Q. 98 – R.

COMMISSION INTERNATIONALE
DES GRANDS BARRAGES
-------
VINGT-CINQUIÈME CONGRÈS
DES GRANDS BARRAGES
Stavanger, Juin 2015
-------

LABORATORY TESTS FOR EVALUATION OF THE ACTIONS LIMITING

THE PROGRESSION OF INTERNAL EROSION IN ZONED DAMS (*)

Ricardo CORREIA DOS SANTOS

PhD student, Laboratório Nacional de Engenharia Civil (LNEC)

Laura CALDEIRA
Head of Geotechnical Dept., Laboratório Nacional de Engenharia Civil (LNEC)

Emanuel MARANHA DAS NEVES

Jubilee Full Professor, Instituto Superior Técnico (IST), University of Lisbon (UL)

PORTUGAL

1. INTRODUCTION

Internal erosion and overtopping are the two main failure modes of large
embankment dams. The overall statistics [1] indicate that they account, in almost
equal parts, for about 94% of the total failures. In particular, the failure by internal
erosion phenomena is almost ten times more frequent than by slope instability,
and thirty times more frequent than by sliding of the embankment body induced
by earthquake/liquefaction.

Over the last three decades there has been a growing interest in finding
ways to estimate the probability of failure by internal erosion. This has been done
mainly by using event tree analysis [2]. Event trees allow the decomposition of
the internal erosion failure process into smaller events, which makes them more
easily understood and makes it easier to estimate probabilities. In addition, it
enables greater understanding of how particular zones of the dam contribute to
the assessed probability of failure.

(*)Tests de laboratoire pour l'évaluation des actions limitant la progression de l'érosion

interne dans les barrages zonés.

113
Q. 98 – R. 8

When applied to zoned dams, event tree analysis allows us to take into
account the potential beneficial mechanisms arising from the presence of certain
types of materials in zones immediately upstream of an erosion path in the core
(e.g., the upstream shell, transition zone or upstream filter/drain). These
mechanisms are usually referred to as the flow-limiting action and the crack-filling
action [3]. The former is related with the ability of the upstream soil to limit the
flow that can pass through a flaw (crack or pipe) in the core. The latter involves
transport of particles from the upstream zone into a flaw in the core, filling it all
the way up to a downstream granular zone. There are very few studies focused
on these mechanisms. A detailed literature review about these actions is
presented by Correia dos Santos et al. [4].

Fig. 1 shows the event tree of an internal erosion failure mode in a zoned
dam, highlighting the branches allocated to the progression of erosion phase.
This event tree assumes that the reservoir level is above a transversal crack in
the core, caused, for example, by large differential settlements and/or seismic
action. In addition, the filter is unable to retain the finer particles detached from
the crack in the core. This is particularly valid for dams that have filters that do
not satisfy modern design criteria [e.g. 5]. Foster and Fell [6] identified that the
filters in many existing dams do not meet Sherard and Dunningan’s [7] retention
criterion, being too coarse by design or having been segregated during
construction. A developing continuous pipe/crack has formed in the core, with
stable roof/side walls. If the upstream material fails to limit the flows converging to
the flaw in the core and fails to fill the crack, then the dam may breach if no
mitigating actions are implemented in a timely and effective manner.

Fig. 1
Event tree for internal erosion in the embankment leading to dam failure
Arbre d’événements de l'érosion interne dans le remblai conduisant
à la rupture d'un barrage

114
 Q. 98 – R. 8

PL and PC (indicated in Fig. 1) are the likelihoods of the upstream zone

being effective at stopping internal erosion due to flow limiting and crack filling
action, respectively. Fell et al [8] give a preliminary indication for estimation of PL
and PC. However, these guidelines result from expert experience and judgment,
and has not validated experimentally. These guidelines are associated to various
uncertainties that need to be evaluated.

This paper, first, includes a brief overview of the laboratory tests available
in the literature studying phenomena related with internal erosion of soils. Then, a
description of two laboratory tests is made: the FLET - Flow Limiting Erosion
Test; and the CFET - Crack Filling Erosion Test. FLET and CFET have different
concepts. The FLET is used to evaluate the isolated influence of the flow limiting
action. The CFET is used to evaluate the potential of an upstream fill to be
washed into a pipe in the core, filling it up to a downstream filter. The test
apparatuses and testing procedures are presented. A series of tests performed
with the FLET and the CFET using various types of soils made it possible to
prove its usefulness in the study of the phenomena under consideration. A
summary of the different types of behaviour observed in the FLET and CFET are
presented. Finally, a preliminary evaluation of the factors influencing the flow-
limiting and crack-filling actions is made.

