Sei sulla pagina 1di 6

Rock Engineering in Difficult Ground Conditions Soft Rocks and Karst Vrkljan (ed)

2010 Taylor & Francis Group, London, ISBN 978-0-415-80481-3

Non-linear fluid flow through rough-walled fractures


M. Javadi, M. Sharifzadeh & K. Shahriar
Department of Mining & Metallurgical Engineering, Amirkabir University of Technology, Tehran, Iran

M. Mehrjooei
Department of Mechanical Engineering, Amirkabir University of Technology, Tehran, Iran

ABSTRACT: Fractures in rock mass are the main flow paths and introduce as a most important
attribute in rock mass hydraulic behavior. Hydraulic properties of rock mass are important in civil, mining, and environmental such as Isolation of hazardous and nuclear waste, petroleum and geothermal
energy. In this paper, single-phase fluid flow through a rock fracture was studied. Computational domain
for an artificial three-dimensional fracture was generated and used for numerical fluid flow simulations.
Both laminar and turbulent flow simulations were performed for a wide range of inlet velocities. The
results show that the relation between pressure drop and flow rate has a quadratic polynomial form and
Forchheimer law was fitted very well to flow simulation results. Also the predicted static pressure drop for
turbulent flow simulation was roughly 3% to 17% more than those predicted with laminar flow simulation
at Reynolds number of 4.5 to 89.5, respectively.
1

INTRODUCTION

In many geological structures, the matrix permeability is negligible compared to permeability


of fractures and rock mass hydraulic behavior is
controlled by fractures. In such situations, the
development of realistic and robust predictive
models of flow and transport requires a thorough
understanding of the physical processes that govern flow in individual fractures. Quantification
and understanding of fluid flow through fractured
rock are important in several industrial activities
and research topics, such as in the reservoir exploitation for water supply, petroleum, geothermal,
and heat storages and in the safe design of disposal
sites for domestic, industrial, and nuclear wastes.
The simplest model of fracture flow was that
of viscous flow between two parallel plates, which
is governed by the well-known cubic law (Witherspoon et al. 1980). The single fractures are roughwalled conduits with variable apertureand the
classical view of a rock fracture is not adequate for
the description of flow. In order to deal with the variations of the aperture of the fracture, the Reynolds
lubrication equation (Brown 1987) was introduced
as an alternative to the considerably more computationally intensive solution of the Navier-Stokes
(NS) equations. The Reynolds lubrication equation is just valid for laminar flow through fractures
and needs some especial geometrical and kinematical conditions of fracture and fluid flow respectively (Zimmerman & Bodvarsson 1996, Ge 1997).

Therefore, in recent studies, direct numerical calculation of NS equations is used for investigation
of fluid flow through fractures (Zimmerman et al.
2004, Brush & Thomson 2003, Koyama et al. 2008).
Several researchers evaluated the behavior of fluid
flow through rough-walled fractures, experimentally
(Sharifzadeh 2005, Qian et al. 2005) or numerically
(Javadi et al. in press a, b, Brush & Thomson 2003,
Zimmerman et al. 2004) all of which indicate that
the effect of nonlinear flow becomes more evident
in highest values of Reynolds number.
The laminar flow through a rock fracture is usually assumed a linear relationship between the flux
and the pressure gradient and, it is known that, at
sufficiently high values of the Reynolds number,
this relationship becomes nonlinear. However, few
studies have been implemented on the numerical
calculation of NS equations for turbulent flow
through three-dimensional fractures and this subject is studied in this paper.
In present study, first a three-dimensional geometrical domain of a hypothetical rough-walled
fracture is generated. Computational domain of
this fracture is generated and both laminar and
turbulent simulations of fluid flow are performed
through the void specimen. Both laminar and turbulent simulations of fluid flow are performed
three-dimensionally for a wide range of flow rates
based on a finite volume method. The calculated
average pressure drops, between consecutive vertical sections were compared to describe the flow
rate dependant pressure drop.

261

2.3 Computational domain and boundary


conditions

METHODOLOGY

2.1 Governing equations


The general description of fluid flow in a single
fracture is given by the NS equations which
express momentum and mass conservation at the
microscopic level at points within the fluid continuum over the fracture void space. Considering
the steady laminar flow of a Newtonian fluid with
constant density and viscosity through a fracture
with impermeable walls, the NS equations may
be written in vector form as (Brush & Thomson
2003),

