Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 103 (2015) 237 245

The 13th Hypervelocity Impact Symposium

Strength of Granular Materials in Transient and Steady State Rapid

Shear
Ryan C. Hurleya, Jos E. Andradea*
a

California Institute of Technology, Pasadena, CA 91125, USA

Abstract
This paper discusses relationships between the frictional strength of a flowing granular material and quantities including porosity and
grain-scale energy dissipation. The goal of the paper is to foster an understanding of frictional strength that will facilitate the development
of constitutive laws incorporating important physical processes. This is accomplished in several steps. First, a friction relationship is
derived for a steady state simple shear flow using an energy balance approach. The relationship shows that friction is explicitly related to
porosity, grain connectivity, and grain-scale dissipation rates. Next, the friction relationship is extended to describe transient changes in
frictional behavior. The relationship shows that, in addition to the processes important for steady flows, the rate of dilatation and changes
in internal energy play a role in the frictional strength of a granular material away from steady state. Finally, numerical simulations are
performed to illustrate the accuracy of the friction relationships and illuminate important scaling behavior. The discussion of numerical
simulations focuses on the rate-dependence of frictional strength and the partition of macroscopic energy dissipation into its grain-scale
components. New interpretations of existing constitutive laws and ideas for new constitutive laws are discussed.
2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the Curators of the University of Missouri On behalf of the Missouri University of Science and Technology
Keywords: Granular flow, rheology, friction, strength

Nomenclature
I
d
P

the inertial number, a non-dimensional shear rate equal to

characteristic grain size or grain diameter for spherical grains
confining pressure
grain density
solid fraction
Z
coordination number, average number of contact points per grain
Uc
elastic emerged stored at grain contact
Tp
kinetic energy of grain
Greek symbols
shear rate in simple shear or strain rate tensor in 3D
rate of dilatation
dissipation rate at grain contact
friction coefficient
steady state portion of the friction coefficient
transient portion of the friction coefficient
portion of the friction coefficient furnished by grain viscoelasticity

* Corresponding author. Tel.: 626-395-4141

E-mail address: jandrade@caltech.edu.

1877-7058 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the Curators of the University of Missouri On behalf of the Missouri University of Science and Technology

doi:10.1016/j.proeng.2015.04.032

238

Ryan C. Hurley and Jos E. Andrade / Procedia Engineering 103 (2015) 237 245

portion of the friction coefficient furnished by grain sliding

1. Introduction
The behavior of flowing granular media has long been of interest in the engineering and physics communities.
Researchers in these fields have sought to identify the numerical schemes and material properties required to accurately
model dense and rapid granular flows in various scenarios. Such scenarios include powder transport [1], rapid soil
penetration [2], the flow of brittle ceramics [3], and planetary-scale asteroid impact [4]. A diverse set of modeling
techniques has emerged to address this diverse set of application. For instance, discrete element modeling [5] and its multiphysics variants [6] have been used for applications in which only hundreds or thousands of grains need to be modeled.
Rate-dependent plasticity theories, kinetic gas theories, and Navier-Stokes solvers have been reserved for field-scale
modeling [7]. This latter set of continuum approaches is the only computationally tractable means of modeling many
problems and remains a prime focus of on-going research. In these approaches, material strength is captured by a friction
coefficient. This paper focuses on the friction coefficient, also referred to as frictional strength or simply friction.
This work explores the relationship between friction in granular flows and important physical processes occurring at the
grain-scale and macro-scale. Friction in granular materials is known to be strongly rate-dependent. A prime goal is therefore
to understand how the various grain-scale and macro-scale processes contribute not only to the absolute value of friction but
also to frictions evolution as a function of shear rate. This understanding is important for rapid impact and penetration
scenarios, where material strength may evolve significantly throughout the event. Two expressions relating steady state and
transient friction to physical processes are derived for shear flows. Numerical simulations using the Discrete Element
Method (DEM) [5] are then used to study the scaling behavior of the variables in these relationships. The influence of
distinct grain-scale energy dissipation mechanisms on steady state friction is investigated. The future development of new
constitutive laws incorporating the derived links will be discussed at the end of the paper.
In past work, simple shear experiments and simulations have been used to establish empirical relationships between
steady state friction and shear rate in a granular flow. For instance, in [8] the authors proposed the linear empirical
for flows of 2D grains, where a and b are empirical constants and I is a non-dimensional shear rate
relationship
to be described in section 2.1. In [9] the authors proposed the nonlinear empirical relationship
for flows of 3D grains, where
,
, and
are empirical constants. The latter
relationship has been validated against experiments [9] and used in a rate-dependent constitutive law for field-scale
simulations of planetary asteroid impact [4]. Despite the success of these empirical relations, including friction laws
incorporating physical processes and material properties into constitutive laws remains a challenge.
Some understanding of the physical processes giving rise to frictional rate-dependence has been found in studies on
inertial granular flows. Inertial flows are characterized by a dense packing of grains and an inertial number I
) between 0.001 and 1. The inertial flow regime is characterized by a marked rate-dependence of steady(
state frictional strength and dramatic changes in the internal structure of the material [10]. Most significantly, [8,10] have
shown that the anisotropy of the contact force network between grains changes in the inertial flow regime and is directly
related to frictional shear strength in the granular flow. This finding provides insight into the microstructural response of the
material necessary to resist certain macroscopic loads. In this paper, alternative interpretations of frictional strength and its
dependence on shear rate are provided. The inertial flow regime is the primary focus of this paper since, as defined by the
inertial number, it is relevant to very rapid impact and shear events (strains up to 105 for sub-millimeter sand grains
subjected to pressures in the 10s of MPa).
2. Friction Relationships for Granular Shear Flows
2.1. The Inertial Number
An important dimensionless measure of shear rate in granular flows is the inertial number
(1)
The inertial number can be interpreted to be the ratio of two time scales: a confinement timescale
proportional to the time required to fully compact a grain after it dilates past its neighbor during the shearing process, and a
related to the frequency at which a grain is forced past its neighbors. In simple shear
macroscopic shear timescale
flows, is the shear rate in the direction of flow and P is the confining stress. In general 3D flows, is replaced by , the

