ratios. The solid lines in diagrams (d)–(f) have been generated using Eq.
(1) with the ucc values in Table 2.
e_ acc
2
v M cc 2 ðgav Þ 4.3. Intensity of accumulation
acc ¼ ð1Þ
e_ q 2gav
The curves of accumulated strain eacc versus the number
6 sin ucc
with M cc ¼ for triaxial compression tests. Eq. (1)
3sin ucc
of cycles N measured for the three different test materials in
means that the strain accumulation is purely volumetric the four different test series are provided in Fig. 13. The
(_eacc
q = 0) at g
av
= 0 and purely deviatoric (_eacc v = 0) at
curves obtained for the crushed sand run almost propor-
g = Mcc, i.e. at an average stress ratio corresponding to
av tional to ln(N) up to the maximum number of cycles
the critical friction angle ucc. The ucc and corresponding applied in the tests (N = 105). Most of the data collected
Mcc values delivering the optimum approximation of the for the natural sand and the glass beads obey eacc(N)
cyclic test data are summarized in Table 2. The eacc acc ln(N) only up to N = 104. At larger number of cycles, the
q –ev
rate of increase in the residual strain is faster than a loga-
strain paths predicted by Eq. (1) with these ucc values have
rithmic increase with the number of cycles, i.e. the inclina-
been added as thick solid lines in Fig. 12d–f. The data in
tion of the curves in the eacc–N diagrams with a semi-
Table 2 reveals that ucc significantly increases with increas-
logarithmic scale increases. The test data in Fig. 13 reveal
ing angularity of the grains. This is in accordance with the
that beside the grain size distribution curve (Wichtmann
higher peak friction angles uP observed in the drained
et al., 2015; Wichtmann and Triantafyllidis, 2015), the
monotonic tests (Section 3) and can be again explained by
grain characteristics influence the shape of the strain accu-
an increased interparticle friction and interlocking. Conse-
mulation curves eacc(N). The stronger decay of the rate of
quently, at a certain average stress ratio gav the ratio of vol-
strain accumulation with increasing number of cycles
umetric and deviatoric strain accumulation rate is higher
observed for the more angular materials, in particular at
for a more angular material, due to the larger distance of
N > 104, may be again the result of the interlocking
the average stress to the critical state line in the p-q plane.
between the particles, which increases due to small rear-
For comparison, the angles of repose ur obtained as the
rangements of the grains during cyclic loading and
inclination of a pluviated cone of sand and the correspond-
restrains further particle mobility.
ing Mr values are also provided in Table 2. Obviously, for
For all three materials, the intensity of strain accumula-
all three tested materials the ucc and ur values agree well.
tion grows with increasing values of amplitude (Fig. 13a–c)
This corroborates the recommendation for natural sands
and stress ratio (Fig. 13j–l) while it decreases if the sand
by Wichtmann et al. (2014) that ucc can be estimated from
becomes denser (Fig. 13d–f). In case of the natural sand,
the relationship ucc ur in the absence of cyclic test data.
the strain accumulation curves measured for the different
Based on the present test data, such simplified calibration
average mean pressures are similar if the amplitude-
seems feasible irrespective of grain shape.
Table 2
Critical friction angles ucc derived from the cyclic test data, angles of repose ur obtained from the inclination of a pluviated cone of sand, corresponding
inclinations Mcc and Mr and HCA model parameters used in the equations for the intensity of accumulation for the three tested materials.
Material ucc [°] ur [°] Mcc [–] Mr [–] Campl [–] Ce [–] Cp1 [–] Cp2 [–] CY [–] CN1 [104] CN2 [–] CN3 [–]
Glass beads 24.9 25.4 0.98 1.00 2.15 0.50 0.25 2.46 2.73 1.20 0.17 3.5
Natural sand 32.8 33.2 1.32 1.34 1.80 0.56 0 0.30 2.20 2.80 0.18 3.0
Crushed sand 37.3 37.3 1.52 1.52 0.95 0.48 0 0.19 2.33 2.50 0.15 0
pressure ratio qampl/pav is chosen identical in all tests (1990) derived analogous formulas for a contact of a cone
(Fig. 13h). For the other tested materials, however, the and a sphere, which comes closer to the contact character-
residual strains were found either to decrease (glass beads, istics of reals soils. Goddard (1990) demonstrated that
Fig. 13g) or increase (crushed sand, Fig. 13i) with growing when a certain stress level acting on the grain packing is
pressure. This is due to the differences in the pressure- exceeded, the contact of a cone and a sphere behaves sim-
dependence of the cumulative rates discussed in the next ilar to a contact of two spheres, i.e. a Hertz contact. Thus,
section. at elevated stresses, the original shape of the contact does
Fig. 14 compares the accumulated strains eacc of the not affect the contact behaviour anymore. According to
three different materials after N = 105 cycles. The data Goddard (1990) the transition stress depends on the grain
are plotted as functions of the parameters qampl, ID0, pav material and the angle of the cone, but it lies well below
and gav/Mcc (normalized average stress ratio) varied in the pressures which result in grain breakage, i.e. in the
the four test series. While the residual strain of the glass range of pressures typically applied in triaxial tests on
beads and natural sand grows approximately according sands. Furthermore, studies with uniaxial compression
to a square function with increasing stress amplitude, the tests on single grains (Cavarretta et al., 2010) have shown
increase is slightly lower than proportional (eacc qampl) that plastic deformation of contacts can occur at low con-
in the case of the crushed sand (Fig. 14a). With increasing tact stresses due to the plastification or damage of asperi-
stress and strain amplitude, particle rearrangements due to ties. Plastic deformations can be explained by the fact
rolling and sliding are facilitated in the materials with more that at low stresses applied to the grain skeleton and in par-
rounded grains, while they are restrained by interlocking in ticular in case of angular particles, the initial contact area
the case of more angular particles. between two particles is very small. Therefore, the stresses
The largest differences are obvious in the tests with dif- in the contact zone may be large even when the contact
ferent average mean pressures (Fig. 14c). The tests on glass force is low. Based on these considerations, it can be
beads showed a significant decrease in residual strain with assumed that an increase in pressure will lead to both elas-
increasing values of pav. Also, natural sand exhibited a tic and plastic deformations at the grain contacts, and that
moderate decrease while the opposite tendency was with increasing pressure applied to a specimen, the original
observed for the crushed sand. At low pressures (pav = 50 - shape of the particles involved in the contacts becomes less
kPa), the intensity of strain accumulation was higher in the important regarding the behaviour of the contacts, and
glass beads than in the natural sand, while it was particu- thus, the response of the whole granular packing. This
larly low in the crushed sand (in accordance with leads to a similar cumulative response of the materials with
undrained cyclic tests (Ahmadi and Paydar, 2014; different grain shape at elevated pressures. At lower stress
Kramer, 1996; Vaid et al., 1985)). At higher pressures levels, reorientations of the particles due to sliding and
(pav = 300 kPa), however, no significant differences rotation are easier in the assemblies of round glass beads
between the residual strain measured for the three different with their smooth surface than in the angular crushed sand
materials was detected. It should be noted that in the tests with higher interparticle friction and distinct interlocking
with pav = 300 kPa, the initial relative density was the same between adjacent grains.
(ID0 = 0.56) for glass beads and natural sand, but some- A similar tendency for the liquefaction resistance of
what higher (ID0 = 0.70) for the crushed sand, due to the materials with different grain shape is reported by Vaid
higher compressibility of the latter material during pressure et al. (1985). However, even larger pressures, up to
increase. If the crushed sand had also been tested at 800 kPa, were used in their investigation, and the observa-
ID0 = 0.56, it is likely that the residual strain would have tions were partly attributed to particle breakage effects. In
been somewhat higher; i.e. the point for the crushed sand contrast, no noticeable particle breakage was detected in
would lie slightly above those for the two other materials the present study (Section 5).
in Fig. 14c. The relationship between residual strain and relative
A possible explanation for the tendencies in Fig. 14c density (Fig. 14b) is similar for glass beads and natural
may be given based on micromechanics. Considering two sand, while the decrease in eacc with ID0 is somewhat
particles in contact, a contact force generates elastic defor- weaker for the crushed sand. Also, the dependence of resid-
mations in the contact zone, as originally demonstrated by ual strain on the normalized stress ratio gav/Mcc seems to
Hertz (1881) for the contact of two spheres. Goddard be slightly less pronounced when the angularity of the
a) 1.6 qampl [kPa] = Glass b) 2.0 qampl [kPa] = Natural c) 1.2 qampl [kPa] = Crushed sand
80 beads 80 sand 80 ID0 = 0.62
1.0
Acc. strain εacc [%]

