Professional Documents
Culture Documents
1-s2.0-S0032591015302357-main
1-s2.0-S0032591015302357-main
Powder Technology
a r t i c l e i n f o a b s t r a c t
Article history: We present uniaxial (oedometric) compression tests on two cohesive industrially relevant granular materials
Received 8 September 2014 (cocoa and limestone powder). A comprehensive set of experiments is performed using two devices – the FT4
Received in revised form 17 November 2015 Powder Rheometer and the custom made lambdameter – in order to investigate the dependence of the powders'
Accepted 20 December 2015
behavior on the measurement cell geometries, stress level, relaxation time and applied strain rate. The aspect
Available online 30 December 2015
ratio α, tested with the FT4, is found to play no role for vessels with α ≲ 1 while material characteristics strongly
Keywords:
affect the stress–strain response. After compression is stopped, the constant volume stress relaxation is found to
Cohesive powders follow a power law, consistently for both cohesive powders and for the different testing equipments. A simple
FT4 rheometer (incremental, algebraic) stress evolution model is proposed to describe the relaxation of cohesive powders,
Lambdameter which includes a response timescale along with a second, dimensionless relaxation parameter that sets the
Uniaxial compression very small power law, i.e. extremely slow stress relaxation.
Relaxation theory © 2016 Elsevier B.V. All rights reserved.
Aspect ratio
1. Introduction and background restricted to specific applications, e.g., to the testing of geophysically rel-
evant materials at relatively high consolidation stress.
Granular materials are omnipresent in nature and widely used in The testing and characterization of dry, non-sticky powders are well
various industries ranging from food, pharmaceutical, agriculture and established. For example, rotating drum experiments to determine the
mining — among others. In many granular systems interesting phenom- dynamic angle of repose have been studied extensively as a means to
ena like dilatancy, anisotropy, shear-band localization [49], history- characterize non-cohesive powders [39,6,8], even though these tests
dependence [52], jamming and yield have attracted significant scientific are not well defined with respect to the powder stress and strain condi-
interest over the past decade [1,19,32]. The bulk behavior of these ma- tions. The main challenge comes when the powders are sticky, cohesive
terials depends on their constituents (particles) interacting through and less flowable like those relevant in the food industry [21]. For these
contact forces [42,50]. To understand their deformation behavior, vari- powders, dynamic tests at low stress, like in a rotating drum, are diffi-
ous laboratory element tests can be performed [47,35]. Element tests cult to perform due to contact adhesion and clump formation. In order
are (ideally homogeneous) macroscopic tests in which one can control to overcome this problem, it is preferable to perform confined tests at
the stress and/or strain path. Such macroscopic experiments are higher consolidation stresses, where the stress and strain paths can be
important ingredients in developing and calibrating constitutive rela- well defined and better controlled, so that reproducibility is enhanced.
tions and they complement numerical investigations [32]. Element One element test which can easily be realized (experimentally
test experiments on packings of bulk solids have been studied experi- and numerically) is the uniaxial (or oedometric) compression (in a cy-
mentally in the biaxial box [37,40,36] along with uniaxial compression lindrical or box geometry) involving deformation of a bulk sample in
and volume conserving shear [43], [58,38]. A detailed review of different the axial direction, while the lateral boundaries of the system are
testers has been presented in [47]. Additionally, element tests carried fixed. This test is particularly suited for determining the poroelastic
out with more complex, non-commercial testers have been reported properties of granular materials [18,19,20,4]. While most uniaxial tests
in the literature [15,23,17,5], even though their utilizations are on dry bulk solids have been devoted to studying the relationship be-
tween pressure and density and the bulk long time consolidation be-
⁎ Corresponding author. havior, the dynamics of the time-dependent phenomena have been
E-mail address: o.i.imole@utwente.nl (O.I. Imole). often neglected in experimental and practical applications [59]. For
http://dx.doi.org/10.1016/j.powtec.2015.12.023
0032-5910/© 2016 Elsevier B.V. All rights reserved.
O.I. Imole et al. / Powder Technology 293 (2016) 82–93 83
example, in standard shear testers like the Jenike shear tester [26] and
the Schulze ring shear tester [45], during yield stress measurements,
the focus is usually not on the relaxation behavior. Considerable
stress-relaxation of bulk materials can even disturb yield stress
measurements.
The stress relaxation phenomenon is similar to those observed in
viscoelastic materials such as polymers [13,14,34], gels [9,57], in dielec-
tric relaxation [24,25] and in the attenuation of seismic waves [27]. For
viscoelastic materials, relaxation implies a memory effect and can be de-
scribed using convolution integrals transformed to their fractional form
and containing a relaxation modulus that describes the response of the
system to stress [44]. For these materials, phenomenological models in-
volving the combination of springs and dashpots, such as the Maxwell,
Zener, anti-Zener, Kelvin -Voigt, and the Poynting -Thomson models
have been developed (see Refs. [2,3,28,33] and references therein).
Even though stress relaxation has also been observed in granular
media [4,41,45,59], not much work has been done in providing a theo-
retical description of this phenomenon. Additionally, most cohesive
contact models [32,53,54,56] used in discrete element simulation [21]
of granular materials do not account for the time dependent relaxation
behavior. For the improvement of both discrete element contact models
and constitutive macromodels for cohesive powders, it is necessary to
have an experimental and theoretical understanding of the long-term
stress response of cohesive materials.
In the present study, using two simple devices, we perform
oedometric compression tests with the main goal of investigating the
relaxation of industrial powders at different stress levels. Another goal
is to provide a quantitative comparison between the relaxation behavior
as observed in two testers, namely the lambdameter [30,31,29] and the
FT4 Powder Rheometer [12], in order to confirm that relaxation occurs
irrespective of the device used. The lambdameter has the peculiar ad-
vantage that both vertical and horizontal stresses can be obtained si-
multaneously — unlike the FT4 Powder Rheometer and other simpler
oedometric setups. Finally, we will propose a simple model that cap-
tures the stress relaxation of cohesive powders at different compaction
levels.
The work is structured as follows: in Section 2, we provide a charac-
terization of the material sample, and in Section 3 the description of the
experimental devices and the test protocols. In Section 4, we present the
theoretical model for stress relaxation. Section 5 is devoted to the dis-
cussion of experimental and theoretical results, while the conclusions Fig. 1. Scanning electron microscope images of the cohesive samples (a) cocoa powder
and outlook are presented in Section 6. with 12% fat content (b) Eskal 500 limestone powder. Note the different scales at the
bottom right.