Ryan C. Hurley and Jos E. Andrade / Procedia Engineering 103 (2015) 237 245

239

second invariant of the strain rate tensor, and

is the isotropic pressure. The inertial number is used during the
derivation of the two friction relationships, which now follows.
2.2. Friction Relationship for Steady State Shear Flows
A steady state friction relationship can be derived for a granular shear flow (see Fig. 1a) by considering an energy
balance equation
(2)
where T and U are the kinetic and potential energy densities, respectively,  is the velocity gradient tensor, and
is the dissipation rate per unit volume in the material. In steady state simple shear, only one component of the velocity
) and the left hand side of Eq. (1) is zero when both sides are time-averaged, yielding
gradient tensor is non-zero (
(3)
, assuming all dissipation to occur at grain
where the bars indicate a time average. Defining friction to be
contact points, and neglecting correlations in the fluctuations of the terms in Eq. (3), the expression becomes
(4)
where all quantities should be assumed to be averaged unless otherwise specified. The act of neglecting correlations in the
fluctuations of the terms in Eq. (3) is justified by simulation results, which demonstrate correlations to be negligible. In Eq.
is the average per(4), Nc is the number of inter-particle contact points in the system, V is the volume of the system,
contact dissipation rate, and the identity
has been used. Eq. (4) is simplified by noting that: (1)
where Z is the coordination number and Np is the number of particles in the system; (2) the characteristic
grain diameter d in the inertial number can be defined such that
, where is the volume of solids; (3)
where is the solid fraction. These simplifications yield
(5)
where
is a pressure-dependent characteristic work rate, a constant for constant-pressure experiments.
Equation (5) is a new relationship between steady state friction in a granular shear flow, solid fraction, coordination
number, shear rate, and grain-scale dissipation rates. Shearing dilation, a process by which a granular material reaches a less
compact state with a higher solid fraction at higher steady state shear rates, leads to an decrease in coordination number,
in Eq. (5). However, the increasing shear rates that induce shearing dilation also increase
solid fraction, and therefore
the average dissipation rate at a grain contact point
. Thus, rate-dependence of frictional strength is dictated by a
competition between shearing dilation and grain-scale dissipation rates, each of which have opposing effects on friction as
seen in Eq. (5).
2.3. Relationship for Transient Shear Flows
Beginning with the energy balance in Eq. (2), a relationship for the transient evolution of friction can be derived, yielding
(6)
where
is the average stored energy (assumed to be stored as elastic energy at contact points),
is the average kinetic
energy of the particles, and
is the rate of dilatation. The first term in Eq. (6) can be thought of as a steady state
and the second as a transient energy term .
term
Equation (6) is a new relationship for the evolution of friction in non-steady flows. It is similar to Eq. (5), but includes
terms involving rate of dilatation and changes in internal energy of the system. It is seen from this expression that a positive

