You are on page 1of 11

This article was downloaded by: [200.87.111.

21]
On: 25 April 2014, At: 14:49
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Hydraulic Research


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tjhr20

Granular flows driven by gravity


a
Aronne Armanini
a
CUDAM-Department of Civil , Environmental and Mechanical Engineering, University of
Trento , Trento, via Mesiano, 77, I-38100 , Trento , Italy
Published online: 20 Apr 2013.

To cite this article: Aronne Armanini (2013) Granular flows driven by gravity, Journal of Hydraulic Research, 51:2,
111-120, DOI: 10.1080/00221686.2013.788080

To link to this article: http://dx.doi.org/10.1080/00221686.2013.788080

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Journal of Hydraulic Research Vol. 51, No. 2 (2013), pp. 111–120
http://dx.doi.org/10.1080/00221686.2013.788080
© 2013 International Association for Hydro-Environment Engineering and Research

Vision paper

Granular flows driven by gravity


ARONNE ARMANINI (IAHR Member), Professor, CUDAM-Department of Civil, Environmental and Mechanical Engineering,
University of Trento, Trento, via Mesiano, 77, I-38100 Trento, Italy.
Email: aronne.armanini@ing.unitn.it, aronne.armanini@unitn.it

ABSTRACT
Two-phase fluids, consisting of solid particles in water or in the air, are present in many fields of hydraulics and specifically in the
morpho-hydrodynamics of rivers and mountain streams. In particular, when these mixtures contain a high concentration of sediments, as in debris
flows or in snow avalanches, they must be treated with the methods of granular fluid mechanics. Most recent theories of granular fluid mechanics,
applied to the two-phase flows driven by gravity, are discussed first. Then some of the many unsolved issues on which presumably the research effort
Downloaded by [200.87.111.21] at 14:49 25 April 2014

will focus in the forthcoming years will be identified and outlined.

Keywords: Debris flow; granular flow mechanics; sediment transport; two-phase flow

1 Introduction Debris and mud flows are phenomena which have been known
for a long time to engineers in charge of mountain stream
Although the mixing of water and sediments characterizes the restoration, but the first attempts at systematic framing of these
morphology and dynamics of all water streams regardless of phenomena were more or less empirical descriptions and clas-
their size, we utilize different tools and conceptual methods to sifications without any physical basis. Nowadays there are still
deal with the dynamics of these mixtures. These flows, in fact, a great number of classifications (Jakob 2005). Oldrich Hungr,
present very different behaviours, depending on the sediment one of the main experts on these phenomena, in his recent book
concentrations. Even with reference to rivers characterized by “Debris-flow hazard and related phenomena” (Jakob and Hungr
relatively small volumetric concentrations (<∼ 10−2 ), we gen- 2005), identified at least 10 types of debris flows. Tamotsu
erally discern between bed load and suspended load and treat Takahashi, in his monograph “Debris flow” (Takahashi 1991),
these phenomena with different approaches. In a recent keynote identified four types and gives a specific rheology to each of
lecture to Riverflow 2010, Stephen Coleman observed that in them. In any of these classifications the rheological relationship
general for the suspended load we use an Eulerian approach, should be, unfortunately, known a priori.
while for the bed transport we adopt the Lagrangian point of From a physical point of view it is, then, an urgent task to make
view. As long as the problem is confined to fluvial sediment available new convincing conceptual and numerical tools, based
transport, which is characterized by relatively low concentra- on physical principles, to predict the hazard induced by these
tions, the question of sediment transport has been addressed catastrophic events and to optimize the strategy of defence. A
with a variety of formulas and ad hoc theories, albeit with sub- particularly promising approach is based on the hypothesis that
stantially different methods and approaches. When we consider debris flow could be viewed as a hyper-concentrated flow formed
regimes with much higher particle concentrations, such as debris by two fluids. One is the interstitial fluid, which follows the laws
flows, these approaches are completely unsuitable. Recently, of fluid mechanics, and the other is the granular fluid, represented
the renewed interest in hyperconcentrated flows, and in partic- by the solid phase, also with its specific rheological relationships.
ular the social pressure exerted by the increasing frequency of The scientific literature on granular fluids is abundant, espe-
catastrophic events of debris flows and snow avalanches, has cially on the industrial aspects of these fluids. The industries of
motivated the scientific community to search for a deeper under- pharmaceuticals, plastics, food and concrete frequently employ
standing of the mechanism of interaction between liquid and conceptual models that can be reasonably ascribed to the area
solid phases of geophysical flows. of granular fluid mechanics. In civil engineering, in addition to

Revision received 18 March 2013/Open for discussion until 30 October 2013.

ISSN 0022-1686 print/ISSN 1814-2079 online


http://www.tandfonline.com
111
112 A. Armanini Journal of Hydraulic Research Vol. 51, No. 2 (2013)

debris flows, snow and debris avalanches also fall into this cat- 2 Rheology of granular flows
egory. Few, but significant, are also the attempts to frame bed
load and suspended load as granular flows. 2.1 Theoretical frameworks and approximations
In this paper we will primarily focus on the hydrodynamic
The theory of dispersive pressure, proposed by Bagnold (1954),
approach to granular flows. The approaches based on the molec-
was probably the first conceptual approach to the mechanics of
ular dynamics simulations will not be considered as well as
granular flows. Although Bagnold’s experiments are affected by
the flow initiation due to the incipient failure of a granular
the suspicion that the results were distorted by the presence of
deposit, which lies at the borders between hydrodynamics and
secondary circulations within the rheometer in which the mea-
geotechnics.
surements were made (Hunt et al. 2002), this theory has the
The liquid phase of debris flows is usually composed of water,
advantage of great simplicity. For this reason it has been widely
treated as a Newtonian fluid with the Navier–Stokes equations.
applied to geophysical flows driven by gravity, and in partic-
It is also common to consider the mixture of water and small
ular to debris flows (Takahashi 1991). According to Bagnold,
particles (smaller than ∼30 μm), dispersed in the liquid, as a
the collisions between the particles, averaged over time, con-
homogeneous fluid. In this case we have to distinguish between
tribute to dispersive pressure in the flow field that substantially
silt and clay. If the particles are from the silt fraction and their con-
depends on the relative average distance between the grains. On
centration is low, then their presence will not affect the Newtonian
the basis of experimental measurements made in a viscosimeter
behaviour of the mixture. In the case of clay or mud, the dis-
with light spherical particles, Bagnold arrived at the conclusion
persed particles confer a Binghamian behaviour to the interstitial
Downloaded by [200.87.111.21] at 14:49 25 April 2014