2. BACKGROUNG ON PREVIOUS TESTS

The laboratory test apparatuses used by researchers to study internal

erosion phenomena focus on particular aspects of the process in embankment
dams, including: the effectiveness of filters [9-11]; the dispersivity of clayey
particles [12]; the erodibility of soils [13]; the potential for internal instability and
initiation of suffusion, in particular of coarse grained soils [14, 15]; the ease of
initiation of backward erosion in granular soils [16] or of soil contact erosion in
materials with high permeability contrast [17]. It is notable however that most of
these laboratory tests have been focused on the behaviour of the downstream
filter and/or the base (core) material, disregarding the influence of the presence
of upstream materials. In the majority of the test apparatuses presented in the
literature, an isolated soil specimen simulating the downstream filter or the base
material is used or both are placed in sequential layers.

No previous laboratory tests focused on flow limiting action by upstream

zones was found in literature. In addition, limited previous work has been
conducted on laboratory testing on crack-filling action in zoned dams. In
particular, Maranha das Neves [9] used the crack erosion test (CET), developed
at the National Laboratory of Civil Engineering (LNEC), in Portugal, for the
evaluation of the effectiveness of an uniform fine sand as an upstream crack
filler. Fig. 2 (a) shows the grain size curve of tested materials, as well as the
setup of the specimen. The upstream sandy soil was examined together with a

115
Q. 98 – R. 8

clayey core, and a sandy downstream filter (Filter A or Filter B). Five tests were
performed on four crack orientations: horizontal, with a +/–20° slope with
horizontal (for upward and downward flow), and vertical (simulating a vertical
transversal crack). The four selected crack orientations were assessed with a
crack of 5 mm in width. The fifth test was carried out with a horizontal crack of
2.5 mm in width. The test with the vertical crack (and horizontal flow) was carried
out with the coarser filter A. All the other tests were performed with filter B. His
main finding was that significant crack filling occurs for all the tested cases and
materials, even at low flow velocities (around 0.02 m/s). For example, Fig. 2 (b)
shows the test conducted with the sloping crack at 20º for downward flow. In this
test, a complete crack filling was observed, although it was necessary to halt the
test in order to open the cover Perspex plate and refill the upstream zone of the
cell with more sand.

Fig. 2
Crack erosion tests by Maranha das Neves [11]:
Essais d'érosion en fissures par Maranha das Neves [11]

The following chapter shows the details of a new test cell, which was
developed at LNEC specifically for the evaluation of both the flow limiting action
and the crack-filling action. This device is able to test specimens composed of an
upstream material, a core (base) material, with a preformed erosion path, and a
downstream filter material. In addition, its large dimension allows coping with the
use of materials composed by some particles with size up to the gravel.

116
 Q. 98 – R. 8

3. NEW LABORATORY TESTS (FLET AND CFET)

3.1. CONCEPT OF AND TERMINOLOGY OF TESTS

The upstream flow limiting action and the crack-filling action may occur
simultaneously. However, due to the complex interactions between them,
experimentally, it has been found more practical to model each action separately.
This approach provides a better initial understanding of the individual contribution
of each action in the internal erosion process. Thus, two laboratory tests with
distinct concepts are developed in this research, referred as:

 The Flow Limitation Erosion Test (FLET), to study the isolated influence of
the flow restriction action due to the presence of materials located
upstream of the core.
 The Crack Filling Erosion Test (CFET), which is focused on evaluating the
ability of crack-filling action by particles eroded from granular materials
located upstream of the core.

Two basic assumptions are considered in the FLET: (1) No downstream

filter is used for the evaluation of the isolated action of upstream flow restriction.
The role of the filter might overlap the influence of the upstream material. In other
hand, this can simulate a scenario in which the downstream filter also holds a
crack when subjected to deformations that cause cracks in the adjacent core.
Some authors [11] have conducted laboratory tests in which this behaviour was
observed, in particular, in filters with some percentage of plastic fines; (2) The
developing pipe through the core is usually extended to the upstream fill zone.
This simulates a scenario in which the mechanism that causes cracking/high
permeable zone in the core is also likely to affect the upstream zone. This
condition corresponds to the worst-case scenario, that is, it is related to a higher
likelihood that flow will not be limited, according with [8].

There are three basic assumptions that are considered in the development
of the CFET: (1) very little benefit to be had from crack-filling action when there is
no downstream filter/transition zone. Thus, the CFET is always carried out with a
downstream filter layer. (2) the downstream filter layer may not properly filter the
particles eroded from the core, eventually for being too coarse to fulfil the no
erosion filter criteria. However, the filter layer has an important role in the crack-
filling action, by retaining some of the particles that are washed from the
upstream material. The potential benefits of crack-filling action may arise from the
compatibility between the particle sizes of the upstream material and of the
downstream filter; and (3) the upstream material is not drilled. The crack-filling
action is more likely to occur whenever a granular material is present upstream of
the core. Thus, the majority of the upstream materials tested are granular soils
with no fines or a low fines content that cannot sustain an open crack or pipe.

117
Q. 98 – R. 8

3.2. DEVELOPED TEST CELL

Fig. 3 shows the new test cell developed. It was designed so that it can be
used either as FLET or CFET. The construction of all the pieces of the cell was
made entirely at LNEC, in Lisbon.