240

Ryan C. Hurley and Jos E. Andrade / Procedia Engineering 103 (2015) 237 245

rate of change in internal energy or a positive rate of dilatation invoke an increase in the shear resistance of the material.
Equation (6) is similar in form to friction laws proposed in the soil mechanics literature [11,12] which are composed of a
), except that it contains an additional energy term . In the
steady-state portion and a dilatation portion (e.g.
following section, DEM simulations illustrate how the quantities in Eqs. (5) and (6) evolve with shear rate and material
properties and give rise to some of the observed behaviors of frictional strength in granular materials.
3. Numerical Simulations
3.1. Discrete Element Modeling Code
A DEM code is used to study the influence of particle properties and shear rate on the variables in Eqs. (5) and (6). The
code is a modified version of the molecular dynamics code LAMMPS. Grains are modelled as spheres and interact with a
Hertz-Mindlin contact law described in [13]. The contact model ignores effects of grain fracture and melting and can
therefore only be assumed to give an approximation of the true physics of rapidly sheared systems. The simulations
discussed here involve spherical grains with 20% bidispersity in diameter that are sheared with periodic x and y boundary
conditions (see Fig. 1a). The grains are sheared between rough walls composed of grains fixed to move together. The
parameters held constant during all simulations include the confining pressure, which is chosen in conjunction with the
grain stiffness (normal spring constant) to ensure fairly rigid grains as discussed in [8]. The confining pressure is held
constant by moving the rough walls in the y direction as required. The parameters varied throughout the simulations include
the grain damping coefficient, chosen to produce a collision-velocity-dependent coefficient of restitution , and an interthat restricts the ratio of the tangential mechanical force to the normal mechanical force at a
particle friction coefficient
contact before truncation of the tangential force. The three data sets investigated are detailed in Fig. 1c.

Figure 1 (a) Illustration of simple shear flow simulation. (b) Illustration of coefficient of restitution used in the three data sets. (c) Quantities used for
material properties in the three data sets. The description of coefficient of restitution in (c) relates to the curves in (b).

3.2. Steady State Simulations

Steady state simple shear flows are investigated by running simulations until a steady state is reached in the time-averages
of all quantities in Eq. (5). The averaged quantities are used to compute Eq. (5). The result is compared to the equation for
, where the stress components are computed using the coarse-graining expression [14]
friction
(7)
where is a vector pointing from the centroid of particle j to the centroid of i, is a contact force from particle j to i, and
is a dyadic (tensor) product. The resulting comparison for all the three data sets investigated here is shown in Fig. 2a. An
excellent agreement is seen for each data set in Fig. 2a, indicating that Eq. (5) accurately links friction to important
processes such as shearing dilation and grain-scale dissipation processes.
, one furnished by viscoelastic processes with the other
A decomposition of Eq. (5) into two components
be the
furnished by grain-sliding, can be derived by considering the contact law described in section 3.1. Letting
average per-contact energy dissipation rate by viscoelastic processes within the grains and
to be the average per-contact
energy dissipation rate at grains sliding past surrounding grains, the two components in the decomposition can be written as

Ryan C. Hurley and Jos E. Andrade / Procedia Engineering 103 (2015) 237 245

and

241

(8)

An illustration of the evolution of each of these terms is given in Fig. 2b. For all data sets, grain sliding furnishes all of
the granular materials frictional strength at low shear rates. As shear rates increase, the viscoelastic energy dissipation
mechanism contributes increasingly to frictional strength. For DS1, the viscoelastic contribution increases rapidly,
overtaking the grain-sliding contribution at approximately I=0.62. For DS2, which features a higher coefficient of
restitution for the grains, the viscoelastic contribution never overtakes the grain-sliding contribution. This is likely because
less energy is dissipated due to viscoelasticity in each collision, causing higher impact velocities and leading more energy to
be dissipated through grain sliding. For DS3, the viscoelastic contribution and grain sliding contribution are very similar to
those of DS1, except that the viscoelastic contribution overtakes the grain-sliding contribution at a lower shear rate, around
I=0.55. Interestingly, both DS1 and DS3 feature very similar behaviors in Figs. 2a and 2b. The increased inter-particle
friction coefficient in DS3 only appears to shift the total friction curve upward (see Fig. 2a) but does not change its general
curve at a given shear rate is primarily accomplished by varying
shape. Hence, it seems that changing the slope of the
the restitution coefficient of the grains. This finding is supported by other work that finds a transition from rate-dependent
inertial flows to rate-independent gas-like rapid flows to be determined mainly by the coefficient of restitution [15].

