Documentos de Académico
Documentos de Profesional
Documentos de Cultura
GRAVEL MATERIALS
By B. ~berg ~
dss ~ 5 and D15/d15-> 4, both the base and filter materials must possess grading
stability when exposed to seepage. There are two reasons for grading instability;
scantiness of intermediate grain sizes, which interrupts an internal filter formation
process; and loose grains, which move through the void space between fixed grains.
A mathematical model for washout of grains and internal filter formation (self-
filtration) in cohesionless materials is developed on the basis of grain and pore
geometry considerations and found to agree with results from laboratory tests on
sand and gravel mixtures. The model discloses both scantiness and loose grains and
predicts the extent of grain washout and the volumedecrease causedby the washout.
For soils with a large content of fine grains, the hydrodynamicnumber is shown
to have considerable influence on the extent of the washout of grains.
INTRODUCTION
FUNDAMENTALEQUATIONS
This section summarizes some definitions, equations, and other results
from ~ b e r g (1992a) that will be used in the following.
A grain chord is the intersection of an imaginary straight line and the
volume of a grain. A void chord is the intersection of such a line and the
void space between two grains. The average length of all the chords of a
single grain of naturally rounded hard-rock gravel is close to g = x/2, where
x is the grain size (square-mesh-sieve aperture). The mean grain chord of
a whole material is
.............................................. (1)
= 2 2"B0
where y = 1/Bo = the harmonic mean grain size of the material. B0 is given
by (4) with Ya = 0.
Dependent upon the grain-size distribution, there are two types of gran-
ular materials; type A with only fixed grains; and type B with both large
fixed grains and small loose grains, that can move in the void space between
the fixed grains. The void ratio of a type A or type B material is
Aa
e = 2-c-~-~ - Ya + 2 - d ' ( 1 - Ya) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
Aa = o-~dy-ya" 1o x (dy
y) ................................. (3)
LdY
Ba =
fv
.o xC~)
............................................... (4)
where y = the ordinate of the grading curve of the material; 0 -< y -< 1.0;
x(y) = the grain size corresponding toy; andy~ = the ordinate that separates
fixed and loose grains in type B materials and gives the largest value of e
as a function of Ya. The corresponding grain size is
37
ILILI
Downloaded from ascelibrary.org by Cambridge University on 07/13/15. Copyright ASCE. For personal use only; all rights reserved.
hB
hA
ho
B hB
C hC
h D~
D hD
J, DO00001 000000
FIG. 1. Material Before and After Washout of Grains and Internal Filter Formation
Y
1.0'
YD
YC
. . . xf
I
i BC
'I
!
I
D
I I
YB
0
x0 xB xC xD xlO0
X
FIG. 2. Grading Curves for Original Material (A) and Filter Layers after Washout
(B, C, and D)
The size, XD, of the smallest grain, which will remain in the layer D after
washout, is determined by the pore size of the drainage layer or the aperture
of the screen. The size, Xo of the smallest grain, which will remain in the
layer C, is determined by the size of the pores in the layer D. The size, xB,
of the smallest grain, which will remain in the layer B, is determined by the
size of the pores in the layer C, and so on.
The layers D, C, and B can be considered filter layers, which successively
develop during the washout of grains from the original material A. In general
terms, the size, xr, of the smallest grain, which can be retained by such a
filter layer (the controlling pore size of the filter layer) was assumed to be
given by
The term a'~0s in (7) was obtained from base-filter compatibility tests
made in a permeameter. Five filter materials consisting of sand and gravel,
densified by means of the modified Proctor method, were tested with very
uniform base materials (Fig. 3). With xr = d85, the test results gave a =
0.4 (Fig. 4). The void ratios of the filter materials gave by (2), with y, =
0 and d = 0, the average value c = 0.73.
The term 2. dr. gr in (7) was based on the assumption that in less-densified
filter materials, the imperfect relative positions and directions of the irreg-
ularly shaped grains increase both the mean void chord and the controlling
pore size with the same increment [cf (6)].
