Rockfill Headloss

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Head Loss Equation to Predict Flow Transmissibility of

Rockfill Structures
Toshihiro Morii1

Rock has been advantageously employed in the hydraulic structures such as


flowthrough embankments, cofferdams, gabion weirs and drain works. To predict a
flow transmissibility of rockfill structures, an accurate evaluation of head loss
property of flow through rockfill is required. In the present study, a head loss
equation of flow through rockfill that is applicable to a wide range of rock particle
size is proposed. Forchheimer equation is selected to describe the head loss
property of flow through rockfill. Two coefficients dependent only on void
structure of the rockfill are included in the equation. These two coefficients are
effectively determined as a function of a hydraulic mean radius of void based on
the results of three series of laboratory tests: One-dimensional permeability tests
using river gravel 5-25mm in sieved diameter, embankment tests using river gravel
5-25mm in sieved diameter conducted in a laboratory 20cm-in-width water flume,
and embankment tests using river gravel 25-75mm in sieved diameter conducted in
a laboratory 50cm-in-width water flume. In the one-dimensional permeability tests
two coefficients included in the head loss equation of flow are determined directly
by a regression analysis of a hydraulic gradient applied to a rockfill column and a
discharge velocity of flow through the column. In the latter two tests, two
coefficients are determined by a parameter estimation procedure that combines a
genetic algorithm with a finite element calculation of nonlinear flow from
stage-discharge rating data observed in the tests. Unique relationships between two
coefficients and the hydraulic mean radius of voids are found. It is shown that the
head loss equation of flow proposed can be well applied to rock particle size up to
about 10cm to predict the flow transmissibility of rockfill structures.
Key words: Rockfill, Flow transmissibility of rockfill, Head loss equation of flow

1. Introduction

Rock has been advantageously employed in hydraulic structures such as flowthrough


dams, cofferdams, gabion weirs and drain works. To predict flow transmissibility of the
rockfill structures, an accurate evaluation of hydraulics of flow through rockfill is required.
Analysis of flow through rockfill is complicated because of nonlinear relationship between
discharge velocity through rockfill and hydraulic gradient applied.
In the present study, a head loss equation of flow through rockfill is proposed to predict
flow transmissibility of the structures based on laboratory tests and numerical calculations. A
special feature of the head loss equation of flow proposed is that it can be applied to 5 to
about 100mm rockfill particles in diameter without any consideration of scale effects of flow.
Firstly a Forchheimer-type equation is recommended to describe the head loss property of
flow through rockfill, that is, a nonlinear relationship between discharge flow through the
rockfill and a hydraulic gradient applied on it. And a numerical procedure that employs a
simple genetic algorithm (GA) combined with a nonlinear flow analysis formalized by a finite
element method (FEM) is developed to determine the nonlinear parameters included in the
head loss equation of flow. Secondly, by applying the GA-FEM to results of laboratory
one-dimensional permeability tests and embankment tests using river gravel, a hydraulic
mean radius of voids is effectively introduced into the head loss equation of flow to describe

1
Associate Professor, Faculty of Agriculture, Niigata University, Niigata 950-2181,Japan.
T/F: +81-25-262-6652, EM: morii@agr.niigata-u.ac.jp
effects of shape and size of rock particles as well as voids within the rockfill on the flow
hydraulics. The effect of water temperature is also included in the head loss equation of flow.
The head loss property of flow calculated by the proposed equation is compared with the
Wilkins’ equation to show a practical applicability of the former equation. Lastly concluding
remarks are given and some remaining problem left is commented.

2. Head loss equation of flow through rockfill

2.1 Head loss equation of flow


It has been well known that the discharge velocity through rockfill, V, has a nonlinear
relationship with the hydraulic gradient applied on it, i. Some examples of the nonlinear
relationship between V and i is shown in Figure 1 which were observed in the laboratory
one-dimensional permeability tests shown later. Solid lines are quadratic regression curves.

1.0

D4
0.8 D3 D1
D2
Hydraulic gradient i

0.6

0.4

Particle size
0.2 D1 20-25mm
D2 15-20mm
D3 10-15mm
D4 5-10mm
0.0
0 2 4 6 8 10 12
Discharge velocity V, cm/s
Figure 1 Typical results of the one-dimensional permeability test of rockfill.