(u )u = 2u p

(1)

where is the fluid density, is the fluid viscosity,


u is the flow velocity vector, and p is the hydrodynamic pressure. Eq. (1) is composed of a set of
coupled nonlinear partial derivatives of varying
orders. In order to have a closed system of equations, they must be supplemented by the continuity
equation, which represents conservation of mass.
For an incompressible fluid, conservation of mass
is equivalent to conservation of volume, and the
equation takes the form

u=0

A three-dimensional geometrical domain of an


arbitrary fracture with 3 mm in width and 12.5 mm
in length, in x- and y-directions respectively, was
used for fluid flow simulations. For evaluating
the impacts of surface roughness and aperture on
fluid flow through void specimen, the lower and
upper walls of the fracture (Fig. 1a) were applied
as a smooth and very rough surface, respectively.
Three-dimensional void space was generated over
an xy grid with uniform spacing x and y (equal
to 0.5 mm), and variable aperture (z). The fracture
geometrical domain consisted of 150 volumetric
elements (6 rows and 25 columns). Each volumetric elements height, in z-direction, represents the
aperture.
As shown in Figure 1b, the aperture of fracture
varies from a minimum of 0.35 mm to a maximum
of 0.65 mm and its average and standard deviation
are 0.485 mm and 0.106 mm, respectively.
Computational grids of about 890,000 tetrahedral meshes were generated using GAMBITTM
preprocessor code. The numerical simulations were
performed for water with a density of 998.2 kg/m3
and a viscosity of 0.001 kg/ms and the gravitational
effect was neglected. The velocity inlet boundary

(2)

The relevant boundary conditions for the NS


equations include the no-slip conditions, which
specify that at any boundary between the fluid and
a solid, the velocity vector of the fluid must equal
that of the solid (Zimmerman & Bodvarsson,
1996).
2.2 Numerical simulation
Finite volume method was employed to solve
the NS equations (Eq. (1)) and continuity equation (Eq. (2)) for a three-dimensional problem.
Direct numerical simulation (DNS) and Reynoldsaveraged Navier-Stokes (RANS) equations were
used for laminar and turbulent flow simulation,
respectively. The numerical solution of the NS
equations for turbulent flow is extremely difficult,
and due to the significantly different mixing-length
scales that are involved in turbulent flow, the stable
solution of this requires such a fine mesh resolution that the computation time becomes significantly long. Attempts to solve turbulent flow using
a laminar solver typically result in a time-unsteady
solution, which fails to converge appropriately. To
encounter this, a time-averaged equation, RANS,
supplemented with k turbulence model was
used in practical computational fluid dynamics
applications when modeling turbulent flow. In this
study, the FLUENTTM was used for both laminar
and turbulent flow simulations.

Figure 1. Geometrical domain of the hypothetical


fracture which used for fluid flow simulation: (a) threedimensional geometrical domain with boundary conditions and (b) frequency of aperture.

262

condition was used for the inlet region of the


domain. The inlet velocity was considered to distribute uniformly in y-direction and low turbulence
intensity (1%) was assumed in all numerical turbulent flow simulations. For the outlet of the domain,
an outflow boundary condition was assumed.
All other solid surfaces of the domain (boundaries of the fracture) were defined as impermeable
walls with the no-slip velocity boundary condition (Fig. 1a). Flow simulations through threedimensional fracture were performed within a
range of inlet velocity from 0.01 to 0.2 m/s for
laminar flow simulation, and from 0.01 to 1 m/s
for turbulent flow simulation.
3

SIMULATION RESULTS

Both laminar and turbulent flow simulations


were performed for inlet velocities in the range
of 0.01 m/s to 0.2 m/s. For each simulation, the
static pressure drop was calculated as the difference between static pressure on inlet and outlet
boundaries. These static pressure drops were plotted as inlet velocity in Figure 2a.
In order to compare more closely and clearly
the difference between laminar and turbulent flow
simulations, for each inlet velocity the relative error
between laminar and turbulent flow simulations
was calculated as difference between the predicted
static pressure drops. The relative error between
laminar and turbulent flow simulations was plotted as Reynolds number in Figure 2b.
The flow rate is usually expressed in terms of
the dimensionless Reynolds number, which quantifies the relative strength of inertia forces as compared to viscous forces. For fluid flow through
fractures, the Reynolds number can be defined as
(Zimmerman et al. 2004)
Re =

Q
w

(3)

where Q is the volumetric flow rate, is the fluid


viscosity, is the fluid density, and w is the width
of fracture.
The quadratic polynomial was used for curve
fitting of both laminar and turbulent results
(Fig. 2a), that leads to a good correlation factor
and appropriately low constant term of expected
regression equations. This type of relation between
static pressure drop and inlet velocity (or flow rate)
is recognized as Forchheimer law (Elsworth &
Goodman 1986) which uses a polynomial expression to describe the flow rate dependant pressure
gradient, or pressure drop, as

p = AQ + BQ2

(4)