60 1.6 60 60 - 0.69
1.2
40 40 0.8 40
pav = 200 kPa 1.2 pav = 200 kPa 20
0.8 ηav = 0.75 ηav = 0.75 0.6 pav = 200 kPa
ID0 = 0.56 - 0.59 0.8 ID0 = 0.56 - 0.61 ηav = 0.75
0.4
0.4
0.4 0.2
0 0 1 2 3 4 5
0 0 1 2 3 4 5
0 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
Number of cycles N Number of cycles N Number of cycles N
d) 1.6 ID0 = Glass e) 1.6 ID0 =
f) 1.4 ID0 =
pav = 200 kPa pav = 200 kPa
0.78 beads 0.81 ηav = 0.75 1.2 1.18 ηav = 0.75

1.2 0.56 1.2 0.60 qampl = 60 kPa 0.83 qampl = 60 kPa
0.49 1.0
0.49 0.68
pav = 200 kPa
0.8 0.55
0.8 0.8 Natural
ηav = 0.75 0.6 Crushed
qampl = 60 kPa sand
0.4 0.4 0.4 sand
0.2
0 0 1 2 3 4 5
0 0 1 2 3 4 5
0 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
g) 2.4 h) 1.4 i) 1.2
pav [kPa] = ηav = 0.75 pav [kPa] = av pav [kPa] = ηav = 0.75
η = 0.75
2.0 300 qampl/pav = 0.3 1.2 300 1.0 300 qampl/pav = 0.3

qampl/pav = 0.3
200 I = 0.56 200 200 ID0 = 0.59

D0 1.0 ID0 = 0.52
1.6 150 - 0.58 100 0.8 100 - 0.70
- 0.61
100 0.8 50 50
1.2 50 0.6
0.6
0.8 0.4
0.4
0.4 0.2 0.2
Glass beads Natural sand Crushed sand
0 0 1 2 3 4 5 0 0 1 2 3 4 5
0 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
j) 1.4 k) 2.0 ηav = l) 1.6
ηav = pav = 200 kPa pav = 200 kPa ηav = pav = 200 kPa
1.2 0.875 qampl = 60 kPa 1.25 qampl = 60 kPa 1.25 qampl = 60 kPa
0.75 I = 0.56 1.6 1.00 I = 0.58 1.00

1.2
1.0 0.50
D0 0.75 D0 - 0.65 0.75
- 0.62
0.8 0 1.2 0.50 0
0 0.8
0.6 Glass 0.8 ID0 = 0.67
beads Natural - 0.75
0.4 0.4
0.4 sand
0.2
Crushed sand
0 0 1 2 3 4 5
0 0 1 2 3 4 5
0 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 103 104 105
Fig. 13. Strain accumulation curves e (N) measured for glass beads (1st column), natural sand (2nd column) and crushed sand (3rd column) in the four
acc
test series with a variation of stress amplitude (first row), initial relative density (second row), average mean pressure (third row) and average stress ratio
(fourth row). The red curves stem from simulations of the tests with the HCA model using the optimum parameters in Table 2.
particles increases (Fig. 14d). Generally, a higher interpar- four varied parameters. As expected and in accordance
ticle friction and interlocking between adjacent particles in with earlier work (Wichtmann et al., 2009, 2015), the strain
materials with angular grains seems to weaken the depen- amplitude grows with increasing stress amplitude, decreas-
dencies of the strain accumulation rate on stress or strain ing density, increasing pressure (for f = qampl/p0 = con-
amplitude, relative density and average stress ratio. stant) and decreasing stress ratio for all three test
The results of the analysis of the elastic portion of strain materials. For all test conditions, the strain amplitudes
are shown in Fig. 15. In those diagrams mean values of the measured for the crushed sand exceeded those observed
strain amplitude during 105 cycles are plotted versus the for the two other materials. The elastic strains for the glass
a) 2.0 b) 1.6
Glass beads Glass beads
1.6 Natural sand Natural sand
εacc(N = 105) [%]
εacc(N = 105) [%]

Crushed sand 1.2 Crushed sand
1.2 pav = 200 kPa pav = 200 kPa
ηav = 0.75 0.8 ηav = 0.75
0.8 ID0 = 0.56 - 0.69 qampl = 60 kPa
0.4
0.4
0 0
0 20 40 60 80 100 0.4 0.6 0.8 1.0 1.2
Stress amplitude qampl [kPa] Relative density ID0
c) 2.4 d) 2.0
Glass beads Glass beads
2.0 1.6
Natural sand Natural sand
εacc(N = 105) [%]
εacc(N = 105) [%]

Crushed sand Crushed sand
1.6
1.2
1.2
0.8
0.8
ηav = 0.75 pav = 200 kPa
0.4 qampl/pav= 0.3 0.4 qampl = 60 kPa
ID0 = 0.52 - 0.70 ID0 = 0.56 - 0.75
0 0
0 100 200 300 0 0.2 0.4 0.6 0.8 1.0
Average mean pressure pav [kPa] Normalized stress ratio ηav/Mcc
Fig. 14. Accumulated strain eacc after N = 105 cycles as a function of (a) stress amplitude qampl, (b) initial relative density ID0, (c) average mean pressure
pav and (d) normalized average stress ratio gav/Mcc.
beads and the natural sand were quite similar. Analogously of the HCA model (see Table 3) in order to eliminate the
to the stiffness E50 in the monotonic tests (Section 3), the influence of slightly different initial densities and the higher
larger strain amplitudes, and thus the lower values of compaction rates e_ ¼ @e=@N in the tests with larger stress
secant stiffness in case of the crushed sand, can be partially amplitudes. The void ratio e entering the void ratio func-
attributed to the higher void ratios and lower coordination tion is also a mean value over the N previously applied
numbers at same relative density. Furthermore, the stiffness cycles. A comparison of the diagrams in Fig. 16a–c reveals
at small to moderate strains reflects the nature of interpar- that the increase in the residual strain eacc =f e with increas-
ticle contacts. Below the transition pressure, contacts ing strain amplitude eampl was more pronounced for the
between angular particles are more deformable, as can be glass beads than for the natural sand. In agreement with
readily shown by comparing a cone-to-plane contact Fig. 14a, in case of the crushed sand at large numbers of
(Goddard, 1990) versus a sphere-to-plane contact (Hertz, cycles, the growth of the residual strain with increasing
1881), cf. Cho et al. (2006). strain amplitude was slower than linearly. The parameters
Campl of the amplitude function fampl (Table 3) were gath-
ered from the diagrams in Fig. 16a–c by fitting the function
4.4. HCA model parameters C
f ¼ k ðeampl Þ ampl to the data for each N value (solid curves
in Fig. 16a–c), resulting in the parameter Campl and a con-
The diagrams provided in Fig. 16 were used to calibrate
stant k which is not used further. The mean value of the
the HCA model for the three different test materials (the
parameters collected for the different numbers of cycles
procedure is explained in detail by Wichtmann et al.
was chosen as Campl (the Campl value given in Table 2
(2010a) and Wichtmann (2016)). The first row of diagrams
results from a further optimization as explained below).
shows the residual strain eacc as a function of the strain
The second row of diagrams in Fig. 16 contains the data
amplitude eampl for different numbers of cycles. Since the from the test series with a variation of initial density. In
tests were performed with stress cycles, the strain amplitude those diagrams the residual strain is plotted against the
slightly varied (usually decreases) with increasing number void ratio e. The residual strain was divided by the ampli-
of cycles. The strain amplitude eampl is a mean value up tude function f ampl of the HCA model with the aim of elim-
to the N value under consideration. On the ordinate the
inating the effect of the moderately different strain
eacc data have been divided by the void ratio function f e
a) b)
7 7
Strain amplitude εampl [10-4]

