Sunteți pe pagina 1din 12

Computers and Mathematics with Applications 69 (2015) 89–100

Contents lists available at ScienceDirect

Computers and Mathematics with Applications


journal homepage: www.elsevier.com/locate/camwa

VOF/FVM prediction and experimental validation for


shear-thinning fluid column collapse
Nelson O. Moraga a , Luis A. Lemus b , Mario A. Saavedra c ,
Roberto A. Lemus-Mondaca d,∗
a
Departamento de Ingeniería Mecánica, Universidad de La Serena, Benavente 980, La Serena, Chile
b
Departamento de Ingeniería en Construcción y Prevención de Riesgos, Universidad Santa María, Av. Santa María 6090,
Viña del Mar, Chile
c
Departamento de Ingeniería Mecánica, Universidad de Santiago de Chile, Av. Bernardo O’Higgins 3363, Santiago, Chile
d
Departamento de Ingeniería en Alimentos, Universidad de La Serena, Av. Raúl Bitrán Nachary 1305, La Serena, Chile

article info abstract


Article history: Dam break problems for non-Newtonian fluids can be found in sudden collapse of mine
Received 30 December 2013 tailings, snow avalanches, debris and lava flows, and casting solidification. A numerical
Received in revised form 15 August 2014 simulation and experimental validation of collapse of a shear-thinning fluid column with
Accepted 27 November 2014
a high viscosity is presented. The 2D fluid mechanics, described in terms of the non-linear
Available online 18 December 2014
coupled continuity and momentum equations, was solved by the finite volume method
(FVM) with the Pressure Implicit with Splitting of Operators (PISO) coupling algorithm and
Keywords:
Shear-thinning fluid
the volume of fluid method (VOF). The shear-thinning fluid column that collapses was
CFD described by the rheological model of Carreau–Yasuda. The numerical results obtained
Carreau–Yasuda model for the instantaneous position of the free surface were validated with 7% accuracy in
VOF method comparison with experimental measurements, determining that when a small container
is used the walls affect the transport of fluid, causing a creeping flow.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The comprehension of fluid dynamics to describe the space–time evolution of mobile interfaces and free surfaces are of
great interest for many engineering problems in manufacturing processes such as flow in canals, filling moulds and deposits,
pulp transport, concrete mixing, transfer of liquids confined in deposits, fluid spills, mud slides, debris and lava flows,
and snow avalanches, among other applications [1–3]. Also, there are some applications that evaluate the fluid mechanics
behaviour and moving front by using analytical methods [4,5] and numerical techniques [6,7] based on physic experiments,
causing the flow characteristics to change due to the closeness of the walls.
Numerous studies have been used to describe the rheological behaviour of non-Newtonian fluid materials. The examples
found include Ostwald-de Waele, Binhgam, and Herschel–Bulkley models and other more complex rheological models such
as the Casson, Ellis, Cross, Van Wazer, Carreau and Carreau–Yasuda models [8]. Shamekhia and Sadeghy [9] studied the
non-Newtonian fluid behaviour inside a cavity obeying at Carreau–Yasuda rheological model. These researchers reported
that this model revealed the strong effect of the shear-thinning behaviour and flow kinematics within the cavity. In addition,
the results showed a suitable agreement with other results published in the literature. Balan et al. [10] carried out physical

∗ Corresponding author. Tel.: +56 512204446; fax: +56 51 2204446.


E-mail address: rlemus@userena.cl (R.A. Lemus-Mondaca).

http://dx.doi.org/10.1016/j.camwa.2014.11.018
0898-1221/© 2014 Elsevier Ltd. All rights reserved.
90 N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100

Nomenclature
D Deformation rate (1/s)
g Gravitational acceleration (m/s2 )
k Thermal conductivity (W/m K)
K Consistency index (Pa s) (Pa sn )
L Height of the cavity (m)
n Power-law index (dimensionless)
p Pressure (Pa)
Rg Universal gas constant (kJ/mol K)
t Time (s)
u, v Velocity components
WL Left wall
WR Right wall
WB Bottom wall
x, y Coordinates