The calculation of xI proceeds layer by layer, D, C, B, etc., until for some
layer (in Figs. 1 and 2 the layer B) xi becomes smaller than the smallest
grain size, x0, of the original material A. If this material should be of type
B, then the calculation shall stop when xi < xa, where xa is the grain size
that separates loose and fixed grains in the original material [(5)]. The
washout of most of the grains between x0 and x, will never stop.
Y 0,125 0.25 2 4 16
o./ /
1.0
0 i :
0.031 0.063 0,125 0.25 0.5 1 2 4 8 16 32
xmm
FIG, 3. Grading Curves for Tested Base and Filter Materials
0,3 -
9 StabLe
E 0,2- o UnstabLe 9 9 ~o !
E
L~
~o O.lJ
0
0 oJ1 o J2 o5 o'.4 015
Tomm
40
material, which remains in the filter layer after washout. For example, if
bo is a coefficient, we get the following thickness of the layer D after washout
(Fig. 1):
hD = b D ' ( g D + r9) = b o ' ( 1 + eD)'gD ......................... (8)
Fig. 1 also shows from which parts B', C', and D' of the original material
that the materials in the filter layers B, C, and D originate. The thickness
e.g., of that part of the original material, from which the material in the
layer D originates is
where eyj) = a fictitious void ratio for the grains XD--XIoo, when they are
situated in the original material A. By considering the grains Xo to XD as
part of a void space between the grains x~ to Xlo0, we get
eA + Y D
eyD -- 1 -- YD ............................................. (10)
where e A = the void ratio of the original material A; and YD is the ordinate
of the grading curve of the material A, which corresponds to the grain size
XD (Fig. 2). eA can be determined experimentally or calculated from (2).
Eqs. (8)-(10) are valid also for the filter layers C and B. One only has
to change all the indices D into indices C and B, respectively.
The total shrinkage of the material caused by the washout of grains is s
= ho - h = (h'8 - h s ) + ( h c - h c ) + (h'D -- hD), which, with (8)and
(9), gives
s = [bs"(e,8 - e~).g~ + bc'(eyc - ec)'gc + b D ' ( e , D -- eD)'go] 9 (11)
For
8Ng
> 1 Ng
b = 2.5-~N g ................................... (12b)
o x(y)
where Yc and Yo = the ordinates of the grading curve of the original material
A that correspond to the grain sizes Xc and xo, respectively (Fig. 2). Similar
discussions and calculations of ~Ng/Ngcan be made for the other filter layers;
C, B, etc.
LABORATORYTESTS
The mathematical model was compared with results from grain washout
tests made by the writer about 20 years before the model was developed,
and also with grain washout tests reported by Kenney and Lau (1985, 1986)
and Lafleur et al. (1989),
Writer's Tests
The tests were made in a permeameter with the diameter of 190 mm. A
125-mm thick layer of a test material, consisting of sand and gravel, was
placed on top of a square mesh screen, the aperture of which varied between
different tests. The test material was densified by means of a procedure and
compaction effort corresponding to the modified Proctor method. The test
material was then loaded by an about 400-mm-thick layer of 30-50-mm
pebbles placed on a flexible metal cloth with 1.5-mm mesh size. The flow
of water was directed downward. The total drop of head in the permeameter
was 2.7 m. The permeameter could be mildly vibrated by means of a vi-
brator, which was fastened to it.
Each test started with 60 min without vibration, followed by two periods
of 15 min each with continuous vibration. Washed out grains were collected
after each period. In all but two tests, the amount of grains washed out
during the second period of vibration was smaller than 5% of the total
amount washed out during the test, and the discharge of water was nearly
the same at the ends of the two periods with vibration. The filter-formation
process was thus close to finished already after the first period with vibration.
The amount of grains washed out during the period without vibration was
of the order of 10-50% of the total amount washed out during the test.
The evaluation of the test results was based on the total amount.
The vibration was used to compensate for the short duration of the tests
as compared with prototype conditions, but, in spite of that, the vibration
probably implied more severe conditions in the direction towards washout
of grains than only long-duration seepage would have done.
