You are on page 1of 6

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

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

A numerical model for analyses of flow through rock discontinuities

J.F. da Silva
CMEC Consulting Engineers, Belo Horizonte, MG, Brazil

M.V. Vidigal
EMBRAER, Sao Jose dos Campos, SP, Brazil

ABSTRACT: Water flow in open smooth or rough rock joints, under laminar or turbulent regime, is
discussed. The use of a “virtual joint” concept has allowed the development of a numerical model for anal-
yses of non-linear flow in open rock joints by means of the finite element method. The model was verified
against published results of laboratory tests on plates of slate and charnockite, carried out to check the
validity of existing flow laws. The joints flow analyses, realized to verify the numerical model, have shown
very good agreement with the laboratory tests’ results. This, together with existing means of determining
the joint’s opening and relative roughness through field water pressure tests will permit adequate flow
analyses of the foundation drainage systems of concrete gravity dams, lying on jointed rock.

1 INTRODUCTION D = drain’s diameter; g = gravitational acceleration;


and i = hydraulic gradient in the drain.
In nature, the flow of water in rock masses occurs A numerical model developed by means of the
basically through the rock joints or, in the case of finite element method to incorporate the drains’
permeable rocks, water flows through the rock influence on three-dimensional flow analyses of
pores in addition to the rock joints, if present. concrete gravity dam foundations in continuous
In some structures like concrete gravity dams permeable rocks, named DW3D, has been dis-
(Figure 1) water also flows though a third, man- cussed in previous publications (da Silva & da
made imposed media, the drains, consisting of Gama, 2003, da Silva, 2005, da Silva, 2006).
cylindrical holes drilled from the drainage gallery In order to allow for the analyses of seepage
into the rock, to reduce the water pressures in the through the foundations of structures like the one
dam’s foundation. shown in Figure 1, it was also necessary to develop
The general flow in continuous permeable an approach to incorporate the flow, in open rock
rock media is represented by Darcy’s law, in three joints, in model DW3D, and this is the object of
dimensions: this paper.

Vj = kjij (1)

where Vj = velocity of flow; kj = coefficient of per-


meability; and ij = the hydraulic gradient. The sub-
script j represents the directions of three mutually
perpendicular axes X, Y and Z.
The flow through rough or smooth drains under
laminar or turbulent regime can be represented by
DRAINAGE GALLERY
Darcy-Weissbach’s law for flow in pipes:
DRAINS
LV2 V2
hf = f or i = f (2)
D 2g 2 gD JOINTS

where hf = energy loss; f = roughness coefficient;


L = drain’s length; V = flow velocity in the drain; Figure 1. Concrete gravity dam lying on fissured rock.

255
2 HYDRAULICS OF OPEN ROCK JOINTS expressions for the determination of the roughness
coefficients, for the various types of flow that occur
2.1 Hydraulic diameter in open rock joints, are indicated in Tables 1–3.
The joints’ hydraulic diameter is given by the fol-
lowing expression: Smooth Laminar
ha
(
Dh
< 0.0168 ) Hydraulically smooth Turbulent
Dh = 2e (3) Parallel flow Completely rough Turbulent
(
ha
< 0.0320 ) ha
Laminar
Dh ( 0.0168 <
where Dh = hydraulic diameter; and e = joint open-
< 0.0320 )
Dh Turbulent
ing or aperture.
Laminar
Non parallel flow
ha
Turbulent
( > 0.0320 )
2.2 Darcy—Weisbach’s equation for joints Dh

The flow in either smooth or rough joints, under Figure 2. Flow in open rock joints (Apud Louis, 1967).
laminar or turbulent flow, can be represented by
Darcy-Weisbach’s equation:
Table 1. Roughness coefficients for rock joints—ha/Dh <
0.0168 (After Louis, 1967).
L V2
hf = f (4) Flow regime Flow law
Dh 2 g ha /Dh < 0.0168 – Parallel flow
Laminar – smooth Poiseuille:
where L = joint’s length in the flow direction. NR<2300 96
f (7)
NR
Turbulent – hydraulically smooth Blasius:
2.3 Reynolds’ number 8 1
3.71 f 0.316.NR 4 (8)
The roughness coefficient f, both for smooth or 2300 NR 2.552 log
h a Dh
rough joints, is a function of Reynolds’ number
Turbulent – completely rough Nikuradse:
NR, which is expressed by the following non-
h a / Dh
8
1
dimensional relationship: NR 2.552 log
3.71 2 log (9)
h a Dh f 3.71