Greek symbol
φ Current variable
γ̇ Shear rate (1/s)
η Apparent viscosity (Pa s)
µ Dynamic viscosity (Pa s)
ρ Density (kg/m3 )
τ Shear stress (Pa)
ν Kinematic viscosity (m2 /s)

Subscripts
i, j Node position
in Inlet
ref Reference property
f Fluid
s Shear-thinning fluid
o Initial
exp Experimental data
num Numerical data
max Maximum
min Minimum

Superscripts
k − 1, k Iteration number
→ Vector
⇒ Tensor

and numerical experiments between immiscible Carreau–Yasuda fluid flows inside Y-type bifurcations in micro-channels.
The physical experiments performed, were validated using VOF technique implemented in a commercial code (Fluent v.6).
The numerical results adequately described the vortexes formation in the interface vicinity. Janßen et al. [11] simulated free
surface flow problems by volume of fluid based model with FVM and Piecewise Linear Interface Reconstruction algorithm.
They validated the model by some benchmarks such as: breaking dam and free falling jet. Thus, all validation and benchmark
tests confirm the accuracy of the proposed model.
A large variety of numerical techniques have been developed in recent years to analyse flow problems with moving
inter faces and free surfaces. The calculation, definition and evolution of moving fronts can be classified into two groups:
fixed grid and mobile grid techniques. The deformable-spatial-domain/stabilized space–time method [12], finite calculus
formulation [13] and dynamically adaptive quad tree grids [14] are mobile grid techniques using the methods of finite
elements and finite volumes. Although these methods have been shown to describe favourably the moving material front,
the computational cost of remeshing and the difficulty of estimating the velocity of the motion of the grid have enhanced
the development of fixed grid formulations such as the midpoint value of that interface function using the finite elements
method [15], and volume of fluid using the finite volume method [16].
N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100 91

Fig. 1. Physical situation for collapse of a shear-thinning fluid column: (a) t = 0 and (b) t > 0.

The numerical difficulty for tracking the interface position in time, in addition to the problem of discretization of the
transport equation for the volume fraction using fixed grid techniques, causes numerical interface diffusion problems.
To improve the solution of the multiphase problem various schemes have been proposed. Within the framework of the
finite volume method, high resolution schemes have been developed [17,18]. Waclawczyk and Koronowicz [19] presented
efficient results to use ‘‘Compressive Interface Capturing Scheme for Arbitrary Mesh (CICSAM)’’ high resolution scheme to
discretize the convective terms of the volume fraction transport equation.
The objective of this paper is to present novel results related to describe the sudden fluid column collapse of shear-
thinning fluids with the Carreau–Yasuda model inside a square cavity. Mathematical modelling and computer simulations
are accomplished by using the Finite Volume Method (FVM) with the sequential coupling PISO algorithm, the VOF
two immiscible phases method and the CICSAM high resolution scheme. Accuracy of numerical results for tracking the
instantaneous position of the moving front is calculated by comparison with the experimental data. The obtained results
can be useful in many industrial applications such as: floods from mine tailing [20] and non-Newtonian metals and alloy
casting solidification [21], and liquid food injection in containers.

2. Experimental setup

2.1. Experimental model

The experiment corresponds to the collapse of a shear-thinning fluid column caused by opening the gate that contains it
(Fig. 1). The working device was a transparent plastic tank of the size 4L × 4L (L = 6.7 cm) as it is shown in Fig. 1(a). The box
filled shear-thinning fluid was separated by a 1 mm thick plastic gate that was removed vertically. The domain is initially
separated into zones A and B by a gate, and the top surface is exposed to the atmosphere. In zone A is the shear-thinning fluid
(Fluid A: viscous shampoo), where it coexists with the Newtonian fluid (Fluid B: air) in the zone B. The fluid A is contained
between the left wall WL and the gate at a distance L, and it has a column height of 2L. The gate is removed instantaneously
and fluid A begins its sudden collapse, moving along the lower wall WB and occupying part of the space of zone B (Fig. 1(b)).
The whole bottom back face was covered with blue graph paper to follow up the mobile front, while the side faces had
a 1 cm of paper from the bottom towards the front face. The left side of the separation (zone A) was initially filled with
green-coloured shampoo that was represented by the rheological properties of the Carreau–Yasuda model, and the right
side (zone B) had air under ambient conditions. A black marker was used to draw the fluid outline on the front face, in order
to assist in controlling the initial filling.

