You are on page 1of 9

DOI 10.

1007/s11204-015-9319-7
Soil Mechanics and Foundation Engineering, Vol. 52, No. 3, July, 2015 (Russian Original No. 3, May-June, 2015)

EXPERIMENTAL STUDY ON SUFFUSION


OF GRAVELLY SOIL

Liang Chen,1 Jingchuan Zhao,2 Hongyu Zhang,3 Wen Lei4 UDC 624.131.212:624.131.63
Hohai University, Nanjing, China; 2Guangzhou Metro Group Co., Ltd.
1

Headquarters of Construction Undertaking, Guangzhou, China; 3Jiangsu


Electric Power Design Institute, Nanjing, China; 4China Railway Shanghai
Design Institute Group Co.,Ltd, Shanghai, China

Although internal erosion causes serious hazards to embankments and dams, studies on
the emission of fine soil particles and the development of non-uniform penetration of the
soil foundation after suffusion do not go far enough. Given that suffusion development is
random and complex, this study employed self-designed equipment to conduct tests under
specific conditions. Twelve samples with three different pore size distributions were used
to make a comparative analysis on aspects such as piezometric head, flow rate, coeffi-
cient of permeability, trend of fine particles, and accumulated sand emissions. Further-
more, the coefficient of permeability, void ratio, movement trend of fine particles in sam-
ples, and changes in content of the remaining fine particles were studied. Combined with
theoretical calculations, correlations among the movement trend of fine particles in the
samples, content of remaining fine particles and uneven changes in the permeability coef-
ficient were shown. Using the relationship between permeability and porosity, we derived
a theoretical formula was deduced from the changing amount of accumulated sand emis-
sions over time under fixed upstream total head.

1. Introduction
Flooding is one of the most harmful natural disasters in China, resulting in significant economic
loss. Flooding can also result in suffusion, landslides, and bank collapse, with suffusion showing the
highest occurrence.
Previous scholars have studied suffusion by means of laboratory experiments, theoretical deriva-
tion, and numerical simulations. Mao et al. [1] simulated suffusion in a glass tank under different sand
thickness, embedded depth, soil layer structure, and contact surface roughness using fine sands of a typ-
ical levee foundation of the Beijiang River. An indoor unsteady seepage flow model was established to
investigate the effects of flood peaks and tidal amplitude on suffusion occurrence in an embankment [2].
Liang et al. [3] created a fluid-structure interaction model for the process of suffusion development.
Sinco et al. [4] studied mud-caused sand suffusion. Ming et al. [5] reported on the effect of suffusion
on soil shear strength. Zhou et al. [6] established a three-dimensional suffusion mechanism model using
the discrete element method. Richards et al. [7] studied the potential existence of suffusion in the soil
structure. Koenders et al. [8] established a mathematical model for piping. Richards and Reddy [9] pre-
sented a comprehensive literature review on soil suffusion. Fell et al. [10] proposed a method to esti-
mate dam foundation erosion and suffusion time. Chen et al. [11] applied well flow theory to the study
of piping. Zhou et al. [12] simulated the occurrence and development of sand piping in combination

Translated from Osnovaniya, Fundamenty i Mekhanika Gruntov, No. 3, p. 30, May-June, 2015.
©
0038-0741/15/5203-0135 2015 Springer Science+Business Media New York 135
9 1
7 2
5 3
3
1 4
mm z x 5

6
Fig.1. Apparatus of model tests: 1) sand and water collector; 2) outlet;
3) sample; 4) sand trap plate; 5) buffer area; 6) inlet.

