You are on page 1of 60

Project Report

Isaac Cornish

Bachelor of Engineering (Honours)


Civil and Geotechnical Engineering

The University of Queensland


2020
Isaac Cornish
CIVL4560

Investigating Initiation of Internal


Erosion Within Laboratory
Embankment Dams
by Isaac Cornish

Student Number: 44777128


Course Code: CIVL4260
Supervisor: Adnan Sufian
Submission Date: 27/10/2021

School of Civil Engineering


The University of Queensland
Isaac Cornish
CIVL4560

Acknowledgements
I would like to acknowledge Dr Adnan Sufian for his continuous and ongoing mentoring and
support for this project. I would also like to thank my colleagues ATCwilliams for inspiring me
to pursue geotechnical engineering.
Isaac Cornish
CIVL4560

Abstract
Embankment Dams are susceptible to a plethora of general failure mechanisms. The most
common failure mechanisms are initiated via a process known as internal erosion. This
process is caused through the seepage of water through embankment materials, leading to
the detachment of soil particles at critical interfaces/zones or simply in gap-graded soils that
segregate. The consequences associated with this phenomenon can be devastating to
human life, infrastructure, and the environment. At present, there is a deficiency in
understanding the cause of internal erosion because it is difficult to observe and happens
locally within an embankment. Evaluating the initiation phase of internal erosion is crucial, as
it has the greatest impact on estimated failure probability. This gap has led to design and
construction of embankments being based on empirical methods, past-experience, and
engineering judgement. This project’s primary objective is to verify that the current
understandings of internal erosion are valid whilst simultaneously offering recommendations
and discussing the limitations of these methodologies. This is done though the construction
of laboratory embankment dams with a plethora of zoning schematics under a hydraulic load.
The geometric and hydraulic criterion which dictates internal erosion initiation is verified using
the observations gathered in the laboratory. A qualitative assessment of the trial
embankments showed that grain-size criterion is acceptable in the design of filters, but that
caution should be used since no criteria satisfied all conditions. Moreover, hydraulic design
criterions were unable to be accurately assessed, prompting the need for an improved
laboratory setup. Overall, the initiation of internal erosion is physically understood, but further
research would be required to create a method of predicting how and when it will occur, that
is not based on empirical or analytical evidence.

1
Isaac Cornish
CIVL4560

Table of Contents
1. Introduction .................................................................................................................... 6
1.1 Objective and Aims.................................................................................................. 8
1.2 Project Scope .......................................................................................................... 8
2. Background Research and Literature Review ............................................................... 9
2.1 Embankment Dam Zoning ....................................................................................... 9
2.2 Internal Erosion ..................................................................................................... 10
2.3 Geometric, Hydraulic and Mechanical Conditions ................................................. 11
2.4 Physical Processes of Internal Erosion ................................................................. 12
2.4.1. Concentrated Leak Mechanism .......................................................................... 12
2.4.2. Suffusion Mechanism ......................................................................................... 14
2.4.3. Contact Erosion Mechanism .............................................................................. 14
2.4.4. Backward Erosion Mechanism ........................................................................... 15
2.5 Initiation Criteria of Internal Erosion....................................................................... 16
2.5.1. Concentrated Leaks ........................................................................................... 16
2.5.2. Contact Erosion - Den Adel (1994) and Beguin (2011) ...................................... 17
2.5.3. Suffusion – Witt Method ..................................................................................... 18
2.6 Historic Filter Design ............................................................................................. 19
2.6.1. Basic Theory of Terzaghi Geometric Criterion ................................................... 19
2.6.2. Segregation Criterion ......................................................................................... 22
2.6.3. Basic Theory of Hydraulic Criterion .................................................................... 23
2.6.4. Vertical Critical Gradient (Buoyancy).................................................................. 24
2.6.5. Horizontal Critical Gradient ................................................................................ 25
2.6.6. Effective Filter Criterion based on Vertical Hydraulic Gradient ........................... 25
2.7 Modern Filter Design Criteria ................................................................................. 26
2.7.1. Sherard and Dunnigan - Geometric.................................................................... 26
3. Experimental Methodology .......................................................................................... 27
3.1 Experimental Apparatus ........................................................................................ 27
List of Apparatus.......................................................................................................... 27
3.2 Methodology .......................................................................................................... 28
4. Results and Discussion ............................................................................................... 29
4.1 Results .................................................................................................................. 29
4.2 Limitations ............................................................................................................. 37
4.3 Improvements ........................................................................................................ 37
2
Isaac Cornish
CIVL4560