Washout tests were made with eight materials, I, II, III, C, E, F, G, and
H, having different grain-size distributions (Figs. 3 and 5). One purpose of
the evaluation of the test results was to determine values of the coefficients
42
C
Downloaded from ascelibrary.org by Cambridge University on 07/13/15. Copyright ASCE. For personal use only; all rights reserved.
0.5
0
0.016 0.031 0.063 0.125 0.25 0.5 1 2 4 8 16
xmm
a, b, c, d, and dr for the filter layers, which were formed by the washout.
The determination was somewhat uncertain, but the best general fit between
the test results and the theoretical equations was obtained as follows.
The packing of the grains, which remain in a filter layer after washout,
is looser than the packing obtained by modified Proctor compaction. There-
fore, the void ratio, eI, is given by
e[ = e0s + 2"d r . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (14)
For loosest possible filling of sand and gravel. ~ b e r g (1992a) found dr =
0.18. This value was used in (7) and (14) unless
d i = e , r - eor
2 .............................................. (15)
gave a smaller value, eyr is the fictitious void ratio of the filter layer grains,
when they were situated in the original material [(10)]. Eq. (15) means that
e f = e y I.
e0r was calculated by (2) with Ya = 0, d = 0, and a value of c evaluated
from the measured void ratio of the original material. This % was also used
in (7)-(9) and (11), together with the value a = 0.4, which was found in
the previously mentioned base-filter compatibility tests.
Figs. 6 and 7 show results from the washout tests and the theoretical
calculations. All the diagrams in Figs. 6(a and b) include a curve, y, that
shows the grading of the original test material, and four theoretically cal-
culated curves, Ra, Rf, h', and s.
The curve Ra shows the ratio Ra = x;/x for different grain sizes, x, of
the original material. X'a is given by (5), with Ya in (3) and (4) replaced by
the ordinate of the original material grading curve, which corresponds to
the grain size x. If Ra < 1 for all grain sizes, x, then the original material
is a type A material. Otherwise, it is, in most cases, a type B material, and
the rightmost Ra = 1 corresponds to the grain size, xa, which separates
fixed and loose grains. Further details about how to distinguish between
type A and type B materials are given in Aberg (1992a).
The curve Rfshows the ability of the original material to form filter layers
during the washout. R f = x / x , where xI is given by (7), with the values of
a and dr mentioned earlier. R f -> 1 means that the filter formation process
43
Rf . . . N ~
0.5 ~- _-~. :::.:.. ~ -5o
.: / "~\,~<
0 i , _f,, L_X_ , I0
0.063 0.25 0.5 1 2 4 8 16 0.063 0.25 0.5 1 2 4 8 16
x mm
y Ra,Rf h',s mm
1.5
Material m h' Material c 115~ I
h'--~ e = 0.295 - - - 3~ e=o.28o
J I
R'=15-- 160
1.0
0.5.
0 i
0.063 0.25 0.5 I 2 4 8 16 0.063 0.25 0.5 I 2 4 8 16
xmm
(a)
FIG. 6(a). C a l c u l a t i o n (Ra, Rf, h', s) and Test (h't, st) Results for M a t e r i a l s S h o w n
in Figs. 3 and 5. M o r e in Figs. 7(a and b)
y,Ra,Rf h',s mm
1.5 150
\ , Material E Material F
\~_-~h e=0.227 e=0403
05 h
s t --10 s ~~'-'-~:~ st =30 s - _ ' ~ z - ' " "--""
o~ T - - ~ - ? -% . . . . . . . . .
0.063 0.25 0.5 1 2 4 8 16 0.063 0.25 015 1 2 4 8 16
• mm
y,Ra,R f h',s m m
1.5 150
h' Material G /'"-,, , Material H
...... ~\ e=0.189 -- -- ~ ' - - ~ X -qh e =0.204
1.0 ,
\
\ R'= 2 ' "%1 R,:52
0.5 ~ .
__ _ s__ _ \ " ~ - -':
0
0.063 0.25 0.5 1 2 4 8 16 0.063 0.25 0.5 I 2 4 8 16
X mm
(b)
FIG. 6(b). Calculation (Ra, R[, h', s) and Test (h't, st) Results for Materials S h o w n
in Fig. 5.