There are two types of equation describing the nonlinear head loss of flow through coarse
materials such as rockfill: The Forchheimer equation and the Missbach equation (Basak, 1977,
Parkin, 1990, and Kazda, 1990). A superior feature of the former is that it can be derived
theoretically from the Navier-Stokes equation of fluid, while the latter has been determined
only empirically. The Forchheimer equation is given by

i = AV + BV 2 1

where A and B are coefficients which depend on the void structures of the rockfill as well as
water viscosity of flow, and have their dimensions [T1L-1] and [T2L-2], respectively. According
to the theoretical considerations done by Irmay (1958), Scheidegger (1974), and Bear and
Verruijt (1987), the coefficients A and B in Equation 1 are described as

A0 ν B
A= , B= 0 2
g g

in which g is acceleration of gravity, ν is a kinematic viscosity of water, and both A0 and B0


are the coefficients depending only on the void structures of rockfill. A0 and B0 have their
dimensions [L-2] and [L-1], respectively. A reciprocal of A0 corresponds to a physical or
intrinsic permeability of porous media that has been well established in a theory of Darcy
flow through soil.
Effects of void structures on the head loss property of flow through rockfill should be
evaluated by taking the size and shape of the rock particles as well as the size and distribution
of voids within the rockfill into account. Martins (1990) shows that all the effects mentioned
above could be well described by hydraulic mean radius of voids, m, which is defined as

eb
m= 3
6r

in which e is a void ratio of rockfill, b is a representative diameter of rock particles, and r is


shape coefficient of rock particles (Hansen, Garga and Townsend, 1995). It may be right to
think that the coefficients A0 and B0 in Equation 2 are closely related to m.

2.2 Laboratory tests to study head loss equation of flow


The one-dimensional permeability tests of a rockfill and the laboratory tests of
embankment constructed in the water flumes were conducted to study the head loss property
of flow through rockfill. In the former tests the river gravel 5 to 25 mm in diameter was used,
while in the latter both 5 to 25 mm and 25-75 mm river gravel were employed.
In the one-dimensional permeability tests, rock particles were filled and compacted
lightly by hand into a cylindrical column 10 cm diameter and about 100 cm length. Mass and
volume of the rockfill column were measured to calculate porosity, n, of the rockfill specimen.
The rockfill column being placed vertically, water was supplied into the bottom of the column
from a water reservoir. Flow rate through the rockfill column was measured by mass, and
divided by a cross-sectional area of the column to determine V through the rockfill. Total head
along the rockfill column was measured by eight piezometers connected 10cm interval to the
wall of the column. i applied to the rockfill column was calculated from the slope of the linear
regression line through a plot of the total head along the rockfill column. i was increased
successively from about 0.05 to 0.8 by seven to eight steps in each test of the rockfill column.
Water temperature of flow through the rockfill column, T, was measured at beginning and
completion of the test.
Two types of the acrylic-walled water flume were employed in a series of the laboratory
embankment tests: One is 20 cm wide and 40 cm deep, and the other 50 cm wide and 60cm
deep. The same rock particles as used in the one-dimensional permeability tests were filled
into the water flume in layers and compacted lightly with a steel rod to construct the
laboratory embankment. The mass and the volume of the laboratory embankment were
measured during and after construction, respectively, to calculate n and e. Both upstream and
downstream slopes of the embankment were retained by a wire mesh with squared openings
of 2 mm to ensure stability of the slopes and to prevent any erosion of rock particles from the
downstream slope. Total head H within the laboratory embankment were measured from
tapping points installed along the base and the wall of the water flume. A free surface of flow
within the laboratory embankment and its exit point on the downstream slope were observed
through the acrylic wall of the water flume. The upstream water level, hu, was raised
successively in four to six steps. After the steady state of flow was attained at each step, bulk
discharge through the laboratory embankment was measured by mass at an outlet of the water
flume. The flow discharge per unit width of the water flume, q, was calculated by dividing the
measured bulk discharge by the width of the water flume. No regulation was given on the
downstream water level in the laboratory embankment tests.
Being washed and air-dried, the river gravel was sieved into six classes of size, D1 to D6.
Their representative diameter, rock particle shape and physical properties of rock particles in
each class of size are given in Table 1. Representative diameter of rock particles, d, is defined
as an arithmetic mean of particle size. The rock particles were classified according to a Zingg
diagram (Garga, Townsend and Hansen, 1991) into four shapes of blade, disk, spheroid and
rod based on measurements of three orthogonal lengths of selected rock particles. Their
frequencies are given in the rows (4)-(7) of Table 1. Mean shape coefficient of rock particles
in each class of size, r, in the row (8) is calculated as a mean of re weighted by the frequencies
in the rows (4)-(7), in which re is the shape coefficient for each shape of rock particles
measured by Sabin and Hansen (1994).

