Powder Technology
a r t i c l e i n f o a b s t r a c t
Article history: The TwoFluid Method is capable of modeling largescale (i.e., lab scale or larger) multiphase (particleuid)
Received 8 December 2016 ows by treating both the uid and particle phase as interpenetrating continua and solving mass and momentum
Received in revised form 25 April 2017 balances for each phase. To solve for the ow of the solids phase the momentum balance requires constitutive
Accepted 5 May 2017
relations in the form of normal and shear stresses i.e., pressures and viscosities. However, the stresses that ac
Available online 16 May 2017
count for frictional contacts in dense particle systems, and are relevant to this work, are empirically based. A
Keywords:
study of the effects of adjusting the frictional model formulation (the empirical parameters of the model), by
Twouid model changing the overall frictional stress magnitude and the relative magnitude of the frictional viscosity to the fric
CFD tional pressure, on the behavior of the bed is presented here. It was found that the magnitude of the frictional vis
Friction cosity relative to the frictional pressure affects the crater growth prediction almost as much as the magnitude of
Crater formation the overall frictional stress. Additionally, a frictional model formulation is validated for sand particles, and predic
tions are compared with existing experimental data for the crater formation of a sand bed under a vertical, im
pinging jet of gas (Metzger et al. J Areo Eng (2008) v22, p2432). In the low jet velocity regime (subsonic,
turbulent jet), the model predicts the salient features previously measured for the growth rate of the crater of
time, the prole of the crater, and the response of the crater to turning the jet off. In the high jet velocity regime
(compressible, near sonic jet ow) the prediction agrees qualitatively with prior experimental observations.
2017 Published by Elsevier B.V.
1. Introduction of different rocket exhaust designs and the particleladen surfaces. Fur
thermore, space exploration is not the only application that would ben
The interaction of rocketexhaust with the dense particle layer that et from a software package that can simulate the gassolid and solid
covers the surfaces of the moon, Mars and some asteroids can result in solid interactions as transporting particles is commonplace in many in
potentially hazardous soil erosion in the form of highspeed particle dustrial operations, such as hoppers, uidized beds, silos, pneumatic
spray or crater formation [1]. During the Apollo missions, the eroded conveying, etc. [3].
particle spray obscured the astronauts' view of the lunar surface [1]. Ad Two main model frameworks are typically used to model multi
ditionally, in an attempt to mitigate the effects of the plume impinge phase (particleuid) systems, namely, Discrete Element Method
ment interacting with the Martian surface, several landing techniques (DEM) and the TwoFluid Method (TFM). In DEM, Newton's equations
were tested for the Martian rovers including the use of an umbilical of motion are solved for every particle in the domain, and therefore
structure that allowed for the rockets to sit 6.5 m above the lander [2]. adjusting the particle properties for each particle is straightforward
Performing experiments in conditions that match a spacecraft landing [4]. However, tracking the positions of every particle in the domain re
on an extraterrestrial surface are technologically and economically pro quires a large computational overhead, and consequently the number
hibitive, which is evident when considering the tests would require r of particles, or size of the system, that can be simulated with DEM is lim
ing a rocket, in vacuum, in variable gravity, onto a bed of extraterrestrial ited to orders of magnitude smaller than realistic processes [5]. In TFM,
particles. Thus, designing spacecraft and surface architecture for future the particle phase and uid (liquid or gas) phase are treated as interpen
space exploration will benet from the ability to predict the interaction etrating, interacting continua, hence the particle phase is treated as a
uid (i.e., individual particles are not considered), and mass and mo
Corresponding author. mentum balances are solved for each phase. The solids phase mass
Email address: clamarche@u.edu (C.Q. LaMarche). and momentum balances are derived by ensemble averaging the
http://dx.doi.org/10.1016/j.powtec.2017.05.008
00325910/ 2017 Published by Elsevier B.V.
C.Q. LaMarche et al. / Powder Technology 318 (2017) 6882 69
equations of motion so larger systems can be considered than with DEM pushes them away [37]. When the stagnation pressure of the jet ex
(since individual particles are not tracked), but constitutive relations are ceeds the bearing capacity of the soil, as with BCF, the soil is shoved
required to solve the averaged equations [69]. downward, producing a shallow crater [36]. In the DDF mechanism,
The constitutive relations solved in TFM in the form of in the granu the uid is pushed by the jet into the void spaces of the soil, unjams
lar phase equation of motion as collisional, kinetic, and frictional stress the material below the surface and shears it [31]. Simulating the crater
es. The collisional and kinetic stresses dominate in dilutesolids ows formation under high and lowvelocity jets could provide insight into
and account for inelastic collisions and particles crossing imaginary the experimentally observed cratering mechanisms by Metzger et al.
shear planes, respectively [10,11]. Additionally, the collisional and ki [31]. Metzger et al. [31] also report measurements of transient cratering
netic stresses are generally derived from rst principles, e.g., derived behavior such as crater growth through the cratering experiment. They
for dilute systems of hardsphere particles [12,13]. At the denser limit present a rigorous parametric study of the effect of the properties of the
of particle ows (high solids volume fractions), particles sustain endur gas jet on the cratering of the particle bed, which offers a set of measure
ing contacts and therefore friction between particles plays a signicant ments that can be compared to TFM predictions of a jet interaction with
role in the behavior of the particle bed [14]. Unlike the stress contribu a particle bed surface [31].
tions that dominate in the dilute limit, the models that account for fric The main aim of this work is to validate the predictions from TFM
tional contacts is empirically based [9]. (Newer frictional models relate using frictional stress models for a gas jet impinging on a bed of particles
the frictional stress to the particleparticle friction coefcient [15], but and to evaluate the sensitivity of the bed behavior to the formulation of
are not the focus of this work since the frictional coefcient of the par the empirical, frictional stress model. The experimental results of
ticles was not known). Additionally, the frictional contacts are Metzger et al.'s [31] parametric study were used to validate TFM as a
accounted for by the frictional stresses [14], which are orders of magni tool to predict crater formation. The importance of correctly dening
tude higher than the collisional and kinetic stress contributions [9]. Ac the frictional stress is shown here via a sensitivity study of the empirical
cordingly, the frictional stresses will play a signicant role in the values used in the frictional model on the overall prediction of the
behavior of dense layer of particles covering the surfaces of the Moon model. This work provides a rigorous analysis of the frictional model
[16], Mars (based on images in [17]), and asteroids. used for modeling the dense particle TFM. Finally, a frictional model for
Previous studies illustrated the importance of correctly im mulation for sand is presented that correctly predicts the salient fea
plementing the frictional stress to achieve accurate predictions, and in tures that were measured [31] for the bed of sand impacted by a
dicate caution should be used when applying frictional models due to turbulent, subsonic jet.
their inherent empiricism [9,1820]. TFM has been shown to be a viable
approach to modeling dense particle, gasparticle systems, as long as 2. Model equations
the appropriate models are implemented correctly [9,1829]. Thus,
TFM is expected to be a viable approach to predicting the interaction 2.1. Governing equations
of a gas plume with a dense particle bed.