with small-scale micromechanical model tests. In addition, state-of-the-art digital camera visual tracking
technology and digital image processing technology was introduced to indoor model tests of piping [13,
14]. Li [15] took a model-based test using a Plexiglas tank with a ratio of 1:100 to simulate the occur-
rence and development of piping in levee foundations of double layers and different cutoff walls. Zhang
et al. [16] proposed a predictive mechanism model to assess piping possibility. Skempton and Brogan
[17] tested two types of internal instability in soil, simulating the occurrence of the piping. Cundan and
Strack [18] determined that a sand body was constituted by a series of discontinuous soil particles,
which was simulated effectively by the discrete element method.
Most previous research has examined suffusion from either a microcosmic or a macroscopic view
only, ignoring the continuity of the occurrence and development of suffusion over time. Numerical sim-
ulation has considered the continuity of time, but the assumptions and results have shown tremendous
variation with actual situations. At the same time, studies on the emission of fine soil particles and the
development of non-uniform penetration of the soil foundation after suffusion do not go far enough.
Combining prior research results and identifying their limitations, we made certain improvements
in this study, including the development of novel experimental apparatus and analytical methods. Specifi-
cally, the experimental apparatus used in the present research was self-developed, and our experimental
analysis combined the microcosmic and macroscopic view. We divided the experimental samples into eight
units using independent separation but an integral analysis method. Further, based on the continuity of
time, we studied the emission of fine soil particles and development of non-uniform penetration of the soil
foundation after suffusion and considered relevant factors in the process of suffusion changing over time.

2. Suffusion experiment in cylindrical tank


A self-designed cylindrical tank (Figure 1) 50 cm high, with a diameter of 14 cm, was con-
structed. Eight piezometric tubes, numbered 1 to 8 respectively, were installed on the sidewall from the
bottom up. The water flowed into the tank from an inlet and flowed out from an upper outlet. To create
uniform water inflow, a buffer area was located at the bottom of the tank, and a permeable sand trap
plate was set up near Piezometric Tube 1. Geotextile was laid over the sand trap plate.
The coarse particles used in the experiment were quartz sand, with a grain diameter of 2.0 ~
10.0 mm. The fine particles were yellow sand, with grain diameters of 0.1 ~ 0.3 mm for A group sam-
ple, 0.3 ~ 0.5 mm − for B, and 0.5 ~ 1.0 mm − for C. They were mixed at a ratio of 4:1, with the fine
material accounting for 20%. The suffusion type soil was used as the soil sample. Group A was the
product of mixing quartz sand and yellow sand with particle sizes of 0.1 ~ 0.3 mm. The experiment was
conducted under different void ratios. Table 1 lists the blending parameters.
The experiment steps were as follows:
1. The samples were baked in an oven for 24 h at a temperature of 105°C.

136
TABLE 1
Sam- Fine parti- Nonuniform Curvature Specific gravity of Void Maximum Effective Re Critical Permeability
ple cle quan- coefficient coefficient soil particles Gs ratio e flow Q, grain size hydraulic coefficient,
tity, g Cu Cc ml/sec d10, mm gradient cm/sec
fine coarse
A1 2,427.8 0.43 51.56 0.198 0.51 0.9 0.072
A2 2,394.0 0.45 60.78 0.198 0.60 0.575 0.113
A3 34.34 11.88 2.57 2.88 48.97 0.198 0.48 0.455 0.145
2,357.6 0.48
A4 2,323.0 0.50 68.11 0.198 0.67 0.4 0.187
B1 2427.0 0.44 84.34 0.38 1.59 1.55 0.135
B2 2396.0 0.46 117.78 0.38 2.23 1.25 0.109
B3 17.89 6.19 2.60 2.88 36.92 0.38 0.70 0.95 0.161
2357.0 0.48
B4 2304.8 0.51 50.56 0.38 0.96 0.7 0.252
C1 2431.4 0.44 58.31 0.70 2.03 2.565 0.141
C2 2390.0 0.46 84.77 0.70 2.95 2.458 0.178
C3 9.71 3.36 2.62 2.88 93.28 0.70 3.25 1.805 0.251
2356.0 0.49
C4 2280.0 0.53 89.63 0.70 3.12 1.49 0.364

a b c

d e f

Fig. 2. Process of suffusion development in the sample on the


sidewall of cylindrical test tank.