Table 1 Size, shape and physical properties of rock particles (river gravel) used in the laboratory tests.

Class Arithmetic Particle Rock particle shape a) Specific Water


of mean of size, Mean shape gravity of absorption
Particle particle sizemm Blades Disks Spheroids Rods coefficient dried particle of particle,
d, mm r b) G %
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
D1 20-25 22.5 0.32 0.22 0.14 0.32 1.9314 2.563 1.4
D2 15-20 17.5 0.22 0.18 0.26 0.34 1.8600 2.550 1.5
D3 10-15 12.5 0.22 0.24 0.20 0.34 1.8774 2.557 1.4
D4 5-10 7.5 0.32 0.16 0.06 0.46 1.8944 2.575 1.6
D6 25-50 37.5 0.17 0.40 0.16 0.27 1.9051 2.550 1.2
D5 50-75 62.5 0.18 0.38 0.14 0.30 1.9008 2.601 0.9
a) Rock particle shapes such as blades, disks, spheroids and rods are classified according to the Zingg classification (Garga,
Townsend and Hansen, 1991). Numerals in the cells show frequency of rock particle shapes determined by measuring three
orthogonal axes of rock particle. Fifty rock particles are selected from each class of rock particle size to determine the
frequency of rock particle shapes.
b) r is a mean shape coefficient of rock particles, and is calculated as a mean of re weighted by the frequency given in the rows
(4)-(7). re is the shape coefficient for each shape of rock particles, and is given by Sabin and Hansen (1994).

2.3 Numerical procedure to estimate nonlinear parameters of flow


A numerical parameter estimation procedure, GA-FEM, that employs the simple genetic
algorithm combined with the FEM flow calculation is developed to determine the nonlinear
coefficients included in the head loss equation of flow, Equation 1. In the parameter
estimation, the nonlinear coefficients are coded into binary genetic chromosomes with a
length of 12 bits, and then an optimal set of the nonlinear coefficients is determined by
iterative genetic operations so that it minimizes an object function. Note that the object
function is defined as a squared deviation of flow discharge through rockfill between
observation and FEM calculation as shown in Figure 2. A fitness value of the chromosome is
calculated as a reciprocal of the value of the object function. The chromosomes with the large
Discharge per unit width

150
Observed (q obs )
q, cm 3 /s -1/cm -1

100 Calculated (q cal )


q cal -q o bs
50

0
0 10 20 30
Upstream water depth h u , cm
Figure 2 Stage-discharge relations showing a deviation of flow discharge through rockfill between observation
and FEM calculation that is included in the object function.
Start

Control parameters for genetic operation:


Number of chromosomes (N), values of
probability of crossover and mutation, bit
numbers of coded A0 and B0, and lower and
upper limitations of A0 and B0.
Control parameters for flow calculation:
Bottom width and upstream/downstream
slopes of embankment, number of steps of
impounding (L), water temperature of flow,
and L sets of measured data (upstream and
downstream water depth, and discharge
qobs).

Generation=0

Random generation of initial N chromosomes.

Generation=generation+1

n=1 to N

Decoding of i-th chromosome into A0 and B0.

A=A0・ν/g and B=B0/g


where ν: kinematic viscosity of fluid and g:
gravity acceleration.

FEM calculation of flow qcal using A and B.

L 2
O ( n) = å [qcal (l ) - qobs (l )]
l =1

Calculation of fitness of n-th chromosome


based on linear scaling.

Chromosomes are converged?

Yes
End

Figure 3 A flow chart showing the numerical procedure to determine the nonlinear coefficients included in the
head loss equation of flow through rockfill. O is the object function to be minimized.

value of fitness are selected to reproduce a next generation of offspring. Figure 3


schematically shows a flow chart of the parameter estimation procedure, in which O is the
object function, and qcal and qobs are the flow discharge through the rockfill embankment
observed and calculated by the FEM, respectively. Control parameters commented in Figure 3
are given in Table 2. Lower and upper limitations of A0 and B0 are calculated from the bit
numbers and the accuracy of estimation given in Table 2.