Due to the technological and economic limitations associated with As discussed above, the twouid approach models both the uid
performing experiments that match the conditions of a spacecraft land and solid phases as interacting and interpenetrating continua, and a
ing on an extraterrestrial surface, performing experiments on Earth  in mass and a momentum balance are solved for both the solids and
atmospheric conditions, with turbulent, subsonic jets  to validate TFM uid phases. The ensemble averaged equations of motion of Anderson
for predicting crater formation is a rst step to building a software pack and Jackson [8] and Jackson [6,7] are provided in Table 1 and were
age capable of predicting rocket exhaust impingement onto extraterres used to model the uid (gas) and particle phases. As discussed above,
trial surfaces. Moreover, TFM has been used extensively to model constitutive relations are needed to close the averaged solids phase mo
spouting and jetuidized beds, which are analogous to a jet impinging mentum equation (i.e. solidsphase stress). Somewhat analogous to the
from below [9,2129]. However, for the case of a jet impinging down Reynolds stress in turbulence that depends on the uctuating velocity of
ward onto a bed of particles, which is relevant to landing spacecraft, the the uid, the collisional and kinetic particlephase stresses depends on
motion of the particle phase was neglected and only the singlephase the magnitude of the particle velocity uctuations (v's). The uctuating
uid (liquid or gas) was simulated. In particular, experimental studies velocity is the difference between the instantaneous (vs) and mean par
demonstrated that when a turbulent, subsonic jet impinges onto a parti ticle velocity (Vs), namely v's = vs Vs). A granular temperature, , can
cle bed in atmospheric conditions, a crater is produced (details available be introduced, and is given by [10],
elsewhere, e.g., [3033]), which is different than the spout and uidized
beds beds that occur when the jet is at the bottom of the bed [9,2129]. 1 D 0 2E
vs 1
Previous computational uid dynamics simulations were used to study 3
the ow patterns in horizontal scour holes by modeling a jet impinging
into a solid wall shaped as a crater (i.e. the crater shape was frozen in which is similar to the thermodynamic temperature of a gas. A balance
time and simulated using noslip boundary conditions) [31,34,35]. For in of granular energy, 3/2, is used to close the solidsphase governing
stance, Metzger et al. [31] performed a gasphase simulation of a vertical equations (Table 1). Granular energy is generated by shear in the parti
jet impinging inside of a wall in the shape of a crater in order to investi cle phase, diffuses along gradients of granular energy, dissipates due to
gate the properties of the jet inside of the crater. These previous scour inelastic particleparticle collisions and is generated and/or dissipated
and crater studies only accounted for the interaction of the uid jet with from interactions of the uid phase with the particle phase [9]. Lun
a solidwall and therefore did not account for the interaction and inter et al. [12] derived expressions for these relations based on the kinetic
penetration of the uid with the particle phase. theory of dense gas [38] allowing for inelastic particleparticle collisions.
Nonetheless, the formation of craters under impinging jets was in Only the collisional and kinetic contributions of the solids phase stress
vestigated experimentally (see elsewhere for details, e.g., [3033]). For are taken into account for the granular energy balance, as frictional
example, Metzger et al. [31] experimentally investigated the cratering stress is considered to dissipate energy to true thermal energy, rather
mechanisms dominating in beds of sand cratered by both high and than uctuating energy [14]. van Wachem et al. [9] showed that for
lowvelocity jets. They identied diffusiondriven ow (DDF) [31] and dense systems, is typically conserved locally thus convection and dif
bearing capacity failure (BCF) [36] as possible mechanisms that domi fusion can be neglected, and therefore an algebraic assumption can be
nate the in highvelocity ow regime and viscous erosion (VE) [37] to employed in general for the granular energy balance in dense particle
dominate the lowvelocity ows. VE occurs at the top layers of the par ows. The governing equations and constitutive relations are provided
ticle bed where the gas torques particles over their neighbors and in Table 1.
70 C.Q. LaMarche et al. / Powder Technology 318 (2017) 6882
Table 1
Governing equations and constitutive relations.
Governing equations
Constitutive relations
h i
Gas phase stress tensor, g (N/m2) g g Vg Vg T 23 Vg I
Solids phase stress tensor, S (N/m2) s s Vs Vs T s 23 s Vs I
Solid phase pressure, Ps (N/m2) [14] Ps = Ps,col + Ps,kin + Ps,fr
Solids collisional pressure, Ps, col (N/m2) [12] Ps,col = 2gos2s (1 + e)
Solids kinetic pressure, Ps ,kin (N/m2) [12] Ps,kin = ss
(
Solids frictional pressure, Ps, fr (N/m2) [14] 0 s s; min
P s;fr min n
Fr s s; p
s Ns; min
s; max s
Gas phase Reynolds stress (v'gv'g for RSM model), Rg(m2/s2) Rg = v'gv'g
Boussinesq hypothesis (v'gv'g for k model, used only for high velocity simulation) (m2/s2) hv0 g v0 g i 2 kI t;g Vg Vg T
3 g
Turbulent kinetic energy (highvelocity simulation only), k (m2/s2) [44]
k g g Vg k g t;gk k g t;g S2
t g g
g g g k;g
q
Mean rate of strain tensor, S (1/s)
S 12 Vg Vg T : Vg Vg T
Turbulent dissipation rate (k model; highvelocity simulation only), (m2/s3) [44]
g g Vg g t;g
kg C 1 t;g S2 g C 2
t g g
g ;g
Turbulent viscosity, t,g (Pa s) t;g g C k
2
b
Covariance of gas phase velocity and particle phase velocity, ksg (m2/s2) [45] ksg 2k 1sg , with b 1 C v s C v
1
sg g
Ratio of characteristic time scales, sg [45] sg t;sg
F;sg
Lagrangian integral time scale calculated along particle trajectories, t, sg (s) [45] t;sg p
t;g
, with
2
1C
vg vs t;g
Lt;g and C = 1.8 1.35cos2(t),
t is the angle between the mean particle velocity and the mean relative velocity
Particle relaxation time due to inertial effects acting on particle phase, F, sg (s) [45] F;sg s s 1 s C v
g
Table 1 (continued)
The mass and momentum balances solved for the gas phase are also Johnson and Jackson [14] and Johnson et al. [54] developed the
provided in Table 1. The Reynolds Stress Model (RSM) was used to sim commonlyused, empirical equation for the frictional pressure that is
ulate the turbulent ow of the jet. For the case of a jet impinging on a given by [54],
dense particle bed simulated here, it is assumed that the effects of tur
8
bulence are important in diluteparticle regions and negligible in >
< 0 s s; min
n
dense particle regions. Accordingly, Cokljat et al.'s [51] RSM is used to P s;fr s s; min 2
account for the effects of particles on turbulence here. Transport equa : Fr
> p s Ns; min
s; max s
tions and for the RSM employed, i.e., the Reynolds stresses Rg and turbu
lence dissipation rate , are provided in Table 1.
where Fr, n and p are materialspecic, empirical constants. The form of
Ps,fr in Eq. (2) captures the measured trend of rapidly increasing Ps,fr with
increasing s that diverges as s,max is approached [14,54]. Additionally,
2.2. Frictional stress model
the form of Ps,fr in Eq. (2) ensures that frictional stresses are not consid
ered for s bs,min [54]. Johnson and Jackson [14] and Johnson et al. [54]
The aforementioned frictional stress models account for the effect of
use Coulomb's law (the frictional force is proportional to the normal
friction on particle contacts, and is added to the kinetic and collisional
force) to relate Ps,fr to s,fr. The rst s,fr model analyzed in this paper
stress contributions in the solids phase momentum balance [52]
was proposed by Schaeffer, is also based on Coulomb's law, and is
(Table 1). Frictional stress models are based on critical state theory,
given by [40],
which is rooted in soil science, and the resulting stress equations are
empirical [14,20,53]. Johnson and Jackson [14] considered the frictional 8
> 0 s s; min
stress in Newtonian form, i.e., it is represented in terms of a pressure P s;fr sin < n
(Ps,fr; normal stress) and viscosity (s,fr; shear stress), and proposed s;fr p s s; min sin 3
2 I2D : Fr
> p p s N s; min
forms of the Ps,fr and s,fr based on the assumption that the medium s; max s 2 I 2D
does not dilate or contract while deforming i.e., the critical state as
sumption. Srivastava and Sundaresan [20] found that the critical state where I2D is the second invariant of the deviatoric stress tensor and is
assumption did not affect the prediction of particles discharging from the internal angle of friction. To avoid singularity in calculating s,fr a
a hopper. minimum value for I2D (1012) was applied, but did not affect the re
The frictional stress is used to account for the effects of friction that sults as I2D was always N10 12 in the regions of interest. Srivastava
occur result from sustained contacts, which are more likely to occur at and Sundaresan [20] proposed a model for s,fr to account for uctua
higher solids volume fractions [54]. For solids volume fractions above tions in the strain rate associated with the formation of shear layers
a critical limit, s N s,min, enduring contacts are assumed and friction by combining Schaeffer's frictional viscosity (Eq. (3)) [40] with Savage's
must be considered [54]. In general, enduring contacts are assumed to estimate of the rootmean square strainrate uctuation [52]. Srivastava
occur for s,min = 0.5, as particles do not touch for uniformly spaced and Sundaresan's [20] frictional viscosity model is given by [20],
grains with s b s,min [54]. A model that does not account for enduring
contacts, or does not accounts for friction until an unrealistically high P s;fr sin
value of s, the bed will act too much like a liquid. On the other hand, s;fr q 4
2
if a model accounts for enduring contacts when a system is too dilute, 2 I 2D =ds
i.e., s,min is taken as a value too low, the particles cannot be assumed
to endure these long contacts (i.e., collisions should be considered in and includes the dependence on granular temperature to account for
stantaneous) and the ow will be incorrectly predicted. Accordingly, the formation of shear layers, i.e., /d2s where ds is the particle diameter.