Strain amplitude εampl [10-4] 6 Glass beads 6 Glass beads
5 Crushed sand 5 Crushed sand
4 4
3 3
2 2
1 1
0 0
0 20 40 60 80 100 0.4 0.6 0.8 1.0 1.2
Stress amplitude qampl [kPa] Relative density ID0
c) d)
6 7

5 6
5
4
4
3
3
2 Glass beads Glass beads
2
1 Crushed sand 1 Crushed sand
0 0
0 100 200 300 0 0.2 0.4 0.6 0.8 1.0
Average mean pressure pav [kPa] Normalized stress ratio ηav/Mcc
Fig. 15. Strain amplitude eampl (mean values over 105 cycles) as a function of (a) stress amplitude qampl, (b) initial relative density ID0, (c) average mean
pressure pav and (d) normalized average stress ratio gav/Mcc.
amplitudes eampl (Fig. 15b). The increase of the residual The residual strain has been divided by f ampl and f e in
strain with increasing void ratio was similar for all three order to purify it from the influences of strain amplitude
tested materials. The HCA model parameter Ce of the void (Fig. 15c) and void ratio. In contrast to Fig. 14c, the data
ratio function fe, corresponding to the void ratio for which in Fig. 16g–i thus refers to a constant strain amplitude. The
e_ acc ¼ 0 holds, was obtained by fitting the function f = k data reveal that the decrease in the strain accumulation rate
(Ce e)2/(1 + e) to the data for each number of cycles in with increasing pressure was much more pronounced for
Fig. 16d–f (solid curves). While the parameter k was not the glass beads than for the natural sand. In the case of
further used, the Ce values derived for the various N values crushed sand, the opposite was found, i.e. e_ acc slightly
were finally averaged. Since the function f e with the increases with increasing pressure. The data for the glass
parameter Ce is necessary on the ordinate of the diagrams beads in Fig. 16g could not be fitted sufficiently well by
in Fig. 16a–c and the function f ampl with its parameter the original function fp of the HCA model (Table 3, for
Campl is used in the diagrams in Fig. 16d–f, the simultane- the fitting the parameter k is again set in front of fp). Thus,
ous determination of the material constants Campl and Ce the function was extended by a constant portion according
has to be done by iteration. As a first estimate for Ce, the to:
correlation Ce = 0.95 emin can be applied. With this value f p ¼ Cp1 þ ð1 Cp1 Þexp½Cp2 ðpav =100 1Þ ð2Þ
for Ce, the first determination of Campl was conducted. This
Campl value was used in the first curve-fitting for Ce. With with two parameters Cp1 and Cp2. For Cp1 = 0 the original
the obtained Ce, the determination of Campl was refined, function is regained with Cp = Cp2. In accordance with the
and so on. Usually, two or three iterations were sufficient original function, also the modified fp according to Eq. (2)
to obtain the final values of Campl and Ce. The solid curves delivers fp = 1 at the reference pressure pav = 100 kPa. In
shown in Fig. 16a–c and d–f stem from the last iteration. Ce case of the natural sand and the crushed sand, the original
corresponds to the void ratio where the curves in the dia- function was sufficient to approximate the data in Fig. 16h,
grams in Fig. 16d–f would intersect the ordinate. i and thus Cp1 = 0 holds for these two materials. The
The third row of diagrams in Fig. 16 was used to analyze increasing trend of e_ acc with pav observed for the crushed
the pressure-dependence of the strain accumulation rate. sand results in a negative Cp2 value. The procedure of
a) 5 legend for b) 10 c) 2.5

all diagrams: Glass beads Natural sand Crushed sand
4 N= 8 2.0
100000
εacc / fe [%]
εacc / fe [%]
εacc / fe [%]
3 10000 6 1.5
1000
2 100 4 1.0
10
1 2 0.5
0 0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6 0 2 4 6 8
Strain amplitude εampl -4
[10 ] Strain amplitude εampl [10 ] -4
d) 0.20 e) 0.20 f) 0.25
0.16 εacc / fampl [%] 0.16 0.20
εacc / fampl [%]

εacc / fampl [%]
0.12 0.12 0.15
0.08 0.08 0.10
0.04 0.04 0.05
0 0 0
0.54 0.56 0.58 0.60 0.62 0.64 0.60 0.64 0.68 0.72 0.76 0.6 0.7 0.8 0.9 1.0
Void ratio e Void ratio e Void ratio e
g) 2.0 h) 1.6 i) 0.8
εacc / (fampl fe) [%]

1.6
1.2 0.6
1.2
0.8 0.4
0.8
0.4 0.2
0.4
0 0 0
0 100 200 300 0 100 200 300 0 100 200 300
Av. mean pressure pav [kPa] Av. mean pressure pav [kPa] Av. mean pressure pav [kPa]
j) 0.5 k) 2.0 l) 1.2
1.0
0.4 1.6
0.8
0.3 1.2
0.6
0.2 0.8
0.4
0.1 0.4 0.2
0 0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8
Average stress ratio Y av
Average stress ratio Y av
Average stress ratio Yav
Fig. 16. Accumulated strain (normalized with fe and/or fampl to purify the data from the influences of void ratio and/or amplitude, respectively) versus (a)
strain amplitude eampl (mean value up to cycle number

Nunder consideration), (b)
void ratio e (mean value up to cycle number N), (c) average mean
pressure pav and (d) normalized average stress ratio Y av (Y av = 0 for gav = 0 and Y av = 1 for gav = Mcc).
curve-fitting to obtain Cp1 and Cp2 was identical to that in ting of the function k fY to the data in those diagrams
case of Campl and Ce, but no iterative procedure was neces- delivered CY.
sary. The final Cpi values result from an averaging of the Afterwards, an element test program has been used in
individual Cpi values determined for the various numbers order to determine the parameters CN1, CN2 and CN3
of cycles. describing the increase of the residual strain with increasing
The last row of diagrams in Fig. 16 was used to deter- number of cycles (Table 3), and in order to optimize the
mine the parameter CY in the stress ratio function fY of parameters Campl, Ce, Cp1, Cp2 and CY. All cyclic tests per-
the HCA model (Table 3). The residual strain divided by formed in this study were simulated with the element test
the void ratio and amplitude functions is plotted versus program. The parameters derived from the diagrams in

the normalized average stress ratio Y av (Y av = 0 for Fig. 16 were used as the starting point. The parameters
were varied in consecutive simulations until the best con-
gav = 0 and Y av = 1 for gav = Mcc, see Appendix A). A fit-
Table 3 tested in the present study. Grain crushing was quantified

Summary of the functions and material constants of the HCA model. by accurate sieve analyses of the whole triaxial sample once
Function Material constants before and once after a test with 105 cycles. A comparison