Fig. 3
Design drawings of the cell test after assembly [18]
Dessins de conception de la cellule d'essai assemblée [18]

The cell is composed mainly of aluminium parts to provide light weight and
corrosion resistance. Its main structural parts are three aluminium plates, three
aligned aluminium tubes (with an inner diameter of 280 mm), which are also used
as moulds for easy compaction, and a circular cover plate of acrylic glass. The
different pieces are sequentially assembled, as the core and upstream materials
are being compacted, using threaded steel rods and nuts. Sealing is achieved
using circular O-rings placed inside grooves on the plates. The test cell allows the
compaction of a specimen composed of a core 170 mm long, a filter 120 mm
long, and upstream material up to 250 mm long.

The cell has a water inlet chamber, where, to ensure lateral support for the
upstream material, two concentric springs of similar stiffness and a rigid annular
plate (5 mm thick and 50 mm inner diameter) are installed. In tests without a
drilled axial hole on the upstream material, this plate is replaced by a porous one,

118
 Q. 98 – R. 8

for a more homogeneous distribution of the flow at the entrance of the specimen.
In the FLET, at the downstream side, an outlet chamber is adopted to collect the
eroded material from the pipe in the core. The acrylic glass plate allows direct
visualization into this chamber. In the CFET, the downstream filter is compacted
within the outlet chamber. The inlet and outlet chambers have a purge valve used
for air release/entrance during the filling/empting of the test cell.

3.3. SETUP OF TESTS AND MEASUREMENTS

Fig. 4 (a) and Fig. 4 (b) show a schematic diagram of the setup of the FLET
and of the CFET, respectively. The soil specimen is composed of an upstream
material, a core, and, in the CFET, additionally of a downstream filter. These are
compacted sequentially inside the cylindrical cell. A hole is drilled in the centre of
the core to model a concentrated leak. In the FLET, the hole is also drilled in the
upstream material.

The specimen is subjected to water flow imposed through a constant

hydraulic head loss, H = Hu – Hd. Hu and Hd are the total head at the entrance
and exit of the test cell, respectively. Control valves are placed at upstream and
at downstream of the cell to assist the test procedures.

The measurements made during the tests include piezometric heads, flow
rates, and visual observations through the downstream acrylic glass cover plate.
The piezometric heads are measured, at the pipe level, using plastic tubes
immediately upstream of the specimen, hU/S, near the upstream–core interface,
hINT, and inside the downstream filter, hD/S. hu = hU/S – hINT, and hcf = hINT – hD/S are
the piezometric head losses along the upstream material, and along the core and
core–filter interface, respectively. The flow rate, Q, is measured by an
electromagnetic flow meter placed upstream of the test cell. The erosion
phenomena and the turbidity of the effluent observed through the acrylic cover
plate are recorded automatically using a digital camera, which are then sent to a
laptop at regular intervals for further analysis.

3.4. TESTS PROCEDURE

The core and the upstream material are compacted sequentially in lifts of
35 mm and 50 mm, respectively, with a moist tamping procedure. Dry soil
material, or soil with previously selected water content, is thoroughly mixed with a
known percentage of water. Care should be taken to distribute evenly the soil in
each layer and perform the compaction to the intended density, measuring
indirectly the thickness of each lift using a depth gauge. In the CFET, the granular
filter layer is placed in dry conditions to a desired relative density.

119
Q. 98 – R. 8

Fig. 4
Setup of tests: (a) FLET; (b) CFET
Mise en place des essais: (a) FLET; (b) CFET

Fig. 5 illustrates the procedure used to assemble the device and to

compact the materials in a typical FLET, which follows five main steps. In step 1
the core is compacted inside the mould. An aluminium tube is used as the
compaction collar. In step 2 the excess core material is trimmed. An electric
drilling device and a guide are used to drill a central hole along the core sample.
A rod is inserted in the predrilled hole to prevent it from becoming filled with soil
in the next step. In step 3 the upstream material is compacted on the core
material. In step 4, a central hole is drilled in the upstream material whenever the

120
 Q. 98 – R. 8

material can sustain an open hole. A rigid annular (or porous) plate and two
concentric springs are then inserted. In step 5 the aluminium upstream cover
plate is assembled, exerting force on the springs. The test cell is rotated 90º and
placed in a support. Finally, the aluminium tube, used in step 1 as the compaction
collar, and the acrylic glass cover plate are assembled. The piezometer tubes are
inserted upstream and downstream of the specimen.