the frictional stress is considered to be zero until a critical solids volume Previous TFM simulations using Johnson et al.'s [54] Ps,fr (Eq. (2)) and
fraction, s,min, is achieved [54]. The frictional stress increases with in Schaeffer's [40] s,fr (Eq. (3)), predicted dense beds of glass beads with
creasing s until the maximum random close packing fraction s,max for too high of a frictional stress in certain ow geometries [18,20]. By in
the solids is reached [54], i.e., the maximum solids volume fraction cluding the granular temperature in Eq. (4), s,fr is effectively reduced
that randomly oriented grains can reach [55]. From a physical stand relative to the s,fr predicted with Eq. (3) for the same Ps,fr. Hence, s,fr
point, the frictional stress accounts for the frictional contacts that op can be viewed as an effective friction, and this same idea is extended
pose compaction of a particle bed. in this work to account for rolling friction effects.
72 C.Q. LaMarche et al. / Powder Technology 318 (2017) 6882
The frictional viscosity theories based on Coulomb's law (Eq. (3) and 3.1. Experimental setup of Metzger et al. (2009)
Eq. (4)) are derived for particle contacts that only experience sliding
friction [40], and therefore do not account for other interactions that Metzger et al.'s [31] cratering experiments were used to validate the
can occur for particles in enduring contact such as rolling friction be TFM used here, and the experimental apparatus used to measure the
tween grains. However, the resistance to rolling at particle contacts crater formation is schematized in Fig. 1. A brief description of their ex
can have a signicant effect on the shearing behavior of granular mate periments is provided here, but the details can be found elsewhere [31].
rials [56]. Rolling resistance is expected to increase the bulk frictional In Metzger et al.'s [31] experiments, a gas jet was directed downward,
behavior of granular materials for the same normal force [56]. Nonethe towards a bed of particles. The jet was produced by gas ow expanding
less, a wellestablished, continuum, frictional stress theory that incorpo out of the exit of a long, straight pipe. The experiment was designed
rates rolling friction does not exist. Accordingly, as mentioned above, such with the intent that the jet would be split by the bevelededge of
s,fr is treated as semiempirical here and recast as, the window and a halfjet would impact the particle bed. Then, the cra
ter development could be monitored through the window [31]. Metzger
8 et al. [31] performed experiments on Jetty Park Beach sand sieved to a
>
< 0 s s; min particle size range of 100180 m.
n
s;fr s s; min sin 5
>
: Fr p p s Ns; min
s; max s 2 I2D 3.2. Simulation setup
tested, such as the drag, kinetic and collisional solidsphase stress, and
turbulence model.
The accuracy of the RSM and k model were veried by simulating a
jet impacting a wall placed orthogonally to the exit of the jet, and com
paring the predictions to the measurements of Cooper et al. [58] and
simulations of Craft et al. [59], respectively. The turbulence model pre
dictions were consistent with the predictions of Craft et al. [59]. In par
ticular, both turbulence models have high errors when predicting
stagnation regions, where strainrates are often high, but in general,
the RSM is more accurate than the k model [59]. A thorough discus
sion of the limitations of turbulence models predicting the stagnation
region is [59].
Once the most accurate frictional stress model formulation was de
termined by comparing predictions to measurements [31] for the base
Fig. 2. Axisymmetric formulation (a) computational mesh and (b) simulation domain. case conditions, it was used to predict the crater formation in conditions
Unless otherwise noted, A = 15.2 cm, Hp = 7.62 cm, Dp = 1.02 cm. that match Metzger et al.'s [31] other experiments. Metzger et al. [31]
reported the results of parametric tests of the properties of the jet in
s = 105, which was veried by lowering the value by two orders of the lowvg regime. Here, only the smallest and largest extremes of
magnitude. each test (e.g., smallest and largest gas density) were simulated for
To simulate the gas expansion from the jet, a short length of pipe this work. It is important to note that the frictional model was not
was simulated in the domain (the gas inlet boundary was designated changed when simulating conditions other than the base case; only
as a velocityinlet). For each simulation, the magnitude of the gas veloc the simulated jet operating conditions were changed (gas velocity, gas
ity at the velocityinlet i.e., jet velocity, vg was held constant species, etc.) to determine the accuracy of the frictional model in
throughout each simulation, and for a given set of simulation condi predicting the behavior of the sand. To avoid confusion, the conditions
tions, the velocity of the gas at the inlet was set to the experimentally and methods used to set up the simulations for the highvg regime are
published value [31]. The turbulence conditions of the jet were set discussed in the relevant section below.
using turbulence intensity and length scale. Table 2 provides all of the
values used for the simulations. 4. Results and discussion
3.3. Model sensitivity In this section, the sensitivity of crater formation to various numeri
cal and physical aspects of the simulation is discussed. Then, the effect of
Sensitivity to mesh size as well as time step were tested to ensure the overall magnitude of the frictional stresses as well as the relative
that accurate resolution of gradients were captured. The effect of the magnitude of Ps,fr and s,fr are investigated, and the results are presented
presence of the sidewall and side and topexit boundaries on the crater with respect to the cratering phenomena. A model is presented, Model
predictions was investigated by implementing a mesh geometry with A, which accurately predicts the salient features of crater growth mea
walls twice as far as in the original mesh geometry. Additionally, the sured by Metzger et al. [31]. Additionally, the new model, which is an
sensitivity of the crater growth rate to the physical models was also extension of previous frictional models, is employed to explore the
cratering mechanism for both lowvg and highvg jet regimes.
Table 2
Simulation parameters. 4.1. Sensitivity results
Property Value
The sensitivity of the cratering prediction was tested against various
Particle density, s (kg/m3) 2650 aspects of the simulation domain, physical models, and numerical
Particle diameter, ds (m) 140
methods, as discussed above. In terms of the sensitivity of the crater
Particleparticle coefcient of restitution, e 0.9
Particlewall coefcient of restitution 0.9 growth to the computational mesh, if the mesh element size was too
Angle of internal friction, () 28 large in the impingement region, the crater formation began much
Max packing solids volume fraction, s,max 0.63 later in time and did not grow at the correct rate. The size of the domain
Min solids volume fraction, s,min 0.5 was large enough to not affect the crater growth, as the predictions
Initial solids volume fraction 0.58
were not affected when the size of the domain was doubled (regardless
74 C.Q. LaMarche et al. / Powder Technology 318 (2017) 6882
of the frictional stress model formulation). Additionally, the cratering in Table 3). In Model B, s,fr and Ps,fr are increased compared to Model
predictions were not affected when the length of the pipe in the domain A. Using Model C, the effect of increasing Ps,fr relative to s,fr with respect
was increased, as long as the mesh was resolved enough at the bedjet to Model A can be investigated. The Ocone et al. model [63] accounts
interface, indicating that the domain/mesh used was adequate for all only for sliding friction (Eq. (3)) and reduces the overall frictional stress
simulations. by two orders of magnitude with respect to the Johnson et al. model.