ampl C ampl Campl of the grain size distribution curves before and after the test
f ampl ¼ min e
104
; 10C ampl
(as recommended by Hardin (1985) to evaluate the relative
A B
f_ N ¼ f_ N þ f_ N CN1, CN2, CN3 breakage ratio Br) showed only a very small increase of the
A
h i
f_ N ¼ C N1 C N 2 exp CN 1gf
A
amount of fines. The low amount of particle breakage is
ampl
B probably due to the moderate average and cyclic stresses
f_ N ¼ C N1 C N 3
2
applied in the tests. Based on this test series, no significant
f e ¼ ðC1þe
e eÞ 1þemax
ðC e emax Þ2
Ce particle breakage or abrasion was expected for the quartz
h av i grains tested in the present study.
p Cp
f p ¼ exp C p 100kPa 1
Sands with crushable particles like carbonate sands are
f Y ¼ exp C Y Y av CY often composed of grains with a rather angular or irregular
fp = 1 for constant polarization, Wichtmann et al. (2007a).
shape. Therefore, the material response under monotonic
or cyclic loading is influenced by both the crushability
and the particle shape. It is reasonable to expect that crush-
able sands show larger cumulative rates, not only because
gruence between the measured and the predicted curves
of particle breakage but also due to the higher void ratios
eacc(N) was achieved. The optimum HCA model parame-
and thus larger compressibility typically observed for such
ters are summarized in Table 2. The eacc(N) curves pre-
granular materials. However, several studies in the litera-
dicted by the HCA model with these parameters were
ture with undrained cyclic loading have found the opposite
added as red solid lines in Fig. 13, confirming the good
tendency. Crushable sands do not liquefy as easily as
agreement between the measured and calculated data for
harder-grained sands of same density (e.g. Duku et al.,
most tests.
2008; Hyodo et al., 1996; Hyodo et al., 1998; López-
Finally, the HCA model parameters are plotted versus
Querol and Coop, 2012; Salem et al., 2013; Salleh, 1992),
the grain shape parameters Circularity and Aspect Ratio
primarily due to interlocking effects resulting from the
(Table 1) in Fig. 17. The exponent Campl of the amplitude
more angular particle shape, rendering the fabric more
function increases with increasing Circularity or decreasing
stable against undrained cyclic loading. The opposite beha-
Aspect Ratio of the grains (Fig. 17a and b). No clear
viour, i.e. a higher liquefaction susceptibility for crushable
dependence of Ce on the grain shape parameters can be
sands compared to harder-grained materials has been,
detected in Fig. 17c and d. The parameters Cp2 in
however, sometimes observed at higher densities and larger
Fig. 17e and f and CY in Fig. 17g and h exhibit an increase
stresses (Hyodo et al., 1998; Salem et al., 2013). The latter
with increasing Circularity and decreasing Aspect Ratio.
may be partly attributed to the larger amount of particle
No diagram for Cp1 is provided in Fig. 17 since only one
data point unequal to zero is available at present. While breakage occurring at elevated pressure levels.
CN1 is lower for higher values of Circularity and lower val- Some studies on calcareous sands in the literature (e.g.
ues of Aspect Ratio (Fig. 17i and j), no clear trend can be Donohue et al., 2009) report a considerable amount of
observed for CN2 (Fig. 17k and l) and the opposite relation- breakage of particles caused by a cyclic loading. It grows
ship was obtained for CN3 (Fig. 17m and n). Note that the with increasing number of cycles and correlates with the
permanent volumetric strain. In contrast, negligible break-
parameter CN3 could be better judged based on tests with
age was observed in other experimental work (López-
even higher numbers of cycles (N 106) (Wichtmann
et al., 2015). Summing up, Fig. 17 demonstrates that sev- Querol and Coop, 2012; Salem et al., 2013; Sandoval and
eral HCA model parameters strongly depend on grain Pando, 2012). For a monotonic loading of sands partly
shape. or completely composed of crushable particles the amount
of crushing has been found dependent on stress level, void
ratio and particle characteristics (size, grading, shape,
5. Remarks on grain crushing strength, mineral composition) (Altuhafi and Coop, 2011;
Coop, 1990; Coop et al., 2004; Hardin, 1985; Lade et al.,
Neither grain crushing nor particle abrasion have been 1996; Leleu and Valdes, 2007; McDowell and Bolton,
explicitly investigated in the present study. Both mecha- 1998; Nakata et al., 1999; Nakata et al., 2001).
nisms are widely believed to lead to a certain amount of
fines in the mixture at the end of a test. This, however, 6. Influence of sample dimensions
has not been observed in the present experiments.
Another series of drained cyclic tests with similar ranges In the present study, samples with a height-to-diameter
of relative densities and stresses (i.e. stress paths as shown ratio of h/d = 1 were tested, while samples with h/d = 2
in Fig. 10) was recently performed by the authors on a car- are more common internationally. While both geometries
bonate sand from The Philippines, composed of grains are standard in the German code DIN 18137, the 1:1 sam-
more susceptible to crushing than the quartz particles ples are more frequently used in experimental studies ded-
a) 3 b) 3 c) 0.6 d) 0.6
2 2
Campl
Campl
Ce
Ce
0.5 0.5
1 1
0 0 0.4 0.4
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
Circularity Aspect ratio Circularity Aspect ratio
e) 3 f) 3 g) h)
2 2 2.8 2.8
Cp (Cp2)
Cp (Cp2)
CY
CY
1 1
2.4 2.4
0 0
-1 -1 2.0 2.0
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
i) 4 j) 4 k) 0.3 l) 0.3
3 3
CN1 [10-4]
CN1 [10-4]
0.2 0.2
CN2
CN2
2 2
0.1 0.1
1 1
0 0 0 0
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
m) 5 n) 5
4 4
CN3 [10-5]
CN3 [10-5]
3 3
2 2
1 1
0 0
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
Circularity Aspect ratio
Fig. 17. HCA model parameters as a function of grain shape parameters Circularity and Aspect Ratio.
icated to constitutive modeling since they are believed to samples may be due to the different pattern of shear bands
deform more homogeneously. In the following the influ- in the samples (Desrues et al., 1996). The higher uP values
ence of the sample dimensions is discussed based on an for the smallest sample geometry may be the result of a lar-
additional test series performed on the natural sand. Beside ger influence of end restraints. In contrast to Bishop and
the 1:1 and 2:1 samples with a diameter of 100 mm, also Green (1965), increasing the number of lubrication layers
smaller (d = 50 mm) and larger (d = 150 mm) 1:1 samples (e.g. using two layers of grease and rubber disk at each
were tested. Schemes of the different sample dimensions end plate) was found ineffective regarding a further reduc-
are provided at the top of Fig. 18. tion of uP for the 1:1 samples with d = 100 mm in another
The first row of diagrams in Fig. 18 presents results of series of tests on a fine sand (Wichtmann, 2016).
drained monotonic triaxial tests. A loose and a dense sam- Since the present study concentrates on the cumulative
ple were tested for each sample geometry. The principal behaviour under cyclic loading, the influence of the sample
shape of the curves q(e1) and ev(e1) (Fig. 18a and b) is quite geometry in cyclic tests is of larger interest. The second row
similar for the different sample dimensions, with the only of diagrams in Fig. 18 presents a comparison of cyclic tests
exception that the curve q(e1) measured for the dense 2:1 performed with the four different sample dimensions. For
sample reaches its peak at a slightly lower axial strain. each geometry several samples with different initial relative
The deviations of the curves in Fig. 18a and b are mainly densities were tested. Neither the strain accumulation
due to the variations in initial relative density. The uP- curves eacc(N) for medium dense samples in Fig. 18d nor
ID0 diagram in Fig. 18c reveals, however, an influence of the relationship between the accumulated strain eacc after
the sample geometry on the peak friction angle. At low 105 cycles and the initial relative density ID0 in Fig. 18e

densities, the 2:1 samples with d = 100 mm show an about nor the plot of deviatoric strain eacc
q N ¼ 105 versus volu-
2° lower peak friction angle than the 1:1 samples of same
metric strain eacc
v N ¼ 105 in Fig. 18f show a distinct influ-
diameter. These differences decrease with increasing den-
ence of the sample dimensions. Therefore, it can be
sity. The uP data for the samples with d = h = 150 mm
concluded that the results of the cyclic tests presented in
agree well with those for d = h = 100 mm, while the values
this paper are not significantly affected by the (unusual)
for d = h = 50 mm are again about 1–1.5° higher. The dif-
sample geometry.
ferences in the peak friction angles between the 1:1 and 2:1
Legend for c)-f):

Legend for a)+b):
Sample dimensions
20 cm
Dimensions h x d / ID0 loose / ID0 dense = hxd=
15 cm
50 x 50 / 0.47 / 1.03
10 cm
50 x 50 mm
5 cm
100 x 100 / 0.22 / 0.98 100 x 100 mm
150 x 150/ 0.20 / 0.92 150 x 150 mm
200 x 100/ 0.21 / 0.95 200 x 100 mm
5 cm 10 cm 15 cm 10 cm
a) b) c)
500 46
Peak friction angle ϕP [°]