Figure 2 (a) Comparison of the friction coefficients computed using Eq. (5) (dashed red curves) with those computed using Eq. (7) (solid black curves
with symbols) for each of the three data sets. The inset shows the same data with a logarithmic scale for shear rate, illustrating the onset of rate-dependence
in the inertial flow regime. (b) Evolution of distinct grain-scale contributions to macroscopic friction for each data set. Error bars are not shown for either
figure to improve readability. However, error bars marking one standard deviation on each friction measurement are typically span less than 0.05 from the
shown data points.

Figure 3 illustrates the evolution of the components in Eq. (5) as a function of shear rate for the three data sets
investigated. The finding indicates that shearing dilation and grain-scale dissipation rates play competing roles in
determining the strength of a sheared granular material. Shearing dilation decreases the internal area over which energy can
be dissipated by decreasing the coordination number Z and the solid fraction . Simultaneously, grain-scale dissipation rates
increase with shear rates due to higher grain collision velocities and inter-particle forces, increasing
.

242

Figure 3 (a) Evolution of the term

and faster than for

Ryan C. Hurley and Jos E. Andrade / Procedia Engineering 103 (2015) 237 245

shows two distinct scaling regimes in the inertial flow regime. (b) Dissipation rate terms scale close to

for

shown in Fig. 3a has two distinct scaling behaviors in the inertial flow regime. All data sets appear to
The term
collapse onto the scaling
at low shear rates, consistent with the fact that and maintain maximum quasiat high shear rates due to a
static values at low shear rates. Similarly, all data sets collapse onto the scaling
decrease in and . The transition between these two behaviors occurs at a shear rate (I=0.1) at which dramatic changes in
internal structure of the material have been observed to occur [10]. The energy dissipation rate terms in Fig. 3b also have a
and faster than for
. The latter scaling can be derived
characteristic scaling behavior close to for
from the fact that collision velocities scale with the inertial number and the contact law requires viscous forces to scale with
and
observed in these simulations gives rise
collision velocities. When combined, the competing scaling of
to the rate-strengthening of the friction coefficient shown in Fig. 2a. This is a common feature of inertial flows seen in the
literature [8-11] and used in modelling of rapid impact events [4]. From the preceding analysis, it is understood to arise
decrease.
because grain-scale dissipation rates increase faster with shear rate than dilation-induced quantities
3.3. Transient Simulations
The transient simple shear flow simulations are performed only with grain properties used in DS1 above. The simulations
are performed as follows. First, a confining pressure is applied to the simple shear flow configuration shown in Fig. 1a. A
slow shear rate (I=0.001) is then applied to the material to allow quasi-static values of and to be reached. Once a
steady-state is reached at the slow shear rate, a faster shear rate (I=0.5) is instantaneously applied to the material by
changing the velocity of the boundaries. The boundaries are allowed to expand in the y direction while the material seeks
vertical stress equilibrium. The boundaries are given mass equal to ten times the sum of the masses of the particles that
compose them. A small global damping parameter is used to prevent the boundary from accelerating upward too quickly or
oscillating while the material seeks vertical stress equilibrium [5]. The mass of the boundaries has a minimal effect on the
rate at which the material reaches a new equilibrium. The damping parameter has an influence on the rate at which the
material reaches a new equilibrium, but is chosen to be the smallest value below which a damping parameter would allow
the boundaries to oscillate.
Variables in Eq. (6) are monitored in a thin layer of grains within the flow to ensure the shear rate is uniform (see Fig
4a). Only one layer is analysed in this work for brevity. Considering the assembly as a whole is not appropriate since the
flow profile evolves according to a momentum-diffusion type process, making the full flow profile non-linear and
invalidating the homogeneous shear rate required in the inertial number. The evolutions of spatially averaged quantities in
Eq. (6) are therefore averaged over fewer than the total 10,000 modelled grains and possess considerably more variability
than in the steady state case. Some variables are fit with a smooth curve before being used in Eq. (6), as discussed in
Appendix A.