2. The coarse and fine sand particles were blended with a mass ratio of 4:1, and a small amount
of water was added evenly.
3. The sand was hierarchically compacted every 2 ~ 3 cm, and the test began after 24 h saturation.
4. The hydraulic head was slowly elevated. Total head (namely, the head in piezometric tube 1)
was controlled during the test. The selected time steps were 60 s. In the experiment, the hydraulic heads
were slowly increased to 1-2 cm within 60 sec, and the corresponding hydraulic gradient was increased
by 0.025-0.05. During the early period, the hydraulic heads increased every 2 cm and then decreased to
1 cm when approaching suffusion. Thus, the critical hydraulic gradient was not seriously misjudged. At
each hydraulic head level, the values of the piezometric tubes and corresponding times were recorded,
as were the flow values. After suffusion, with the upstream head kept constant, the hydraulic head and
flow changes in the piezometric tubes were recorded frequently, and the sand out of the tank was col-
lected.
5. The test was stopped once the sample was destroyed, when fine grains beating and springs
rolling occurred at two-thirds of superficial area of the sample.
The suffusion development process of the samples on the sidewall of cylindrical test tank was
recorded with a digital camera [19] as shown in Figure 2: a) distribution of fine and coarse particles was
uniform before test; b) sands moved within a certain range when reaching a certain hydraulic head; c)

137
Water stream Permeability coefficient, cm/sec Water stream

A1 A2
a a
a
b a
b
b a a
b

Layers Layers

A3 A4
a
a
a a
a
b b
a b
b

Water stream Permeability coefficient, cm/sec Water stream

Fig. 3. Movement trends of fine particles within samples (Group A): a) for
loss; b) for accumulation; _ _) at the beginning of the test; ____) upon
suffosion; _ _) at the end of the test.

the range of motion for fine particles increased with increasing hydraulic head; d) fine particles in the
sample were removed by water after suffusion; e) after removal of fine particles, suffusion paths were
formed and developed; f) the concentrated seepage paths were formed eventually.
Suffusion occurred at the outlet when reaching critical hydraulic head; at the later stage of suf-
fusion, large amounts of sand surged and tumbled.

3. Processing and analysis of experimental data


The ambient temperature of the test was 10°C, and kinematic viscosity v = 1.306  10-6 m2/sec.
The Reynolds number of each sample was less than 5 (Table 1). The basic requirements for the occur-
rence and development of suffusion were met: the soil was suffusion type soil, the soil particles were
divided into framework grains and movable particles, and the pores among the framework grains formed
a path through which fine particles could be drained.
Taking Group A samples as an example, comparative analysis was conducted on piezometric
head, flow, and accumulated amount of sand.
Suffusion occurred in four samples at 25, 37, 25, and 16 min after the start of the test. At the
beginning, the hydraulic head was evenly distributed but became unevenly distributed when suffusion was
about to occur. The smaller the void ratio, the more uneven was the head distribution. After the occurrence
of suffusion, with the upstream head fixed, the piezometric head began to decline. The smaller the sample
void ratio, the greater the decline in the piezometric head. The range of variation was 10, 4, 3, and 1.5 cm.
The samples were divided into eight 5 cm thick layers. Using Darcy's law, we obtained the aver-
age coefficient of permeability of each soil layer ki. Since the permeable sand trap plate and geotextile
were laid at the bottom of the sample, the coefficient of permeability of the first layer soil is not dis-
cussed here. The variation in permeability coefficients of the remaining soil layers can be expressed as
follows: At the beginning of the test, the soil permeability coefficients were roughly the same; when suf-
fusion was approaching, the permeability coefficients varied unevenly, and the permeability coefficients
of soil layers at both ends increased significantly; after suffusion, curve peaks and troughs appeared and
became more and more evident with the development of suffusion. During the suffusion process, the