When the jet initially impacts the bed, a small time step, 107 s, was The last model in Table 3, implements the frictional viscosity model of
required to ensure convergence. The time step was slowly increased to Srivastava and Sundaresan [20] (Eq. (4)) that effectively reduces s,fr rel
106 s by 0.1 s. All times reported, associated with the simulations, are ative to the Johnson et al. model to account for the formation of shear
in seconds of real time and from here on will be referred to with the unit layers.
of seconds, to be concise. To reduce error in the solution, the residuals The rst second of crater growth predicted under the base case con
were set to 105 in order to set a strict tolerance on the convergence ditions by the various frictional model formulations are compared in
of all variables. Convergence was checked for every time step as the tra Fig. 3. The solids volume fraction of 0.02 was used to track the bed sur
jectory of the solution changed if the convergence criteria were not met, face, which is necessary to determine the predicted crater depth for all
even for only a few time steps. simulations (the crater depthprediction was insensitive to the s
The frictional stress model formulation was found to dominate the value used to track the surface of the bed). The Johnson et al. frictional
cratering predictions. Setting the collisional and kinetic stress compo model formulation, the empirical parameters of which were originally
nents to a constant value did not affect the growth rate of the crater. Ad determined for glass spheres [54], predicts a crater that grows too
ditionally, the cratering rate was independent of the drag model when quickly compared to the measured depth. However, as stated earlier,
comparing predictions of the Wen and Yu [42] and Gidaspow [39] Metzger et al.'s [31] experiments were performed with sand, which is
drag models. Since crater formation is insensitive to the drag force the more angular and rougher than glass spheres, and therefore the same
lift force was neglected as it is small compared to the drag and gravita empirical (materialspecic) parameters would not be applicable for
tional forces [6062]. both particle types. The prediction of the particle bed by Model A is
The effect of the turbulence model on the model prediction was ex more accurate than the Johnson et al. model, as illustrated in Fig. 3,
plored by comparing simulation predictions employing the k model to and was produced by (empirically) accounting for rolling friction via a
the RSM. The RSM was less sensitive to time step size and predicted bet factor of two increase in s,fr relative to the Johnson et al. model. Accord
ter resolution of gasphase vortical structures. Accordingly, for the ingly, the model adjustments for Model B and Model C are based on
quantitative validation of the TFM the RSM was used for simulations Model A, rather than the Johnson et al. model.
of lowvg conditions. The k turbulence model did affect the quantita The predictions of frictional model formations Model A and Model B
tive predictions of the twouid model for the lowvg base case, are similar, even though p and p increase from 5 to 5.5 for Ps,fr (Eq. (2))
but the qualitative predictions were similar. However, since only a and s,fr,(Eq. (5)) respectively (Table 3). Thus, simulations were per
qualitative comparison of the cratering predictions and experiments is formed with other frictional model formulations that simultaneously
performed for the highvg (nearsonic ow) regime, the more computa increased the frictional stress for Ps,fr and s,fr further from the magni
tionally efcient the k model. tude of Model A (i.e. Fr = 0.1 and p = 5.9; Fr = 1.0 and p = 5), which
The results of a simulation implementing the full partial differential predicted a negligible difference in the crater growth as compared to
equation for the granular temperature balance, found in Table 1, show Model B. Additionally, a simulation was performed with a model that in
that the algebraic assumption for determining the granular temperature creased s,fr further relative to Ps,fr compared to Model A e.g., to inves
is sufciently accurate for the present case. The algebraic form of the tigate the effects of further increasing the rolling friction but the crater
granular energy balance (neglecting convection and diffusion) was prediction is again negligibly different than that of Model A. The similar
found to be acceptable by comparing the results to those predicted by bed prediction of increased frictional models indicates that once the
the full partial differential equation balance. frictional stress is of a certain order, or the frictional viscosity is
Table 3
Summary of frictional models.
Fr p Fr p
relatively large compared to the frictional pressure, the cratering predic 4.3. Parametric study: comparing crater predictions to measurements
tions become insensitive to further increases.
Decreasing the frictional stress by two orders of magnitude with the The conguration of frictional models Model A and Model B predict
Ocone et al. model causes large changes in the cratering predictions ed similar crater growth and were the most accurate predictions of the
compared to the Johnson et al. model. The Ocone et al. model over pre crater development for the base case (Fig. 3). However, since Model A
dicts the crater depth, as this model signicantly underestimates the was slightly more computationally efcient, Model A was used to simu
frictional stresses for sand. When the frictional stresses are too low, late the various jet conditions reported by Metzger et al. [31]. In partic
the bed behaves too liquidlike and the jet drives deep into the bed ular, Model A was used to predict crater formation to t = 1 s with
followed by the bed sloshing back. This splashing of the bed can be iden conditions that match the parametric investigation by Metzger et al.
tied in Fig. 3, where the crater depth quickly decreases after reaching a [31]. The predicted growth of the crater depth by Model A for the case
maximum at around 0.5 s, which corresponds to a jump in the crater of Argon jets with vg = 37 m/s and vg = 56 m/s plotted against Metzger
width after the crater sloshes. et al.'s [31] measurements in Fig. 4. The predictions of Model A and mea
Frictional model with increased Ps,fr only (Model C), Srivastava and surements [31] for the crater depth over time as a function of gas species
Sundaresan model, and the Johnson et al. model demonstrate increased are presented in Fig. 5. In Fig. 5, the simulation and experimental data
craterdepth growth over time compared to Model A. The three former are for 56 m/s jets of Argon, Nitrogen and Helium. The effect of Hp on
models have reduced s,fr with respect to the Ps,fr relative to the latter. As the predicted crater depth over time is compared to Metzger et al.'s
mentioned above, Srivastava and Sundaresan model has the same form [31] measurements in Fig. 6. The conditions of the experiment and sim
of Ps,fr as the Johnson et al. model, but s,fr is effectively decreased to ac ulation in Fig. 6 are a 40 m/s Nitrogen jet with nozzle heights (Hp in
count for uctuations in the strain rate (Eq. (4) vs. Eq. (3), respectively). Fig. 2) of 7.62 cm and 10.16 cm. The simulation predictions illustrated
Model C predicts a bed that craters at the second fastest rate, even in Figs. 46 were computed in an axisymmetric domain. In Figs. 4
though s,fr is the same as Model A, and Ps,fr is higher. If Ps,fr is increased, and 5, Model A predicts the crater formation to t = 1 s well for a variety
at the expense of the s,fr (with respect to Model A, i.e., Model C, Johnson of gas species and jet velocities. However, the predicted crater depth by
et al. model and Srivastava and Sundaresan model), the bed behaves too Model A, Fig. 6, is more sensitive to the pipe height than what was found
much like a liquid and the cratering rate exceeds experimental observa experimentally. However, the experimental data does not have error
tions. Conceptually, a bed with increased Ps,fr relative to s,fr, can be bars, so it is difcult to tell if the predictions are within the experimental
thought as more similar to a true (molecular) liquid than a granular ma uncertainty. Hence, the simulations performed with Model A predict the
terial because the increase in Ps,fr makes the bed more incompressible, salient features of Metzger et al.'s [31] parametric study.