European project, c.f. www.pardem.eu [51]. Scanning Electron Micro- Property Unit Cocoa (12%) Eskal 500
scope (SEM) images obtained using a Hitachi TM 1000 Instrument limestone
(Hitachi Ltd., Japan) for both powders are displayed in Fig. 1 and prop- Size distribution D10 μm 2.14 1.34
erties are displayed in Table 1. D50 μm 9.01 4.37
The particle size distributions are measured using the Helos testing D90 μm 37.40 8.24
Particle density ρp kg/m3 1509 2710
instrument (Sympatec GmbH, Germany). While limestone powder is
Moisture content w % 5.68% 0.25%
dispersed with air pressure, we use the wet mode to disperse cocoa Bulk cohesion σc kPa 1.8 at 7.4 kPa 1.3 at 4.6 kPa
powder since it forms agglomerates. For the wet mode, cocoa powder 9.6 at 41.8 kPa 3.3 at 12.7 kPa
is dispersed in dodecane, an oily liquid, in order to retain the fat layer Wall friction (lambdameter) μw [−] 0.25 0.58
while ultrasound (vibration) is applied to stress the dispersion and Initial bulk density ρ0 kg/m3 388 (L) 717 (L)
450 (F), 467 (f) 855 (f)
break off the agglomerates. The particle density is measured by helium
84 O.I. Imole et al. / Powder Technology 293 (2016) 82–93
shear experiments with a Schulze ring shear tester (RST-01.pc by 3. Experimental setup
Dietmar Schulze Schüttgutmesstechnik, Germany).
For completeness, to obtain the coefficient of wall friction, tests with In this section, we describe the FT4 powder rheometer and lamb-
a Jenike shear cell at a maximum normal stress of 8 kPa and a bottom dameter along with the protocols used in performing the tests.
plate made of the same material as that of the lambdameter were car-
ried out. The coefficient of wall friction for limestone was found to be 3.1. FT4 powder rheometer
about twice as high as that of cocoa. A summary of sample characteris-
tics is provided in Table 1 and a more specific description of the exper- The first experimental equipment used in this work is the FT4 pow-
imental samples is discussed in the following Section. der rheometer (Freeman Technology Ltd., UK), illustrated in Fig. 2(a).
Standard accessories for the compressibility test include the 50 mm di-
2.1. Cocoa powder ameter blade for conditioning, the vented piston for compression and
the 50 mm height by 50 mm diameter borosilicate test vessel. One ad-
One cohesive sample used in this work is cocoa powder with 12% fat vantage of the commercial FT4 Powder Rheometer is the automated na-
content — which is the material used as basic ingredient in the produc- ture of the test procedure requiring minimal operator intervention.
tion of chocolate and related beverages. The material properties includ- The compression test sequence is as follows: the sample is carefully
ing size distribution, particle density and moisture content are shown in filled into the test vessel using a spatula after the tare weight of the ves-
Table 1 along with a scanning electron microscope visualization of its sel has been obtained. The weight of the powder is measured and the
morphology in Fig. 1(a). We note that even though the powder is rela- conditioning cycle is initiated. The conditioning procedure involves
tively hygroscopic, its moisture content does not change significantly the movement of the conditioning blade into the test sample to gently
during the experiments. Nevertheless, the set of experiments are per- disturb the powder bed for a user pre-defined number of cycles before
formed as fast as possible over a relatively short period under ambient it is removed slowly where a cycle consists of the inward and outward
conditions and samples are sealed in air-tight bags when not in use to movement of the conditioning blade into the powder bed. To prevent
minimize effects that could arise due to changes in the moisture content the conditioning blade from touching the base of the vessel, the direc-
of the product. tion of the blade movement is reversed as soon as it is within 1 mm of
the vessel base. It is believed that this action creates a uniform, loosely
2.2. Eskal 500 limestone packed test sample that can be readily reproduced [12]. In this study,
we allow three pre-conditioning cycles before the uniaxial compression
The other industrial powder used in this work is Eskal 500 limestone tests are carried out. Neither evidence of size segregation in which
(KSL Staubtechnik, Germany). Eskal 500 limestone is a commercially coarser particles are brought to the top of the cylinder, nor fluidization
available powder that has wide applications in architecture, road con- of the cohesive samples due to the action of the conditioning blade
struction, blast furnaces, medicines and cosmetics. It is also considered was observed during the cycles.
a suitable reference material for calibration and standard testing [11, The portion of the base insert of the test vessel has been excluded
29,59]. One advantage of this material over other grades is its inability from the calculation of the vessel height, leading to a maximum vessel
to absorb humidity from air. During long term storage under stress, height of 44 mm instead of 50 mm and an aspect ratio α of 0.87. In
Eskal 500 limestone shows no degradation as confirmed by repeatable order to achieve a lower aspect ratio (α= 0.4) with the FT4, a cylindrical
results from experiments carried out under different conditions. The solid insert with a height of 30 mm was machined and added to the base
material properties and SEM morphology are shown in Table 1 and of the 50 mm vessel, leaving a smaller volume with a height of 20 mm,
Fig. 1(b). containing the sample material. Accordingly, the starting and end posi-
Comparing the physical features of the powders, cocoa powder is tions of the blade during the conditioning cycle are adjusted to accom-
brownish while Eskal is whitish in color. Secondly, while cocoa powder modate the inserts.
contains some 12% fat, Eskal 500 limestone does not. This distinction is Subsequent to pre-conditioning, the vessel assembly is split
important for a understanding of their different relaxation behavior. (and thus leveled) to provide precise volume measurement and the
Fig. 2. (a) The FT4 Powder Rheometer [12] and (b) the lambdameter apparatus used for the experimental tests. (c) A schematic representation of the lambdameter test set-up.
O.I. Imole et al. / Powder Technology 293 (2016) 82–93 85
powder mass is recorded after splitting to compute the bulk density. recorded with a data logger on a computer connected to the experimen-
The blade is also replaced with a vented piston, which incorporates a tal setup.
stainless steel mesh to allow the entrained air in the powder to escape A detailed comparison between the main features of both testers is
uniformly across the surface of the powder bed. shown in Table 2 and a comparison of the stress–strain evolution in
The compression test then begins with the distance traveled by the both testers in Appendix A. We only note that the FT4 is more automat-
piston measured for each applied normal stress. In all cases, the first ed and requires less human intervention compared to the lambdameter.
few lines of the data output of ≈50 Hz rate typically were not included On the other hand, in contrast to the FT4 Powder Rheometer, higher
due to low reliability. A description of the vessel split and leveling pro- axial stresses and, thus volumetric strains can be reached with the
cedure has been discussed in [12]. lambdameter.