Deviatoric stress q [kPa]
Volumetric strain εv [%]

dense 0
400
42
-4
300 loose
38
-8
200
loose dense
-12 34
100
ϕr
0 -16 30
0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 0.2 0.4 0.6 0.8 1.0 1.2
Axial strain ε1 [%] Axial strain ε1 [%] Initial relative density ID0
d) e) f)
1.0 3.0 2.5
q [%]
pav = 200 kPa pav = 200 kPa pav = 200 kPa
2.5
0.8 ηav = 0.75 ηav = 0.75 2.0 ηav = 0.75

qampl = 60 kPa 2.0 qampl = 60 kPa qampl = 60 kPa
0.6 ID0 = 0.59 - 0.61 1.5
1.5
0.4 1.0
1.0
0.2 0.5 0.5
0 0 0 0
1 2 3 4 5 0 0.2 0.4 0.6 0.8 1.0 1.2 0 0.5 1.0 1.5 2.0 2.5
10 10 10 10 10 10
Number of cycles N Initial relative density ID0 Acc. vol. strain εacc
v [%]
Fig. 18. Influence of sample geometry on the results of drained monotonic triaxial tests on the natural sand: (a) Stress-strain curves q(e1), (b) volumetric
strain ev(e1), (c) density-dependent peak friction angle u(ID0); Influence of sample geometry on the results of drained cyclic triaxial tests on the natural
sand: (d) Strain accumulation curves eacc(N) for eacc
medium density, (e) accumulated strain acc
5
after N5 = 10 cycles as a function of initial relative density ID0,
(f) accumulated deviatoric strain eq N ¼ 10 versus accumulated volumetric strain ev N ¼ 10 after 105 cycles.
acc 5
7. Summary, conclusions and outlook interparticle friction and interlocking between adjacent
grains.
Three granular materials with significantly different The curves of accumulated strain eacc versus the number
grain shape and surface characteristics were studied in of cycles N are also affected by the grain shape. For the
drained cyclic triaxial tests. The glass beads, the natural crushed sand they run almost proportional to ln(N) up to
sand with its subrounded grains and the crushed sand with N = 105. For the natural sand and the glass beads
its very angular particles were tested with the same spe- eacc(N) ln(N) is obeyed only up to N = 104. At larger
cially mixed grain size distribution curve. Four test series number of cycles the inclination of the curves in eacc-N
with different amplitudes, densities, average mean pressures diagrams with semi-logarithmic scale increases.
and average stress ratios were performed on each material. The higher interparticle friction and interlocking in the
The maximum number of cycles (105) was the same in all more angular materials affects also the intensity of strain
tests. accumulation. While for glass beads and natural sand the
The direction of accumulation, i.e. the ratio e_ acc
v =_eacc
q of residual strain at N = 105 grows almost proportional to
the volumetric and deviatoric strain accumulation rates, the square of the stress amplitude, an approximately linear
was found to be almost independent of amplitude, density relationship between eacc and qampl was observed for the
and average mean pressure for all three tested materials. crushed sand. The largest differences between the various
The direction of accumulation was mainly determined by grain shapes are obvious in the pressure dependence of
the average stress ratio gav. While e_ acc
q = 0 (purely isotropic the strain accumulation rates. While the glass beads and
strain accumulation) was approximately fulfilled for isotro- (less pronounced) the natural sand showed a decrease in
pic average stresses (gav = 0) independently of grain shape, the residual strain with increasing values of average mean
the stress ratio gav = Mcc corresponding to a purely devia- pressure pav, the tendency for the crushed sand was
toric strain accumulation (_eacc
v = 0) grew with increasing the opposite. Due to the larger potential for grain
angularity of the particles. This was attributed to the higher rearrangements by rolling and sliding, at low pressures
(pav = 50 kPa) the accumulation of strain was higher in the The basic equation of the HCA model reads
glass beads than in the natural sand and in particular in the
r_ ¼ E : e_ e_ acc e_ pl ð3Þ
crushed sand. The differences diminished at higher pres-
sures (pav = 300 kPa), however. This is probably a result with the stress rate r_ of the effective Cauchy stress r (com-
of the increase in the deformations at the grain contacts pression positive), the strain rate e_ (compression positive),
with growing pressure, which render the contact behaviour the accumulation rate e_ acc , a plastic strain rate e_ pl (necessary
almost independent of the original shape. The relationships only for stress paths touching the yield surface) and the
between residual strain and relative density or average barotropic elastic stiffness E. In the context of HCA models
stress ratio were found to be quite similar for the three the dot over a symbol means a derivative with respect to
tested grain shapes, with the dependencies being slightly the number of cycles N (instead of time t), i.e.
less pronounced for the crushed sand. For all test condi- F_ ¼ @ F =@N . Depending on the boundary conditions,
tions, the strain amplitudes (elastic portion of deformation) Eq. (3) predicts either a change of average stress (r–0) _
measured for the crushed sand were larger than those or an accumulation of residual strain (_e–0) or both.
observed for the other two materials. This was due to the For e_ acc in Eq. (3) the following multiplicative approach
higher void ratio at same relative density, the lower coordi- is used:
nation number and the weaker response of the angular
contacts. e_ acc ¼ e_ acc m ð4Þ
Based on the experimental data the parameters of the with the direction of strain accumulation (flow rule)
high-cycle accumulation (HCA) model of Niemunis et al. m ¼ e_ acc =k_eacc k ¼ ðe_ acc Þ! (unit tensor) and the intensity of
(2005) were calibrated. The function fp was extended by strain accumulation e_ acc ¼ k_eacc k. Based on the own test
an additional term in order to adequately describe the data results (Wichtmann, 2005; Wichtmann et al., 2006;
collected for the glass beads. The HCA model parameters Wichtmann et al., 2014) and corroborated by the literature
were analyzed in dependence of grain shape parameters (Chang and Whitman, 1988; Luong, 1982) the flow rule of
determined from an automated analysis. The parameters the modified Cam clay (MCC) model is adopted for m:
Campl, Cp, CY and CN3 were found to increase with increas- " ! #!
2
ing circularity and decreasing aspect ratio of the particles, 1 av ðqav Þ 3 av
m¼ p 2 av 1 þ 2 ðr Þ ð5Þ
while the opposite tendency (CN1), or no clear correlation 3 M p M
(Ce, CN2), was observed for the remaining parameters. F! F F
Currently, the experimental investigation is being where ¼ =k k denotes the normalization of a ten-
extended to several other natural sands with varying grain sorial quantity. For the triaxial case the critical stress ratio
shape. Based on the available and complemented data the M = F Mcc is calculated from
simplified calibration procedure for the HCA model 8 M ec
parameters will be extended by the influence of the grain <1 þ 3
> for gav M ec
av
shape in future. F ¼ 1 þ g3 for M ec < gav < 0 ð6Þ
>
:
1 for g av
0
Acknowledgements
wherein
The presented study has been funded by the German
6 sin ucc 6 sin ucc
Research Council (DFG, project No. TR 218/18-1/WI M cc ¼ and M ec ¼ ð7Þ
3 sin ucc 3 þ sin ucc
3180/3-1). The authors are grateful to DFG for the
financial support. The monotonic and cyclic tests have with parameter ucc.
been performed by the technicians H. Borowski, P. Gölz The intensity of strain accumulation e_ acc in Eq. (4) is cal-
and N. Demiral in the IBF soil mechanics laboratory. culated as a product of six functions:
Appendix A. Equations of the HCA model e_ acc ¼ f ampl f_ N f e f p f Y f p ð8Þ

each considering a single influencing parameter (see
The HCA model has been originally proposed by Table 3), i.e. the strain amplitude eampl (function fampl),
Niemunis et al. (2005). The model is based on the compre- the cyclic preloading gA (f_ N ), void ratio e (fe), average
hensive experimental parametric study documented in
(Wichtmann, 2005; Wichtmann et al., 2005, 2006, 2007a, mean pressure pav (fp), average stress ratio gav or Y av (fY)
2007b). Deficits (lack of generality, missing influencing and the effect of polarization changes (fp = 1 for a constant
parameters, 1D formulation) of older HCA models pro- polarization as in the case of the test series presented in this
posed in the literature (Bouckovalas et al., 1984; Diyaljee paper).