138
upstream permeability coefficient showed the greatest change, while the downstream permeability coeffi-
cient increased with increase in the void ratio. In the middle of the sample, peaks appeared at the fourth
sample layer with a relatively small void ratio, and at the fifth sample layer with a larger void ratio.
Before the occurrence of suffusion, flow increased linearly with hydraulic gradient; after suffu-
sion, even when the hydraulic gradient was constant, flow still increased linearly with time. For samples
A1 and A2, with relatively small void ratios, the flow rate increased significantly compared with the
curves before suffusion. However, for samples A3 and A4, with relatively large void ratios, flow main-
tained linear changes before and after suffusion.
According to the assumption that soil volume does not change if fine particles are lost, i.e., the
volume of lost particles transforms into pore volume, the movement of fine particles (Figure 3) was esti-
mated based on the change in permeability coefficients.
The peaks and troughs of the curves suggest the dominant movement of fine particles, namely,
loss and accumulation. Initially, the fine particles did not move, but as the test proceeded, fine particle
movement occurred place within a certain range, and movement of the top and bottom soil layers
became evident. After the occurrence of suffusion, fine particles moved within a greater range, and par-
ticle loss and accumulation evidently increased. The loss of fine particles was most significant upstream,
downstream, and in the fourth and fifth layers of the samples. In the middle of the sample, peaks
appeared at the fourth sample layer with a relatively small void ratio, and at the fifth sample layer, with
a larger void ratio. For samples with relatively small void ratios, since the diameter of aisle space
formed by large particles was small, the loss and accumulation of fine particles was most obvious, and
it was much easier for fine particles to accumulate within a relatively short distance. This is shown in
the A1 and A2 curves.
The variation in cumulative sand production with time is elaborated as follows: Sand emission
was always large for sample A1 due to its large void ratio. The accumulated amount of sand emission
for sample A3 was the smallest, except for the amount accumulated during a short period after suffusion.
If sand composition was the same, the accumulated amount of sand emission was not only related to the
void ratio but also the critical hydraulic head. The greater the void ratio, the looser the particles were,
and the more prone the fine particles were to erosion. Thus, sand emission was relatively large despite
the small critical hydraulic head. In contrast, fine particles were more difficult to erode if the void ratio
was smaller. At the beginning of suffusion, sand emissions were not large, but a large amount of fine
particles ultimately effused under large hydraulic head pressure.
For three sample groups, void ratios were roughly the same, coarse particle grain size and fine
particle content were equal, but fine particle grain size varied.
The larger the fine particle size, the greater the penetration flow and critical hydraulic gradient
(see Table 1) in the event of suffusion. When the hydraulic gradient was fixed, the flow increment of
sample B1 in unit time was greatest while that of C1 was the smallest. The increase in flow was not only
related to particle size but also to the hydraulic gradient. Under a relatively high hydraulic gradient,
flow increased most significantly due to the great loss of fine particles.
For samples in the same group, the larger the void ratio, the greater was the permeability coef-
ficient. For samples with the same void ratio in different groups, the larger the particle size, the greater
was the permeability coefficient.

4. Law of change of permeability coefficient and sand emission amount


Assuming the overall average gradient i for the sample after suffusion was kept constant, Darcy's
Law was applied:
Q = ikA, (1)

where Q is the seepage, k is the average permeability coefficient for the overall sample, and A is the
cross-section area.

139
k10, m/sec
1.2
A1
A2
0.9 A3
A4

0.8

0.3

0.0
0 10 20 30 40 50 min
Fig.4. Variation in the coefficient of permeability with
time during suffusion of samples (Group A).

TABLE 2
Sample a b R2
A1 00.0117 0.0754 0.9855
A2 0.0121 0.1691 0.9857
A3 0.0113 0.2333 0.9909
A4 0.0197 0.2245 0.9951
B1 0.0323 0.1437 0.9615
B2 0.0249 0.1379 0.9862
B3 0.0033 0.1573 0.9832
B4 0.0091 0.2273 0.9712
C1 0.0016 0.1218 0.9117
C2 0.0023 0.1667 0.8053
C3 0.0016 0.2177 0.8194
C4 0.0021 0.3217 0.8930

The permeability coefficient k10 formula [20] deduced by IWHR was also adopted:
2
k10 = 234n3d 20, (2)

where n is the porosity, and d20 is the equivalent grain size varied with time.
The time suffusion occurred was set to zero, and the variation in the permeability coefficient for
Group A with time is shown in Figure 4. See Table 2 for curve fitting results.
The permeability coefficient at other temperatures kt was converted by the following formula:

kt = k10(η10 /ηt), (3)

where ηt and η10 are dynamic viscosity coefficients of water at T°C and 10°C, respectively.
After suffusion, the permeability coefficient increased linearly with time:

k = at + b, (4)

where t is time after suffusion (min), and a and b are constants related to the test and can be determined
via the test.
Sand emission amount at a certain time was recorded. Taking other samples as uniform samples,
we plotted the particle size distribution curve was plotted to obtain d20 at different times. Formula (4)
can be changed as follows:

d10 = 234n3[(d20)t]2. (5)