that for a sufficiently high shear rate, the relationship between the
fluid, even if the clay content is relatively small (Coussot 1977,
dispersive pressure and the shear rate is quadratic, and in a 2D
Ancey 2007). Under the hypothesis that the particle dimensions
shear flow it is pg−disp = a cos φρs (ldp γ̇ )2 where dp is the size
are smaller than the infinitesimal control volume, the granular
of the particles, γ̇ is the shear rate, and φ is the dynamic friction
phase can be assimilated to a continuous medium with its own
angle, which according to Bagnold is substantially constant in
rheological properties due to the interactions between the grains.
this regime. The parameter l is the linear concentration, which
However, if the particle size is of the same order of magnitude as
represents the ratio between the diameter of the particle and the
the boundaries of the flow, then this simplification will not work.
average distance between the particles, and a is an experimental
If the two phases are treated as two fluids without any mass
constant. According to Bagnold’s model, the dispersive pressure
exchange between them, their dynamics is described by the same
depends, therefore, only on the shear rate (known as Bagnold
equations of mass and momentum conservation known as the
scaling). The dispersive pressure induces in turn (or rather is
Cauchy equations:
induced by) dispersive shear stress τ g−disp , related to the pressure
by a Coulombian relationship: τ g−disp = a sin φρs (ldp γ̇ )2 .
∂ρ β ∂(ρ β uβ ) It is easy to prove (Armanini et al. 2008) that Bagnold’s theory,
+ =0
∂t ∂xi applied to a gravity-driven flow of rigid grains in water, implies
β β β β (1) that the concentration is constant across the flow depth. This
∂ρ β ui ∂(ρ β ui uj ) β ∂Tij β
+ = ρ β gi + + Fi is an obvious inconsistency. In fact, experiments show that the
∂t ∂xj ∂xj
concentration assumes the maximum c∗ (random packing con-
centration) near the bed, and tends to zero near the free surface
where β = g, f represents, respectively, the granular phase and (Armanini et al. 2005). This limitation of the Bagnold theory is
the interstitial fluid; ρ g = cρs and ρ f = (1 − c)ρw are the den- probably due to the fact that he assumed the same velocity scale
sity of the granular and the fluid phase, respectively; c is the (ldp γ̇ ) and the same length scale for both the dispersive pres-
volumetric concentration; ρs is the material density of the grains sure and the shear stress. This limitation has been overcome with
β
and ρw is water density; ui is the generic component of the veloc- the introduction of the kinetic theory of granular flows (Chap-
β
ity vector of each phase; Tij is the tensor of the internal stresses man et al. 1971, Jenkins and Savage 1983, Lun et al. 1984,
β
of each phase; and Fi is the vector of the forces of interaction Jenkins and Richman 1985), derived from the kinetic theory of
between the two phases (because of the principle of reciprocity ideal gases and based on the concept of granular temperature
 = u i u i /3, in which the symbol  represents the average
g f β p p
Fi = −Fi ); gi is the component (per unit mass) of the mass
force acting on each of the phases, which in the present case is value of all particles that are in the control volume in a certain
instant (u i = ui − ui  is the fluctuating component of the par-
β p p p
the vector of the gravity acceleration g, namely gi = −g∂z/∂xi ,
p g−coll
where z represents the vertical rising direction. ticle velocity, ui  = ui is the average velocity component).
Below I briefly consider the most recent theories of granu- The volume dimensions should be sufficiently large compared
lar fluid mechanics, applied to the two-phase flows driven by with the particle size and the mean distance between particles
gravity. Then, some of the many unsolved issues on which pre- but sufficiently small with respect to the dimensions of the flow
sumably the research effort will focus in the coming years will be field under consideration (though the theory is almost always
identified. applied beyond these limits).
Journal of Hydraulic Research Vol. 51, No. 2 (2013) Granular flows driven by gravity 113

Table 1 Expressions of the different functions adopted in the kinetic energy theory equations, according to Lun and Savage (1986)

f1 = c(1 + 4cηp go )
√   
5 π 8 1 8 8
f2 = 1 + cηp go + cηp (3ηp − 2) + √ c2 ηp go
96ηp (2 − ηp ) 5 go 5 5 π
√   
25 π 12 1 12 2 4
f4 = 1 + cηp go + cηp (4ηp − 3) + √ c2 ηp g
16ηp (41 − 33ηp ) 5 go 5 π
12
f5 = √ c2 go (1 − ep2 )
π
(1 + ep )
ηp =
2
 
c −2.5c∗
go = 1 −
c∗
g−coll

μ = ρ s f 2 dp 

k = ρ s f 4 d p 
pg−coll = ρs f1 

Note: Other expressions, differing slightly, are available in the literature.