and Raymond, 1982; Gotschol, 2002; Kaggwa et al., The normalized stress ratio Y av used in fY is zero for iso-
1991; Marr and Christian, 1981; Sawicki and Swidzinski, tropic stresses (gav = 0) and one on the critical state line
1987; Sawicki and Swidzinski, 1989) have been discussed (gav = Mcc). The function Y of Matsuoka and Nakai
by Wichtmann (2005). (1982) is used for that purpose:
Y av 9 9 sin2 ucc Bishop, A.W., Green, G.E., 1965. The influence of end restraint on the
Y av ¼ with Yc ¼ ð9Þ compression strength of a cohesionless soil. Géotechnique 15 (3), 243–
Yc 9 1 sin2 ucc 266.
27ð3 þ gav Þ Bouckovalas, G., Whitman, R.V., Marr, W.A., 1984. Permanent dis-
Y av ¼ ð10Þ placement of sand with cyclic loading. J. Geotech. Eng. 110 (11), 1606–
ð3 þ 2gav Þð3 gav Þ 1623.
Brown, S.F., 1974. Repeated load testing of a granular material. J.
For a constant strain amplitude, the function fN simplifies
Geotech. Eng. Div. 100 (7), 825–841.
to: Cavarretta, I., Coop, M., O’sullivan, C., 2010. The influence of particle
f N ¼ CN1 ½lnð1 þ CN2 NÞ þ CN3 N ð11Þ characteristics on the behaviour of coarse grained soils. Géotechnique
60 (6), 413–423.
The formulations of the functions fampl, fN, fe, fp, fY, fp Chang, C.S., Whitman, R.V., 1988. Drained permanent deformation of
have been chosen based on the own experimental results sand due to cyclic loading. J. Geotech. Eng. 114 (10), 1164–1180.
Cho, G.-C., Dodds, J., Santamarina, J.C., 2006. Particle shape effects on
and documented data in the literature (Brown, 1974;
packing density, stiffness, and strength: natural and crushed sands. J.
Diyaljee and Raymond, 1982; Raymond and Williams, Geotech. Geoenviron. Eng. 132 (5), 591–602.
1978; Sawicki and Swidzinski, 1987; Sawicki and Coop, M.R., 1990. The mechanics of uncemented carbonate sands.
Swidzinski, 1989; Silver and Seed, 1971a; Silver and Seed, Géotechnique 40 (4), 607–626.
1971b; Timmermann and Wu, 1969; Youd, 1972). They Coop, M.R., Sorensen, K.K., Bodas Freitas, T., Georgoutos, G., 2004.
Particle breakage during shearing of a carbonate sand. Géotechnique
are also corroborated by more recent studies (Duku
54 (3), 157–163.
et al., 2008; Indraratna et al., 2005; Lackenby et al., Cox, M.R., Budhu, M., 2008. A practical approach to grain shape
2007; Suiker et al., 2005; Yee et al., 2014). Since several quantification. Eng. Geol. 96, 1–16.
experimental investigations in the literature reveal that Desrues, J., Chambon, R., Mokni, M., Mazerolle, F., 1996. Void ratio
the frequency does not influence the cumulative rates in evolution inside shear bands in triaxial sand specimens studied by
computed tomography. Géotechnique 46 (3), 529–546.
sand under drained or undrained cyclic loading (Duku
Diyaljee, V.A., Raymond, G.P., 1982. Repetitive load deformation of
et al., 2008; Karg and Haegeman, 2009; Kokusho et al., cohesionless soil. J. Geotech. Eng. Div. 108 (GT10), 1215–1229.
2004; Peacock and Seed, 1968; Quadimi and Coop, 2007; Donohue, S., O’Sullivan, C., Long, M., 2009. Particle breakage during
Shenton, 1978; Tatsuoka et al., 1986a; Tatsuoka et al., cyclic triaxial loading of a carbonate sand. Géotechnique 59 (5), 477–
1986b; Wong et al., 1975; Yasuda and Soga, 1984; 482.
Duku, P.M., Stewart, J.P., Whang, D.H., Yee, E., 2008. Volumetric
Yoshimi and Oh-Oka, 1975; Youd, 1972) the loading fre-
strains of clean sands subject to cyclic loads. J. Geotech. Geoenviron.
quency is not considered as a parameter of the HCA Eng. 134 (8), 1073–1085.
model. While the parameter Campl of the amplitude func- Flemming, L., 2015. Studies of the surface characteristics of sand particles
tion fampl was fixed to a value of 2.0 in earlier versions of (in German). Diploma thesis, Institute for Soil Mechanics and Rock
the HCA model (Niemunis et al., 2005; Wichtmann, Mechanics, Karlsruhe Institute of Technology (KIT).
Galindo, R., Illueca, M., Jimenez, R., 2014. Permanent deformation
2005; Wichtmann et al., 2010a), later experimental results
estimates of dynamic equipment foundations: application to a gas
for various sands exhibited the need to introduce it as a turbine in granular soils. Soil Dyn. Earthq. Eng. 63 (1), 8–18.
material constant (Wichtmann et al., 2015). The extension Goddard, J.D., 1990. Nonlinear elasticity and pressure-dependent wave
of fp as proposed by Eq. (2) represents another modifica- speeds in granular media. In: Proceedings of the Royal Society
tion of the original HCA model formulation. Detailed London, vol. 430, pp. 105–131.
Gori, U., Mari, M., 2001. The correlation between the fractal dimension
investigations on the elastic stiffness E in Eq. (3) may be
and internal friction angle of different granular materials. Soils Found.
found in (Wichtmann et al., 2010b, Wichtmann et al., 41 (6), 17–23.
2013). A more general formulation for the flow rule m Gotschol, A., 2002 Veränderlich elastisches und plastisches Verhalten
enabling to consider anisotropy has been published by nichtbindiger Böden und Schotter unter zyklisch-dynamischer Bean-
Wichtmann et al. (2010d). The large experimental effort spruchung. PhD thesis, University Kassel.
Guo, P., Su, X., 2007. Shear strength, interparticle locking, and dilatancy
for the calibration of the material constants of the HCA
of granular materials. Can. Geotech. J. 44 (5), 579–591.
model lead to the development of a simplified calibration Ham, A., Wang, J., Stammer, J.G., 2012. Relationships between particle
procedure based on correlations with the grain size distri- shape characteristics and macroscopic damping in dry sands. J.
bution curve and index test results (Wichtmann et al., Geotech. Geoenviron. Eng. 138 (8), 1002–1011.
2009; Wichtmann et al., 2015). Hardin, B., 1985. Crushing of soil particles. J. Geotech. Eng. 111 (10),
1177–1192.
Hertz, H., 1881. Über die Berührung fester elastischer Körper. Journal
References reine und angewandte Mathematik 92, 156–171.
Hyodo, M., Aramaki, N., Itoh, M., Hyde, A.F.L., 1996. Cyclic strength
Ahmadi, M.M., Paydar, N.A., 2014. Requirements for soilspecific and deformation of crushable carbonate sand. Soil Dyn. Earthq. Eng.
correlation between shear wave velocity and liquefaction resistance 15 (5), 331–336.
of sands. Soil Dyn. Earthquake Eng. 57, 152–163. Hyodo, M., Hyde, A.F.L., Aramaki, N., 1998. Liquefaction of crushable
Altuhafi, F.N., Coop, M.R., 2011. Changes to particle characteristics soils. Géotechnique 48 (4), 527–543.
associated with the compression of sands. Géotechnique 61 (6), 459– Indraratna, B., Lackenby, J., Christie, D., 2005. Effect of confining
471. pressure on the degradation of ballast under cyclic loading. Géotech-
Andersen, K.H., Schjetne, K., 2013. Database of friction angles of sand nique 55 (4), 325–328.
and consolidation characteristics of sand, silt, and clay. J. Geotech. Jänke, S., 1968. Zusammendrückbarkeit und Scherfestigkeit nichtbindiger
Geoenviron. Eng. 139 (7), 1140–1155. Erdstoffe – ihre quantitative Ermittlung mit Hilfe einfacher Kennwerte
und Feststellung der sie bestimmenden Einflußfaktoren. Baumaschine Powers, M.C., 1953. A new roundness scale for sedimentary particles. J.
und Bautechnik 15 (3), 91–101. Sediment. Petrol. 23 (2), 117–119.
Kaggwa, W.S., Booker, J.R., Carter, J.P., 1991. Residual strains in Quadimi, A., Coop, M.R., 2007. The undrained cyclic behaviour of a
calcareous sand due to irregular cyclic loading. J. Geotech. Eng. 117 carbonate sand. Géotechnique 57 (9), 739–750.
(2), 201–218. Raymond, G.P., Williams, D.R., 1978. Repeated load triaxial tests on a
Karg, C., Haegeman, W., 2009. Elasto-plastic long-term behavior of dolomite ballast. J. Geotech. Eng. 104, 1013–1029.
granular soils: experimental investigation. Soil Dyn. Earthq. Eng. 29, Rittenhouse, G., 1943. A visual method of estimating two-dimensional
155–172. sphericity. J. Sediment. Petrol. 13 (2), 79–81.
Koerner, R.M., 1970. Effect of particle characteristics on soil strength. J. Rousé, P.C., Fannin, R.J., Shuttle, D.A., 2008. Influence of roundness on
Soil Mech. Found. Div. 96 (SM4), 1221–1234. the void ratio and strength of uniform sand. Géotechnique 58 (3), 227–
Kokusho, T., Hara, T., Hiraoka, R., 2004. Undrained shear strength of 231.
granular soils with different particle gradations. J. Geotech. Geoenv- Sadrekarimi, A., Olson, S.M., 2011. Critical state friction angle of sands.
iron. Eng. 130 (6), 621–629. Géotechnique 61 (9), 771–783.
Kramer, S.L., 1996. Geotechnical Earthquake Engineering. Prentice-Hall, Salem, M., Elmamlouk, H., Agaiby, S., 2013. Static and cyclic behavior of
Upper Saddle River, N.J.. North Coast calcareous sand in Egypt. Soil Dyn. Earthq. Eng. 55 (10),
Krumbein, W., 1941. Measurement and geological significance of shape 83–91.
and roundness of sedimentary particles. J. Sediment. Petrol. 11 (2), 64– Salleh, S.B.M., 1992. Cyclic loading of carbonate sand. PhD thesis.
72. University of Bradford.
Krumbein, W.C., Sloss, L.L., 1963. Stratigraphy and Sedimentation, Sandoval, E.A., Pando, M.A., 2012. Experimental assessment of the
second ed. Freeman, San Francisco. liquefaction resistance of calcareous biogeneous sands. Earth Sci. Res.
Lackenby, J., Indraratna, B., McDowell, G., Christie, D., 2007. Effect of J. 16 (1).
confining pressure on ballast degradation and deformation under Santamarina, C., 2012. Soil phenomena at the particle/pore scale.
cyclic triaxial loading. Géotechnique 57 (6), 527–536. Presentation at IBF, KIT.
Lade, P.V., Yamamuro, J.A., Bopp, P.A., 1996. Significance of particle Sawicki, A., Swidzinski, W., 1987. Compaction curve as one of basic
crushing in granular materials. J. Geotech. Eng. 122 (4), 309–316. characteristics of granular soils. In: Flavigny, E., Cordary, D. (Eds.),
Leleu, S.L., Valdes, J.R., 2007. Experimental study of the influence of 4th Colloque Franco-Polonais de Mechanique des Sols Appliquee,
mineral composition on sand crushing. Géotechnique 57 (3), 313–317. Grenoble, Vol. 1, pp. 103–115.
Liu, Q.B., Lehane, B.M., 2012. The influence of particle shape on the Sawicki, A., Swidzinski, W., 1989. Mechanics of a sandy sub-soil subjected