VDh
NR= (5)
ν
Table 2. Roughness coefficients for rock joints—0.0168 <
where ν = kinematic viscosity. ha/Dh < 0.0320 (After Louis, 1967).
Depending on the values of NR the flow in open Flow regime Flow law

joints can be laminar, turbulent or a transition 0.0168 < ha /Dh < 0.0320 – Parallel flow
regime. In a transition regime, the flow oscillates Laminar Poiseuille:
between laminar and turbulent. 1
2 1.76
3.71 96
NR 142000 log f (7)
h a Dh NR
2.4 Absolute and relative roughness
Turbulent Nikuradse:
The absolute roughness corresponds to the height 1
2 1.76
of the wall projections or asperities ha in the joints’ 3.71
1 h a / Dh
NR 142000 log
faces and the relative roughness RR is given by the 2 log (9)
h a Dh
f 3.71
expression:

RR = h a = h a (6) Table 3. Roughness coefficients for rock joints—ha/Dh >


Dh 2e 0.0320 (After Louis, 1967).
Flow regime Flow law
ha /Dh > 0.0320 – Non-parallel flow
2.5 Types of flow Laminar
1 Louis:
According to Louis (1967), flow in open rock joints 2 1.76
96
NR 142000 log
1.9 f 1 8.8(h a /Dh )1.5 (10)
can be classified as shown in Figure 2. h a Dh NR

Turbulent
Louis:
2.6 Flow laws for open joints 1

h a /Dh
2 1.76
1.9 1 (11)
NR 142000 log 2 log
The roughness coefficient f for joints is both a func- h a Dh f 1.90

tion of NR and RR. According to Louis (1967), the

256
Putting Expression 7, from Table 2, into A quadrilateral finite element formed by four
Expression 4, Poiseuille’s equation for smooth triangles was adopted. Using Hubbert’s (1940)
joints is obtained: approach, the flow velocities in a triangular ele-
ment are given by the expressions:
g e2
V= I or V = K joint I (12)
⎛ ∂P ⎞
12ν ⎛ ∂P ⎞
V x′ = −K x′ ⎜ + ρ g x ′⎟ ; V y ′ = − K y ′ ⎜ + ρ g y ′⎟ (16)
⎝ ∂x′ ⎠ ⎝ ∂ y′ ⎠
where Kjoint = “coefficient of permeability” of the
smooth joint, “equivalent” to k in Expression 1
and equal to: where Vx' and Vy' = flow velocities in the x' and y'
directions; P = pressure at a point inside the joint;
Kx' and Ky' = “equivalent permeabilities” of the
g e2
K joint = (13) joint in the x' and y' directions; and ρ = mass den-
12ν sity of the fluid.
In matrix form Expression 16 can be written as:
Expression 12 indicates that, under laminar
regime, the flow is independent of the joint’s
⎧V x ′ ⎫ 1 ⎡K x′ 0 ⎤ ⎧Pi ⎫
roughness. ⎪ ⎪
⎨ ⎬=− ⎢ ⎥ [ B] ⎨ P j ⎬
The quantity of flow is given by: V
⎩ ⎭y ′ Δ ⎣ 0 K y′⎦ ⎪⎩ P k ⎪⎭
(17)
Q = VA = VeW (14) ⎡K x′ 0 ⎤ ⎧⎪ ρ g a x ′z ⎫⎪
−⎢ ⎥⎨ ⎬
where Q = quantity of flow; A = joints’ cross- ⎣ 0 K y ′ ⎦ ⎪ρ g ⎪
⎩ a y ′z ⎭
sectional area; and W = width.
The flow per unit width is then equal to:
where ax'z = direction cosine of the angle x'Z;
Q ge3 ay'z = direction cosine of the angle y'Z; Δ = trian-
q= or q = (15) gle’s area; and [B] = triangle’s geometric matrix
W 12v (Desai & Abel, 1972).
For a smooth joint under laminar flow, the value
where q = flow per unit width. Equation 15 is
of Kx' = Ky' is constant and determined by means
sometimes referred to as the “cubic law”.
of Expression 13.