Downloaded by [200.87.111.21] at 14:49 25 April 2014

In the kinetic theory, the rheological constitutive relationship The rate of change of the kinetic energy (left side of Eq.2) is
g−coll
is also shear dependent (Bagnold scaling), i.e. Tij = balanced by (1) kinetic energy diffusion due to particle fluctua-
g g
−p g−coll
δij + μ g−coll
(∂uj /∂xi + ∂ui /∂xj ). The “collision” vis- tions (first term on the right), (2) work of the granular stresses
cosity μg−coll√ is scaled by the square root of the granular (second term on the right), and (3) dissipation rate due to the
temperature  (served here as a velocity scale) and √ by the non perfectly elastic collisions among particles (third term). The
grain size dp as a length scale, i.e.: μg−coll = ρs f2 dp . The expression for the diffusion coefficient √ k is similar to that for
same velocity scale is used to express the granular pressure, the viscosity coefficient, i.e. k = f4 ρs dp , and f5 is a func-
i.e. pg−coll = ρs f1 , where f1 and f2 are two parametrical func- tion of the radial distribution function go , concentration c, and the
tions, which depend on the particle concentration c (Table 1). inelasticity restitution coefficient of the particles ep (Table 1). In
Both functions are in turn proportional to the radial distribu- addition, it is important to note that the production and dissipa-
tion function go , which takes into account the fact that in the tion terms are scaled by the same length scale dp . This means that
determination of the probability of collisions between two parti- granular flows lack a scale separation (Goldhirsch 2003), unlike
cles, the net volume surrounding the first particle is reduced due the turbulent Newtonian fluid, in which the small-scale fluctua-
to the presence of the second particle (Carnahan and Starling tions have an internal scale depending on the viscosity, while the
1970). Various formulations have been proposed for this func- larger fluctuations are scaled by the mixing length scale.
tion. They are especially different at high concentrations where The limitations of the kinetic theory in relation to flows char-
the radial distribution function diverges, and the kinetic theory acterized by high particle concentrations become particularly
should lose its validity. One of these expressions is shown in severe when the granular flows are driven by gravity and occur
Table 1. on a loose bed composed of the same particles at rest. The flow in
The main difference between Bagnold’s rheology and the the proximity of this boundary is characterized by prolonged con-
kinetic theory is in the equation of state, which in this case is tacts (frictional regime), even though a certain degree of granular
not directly related to the shear rate. In addition to the mass and temperature is always present (Fig. 1).
momentum conservation, the balance equation for the granular
kinetic energy (here in terms of granular temperature; Jenkins and
Savage 1983) permits us to close the set of necessary equations:

 
3 ∂ g ∂
ρs c + uj
2 ∂t ∂xj
  
advection
   g g
∂ ∂ g−coll ∂ui δuj 1.5
= k + τij + − ρs f5 (2) Figure 1 Sketch of the instantaneous distribution of a submerged gran-
∂xj ∂xj ∂xj ∂xi dp
         ular flow driven by gravity. The boundary between the frictional and the
diffusion production dissipation collisional regimes is intermittent
114 A. Armanini Journal of Hydraulic Research Vol. 51, No. 2 (2013)

Figure 2 Sketch of submerged granular flow on a fixed wall (left) and on a loose bed (right)

In addition, the collisions could involve several particles at the


same time. The particle concentration at this erodible boundary
assumes its maximum value c∗ . The loose static bed is formed
because the solid concentration increases downward up to a value
(random packing concentration) at which the frictional shear
stress is so high that no granular motion is possible.
Dealing with “free surface” granular flows, it is important to
Downloaded by [200.87.111.21] at 14:49 25 April 2014

distinguish flows on a rigid wall (fixed bed) and flows on a loose


bed. In the second case, towards the bed all kinematic variables
(velocity, concentration, and granular temperature) asymptoti-
cally tend to their static values. This also means that the granular Figure 3 Instantaneous networks of the force chains. Reprinted with
shear stress tends to become shear-rate independent. In the first permission from Da Cruz et al. (2005). Copyright 2005 by the American
Physical Society. http://link.aps.org/abstract/PRE/v72/e021309*.
case, on the contrary, the same variables exhibit at the wall
finite gradients. In particular, the concentration near the wall may
exhibit positive gradients (Fig. 2). This behaviour is due to the (in analogy with the boundary layer theory), within which the
fact that the solid wall is much “stiffer” than a bed formed from proper rheology of the frictional regime is valid. The experiments
loose particles and therefore the coefficient of inelastic restitu- (Armanini et al. 2009) show that this separation is somewhat arbi-
tion of the solid wall is typically larger. This effect applies even trary and that the two regimes (frictional and collisional) are not
if the wall is formed by a solid plate with glued particles identical stratified, but coexist (Fig. 4) and are intermittent (Fig. 1), just as
to the particles of the flow. In particular, in a uniform channel in the buffer layer between the viscous sublayer and the logarith-
flow on a solid wall the free surface is parallel to the wall. How- mic region in the turbulent boundary layer. However, unlike the
ever, in the case of loose particles the slope of the free surface fluid boundary layer, in which each sub-layer is characterized by
is independent of the slope of the channel bottom (Fig. 2). In the its own specific length scales, the granular flows lack separation
case of a rigid wall, the particles colliding with the wall could of scales, since the particle diameter turns out to be the only
bounce more than those that collide with each other at a certain geometrical scale that dominates both collisional and frictional
distance from the wall, and this mechanism is responsible for the regimes.
reduction of the concentration in the vicinity of the rigid wall. This is likely one of the major reasons for the lack of a clearly
The rheology of the frictional regime has not yet found a con- layered structure of such flows. In general, it can be consid-
vincing solution, although different formulations of an empirical ered reasonable that stresses related to the two regimes can be
nature are available in the literature. The most frequently adopted summed, so that the total stress tensor of the granular phase can
constitutive models can be grouped into four classes (Lois et al. be expressed as
2005): (i) shear rate-independent models, based on a Coulombian
relationship between granular shear stress and granular pressure g g−coll g−fric
Tij = Tij + Tij (3)
through a dynamic friction angle that mainly depends on the
particle concentration; (ii) shear rate-dependent relationships, g−coll g−fric
where Tij and Tij are the collisional and frictional
mostly based on the extension of the kinetic theories; (iii) models
components of the granular stress tensor, respectively.
based on the idea of the force chains mechanism (Fig. 3), leading
to a non-local rheology; this scheme is mainly adopted to explain
the jamming transition; and (iv) models based on the Ginzburg– ∗ Readers may view, browse, and/or download material for tempo-
Landau differential equation treating the transition between fric- rary copying purposes only, provided these uses are for noncommercial
tional and collisional regimes as a phase transition phenomenon. personal purposes. Except as provided by law, this material may not
be further reproduced, distributed, transmitted, modified, adapted, per-
In almost all the models adopted in the literature, it is assumed formed, displayed, published, or sold in whole or part, without prior
that the frictional regime is confined within a layer of finite height written permission from the American Physical Society.
Journal of Hydraulic Research Vol. 51, No. 2 (2013) Granular flows driven by gravity 115