Fig. 5
Steps of device assembly and compaction/drilling in the FLET [4]
Étapes de l'assemblage de l'appareil et compactage/forage dans FLET [4]

Additional procedures are required for the preparation of the specimen and
for cell assembly in the CFET. These include: (1) the installation of a geotextile at
the interface between the core and the upstream material (see Fig. 4 (b)). The
geotextile is an annular element (100 mm outer diameter and 40 mm inner
diameter) used to prevent parasitic erosion paths along the interface with core-
upstream material; (2) the installation of a circular wire meshes with a diameter of
100 mm (see Fig. 4 (b)). The wire mesh at the filter-core interface allows
compaction of the filter layer without filling the drilled hole in the core. To
minimize influence of the wire mesh during the erosion process, its opening size
is larger than the controlling constriction size of the filter being tested; and (3) the
vertical compaction of the filter layer to the desired relative density.

The tubing system is connected to the inlet and outlet cambers, ensuring
that the control valves are closed. The FLET/CFET is conducted to a predefined
head differential, H. The piezometric tubes are connected to the readings panel.
After the opening of the air release purges, the cell is slowly filled to prevent soil
erosion caused by rapid passage of water along the pipe. After complete filling of
the cell, the upstream control valve is closed, the downstream control valve is
fully open, and the air release purges are closed. The timer starts at the instant
when the inlet control valve is fully open.

A test is deemed complete when any of the following criteria is verified: (i)
Q < 50 litres/hour during 30 minutes; (ii) all readings are stable during 30 minutes
with clear outflow; (iii) hu or hc below 20 mm; (iv) Q reaches the maximum
discharge capacity that the system withstands (about 1700 l/h); and (v) maximum
test duration of 3 hours. The test stops closing the upstream control valve.

121
Q. 98 – R. 8

After careful disassembly of the cell, the core, upstream material, and filter
samples are obtained. Whenever possible the resulting pipe in the core (and in
the upstream soil in the FLET) is filled with melted paraffin wax. The paraffin
moulds allows the inspection of the pipe shape and roughness. In addition, the
volume of the paraffin moulds is used for estimation of the final equivalent
diameter of the pipe. In the CFET, to evaluate the erosion loss in upstream zone,
grain-size distribution of samples taken from different zones is performed. The
quantification of the amount of particles retained in the filter is also carried out,
considering the dry weight difference in relation to the initial filter weight.

4. EXPERIMENTAL STUDY PERFORMED WITH THE FLET/CFET

4.1. MATERIALS SELECTED

Fig. 6 shows the grain-size distribution curves of the various soils used in
the laboratory tests as core, as upstream material, and as filter (in the CFET).
Table 1 summarizes the main physical and compaction properties of the soils that
can hold an open pipe used as core and as upstream material. Table 2
summarizes the main physical properties of the granular soils used as upstream
material and as filter. Table 2 also shows the maximum and minimum dry unit
weights of the granular soils, obtained from standard density tests.

Fig. 6
Soils tested in the FLET/ CFET
Sols testés dans le FLET/CFET

122
 Q. 98 – R. 8

The upstream materials tested in the FLET (N1, N2, N3, P1 and P2) are
coarse well graded soils (coefficient of uniformity Cu > 6), obtained by sieving
(with 2”, 3/4”, or 3/8” ASTM sieves) of the materials collected during the
construction of two zoned dams in Portugal. They have been selected so that
their fines content (pf200) is within the limits referred to guidance information
listed available in [8]. Upstream materials N1, N2, and N3 are from Ribeiro
Grande Dam (with non-plastic fines) and upstream materials P1 and P2 are from
Odelouca Dam (with plastic fines).

Table 1
Main properties of tested soils that are capable to hold an open pipe, used in the
upstream zone (N1, N2, N3, P1 and P2), and the core (Core#4 and Core#20)

Compaction
Atterberg limits (c)
Soil pf200 (a) pc4 (b) ASTM parameters (d)
types (%) (%) D2487 wopt d,max
wL (%) Ip (%)
(%) (kN/m3)
N1 35.1 14.3 NP NP SM 11.2 19.1
N2 29.9 27.1 NP NP SM 10.9 19.5
N3 20.9 49.0 NP NP GM 8.6 20.6
P1 16.9 56.1 38 14 GC 12.7 19.4
P2 11.8 69.2 38 14 GP-GC 10.0 20.2
Core#4 37.0 0.0 38 14 SC 14.4 19.0
Core#20 63.0 0.0 38 14 CL 17.2 17.8
(a) pf200= Fines content, i.e., the mass fraction, in percentage, of soil particle finer than 0.075 mm
(no. 200 sieve). (b) pc4 = Gravel content, i.e., the mass fraction, in percentage, of soil particles
coarser than 4.75 mm (no. 4 sieve). (a) NP= Non-Plastic. (a) ASTM D698 was used.