Table 3
Table of experimental protocols performed. Note that for experiments with the FT4 Powder Rheometer, the maximum stress reached is 22 kPa. Protocols 1–4 represent a variation of the
relaxation time, while 5–9 are different compression rates. Crosses (x) indicate the device used in performing the experiment.
Protocols Velocity [mm/s] No. of steps Vertical stress [kPa] Relaxation time [s] FT4 Lambdameter
Fig. 4. (a) Relaxation at different stress levels R1–R5 during the uniaxial compression of
cocoa powder in Fig. 3. The subscripts f (solid symbols) and F (open symbols) represent
data for α=0.40 and 0.87, respectively. The symbols represent experimental data while
the solid lines represent the theoretical fit using Eq. (3) with parameters listed in
Table 4. Even though the data output was at 50 Hz, we show only points at intervals of
≈0.2 Hz to allow for a clear visualization of the relaxation process. (b) Dimensionless
parameter C, as displayed in Table 4, plotted as function of stress level. The lines
represent Eq. (4); solid and open symbols represent the σvmax and C values obtained
when kinks are included and excluded, respectively.
stronger relaxation in the test with the higher aspect ratio (open sym-
bols in Fig. 4(a)), with the difference between the two aspect ratios in-
creasing with stress level.
The stress relaxation law proposed in Eq. (3) describes well the re-
laxation for the two aspect ratios at all stress states after the stress in
each state has been normalized by its maximum value σvmax such that
σv⁎ = σv/σvmax has a maximum value of 1. The stress levels reached be-
fore the 300 s relaxation, and the other parameters are displayed in
Table 4.1 Note that the fitting ranges used for the different stress levels
are chosen in such a way that discontinuities in the experimental data
(jumps/kinks) are excluded leading to higher quality fits. The ranges
are also different for each stress level. Also, the initial few seconds
were disregarded. When kinks are excluded from the data, the stress
level is changed and the fit parameters obtained are also different (see
open symbols in Fig. 4(b)).
As for the parameters, the timescale t0 determines the inverse initial
slope, i.e., larger t0 corresponds to slower relaxation. The power deter-
mines the longer term relaxation, i.e. the slope after long relaxation.
Comparing the parameters of the model, we observe that the response
time t0 fluctuates, around order of 0.7 s, especially for the higher aspect
ratio, and increases non-linearly from R1–R5 for the lower aspect ratio
(Table 4). On the other hand, the parameter C displayed in Table 4
Fig. 3. Comparison of the vertical (axial) stress plotted against (a) time (b) volumetric and plotted in Fig. 4(b) shows a mostly decreasing trend for both α,
strain (c) bulk density for different aspect ratio. (d) Bulk density plotted against the
volumetric strain. Experiments are carried out using cocoa powder and the FT4 with
carriage velocity 0.05 mm/s. R1–R5 represent the intermediate relaxations during
loading. The dashed blue line (α = 0.4 ⁎) in (c) and (d) above represents a projection of
the α = 0.4 data on the α = 0.87 data. 1
Note that the second-last test R4f had experienced a very strong kink, which possibly
affects the following R5f.
88 O.I. Imole et al. / Powder Technology 293 (2016) 82–93
Table 4 In Fig. 5(a), we show the time evolution of stress during com-
Fit parameters for the analytical predictions of the relaxation model Eq. (3) for relaxation pression for cocoa powder and Eskal 500 limestone. Both powders
at different stress levels σmax
v . The subscripts F and f represent data for aspect ratio α=0.87
and 0.40, respectively while R1–R5 are the relaxation steps. Experiments are carried out
show qualitatively identical relaxation behavior under applied stress.
with the FT4 using cocoa powder as test sample. Comparing the stress–strain response for both materials, as presented
in Fig. 5(b), a similar response for both materials within the small
σFmax (kPa) t0F (s) CF σfmax (kPa) t0f (s) Cf
strain region (εvol b 0.15) is observed. However, at larger strains the
Step α = 0.87 α = 0.40 response differs and limestone responds softer to strain, evidenced
R1 4.75 0.719 0.044 4.75 0.203 0.034 by the slower increase in vertical stress. Secondly, we confirm that
R2 9.5 1.043 0.036 9.51 0.463 0.027 the stress relaxation at R1–R5 occurs at constant strain as shown by
R3 14.25 1.010 0.028 14.26 0.518 0.020 the vertical drops along the deformation strain path. For the same inter-
R4 19.01 0.547 0.020 19.01 1.508 0.022
R5 20.9 0.636 0.019 20.93 13.122 0.019
mediate stress, the onset of relaxation occurs at a higher strain in lime-
Error [%] – 0–0.5 0–1 – 0–0.9 0–0.8 stone as in cocoa. These differences are explained by the finer particle
size of limestone resulting in much more contact points, relatively
higher van der Waals forces between the particles, and higher air
the values being higher for α = 0.87. The decrease of C with stress is ob- entrapment.
served to follow the relation: In Fig. 5(c), the vertical stress is plotted against the respective bulk
densities (ρb) for both materials. The initial bulk densities for Eskal
β
C∝C σ v =σ vmax ð4Þ 500 and cocoa are approximately 717 kg/m3 and 388 kg/m3, respective-
ly. The stress response to increasing bulk density of both materials is dif-
where C ⁎ = 0.06, β = 0.37 for α = 0.4 and C⁎ = 0.10, β = 0.52 for α = ferent as shown by the projection of the cocoa data on the Eskal 500
0.87. The vertical stress σ0v is set to a target value of 1 kPa for both aspect data (dashed blue line), as observed in Fig. 5(b).
ratios. In contrast to the aspect ratio described in Section 5.1, when bulk
In summary, both t0 and C are influenced by the stress level and as- density is plotted against strain in Fig. 5(d), the two powders show sig-
pect ratio of the experimental setup and thus also by the history of the nificant differences. This implies that the deviations observed in the
sample. For example, a higher value of C could be expected if a fresh stress–strain response cannot be attributed only to the initial bulk den-
powder sample was compacted to 20 kPa without intermediate relaxa- sities but is also related to the characteristics of the two materials.
tion. This will be discussed further in Section 5.3. In Fig. 6(a), the relaxation phases of the experiments shown in Fig. 5
Eskal (E) and cocoa (C) are extracted and plotted against the relaxation
5.2. Dependence on material characteristics time. At the same stress and using the same driving velocity, cocoa
powder relaxes more and much faster than the Eskal limestone. For ex-
As a second step, in order to compare the response of different ample, the relaxation after 300 s under the lowest compressive stress
cohesive powders, we introduce the second powder (Eskal 500) and (R1) shows a 33% decrease in stress for Eskal compared to a 43% de-
repeat the same protocol as described in Appendix A. For the sake of crease for cocoa. At the highest stress, the respective changes are more
brevity, this comparison is done using only the lambdameter setup, different, i.e. 20% and 35%, as possibly arising from the fat present in
with α = 0.53. the cocoa powder.