1 1

pg 0.8 t g12
0.8
pg−coll t g-coll
0.6 0.6
h h
0.4 0.4

0.2 0.2

0 0
0.0 0.1 0.2 0.3 0.4 0.00 0.05 0.10 0.15
pg/(rs gh) t g/(rs gh)

Figure 4 Distribution along the dimensionless normal distance η = x2 /h (h is the flow depth) from the bed of the total and collisional shear stresses
(right) and pressure (left) in a uniform channel flow of mixed spheres and water. Note the significant thickness of the region where the two regimes
coexist. The collisional regime dominates only in the proximity of the free surface (η → 1) (Larcher et al. 2007)

A widely used method of modelling of the frictional regime thin layer of particles moving down a slope, the angle at which the
Downloaded by [200.87.111.21] at 14:49 25 April 2014

is based on a visco-plastic rheology (Meruane et al. 2010): material begins to flow (static angle, θs ) is greater than the angle
at which the layer stops (dynamic repose angle, θ2 ). This effect
Tij
g−fric g−fric
= −pg−fric δij + τij (4) invokes the idea that in granular flows, the hysteretic behaviour
may play a significant role. The μ(I ) model is, therefore, a
binomial Columbian model in which the friction coefficients
where pg−fric represents the (non-deviatoric) pressure due to the
depend on the inertial parameter (Jop et al. 2006):
long-lasting contacts between the particles. As the frictional pres-
sure pg−fric depends on the particle concentration, it must vanish tan θs (Io + I ) + (tan θ2 − tan θs )Io
when the concentration is less than that at which the contacts τ g = pg (5)
Io + I
become instant. The deviatoric part of the tensor is expressed
through a Coulombian scheme, and so directly linked to the The model has been extended beyond this frame, giving the angle
g−fric
frictional pressure, i.e. |τij | = (1 − δij )pg−fric tan φ fric , where θ2 the meaning of the “friction angle at high I ”, the nature of
φ fric is the internal friction angle. which should be better investigated and in any case has to be
Different expressions have been proposed to link the pressure determined experimentally together with the parameter Io . As
or/and the friction angle to the particle concentration (Johnson the inertial parameter, I is a function of the shear rate, the rheol-
and Jackson 1987, Johnson et al. 1990). All these models contain ogy of this model is in some way also shear dependent. The same
from three to four empirical constants that have to be experi- Group GDR (Forterre and Pouliquen 2008) has coupled the rhe-
mentally obtained with an ex post method. These models very ological relationship (5) with an equation of state, consisting of
often lack convincing physical foundations and generality, per- a linear relationship between the inertial parameter and the parti-
haps with an exception of the model of Savage (1998), which is cle concentration c = cmax − (cmax − cmin )I , in which cmax is the
probably the most general of these approaches although it con- random packing concentration and cmin is the minimum concen-
tains a large number of empirical parameters. This model has not tration at which the model is valid. By substituting this equation
been widely used and its limits have to be further checked. of state into Eq. (5), it is easy to realize that this model can also be
The experimental evidence suggests that the shear-rate ascribed to the category of the Coulombian models in which the
dependency is preserved across the whole flow depth and thus friction angle depends on the solid concentration while the over-
needs to be taken into consideration (Armanini et al. 2005, all model requires four experimental parameters. Nonetheless,
Lois et al. 2005, Armanini et al. 2011). In this respect, it is this model has been applied to a variety of scenarios, particu-
worth mentioning that the μ(I ) model recently proposed by larly related to gravity-driven flows such as debris flows or dry
the GDR Milieux Divisès Group (GDR Midi 2004) is currently granular flows, rock avalanches, and powder snow avalanches
being widely adopted. This model is based on the parameter (Rognon et al. 2008b, Marks et al. 2011). The model was origi-

I = dp γ̇ / pg /ρs , called the inertial parameter, which repre- nally formulated for a granular fluid without the interstitial fluid

sents the ratio of two temporal scales: a micro-scale dp / pg /ρs and later extended to the granular flow with an interstitial fluid
that is the time required for a particle to fall the distance dp by (Cassar et al. 2005, Berzi and Jenkins 2008), including cohesive
the action of a pressure pg , and a macro-scale proportional to the materials (Rognon et al. 2008a).
inverse local shear rate γ̇ . As previously mentioned, the rheology of the frictional regime
The inertial parameter is the square of the Savage number remains one of the most important problems of the dynamics of
(Savage 1994). The model is based on the observation that for a granular fluids that is still unresolved.
116 A. Armanini Journal of Hydraulic Research Vol. 51, No. 2 (2013)

A vital observation related to the frictional regime is that the 30


granular temperature plays a significant role in this region too,
highlighting an area of new conceptual developments. In this 28

respect, it is worth mentioning a model proposed by Jenkins


26
(2007). His model is based on the extension of the kinetic theory

jd (°)
to dense granular flow, in which the length scale corresponds 24
to a cluster of particles rather than to a single particle size. The
interpretative picture given by Lois et al. (2005) in this regard 22
clarifies the physics of the frictional regime: “… a dense granular
flow of perfect hard grains is animated with intense rattling and 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
changes of the contact network orientation. The dynamics of the
c
system involves many minute events which generalize binary
Figure 5 Behaviour of the apparent friction angle ϕd = arctan(τ g /pg )
collisions: these are creation and opening of contacts, rearrange-
when neglecting the diffusive term in the kinetic energy balance, as a
ments of existing force network, and sudden changes in the nature function of particle concentration c for ep = 0.75
of the frictional contact.” This picture also inspired the model by
Armanini et al. (2011).
It is also worth mentioning that Aranson and Kramer (2002)
and Aranson and Tsimring (2006) introduced an equation for of momentum, thus considerably reducing the degree of com-
Downloaded by [200.87.111.21] at 14:49 25 April 2014