140
TABLE 3
Sample A2 Sample B1 Sample C4
Time, Accumulated sand emission, g Time, Accumulated sand emission, g Time, Accumulated sand emission, g
min measured calculated min measured calculated min measured calculated
5 17.8 8.4 4 20.6 26.1 3 3.4 4.9
10 47.1 49.9 6 41.3 54.3 10 7.5 6.8
14 95.6 83.4 7 117.2 118.1 18 12 17.9
20 190.6 183.8 8 161 158.2 49 25.8 49.1
25 241 245 10 191.4 243.6 66 40.3 67.6
29 321.5 315.4 14 323.4 374 72 80.1 83.6
34 363.5 357.1 16 413.8 415.6 74 91 91.6
37 392.9 396.9 / / / / / /
44 459.6 443.1 / / / / / /
48 493.7 486.4 / / / / / /

Assuming the sample does not deform after suffusion, then porosity n can be expressed as follows:

Vv + ΔVv V + ΔVsx
n= = v , (6)
Vv + Vsx + Vsc V

where Vv is the pore volume before suffusion, ΔVv is the pore volume increment after suffusion, Vsx is
the volume of fine particles before suffusion, ΔVsx is the volume of eroded particles, and Vsc is the vol-
ume of coarse particles. It is assumed that ΔVv = ΔVsx.
At certain times, the mass of accumulated sand emission m can be obtained as

m = ΔVsx ρs = ΔVsx(ρw)4 CGs, o (7)

where ρs is the particle density of fine particle soil, (ρw)4 C is the density of pure water (1 g/cm3), and
o

Gs is the specific gravity of fine particle soil.


In combination with Formula (4.6) and (4.7), the following can be obtained:

m = (nV − Vv)(ρw)4 CGs. o (8)

Using (3), (4), (5), and (8), we can write the mass of accumulated sand emission amount at time
t as follows:
⎡⎛ at + b ηt ⎞ ⎤
m = ⎢⎜ 3 ⎟ V − Vv ⎥ ( ρw ) 4 o C Gs .
⎢⎣⎝ 234[( d20 ) t ] η10 ⎠
2
⎥⎦

The content of fine particles P can be calculated by the following formula:

⎡⎛ at + b ηt ⎞ ⎤
mx − ⎢⎜ 3 ⎟ V − Vv ⎥ (ρw )4o C Gs
2
η
⎣⎢⎝ 234[( d20 ) t ] 10 ⎠ ⎦⎥
P= × 100%,
M

where mx is the mass of fine particles in the sample before suffusion, M is the total mass of the sample
before suffusion, and other notations as are above.
Samples A2, B1, and C4 were randomly selected to validate the results (Table 3 and Figure5). The
measured and calculated values of accumulated sand emission for each sample had the same variation

141
500
1

Accumulated sand emission, g


5
4
400 2

300

200

6
100
3

0 10 20 30 40 50 60 70 80 min
Fig. 5. Comparison of theoretical calculations and experimental results: measured values for 1) A2;
2) B1; 3) C4, and calculated values for 4) A2; 5) B1; 6) C4.

trend and fitted well in general. Therefore, the correctness and feasibility of the assumptions and
deduced formulas in this study were justified.

Conclusions
1. For samples with the same particle composition but different void ratios: At the beginning, the
hydraulic head and permeability coefficient were uniformly distributed but became unevenly distributed
when suffusion was approached. With the development of suffusion, the peaks and troughs of the per-
meability coefficient curve became increasingly evident, and the loss and accumulation of fine particles
was enhanced. At this time, even when the hydraulic gradient was constant, flow still increased linearly
with time.
2. For samples with the same void ratio but different particle composition: The more uniform the
particle composition, the greater was the critical head; the more non-uniform the particle composition,
the smaller the difference among permeability coefficients of soil layers before suffusion, and the more
significant the increase in the permeability coefficients of soil layers after suffusion. For samples with
relatively non-uniform composition, the range of motion of fine particles was larger. With increasing
non-uniformity, accumulated sand emission increased, and discreteness of particle loss per unit of time
was reduced.
3. With the hydraulic gradient kept constant, assuming that deformation did not occur in the
sample after suffusion, Darcy's law and relationships between the permeability coefficient and porosity
were applied to derive the theoretical formula for the variation of accumulated sand emission and the
content of residual fine particles over time. The feasibility of the formula was verified.