but without a comparable change in s,fr i.e., the bed ows like a liquid
because the level of friction (s,fr is too low to keep the highpressure re 4.4. Long time simulations
gions of the bed from owing to lower pressure regions. In addition, for
very short times, the bed expands because Ps,fr is larger than the weight Metzger et al. [31] describes the experimentally observed crater evo
per unit area of the particles in the cell i.e., gravitational forces are too lution as the initial formation a single parabolic crater that later col
small to keep particle phase from expanding [54] which is expressed lapses into a dual crater consisting of an inner (parabolic) and outer
by the dip in the crater depth in Fig. 3 at early times for these three (conical) crater. The walls of the outer crater oscillate between the
models. Similar to the case of increased Ps,fr relative to s,fr, the bed be angle of failure and the angle of repose of the sand [31]. The crater
haves too liquidlike when the overall magnitude of the frictional stress growth rate is altered at the moment when the outer crater forms, as
es are too low, for instance with the Ocone et al. model. To better particles begin to recirculate inside of the crater (between the inner
illustrate the correlation of the magnitudes of s,fr and Ps,fr on the crater and outer craters) [31]. For the relevant experimental cases, the crater
depth prediction, Table 4 provides the orders of magnitude of these split occurs well after 1 s and therefore simulations were carried out
stresses in the bed at 0.5 s for the models provided in Table 3. Addition with Model A to longer times. Fig. 7 demonstrates such longtime pre
ally, Table 4 illustrates the bed compressibility by providing the maxi dictions of crater growth compared to Metzger et al.'s [31] base case
mum (peak) predicted s in the bed by the various models. The s measurements. The predicted crater depth of Model A diverges from
values in Table 4 are outputs from the simulation that depends on the the experimental results at around 4.2 s (Fig. 7), which is slightly later
specic frictional model formation and should not be confused with than the time it took for the crater to split in the experiments, namely,
s,max, which is a constant (Table 2) and an input to Ps,fr (Eq. (2)) and ~3.7 s [31].
s,fr (Eqs. (35)). The s values is Table 4 demonstrate that the compress To understand the underprediction of crater depth at later times,
ibility of the bed is determined by the magnitude of Ps,fr. For instance, crater proles from experiments [31] and predictions are compared in
the orders of magnitude smaller Ps,fr and s,fr of the Ocone et al. model Fig. 8. In Fig. 8, the column of images on the left are experimental snap
compared to the of Johnson et al., results in a bed that is much more shots of the crater for the base case conditions, in the middle and on the
compressible. In sum, the cratering prediction can be greatly affected right are the contour plots of s predictions for corresponding times by
by adjusting the overall frictional model magnitude (e.g., Ocone et al. Model A and the Ocone et al. model, respectively. The crater proles
model and Johnson et al. model) and the relative magnitude of s,fr to predicted by Model A does not properly predict the formation of the
Ps,fr (e.g., Model A and Model C). outer crater, Fig. 8, which could impact the solidphase recirculation
Table 4
Predicted frictional viscosity and pressure associated with different frictional models.
Fig. 4. Crater depth predictions with varying Argon jet velocities (Model A). Fig. 6. Crater depth prediction with varying nozzle height (Model A).
Fig. 5. Crater depth prediction with varying gas type (Model A). Fig. 7. Long time crater depth prediction with 37 m/s Argon jet (Model A).
C.Q. LaMarche et al. / Powder Technology 318 (2017) 6882 77
Fig. 8. Experimental crater images and corresponding simulations at different times (left column, experimental images; middle column, Frictional Model A; right column, Ocone et al.
frictional model).
In Fig. 10, the crater depth measurements for the base case [31] are The effect of the wall on crater growth is further demonstrated in
compared to the predictions with the new frictional model formulation, Fig. 11, which provides s contours within the bed predicted by Model
Model A, for three different simulations: axisymmetric without a wall A at various simulated times. The crater depth is tracked at the center
(Axisymmetric Mesh), Cartesian without a wall (Cartesian: No Wall), of the jet, at the surface of the wall. The jet is has initially defected
and Cartesian with a wall (Cartesian: with Wall). At early times, without when it rst impinges on the top of the wall, causing it to be deected
considering the wall, the crater growth rate is underpredicted in both towards the center of the bed, resulting in erosion slightly away from
the Cartesian and axisymmetric coordinate systems. However, by the centerline (Fig. 11a). The predicted drop in the crater depth (crater
~0.5 s the crater depth is accurately predicted in the Cartesian simula becomes shallower) at ~0.1 s in the presence of the wall (dashed gray
tion without considering the wall, but overpredicted in the axisymmet line in Fig. 10) is explained by a pocket of eroding solids that collides
ric simulation. Nonetheless, the crater depth is generally under with the wall, resulting in a temporary increase in the bed depth
predicted in the presence of a wall (except for the period of t ~ 0.2 to (Fig. 11b). As the jet develops over time, it expands towards the wall,
0.35 s). To understand the effect of the splitting the jet with the window causing the erosion to increase closer to the centerline, as illustrated
in experiments, the predictions of Cartesian simulations with and with in Fig. 11c and d. Hence, the interaction of the jet and the wall decreases
out the wall are compared. Specically, the wall increases the trajectory the crater growth rate compared to when a wall does not split the jet.
of the crater depth growth at early times compared to the nowall pre
dictions, but decreases crater growth at later times. This qualitative 4.6. Simulations of turning the jet off
change in crater growth due to the presence of the wall provides a pos
sible explanation for the overprediction of the crater depth near ~1 s in The dynamic behavior of the crater formation when the jet is on pro
the axisymmetric coordinate system (Fig. 3) because the wall was not vides a validation of the frictional model under high shear rates, which
simulated. Nonetheless, the predictions indicate the effects of the wall is consistent with the conditions under which the s,fr models were de
are small, but that it slightly increases the initial cratering rate and re rived [40]. However, in experiments, after the jet is extinguished, a cra
duces the erosion rate at later times. ter remains with walls at the angle of repose. The behavior of the bed
78 C.Q. LaMarche et al. / Powder Technology 318 (2017) 6882
Fig. 9. (a) Domain and (b) computational mesh for the twodimensional, Cartesian simulations without a wall splitting the jet. (c) Domain and (b) computational mesh for the two
dimensional, Cartesian simulations with a wall splitting the jet.
after turning off the jet was investigated with the various frictional velocity at the inlet of the jet to a very small value and modeling the
model formulations. It is not expected that TFM with the frictional ow as laminar in the axisymmetric simulation. Additionally, the top
models implemented here is capable of predicting a crater with walls surface of the particle bed was tracked using s = 0.58, as this s is
at the angle of repose (see Fig. 8) or a stable crater after the jet is above s,min, such that the ow behavior is controlled by frictional
extinguished. Nonetheless, it is expected that the a model with a suf stresses. Immediately after extinguishing the jet, the crater predicted
ciently high s,fr can predict the a crater that will ll in slowly and the by Model A initially remains at a nearly constant size, but begins to
surface of the particle bed will not splash above its original level (as slowly ll in after a few seconds of simulation time.
would occur for the surface of a true liquid after extinguishing a jet). Ex Conversely, the frictional models that predicted a particle bed that
periments performed with analogous particles in a similar experimental behaved much like a liquid while the jet was on (the Ocone et al.
apparatus show that the crater remains when the jet is terminated at model and Model C), also predicted a bed that behaved too liquidlike
~1 s. Hence, the effect of turning off the jet on the behavior of the crater after turning off the jet, as the crater immediately collapsed and made
was explored using the predictions from Model A by turning the gas a splash. Hence, the TFM with improved frictional stress description,
C.Q. LaMarche et al. / Powder Technology 318 (2017) 6882 79
Fig. 11. Contour plots of solids volume fraction predicted using the Cartesian simulation with thin wall after (a) t = 0.05 s, (b) t = 0.10 s, (c) t = 0.13 s, and (d) t = 0.40 s. The solid black
(vertical) line indicates the wall and the dotted (horizontal) line represents the initial, undisturbed (t = 0 s) bed height.