44
I00, E] ioo
51
50 - ~~ ' ~
31
50 5O
-40
Downloaded from ascelibrary.org by Cambridge University on 07/13/15. Copyright ASCE. For personal use only; all rights reserved.
15 ~ 1 2 %
\ 0 , - ;
0.25 0.5 1 2 4 8 16 0.25 0.5 1 2 4 8 16
xmm
100- []
6 ~ ~z.2 8
4~% 5 24 - - ~ 4 " N ~ -
50
0 ,
o
0.25 0.5 1 2 4 8 16 0.25 0.5 1 2 4 8 16
X fnm
(a)
FIG. 7(a). C a l c u l a t i o n and Test Results, D a s h e d Lines = h'. Full Lines = h't.
N u m b e r s = s and st, M o r e in Fig. 6(a)
h,ht mm h,ht mm
1001 2 ~ 2 0 [] -100 []
"\ K \ o
0.25 0.5 1 2 4 8 16 0.25 0.5 1 2 4 8 16
(b) • mm
FIG. 7(b). C a l c u l a t i o n and Test Results. D a s h e d Lines = h'. Full Lines = h't.
N u m b e r s = s and st. M o r e in Fig, 6(a)
cannot proceed towards finer grain sizes. An Rf curve has no meaning for
grain sizes that correspond to loose grains, x < x~. Washout of loose grains
is not covered by the present theory, but washout of grains x > Rfa'x, is
probably prevented, because such grains are larger than the voids of a filter
layer, which consists of the grains between Xa and xl00. Rf,, is the value of
xr for x = x~. Whether loose grains smaller than Rf,,. Xawill become washed
out, or possibly stopped by further filter formation amongst the loose grains,
cannot be predicted by the present theory. However, filter formation amongst
loose grains seems unrealistic, in case vibrations or alternating seepage
current directions or turbulence should make loose grains move around in
the void space between fixed grains. The change from dI = 0.18 to ds <
0.18 usually gives a peak in the Rf curve.
45
ho he, etc., as function of the grain sizes xD, Xc, xe, etc. For the materials
C, E, F, G, and H, Figs. 6(a and b) also show curves of the measured
Downloaded from ascelibrary.org by Cambridge University on 07/13/15. Copyright ASCE. For personal use only; all rights reserved.
average height, h't, up to which grains of different sizes were washed out
in the test. h't was evaluated from the amounts of differently sized washed
out grains. For each grain-size fraction, the thickness of the original material,
125 ram, was multiplied by the mass of washed out grains and divided by
the mass of grains in the original material. The measured total shrinkage,
st, of the original material is also shown. For the materials I, II, and III,
which were tested with different screen apertures below the test material,
curves of h' (dashed lines) and h't (full lines) are shown in Figs. 7(a and
b). Numbers at the ends of the curves show the calculated and measured
total shrinkages, s and st, respectively.
Figs. 6(a and b) and 7(a and b) show that the calculation and test results
agree fairly well. In this kind of tests, some erratic spread of the test results
is natural, but in this case, some of the deviations between the calculated
and measured washout heights and shrinkages also seem to be systematic
and correlated with the hydrodynamic conditions within the tested materials.
According to Kenney and Lau (1985), these conditions can be characterized
by means of a dimensionless hydrodynamic number, R' = qxJnv; where q
is the unit flux (discharge rate divided by total cross sectional area), x5 is
the 5% grain size of the tested material, n is the porosity of the material,
and v is the kinematic viscosity of water. According to Kenney and Lau
(1985), a minimum value of R' -~ 10 is required in order to obtain washout
of the largest grains that should become washed out, if only geometric
conditions were determining.
Figs. 6(a and b) show values of R', which were calculated on the basis
of the gradings of the original materials and the discharges measured at the
ends of the washout tests. The smaller-than-calculated washout heights for
the materials C and G seem to be dependent upon the extremely small
values of R'. For the rest of the materials (except the material F), values
of R' between 17 and 235 gave washout heights and shrinkages, which agree
fairly well with the calculated values. These results confirm that small hy-
drodynamic numbers decrease the amount of washout.