ACKNOWLEDGMENTS
This paper was supported by the Natural Science Foundation of Jiangsu Province (Study on the
Mechanism of Seepage Deformation of Soil under Variable Head, 2015), Post-Doctoral Fund of Jiang-
su Province of China (NO.1402012A) and Fundamental Research Funds for the Central Universities
(No. 2015B06014).

REFERENCES

1. Chang-xi Mao, Xiang-bao Duan, Jin-bang Cai, and Hai-ying Yu, "Experimental study and analysis on piping
of levee in Beijiang River," J. Hydraul. Eng. (2005).
2. Chang-xi Mao, Xiang-bao Duan, Jin-bang Cai, and Jian-hui Ru, "Experimental study on harmless seepage
piping in levee foundation," J. Hydraul. Eng., 11, 46-53 (2004).
142
3. Yue Liang, Jiansheng Chen, and Liang Chen, "Mathematical Model for Piping Erosion Based on Fluid-
Solid Interaction and Soils Structure," GeoHunan Int. Conf., 109-116 (2011).
4. Exequiel Sinco, Xiao Ming, Louay M. Owaidat, and Larry Smith, "Slurry Induced Piping Progression of a
Sand," Int. Conf. Scour and Erosion 2010, ASCE, 408-417 (2010).
5. Xiao M. and Gomez J., Effect of piping on shear strength of levees, Geoenvironmental engineering and
geotechnics: progress in modeling and applications, " Proc. of Sessions of GeoShanghai 2010, Shanghai,
China, 3-5 June, 51-56 (2010).
6. Jian Zhou and Kai-min Zhou, "3D Modeling of Piping Mechanism Using Distinct Element Method,
GeoShanghai Int. Conf. 2010 (2010).
7. K.S. Richards and K.R. Reddy, Experimental Investigation of Piping Potential in Earthen Structures,
Geocongress 2008: geosustainability and geohazard mitigation (2008).
8. M. A. Koenders and J. B. Sellmeijer, "Mathematic Model For Piping," J. Geotech. Eng., 118:943-946 (1992).
9. R. Fell, Chi Fai Wan, J. Cyganiewicz, and M. Foster, "Time for Development of Internal Erosion and Piping
in Embankment Dams," J. Geotech. Geoenviron. Eng., 129, 307-314 (2003).
10. Jiansheng Chen, Xinwen Li, and Weibing Zhao, "Study on piping leakage mechanism," J. Hydraul. Eng.,
9, 48-54 (2000).
11. Jian Zhou, Zhixiong Yao, and Gang Zhang, "Research on piping mechanism in sandy soils based on discrete
element theory," J. Rock Mech. Eng., 4, 749-756 (2008).
12. Zhou Jian, Yao Zhixiong, Bai Yanfeng, and Zhang Jiao, "Meso-mechanical Study on Piping in Sandy
Soils," J. Tong ji Univ. (Nat. Sci.), 36, 733-737 (2008).
13. Jian Zhou, Zhi-xiong Yao, and Gang Zhang, "Meso-laboratory Study on Initiation and Evolution of Piping,"
Chin. J. Undergr. Space Eng., No.5, 842-848 (2007).
14. Guang-xin Li and Xiao-jie Zhou, "Laboratory simulation on generation and evolution of piping in
embankment foundation," Adv. Sci. Technol. Water Resour., No.12, 21-24 (2005).
15. Wo-hua Zhang, Gong-shuan Yu, and Yuan-qiang CAI, "Mechanism model and artificial intelligence method
for prediction and judgment of piping occurring in embankment," Zhejiang Univ. (Eng. Sci.), 7,
902-908 (2004).
16. L. MeDowell-Boyer, J.R. M-Hunt, and N. Sitar, "Particle transport through porous media, "Water Resour.
Res., 22(13), 1901-1921 (1986).
17. C.S.P. Ojha and V.P. Singh, "Determination of critical head in soil piping, J. Hydraul. Eng.," 129,
511-518 (2003).
18. P.A.Cundall, and O.D.L. Strack, "A discrete numerical model for granular assemblies," Geotechnique,
29, 47-65 (1979).
19. Yue Liang, "Reseaches on the charasticstics of pore flow in seepage field and the mathmatical models of piping
erosion," PhD diss., Hohai University, Nanjing, China (2011).
20. Jie Liu, Seepage stability and seepage control on soil in Chinese, Monograph, Beijing, China (1992).

143

You might also like