5. Conclusion ................................................................................................................... 38
6. Bibliography ................................................................................................................. 39
7. Appendix II – Experimental Results ............................................................................. 40
8. Appendix III - (Slope Stability, Geo5 Iterations) ........................................................... 46
8.1 Rotational Failure Analysis - The Simplified Bishop Method ................................... 0

3
Isaac Cornish
CIVL4560

Table of Figures
Figure 1-1: Statistics of Embankment Dam Failures (ICOLD, 2015) .................................... 7
Figure 1-2: Historical Frequencies of Failures and Accidents in Large Embankment Dams
(ICOLD, 2015) ...................................................................................................................... 7
Figure 2-1: Homogenous Embankment Dam (Fell, 2015) .................................................... 9
Figure 2-2: Zoned Embankment Dam (Fell, 2015) ............................................................. 10
Figure 2-3: Central-Core Embankment Dam (Fell, 2015) ................................................... 10
Figure 2-4: Interplay of Geometric, Hydraulic and Mechanical Conditions on Internal Erosion
Triggers (Garner and Fanin, 2010) ..................................................................................... 12
Figure 2-5: Concentrated Leak Example (Foster, 2000) ..................................................... 13
Figure 2-6: Hydraulic Fracture Example (Foster, 2000) ...................................................... 13
Figure 2-7: Suffusion Process Example (Ziems, 1969)....................................................... 14
Figure 2-8: Contact Erosion Process (Foster, 2000) .......................................................... 14
Figure 2-9: Backwards Erosion Piping Model (Van Beek et al. 2011) ................................ 15
Figure 2-10: Hydraulic Criteria of Contact Erosion (ICOLD, 2015) ..................................... 17
Figure 2-11: Witt Method for Suffusion Initiation (ICOLD, 2015)......................................... 18
Figure 2-12: Boundary Conditions Between Soil and Filter (Douglas, 1968) ...................... 21
Figure 2-13: Coefficient of Uniformity VS Permeability Criterion (Segregation Limit) ......... 22
Figure 2-14: Hazen's Relationship (Barnes, 2016) ............................................................. 23
Figure 3-1: Experimental Setup .......................................................................................... 27
Figure 4-1: Limits for Acceptable Filters (Permeability Criterion) ........................................ 30
Figure 4-2: Particle Size Distribution of Experimental Soils ................................................ 31
Figure 7-1: Supplier Soil Information (18/40 and 30/60) ..................................................... 45

4
Isaac Cornish
CIVL4560

Table of Tables
Table 2-1: Analytical Expressions for Hydraulic Shear Stress (Wan, 2006) ....................... 16
Table 2-2: Extreme Filter Particle Assemblies .................................................................... 20
Table 2-3: Sherard and Dunningan - Geometric Criterion (ICOLD, 2015 p82) ................... 26
Table 2-4: Early Filter Design Criterion ............................................................................... 26
Table 4-1: Legend .............................................................................................................. 29
Table 4-2: Terzaghi Filter Criterion Applied to Experimental Soils ...................................... 29
Table 4-3: Segregation Criterion of Experimental Soils ...................................................... 30
Table 4-4: Percent of Fines Passing at Different Limits ...................................................... 31
Table 4-5: 7C Homogenous Embankment – Experiment Number 1 ................................... 32
Table 7-1: Links to Experiment Videos ............................................................................... 40
Table 7-2: Percent Finer - Limits ........................................................................................ 43