According to Fig. 6(a), the material C should be on the verge of being
unstable, because Rfis slightly larger than 1.0 for grain sizes between 0.15
and 0.25 mm. In the material E, grains smaller than 0.4 mm should be loose
(Ra > 1), and all grains smaller than Rfa'xa = 0.5"0.4 = 0.2 mm should
become washed out. For the material G, both Ra and Rf are smaller than
1.0 for all grain sizes x > 0.06 mm. In spite of that, the h' curve shows that
grains smaller than 0.7 mm should become washed out from the whole
thickness, 125 mm, of the test material. From the material H, all grains
smaller than 1.1 mm should become washed out because Rf > 1 for these
grains. The reason is the scantiness of grains between 0.5 and 4 mm.
In Fig. 8, the calculated and measured total weights of the washed-out
grains are compared. The calculated weight was obtained by means of mul-
tiplying the area under a h' versus y curve by the area of the permeameter
and the solid density of the grains and then dividing by (1 + e), where e is
the void ratio of the original material.
46
g•
0.5 9 o x
0.1 / 111 x
I - I I I I I I I II I I I I
0.1 0.2 0.5 1 2 5
CALCULATED
Kenney's Tests
The tests reported by Kenney and Lau (1985, 1986) were made in per-
meameters, which were lightly vibrated by manual tapping with a rubber
hammer. Information about grain shape, initial dry density, and shrinkage
of the test materials and controlling pore size of the underlying drainage
layers is incomplete or missing. The writer's calculations, therefore, became
more uncertain than the calculations concerning his own tests.
Figs. 9(a and b) show calculation results for eight of the tested materials.
It was assumed, that the materials consisted of sand and gravel, and the
coefficients a, b, and df were chosen in the same way as mentioned for the
writer's tests, c was calculated from (2) with d = 0. If c became larger than
0.80, then it was put equal to 0.80, and a value of d > 0 was calculated
from (2). e was based on initial dry densities received from Kenney. The
pore size of the drainage layer was assumed to be 2 mm or 5,6 ram, de-
pendent upon the grain size of the drainage layer.
Kenney and Lau (1985) classified the material A as unstable. A large part
of the grains smaller than 2.5 mm were lost from the central zone of the
material. Also, the calculations showed that the material A should be un-
stable, because Rf > 1.0 for a large interval of grain sizes smaller than 4
mm. Because Ra < 1.0 for all grain sizes, the material A has no loose grains,
but the instability is caused by scantiness of grain sizes between 1.4 and
11.3 ram.
Kenney and Lau (1985) classified the material As as unstable, but Kenney
and Lau (1986) changed the classification into stable. According to the
calculations Rf < 1.0 for all grain sizes. Ra > 1.0 for grains smaller than xa
0.18 mm. Hence, the very smallest grains of the material As should be
47
FIG. 9(a). Calculation Results for Materials Tested by Kenney and Lau (1985,
1986)
y, Ra,Rf h~,smm
1.5 150
\ Material U 1 Materiat 2
h' k e = 0.251 h'I e:0.202
R'= 7 R'=IO
1.0 -- \ l 100
"" Ra --'~ R . R f / f
. j/ 1
\ R~\'-../
0.5
, . . . . . . . . . . 0
0.125 0.5 1 2 4 8 163264126 0.125 0.5 1 2 4 8 163264126
xmm
y, Ra,Rf h',s mm
1.5 ( f ~ ' \ l Material 3 \,171 MateriaL20A 150
\[
'" I e=0199 "\4 r e = 0.359
1.C "4 , ,\ . R'=9 - ',', '\~'<
\ -\ R'=I~ - 100
i5o
-7.1. --- / " " ~ S "~
o
0.125 0.5 1 2 4 8 16 32 64128 0.125 0.5 1 2 4 8 163264126
xmm
FIG. 9(b). Calculation Results for Materials Tested by Kenney and Lau (1985,
1986)
48
out height, h' -- 49 mm, is the average height corresponding to the amount
of washed-out grains of size xo. On the other hand, ht' ~ 90-135 mm, is
the maximum height, at which any loss of grains was found in the material
remaining in the permeameter after the washout.