Fig. 5. Comparison of the vertical (axial) stress plotted against (a) time (b) volumetric strain (c) bulk density for experiments with cocoa powder and Eskal 500 limestone, (d) bulk density
plotted against volumetric strain. Experiments are carried out using the lambdameter with carriage velocity 0.05 mm/s while R1–R5 represent the intermediate relaxation steps during
loading. The dashed blue line (cocoa ⁎) in (c) and (d) above represents a projection of the cocoa data on the Eskal data.
O.I. Imole et al. / Powder Technology 293 (2016) 82–93 89
Fig. 7. Comparison of the relaxed stress states after different relaxation duration tR, as
given in the inset. Here, only the relative change of stress within the first 300 s of the
relaxation period is considered.
In order to compare the changes in the vertical stress drop due to the
relaxation duration, using the lambdameter, we perform several exper-
iments in which the relaxation time is varied between 300 s and 1800 s
(protocols 1–4 in Table 3). Afterwards, for each achieved stress state,
only the first 300 s of relaxation are chosen for comparison. This is be-
cause the biggest changes in stress occur during this time interval. Fur-
thermore, this allows us to objectively compare the effect of previous
history, where the powders with larger tR, previously could relax longer.
In Fig. 7, we plot the relative stress reduction, κ = 1 - (σv(t300)/σvmax)
as function of the maximum intermediate stress level σmax v for different
previous relaxation times tR. The longer relaxation time at the lowest
Fig. 6. (a) Extract of the 5 relaxation steps R1–R5 for both experimental specimens in
stress of 5 kPa has no effect on the stress reached since all data are
Fig. 5. The subscripts E and C represent data from experiments with Eskal 500 (E) and
cocoa (C), respectively. The symbols represent the experimental data while the solid
taken at t = 300 s, as evidenced by the collapse of the data at κ ≈ 0.32,
lines represent the analytical fits to Eq. (3). (b) Variation of the dimensionless since there is no effect yet of the loading history; this rather confirms
parameter C in Eq. (3) with intermediate maximum stress level σvmax for the two the repeatability of our measurements.
powders. The lines represent Eq. (4) with σv =1 kPa for both powders. For subsequent relaxations at higher stresses (10, 15, 20 and 25 kPa),
differences due to the longer previous waiting times become visible.
Consistently, at all stresses, an increase in relaxation time tR results in
The fit parameters of Eq. (3) are shown in Table 5 for the five relax- a lower relative stress reduction. The effects of previously experienced
ation shown in Fig. 6(a) for limestone and cocoa. The response time t0 longer relaxation kick in at higher stresses, i.e., a longer active previous
and dimensionless parameter C for each powder generally show a de- load reduce the possible relaxation amplitude in the present state.
creasing trend with the maximum stress at which the relaxation is ini- The fit parameters for the different relaxation duration are displayed
tiated. The decreasing trend of both parameters t0 and C is confirmed in Table 6. t0 is found to show a decreasing trend with stress for all
also for Eskal 500, however, the time-scale is orders of magnitude small-
er than cocoa while C is of the same order but 3–5 times smaller, as sum-
marized in Table 5 and plotted in Fig. 6(b). The parameter C also follows
the relation displayed in Eq. (4) with C⁎ =0.086, β = 0.27 for cocoa and Table 6
Fit parameters for the analytical predictions of the relaxation model Eq. (3) for different
C ⁎ = 0.030, β = 0.47 for Eskal 500 limestone.
relaxation duration. R1–R5 are the relaxation steps.
In summary, we conclude that even though both Eskal and cocoa
powder show qualitatively similar relaxation at constant strain at vari- Duration Step σvmax [kPa] t0 [s] C
−4
ous stress levels, their responses are quantitatively very dissimilar. The 300 s R1 5.00 8.01 × 10 2.99 × 10−2
surprising fact that the parameters for cocoa are also dissimilar for the R2 10.02 5.04 × 10−4 2.28 × 10−2
R3 15.06 2.46 × 10−4 1.80 × 10−2
two devices will be addressed in Appendix A.
R4 20.00 1.99 × 10−4 1.58 × 10−2
R5 25.03 3.59 × 10−4 1.61 × 10−2
600 s R1 5.00 2.11 × 10−3 3.28 × 10−2
R2 10.02 1.01 × 10−3 2.39 × 10−2
R3 15.03 3.42 × 10−4 1.91 × 10−2
Table 5 R4 20.12 1.89 × 10−4 1.59 × 10−2
Fit parameters for the analytical predictions of the relaxation model Eq. (3). The subscripts R5 24.94 1.94 × 10−4 1.55 × 10−2
C and E represent data for cocoa and Eskal, respectively, while R1–R5 are the relaxation 1200 s R1 5.00 1.80 × 10−3 3.27 × 10−2
steps. Experiments are carried out with the lambdameter with α=0.53. R2 10.02 3.99 × 10−4 2.15 × 10−2
R3 15.01 2.39 × 10−4 1.77 × 10−2
Step σCmax (kPa) t0C (s) CC σEmax (kPa) t0E (s) CE
R4 19.87 2.81 × 10−4 1.29 × 10−2
R1 4.90 0.5149 0.0869 5.0027 0.00085 0.03010 R5 24.51 5.86 × 10−4 1.49 × 10−2
R2 9.97 0.2420 0.0708 10.0246 0.00041 0.0224 1800 s R1 5.03 5.41 × 10−4 2.97 × 10−2
R3 14.91 0.2698 0.0659 15.0648 0.00021 0.0178 R2 10.01 1.80 × 10−4 2.06 × 10−2
R4 20.04 0.2207 0.0612 20.0004 0.00009 0.0151 R3 14.87 8.57 × 10−5 1.62 × 10−2
R5 25.07 0.1123 0.0538 25.0253 0.00007 0.0144 R4 19.79 6.98 × 10−5 1.31 × 10−2
Error[%] – 0–3 0–0.5 – 0–10 0–0.7 R5 24.58 5.70 × 10−5 1.24 × 10−2
90 O.I. Imole et al. / Powder Technology 293 (2016) 82–93
Fig. 8. Dimensionless parameter C from Eq. (3) plotted against the intermediate maximum
stress level σvmax for different relaxation duration. The lines represent Eq. (4) with σv =
1 kPa for all relaxation durations.