Figure 2. Comparison between the predicted static


pressure drops for laminar and turbulent flow simulation: (a) the predicted total pressure drop for different
inlet velocities with quadratic polynomial curve fitting
(b) the relative error between laminar and turbulent flow
simulations for different Reynolds numbers.

where p is the pressure drop, A and B are


constants and Q represents the flow rate in the
direction of maximum pressure gradient.
As shown in Figure 2, in the same inlet velocities, the static pressure drops predicted by turbulent
flow simulation are higher than those predicted by
laminar flow simulation. In low Reynolds number
the difference between predicted static pressure
drops is relatively low and by increasing Reynolds
number, the difference increases as the relative
error from 3.2% to 17.3% for the Reynolds number
of 4.5 to 89.5, respectively. Comparison between
relative errors shows that in low inlet velocities the
flow regime can be considered as laminar flow and
by increasing inlet velocity (or Reynolds number)
the regime of flow propones to turbulent. Therefore, the results of turbulent flow simulation were
selected for evaluating the impact of roughness
and inlet velocity on flow behavior. Moreover,
turbulent flow was simulated for inlet velocities of
0.5 and 1 m/s.
In order to evaluate the effects of roughness and
aperture variations on the total pressure drop, the
fracture geometrical domain is discretized to the
smaller sub-fractures. By considering the stepped
profile for fractures, as shown in Figure 1a, the
rough-walled fracture in the longitudinal main
flow direction (y-direction) can be discretized to

263

n smaller sub-fractures. Each sub-fracture has


no varia and has a variable aperture in perpendicular main flow direction (x-direction).The
sub-fractures are consecutively located and form
the main fracture. The 126 vertical sections, normal to y-direction and perpendicular to the main
flow direction, with 0.1 mm consecutive distances
in y-direction were selected through geometrical
domain. By using these vertical sections, the geometrical domain was divided to 125 sub fractures
and the space between two consecutive sections
represents a sub-fracture.
The average total pressure on each vertical section pi (i from 1 to 126) was calculated and then
the average total pressure drop between tow consecutive sections sp j ( j from 1 to 125) was calculated as sp j = pi pi +1 (for i = j ) which represents
the total pressure drop of the jth sub-fracture.
Figure 3 shows the sp j for different inlet velocities which appears as a semi-undulation graph.
The uniformly velocity distribution and low flow
turbulency were considered in all numerical turbulent fluid flow simulations. Low flow turbulency in
the inlet flow leads changes on sp j for four initial
sub-fractures, j = 1 to 4, which can be seen in the
Figure 3a, b. The diagrams in Figure 3 show a rise
and fall behavior and the sp j appears as a semiundulation graph with abrupt peaks.
The fluctuations in Figure 3 can be analyzed
with the fracture geometrical domain. As presented above, the apertures were distributed over
an xy grid with uniform spacing which are suddenly changed in the y-direction in the magnitudes of y = 0.5, 1, 1.5, 12.5 mm. Therefore,
the flow section changes abruptly after the subfractures which their j index are multiples of 5
(hereafter named SA sub-fractures). For another
sub-fractures (hereafter named NA sub-fractures)
the flow section between tow consecutive subfractures remains constant.
The sp j of NA sub-fractures was induced due
to the viscous effects and increases by decreasing
the area of sub-fracture. For SA sub-fractures,
an additional pressure drop (local pressure drop)
occurs during fluid flow through fractures due
to flow section changing. By change in the subfracture section area, flow section changes and the
fluid static pressure decreases suddenly. Therefore
the sp j of SA sub-fractures increase dramatically and whatever the flow section changes more
intensively, then, the pressure drop is more intensive. These phenomena can be proposed as the
effect of fracture roughness on fluid flow behavior
through fractures. As shown in Figure 3, the sp j of
both SA and NA sub-fractures increase with the
inlet velocity.
In order to compare more closely and clearly
about the effect of inlet velocity (or flow rate) on

Figure 3. Variation of sp j for inlet velocities of


(a) 0.02 m/s (b) 0.1 m/s (c) 0.5 m/s and (d) 1 m/s.

sub-fractures pressure drop, three SA sub-fractures


with j = 42, 87 and 117 and three NA sub-fractures
with j = 15, 55 and 115 were selected from geometrical domain and the sp j of each sub-fracture was
calculated for different inlet velocities (Figs. 4, 5).
Figure 4 shows the variation of the sp j of the
selected NA sub-fractures for different inlet velocities in the range of 0.01 to 1 m/s. this figure shows
a roughly linear variation of the sp j and then linear
regression was selected for describing the pressure
drop dependant inlet velocity.
Comparing the sp j of the selected SA subfractures for different inlet velocities shows a
non-linear variation and then polynomial regression was selected for describing the pressure drop
dependant inlet velocity. As presented above, the

264

Figure 4. The effect of inlet velocity on variation of


sp j for NA sub-fractures for (a) j = 42 (b) j = 87 and
(c) j = 117.