As mentioned before, the assumption of formation of discrete filter layers
(Figs. 1 and 2) is an idealization. In reality, individual washed-out grains
of a certain size, x, come from heights that vary considerably about the
average height, h'. The grading curves observed by Kenney and Lau (1985,
1986) for filter layers after washout show that some fine grains remained in
coarse filter layers. The average height of washout, therefore, is smaller
than its maximum height. This must be considered in practical design by
means of multiplying the calculated height of washout, h', by a factor of
safety. There is not yet sufficient basis for establishment of the magnitude
of this safety factor but, just as an example, the test with the material As
gives a factor ht'/h' = 1.8-2.8.
According to the calculations, the material Ds should be stable, but for
a small fraction of loose grains smaller than Rfa.x~ = 0.47.1.41 = 0.66
ram. Possible washout of these grains in the test may be hidden by the
spread of the grading curves shown by Kenney and Lau (1985) for the layers
2 - 6 , inclusively.
For the five materials, F, U, 2, 3, and 20A, included in Figs. 9(a and b),
the calculated washouts were larger or much larger than the washouts in-
dicated by the test results. The exact reason is unknown, but it may be the
small hydrodynamic numbers, R', which for the materials F, U, 2, and 3
were smaller than 10, and for the material 20A equal to 18. On the other
hand, R' = 29, 25, and 76 for the tests with the materials A , As, and Ds,
respectively. A conclusion from Figs. 6, 7, and 9 is then that hydrodynamic
numbers, R' > 15 or 20, and vibrations are required in order to get washout
to the extent predicted by the present mathematical model.
Lafleur's Tests
Lafleur et al. (1989) made two series of tests. The materials tested in the
series called screen tests consisted of spherical glass beads with smooth
surfaces. A 230-mm-thick layer of test material was placed on top of a square
mesh screen in a permeameter and submitted to downward hydraulic gra-
dients ranging from 2.5 to 6.5 and periodical vibrations.
Fig. 10 shows calculation results corresponding to the screen tests. The
initial dry-unit weights reported for the test materials gave by (2) c = 0.6
and d = 0.1. c = 0.6 also corresponds to the porosity, n ~ 0.38, of well-
densified random packings of unisized spheres and fits a scale, c = 0.6,
0.73 and 1.00, for spherical, naturally rounded and crushed grains, respec-
tively [cf. text after (2)]. The relative density indices varied between 5%
and 28%, which explains why d > 0. Because of the limited number of tests
and test materials, the choice of the coefficients a, b, and df was uncertain.
a = 0.67 was chosen, only because it is the average value for the loosest
(cubic) and densest (rhombohedral) stable packings of monosized spheres.
b = 2.5 was chosen in conformity with the previous evaluations, dI was
49
0.5
0 ~ I - ~ - ~ . ~=- : '. ,0
0.063 0.250-5 1 2 4 8 1632
xmm
FIG. 10. Calculation Results for Materials Tested by Lafleur et al. (1989)
determined by (15), but this time with a largest value, which made the
calculated and measured total weights of the washed-out grains equal for
each test. dI -- 0.113 for the test M42-T4, and dI = 0.040 to 0.047 for the
tests M8-T5, M6-T20, and M6-T6. (In the test M8-T20, the washout of
grains does not seem to have finished.) Compared with d = 0.1 and with
regard to the final and initial relative density indices reported for the tests,
these values of dy are logical. Loose initial packing, spherical grain shape,
and smooth grain surfaces obviously facilitated rearrangement of the grams
into a more dense structure during the tests. In the writer's tests, dense
initial packing, irregular grain shape, and rough grain surfaces probably
brought about interlocking effects, that hindered rearrangement of the grains
towards dense packing.
According to Fig. 10, both the materials M42 and M6 should be stable,
because Ra < 1 and Rf < 1 for all grain sizes, x. The material M8 should
be unstable, because Rf > 1.0 for x < 0.5. The test results agree with these
conclusions.