Table 2
Main properties of granular soils used in the upstream zone (A0, G1-GA4, GN,
and GP) and the filter layer (Filter S, and Filter G)

Grading Atterberg Grading ASTM ASTM ASTM D4254

Soil properties (a) limits (b) coefficients (c) D2487 D854 D4253
types pf200 pc4 pA0 a wL Ip Soil (d)
d,min d,max
Cu Cc G
(%) (%) (%) (%) (%) class. (kN/m3) (kN/m3)
A0 0 0 100 – – 1.9 1.0 SP 2.65 15.1 16.6
GA1 0 85 10 – – 8.6 2.6 GW 2.72 15.2 18.1
GA2 0 80 15 – – 59 14 GP 2.72 16.6 18.7
GA3 0 74 20 – – 66 10 GP 2.72 17.3 19.6
GA4 0 68 30 – – 69 0.4 GP 2.72 17.6 20.0
GN 5 68 25 NP NP 90 0.3 GP-GM 2.72 17.7 20.2
GP 5 68 25 38 14 90 0.3 GP-GC 2.72 17.6 20.1
Filter S 0 60 0 – – 2.6 0.5 GP 2.60 14.5 16.9
Filter G 0 90 0 – – 1.4 0.8 GP 2.60 14.4 16.4
(a) pf200 = Fines content; pc4 = Gravel content; pA0 = Percentage of fine sand (soil A0) in mixture.

(b) Atterberg limits (of the fines); w = Liquid limit, I = Plasticity index, NP = Non-plastic. (c) C =
L p u
Coefficient of uniformity; Cc = Coefficient of curvature. (d) G = Specific gravity.

123
Q. 98 – R. 8

The upstream materials tested in the CFET (A0, GA1 to GA4, GN, and GP)
are granular soils with no or very few fines content. Soil A0 is similar to the
uniform fine sand (silica) used as upstream filler by Maranha das Neves [9],
which he proved to be highly effective at filling in cracks. The remaining soils are
gap-graded soils which do not comply with the Kenney and Lau’s criterion [15],
and thus present high susceptibility to suffusion (are internally unstable soils).
GA1 to GA4 are fines free, GN contains 5% of non-plastic fines, and GP contains
5% of fines with some plasticity (Ip = 14%). Soils GA1, GA2, GA3 and GA4 are
mixtures containing a content of fine sand (soil A0), pA0, respectively, 10, 15, 20
and 30%. Upstream gap-graded soils GN and GP have pA0 = 25%.

Two soils were used as core (Core#4 and Core#20), which were obtained
by sieving of a material with fines and resulting from schist weathering. This
material was collected during the construction of Odelouca Dam, which has
plastic fines. All FLETs and most of the CFETs were carried out with soil Core#4
(37% of fines), which is the fraction of the material collected that is finer than No.
4 sieve (4.75 mm). A few CFETs were performed with a finer soil, termed
Core#20 (63% of fines), corresponding to the fraction passing the No. 20 sieve
(0.84 mm).

Each CFET was carried out with one of the two coarse-grained granular
soils selected as downstream filter. One is classified as gravelly sand (filter S)
and the other as sandy gravel (filter G). Both these soils are fine-free.
Excessively coarse filters are selected intentionally to not meet the USBR
retention criterion for no erosion [5] of either Core#4 or Core#20 (D15F of filters is
higher than the allowable maximum value, as shown in Fig. 6)

4.2. LOADING AND SPECIMEN COMPACTION CONDITIONS

4.2.1. FLET

A total of 14 FLETs were conducted. The tests were carried out with a head
loss of H = 2050 mm. This corresponds to the application of a hydraulic gradient
through the hole of the core/upstream material specimen of about 4.8. All FLETs
were carried out with a drilled hole, with an initial diameter of Di = 12 mm.

The influence of the following parameters of the upstream material was

evaluated: gradation (pf200 and pc4); the nature of the fines (plastic or non-
plastic fines); and compaction conditions (water content ranging between about
wopt – 2% and wopt + 2%, in specimens prepared to degree of compaction of 95%).

124
 Q. 98 – R. 8

4.2.2. CFET

A total of 26 CFETs were conducted. The majority of the CFETs were

carried out on specimens with a predrilled hole of 12 mm diameter, Di. In a few
tests, however, a preformed hole of larger diameter was also considered. In
particular, diameters of 14 and 16 mm were examined. Most of the tests were
performed for total head loss of 2050 mm. In a few tests using the upstream soil
A0, head loss of 1050 mm was also tested.

Upstream soils A0, and GA1 to GA4 were always prepared with water
content of 3.5%. Upstream soils GN and GP (with 5% fines) were prepared with
water content of about 7%. Soil A0 was compacted to a target relative density of
60% (maximum compaction achieved using a tamping method). Upstream soils
GA1 to GA4, GN and GP were compacted to a relative density around 100%.
Core#4 was always prepared near the optimum water content, and core#20 was
always prepared 2.5% wetter than the optimum water content. Both core soils
were compacted to a degree of compaction of 95%. For these compaction
properties, core#4 and core#20 have an erosion rate index in the hole erosion
test [13], IHET, around 4.1 [19], which is associated to a moderate soil erosion
behaviour. All the CFETs were carried out with a filter (S or G) prepared to a
relative density, Dr, of 60 +/- 8%. It was considered the conservative approach to
test filters with a relative density somewhat smaller than the minimum 70%
proposed by USBR [5].