posed to the dense, slow, quasi-static regime for the slowest rate at Rate Step σmax
v [kPa] t0 [s] C
I ≈10-4. 0.01 mm/s R1 4.99 1.76 × 10−2 2.67 × 10−2
In Fig. 9(a), the vertical stress is plotted as function of strain for dif- R2 10.02 8.80 × 10−3 1.97 × 10−2
ferent loading rates on a semi-logarithmic axis. For clarity, we show R3 14.94 6.20 × 10−3 1.61 × 10−2
only points at intervals b 1 Hz. For small strain (εvol b 0.05), we observe R4 19.53 8.20 × 10−3 1.27 × 10−2
R5 24.98 6.50 × 10−3 1.39 × 10−2
much higher stress develops at faster loading, while the strains needed
0.3 mm/s R1 4.41 2.77 × 10−5 3.40 × 10−2
to reach the first stress level σmax v = 5 kPa decrease. However, for all R2 10.00 3.60 × 10−6 2.68 × 10−2
loading rates, about the same strain is reached at the highest stress R3 15.04 5.02 × 10−6 2.22 × 10−2
level. This suggests that once the relaxation process is fully underway, R4 19.99 4.68 × 10−6 1.89 × 10−2
at the highest stress levels, and the air/fluid entrapped in the pores in R5 22.57 5.44 × 10−3 1.75 × 10−2
0.7 mm/s R1 5.01 3.07 × 10−10 4.77 × 10−2
and between the particles is released, they approach identical state re- R2 9.97 1.78 × 10−10 2.95 × 10−2
gardless of the loading rate employed. R3 15.00 1.17 × 10−9 2.25 × 10−2
R4 20.00 7.97 × 10−8 1.95 × 10−2
R5 24.54 5.54 × 10−7 1.72 × 10−2
1.0 mm/s R1 5.02 1.32 × 10−12 4.86 × 10−2
Table 7 R2 10.01 7.95 × 10−12 3.32 × 10−2
Fit parameters of the analytical prediction of C with stress Eq. (4) for different relaxa- R3 14.92 1.80 × 10−11 2.19 × 10−2
tion duration shown in Fig. 8. R4 20.02 3.25 × 10−7 2.30 × 10−2
R5 24.13 1.14 × 10−6 1.75 × 10−2
Duration [s] β C⁎
1.3 mm/s R1 4.65 9.33 × 10−13 6.19 × 10−2
300 0.43 0.06 R2 10.10 3.93 × 10−12 3.52 × 10−2
600 0.49 0.07 R3 15.02 3.41 × 10−9 2.74 × 10−2
1200 0.57 0.08 R4 18.77 5.51 × 10−8 1.85 × 10−2
1800 0.57 0.07 R5 24.23 6.10 × 10−8 1.51 × 10−2
O.I. Imole et al. / Powder Technology 293 (2016) 82–93 91
stress levels are listed in Table 8, and plotted in Fig. 9(b). For all loading trend with stress for both materials and is higher for cocoa than for
rates, C decreases with increasing stress — consistent with the findings limestone, accounting for the latter (ii).
discussed in Section 5.2 and Appendix A. At a given stress level, C is In terms of the relaxation duration, we find that longer previous re-
found to increase with increasing loading rate. This can be attributed laxation leads to observable differences in relative stress reduction, i.e.
to faster compression which allows for insufficient response and relax- reducing the presently possible relaxation amplitude. Faster loading
ation during stress changes. However, the relative increase in C with rates do not give sufficient time for relaxation leading to differences in
increasing loading rate is found to decrease with increasing stress. At the dimensionless parameter C most visible for relaxation at low stress
the highest stress (25 kPa), except for the rate 0.01 mm/s, all C values levels.
collapse on each other indicating that the final relaxed state is indepen-
dent of the previously experienced rates. This suggests that smaller 7. Outlook
initial relaxation time is needed when the stress becomes higher. Also,
Eq. (4) predicts well the trend of C with stress for all loading rates Future studies will focus on:
with parameters listed in Table 9.
i. the comparison between the two testing devices for exactly iden-
In summary, we find that faster loading rates lead to insufficient
tical aspect ratios to understand the effect of system size,
time for relaxation with differences most visible at lower stress levels,
ii. measurement of the horizontal stress in the lambdameter,
diminishing at higher stress levels. This is true for the Eskal powder
iii. the solution of the model for finite compression rates,
samples that have undergone sufficient (earlier) relaxation steps at
iv. the validation of the relaxation model for other materials, loading
lower stress levels. Other protocols are subject for ongoing research.
rates and geometries,
v. the confirmation of the trend in the dimensionless parameter C
with stress in Eq. (4) which, if confirmed will make Eq. (1) be-
6. Conclusions
come:
We have performed oedometric experiments to study the slow
stress relaxation of cohesive powders under different consolidation ∂ C σ v =σ vmax β
σv ¼ − σ v; ð5Þ
levels, in different aspect ratios and experimental devices, namely the ∂t t0 þ t
custom-built lambdameter and the commercially available FT4 Powder
Rheometer. A comparison of the relaxation behavior of two industrially with a considerable non-linearity β in the range 0.3 ≲ β ≲ 0.8.
relevant cohesive powders, namely cocoa powder with 12% fat content vi. comparison of the findings to relaxation experiments at constant
and Eskal 500 limestone powder, was carried out. stress (i.e. strain creep), and
The relaxation, i.e., the decrease in stress occurring at constant vol- vii. comparison with DEM simulations in order to understand the
ume, is similarly reproduced in the two testing equipments. When possible microscopic origin (e.g. force chains in Section 5.1) of
two aspect ratios are tested with the FT4, differences in the stress– the stress relaxation behavior as observed at the macroscale.
strain and stress-density behavior appear. However, a closer analysis
of the bulk density-strain relation reveals that the differences are only
related to the initial density of the samples. This implies that the aspect Acknowledgment
ratio has no influence on the material behavior for α ≲ 1.