the generic first-order phase transition (i.e. Ginzburg-Landau plexity of the system of equations. Following this simplification
equation) to explain the transition from the solid phase to the we obtain:
gas (kinetic) granular flow, i.e.:
τ g−coll f2 f5
= (8)
2 pg−coll f1
∂ρ g ∂ρ ∂ ρ ∂F (ρ, δ)
+ ui =D − (6)
∂t ∂xi ∂xi ∂xi ∂ρ This result corresponds to Bagnold’s scaling for both granular
shear stress and pressure. The behaviour of the ratio τ g /pg ,
where the parameter ρ represents the ratio between the total thus calculated, is reported in Fig. 5 as a function of the par-
number of static (or persistent) contacts and the total number ticle concentration for the case of constant inelastic restitution
of contacts. This parameter controls the ratio between the static coefficient ep .
part of the stress tensor and the full stress tensor of granular The plot highlights the linear asymptotic behaviour for
flow. The parameter D is a suitable diffusion coefficient; and c → c∗ and c → 0; this trend explains the linear behaviour of
F(ρ, δ) is the free energy density, which is assumed to have two
local minima at ρ = 1 (solid phase) and ρ = 0 (granular fluid
phase), Aranson and Tsimring (2006). Their model has been
1.0
successfully applied to powder snow avalanches to explain an
instability phenomenon, observed by Goldfarbs et al. (2002) and kinetic
0.8
known as fingering. The model has not yet been applied to sub- theory
merged granular flows, but it could give interesting results for 0.6
such flows also. h
Bagnold

0.4
Coulomb
2.2 Simplified solutions
0.2
There is a strong need to find simplified solutions of the system
of equations for granular flows, especially for the applications of 0.0
the kinetic theory to debris flows. It is easy to verify that if, in a 5 10 15 20 25 30 35 40
j
uniform flow, we neglect the diffusive term of the energy balance
(Eq.2), then it is possible to find a relationship of proportionality Figure 6 Distribution along the dimensionless normal distance
η = x2 /h of the measured arctangent of the ratio between shear stress
between the granular temperature  and the square of the shear and pressure (apparent friction angle) in a uniform flow of mixed spheres
rate γ̇ 2 : and water (data from Armanini et al. 2009). It is possible to observe
the smooth transition from the pure Coulomb model on the loose bed
f2 (η = 0) to the collisional regime at the free surface (η = 1). The figure
= (dp γ̇ )2 (7)
f5 also shows the constant friction angle suggested by Bagnold for his
experiments (Bagnold 1954). Note that the apparent friction angle of
In this case, we could drop the granular temperature in the expres- the collisional regime is bigger than the corresponding angle of the
sion for the collision viscosity (Table 1) and hence in the budgets frictional regime
Journal of Hydraulic Research Vol. 51, No. 2 (2013) Granular flows driven by gravity 117

1 3 Interaction forces between granular fluid and


interstitial fluid
0.8

0.6
3.1 Critical issues
h
In principle, the interaction between interstitial fluid and granular
0.4
phase is relatively simple and can be related to two actions: (a)
0.2 Dffusion the buoyancy forces component, independent of the difference in
Production
Dissipation velocities between liquid and solid, and attributable to the fact
0 that a portion of volume is occupied by the solid instead of being
-0.08

-0.06

-0.04

-0.02

0.00

0.02

0.04

0.06

0.08
occupied by the fluid and (b) the drag and lift forces component,
Energy balance terms: (-) induced by the difference in velocity between the two phases.
Figure 7 Vertical distribution of the different terms composing the The general relation is
kinetic energy balance (Eq.2) measured in a uniform flow of spheres
and water (Armanini et al. 2009). All the terms are made dimensionless, f
f ∂pf ∂τij
by dividing them by ps h1/2 g 3/2 Fi = c −c − Di (9)
∂xi ∂xj 
   drag
buoyancy
Downloaded by [200.87.111.21] at 14:49 25 April 2014

the apparent friction angle in Fig. 6, when approaching the free f


Here, Fi is the interaction force exerted by the fluid phase on
surface. g f
the granular phase, and vice versa for Fi (i.e. Fi = −Fi ), and
g

The simplification k ≈ 0 is however problematic, since, as Di is the drag component exerted by the fluid on the particles
shown in Fig. 7, the diffusive term in the kinetic energy bal- (Iverson 1997).
ance (Eq.2) is of the same order of magnitude as both the In granular flows at high concentration (which correspond to
production and the dissipation, especially in the intermediate the fully developed debris flow), the terms related to the drag and
region in which the two regimes have more or less the same lift are usually negligible in the momentum balances, while in
significance. the energy balance their effect may be significant. These forces
However, the possibility of simplifying the energy balance become increasingly important with decreasing concentrations,
has not yet been fully explored, nor have all the limits been and are decisive in the bed load (Seminara et al. 2002) and the
investigated. It is probably still possible to find new simplified suspended particles transport (Greimann et al. 1999).
solutions that are useful for applications in the field. Other kinds Apart from the difficulty of the determination of the appro-
of simplified solutions are based on the a priori assumptions priate coefficients of lift and drag in the presence of clusters of
regarding some of the dependent variables of the system, which particles at high concentration, the problem becomes even more
usually include the volume concentration. complicated when we consider turbulent flow of the intersti-
tial fluid. In this situation, the particles introduce a turbulence
length scale comparable to the particle size, which is added to
2.3 Research challenges
the scale generated by the rough wall. The problem of turbu-
From the previous section, it is clear that a general framing of lence closure has been addressed with zero equation models
the problem of the rheology of granular fluids is still lacking, and empirical relationships for the mixing length, expressed as a
particularly for the frictional regime. While at least one con- function of the solid concentration (Takahashi 1991). Although
ceptual scheme for the collisional regime exists (based on a such models provided reasonable results, they have been heavily
micro-structural constitutive model), the rheological formula- empirical and therefore difficult to generalize. There have been
tions for the frictional regime are essentially empirical. The μ(I ) also several attempts to tackle the problem with more elaborate
model has shed new light on this topic, but many questions are closure models (one- or two-equation closure models), in par-
still open. This issue is probably the major challenge of research ticular with the k − ε model. In this case, however, the problem
on granular flows. of the formulation of the terms relevant to the mechanisms of
From an engineering point of view then, it is always interaction between the solid and (turbulent) fluid phases has not
challenging to find simplified solutions to complex problems. yet found a definitive solution. In the literature, there are dif-
The idea of neglecting the diffusive terms, as has been said, ferent approaches: some authors (see, e.g. Meruane et al. 2010)
represents a remarkable possibility. It could be used to derive consider the effects of drag by adding a specific production term
bulk relationships, useful in depth-integrated models applicable into the equations of the turbulent energy, while other authors
to field cases. (see, e.g. Hsu et al. 2003) consider this effect by adding a special
As mentioned in the previous section, the model based on the dissipation term into the turbulent energy balance.
first-order phase transition also deserves attention and an attempt It should be underlined that the presence of the interstitial fluid
to apply it to submerged granular flows is welcome. is involved in the energy balances for the granular phase (Eq.2). A
118 A. Armanini Journal of Hydraulic Research Vol. 51, No. 2 (2013)