243

Ryan C. Hurley and Jos E. Andrade / Procedia Engineering 103 (2015) 237 245

Figure 4 (a) Location of region where variables in Eq. (6) were extracted to compute the transient evolution of the friction coefficient. Grain colors are
the same as those in Fig. 1a. (b) Comparison of the transient friction coefficient in the specified layer computed using the proposed friction law (Eq. (6),
solid blue curve) and the coarse-grained stress formula (Eq. (7), dashed red curve). (c) Comparison of the three terms in Eq. (6). The steady-state term
(solid black line) evolves slowly to the its new value while the rate of dilatation (dotted green curve) provides part of the transient frictional response at
early times while the transient energy term (dot-dashed black curve) is negligible at all times.

Figure 4b shows the evolution of the friction coefficient in the layer of interest computed using the friction relationship in
Eq. (6) and the coarse-graining expression in Eq. (7). The time t=0 corresponds to the time at which the velocity step is
applied to the boundaries. The friction relationship in Eq. (6) is seen to provide a decent estimate of the friction coefficient
as a function of time. The major difference is the slight offset of the curves at early times and the large variability in the
curve computed using Eq. (6).
Figure 4c shows the evolution of the 3 terms in Eq. (6) with time. At early times, the steady state term (first term in Eq.
(6)) decreases due to an increase in the confining pressure in the layer (e.g., an increase in ). The rate of dilatation
(third term in Eq. (6)) partially compensates for this decrease, increasing rapidly before quickly decaying to zero. The
transient energy term (second term in Eq. (6)) is negligible for the entire simulation. Remarkably, this suggests that the
transient evolution of friction in some rapid shear flows is similar in nature to the observed evolution of friction in soil
mechanics [11,12]. More work is needed to determine how this behavior changes in simulations employing other material
properties or shear rates.
4. Discussion and Future Work
The theoretical relationships derived in section 2 provide a link between frictional strength and the physical processes
occurring in inertial granular flows. The relationships also provide insight into the influence of grain properties on frictional
strength. Figure 2 illustrated that grain properties related to the restitution coefficient (e.g. the Youngs modulus and
plasticity law of the bulk grain material) contribute more to shaping frictional rate-dependence than inter-particle friction
coefficient of the grains. Additional simulations are needed to investigate the influence of grain shape on frictional strength
and rate-dependence in granular shear flows. Additional work is also needed to extend Eqs. (5) and (6) to include the effects
of additional physics, such as grain fragmentation and melting, that may influence a granular materials response in certain
rapid shearing and impact events.
A prime future goal of this work is to foster the development of constitutive laws for continuum modelling of rapid shear
flows and impact events. Some preliminary work has been done on incorporating Eqs. (5) and (6) into the constitutive laws
developed in [9]. For instance, using the stress tensor decomposition proposed in [9] and a derivation similar to that for Eq.
(4), the stress tensor in a granular material undergoing a general 3D steady state deformation can be written as
(9)
Equation (9) can be used in continuum numerical simulations with an appropriate equation of state relating P to and
. The influence of grain
equations governing the evolution of coordination number and grain-scale dissipation rates
to create a physics-based constitutive law suitable for
properties can be incorporated into evolution laws for and
field-scale simulations. It may be possible to simulate non-steady flows with Eq. (9) if the response on short time scales like
those seen in Fig. 4c is not needed. This is the subject of on-going research.
The simulations probing the transient behavior of friction reveal that, in certain flows, friction has the simple form
, as in soil mechanics [11,12]. However, more work is needed to understand the influence of shear rates and