The base materials tested in the series called compatibility tests had a
large content of small grains (15-50% smaller than 0.10 mm). Therefore,
the hydrodynamic numbers of the tests must have been very small. Tentative
calculations for the base materials B5-B8 in combination with the filter
material F4 also gave amounts of washed-out grains, which were 25-80
times larger than the amounts measured in the tests. Evidently, the present
mathematical model cannot be used when the hydraulic number is small.
The hydrodynamic conditions, and not the grain and pore geometry, de-
termine the extent of grain washout.
CONCLUSIONS
A mathematical model for filter formation (self-filtration) during washout
of grains from cohesionless materials was developed and found to satisfac-
50
ACKNOWLEDGMENT
The preparation of this paper was supported by Grant 890513-5 from the
Swedish Council for Building Research.
APPENDIX I. REFERENCES
Aberg, B. (1978). "A theory for calculation of the void ratio of non-cohesive soils
and similar materials." Hydr. Engrg. Studies Dedicated to Prof. Erling Reinius.,
Bulletin No. TRITA-VBI-97, Hydr. Lab., Royal Inst. of Tech., Stockholm, Swe-
o den, 25-46.
Aberg, B. (1984a). "Kritisk analys av dimensioneringsregler f6r erosionsskyddande
filter (Critical analysis of filter criteria)." Jorddammar (Embankment Dams), Sem-
inar 16th and 17th of Nov. 1983., Bulletin No. TRITA-VBI-126, Hydr. Lab., Royal
Inst. of Tech., Stockholm, Sweden, 211-219 (in Swedish with English summary).
fi,berg, B. (1984b). "Kohesionslrsa jordmaterials portal och permeabilitet (Void ratio
and permeability of non-cohesive granular materials)." Jorddammar (Embankment
Dams), Seminar 16th and 17th of Nov. 1983., Bulletin No. TRITA-VBI-126, Hydr.
Lab., Royal Inst. of Tech., Stockholm, Sweden, 221-237 (in Swedish with English
o summary).
Aberg, B. (1984c). "Filterbildning i sand och grus (Filter formation in sand and
51
Kenney, T. C., and Lau, D. (1985). "Internal stability of granular filters." Canadian
Geotech. J., 22, 215-225.
Kenney, T. C., and Lau, D. (1986). "Internal stability of granular filters: Reply."
Canadian Geotech. J., 23,420-423.
Lafleur, J., Mlynarek, J., and Rollin, A. L. (1989). "Filtration of broadly graded
cohesionless soils." J. Geotech. Engrg., ASCE, 115(12), 1747-1768.
Sherard, J. L., Dunnigan, L. P., and Talbot, J. R. (1984a). "Basic properties of
sand and gravel filters." J. Geotech. Engrg., ASCE, 110(6), 684-700.
Sherard, J. L., Dunnigan, L. P., and Talbot, J. R. (1984b). "Filters for silts and
clays." J. Geotech. Engrg., ASCE, 110(6), 701-718.
Terzaghi, K., and Peck, R. B. (1948). Soil mechanics in engineering practice. John
Wiley and Sons, Inc., New York, N.Y.
Aa : cf. (2);
a = coefficient;
Ba = cf. (3);
B0 = cf. (3)with Ya = 0 ;
b = coefficient;
c = coefficient;
D = filter material grain size;
d = coefficient, base material grain size;
e = void ratio;
ey = fictitious void ratio;
eo = void ratio obtained by modified Proctor compaction;
g = mean chord of single grain;
g = average length of all grain chords in material;
h = total thickness of material after washout thickness of
filter layer after washout;
h0 = total thickness of material before washout;
h' = calculated average height to which grains of size x
should become washed out; thickness of original ma-
terial corresponding to filter layer of thickness h;
h't = average height to which grains of size x were washed
out in test;
= average length of all void chords in material;
i0 = mean void chord obtained by modified Proctor com-
paction;
Ng = number of grain chords;
n = porosity;
q = unit flux (discharge rate divided by total cross sectional
area);
R' = hydrodynamic number, R' = qxs/nv;
Ra = xa/x;
52
Indexes
j: = filter layer;
B, C, and D = specific filter layers; and
5, 15, 50, and 85 = percent grain sizes, D, d, or x.
53