5
Isaac Cornish
CIVL4560

1. Introduction
Embankment dams are those which are constructed of earthfill materials such as sand, clay,
and other soils. The provision of filters is also commonly achieved by using rock or coarse
material to achieve adequate seepage flow in a desired location (rockfill). This may be to
improve the stability of the dam itself, or to control the disposal of unwanted material
(water/sediment).
In recent years, inadequate engineering protocol has led to the continual deterioration of old,
large dams1. Sequentially, this leads to a failure mode criterion being triggered because of
continual dynamic changes to the dam structure such as a change in externally applied
forces, freezing and thawing of semi-plastic material and seismic activity. A failure mode is
usually defined by the design engineers through a qualitative risk assessment, some common
failure modes are shown below:
● Failure of the embankment dam to perform design goal
● Partial collapse or full collapse of the structure
● Unwanted overtopping
● Piping through or along the foundation
● Piping through the embankment
● Slope failure
These failure modes obviously have associated consequences for the responsible engineers,
the population at risk (PAR), the environment downstream and the safety of the structure.
Common consequences of embankment dam failures are shown below:
● Single or multiple fatalities
● Loss of Infrastructure
● Loss of community infrastructure
● Pollution of local water supply
● Food chain contamination
● Loss of life to fauna and flora/biodiversity/nature reserves
● Major clean-up costs/repair costs
● Fines and prosecutions
● Impact on company share price
Understanding how failure modes are triggered because of dynamic changes on the dam
structure can prevent these associated consequences from initiating and more clearly, the
risk management process above can be improved if the initiation and continuation of failure
modes is understood.
Foster analysed the statistics of a survey conducted by the International Committee of Large
Dams (ICOLD), (Foster, Fell, & Spannagle, 2000), and found that internal erosion was
responsible for about half of embankment dam failures in which the failure mode was known.
Internal erosion is the detachment of particles within a soil mass due to loss of shear strength
and leads to common failure modes. This can be seen in Figure 1-1 below:

1 Large damsrefer to dams that have the following criteria “A dam which is more than 15 metres in height” (Fell,
MacGregor, Stapledon, Bell, & Foster, 2015)
6
Isaac Cornish
CIVL4560

Statistics of Embankment Dam Failures


Seismic Instability, 2%
Static Instability, 4%

External Erosion
(overtopping), 48%
Internal
Erosion,
46%

External Erosion (overtopping) Internal Erosion Static Instability Seismic Instability

Figure 1-1: Statistics of Embankment Dam Failures (ICOLD, 2015)


Because of the mechanical process of internal erosion, it is more likely to occur between
incompatible interfaces. Figure 1-2 shows the historical frequencies of failures and
accidents in large embankment dams constructed between 1800-1986. Impervious
conduits when in contact with shearing soil from a seepage flow leads to a higher
probability of failure. Furthermore, summating the total failures in Figure 1-2, it can be
deduced that 36% occurred on the first fill and 64% after the first fill. This implies that the
first wetting and eventual saturation of an embankment is in-turn related to failure
probabilities.

Historical Frequencies of Failures and Accidents (1800 to 1986)

Seepage accidents with no detected 17%


erosion 83%

31%
Internal Erosion Accidents
69%

47%
Internal Erosion Failures
53%

0% 20% 40% 60% 80% 100%

Around Conduits or Walls In Dam

Figure 1-2: Historical Frequencies of Failures and Accidents in Large


Embankment Dams (ICOLD, 2015)
7
Isaac Cornish
CIVL4560

1.1 Objective and Aims


Research into internal erosion is of particular importance as there is a notable gap in
understanding this failure mechanism as it is not easily observable and there may be no
external evidence that it is taking place. Currently the design procedures for mitigating
internal erosion are largely empirical, and this design rationale has led to the realization that
there is a need for analytical design procedures and construction methods. Overall, this would
lead to a better understanding of the mechanics of the internal erosion process’ leading to
responsible engineers incorporating better engineering judgment, and improved safety of
embankment dams.
The aim of this project is to establish how accurate existing approaches to internal erosion
are in determining the initiation and continuation of internal erosion. Accuracy is a broad term,
and it is defined in this project as simply whether a mechanism of internal erosion occurred
because of the reasons it was expected to occur, based purely on literature. In this analysis,
fundamental soil mechanics is used to describe the experiments, and when applicable,
methods outlined in the literature review are used to specifically determine initiation criterions
for internal erosion alongside fundamental physical equations.

1.2 Project Scope


● A literature review is performed to understand the physical requirements of internal
erosion and to determine research gaps.
● Experimentation is performed on embankment dams of various zoning schematics to
analyse possible triggers.
● Analysis of laboratory results is performed, and triggers are compared to existing
models.