The algebraic model, cf. Eq. (3), captures perfectly the decrease in Helpful discussions with N. Kumar and M. Wojtkowski are gratefully
stress during relaxation at different stress levels for both aspect ratios. appreciated. This work is financially supported by the European Union
There is no clear trend for the response time t0 while the dimensionless funded Marie Curie Initial Training Network, FP7 (ITN-238577), see
material parameter C, is systematically decreasing from low to high http://www.pardem.eu/, and the NWO-STW VICI grant 10828.
stress levels. This non-linear behavior can possibly be attributed to the
accumulated relaxation in the previous stages, but requires further Appendix A. Testing equipments — a comparison
study.
In a macroscopic stress–strain analysis, the two cohesive powders A comparison of testers is necessary for several reasons. Apart from
studied show an identical response to axial loading until about 15% the fact that several literatures have reported on comparative studies
strain, where the differences in response begin to manifest. Eskal 500 between different testers used in the characterization of cohesive pow-
limestone is found to produce a softer response to applied vertical stress ders [23,26,51], most differences observed have been attributable to
in comparison to cocoa powder. A closer comparison shows that the de- human errors, differences in the filling procedure and the measurement
viations in the macroscopic behavior of both powders cannot be attrib- conditions. For our experiments, it is important to confirm that the re-
uted to the initial bulk densities (as was the case for aspect ratio with laxation feature can be reproduced in different testers with different
the FT4) only but it also depends on the material characteristics. At preparation protocols and is not due to drift or bias in our testing equip-
the same stress level, cocoa powder is found to (i) relax much slower ments. The material used for this comparison is cocoa powder in the FT4
but (ii) with a larger amplitude than Eskal. In terms of the parameters vessel with α =0.87 or the lambdameter with α= 0.53.
of the model, the former (i) is due to the fact that the response timescale In order to compare the response of the two testing equipments to
for Eskal, t0E, is several orders of magnitude smaller than that of cocoa. vertical strain, we perform uniaxial compression tests with a carriage
On the other hand, the dimensionless parameter C shows a decreasing speed of 0.05 mm/s (protocol 1 in Table 3). Five intermediate relaxation
stages (R1–R5),2 in which the top piston/punch is held in position for
Table 9 300 s at intervals of 5 kPa during the compression tests are recorded.
Fit parameters of the analytical prediction of C with stress Eq. (4) for different loading
rates shown in Fig. 8.
In Fig. 10(a) we plot the vertical stress σv as function of time. During
loading, the axial stress builds up with time until the first target stress
Rate [mm/s] β C⁎ of 5 kPa (at R1) is reached. For both testing devices, no significant veloc-
0.01 0.46 0.24 ity reduction of the top punch or overshoot of stress is observed as the
0.3 0.42 0.13
0.7 0.65 0.14 2
Due to the difference in the final stress level reached in the FT4 (22 kPa) and
1.0 0.62 0.07
lambdameter (25 kPa), the final relaxation R5 for both equipments should be compared
1.3 0.83 0.06
with caution.
92 O.I. Imole et al. / Powder Technology 293 (2016) 82–93
Fig. 10. Comparison of the vertical (axial) stress plotted against (a) time (b) volumetric strain for experiments with cocoa powder using the FT4 Powder Rheometer and the lambdameter.
The carriage velocity is 0.05 mm/s while R1–R5 represent the intermediate relaxations for increasing target stress.
target stress is approached. However, we observe a slower increase of [8] F. Cantelaube, D. Bideau, Radial segregation in a 2d drum: an experimental analysis,
Europhys. Lett. 30 (3) (1995) 133–138, http://dx.doi.org/10.1209/0295-5075/30/3/
the axial stress in the lambdameter in comparison to the FT4 Powder 002.
Rheometer. Interestingly, the vertical compaction and thus strain at con- [9] F. Chambon, Z.S. Petrovic, W.J. MacKnight, H.H. Winter, Rheology of model polyure-
stant vertical stress are also higher in the lambdameter than in the FT4 thanes at the gel point, Macromolecules 19 (8) (1986) 2146–2149.
[10] J.H. Dieterich, B.D. Kilgore, Direct observation of frictional contacts: new insights for
powder rheometer due to its larger height. This is possibly due to the dif- state-dependent properties, Pure Appl. Geophys. 143 (1–3) (1994) 283–302.
ference in size of the experimental molds and filling procedures (e.g. [11] H.J. Feise, A review of induced anisotropy and steady-state flow in powders, Powder
conditioning of the sample by a rotating blade in case of the FT4 Powder Technol. 980 (3) (1998) 191–200 (URL file:///C:/JP/Documents/Papers/Feise - 1998 -
A review of induced anisotropy and steady-state flow in powders.pdf).
Rheometer) resulting in different densities for both equipments. [12] R. Freeman, Measuring the flow properties of consolidated, conditioned and aerated
With the initiation of the first relaxation R1 at 5 kPa, we observe for powders — a comparative study using a powder rheometer and a rotational shear
both equipments a time-dependent stress relaxation during the rest- cell, Powder Technol. 174 (1–2) (2007) 25–33, http://dx.doi.org/10.1016/j.powtec.
2006.10.016 (ISSN 00325910).
time of 300 s where the amplitude is larger for the lambdameter. Ac-
[13] C. Friedrich, Relaxation and retardation functions of the Maxwell model with frac-
cording to the equation of Janssen [22], due to the larger diameter of tional derivatives, Rheol. Acta 30 (2) (1991) 151–158, http://dx.doi.org/10.1007/
the lambdameter, the stress away from the powder at the lid is expected bf01134604 (ISSN 0035-4511).
to be larger, resulting in higher mean stresses. This observation, along [14] W.G. Gloeckle, T.F. Nonnenmacher, Fractional integral operators and Fox functions
in the theory of viscoelasticity, Macromolecules 24 (24) (1991) 6426–6434,
with other observations reported in literature for other granular mate- http://dx.doi.org/10.1021/ma00024a009.
rials [45,59], confirms that the stress relaxation is not due to a drift in [15] J. Harder, J. Schwedes, The development of a true biaxial shear tester, Part. Part. Syst.
the measuring equipments but it is a material feature as discussed in Charact. 2 (1–4) (1985) 149–153, http://dx.doi.org/10.1002/ppsc.19850020127.
[16] R.R. Hartley, R.P. Behringer, Logarithmic rate dependence of force networks in
Section 5.1. From 5 kPa, we observe an approximate 45% relative de- sheared granular materials, Nature 421 (6926) (2003) 928–931, http://dx.doi.org/
crease in stress for the lambdameter compared to 22% in the FT4 Powder 10.1038/nature01394.