Table 2 Control parameters for genetic operation used in the numerical estimation of A0 and B0.

Control parameters For A0 For B0


(1) (2) (3)
Number of chromosome, N 100 100
Probability of crossover 0.25 0.25
Probability of mutation 0.01 0.01
Bit number 12 12
Accuracy of estimation 0.1 cm-2 0.001 cm-1
In the FEM calculation of flow through rockfill, a numerical difficulty due to the
non-linearity in the head loss equation of flow, Equation 1, is successfully solved by a method
of successive approximation (Kazda, 1990). During an iterative calculation, the flow is
assumed to be Darcy one, and a fictitious hydraulic conductivity, Kf, is introduced to calculate
the flow velocity (Kazda, 1990, and Kells, 1993):

- A + A 2 + 4 Bi
Kf = 4
2 Bi

The iterative calculation to determine the position of the free surface of flow was
performed until the difference between the new height of a point of the free surface and its
previous value is less than 0.1 % of the upstream water level.

3. Head loss equation of flow determined by laboratory tests and


parameter estimation

3.1 Nonlinear parameters of flow correlated with hydraulic mean radius of voids
The nonlinear coefficients included in the head loss equation of flow through rockfill
were determined from the one-dimensional permeability tests of rockfill column and the
laboratory tests of rockfill embankment constructed in the water flume. Sixteen rockfill
columns were tested in the one-dimensional permeability tests. T was 15-23 degrees
centigrade. Reynolds number, Re, which was calculated by the following equation, ranged 9 to
323:

Vm
Re = 5

The nonlinear coefficients A and B in Equation 1 were determined directly by a regression


analysis as shown in Figure 1, then converted into A0 and B0 by using Equation 2 with known
values of g and ν. Be sure that ν is a function of T measured. As well fifteen and twelve
embankments with different combination of the rock particle size, the downstream slope and
the top width of the embankment section were tested in the 20cm-in-width and the
50cm-in-width water flumes, respectively. The values of T and Re measured in the tests
ranged 8-25 degrees centigrade and 15-1150, respectively. A and B were estimated from the
hu-q data by using the numerical procedure shown in Section 2.3, and then converted into A0
and B0 in the same manner as the one-dimensional permeability tests.
As stated in Section 2.1, it may be right to think that the coefficients A0 and B0 in
Equation 2 are closely related to m because all the effects of rock particles and void structures
on the flow can be well described by it. Figure 4 shows plots of A0 and B0 determined or
estimated from the test results with the values of m. In calculating m, b in Equation 3 is
replaced by d given in Table 1. A very unique relationship between A0 and m as well as B0 and
m can be noticed, which shows a practical effectiveness of the proposed head loss equation of
flow through rockfill, that is, Equation 1 combined with Equation 2. Functional relationships
of A0 and B0 with m are given as, using a power type equation,

A0 = 24.39 × m -1.60 6a
B0 = 0.630 × m -1.15 6b
5000 30

Coefficient B 0 , cm -1
Determined from 1D tests(1)
Coefficient A 0 , cm -2

4000 Estimated from small flume tests(2)


Estimated from large flume tests(3)
Regressed from (1), (2) and (3) 20
3000

2000 (a) Coefficient A0 (b) Coefficient B0


10
1000

0 0
0.0 0.1 0.2 0.3 0.4 0.0 0 .1 0.2 0.3 0.4

Hydraulic mean radius of void m, cm Hydraulic mean radius of void m, cm


Figure 4 Unique relationships of the nonlinear coefficient A0 and B0 with the hydraulic mean radius of voids
determined and estimated from the laboratory tests using the river gravel 5-75mm in diameter.

in which A0, B0 and m are in cm-2, cm-1 and cm, respectively.


Figure 5 compares the flow discharge per unit width of the water flume measured in the
laboratory tests with the FEM calculations using the estimated values of A0 and B0 in a
logarithmic plot. A good comparison between the measurements and the calculations show a
practical accuracy of the head loss equation of flow proposed above. In Figure 6(a) the free
surface of flow and the total head of water within the laboratory embankment are compared
between the FEM calculations and the water flume tests at the step of the highest position in
the upstream water level. The laboratory embankment shown in Figure 6 was constructed
using the rock particle D1 and D3 in the upstream and downstream halves of the rectangular
section, respectively. Stage-discharge rating curves, hu-q, measured and calculated are given
in Figure 6(b). There can be seen again good comparisons between the measurements and the
calculations in Figure 6.