2.2. Image capture device

A light box system was implemented to assist in viewing and post-processing the data, placing lamps on the upper,
right and left sides of the container. The data were recorded on a Nikon D40 reflex photographic camera placed on a
tripod directly in front of the container. Images were recorded using the burst mode at a rate of 2 pictures per second. The
reported results correspond to the average of five experiments in order to minimize possible errors due to the gate opening
speed and measuring difficulties caused by light diffraction. Similar experiments and simulations have been carried out by
other authors working with water and shampoo, but assumed to be Newtonian fluids, as those results reported by Hu and
Sueyoshi [3] and Cruchaga et al. [6].
92 N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100

3. Mathematical model

3.1. Governing equations

The spilling of the shear-thinning fluid column inside a cavity with air, caused by opening the gate that holds is studied
considering the gravitational effect, insignificant surface tension, laminar and isothermal flow and incompressible and
immiscible fluids. In addition, walls are considered devoid of slippage, so in this way the forefront region is represented
as a real fluid. Then, the fluid dynamics is described by the coupling of the partial differential mass conservation and linear
momentum equations:

∂ρi ∂ (ρ u)i ∂ (ρv)i


+ + = 0; i = air , shampoo (1)
∂t ∂x ∂y
∂ (ρ u)i ∂ (ρ u)i ∂ (ρ u)i ∂ pi ∂ ∂ (ρ u)i ∂ ∂ (ρ u)i
   
+ ui + vi =− + η + η (2)
∂t ∂x ∂y ∂x ∂x ∂x ∂y ∂y
∂ (ρv)i ∂ (ρv)i ∂ (ρv)i ∂ pi ∂ ∂ (ρv)i ∂ ∂ (ρv)i
   
+ ui + vi =− + η + η + ρg (3)
∂t ∂x ∂y ∂y ∂x ∂x ∂y ∂y
where air is assumed to be a Newtonian fluid and hence that η = µair when i = air, and η is the apparent viscosity of sham-
poo, as described by the Carreau–Yasuda model by Eq. (8), in Section 3.3.
The effects of the contact between the shampoo and the walls of the domain are considered in this study, so the
description of the viscous flow and interface position considers walls without slippage. The shear stress tensor can be written
in terms of the shear-thinning fluid apparent viscosity and the deformation rate tensor, which is defined by means of its
second invariant.

∂ us ∂vs
  
⇒ 1⇒ ⇒ 1
τ = −η(γ̇ )γ̇ ; γ̇ = D : D; D = + . (4)
2 2 ∂x ∂y

3.2. VOF method

The free surface is modelled by means of the volume of fluid, fraction method VOF [23]. The principle of the model is based
on the fact that the volume occupied by a fluid cannot be occupied by another fluid at the same instant of time. Therefore,
a cell with a discrete volume contains a fraction, f , of each material. The sum of the fractions of the fluids in a cell must be
equal to 1 for every time and space. Since the fluids are moving, the fluid interface is also modified, forcing the fluid fractions
to be updated. The fractions of fluid in a cell can be classified into three types: if the cell is filled with a fluid i, fraction fi = 1;
if the cell does not contain fluid i, fraction fi = 0; and if the cell shares its volume with two or more fluids besides fluid i, the
fraction of i takes a value 0 < fi < 1.
To represent and update the interface of the material front, use is made of the transport equation for pure convection:

∂ fi ∂ (ufi ) ∂ (v fi )
+ + = 0. (5)
∂t ∂x ∂y
Once the equation of the material front (Eq. (5)) is solved, the interface between the participating fluids can be
determined, implying the conservation of mass of one phase in the mixture. In a multiphase problem Eq. (5) is solved for
n − 1 phases, and the remaining phase is solved by means of Eq. (6).
n

fi = 1. (6)
i=1

The assignment of properties to solve Eqs. (1)–(3) using a fixed grid is done by weighting the properties of the fluid using
the volume fraction distribution for each fluid, calculated using Eqs. (5) and (6). In this way, the fluid properties for the
mixture of the two phases studied can be described from the following constitutive relations:

ρ = (1 − f )ρ1 + f ρ2 ; η = (1 − f )η1 + f η2 . (7)

3.3. Shear-thinning fluid model

The fluids participating correspond to shampoo for the collapsing fluid column and air for the surrounding fluid. The air
is a Newtonian fluid, with properties given by ρa = 1.210 kg/m3 and ηa = 1.812 × 10−5 Pa s. The viscous fluid, shampoo
with a density ρs = 1.030 kg/m3 , is represented using the shear-thinning fluid model of Carreau–Yasuda type (Table 1 [22]),
whose constitutive equation is Eq. (8). The model has five parameters: where ηo corresponds to the Newtonian viscosity at
N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100 93

Table 1
Experimental values for shear stress, shear rate and viscosity for shampoo.
γ̇ (s−1 ) τ (Pa) η (Pa s)
0.62 23.22 37.211
0.70 27.14 38.771
0.79 31.07 39.229
0.83 34.99 42.156
1.17 42.85 36.623
1.31 48.73 37.198
1.70 62.47 36.747
2.26 78.18 34.592
2.70 89.95 33.310
3.77 105.65 28.020
4.61 109.58 23.770
8.92 113.50 12.720
23.20 121.36 5.230
24.70 125.28 5.070
41.35 131.17 3.170
83.49 137.06 1.640
97.19 142.95 1.470
99.97 142.95 1.420
136.07 150.80 1.100
138.75 150.80 1.108

Fig. 2. Relationship between viscosity and strain rate of shear-thinning fluid model for rheological parameters used in computation.

low deformation speed values, η∞ is the Newtonian viscosity for high deformation speed values, λ is a time constant, n
corresponds to the power law model parameter, and a is a dimensionless constant.
 n−1
η(γ̇ ) = η∞ + (η0 − η∞ ) 1 + (λγ̇ )a a .

(8)

The application of the model and the proper constants for its definition were obtained from the experimental develop-
ment made in [4], where the set of experimental data has been fitted by using non-linear least squares. The experimental
values obtained allowed the estimation of the thinning constants for the model whose values are: ηo = 38.97 Pa s, η∞ =
0 Pa s, λ = 0.3248, n = 0.0097 and a = 2.95. The relation between viscosity and deformation rate is shown in Fig. 2.

4. Computational procedure

4.1. Discretization of the volume fraction transport equation

The coupled governing equations (Eqs. (1)–(3)), with the initial and boundary conditions, were solved using finite volume
method, PISO coupling algorithm and volume of fluid method by Ansys/Fluent v.14 software. The discretization of the
diffusive terms was carried out using second order interpolation functions, while the convective terms were calculated by
a third order scheme based on the original Monotone Upstream-Centered Schemes for Conservation Laws (MUSCL), which
combines the schemes of finite centred differences and second order upwind.
A correct discretization of the volume fraction transport equation (7) is essential in the simulations of multiphase flows.
The low resolution schemes used in many numerical discretizations for the convective terms have caused high numerical
diffusion and oscillations. The CICSAM high resolution scheme has two components: the HYPER-C and Ultimate-Quickest
94 N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100

Table 2
Effects of time step and grid size on computational time (h).
Time step Number of nodes
40 × 40 80 × 80 100 × 100

0.0001 8 ∼18 ∼24


0.001 5 10 ∼12
0.01 3.5 5 6.5

(UQ) schemes. The HYPER-C scheme (Eq. (9)) has limitations when the fi gradient changes. If the interface is tangential to the
direction of flow the scheme tends to deform that state artificially. The UQ scheme (Eq. (10)) has a less restrictive behaviour.