80 C.Q. LaMarche et al. / Powder Technology 318 (2017) 6882
are tangential to the edge of the crater, indicating the DDF mechanism
dominates the crater formation. To verify that BCF does not occur for ei
ther the lowvg or highvg cases, the predicted stagnation pressures can
be compared to the bearing capacity of the cratering material. In partic
ular, the gas pressure in the stagnation region are predicted to be
~800 Pa and ~9000 Pa for the lowvg and highvg simulations, respec
tively, but the bearing capacity of sand is on the order of 100,000 Pa
[67], indicating that the BCF mechanism does not occur.
Isosurface contours of gas velocity magnitude, Vg, plotted on top of
isosurfaces of s are rendered in Fig. 13 for predictions of the 37 m/s
Argon jet after t = 0.5 s (Fig. 13a) and for the near sonic compressible
Nitrogen jet (Fig. 13b) by Model A. To compare the penetration of the
Fig. 12. Solid phase mass ux (m/s) vectors and solids volume fraction isosurfaces pre
dicted by Model A for (a) an Argon jet with vg = 37 m/s at t = 0.5 s and (b) a near
sonic nitrogen jet at t = 0.067 s. It is important to note that the plot in (a) is zoomed in
to a smaller scale of axial and radial positions than in (b) in order for the mass ux in
the shear region to be observable for both cases.
of the solids phase in the bulk, i.e., s 0.5 (Fig. 12b), indicates that the
dominating mechanism is either DDF or BCF rather than VE. Metzger
et al. [66] determined that the trajectories of particles will be tangential
to the edge of the crater if DDF dominates the crater formation (the gas
from the jet penetrates into the pores and slightly uidizes the particles,
causing them to move with the gas through the bed), and perpendicular
to the crater wall if the dominating mechanism is BCF (the gas jet acts
like a mechanical force that pushes particles downward). The simula Fig. 13. Gas velocity (m/s) isosurface contours plotted and volume fraction isosurfaces
tion predictions are in agreement with Metzger et al.'s [31] experiments predicted by Model A for (a) Argon jet with velocity 37 m/s at t = 0.5 s and (b) a near
as the trajectories of the solids phase predicted in the highvg simulation sonic nitrogen jet at t = 0.067 s.
C.Q. LaMarche et al. / Powder Technology 318 (2017) 6882 81
jet into the bed for the lowvg and highvg cases, the Vg shown in Fig. 13 friction coefcient [15,68]. Additionally, comparing predictions of crater
is limited to isosurfaces with a range of 0 to 1 m/s. The small Vg values growth for a wide range of particles to measurements [69] in order to
below the bed surface presented in Fig. 13a indicates there is little gas determine the robustness of the materialspecic frictional stress
penetration for the lowvg case, whereas Fig. 13b illustrates deep pene model.
tration of the jet into the particle bed in the form of large Vg values deep
below the surface of the crater for the highvg case. Comparing the gas Nomenclature
velocity magnitude at similar distances below the surface of the dense
CD  Drag coefcient
crater, which is taken here as solids volume fraction associated with ds [L] Particle diameter
the onset of enduring contacts (s = s,min = 0.5), can be used to quan e  Coefcient of restitution
tify the difference in gas penetration from the jet between the lowvg Fr, Fr [MT2 L1] Empirical material parameter for frictional pressure
and highvg cases. For the lowvg case, the penetration of gas is quanti g [LT2] Gravity
go  Radial distribution function
ed by Vg = 0.05 m/s at a distance of ~11 mm below the dense surface I2D [T2] Second Invariant of the Strain Rate Tensor
of the crater, s = 0.5 (Fig. 13a). However, for the highvg case, at the k [L2 T2] Turbulent kinetic energy
same depth below s = 0.5, the gas velocity magnitude is 0.8 m/s. Fur L [L] Length
thermore, for the highvg case, Vg does not reduce to 0.05 m/s until a dis M [M] Mass
n, n, p, p  Exponent in frictional stress model
tance of 58 mm below s = 0.5 (not shown in Fig. 13b). The DDF
P [MT2 L1] Gas pressure
mechanism occurs when the gas diffuses into the bed faster than the Ps,col [MT2 L1] Solids collisional pressure
solids can move, and the gas drags the particles away from the crater Ps,fr [MT2 L1] Solids frictional pressure
[31]. Accordingly, Vg must be large enough for the drag force to be signif Ps,kin [MT2 L1] Solids kinetic pressure
icant compared to the particle weight, which is more likely in the high Rep  Particle Reynolds number
Rg = v'gv'g [L2 T2] Gas phase Reynolds stress
vg case as indicated by the high gas velocities predicted inside of the bed
t [T] Time
(Fig. 13b). Greek symbols
[MT1 L3] Interphase momentum transfer coefcient
5. Conclusions s [MT3 L1] Collision dissipation of Granular Temperature
g  Gas volume fraction

The twouid model (TFM) including frictional stress models is used 
to predict the crater formation of a particle bed impacted by a subsonic, s  Solids volume fraction
turbulent jet. It is determined here that the crater formation is dominat s,min  Solids volume fraction when frictional stress is
ed by frictional stresses, and thus crater formation is a promising system activated
s,max  Max packing solid volume fraction
for studying and validating frictional stress models. The frictional stress
[L2 T3] Turbulent gas dissipation rate
models implemented in this work are based on previously developed [L2 T2] Granular temperature
models [14,20,40,63], which require empirical, materialspecic param s [MT1 L1] Diffusivity of (s = 0, because of steady state
eters. In this work, the effect of the formulation of the frictional stress assumption)
s [MT1 L1] Bulk granular viscosity
model on the behavior of the particle bed is reported for the rst time.
s [MT1 L1] Solids viscosity
In particular, it is demonstrated that the overall frictional stress magni s,col [MT1 L1] Solids collisional viscosity
tude and the relative magnitude of the frictional pressure to the friction s,kin [MT1 L1] Solids kinetic viscosity
al viscosity greatly affect the predicted behavior of the bed, which is s,fr [MT1 L1] Solids frictional viscosity
further evidence that these empirical stress models should be used t,g [MT1 L1] Turbulent gas viscosity
Vg , Vs [LT1] Mean velocity vector
with caution. A particle bed is predicted to behave too liquidlike if
v's [LT1] Solids velocity uctuation
the frictional stresses are too low. Additionally, when the frictional pres s, g [ML3] Density
sure is increased without a comparable increase in the frictional viscos g, S [MT2 L1] Stress tensor
ity, the bed behaves more liquidlike (e.g., the bed splashes while the jet [Deg] Angle of Internal Friction
is on and collapses immediately after the jet is turned off). Alternatively,
when the frictional viscosity is increased without a comparable increase
in the frictional pressure, the bed acts less liquidlike (e.g., the bed does Acknowledgement
not splash and the bed remains after the jet is turned off).
Furthermore, a new frictional stress model is formulated for 100 This work has been funded by NASA STTR (NNK08EB52C) Phase I
180 m beach sand, Model A, and the predicted crater growth is validat and II as well as NSF OISE 0968313. The authors are grateful to Dr. Philip
ed using the measurements of Metzger et al. [31]. The frictional stress Metzger for his insightful conversations related to this work.
formulated in this work is based on the models of Johnson and Jackson
[14] and Schaeffer [40] with frictional viscosity increased to account for References
the higher friction between sand particles relative to glass spheres that [1] P.T. Metzger, J.E. Lane, C.D. Immer, S. Clements, Cratering and blowing soil by rocket
is not accounted for in Coulomb's law of friction (i.e., sliding only). The engines during lunar landings, in: H. Benaroya (Ed.), Lunar Settlements, CRC Press,
new frictional model formulation can predict the cratering behavior of Boca Raton, FL 2010, pp. 551576.