Rheometer. Due to the non-porous lid and the larger diameter of the [17] K.H. Head, Manual of soil laboratory testing: effective stress tests, Manual of Soil
Laboratory Testing, John Wiley & Sons, USA, 1998 (URL http://books.google.nl/
lambdameter, at similar height, the escape of the air trapped and com- books?id=kelRAAAAMAAJ).
pressed in the powder takes more time. With the activation of axial [18] O.I. Imole, N. Kumar, V. Magnanimo, S. Luding, Hydrostatic and shear behavior of
compression after the relaxation, we observe a sharp increase in the frictionless granular assemblies under different deformation conditions, KONA Pow-
der Part. J. 30 (2013) 84–108 (URL http://www.kona.or.jp/search/ab_30_084.html).
axial stress until the next intermediate stress state is reached. [19] O.I. Imole, M. Wojtkowski, V. Magnanimo, S. Luding, Force correlations, anisotropy,
The evolution of stress and strain in both testing equipments is and friction mobilization in granular assemblies under uniaxial deformation, in: A.
shown in Fig. 10(b), where the vertical stress is plotted against volumet- Yu, S. Luding (Eds.),Powders and Grains, AIP Conf. Proc., 1542 2013, pp. 601–604
(URL http://scitation.aip.org/getpdf/servlet/GetPDFServlet?filetype=pdf\&\#38;
ric strain. We observe that the lambdameter initially produces a softer
id=APCPCS001542000001000601000001\&\#38;idtype=cvips\&\#38;doi=10.
response to the applied stress, as evidenced by the slower increase in 1063/1.4812003\&\#38;prog=normal).
the vertical stress during loading. At higher intermediate strain, similar [20] O.I. Imole, M. Wojtkowski, V. Magnanimo, S. Luding, Micro–macro correlations
stress increase with strain is observed in the lambdameter as compared and anisotropy in granular assemblies under uniaxial loading and unloading,
Phys. Rev. E 89 (4) (2014), 042210http://dx.doi.org/10.1103/physreve.89.
to the FT4 Powder Rheometer. The comparison of the response of both 042210 (ISSN 1539-3755).
testing equipments for identical aspect ratio is a subject for future [21] O.I. Imole, D. Krijgsman, T. Weinhart, V. Magnanimo, B.E. Chávez Montes, M.
work and will be presented elsewhere. Ramaioli, S. Luding, Experiments and discrete element simulation of the dosing of
cohesive powders in a simplified geometry, Powder Technol. 287 (January, 2016)
108–120, http://dx.doi.org/10.1016/j.powtec.2015.07.051 (ISSN 00325910).
[22] H.A. Janssen, Versuche über Getreidedruck in Silozellen, Z. Ver. Dtsch. Ing. 39 (35)
References (1895) 1045–1049.
[23] R.J.M. Janssen, H. Zetzener, Measurements on cohesive powder with two biaxial
[1] F. Alonso-Marroquin, H.J. Herrmann, Ratcheting of granular materials, Phys. Rev. shear testers, Chem. Eng. Technol. 26 (2) (2003) 147–151.
Lett. 92 (2004), 054301. [24] A.K. Jonscher, The ‘universal’ dielectric response, Nature 267 (5613) (1977)
[2] T.M. Atanackovic, A modified Zener model of a viscoelastic body, volume 14, 673–679, http://dx.doi.org/10.1038/267673a0.
Springer-Verlag, 2002 137–148, http://dx.doi.org/10.1007/s001610100056. [25] A.K. Jonscher, Dielectric relaxation in solids, J. Phys. D. Appl. Phys. 32 (14) (1999)
[3] R.L. Bagley, P.J. Torvik, On the fractional calculus model of viscoelastic behavior, J. 57–70, http://dx.doi.org/10.1088/0022-3727/32/14/201 (ISSN 0022-3727).
Rheol. 30 (1) (1986) 133–155. [26] S. Kamath, V.M. Puri, H.B. Manbeck, Flow property measurement using the Jenike
[4] M.M. Bandi, M.K. Rivera, F. Krzakala, R.E. Ecke, Fragility and hysteretic creep in fric- cell for wheat flour at various moisture contents and consolidation times, Powder
tional granular jamming, Phys. Rev. E 87 (4) (2013), 042205. Technol. 81 (3) (1994) 293–297 (URL file:///C:/JP/Documents/Papers/Kamath,
[5] J.P. Bardet, Experimental soil mechanics, Prentice-Hall, Upper Saddle River, Puri, Manbeck - 1994 - Flow property measurement using the Jenike cell for
New Jersey, 1997. wheat flour at various moisture contents and consolidation times.pdf).
[6] G. Baumann, T. Scheffler, I.M. Jánosi, D.E. Wolf, Angle of repose in a two-dimensional [27] E. Kjartansson, Constant Q-wave propagation and attenuation, J. Geophys. Res. 84
rotating drum model, in: D.E. Wolf, M. Schreckenberg, A. Bachem (Eds.), Traffic and (B9) (1979) 4737–4748, http://dx.doi.org/10.1029/jb084ib09p04737.
Granular Flow, World Scientific, Singapore 1996, pp. 347–351. [28] M.P. Kruijer, L.L. Warnet, R. Akkerman, Modelling of the viscoelastic behaviour
[7] J. Brujic, P. Wang, C. Song, D.L. Johnson, O. Sindt, H.A. Makse, Granular dynamics in of steel reinforced thermoplastic pipes, Compos. A: Appl. Sci. Manuf. 37 (2)
compaction and stress relaxation, Phys. Rev. Lett. 95 (12) (September, 2005) (2006) 356–367, http://dx.doi.org/10.1016/j.compositesa.2005.04.019 (ISSN
128001, http://dx.doi.org/10.1103/physrevlett.95.128001 (ISSN 0031-9007). 1359835X).
O.I. Imole et al. / Powder Technology 293 (2016) 82–93 93
[29] A. Kwade, D. Schulze, J. Schwedes, Design of silos: direct measurement of stress ratio [45] D. Schulze, Time- and velocity-dependent properties of powders effecting slip–stick
[Auslegung von Silos. Unmittelbare Messung des Horizontallastverhaeltnisses], oscillations, Chem. Eng. Technol. 26 (10) (2003) 1047–1051, http://dx.doi.org/10.
Beton- Stahlbetonbau 89 (3) (1994) 58–63. 1002/ceat.200303112.