4.3. OBSERVED BEHAVIOUR TYPES IN LABORATORY TESTS

4.3.1. FLET

The evaluation of the consequences of the evolution of the internal erosion

process was accomplished visually and by examination of the data collected from
the readings. Four types of behaviours were identified: flow stopping (or
decreasing greatly) due to upstream pipe collapse (Type F1); flow limitation to a
threshold value (Type F2); no flow limitation but slowing down of core erosion
(Type F3); and no flow limitation along with the progression of severe erosion
(Type F4). Fig. 7 (a) shows the trend of the flow rate for each behaviour type, and
illustrates the typical layout of the specimen at the end of the tests. Fig. 7 (b)
shows some pictures for each behaviour type observed.

The tests showing behaviour Type F1 are characterized by an extremely

rapid erosion of the fines in the inner surface of the erosion pipe in the upstream
material, followed by the release of the surrounding coarser particles. At this point
the flow rate increases relatively fast. Then, some of those coarser eroded
particles are trapped at the soils interface and/or inside the core pipe, which has
no time to widen, since the core material has considerably higher erosion
resistance than the upstream material. The flow rate starts to diminish, but even

125
Q. 98 – R. 8

so, the upstream material continues to erode, leading to the successive filling of
the voids between the trapped particles with even smaller particles. This results
in the progressive limitation of the water flow and collapse of the pipe in the
upstream material, due to prolonged soaking (self-healing process).

Fig. 7
Behaviour types observed in the FLET
Types de comportement observés dans le FLET

126
 Q. 98 – R. 8

Behaviour Type F2 occurs when the upstream soil is non-erodible for the
applied hydraulic load. The initial increase of the flow rate is mostly due to the
widening of the pipe in the core. In general, this occurs together with the drop of
the piezometric head in the core, hc, and with a slight increase of the piezometric
head in the upstream material, hu, caused by a redistribution of the hydraulic
head loss. Consequently, the shear stress due to the eroding fluid on the surface
of the pipe in the core decreases below its c core cr). The
erosion process in the core slows down substantially, or can even stop, resulting
in the stabilization of the flow rate.

In tests showing behaviour Type F3, an initial trend toward increased flow
rate occurs. The fines of the upstream material erode at a faster rate than the
core. This results in a rapid release of the coarser particles of the upstream
material, whose maximum dimension is higher than the diameter of the pipe in
the core. As in behaviour Type F1, these particles are transported downstream
and become trapped at the interface with the core. The flow restriction inhibits the
dragging of the coarser particles of the upstream material, resulting eventually in
near-stabilization of the readings. Nevertheless, as opposed to behaviour Type
F1, the flow rate is sufficiently high to continue to erode the core. When the pipe
size in the core becomes too large to retain the gravel particles from the
upstream material the flow rate increases with no great restriction.

In tests showing behaviour Type F4, the flow rate increases continually at a
very fast rate, mainly because of a high erosion rate of the upstream material. A
common characteristic of these tests is that, despite the low erosion resistance of
the upstream soil, the developing erosion pipe sustains its roof during the
progression of erosion.

4.3.2. CFET

The CFET allows three main types of behaviours to be identified: rapid

‘crack-filling’ with almost ‘no erosion’ of the core (Type C1); filtering after ‘some
erosion’ (Type C2a) or after ‘excessive erosion’ (Type C2b) of the core and/or
upstream material; and ‘continuing erosion’ of the core and upstream material
(Type C3). Fig. 8 (a) shows the trend of the flow rate for each behaviour type.
Fig. 8 (b) illustrates the typical layout of the specimen at the end of the tests and
some pictures for each behaviour type.

The tests showing behaviour Type C1 are characterized by an extremely

rapid transport of a considerable amount of particles of the upstream material at
the interface with the core. The washed-in particles travel along the pipe in the
core (forming a ‘sand jet’) up to the downstream filter face. The flow rate falls
abruptly in an earlier stage of the test.

In tests of Type C2a or Type C2b, the flow rate starts increasing fast,
mainly due to the development of suffusion in the upstream material. The

127
Q. 98 – R. 8

downstream filter is unable to retain the material coming from the upstream
material. At a given instant, the hydraulic shear stresses, applied to the inner
surface of the pipe in the core, reach values high enough to start detaching the
sand-size particles, which are then retained at the downstream filter face. This
filtering mechanism leads to a progressive slowdown of the increase of the flow.

Fig. 8
Behaviour types observed in the CFET
Types de comportement observés dans le CFET

128
 Q. 98 – R. 8

In Type C2a the filter seals after ‘some’ erosion of the core. In Type C2b
the filter seals only after the erosion of an ‘excessive’ amount of soil from the
pipe. The eroded particles are transported by flow up to the filter face, and, then,
slowly seep into the filter. This leads to two opposite effects on the flow rate. On
the one hand, the widening of the pipe diameter tends to increase the flow, and,
on the other hand, the decrease of the filter permeability hinders seepage flow.
This balance may result in an increase or stabilization of the flow for a period.