3.2 Comparison of head loss equation of flow with the Wilkins’ equation
As reported by Martins (1990) and Leps (1973), the Wilkins’ equation was established
for a transition range of flow between laminar and turbulent, and has been considered as the
one most commonly used in practice. Wilkins in 1956, as reported by Leps (1973) conducted
the laboratory transmissibility tests using the cylindrical columns of crushed stone. The
columns were 20 to 56 cm in diameter, and 69 cm long. Five sizes of rock particles sieved
from about 2 to 7.6 cm were tested. From his test data, Wilkins established the head loss

1000
esitim ated coefficients, cm 3 /s/cm
q calculated by FEM using

100

10

Large flume tests (12 tests)


Small flume tests (15 tests)

1
1 10 100 1000
q observed in laboratory water
flume tests, cm 3 /s/cm

Figure 5 Comparison of flow discharge measured in the laboratory embankment test with the FEM calculations
using the estimated coefficients A0 and B0.
Test No. D(1+3)-V-60

Discharge per unit width of


D1 D3 120
Observed in model test

water flume q, cm 3 /s/cm


H=25 Calculated by FEM using
hu=27.4cm 22.5 80
20 estimated coefficients
15 3.6cm
10
40
0 10 20 30 40 50 60
Test No. D(1+3)-V-60
Horizontal distance x, cm 0
0 10 20 30
Observed in model test.
Calculated by FEM using estimated coefficients. Upstream water depth h u , cm
H: Total head above base of water flume in cm.
D1 and D3: Rock particle size 20-25 and 10-15mm, respectively.

(a) Free surface and equi-potential lines (b) Stage-discharge curves

Figure 6 Comparison of flow hydraulics between the laboratory embankment test and the FEM calculation using
the estimated values of A0 and B0. Note that the upstream and downstream halves of the embankment are
constructed using the rock particle D1 and D3, respectively.

equation of flow as, in inch-second units as proposed by Wilkins:

V = nW m 0.5 i 0.54 7

where W is an empirical coefficient for a given rockfill material, depending primarily on the
shape and roughness of the rock particles and on the viscosity of water. It has been found that
W varies from about 33 for crushed stone to about 46 for polished marbles (Leps, 1973). Be
sure that the hydraulic mean radius of void m in Equation 7 is defined without a consideration
of the rock particle shape.
To compare the head loss equation of flow proposed by Equation 1 together with
Equations 2 and 6, with the Wilkins’ equation, two laboratory embankments are selected as
examples. The values of parameters required to determine the head loss equation are
summarized in Table 3. The first, second and third terms of the test number represent the rock
particle size given in Table 1, the specified downstream slope (11=1V:1H), and the specified
top width of the embankment in cm, respectively. In Figure 7 the head loss properties of flow
through rockfill calculated by the proposed equations, Equations 1, 2 and 6, for two selected
embankments are compared with ones calculated by the Wilkins’ equation. A pair of dotted
lines represents a range of discharge velocity calculated by Equation 7 corresponding to the
range of W, that is, 33 to 46 mentioned above. It is found that the head loss equation of flow
proposed gives a moderate prediction as compared with the Wilkins’ equation.

Table 3 Parameters used to calculate the head loss equation of flow proposed and the Wilkins’ equation.

Parameters For D1-11-10 For D5-11-10


(1) (2) (3)
Rock particle size b, cm 2.25 6.25
Void ratio e 0.600 0.623
Proposed eq.
Shape coefficient of rock particles r 1.931 1.901
Hydraulic mean radius of voidsa) m, cm 0.117 0.341
Porosity n 0.375 0.384
Wilkins’ eq. Hydraulic mean radius of voids b) m, in. 0.105 0.299
Coefficient W 33-46 33-46
a) m is calculated by Equation 3.
b) m is interpolated from the values given in Table 1 of Leps (1973). Note that m is in inches.
Hydraulic gradient i 1.5
No. D1-11-10
1.0
No. D5-11-10
0.5
Proposed equation
Wilkins' eq., W=33-46
0.0
0 10 20 30 40
Discharge velocity V, cm/s
Figure 7 Comparison of the head loss property of flow calculated by the proposed equation and the Wilkins’
equation for the selected laboratory embankment.