(centrifuge) cone penetration test (CPT) end resistance in uniformly to cyclic loadings. Int. J. Num. Anal. Method. Geomech. 13, 511–529.
graded granular soils. Géotechnique 62 (11), 973–984. Sezer, A., Altun, S., Göktepe, B.A., 2011. Relationships between shape
López-Querol, S., Coop, M.R., 2012. Drained cyclic behaviour of loose characteristics and shear strength of sands. Soils Found. 51 (5), 857–
Dogs Bay sand. Géotechnique 62 (4), 281–289. 871.
Luong, M. P., 1982. Mechanical aspects and thermal effects of cohesion- Shahu, J.T., Yudhbir, 1998. Model tests on sands with different angularity
less soils under cyclic and transient loading. In: Proc. IUTAM Conf. and mineralogy. Soils Found. 38 (4), 151–158.
on Deformation and Failure of Granular materials, Delft, pp. 239– Shenton, M.J., 1978. Deformation of railway ballast under repeated
246. loading conditions. In: Railroad Track Mechanics and Technology,
Mair, K., Frye, K.M., Marone, C., 2002. Influence of grain characteristics Pergamon Press, pp. 405–425.
on the friction of granular shear zones. J. Geophys. Res. 107 (B10), Shin, H., Santamarina, J.C., 2013. Role of particle angularity on the
ECV-4-1–ECV-4-9. mechanical behavior of granular mixtures. J. Geotech. Geoenviron.
Marr, W.A., Christian, J.T., 1981. Permanent displacements due to cyclic Eng. 139 (2), 353–355.
wave loading. J. Geotech. Eng. Div. 107 (GT8), 1129–1149. Silver, M.L., Seed, H.B., 1971a. Deformation characteristics of sands
Matsuoka, H., Nakai, T., 1982. A new failure criterion for soils in three- under cyclic loading. J. Soil Mech. Found. Div. 97 (SM8), 1081–1098.
dimensional stresses. In: Proc. IUTAM Conf. on Deformation and Silver, M.L., Seed, H.B., 1971b. Volume changes in sands during cyclic
Failure of Granular materials, Delft, pp. 253–263. loading. J. Soil Mech. Found. Div. 97 (SM9), 1171–1182.
McDowell, G.R., Bolton, M.D., 1998. On the micromechanics of Suiker, A.S.J., Selig, E.T., Frenkel, R., 2005. Static and cyclic triaxial
crushable aggregates. Géotechnique 48 (5), 667–679. testing of ballast and subballast. J. Geotech. Geoenviron. Eng. 131 (6),
Miura, K., Maeda, K., Fureukawa, M., Toki, S., 1997. Physical 771–782.
characteristics of sands with different primary properties. Soils Found. Sukumaran, B., Ashmawy, A.K., 2001. Quantitative characterisation of
37 (3), 53–64. the geometry of discrete particles. Géotechnique 51 (7), 619–627.
Miura, K., Maeda, K., Furukawa, M., Toki, S., 1998. Mechanical Tatsuoka, F., 2001. Impacts on geotechnical engineering of several recent
characteristics of sands with different primary properties. Soils Found. findings from laboratory stress-strain tests on geomaterials. In:
38 (4), 159–172. Correia, A.G., Brandl, H. (Eds.), Geotechnics for roads, rail tracks
Nakata, Y., Hyde, A.F.L., Hyodo, M., Murata, H., 1999. A probabilistic and earth structures. Balkema, A.A.
approach to sand particle crushing in the triaxial test. Géotechnique 49 Tatsuoka, F., Maeda, S., Ochi, K., Fujii, S., 1986a. Prediction of cyclic
(5), 567–583. undrained strength of sand subjected to irregular loadings. Soils
Nakata, Y., Hyodo, M., Hyde, A.F.L., Kato, Y., Murata, H., 2001. Found. 26 (2), 73–89.
Microscopic particle crushing of sand subjected to high pressure one Tatsuoka, F., Toki, S., Miura, S., Kato, H., Okamoto, M., Yamada, S.-I.,
dimensional compression. Soils Found. 41 (1), 69–82. Yasuda, S., Tanizawa, F., 1986b. Some factors affecting cyclic
Niemunis, A., Wichtmann, T., Triantafyllidis, Th., 2005. A high-cycle undrained triaxial strength of sand. Soils Found. 26 (3), 99–116.
accumulation model for sand. Comput. Geotech. 32 (4), 245–263. Timmerman, D.H., Wu, T.H., 1969. Behavior of dry sands under cyclic
Oda, M., 1981. Anisotropic strength of cohesionless sands. J. Geotech. loading. J. Soil Mech. Found. Div. 95 (SM4), 1097–1112.
Eng. Div. 107 (9), 1219–1231. Tsomokos, A., Georgiannou, V.N., 2010. Effect of grain shape and
Osinov, V.A., 2013. Application of a high-cycle accumulation model to the angularity on the undrained response of fine sands. Can. Geotech. J. 47
analysis of soil liquefaction around a vibrating pile toe. Acta (5), 539–551.
Geotechnica. 8 (6), 675–684. Vaid, Y.P., Chern, J.C., Tumi, H., 1985. Confining pressure, grain
Peacock, W.H., Seed, H.B., 1968. Sand liquefaction under cyclic loading angularity and liquefaction. J. Geotech. Eng. 111 (10), 1229–1235.
simple shear conditions. J. Soil Mech. Found. Div. 94 (SM3), 689– Wei, L.M., Yang, J., 2014. On the role of grain shape in static liquefaction
708. of sand-fines mixtures. Géotechnique 64 (9), 740–745.
Wichtmann, T., 2005. Explicit accumulation model for non-cohesive soils Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2015. Improved
under cyclic loading. PhD thesis. Publications of the Institute of Soil simplified calibration procedure for a high-cycle accumulation model.
Mechanics and Foundation Engineering. Ruhr-University Bochum, Soil Dyn. Earthq. Eng. 70 (3), 118–132.
Issue No. 38. Wichtmann, T., Triantafyllidis, Th., 2005. Über den Einfluss der Korn-
Wichtmann, T., 2016. Soil behaviour under cyclic loading – experimental verteilungskurve auf das dynamische und das kumulative Verhalten
observations, constitutive description and applications. Habilitation nichtbindiger Böden. Bautechnik 82 (6), 378–386.
thesis. Publications of the Institute of Soil Mechanics and Rock Wichtmann, T., Triantafyllidis, Th., 2015. Inspection of a high-cycle
Mechanics, Karlsruhe Institute of Technology, Issue No. 181. accumulation model for large numbers of cycles (N = 2 million). Soil
Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2005. Strain accumu- Dyn. Earthq. Eng. 75, 199–210.
lation in sand due to cyclic loading: drained triaxial tests. Soil Dyn. Wong, R.T., Seed, H.B., Chan, C.K., 1975. Cyclic loading liquefaction of
Earthq. Eng. 25 (12), 967–979. gravelly soils. J. Geotech. Eng. Div. 101 (GT6), 571–583.
Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2006. Experimental Xu, Y.F., Sun, D.A., 2005. Correlation of surface fractal dimension with
evidence of a unique flow rule of non-cohesive soils under high-cyclic frictional angle at critical state of sands. Géotechnique 55 (9), 691–695.
loading. Acta Geotechnica. 1 (1), 59–73. Yasuda, S., Soga, M., 1984. Effects of frequency on undrained strength of
Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2007a. On the sands (in Japanese). In Proc. 19th. Nat. Conf. Soil Mech. Found. Eng.,
influence of the polarization and the shape of the strain loop on strain 549–550
accumulation in sand under high-cyclic loading. Soil Dyn. Earthq. Yee, E., Duku, P.M., Stewart, J.P., 2014. Cyclic volumetric strain
Eng. 27 (1), 14–28. behavior of sands with fines of low plasticity. J. Geotech. Geoenviron.
Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2007b. Strain accu- Eng. 140, 1–10.
mulation in sand due to cyclic loading: drained cyclic tests with triaxial Yoshimi, Y., Oh-Oka, H., 1975. Influence of degree of shear stress reversal
extension. Soil Dyn. Earthq. Eng. 27 (1), 42–48. on the liquefaction potential of saturated sand. Soils Found. 15 (3),
Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2009. Validation and 27–40.
calibration of a high-cycle accumulation model based on cyclic triaxial Youd, T.L., 1972. Compaction of sands by repeated shear straining. J. Soil
tests on eight sands. Soils Found. 49 (5), 711–728. Mech. Found. Div. 98 (SM7), 709–725.
Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2010a. On the Zachert, H., 2015. On the serviceability of foundations for offshore wind
determination of a set of material constants for a high-cycle accumu- turbines (in German). PhD thesis. Publications of the Institute of Soil
lation model for non-cohesive soils. Int. J. Num. Anal. Method. Mechanics and Rock Mechanics. Karlsruhe Institute of Technology,
Geomech. 34 (4), 409–440. Issue No. 180.
Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2010b. On the ‘‘elastic” Zachert, H., Wichtmann, T., Kudella, P., Triantafyllidis, T., Hartwig, U.,
stiffness in a high-cycle accumulation model for sand: a comparison of 2014. Validation of a high cycle accumulation model via FE-
drained and undrained cyclic triaxial tests. Can. Geotech. J. 47 (7), simulations of a full-scale test on a gravity base foundation for
791–805. offshore wind turbines. In International Wind Engineering Conference,
Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2010c. Towards the FE IWEC 2014, Hannover.
prediction of permanent deformations of offshore wind power plant Zachert, H., Wichtmann, T., Triantafyllidis, T., Hartwig, U., 2015.
foundations using a high-cycle accumulation model. In: International Simulation of a full-scale test on a Gravity Base Foundation for
Symposium: Frontiers in Offshore Geotechnics, Perth, Australia, pp. Offshore Wind Turbines using a High Cycle Accumulation Model. In:
635–640. 3rd International Symposium on Frontiers in Offshore Geotechnics
Wichtmann, T., Rondón, H.A., Niemunis, A., Triantafyllidis, Th., (ISFOG), Oslo.
Lizcano, A., 2010d. Prediction of permanent deformations in pave- Zachert, H., Wichtmann, T., Triantafyllidis, T, 2016. Soil structure
ments using a high-cycle accumulation model. J. Geotech. Geoenviron. interaction of foundations for offshore wind turbines. In: 26th
Eng. 136 (5), 728–740. International Ocean and Polar Engineering Conference (ISOPE-
Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2013. On the ‘‘elastic 2016), Rhodos.
stiffness” in a high-cycle accumulation model – continued investiga- Zhuang, L., Nakata, Y., Kim, U.-G., Kim, D., 2014. Influence of relative
tions. Can. Geotech. J. 50 (12), 1260–1272. density, particle shape, and stress path on the plane strain compression
Wichtmann, T., Niemunis, A., Triantafyllidis, Th., 2014. Flow rule in a behavior of granular materials. Acta Geotechnica. 9 (2), 241–255.
high-cycle accumulation model backed by cyclic test data of 22 sands.
Acta Geotechnica. 9 (4), 695–709.