In tests of Type C3, the flow rate increases fast. There is a rapid
progression of suffusion in the upstream material, and the filter is too coarse to
retain those particles. The flow rate may eventually stabilize if the maximum
discharge capacity of the hydraulic system is reached.

4.4. PRELIMINARY DISCUSSION OF FLET/CFET RESULTS

4.4.1. FLET

The flow-limiting action is shown to be strongly dependent on the gradation

of the upstream soil (fines and gravel contents), the nature of the fines (plastic or
non-plastic), and the compaction water content. In general, soils with plastic fines
appear to be associated with higher likelihoods of being effective at limiting the
flow than soils with non-plastic fines. The finer the soil the higher seems to be its
capability to limit the flow. This trend may not occur with upstream soils that are
considerably coarser than those tested, in particular, in soils with high hydraulic
conductivity and that are unable to sustain an open pipe.

Upstream flow limiting is more likely to occur: (1) in soils prepared

considerably on the dry side (2% from optimum water content or drier), because
of their self-healing capabilities; and (2) in soils with plastic fines prepared near
optimum water content or wetter, because they should have high resistance to
erosion along a concentrated leak.

4.4.2. CFET

The analysis of the results of the CFETs showed that the crack-filling action
(Type C1) occurs over a very short period, and is mainly controlled by some of
the properties of the upstream zone and of the filter. Core soils with an erosion
rate index higher than 4 (i.e., moderately slow erosion or less erodible), should
not have an influence on the crack-filling action. Core should influence however if
filtration occurs (Types C2a or C2b) or erosion progresses (Type C3).

The crack-filling action (Type C1) occurred in tests with filter S together with
upstream soil mixtures of pA0 ≥ 20% and pc4 ≤ 74% (soils GA3, GA4, GN, GP
and A0), and in tests using filter G together with soils of pA0 ≥ 30% and

129
Q. 98 – R. 8

pc4 ≤ 68% (GA4 and A0). Gap-graded soils with only 5% of fines appear to have
a lower likelihood of being effective at filling in cracks, even if the fines are of a
non-plastic nature, than a gap-graded soil of similar grain-size distribution curve
but with no fines. The relation between the content of sand that is susceptible to
suffusion of the granular upstream zone and the equivalent diameter D15F of the
filter appears to be the critical parameters. For a given flaw size and loading
condition, the higher the content of sand susceptible to suffusion, and the lower
the D15F, the greater are the chances of the washed-in particles being caught at
the filter face, and filling in the flaw in the core.

5. CONCLUSIONS

This study addresses the mechanisms limiting the progression of internal

erosion in the embankment of zoned dams. When an open erosion crack/pipe is
under development in the impervious element of the embankment, the unwanted
formation of a breach may only be avoided because of the crack-filling action or
the upstream flow-limiting action.

Dam engineers have relied on general guidelines, available in literature, to

evaluate whether upstream zones are effective, or not, to prevent the erosion
progression. These guidelines are based on practical experience and the
judgement of experts, but have not until now been checked using laboratory
testing. For that reason, they have been linked to a great deal of uncertainty.

The development of the FLET and CFET, and the laboratory work
presented constitutes a relevant step forward for the validation and extension of
some of the recommendations in those guidelines, and for practical applications.
Extensive laboratory testing was carried out with the aim of investigating in
particular the potential of upstream materials to provide the flow-limiting action
and the crack-filling action.

ACKNOWLEDGEMENTS

The authors wish to acknowledge the financial support of the sponsors of

the research, LNEC (Laboratório Nacional de Engenharia Civil) and the FCT
(Fundação para a Ciência e a Tecnologia) (PhD SFRH/BD/76006/2011).