[30] A. Kwade, D. Schulze, J. Schwedes, Determination of the stress ratio in uniaxial com- [46] J. Schwedes, Fliessverhalten von Schuettguetern in Bunkern, Verlag Chemie,
pression tests — part 1, Powder Handl. Process. 6 (1) (1994) 61–65. Weinheim, 1968.
[31] A. Kwade, D. Schulze, J. Schwedes, Determination of the stress ratio in uniaxial com- [47] J. Schwedes, Review on testers for measuring flow properties of bulk solids, Granul.
pression tests — part 2, Powder Hand. Process. 6 (2) (1994) 199–203. Matter 5 (1) (2003) 1–43.
[32] S. Luding, Cohesive, frictional powders: contact models for tension, Granul. Matter [48] T. Shinbrot, N.H. Kim, N.N. Thyagu, Electrostatic precursors to granular slip events,
10 (4) (June, 2008) 235–246, http://dx.doi.org/10.1007/s10035-008-0099-x (ISSN Proc. Natl. Acad. Sci. 109 (27) (2012) 10806–10810, http://dx.doi.org/10.1073/
1434-5021). pnas.1121596109 (ISSN 1091-6490).
[33] F. Mainardi, Fractional calculus and waves in linear viscoelasticity, Imperial College [49] A. Singh, V. Magnanimo, K. Saitoh, S. Luding, Effect of cohesion on shear banding in
Press-World Scientific, London, 2010. quasistatic granular materials, Phys. Rev. E 90 (2) (2014), 022202http://dx.doi.org/
[34] R. Metzler, Fractional relaxation processes and fractional rheological models for the 10.1103/PhysRevE.90.022202.
description of a class of viscoelastic materials, Int. J. Plast. 19 (7) (2003) 941–959, [50] A. Singh, V. Magnanimo, S. Luding, A contact model for sticking of adhesive
http://dx.doi.org/10.1016/s0749-6419(02)00087-6 (ISSN 07496419). mesoscopic particles, Powder Technol. (submitted) (2015) 1–30.
[35] G.D.R. MiDi, On dense granular flows, Eur. Phys. J. E: Soft Matter Biol. Phys. 14 (4) [51] S.C. Thakur, O.I. Imole, M.B. Wojtkowski, V. Magnanimo, E.C. Montes, M. Ramaioli, H.
(2004) 341–365, http://dx.doi.org/10.1140/epje/i2003-10153-0. Ahmadian, J.Y. Ooi, Characterization of cohesive powders for bulk handling and
[36] M. Morgeneyer, J. Schwedes, Investigation of powder properties using alternating DEM modelling, in: M. Bischoff, E. Oñate, D.R.J. Owen, E. Ramm, P. Wriggers
strain paths, Task Quarterly 7 (4) (2003) 571–578. (Eds.), III International Conference on Particle-based Methods — Fundamentals
[37] M. Morgeneyer, L. Brendel, Z. Farkas, D. Kadau, D.E. Wolf, J. Schwedes, Can one make and Applications, ICNME 2013, pp. 1–12, http://dx.doi.org/10.1063/1.4812098.
a powder forget its history? Proceedings of the 4th international conference for con- [52] S.C. Thakur, H. Ahmadian, J. Sun, J.Y. Ooi, An experimental and numerical study of
veying and handling of particulate solids, Budapest 2003, pp. 12–118. packing, compression, and caking behaviour of detergent powders, Particuology
[38] P. Philippe, F. Bonnet, F. Nicot, Settlement of a granular material: boundary versus 12 (2014) 2–12, http://dx.doi.org/10.1016/j.partic.2013.06.009 (ISSN 16742001).
volume loading, Granul. Matter 13 (5) (2011) 585–598. [53] J. Tomas, Particle Adhesion Fundamentals and Bulk Powder Consolidation, KONA 18
[39] G.H. Ristow, Dynamics of granular material in a rotating drum, Europhys. Lett. 34 (2000) 157–169.
(4) (1996) 263–268. [54] R. Tykhoniuk, J. Tomas, S. Luding, M. Kappl, L. Heim, H.J. Butt, Ultrafine cohesive
[40] M. Röck, M. Morgeneyer, J. Schwedes, D. Kadau, L. Brendel, D.E. Wolf, Steady state powders: from interparticle contacts to continuum behavior, Chem. Eng. Sci. 62
flow of cohesive and non-cohesive powders, Granul. Matter 10 (4) (2008) (11) (2007) 2843–2864.
285–293, http://dx.doi.org/10.1007/s10035-008-0088-0. [55] P.G. Verzijden, The flow behavior of coffee powder(Masters thesis) TU Delft, 2005.
[41] E. Rojas, M. Trulsson, B. Andreotti, E. Clément, R. Soto, Relaxation processes after in- [56] O.R. Walton, Force models for particle-dynamics simulations of granular materials,
stantaneous shear-rate reversal in a dense granular flow, Europhys. Lett. 109 (2015) in: E. Guazzelli, L. Oger (Eds.), Mobile Particulate Systems, 287, Kluwer Academic
64002, http://dx.doi.org/10.1209/0295-5075/109/64002. Publishers, Dordrecht 1995, pp. 367–380.
[42] S. Roy, A. Singh, S. Luding, T. Weinhart, Micromacro transition and simplified contact [57] H.H. Winter, F. Chambon, Analysis of linear viscoelasticity of a crosslinking polymer
models for wet granular materials, 2015 1–14, http://dx.doi.org/10.1007/s40571- at the gel point, J. Rheol. 30 (1986) 367–382.
015-0061-8. [58] T.S. Yun, T.M. Evans, Evolution of at-rest lateral stress for cemented sands: experi-
[43] M. Saadatfar, A.P. Sheppard, T.J. Senden, A.J. Kabla, Mapping forces in a 3D elastic as- mental and numerical investigation, Granul. Matter 13 (5) (2011) 671–683.
sembly of grains, J. Mech. Phys. Solids 60 (1) (2012) 55–66. [59] H. Zetzener, J. Schwedes, Relaxation and creep of dry bulk solids, Part. Part. Syst.
[44] H. Schiessel, R. Metzler, A. Blumen, T.F. Nonnenmacher, Generalized viscoelastic Charact. 19 (3) (2002) 144–148, http://dx.doi.org/10.1002/1521-4117(200207)19:
models: their fractional equations with solutions, J. Phys. A Math. Gen. 28 (23) 3%25253C144::AID-PPSC144%25253E3.0.CO;2-S.
(1995) 6567–6584, http://dx.doi.org/10.1088/0305-4470/28/23/012.