Figure 5. The effect of inlet velocity on variation of


sp j for NA sub-fractures for (a) j = 15 (b) j = 55 and
(c) j = 115.

sub-fractures are consecutively located and form


the main fracture. The pressure drop of the main
fracture can be calculated by summing up the sp j
of all sub-fractures. Therefore, the pressure drop
dependant inlet velocity can be described with a
polynomial expression as the same as Forchheimer law.

compared with those predicted with turbulent


flow simulation. The results of turbulent flow
simulation were selected for evaluating the impact
of roughness and aperture variations on the total
pressure drop. To reach this goal, the fracture geometrical domain is discretized to the smaller SA
and NA sub-fractures. The variation of sp j for both
SA and NA sub-fractures were evaluated and the
results were used for describing the pressure drop
dependant inlet velocity.
Both laminar and turbulent flow simulation
results show a good agreement with Forchheimer
law for the Reynolds number in the range of 4.5
to 89.5. The non-linearity of fluid flow implies the
turbulency on fluid and therefore, it seems that the
turbulent flow simulation leads to more rational
and reasonable estimations. Comparison between
simulations results shows that the predicted static
pressure drop for turbulent flow simulation was
roughly 3% to 17% more than those predicted with

DISCUSSION AND CONCLUSION

In this paper, non-linear fluid flow through rock


fracture was studied. Computational domain for
an artificial three-dimensional fracture with one
rough surface was generated and used for threedimensional numerical fluid flow simulations.
Both laminar and turbulent flow simulations
were performed for a wide range of inlet velocities. For some inlet velocities, static pressure drops
of laminar flow simulation were calculated and

265

laminar flow simulation at Reynolds number of 4.5


to 89.5, respectively. The difference between turbulent and laminar flow simulations increases by
increasing Reynolds number, which recommends
us to use turbulent flow simulation for non-linear
fluid flow.
Based on turbulent flow simulation results, two
types of pressure drops, viscous and local pressure
drops, can be considered during fluid flow through
rock fractures. Comparing the pressure drop of
the sub-fractures shows that the roughly linear
and non-linear variations of the sp j can be considered for NA and SA sub-fractures, respectively.
The pressure drop of the main fracture can be
calculated as the summation of the sp j of all subfractures. Therefore, the pressure drop dependant
inlet velocity can be described with a polynomial
expression as the same as Forchheimer law.

REFERENCES
Brown, S.R. 1987. Fluid flow through rock joints: the
effect of surface roughness. Journal of Geophysical
Research 92(B2): 13371347.
Brush, D. & Thomson, N.R. 2003. Fluid flow in synthetic
rough-walled fractures: Navier-Stokes, Stokes, and
local cubic law simulations. Water Resources Research
39(4): 10851099.
Ge, S. 1997. A governing equation for fluid flow in rough
fractures. Water Resources Research 33(1): 5361.

Javadi, M. Sharifzadeh, M. & Shahriar, K. Evaluation of


Non-linear fluid flow through rough-walled fractures.
Amirkabir universiy journal, In press.
Javadi, M., Sharifzadeh, M., Shahriar, K. & Mehrjooei, M.
Roughness effect on velocity domain through rock
fractures. Sharif universiy journal, In press.
Koyama, T., Neretnieks, I. & Jing, L. 2008. A numerical study on differences in using NavierStokes and
Reynolds equations for modeling the fluid flow and
particle transport in single rock fractures with shear.
International Journal of Rock Mechanics and Mining
Science 45: 10821101.
Qian. J., Zhan. H., Zhao. W. & Sun. F. 2005. Experimental study of turbulent unconfined groundwater flow in
a single fracture. Journal of Hydrology 311: 134142.
Sharifzadeh. M. 2005. Experimental and theoretical
research on hydro-mechanical coupling properties of
rock joint. PhD thesis, Kyushu University, Japan.
Witherspoon. P.A., Amick. C.H., Gale. J.E. & Iwai. K.
1979. Observations of a potential size-effect in experimental determination of the hydraulic properties of
fractures. Water Resources Research 15(5): 11421146.
Zimmerman. R.W., Al-Yaarubi. A., Pain. C.C. &
Grattoni. C.A. 2004. Non-linear regimes of fluid
flow in rock fractures. International Journal of Rock
Mechanics and Mining Science 41(3): 163169.
Zimmerman. R.W. & Bodvarsson. G.S. 1996. Hydraulic conductivity of rock fractures. Transport Porous
Media 23: 130.

266

Potrebbero piacerti anche