3 A NUMERICAL MODEL FOR OPEN 3.2 Non-linear flow


ROCK JOINTS
To take into account the non-linearity of Darcy-
3.1 Linear flow Weissbach’s equation, a “virtual joint” concept,
similar to the “virtual drain” concept (da Silva &
Figure 3 shows a two-dimensional finite element da Gama, 2003), was used.
representing a joint with water flowing in the princi-
pal directions x' and y'. A global three-dimensional
coordinate system XYZ is also shown. 3.3 The virtual joint
It is assumed that the virtual joint is smooth
and under laminar flow regime as expressed by
Z Equation 12. It is also assumed that in a virtual joint
y'
the “openings” can be different in the x' and in the y'
L K
directions. Therefore, its “equivalent permeabilities”,
Vy'
Vx' determined through Expression 13, are equal to:
I J x'
g Y gex2′ ge 2y ′
K x′ = and K y′ = (18)
12ν 12v

Figure 4 shows a graphical representation of


Equations 4 and 12. In Figure 4, the line with
X inclination αP represents Poiseuille’s linear flow in
a “virtual joint” finite element having at the outset
Figure 3. Two-dimensional joint element with nodal an opening equal to the opening of the real joint
points IJKL. where the flow is represented by Darcy-Weissbach’s

257
V to verify the applicability of the flow laws for open
Darcy - Weissbach (4) joints, indicated in Tables 1–3, to analyses of rock
joints, since Louis’ experiments were performed on
A A' concrete made joints. Although other rocks were
tested, only the experiments realized on slate and
Poiseuille (12) charnockite will be discussed here.

4.1 Laboratory tests


P The general assemblage of the tests is shown sche-
matically in Figure 5.
DW Basically the set up consisted of two plates of
rock sitting one on top of the other. The plates
iP i DW were separated by rubbery cork ribbons placed
i along the edges in order to limit the flow to a sin-
gle direction and provide the desired joint open-
Figure 4. Velocity versus hydraulic gradients in joints.
ing. Piezometers were connected to one of the rock
plates to allow for pressure measurements in two
equation. It is assumed initially that there is flow in points of the joint. The plates were leveled and the
one direction (say x') only. Also, initially, the veloci- values of W, L and e were measured. Water was
ties V=Vx' are the same for both joints, as indicated then introduced under pressure in one of the joint’s
by points A and A', but the gradients are different extremities. The volume of water flowing out on
since the “equivalent coefficients of permeability” the opposite side was collected in a graduated
of the joints are, respectively: bucket and the time of flow was measured. After
stabilization of the flow, the piezometers’ readings
KP = tan(αP)(virtual) and KDW = tan(αDW) (real) were taken. With these values, the joint’s gradients
and flow velocities could be determined using the
(19)
expressions:
However, according to Expression 13, increas-
P
ing or decreasing the virtual joint opening ex' will h= (21)
cause a corresponding increase or decrease in the ρg
value of its equivalent permeability KP . Therefore,
through adequate changes in the virtual joint open- i = h1 − h 2 (22)
ing it is always possible, by successive iterations, to L
make αP approach αDW. The same approach is then vol
applied to the other direction Vy'. Q
Convergence is achieved when: V= = t (23)
A eW
i P − i DW where P and h = pressure and head, respectively,
≤ tol (20)
i DW in the piezometers; vol = volume of water flowing
through the joint; and t = time of flow.
where (iP) is the gradient in the virtual joint, (iDW) is
the gradient in the real joint and (tol ) is the admit- L
ted tolerance. When convergence is achieved, the
flows both in the virtual and in the real joint will
be the same although having different openings.
At this point there will be total correspondence Piezometers
h1
of velocities, gradients and discharges between the
virtual and the real joint. The process is repeated
for all joint elements of the problem. e
Q h2

4 AN ASSESSMENT OF THE MODEL’S


ACCURACY W

Da Gama & Vidigal (2005) have published the Figure 5. Assemblage of the rock plates for the flow
results of a series of laboratory tests carried out tests (Apud da Gama & Vidigal, 2005).