φ̃D  φ̃D < 0, φ̃D > 1



  :
φ̃fCBC = φ̃D (9)
min
 1, : 0 ≤ φ̃D ≤ 1
Cf

φ̃D  φ̃D < 0, φ̃D > 1



  :
φ̃fUQ = 8Cf φ̃D + (1 − Cf )(6φ̃D + 3) (10)
min
 , φ̃fCBC : 0 ≤ φ̃D ≤ 1.
8

In the CICSAM scheme a smooth change between both schemes (Eqs. (9)–(10)) is defined by means of the mixing factor,
0 ≤ γf ≤ 1 (Eq. (11)), with kγ a constant introduced to control the domain of the different schemes, with kγ = 1. The
value of γf depends on the angle θ (Eq. (12)) defined between the unit vector normal to the interface n ⃗ = ∇φD / |∇φD |
and the parallel unit vector determined by the line between the centres of the donor cells D and acceptor cells A, given by
⃗ = DA
d ⃗ /|DA
⃗ |. When the position of the interface is normal to the direction of flow, γf = 1, and the HYPER-C scheme is used,
while if the interface has a tangential orientation, γf = 0, and the UQ scheme is used. This is expressed in Eq. (13).

1 + cos(2θf )
 
γf = min kγ ,1 (11)
2
 
θf = arccos d⃗ n⃗ (12)
 

φ̃f = γf φ̃fCBC + (1 − γf )φ̃fUQ ; 0 ≤ γf ≤ 1. (13)


The nodes were located in specific spatial locations in order to compare the results calculated with the VOF/FVM and
experimental data measured for the instantaneous position of the moving front. The convergence criterion applied to
stop the temperature calculations at each time step, with a 99% convergenceof nodes, was  based in a maximum value
for the difference in the calculated value allowed at two successive iterations φik,j − φik,−
j
1
≤ 10 −6
. Computations were
accomplished in an Intel⃝ R
Core⃝ R
2 Duo T5750/2.0 GHz personal computer with 3.0 GB of RAM. In addition, the fit quality
was evaluated by means of the Relative Error (Eq. (14)), which compares the numerical values calculated by the finite volume
method and the experimental data.

φexp − φnum
RE (%) = . (14)
φmax − φmin

5. Results and discussion

Several calculations changing the mesh from 40 × 40, 80 × 80 to 100 × 100 nodes and the time step, from 0.0001 to
0.01 s, were performed. In each case the variations in the velocities and instantaneous position of the free surface were
evaluated in terms of convergence and computational time. Due to this previous analysis, it was found that the results have
a deviation lower than 2% when a two-dimensional staggered grid with 80 × 80 nodes was used for the spatial discretization,
with a constant time step of 0.001 s (Table 2).
The VOF technique together with the high resolution CICSAM scheme has been applied to describe the collapse of a
column of a shear-thinning fluid in a container. The comparison between experimental and numerical results, describing
the fluid dynamics of the immiscible phases two and the time-spatial position of the shear-thinning fluid–air interface in
terms of volume fraction is shown in Figs. 3 and 4, from t = 0 to t = 40 s. It is seen that at the initial time t = 0 s, the filling
with shear-thinning fluid in zone A corresponds to that bound by the conditions preset for the test, with small air bubbles
appearing in the experimental record.
The image recorded at t = 0.5 s shows the opening of the gate above the level of the initial fluid, determining that
the extraction mean velocity is 0.12 m/s. The effect of the shear-thinning fluid viscosity is seen clearly at that instant from
the trail of material attached to the gate, which falls on the column after the collapse. This effect is not detected in the
N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100 95

Fig. 3. Interface positions at different instants in the bottom wall. Numerical (left side) and experimental result (right side).

numerical solution because the gate is assumed to be removed instantaneously at the beginning of the process. However,
the phase profile is not altered, presenting a shape of similar characteristics between the numerical and the experimental
results. Thus, the column will collapse and eventually come to rest. The initial stages of the flow are dominated by inertia
forces with viscous effects increasing as the shear-thinning fluid comes to rest. On such a large scale, the effect of surface
tension forces is insignificant. The complexity of the velocity field time evolution occurring at different stages of the collapse
phenomena was successfully captured using the VOF technique and the CICSAM scheme.
The results at t = 5 s and t = 10 s show a layer of fluid adhering to the left wall that starts approximately at the initial
upper position (Fig. 3). This phenomenon is due to the high viscosity of the shear-thinning fluid and to the influence of the
wall on the motion of the flow in a tank of the size used. The computer simulation captures efficiently the adherence layer
reported experimentally. The wall effects are seen in the same way on the front face of the container, where the front zone
of the advancing shear-thinning fluid lags behind with respect to the advance on the bottom face. The numerical modelling
captures the front zone of the shear-thinning fluid as it was seen experimentally, where the highest speed is achieved at
the nose of the front and not in the lower zone in contact with the wall. In this way it has been verified that the viscous
96 N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100