[2] A. Sengupta, J. Kulleck, S. Sell, J. Van Norman, M. Mehta, M. Pokora, Mars lander en
a sand bed with adequate agreement with Metzger et al.'s [31] measure gine plume impingement environment of the Mars science laboratory, Proceedings
ments for both the high (near sonic) and low (subsonic, turbulent) jet of the 2009 IEEE Aerospace Conference 2009, pp. 110, http://dx.doi.org/10.1109/
velocity regimes. Predictions of the crater formation by new model AERO.2009.4839345.
[3] M. Rhodes, Introduction to Particle Technology, second ed. John Wiley & Sons,
agree quantitatively with Metzger et al.'s [31] measurements for various Chichester, 2008.
experimental conditions (e.g., jet velocities, jet diameters, gas densities, [4] Y. Guo, J.S. Curtis, Discrete element method simulations for complex granular ows,
etc.) in the lowvelocity regime. In the highvelocity regime, model pre Annu. Rev. Fluid Mech. 47 (2015) 2146, http://dx.doi.org/10.1146/annurevuid
010814014644.
dictions are in qualitative agreement with the experimental observa
[5] P. Liu, C.M. Hrenya, Challenges of DEM: I. Competing bottlenecks in parallelization of
tions of Metzger et al. [31]. Additionally, for both velocity regimes, the gassolid ows, Powder Technol. 264 (2014) 620626, http://dx.doi.org/10.1016/j.
cratering mechanisms predictions agree with the experimentally deter powtec.2014.04.095.
mined mechanisms [31]. Possible future extensions of this work include [6] R. Jackson, Locally averaged equations of motion for a mixture of identical spherical
particles and a Newtonian uid, Chem. Eng. Sci. 52 (1997) 24572469, http://dx.doi.
investigating crater formation using newer models for frictional viscos org/10.1016/S00092509(97)000651.
ity that rely on properties measured from particles, e.g., particleparticle [7] R. Jackson, Erratum, Chem. Eng. Sci. 53 (1998) 1955.
82 C.Q. LaMarche et al. / Powder Technology 318 (2017) 6882
[8] T.B. Anderson, R. Jackson, Fluid mechanical description of uidized beds. Equations [36] J.D. Alexander, W.M. Roberds, R.F. Scott, Soil Erosion by Landing Rockets, Hayes In
of motion, Ind. Eng. Chem. Fundam. 6 (1967) 527539, http://dx.doi.org/10.1021/ ternational Corp, Birmingham, AL, 1966.
i160024a007. [37] Visibility and dust erosion during the lunar landing, in: L. Roberts, Langley Research
[9] B.G.M. van Wachem, J.C. Schouten, C.M. van den Bleek, R. Krishna, J.L. Sinclair, Com Center (Eds.), A Compilation of Recent Research Related to the Apollo Mission,
parative analysis of CFD models of dense gassolid systems, AICHE J. 47 (2001) National Aeronautics and Space Administration, Hampton, VA 1963, pp. 155170.
10351051. [38] S. Chapman, T.G. Cowling, The Mathematical Theory of nonUniform Gases: An Ac
[10] C.S. Campbell, Rapid granular ows, Annu. Rev. Fluid Mech. 22 (1990) 5792. count of the Kinetic Theory of Viscosity, Thermal Conduction and Diffusion in Gases,
[11] B. Van Wachem, J.C. Schouten, C.M. van den Bleek, R. Krishna, J.L. Sinclair, CFD third ed. Cambridge University Press, New York, 1970.
modeling of gasuidized beds with a bimodal particle mixture, AICHE J. 47 [39] D. Gidaspow, Multiphase Flow and Fluidization, Academic Press, 1994.
(2001) 12921302. [40] D.G. Schaeffer, Instability in the evolution equations describing incompressible
[12] C.K.K. Lun, S.B. Savage, D.J. Jeffrey, N. Chepurniy, Kinetic theories for granular ow: granular ow, J. Differ. Equ. 66 (1987) 1950, http://dx.doi.org/10.1016/0022
inelastic particles in Couette ow and slightly inelastic particles in a general ow 0396(87)900386.
eld, J. Fluid Mech. 140 (1984) 223256. [41] J.L. Sinclair, R. Jackson, GasParticle ow in a vertical pipe with particleparticle in
[13] V. Garz, J.W. Dufty, Dense uid transport for inelastic hard spheres, Phys. Rev. E 59 teractions, AICHE J. 35 (1989) 14731486.
(1999) 5895. [42] C. Wen, Y.H. Yu, Mechanics of uidization, Chem. Eng. Prog. Symp. Ser. 62 (1966)
[14] P.C. Johnson, R. Jackson, Frictionalcollisional constitutive relations for granular ma 100111.
terials, with applications to plane shearing, J. Fluid Mech. 176 (1987) 6793. [43] P.N. Rowe, Drag forces in a hydraulic model of a uidised bed part II, Trans. Inst.
[15] S. Chialvo, J. Sun, S. Sundaresan, Bridging the rheology of granular ows in three re Chem. Eng. 39 (1961) 175180.
gimes, Phys. Rev. E 85 (2012), 021305. http://dx.doi.org/10.1103/PhysRevE.85. [44] S.E. Elghobashi, T.W. AbouArab, A twoequation turbulence model for twophase
021305. ows, Phys. Fluids 26 (1983) 931938.
[16] J.K. Mitchell, W.N. Houston, W.D.I. Carrier, N.C. Costes, Apollo Soil Mechanics [45] C. Simonin, P.L. Viollet, Predictions of an oxygen droplet pulverization in a com
Experiment S200: Final Report Covering Work Performed Under NASA Contract pressible subsonic coowing hydrogen ow, Num Methods Multiphase Flows.
NAS911266, 15 ed., Space Science Laboratory Series, 15, issue 7, 1974. FED91 1990, pp. 6582.
[17] R.E. Arvidson, R.C. Anderson, P. Bartlett, J.F. Bell, D. Blaney, P.R. Christensen, et al., [46] D. Cokljat, V.A. Ivanov, F.J. Sarasola, S.A. Vasquez, Multiphase kepsilon models for
Localization and Physical Properties Experiments Conducted by Spirit at Gusev Cra unstructured meshes, Proceedings of the 2000 ASME Fluids Engineering Division
ter, Science 305 (2004) 821824, http://dx.doi.org/10.1126/science.1099922. Summer Meeting, MA, Boston 2000, pp. 18.
[18] S. Benyahia, Validation Study of Two Continuum Granular Frictional Flow Theories, [47] F.S. Lien, M.A. Leschziner, Assessment of turbulencetransport models including
Ind. Eng. Chem. Res. 47 (2008) 89268932, http://dx.doi.org/10.1021/ie8003557. nonlinear RNG eddyviscosity formulation and secondmoment closure for ow
[19] A. Passalacqua, L. Marmo, A critical comparison of frictional stress models applied to over a backwardfacing step, Comput. Fluids 23 (1994) 9831004.
the simulation of bubbling uidized beds, Chem. Eng. Sci. 64 (2009) 27952806, [48] B.E. Launder, On the effects of a gravitational eld on the turbulent transport of heat
http://dx.doi.org/10.1016/j.ces.2009.03.005. and momentum, J. Fluid Mech. (1975).
[20] A. Srivastava, S. Sundaresan, Analysis of a frictionalkinetic model for gasparticle [49] M.M. Gibson, B.E. Launder, Ground effects on pressure uctuations in the atmo
ow, Powder Technol. 129 (2003) 7285. spheric boundary layer, J. Fluid Mech. 86 (1978) 491511.