1 s2.0 S0038080618302099 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0038080618302099 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0038080618302099 Main

Uploaded by

Copyright:

Available Formats

HOSTED BY Available online at www.sciencedirect.

On the inﬂuence of grain shape on the cumulative deformations in

1. Introduction struction processes (e.g. vibration of sheet piles) or

Silt Sand Gravel AGrain

0 largest dimension in the orthogonal direction (see schemes

0 Glass beads Legend

80 Natural sand for c)-f) 1.6

Fig. 7. Photos of sample preparation by air pluviation.

f = qampl/pav = 0.3 were the same as in the reference test.

a) 0.8 Crushed sand b) c) 0.8

Acc. dev. strain εacc

Acc. dev. strain εacc

0.4 qampl [kPa] = 0.4 ID0 = 0.4 pav [kPa] =

Acc. dev. strain εacc

Acc. dev. strain εacc

Fig. 12. eacc acc

Acc. strain εacc [%]

Acc. strain εacc [%]

Acc. strain εacc [%]

Acc. strain εacc [%]

Acc. strain εacc [%]

Acc. strain εacc [%]

200 I = 0.56 200 200 ID0 = 0.59

0.75 I = 0.56 1.6 1.00 I = 0.58 1.00

εacc(N = 105) [%]

εacc(N = 105) [%]

Strain amplitude εampl [10-4]

Strain amplitude εampl [10-4]

a) 5 legend for b) 10 c) 2.5

εacc / fampl [%]

0.12 0.12 0.15

0.08 0.08 0.10

0.04 0.04 0.05

εacc / (fampl fe) [%]

εacc / (fampl fe) [%]

εacc / (fampl fe) [%]

εacc / (fampl fe) [%]

Table 3 tested in the present study. Grain crushing was quantiﬁed

Legend for c)-f):

Peak friction angle ϕP [°]

Volumetric strain εv [%]

0.8 ηav = 0.75 ηav = 0.75 2.0 ηav = 0.75

Acc. dev. strain εacc

Appendix A. Equations of the HCA model e_ acc ¼ f ampl f_ N f e f p f Y f p ð8Þ

You might also like