4. Conclusions

The head loss equation of flow through rockfill was successfully developed based on the
Forchheimer equation to predict the flow transmissibility of the rockfill structures. A special
feature of the head loss equation of flow developed is that it can be applied to 5 to about
100mm rockfill particles in diameter without any consideration of scale effects of flow. The
head loss equation of flow developed requires only two parameters to estimate the hydraulics
of flow through rockfill: The hydraulic mean radius of rockfill voids and the water
temperature of flow. This is another feature of the head loss equation of flow to be stressed.
The followings are summarized:

1) The nonlinear parameters included in the head loss equation of flow were determined
from the one-dimensional permeability tests and the laboratory embankment tests using
the river gravel, and were effectively correlated to the hydraulic mean radius of voids of
rockfill.
2) By comparing the measurements in the laboratory tests with the nonlinear FEM
calculations, the practical effectiveness and accuracy of the head loss equation of flow
developed was examined.
3) The head loss equation of flow developed gives a moderate prediction as compared with
the Wilkins’ equation.
4) The head loss equation of flow developed is valid for the wide range of rock particle size,
5 to about 100mm.

Further trials of the parameter estimation in the laboratory as well as in the field will be
needed to obtain the more accurate and practical relationship of A0 and B0 with a wide range
of m. Once the relationship is determined, it may be possible to calculate the value of m by
inserting the A0 and B0 estimated from the observation into this relationship. The calculated
value of m may provide us the practically important information related to the rock particle
properties, the deposition of rockfill, or the degree of clogging due to soil particles within the
rockfill.
Rockfill structures are economical ones from the geotechnical point of view, and are
friendly to our environment. Understanding the rockfill hydraulics may be helpful to design
safely such structures as the flowthrough dams, the gabion weirs, the drain works and the
river control works, and as well may give us a new idea of mitigation works for our
environment.
Acknowledgements

The present study was supported by the Grant-in-Aids for Scientific Research (C), No.
11660238 and No. 14560196, made by the Ministry of Education, Culture, Sports, Science
and Technology of Japan. The author are grateful to Messrs. T. Sada, T. Kobayashi, S.
Matsumoto, T. Mori and Y. Kaneko for their help in conducting the laboratory tests.

References

Basak P. 1977. Non-Darcy flow and its implications to seepage problems. Journal of the
Irrigation and Drainage Division. Proceedings of the American Society of Civil Engineers.
103(4): 59-473
Bear J. and Verruijt A. 1987. Modeling Groundwater Flow and Pollution. D. Reidel
Publishing. Holland: 41-43
Garga V. K., Townsend R. and Hansen D. 1991. A method for determining the surface area of
quarried rocks. Geotechnical Testing Journal. 14(1): 35-45
Hansen D., Garga V. K. and Townsend R. 1995. Selection and application of a
one-dimensional non-Darcy flow equation for two-dimensional flow through rockfill
embankments. Canadian Geotechnical Journal. 32: 223-232
Irmay S. 1958. On the theoretical derivation of Darcy and Forchheimer formulas.
Transactions, American Geophysical Union. 39(4): 702-707
Kazda I. 1990. Finite Element Techniques in Groundwater Flow Studies: With Applications in
Hydraulic and Geotechnical Engineering. Elsevier Science. New York: 164-183
Kells J. A. 1993 Spatially varied flow over rockfill embankments. Canadian Journal of Civil
Engineering. 20: 820-827
Leps T. M. 1973. Flow through rockfill, Embankment-Dam Engineering Casagrande Volume
edited by Hirschfeld R. C. and Poulos S. J. John Wiley & Sons: 87-107
Martins M. 1990. Principles of rockfill hydraulics. Advances in Rockfill Structures edited by
Maranha das Neves E. Kluwer Academic Publishers. Netherlands: 523-570
Parkin A. K. 1990. Through and overflow rockfill dams. Advances in Rockfill Structures
edited by Maranha das Neves E. Kluwer Academic Publishers. Netherlands: 571-592
Sabin C. W. and Hansen D. 1994. The effects of particle shape and surface roughness on the
hydraulic mean radius of a porous medium consisting of quarried rock. Geotechnical
Testing Journal. 17(1): 43-49
Scheidegger A. E. 1974. The Physics of Flow through Porous Media. Third edition.
University of Toronto Press. Toronto: 152-170

You might also like