130
 Q. 98 – R. 8

REFERENCES

[1] M. Foster, R. Fell, M. Spannagle, Canadian Geotechnical Journal, 37

(2000) 1000-1024.
[2] USBR, USACE, Best practices and risk methodology. Chapter 26:
Internal erosion risks (draft version from November 26, 2012).
[Link] (last access
15 October 2014), in: USBR (Ed.), U.S. Department of the Interior,
Bureau of reclamation, 2012.
[3] R. Fell, C.F. Wan, J. Cyganiewicz, M. Foster, Journal of Geotechnical
and Geoenvironmental Engineering, 129 (2003) 307-314.
[4] R. Correia dos Santos, L. Caldeira, E. Maranha das Neves,
Geotechnical Testing Journal, 37 (2014) 463-476.
[5] USBR, Design standards No. 13. Embankment Dams. Chapter 5:
Protective Filters. Phase 4 (Final), in, US Bureau of Reclamation,
Denver, CO, 2011.
[6] M. Foster, R. Fell, Journal of Geotechnical and Geoenvironmental
Engineering, 127 (2001) 398-407.
[7] J.L. Sherard, L.P. Dunnigan, Journal of Geotechnical Engineering,
ASCE, 115 (1989) 927-947.
[8] R. Fell, M. Foster, J. Cyganiewicz, G. Sills, N. Vroman, R. Davidson, A
unified method for estimating probabilities of failure of embankment
dams by internal erosion and piping (draft guidance document dated
August 21, 2008). in: R.a.f.d. safety (Ed.), The University of New South
Wales, URS, and US Army Corps of Engineers, 2008.
[9] E. Maranha das Neves, Analysis of crack erosion in dam cores: the crack
erosion test in: De Mello Volume: a tribute to Prof. Dr. Victor F.B. de Mello,
Editora Edgard Blucher Ltda, São Paulo, Brazil, 1989, pp. 284-298.
[10] .L. Sherard, L.P. Dunnigan, Filters and leakage control in embankment
dams, in: Volpe and Kelly (Ed.) Symposium on Seepage and Leakage
from Dams and Impoundments, ASCE, 1985, pp. 1-30.
[11] T.L. Brandon, Y. Park, J.M. Duncan, Geotechnical Testing Journal, 30
(2007) 48-59.
[12] J.L. Sherard, L.P. Dunnigan, R. Decker, E.F. Steele, Journal of
Geotechnical Engineering, ASCE, 102 (1976) 69-85.
[13] C.F. Wan, R. Fell, Geotechnical Testing Journal, 27 (2004) 295-303.
[14] C.F. Wan, R. Fell, Journal of Geotechnical and Geoenvironmental
Engineering, 134 (2008) 401-407.
[15] T. Kenney, D. Lau, Canadian Geotechnical Journal, 23 (1986) 420-423.
[16] J.H. Schmertmann, The no-filter factor of safety against piping through
sands, in: Judgement and Innovation - The Heritage and Future of the
Geotechnical Engineering Profession, ASCE, 2000, pp. 65-132.
[17] C. Guidoux, Y.-H. Faure, R. Beguin, C.-C. Ho, Contact Erosion at the
Interface between Granular Coarse Soil and Various Base Soils under
Tangential Flow Condition, in, ASCE, 2010, pp. 741-750.

131
Q. 98 – R. 8

[18] R. Correia dos Santos, Experimental investigation on limitation of the

progression of internal erosion in zoned dams (submitted), PhD, UL,
Instituto Superior Técnico, Lisbon, 2014, pp. 315.
[19] R. Correia dos Santos, L. Caldeira, E. Maranha das Neves, Revista
Geotecnia, 125 (2012) 36.

SUMMARY

Two mechanisms limiting the progression of internal erosion in zoned

dams, potentiated by the presence of materials located upstream of a damaged
core are investigated. These are the flow limiting and crack-filling actions. The
former is related with the ability of the upstream soil limiting the flows passing
through a flaw in the core. The latter involves transport of particles from the
upstream zone into a flaw in the core, up to a downstream granular zone. A new
apparatus is developed from scratch. It allows carrying out two innovative
laboratory tests: the Flow Limitation Erosion Test (FLET) and the Crack Filling
Erosion Test (CFET). Following an extensive testing programme on twelve
upstream materials, two cores, and two downstream filters, experimental
observations and physical descriptions are presented. The potential soil
behaviour patterns are described. The flow limiting action is governed by grading,
plasticity and compaction characteristics of the upstream soil. Crack filling by
granular soils is a rapid mechanism, being governed mainly by the compatibility
of the washed in material from the upstream soil with the downstream filter. In
general, the higher the sand and the lower the fines content of the upstream soil,
and the lower the effective grain diameter D15F of the filter, the higher the
likelihood of the crack-filling action to occur.

RÉSUMÉ

Cet article étude deux mécanismes qui limitent la progression de l'érosion

interne des barrages zonés, induites par la présence de matériaux situés en
amont d'un noyau fissuré. Ces mécanismes sont généralement appelés limitation
de l’écoulement et remplissage des fissures. Le premier est lié à la capacité du
sol amont de limiter les écoulements à travers un défaut dans le noyau. Le
second est associé au transport des particules vers une fissure du noyau et à
leur rétention par le filtre. Un nouvel appareil a été développé permettant la
réalisation de deux types d’essais innovants en laboratoire : le test du contrôle
d’érosion par limitation de l’écoulement (cf. Flow Limitation Erosion Test - FLET)
et le test du contrôle d’érosion par remplissage de la fissure (cf. Crack Filling
Erosion Test - CFET). Un programme d’essais a été réalisé sur douze matériaux
de recharge amont, deux matériaux de noyau et deux types de filtres. Les

132
 Q. 98 – R. 8

comportements observés des sols sont présentés ici. Cette étude permet de
conclure que la limitation d’écoulement est déterminée par la granulométrique, la
plasticité et le degré du compactage du matériau amont et que le remplissage
des fissures par sols granulaires est un mécanisme rapide, surtout gouverné par
la compatibilité des matériaux amont érodés avec le filtre.

133