Fig. 4. Interface positions at different instants in the right wall. Numerical (left side) and experimental result (right side).

shear-thinning fluid is accurately described by the Carreau–Yasuda model and the wall shear stress calculated induces an
opposition to the motion.
Fig. 4 shows the position of the interface at different instants, from the time the flow reaches the right wall of the container
and starts rising on it, until it reaches the quasi-stationary state. The results at t = 22 s shows the instant at which the front
reaches the right wall. The experimental result shows that the viscous shear-thinning fluid fills the bottom right corner,
while the numerical model predicts that there is an air interstice. The simulation for the succeeding instants of time causes
the gradual filling of the air globe, although it is present until the time t = 40 s. The phase contours, calculated after the
shear-thinning fluid reaches the right wall, have a shape in agreement with that found experimentally for the same instants
of time. A curvature of the shear-thinning fluid and air interface for the upper face with a negative slope is seen, causing
that the height of the column on the right side being lower than on the left.
As time goes by, the free inclined surface tends to balance with the horizontal, the viscous fluid layer attached to the left
wall starts getting thinner and to drop at an almost imperceptible rate.
Fig. 5 shows the evolution of the shampoo front in time, determined experimentally and numerically, for the left wall
(Fig. 5(a)), the right wall (Fig. 5(b)) and the lower wall (Fig. 5(c)). The position of the front on the left wall drops almost
linearly in time. The right wall is not reached by the mobile front until t = 22 s. At that instant the numerical results predict
N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100 97

Fig. 5. Shear-thinning fluid front time evolution on: (a) left wall, WL , (b) right wall, WR and (c) bottom wall, WB .

Table 3
Experimental validation for the collapse of a shampoo column in open cavity.
Time Position front (cm)
WL WB WR
Sim. Exp. Sim. Exp. Sim. Exp.

5 3.11 3.1 0 0 4.62 4.88


10 3.06 3.05 0 0 5.61 5.95
20 2.96 2.95 0 0 6.58 6.6
30 2.87 2.85 0.59 0.60 6.62 6.7
40 2.79 2.75 0.71 0.75 6.66 6.7

the contact with the front, as it was verified in the experiment making the values of the position to increase monotonically
with a concave trend. The shampoo starts moving along the bottom wall, reaching the right wall at t = 22 s, a situation that
is verified with the values for the right wall. The motion of the front in time presents an increasing monotonic trend with
a convex shape, converging at a value equal to the length of the container. A summary of the numerical and experimental
values of the position of the front on the walls is given in Table 3. The comparison between the data simulated with the
experimental data measured at the three positions showed a maximum relative error of 1.5% (WL ), 5.3 (WB ) and 5.7% (WR ).
In addition, based on the criterion of E% < 10%, the simulated values showed a good fit quality to the experimental data.
98 N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100

Fig. 6. Transient evolution of shear-thinning fluid front for H = 2L before of contact with right wall.