258
h1
SLATE h1
3 h1
2 6 h2
8 1
1
2
4
9 h2
e = 1,00mm ha/Dh = 0,010 3
12
15
h2
7 4 8
VELOCITY (m/s)

39
6
Laws: Blasius and Nikuradse 7 7
26
42
10 45
13 25 48
30
e = 1,65mm ha/Dh = 0,017
5 37 51
40 32
29 54
43 34
4 Law: Blasius 46
31
33
L 49 53
3 52

2
W
1
0 Figure 8. Finite element mesh to simulate the tests
0 5 10 15 20 25 30 35 40 45 50 55 60 65
GRADIENT conditions.
Hand calculations Laboratory tests

Figure 6. Tests results for flow in slate joints (after da Table 4. Basic data for the analyses of flow in the rock
Gama & Vidigal, 2005). joints.

Joint faces e W L T
material (mm) ha (mm) (mm) (mm) (°)
CHARNOCKITE

7 Slate 1.65 0.033625 75.5 402 20


6 e = 1,20mm ha/Dh = 0,031 e = 0,95mm ha/Dh = 0,039 Charnokyte 0.55 0.074948 75.5 397 21
Law: Louis Turbulent
VELOCITY (m/s)-

Law: Nikuradse
5

e = 0,55mm ha/Dh = 0,068


SLATE
2
Law: Louis Laminar and Turbulent
1 600

500
VELOCITY (cm/s)

0
0 10 20 30 40 50 60 70 80 90 100 110
400
Hand calculations Laboratory tests GRADIENT
300
t
200

Figure 7. Tests results for flow in charnockite joints 100

(after da Gama & Vidigal, 2005). 0


0 2 4 6 8 10 12 14 16 18

GRADIENT
Tests Dw3d

Figure 6 and Figure 7 shows the tests results


obtained for the joints in slate and in charnockite Figure 9. Parallel flow—Flow laws (7), (8) and (9).
respectively.
To compare the tests’ results with the values
indicated by the flow laws (Tables 1–3), the rock
CHARNOCKITE
plates’ absolute roughness ha and the water tem-
peratures T were also measured. Details of these 500
450
VELOCITY (cm/s)

measurements are given in da Gama & Vidi- 400


350
gal (2005). The values obtained from the flow 300
250
laws, by hand calculations, are also indicated in 200
150
Figure 6 and Figure 7. It can be seen that the flow 100
50
laws discussed by Louis (1967) represent very 0
0 20 40 60 80 100 120
well the flow of water both in the slate and in the
GRADIENT
charnockite. Tests Dw3d

4.2 Numerical model analyses


Figure 10. Non-parallel flow—Flow laws (10) and (11).
4.2.1 Finite element mesh
To test the proposed joint numerical model, the
finite element mesh indicated in Figure 8 was 4.2.3 Analyses results
used. The nodal points for the application of the The flow analyses results, realized with the pro-
piezome- ters’ loads are indicated in the mesh. posed joint model, are shown in Figure 9 for the
slate and Figure 10 for the charnockite, together
4.2.2 Basic data with da Gama & Vidigal (2005) tests’ results. It can
The basic data used in the numerical flow analyses, be seen that the proposed numerical model repre-
extracted from da Gama & Vidigal, 2005, are indi- sents quite well the flow laws for rock joints indi-
cated in Table 4. cated in Tables 1–3.