simplified way to account for this effect is to modify the inelastic


restitution coefficient ep inside the coefficient f5 of Eq. (2), mak-
ing it dependent on the particle Stokes number (Armanini et al.
2009, Gollwitzer et al. 2012).

3.2 Research challenges

As we have seen, the problem of the interaction between the


fluid and granular phases still has several unexplained aspects,
especially the turbulence effects. These effects are increas-
ingly important when the solid concentration is reduced, as,
for example, in debris flows during the arrest phase and also
during the debris flow formation. The question becomes cru-
cial if we want to extend the granular flow approach to bedload
and suspended sediment transport. The problem in this case Figure 8 Typical stratification of successive layers of fine and coarse
relates mainly to the closures of the second- and third-order grains observed by Makse et al. (1997); the small white grains start first,
followed by the larger red grains (from Makse et al. 1997)
correlations between the fluctuating component of concentra-
tion and granular and water velocities. These terms are typically
Downloaded by [200.87.111.21] at 14:49 25 April 2014

modelled with diffusion concepts, in which the diffusion coeffi-


successive stratified layers (Fig. 8). This effect, which seems
cients are often expressed using relationships with questionable
likely to be induced by the fact that the particles exhibit different
phenomenology.
angles of repose, has been observed also for submerged material
(Samadani and Kudrolli 2001).
4 Sorting and segregation in granular flows
4.2 Research challenges
4.1 Mechanisms
In the literature there is a tendency to give ample space to the
An important aspect affecting mostly geophysical granular flows problems of granular mechanics, which have much more affin-
is the strong dispersion that characterizes their particle size dis- ity with industrial applications than with civil engineering. In
tribution, which in debris flows ranges from a few millimetres to addition, the majority of the theories developed on this topic
blocks larger than one meter (Takahashi 1991). are more appropriate to the collisional regime. For the frictional
As a consequence, during the flow the particles are segregated regime, the problem of sorted grain size distribution and asso-
in accordance with their size. The finer particles are subject to a ciated mechanisms of percolation of the finer fraction remain
mechanism of downward infiltration (percolation) into the void unsolved.
spaces in the underlying layers; accordingly in the upper lay- As is known, the expulsion of boulders in debris flows has
ers there is a progressive enrichment of particles of larger size important consequences in the dynamic impact on the struc-
(Bridgwater 1976, Ottino et al. 2000). The result is a reverse tures. From a mechanics point of view, it is very likely that this
grading, where larger particles are concentrated in the higher aspect also has consequences on the rheology of the debris flows.
layers. Indeed, it is reasonable to think that an enrichment of the fine frac-
In addition to this effect, there is a second mechanism, linked tion, where the shear stresses are greater, produces a decrease in
to the size segregation due to the imbalance of impact forces the global resistance of these flows. New experiments are nec-
between particles of different size, which favours the migration essary to better understand the sorting mechanisms but, apart
of the particles. The mechanism that governs this second type from the problem of the grain sorting, it is also urgent to clarify
of migration is not clear. According to some researchers (see, the possible scaling effects in the reduced scale experiments on
e.g. Takahashi 1991) this second mechanism is the main cause granular flows in general.
of the effect of expulsion of boulders and their accumulation in
the vicinity of the front observed in the debris flow. But there
is also the hypothesis (Savage and Lun 1988, Legros 2002) that 5 Final remarks on field applications
the direction of this second mechanism is per se independent of
the size of particles, but added to the infiltration process tends to A further question that relates to all aspects of granular fluid
facilitate the upward shift of particles of larger size. mechanics regards the application of the models to unsteady
The angle of repose of the particles also seems to have a and non-uniform flows. This issue concerns primarily the appli-
certain influence on the stratification of the particles. Makse et al. cation of the rheological constitutive relationships of granular
(1997) have observed that when a mixture of graded granular fluid mechanics to three-dimensional problems. Such applica-
material is poured into a container, the mixture tends to form tions in the literature are quite rare in relation to both laboratory
Journal of Hydraulic Research Vol. 51, No. 2 (2013) Granular flows driven by gravity 119