Fig. 6 shows a dimensionless description for the motion of the front, calculated with Carreau–Yasuda model, for the
viscous shear-thinning fluid during the collapse. The dimensionless length, x∗ , corresponds to √ x/L, where L is the width
of the initial column and x is the position changing in time, and the dimensionless time t ∗ is t 2g /L. In Fig. 6 it can be
observed that the increase in process time leads an increase in x∗ dimensionless length, which have been also observed
experimentally by Martin and Moyce [23] and Koshizuka et al. [24]. However from t ∗ equal to 700 s the x∗ value reaches an
asymptotic constant value equal to 3.9. Martin and Moyce [23] exposed the difficulty to determine the experimental exact
position of the leading edge. A thin layer of shear-thinning fluid column shoots over the bottom and the rest of the bulk
flow follow behind it. The accurate numerical results have been obtained by applying intensive sharpening on the moving
interface. From the above discussion the numerical values reported are very similar to the experimental data due to the
use of the VOF technique and to the use of the CICSAM scheme, with a fine mesh of 80 × 80 nodes. Similar results have
been obtained by other authors working with finite element method as reported by Kačeniauskas et al. [25] and Cruchaga
et al. [6]. Thus, the accuracy of the obtained numerical results is strongly influenced by the numerical parameters, but it is
almost independent of the resolution of the mesh, when sufficiently dense finite volumes meshes are used.
An analysis of the fluid dynamics of both fluids and the mixture is made by means of the distribution of velocity vectors,
as shown in Fig. 7 for different times between t = 5 and t = 26 s. The velocity vectors just at the interface of the mixture
when the front is displaced on the bottom wall are shown on the left. Auxiliary lines have been drawn to show the vectors
in the L/4, L/2, 3/4L and 9/10L positions. It is seen how the mobile front loses velocity as it is displaced, as represented
graphically by the decrease of the length of the vectors for each position. The magnitude of the velocity at instants t = 5, 10,
22, 26 s, corresponds to v = 0.0032, 0.0017, 0.0005, 0.0004 m/s, respectively. The velocity vector distribution calculated for
the mixture (Fig. 7, right) determines that the air velocity is v = 0.011, 0.007, 0.003, 0.003 m/s, showing that it is one order
of magnitude larger than the velocity of the shampoo. It is also seen that a counter clockwise (CCW) vortex is created after
the collapse of the viscous fluid column, moving in the direction of the motion near the upper free surface and inducing the
viscous flow near the interface that will be moved. The vortex is transferred spatially at t = 5, 10, 22, 26 s from the positions
(2.82, 2.04), (3.61, 2.09), (4.33, 0.237), (4.37, 2.53) cm, respectively, in the Cartesian coordinates (x, y).

6. Conclusion

The shear-thinning fluid column collapse as a result of the sudden opening of the gate that retains it has been studied
numerically using the VOF/FVM and verified experimentally. The results of the position of the mobile front interface in time
have been estimated efficiently for the left, bottom and right walls. The evolution of the position of the moving front and
the left walls shows a linear trend, while the bottom and right walls present an increasing concave monotonic trend.
The computational consideration of the influence of the walls effect on the viscous flow predicts the fluid motion in an
accurate way, providing a satisfactory prediction for the front end region and the viscous fluid layer on the left wall, as it was
verified experimentally. The velocity vector distribution of the viscous fluid calculated captures the loss of dynamic energy
of the front as it moves along the bottom wall. A CCW vortex appears in the air above the viscous fluid close to the interface,
which moves in the direction of the column motion after its collapse.

Acknowledgements

The authors gratefully acknowledge the DIULS Multi-disciplinary PMU13331 Project (Universidad de La Serena) and
FONDECYT No. 1140074 Project (CONICYT-CHILE), for providing financial support for the publication of this research.
N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100 99

Fig. 7. Velocity vectors in the shear-thinning fluid to L/4, L/2, 3/4L and 9/10L (left side) and shear-thinning fluid–air (right side).