[21] A. Boemer, H. Qi, U. Renz, Eulerian simulation of bubble formation at a jet in a two [50] B.E. Launder, Secondmoment closure: present and future? Int. J. Heat Fluid Flow
dimensional uidized bed, Int. J. Multiphase Flow 23 (1997) 927944. 10 (1989) 282300, http://dx.doi.org/10.1016/0142727X(89)900179.
[22] S. Dan, W. Shuyan, L. Gougdong, W. Shuai, L. Yongjian, W. Lixin, Simulations of ow [51] D. Cokljat, M. Slack, S.A. Vasquez, A. Bakker, Reynoldsstress Model for Eulerian Mul
behavior of gas and particles in a spouted bed using a secondorder moment tiphase, Prog Comp Fluid Dynamics, 6, 2006 168178.
methodfrictional stresses model, Chem. Eng. Sci. 65 (2010) 26352648, http://dx. [52] S.B. Savage, Analyses of slow highconcentration ows of granular materials, J. Fluid
doi.org/10.1016/j.ces.2009.12.042. Mech. 377 (1998) 126.
[23] C.R. Duarte, M. Olazar, V.V. Murata, M.A.S. Barrozo, Numerical simulation and exper [53] A. Schoeld, P. Wroth, Critical State Soil Mechanics, McGrawHill, London, 1968.
imental study of uidparticle ows in a spouted bed, Powder Technol. 188 (2009) [54] P.C. Johnson, P. Nott, R. Jackson, Frictional collisional equations of motion for partic
195205, http://dx.doi.org/10.1016/j.powtec.2008.04.077. ulate ows and their application to chutes, J. Fluid Mech. 210 (1990) 501535.
[24] O. Gryczka, S. Heinrich, N.G. Deen, M. van Sint Annaland, J.A.M. Kuipers, M. Jacob, [55] S. Torquato, T.M. Truskett, P.G. Debenedetti, Is random close packing of spheres well
et al., Characterization and CFDmodeling of the hydrodynamics of a prismatic dened? Phys. Rev. Lett. 84 (2000) 20642067, http://dx.doi.org/10.1103/
spouted bed apparatus, Chem. Eng. Sci. 64 (2009) 33523375, http://dx.doi.org/ PhysRevLett.84.2064.
10.1016/j.ces.2009.04.020. [56] N. Estrada, A. Taboada, F. Radjai, Shear strength and force transmission in granular
[25] R. Hong, H. Li, M. Cheng, J. Zhang, Numerical simulation and verication of a gas media with rolling resistance, Phys. Rev. E 78 (2008), 021301.
solid jet uidized bed, Powder Technol. 87 (1996) 7381. [57] S. Patankar, Numerical Heat Transfer and Fluid Flow, CRC Press, New York, 1980.
[26] S.H. Hosseini, G. Ahmadi, B. Saeedi Razavi, W. Zhong, Computational uid dynamic [58] D. Cooper, D.C. Jackson, B.E. Launder, G.X. Liao, Impinging jet studies for turbulence
simulation of hydrodynamic behavior in a twodimensional conical spouted Bed, model assessmentI. Floweld experiments, Int. J. Heat Mass Transf. (1993).
Energy Fuel 24 (2010) 60866098, http://dx.doi.org/10.1021/ef100612r. [59] T.J. Craft, L. Graham, B.E. Launder, Impinging jet studies for turbulence model
[27] B. Peng, J. Zhu, C. Zhang, Numerical study on the effect of the air jets at the inlet dis assessmentII. An examination of the performance of four turbulence models, Int.
tributor in the gassolids circulating uidizedbed risers, Ind. Eng. Chem. Res. 49 J. Heat Mass Transf. (1993).
(2010) 53105322, http://dx.doi.org/10.1021/ie901902j. [60] J.F. Kok, N.O. Renno, A comprehensive numerical model of steady state saltation
[28] K.G. Santos, V.V. Murata, M.A.S. Barrozo, Threedimensional computational (COMSALT), J. Geophys. Res. 114 (2009), D17204. http://dx.doi.org/10.1029/
uid dynamics modelling of spouted bed, Can. J. Chem. Eng. 87 (2009) 211219, 2009JD011702.
http://dx.doi.org/10.1002/cjce.20149. [61] B.R. White, J.C. Schulz, Magnus effect in saltation, J. Fluid Mech. 81 (1977) 497512.
[29] R.P. Utikar, V.V. Ranade, Single jet uidized beds: experiments and CFD simulations [62] X.Y. Zou, H. Cheng, C.L. Zhang, Y.Z. Zhao, Effects of the Magnus and Saffman forces
with glass and polypropylene particles, Chem. Eng. Sci. 62 (2007) 167183, http:// on the saltation trajectories of sand grain, Geomorphology 90 (2007) 1122, http://
dx.doi.org/10.1016/j.ces.2006.08.037. dx.doi.org/10.1016/j.geomorph.2007.01.006.
[30] R. Haehnel, Particle Entrainment Under nonUniform Flow Created by an Impinging [63] R. Ocone, S. Sundaresan, R. Jackson, GasParticle ow in a duct of arbitrary inclina
jet, Dartmouth College, 2011. tion with particleparticle interactions, AICHE J. 39 (1993) 12611271.
[31] P.T. Metzger, C.D. Immer, C.M. Donahue, B.T. Vu, R.C.I. Latta, M. DeyoSvendsen, Jet [64] N. Reuge, L. Cadoret, C. CoufortSaudejaud, S. Pannala, M. Syamlal, B. Caussat,
induced Cratering of a Granular Surface With Application to Lunar Spaceports, J. Multiuid Eulerian modeling of dense gassolids uidized bed hydrodynamics: in
Aerosp. Eng. 22 (2009) 2432, http://dx.doi.org/10.1061/(ASCE)0893 uence of the dissipation parameters, Chem. Eng. Sci. 63 (2008) 55405551,
1321(2009)22:1(24). http://dx.doi.org/10.1016/j.ces.2008.07.028.
[32] N. Rajaratnam, S. Beltaos, Erosion by impinging circular turbulent jets, J. Hydraul. [65] P.T. Metzger, Personal Communication, 2010.
Div. 103 (1977) 11911205. [66] P.T. Metzger, R.C.I. Latta, J.M. Schuler, C.D. Immer, Craters formed in granular beds by
[33] N. Rajaratnam, K.A. Mazurek, Erosion of a polystyrene bed by obliquely impinging impinging jets of gas, in: M. Nakagawa, S. Ludig (Eds.),Powders and Grains, Proceed
circular turbulent air jets, J. Hydraul. Res. 40 (2002) 709716, http://dx.doi.org/10. ings of the 6th International Conference on Powders and Grains 2009, pp. 14.
1080/00221680209499917. [67] M. Oda, I. Kotshikawa, Effect of strength anisotropy on bearing capacity of shallow
[34] M. Gunal, A. Guven, Prediction of ow patterns in scoured beds caused by sub footing in a dense sand, Soils Found. 19 (1979) 1528.
merged horizontal jets, Can. J. Civ. Eng. 33 (2006) 4148, http://dx.doi.org/10. [68] P. Jop, Y. Forterre, O. Pouliquen, A constitutive law for dense granular ows, Nature
1139/l05079. 441 (2006) 727730, http://dx.doi.org/10.1038/nature04801.
[35] O.A. Karim, K.H.M. Ali, Prediction of ow patterns in local scour holes caused by tur [69] C.Q. LaMarche, J.S. Curtis, Cratering of a particle bed by a subsonic turbulent jet: ef
bulent water jets, J. Hydraul. Res. 38 (2000) 279287, http://dx.doi.org/10.1080/ fect of particle shape, size and density, Chem. Eng. Sci. 138 (2015) 432445.
00221680009498327.