244

Ryan C. Hurley and Jos E. Andrade / Procedia Engineering 103 (2015) 237 245

grain properties. The present results merely illustrate the accuracy of Eq. (6). There may be material properties or shear rates
for which the transient energy term is not negligible as it was in the simulations shown in Fig. 4. Considering thermal
effects in simulations may also significantly alter the contribution of each term in the expression.
5. Conclusion
This work examines the physical processes underlying frictional strength in flowing granular media. Such processes are
crucial to understanding and accurately modelling rapid deformation and impact events. Relationships linking steady state
and transient frictional strength to properties such as coordination number, solid fraction, shear rate, rate of dilatation, and
the rate of change in internal energy have been derived. The steady state expression illustrates the competing mechanisms
governing steady state frictional rate-dependence: shearing dilation and grain-scale dissipation rates. The transient friction
relationship illustrates that the evolution of strength during transient loading is influenced explicitly by the rate of dilatation
and by rates of change in the internal energy of the material. Simulations shown here indicate that the rates of change in the
internal energy of the material are negligible for certain grain properties and shear rates.
The intended consequence of the current work is to facilitate the development of constitutive laws that can improve the
physical accuracy and predictive capacity of numerical simulations. On-going and future work will continue to investigate
the friction laws proposed here and their role in constitutive laws.
Acknowledgements
Support from the Air Force Office of Scientific Research Grant #FA9550-12-1-0091 through the University Center of
Excellence in High-Rate Deformation Physics of Heterogeneous Materials is gratefully acknowledged.
References
[1] Jaeger, H. M., Nagel, S. R., Behringer, R. P., 1996. Granular solids, liquids, and gases. Reviews of Modern Physics, 68(4), 1259.
[2] Omidvar, M., Iskander, M., Bless, S., 2014. Response of granular media to rapid penetration. International Journal of Impact Engineering, 66, 60-82.
[3] Luo, H., Chen, W. W., & Rajendran, A. M., 2006. Dynamic Compressive Response of Damaged and Interlocked SiCN Ceramics. Journal of the
American Ceramic Society, 89(1), 266-273.
[4] Jutzi, M., Asphaug, E., 2011. Forming the lunar farside highlands by accretion of a companion moon. Nature, 476(7358), 69-72.
[5] Cundall, P. A., Strack, O. D., 1979. A discrete numerical model for granular assemblies. Geotechnique, 29(1), 47-65.
[6] Tang, Z. P., Horie, Y., Psakhie, S. G., 1996. Discrete meso-element dynamic simulation of shock response of reactive porous solids. In American
Institute of Physics Conference Series (Vol. 370, pp. 657-660).
[7] Savage, S. B., 1984. The mechanics of rapid granular flows. Advances in applied mechanics, 24, 289-366.
[8] da Cruz, F., Emam, S., Prochnow, M., Roux, J. N., Chevoir, F., 2005. Rheophysics of dense granular materials: Discrete simulation of plane shear
flows. Physical Review E, 72(2), 021309.
[9] Jop, P., Forterre, Y., Pouliquen, O., 2006. A constitutive law for dense granular flows. Nature, 441(7094), 727-730.
[10] Azma, E., Radja, F., 2014. Internal structure of inertial granular flows. Physical review letters, 112(7), 078001.
[11] Andrade, J. E., Chen, Q., Le, P. H., Avila, C. F., Matthew Evans, T., 2012. On the rheology of dilative granular media: Bridging solid-and fluid-like
behavior. Journal of the Mechanics and Physics of Solids, 60(6), 1122-1136.
[12] Wood, D. M., 1990. Soil behavior and critical state soil mechanics. Cambridge university press.
[13] http://lammps.sandia.gov/doc/pair_gran.html
[14] Bagi, K., 1996. Stress and strain in granular assemblies. Mechanics of materials, 22(3), 165-177.
[15] Forterre, Y., Pouliquen, O. 2008. Flows of dense granular media. Annu. Rev. Fluid Mech., 40, 1-24.

Appendix A Curve Fitting in Transient Simulations

Smooth curves are fit to some of the variables extracted for analysis of Eq. (6). Fitting is achieved using least squares curves
fits in MATLAB. In particular, a smooth curve is fit to the rate of dilatation term of the form proposed in [11]
(10)
where asterisks indicate fitting parameters. Similarly, a smooth curve of the following form is fit to the quantity
and its derivative was subsequently found analytically
(11)

Ryan C. Hurley and Jos E. Andrade / Procedia Engineering 103 (2015) 237 245

245

These fits are necessary to make the presentation of the results in Figs. 4b and 4c clear. Without the fitting, large oscillations
dominate these figures. General trends in all fitted quantities are not significantly changed in the fitting process.