models and field studies, requiring real ability of constitutive Aranson, I.S., Tsimring, L.S. (2006). Patterns and collective
relations to simulate three-dimensional phenomena. In the future, behavior in granular media: Theoretical concepts. Rev. Mod.
more efforts should be invested in both laboratory tests and in Phys. 78, 641–692.
mathematical modelling concerning this aspect. Armanini, A., Capart, H., Fraccarollo, L., Larcher, M. (2005).
The situation with the 2D depth-integrated models is also Rheological stratification in experimental free-surface of
rather disappointing. In the field applications of civil and envi- granular-liquid mixture. J. Fluid Mech. 532, 269–319.
ronmental engineering, it is necessary to use depth-integrated Armanini, A., Fraccarollo, L., Rosatti, G. (2006). Two-
models, based on the equations that meet the shallow water crite- dimensional simulation of debris flows in erodible channels.
ria (Cao et al. 2006, Garegnani and Rosatti 2013). The equations Comput. Geosci. 35(5), 993–1006.
governing these models can be derived by the integration of the Armanini, A., Fraccarollo, L., Larcher, M. (2008). Liquid-
three-dimensional equations across the flow depth (Armanini granular channel flow dynamics. Powder Tech. 182,
et al. 2006, Greco et al. 2012). From the previous sections 218–227.
it seems reasonable, therefore, to use a two-phase approach, Armanini, A., Larcher, M., Fraccarollo, L. (2009). Intermittency
which reproduces the equations of conservation of the masses of rheological regimes in uniform liquid-granular flows. Phys.
and momenta of the solid and liquid phases (Garegnani et al. Rev. E 79(5), 051306.
2011, Rosatti and Begnudelli 2013). Armanini, A., Dumbser, M., Nucci, E., Larcher, M. (2013).
We must, unfortunately, note that even today in most applica- Dynamics of submerged gravitational granular flows. In
tions to debris flow and snow avalanches a homogeneous fluid Numerical methods for hyperbolic equations: Theory and
Downloaded by [200.87.111.21] at 14:49 25 April 2014

approach is still adopted, in which it is assumed that the flow is applications, 157–163, E. Vázquez-Cendón, A. Hidalgo, P.
composed of a mixture of fluid (water or air) and sediment that García-Navarro, L. Cea, eds. CRC Press – Taylor and Francis
form a homogeneous non-Newtonian fluid with constant density. Group.
For these models, the conservation of solid mass is guaranteed Bagnold, R.A. (1954). Experiment on a gravity-free dispersion
a priori by the hypothesis of constant density, so no specific dif- of large solid spheres in a Newtonian fluid under shear. Proc.
ferential equation is written for the balance of the solid mass Royal Soc., London, Series A 225(1160), 49–63.
(similar to the Exner equation). These models behave, there- Berzi, D., Jenkins, J.T. (2008). A theoretical analysis of free-
fore, as fixed bed models (see the difference between the fixed surface flows of saturated granular-liquid mixtures. J. Fluid
bed and the loose bed depicted in Section 2), that is, they are Mech. 608, 393–410.
governed by the Saint Venant equations. It is evident that these Bridgwater, J. (1976). Fundamental powder mixing mechanisms.
models are unable to simulate mass changes, which however are Powder Technol. 15, 215–236.
decisive during both the initiation phase of debris flows and the Campbell, C.S. (1990). Rapid granular flows. Ann. Rev. Fluid
arrest phase. In addition, these models cannot be used to verify Mech. 22, 57–92.
the efficiency of defense works based on the sediment retention Cao, Z., Pender, G., Carling, P.A. (2006). Shallow water hydro-
mechanisms, such as check dams. In these models, in fact, the dynamic models for hyper-concentrated sediment-laden flows
erosion/deposition processes cannot be predicted by definition. over erodible bed. Adv. Water Res. 29(4), 546–557.
To make fixed bed models able to simulate the arrest phase Capart, H., Young, D.L., Zech, Y. (2002). Voronoï imaging meth-
of debris flows, it is usual to assume artificially that the flow has ods for the measurement of granular flows. Exp. Fluids 32,
Binghamian properties (Herschel–Bulkley models). In this way, 121–135.
the model predicts the arrest of the flow en masse when the bed Carnahan, N.F., Starling, K. (1970). Equations of state for non-
shear stress assumes the value of the threshold stress. It is evident attracting rigid spheres. J. Chem. Phys. 51, 635–636.
that this is an artefact and that, especially in the absence of clay Cassar, C., Nicolas, M., Pouliquen, O. (2005). Submarine gran-
material, the arrest occurs through a more or less gradual process ular flows down inclined planes. Phys. Fluids 17, 103301.
of sedimentation of the solid phase. In addition, the Herschel– Chapman, S., Cowling, T.G. (1971). The mathematical theory
Bulkley parameters, suitable for calibrating the models, are of non-uniform gases. ed. 3. Cambridge University Press,
subject to very wide variations, and mostly are known only a pos- Cambridge, England.
teriori. This knowledge gap must be urgently filled through new Coussot, P. (1977). Mudflow rheology and dynamics. IAHR
innovative experiments and advanced numerical simulations. Monograph, Balkema, Rotterdam.
Da Cruz, F., Emam, S., Prochnow, M., Roux, J.N., Chevoi, F.
(2005). Rheophysics of dense granular materials: Discrete
References simulation of plane shear flows. Phys. Rev. E 72, 021309.
Forterre, Y., Pouliquen, O. (2008). Flows of dense granular
Ancey, C. (2007). Plasticity and geophysical flows: A review. J. media. Ann. Rev. Fluid Mech. 40, 1–24.
Non-Newtonian Fluid Mech. 142(1), 4–35. Garegnani, G., Rosatti, G. (2013). On the range of validity of the
Aranson, I.S., Kramer, L. (2002). The world of the complex Exner-based models for mobile-bed river flow simulations.
Ginzburg-Landau equation. Rev. Mod. Phys. 74, 99–143. J. Hydraulic Res. (accepted).
120 A. Armanini Journal of Hydraulic Research Vol. 51, No. 2 (2013)