259
5 FIELD DETERMINATION OF THE Hubbert, M.K. 1940. The theory of ground-water motion.
JOINT’S OPENING AND RELATIVE J. Geol., v. 48, No. 8, Part 1. 785–944.
ROUGHNESS Louis, C. 1967. A study of groundwater flow in jointed rock
and its influence on the stability of rock masses. Dok-
tor-Ingenieur Thesis. Faculty of Civil Engineering
In order to use the proposed numerical model in and Surveying, Univertität (TH) Karlsruhe, Germany.
practice, the rock joints opening and relative rough- (English Translation by Mrs. M. Kutter, 1969, Impe-
ness must be determined. rial College, London.)
This problem has been addressed by Riβler Riβler, P. 1978. Determination of the water permeability
(1978) who, following Louis (1967), has developed a of jointed Rock, The Institute for Foundation Engi-
theory to determine the values of rock joint’s open- neering, Soil Mechanics, Rock Mechanics and Water
ing and relative roughness based on field water pres- Ways Construction, RWTH (University), Aachen,
sure test results. Federal Republic of Germany.

6 CONCLUSIONS SYMBOLS AND UNITS

The concept of a virtual joint is adequate and has hf = energy loss (m)
enabled the development of a numerical model for f = roughness coefficient
non-linear flow analyses of open rock joints. Appli- L = drain’s or joint’s length (m)
cations of the model using existing flow laws has V = flow velocity (m/s)
shown very good agreement with published tests’ D = drain’s diameter (m)
results on joints formed by slate plates, for paral- g = gravitational acceleration (m/s2)
lel flow, and charnockite plates for non-parallel i = hydraulic gradient
flow. The inclusion of the joint model proposed in Dh = hydraulic diameter (m)
this paper into the three-dimensional flow model e = joint opening or aperture (m)
DW3D together with the possibility of determin- f = roughness coefficient
ing the joint’s openings and relative roughness NR = Reynolds’ number
from field water pressure test results allows the ν = kinematic viscosity (m2/s)
realization of flow analyses for the design of the ha = absolute roughness (m)
foundation drainage systems of concrete gravity RR = relative roughness
dams lying on jointed rock. Kjoint = “equivalent coefficient of permeability”
(m/s)
Q = quantity of flow (m3/s)
REFERENCES A = joints’ cross-sectional area (m2)
W = joints’ width (m)
da Gama, E.M. & Vidigal, M.V. 2005. Experimental q = flow per unit width (m3/s/m)
verification of water flow laws for rock disconti- x' and y' = principal directions of flow
nuities. Soils & Rocks, v. 28, No. 1, 43–50. Brazil. (In XYZ = global three-dimensional coordi-
Portuguese.) nate system
da Silva, J.F. 2006. Optimization of concrete gravity dams Vx' and Vy' = flow velocities (m/s)
foundation drainage systems. Dams and Reservoirs, P = pressure (kN/m2)
Societies and Environment in the 21st Century. Pro-
h = head (m)
ceedings of the international symposium on dams in
the societies of the 21st century. International congress Kx' and Ky' = “equivalent permeabilities” (m/s)
on large dams (ICOLD), Barcelona, Spain. Ed. Luis ρ = mass density of the fluid (kg/m3)
Berga et al. Taylor and Francis. London. 633–639. ax'z = direction cosine of angle x'Z
da Silva, J.F. 2005. Influence of the geometry of the ay'z = direction cosine of angle y'Z
drainage system and of the foundation anisotropy Δ = triangle’s area (m2)
on the uplift pressures under concrete dams. Info- [B] = triangle’s geometric matrix
geo 2005-5th. Brazilian symposium on applications of iP = gradient in the virtual joint
computational mechanics in geotechnicale engineering. iDW = gradient in the real joint
165–174, Belo Horizonte, Brazil.
tol = tolerance
da Silva, J.F. & da Gama, E.M. 2003. A Three-dimensional
model for seepage analysis of concrete dams founda- vol = volume of water (m3)
tions. 4th International workshop—applications of t = time (s)
computational mechanics in geotechnical engineering.
337–357. Ouro Preto. Brazil.
Desai, C.S. & Abel, J.F. 1972. Introduction to the finite
element method. Van Nostrand Reinhold Company,
New York, 384.

260

You might also like