References

[1] V. Topin, F. Dubois, Y. Monerie, F. Perales, A. Wachs, Micro-rheology of dense particulate flows: application to immersed avalanches, J. Non-Newton.
Fluid Mech. 166 (2011) 63–72.
[2] H. Kohno, T. Tanahashi, Numerical analysis of moving interfaces using a level set method coupled with adaptive mesh refinement, Internat. J. Numer.
Methods Fluids 45 (2004) 921–944.
[3] C. Hu, M. Sueyoshi, Numerical simulation and experiment on dam break problem, J. Mar. Sci. Appl. 9 (2010) 109–114.
[4] H. Chanson, Analytical solutions of laminar and turbulent dam break wave, in: International Conference on Fluvial Hydraulics River Flow, Lisbon,
Portugal, 2006, pp. 465–474.
[5] Ch. Ancey, S. Cochard, The dam-break problem for Herschel–Bulkley viscoplastic fluids down steep flumes, J. Non-Newton. Fluid Mech. 158 (2009)
18–35.
[6] M. Cruchaga, D. Celentano, T. Tezduyar, Collapse of a liquid column: numerical simulation and experimental validation, Comput. Mech. 39 (2007)
453–476.
[7] M. Cruchaga, D. Celentano, T. Tezduyar, Moving-interface computations with the edge-tracked interface locator technique (ETILT), Internat. J. Numer.
Methods Fluids 15 (2005) 65–75.
[8] J. Steffe, Rheological Methods in Food Process Engineering, second ed., Freeman Press, Michigan, USA, 1996.
[9] A. Shamekhi, K. Sadeghy, Cavity flow simulation of Carreau–Yasuda non-Newtonian fluids using PIM mesh free method, Appl. Math. Model. 33 (2009)
4131–4145.
[10] C. Balan, D. Broboana, C. Balan, Mixing process of immiscible fluids in micro channels, Int. J. Heat Fluid Flow 31 (2010) 1125–1133.
100 N.O. Moraga et al. / Computers and Mathematics with Applications 69 (2015) 89–100

[11] C. Janßen, S. Grilli, M. Krafczyk, On enhanced non-linear free surface flow simulations with a hybrid LBM–VOF model, Comput. Math. Appl. 65 (2013)
211–229.
[12] T. Tezduyar, M. Behr, J. Liu, A new strategy for finite element computations involving moving boundaries and interfaces—the deforming-spatial-
domain/space–time procedure: I. The concept and the preliminary numerical tests, Comput. Methods Appl. Mech. Engrg. 94 (1992) 339–351.
[13] E. Oñate, J. García, S.R. Idelsohn, F. Del Pin, Finite calculus formulations for finite element analysis of imcompressible flows. Eulerian, ALE and
Lagrangian approaches, Comput. Methods Appl. Mech. Engrg. 195 (2006) 3001–3037.
[14] J.P. Wang, A. Borthwick, R. Taylor, Finite-volume VOF method on dynamically adaptative quadtree grids, Internat. J. Numer. Methods Fluids 00 (2003)
1–22.
[15] J.A. Sethian, Evaluation, implementation, and application of level set and fast marching methods for advancing fronts, J. Comput. Phys. 169 (2001)
503–555.
[16] M. Kim, W. Lee, A new VOF-based numerical scheme for the simulation of fluid flow with free surface. Part I: new free surface-tracking algorithm and
its verification, Internat. J. Numer. Methods Fluids 42 (2003) 765–790.
[17] T. Wacławczyk, T. Koronowicz, Modelling of the free surface flows with high resolution schemes, Chem. Proc. Eng. 27 (3/1) (2006) 783–802.
[18] O. Ubbink, R.I. Issa, Method for capturing sharp fluid interfaces on arbitrary meshes, J. Comput. Phys. 153 (1999) 26–50.
[19] T. Wacławczyk, T. Koronowicz, Comparison of CICSAM and HRIC high-resolution schemes for interface capturing, J. Theoret. Appl. Mech. 46 (2008)
325–345.
[20] M. Rico, G. Benito, A. Díez-Herrero, Floods from tailings dam failures, J. Hazard. Mater. 154 (2008) 79–87.
[21] D. Stefanescu, Science and Engineering of Casting Solidification, second ed., Springer Editorial, 2009, p. 380.
[22] B.P. Leonard, The ULTIMATE conservative difference scheme applied to unsteady one-dimensional advection, Comput. Methods Appl. Mech. Engrg.
88 (1991) 17–74.
[23] J.C. Martin, W.J. Moyce, An experimental study of collapse of liquid columns on a rigid horizontal plane, Philos. Trans. R. Soc. Lond. Ser. A 244 (1952)
312–324.
[24] S. Koshizuka, H. Tamako, Y. Oka, A particle method for incompressible viscous flow with fluid fragmentation, Comput. Fluid Dyn. J. 4 (1995) 29–46.
[25] A. Kačeniauskas, R. Pacevič, T. Katkevičius, Dam break flow simulation on grid, Mater. Phys. Mech. 9 (2010) 96–104.

S-ar putea să vă placă și