Garegnani, G., Rosatti, G., Bonaventura, L. (2011). Free surface Larcher, M., Fraccarollo, L., Armanini, A. (2007). Set of mea-
flows over mobile bed: Mathematical analysis and numerical surement data from flume experiments on steady uniform
modeling of coupled and decoupled approaches. Comm. Appl. debris flows. J. Hydraulic Res. 45(Supp. 1), 59–71.
Ind. Math. 2(1), doi:10.1685/journal.caim.371. Legros, F. (2002). Can dispersive pressure cause inverse grading
GDR MiDi (2004). On dense granular flows. Eur. Phys. J. E 14, in grain flows? J. Sedimentary Res. 72(1), 166–170.
341–365. Lois, G., Lemaitre, A., Carlson, J.M. (2005). Numerical tests of
Goldfarbs, D., Glasser, B., Shinbrot, T. (2002). Shear instabilities constitutive laws for dense granular flows. Phys. Rev. E 72(5),
in granular flows. Nature 415, 302–305. 051303.
Goldhirsch, I., 2003. Rapid granular flows. Ann. Rev. Fluid Mech. Lun, C.K., Savage, S.B. (1986). The effects of an impact depen-
35, 267–293. dent coefficient of restitution on stresses developed by sheared
Gollwitzer, F., Rehberg, I., Kruelle, C.A., Huang, K. (2012). granular material. Acta Mech. 63, 15–44.
Coefficient of restitution for wet particles. Phys. Rev. E 86, Lun, C.K.K., Savage, S.B. Jeffrey, D.J., Chepurniy, N. (1984).
011303. Kinetic theories for granular flow: Inelastic particles in Cou-
Greco, M., Iervolino, M., Leopardi, A., Vacca, A. (2012). A ette flow and slightly inelastic particles in a general flow field.
two-phase model for fast geomorphic shallow flows. Int. J. J. Fluid Mech. 140, 223–256.
Sediment Res. 27, 409–425. Makse, H.A., Halvin, S., King, P.R., Stanley, H.E. (1997).
Greimann, B.P., Muste, M., Holly, Jr., F.M. (1999). Two-phase Spontaneous stratification in granular mixtures. Nature 386,
formulation of suspended sediment transport. J. Hydraulic 379–381.
Downloaded by [200.87.111.21] at 14:49 25 April 2014

Res. 37(4), 479–500. Marks, B., Rognon, P., Einav, I. (2011). Grainsize dynamics
Hsu, T.J., Jenkins, T.J., Liu, L.P. (2003). On two-phase sedi- of polydisperse granular segregation down inclined planes.
ment transport: Dilute flow. J. Geophys. Res. (1978–2012) J. Fluid Mech. 690, 499–511.
108 (C3). Meruane, C., Tamburrino, A., Roche, O. (2010). On the role
Hungr, O. (2005). Classification and terminology. In Debrisflow of the ambient fluid on gravitational granular flow dynamics.
hazards and related phenomena, 9–23, M. Jakob, O. Hungr, J. Fluid Mech. 648, 381–404.
eds. Springer Berlin Heidelberg. Ottino, J.M., Khakhar, D.V. (2000). Mixing and segregation of
Hunt, M.L., Zenit, R., Campbell, C.S., Brennen, C.E. (2002). granular materials. Annu. Rev. Fluid Mech. 32, 55–91.
Revisiting the 1954 suspension experiments of R.A. Bagnold. Rognon, P.G., Chevoir, F, Bellot, A., Ousset, F., Naaïm, M.,
J. Fluid Mech. 452, 1–24. Coussot, P. (2008a). Rheology of dense snow flows: Infer-
Iverson, R.M. (1997). The physics of debris flows. Rev. Geophys. ence from steady state chute-flow experiments. J. Rheology
35(3): 245–296. 52, 729–748.
Jakob, M. (2005). A size classification for debris flows. Eng. Rognon, P.G., Roux, J.N., Naaïm, M., Chevoir, F. (2008b). Dense
Geol. 79(3–4), 151–161. flows of cohesive granular materials. J. Fluid Mech. 596,
Jakob, M., Hungr, O. (2005). Debris-flow hazards and related 21–47.
phenomena. Springer Verlag, Berlin, Heidelberg, New York. Rosatti, G., Begnudelli, L. (2013). Two-dimensional simulation
Jenkins, J.T., Richman, M.W. (1985). Kinetic theory for plane of debris flows over mobile bed: Enhancing the TRENT2D
flows of a dense gas of identical, rough, inelastic, circular model by using a well-balanced generalized Roe-type solver.
disks. Phys. Fluids 28, 3485–3294. Comput. Fluids 71, 179–195.
Jenkins, J.T., Savage, S.B. (1983). A theory for rapid flow of Samadani, A., Kudrolli, A. (2001). Angle of repose and seg-
identical, smooth, nearly elastic, spherical particles. J. Fluid regation in cohesive granular matter. Phys. Rev. E 64(5),
Mech. 130, 186–202. 051301.
Jenkins, J.T. (2007). Dense inclined flows of inelastic spheres. Savage, S.B. (1994). The mechanics of rapid granular flows. Adv.
Granular Matter 10, 47–52. Appl. Mech. 24, 289–366.
Jenkins, J.T., Hanes, D.M. (1998). Collisional sheet-flow of Savage, S.B. (1998). Analyses of slow high-concentration flows
sediment driven by a turbulent fluid. J. Fluid Mech. 370, of granular materials. J. Fluid Mech. 377, 1–26.
29–52. Savage, S.B., Lun, C.K.K. (1988). Particle size segregation in
Johnson, P.C., Jackson, R. (1987). Frictional-collisional constitu- inclined chute flow of dry cohesionless granular solids. J. Fluid
tive relations for granular materials, with application to plane Mech. 189, 311–335.
shearing. J. Fluid Mech. 176, 67–93. Seminara, G., Solari, L., Parker, G. (2002). Bed load at
Johnson, P.C., Nott, P., Jackson, R. (1990). Frictional-collisional low Shields stress on arbitrarily sloping beds: Failure of
equations of motion for particulate flows and their application the Bagnold hypothesis. Water Resour. Res. 38(11), 31-1,
to chutes. J. Fluid Mech. 210, 501–535. doi:10.1029/2001WR000681.
Jop, P., Forterre, Y., Pouliquen, O. (2006). A constitutive law for Takahashi, T. (1991). Debris flow. IAHR Monograph Series,
dense granular flows. Nature 441, 727–730. A.A. Balkema, Rotterdam-Brookfild.

You might also like