8
Isaac Cornish
CIVL4560

2. Background Research and Literature Review


2.1 Embankment Dam Zoning
In the book Geotechnical Engineering of Dams (Fell, MacGregor, Stapledon, Bell, & Foster,
2015) various embankment dams are classified by their zoning schematic. Zoning is simply
the arrangement of construction materials that form the cross-section of the dam. The choice
of zoning schematic is related strongly to controlling or mitigating seepage.
A homogenous type of embankment dam (Figure 2-1) is composed of a single type of
material, this material should be impervious enough to provide impoundment of water.
Notably, early embankment dams attempted to minimize seepage using highly compacted
fine material in homogenous embankment dams. A homogenous embankment type should
only be constructed in localities where available soils have little variation in permeability.
Improvements in construction practices have led to many homogenous dams being replaced
with a modified homogenous cross-section with impervious materials. This allows for
embankments to have a higher slope design.
On the other hand, zoned embankments (Figure 2-2) are constructed in areas where many
soils are readily available. By placing various soils in different zones, the soil properties
strengths can be utilized more beneficially. For example, clay the most impervious of the soils
would typically be found in the core of the cross-section and would be surrounded by
transition zones of various other less impervious soils to act as drains and filters to control
the seepage flow (U.S. Department of the Interior Bureau of Reclamation, 2014). A limiting
constrain in the design of zoned embankment dams is typically the foundations’ ability to
resist the forces exerted by the weight of the clay centre. Regarding stability and internal
erosion, a zoned embankment dam provides much greater resistance to failure than a
homogenous embankment dam.
Central core embankment dams are the most constructed embankment dams in recent times.
They provide the highest resistance to internal erosion (Foster and Fell, 2001), and are dams
which consist of an earthfill core flanked by rockfill and protected from external erosion with
coarse rockfill. This is shown in Figure 2-3.

Figure 2-1: Homogenous Embankment Dam (Fell, 2015)

9
Isaac Cornish
CIVL4560

Figure 2-2: Zoned Embankment Dam (Fell, 2015)

Figure 2-3: Central-Core Embankment Dam (Fell, 2015)


2.2 Internal Erosion
Internal erosion occurs when a body of water exerts a hydraulic force onto an embankment,
resulting in seepage. When the resulting seepage of water and the force exerted by the water
body are sufficiently high; that is when the frictional drag force is greater than the force holding
the particle grain to its parent material, a soil particle becomes mobile. The resisting forces
depend on the soil cohesion, weight of the soil particles and the particle-interlocking effect.
When the critical hydraulic load is surpassed, particles detach and are transported along a
predetermined path dictated by the void spaces of the soil. Fine grained soils will be the first
to detach in this process and their washout may be blocked by coarser material, resulting in
no volume change, only redistribution within the layer (suffusion). The zone in which these
fine-grained soils wash-out have increased porosity and permeability, whereas the zone of
accumulation (the place where these particles stop) has lower porosity and permeability. The
overall structural change within an embankment dam due to this process will concentrate
flow, increasing the hydraulic load until larger particles also reach a critical state. This
process, if allowed to continue will result in a failure mode being reached.
Internal erosion occurs entirely locally, in other words it may affect only a portion of the dam
and unlike seepage is a process that has been broken down into four critical phases.
Initiation → Continuation → Progression → Breach
1. Initiation: Which can be broken down into four mechanisms
a. Concentrated Leaks
b. Backward Erosion
c. Contact Erosion
d. Suffusion
2. Continuation: Occurs after initiation, as erosion will continue unless the eroding
forces are reduced.

10
Isaac Cornish
CIVL4560

3. Progression: Is the macroscopic structure that forms through the whole system.
Such as a sinkhole or pipe, because of internal erosion.
4. Breach: Refers to total failure of the dam, that is the structure is no longer stable and
a typical failure mechanism has been reached. E.g., the partial collapse of the
structure, and the loss of impounded material.
There are four general failure paths associated with internal erosion. If an embankment dam
fails whilst one of these conditions is true, a failure or incident is deemed true. The four
general failure modes are:
● Internal Erosion through the Embankment
● Internal Erosion along conduits and walls
● Internal Erosion through the foundation
● Internal erosion of the embankment into the foundation

2.3 Geometric, Hydraulic and Mechanical Conditions


Garner and Fannin (2010) argue that there are three conditions that interact, and that dictate
the initiation and behaviour internal erosion. These three conditions are namely, the
geometric, hydraulic, and mechanical condition.
The geometric condition is dictated primarily from the grain size and the particle size
distribution of the relevant soil. This inadvertently will dictate the potential for the soil to
experience loss of its finer fractions.
Secondly, the mechanical conditions of the soil dictate the ability to resist internal erosion. In
other words, when the effective stress in areas of the dam is lowered by applied force it will
reach a critical state. This critical state will be unable to resist triggers (initiation criterions) of
internal erosion.
Thirdly, the hydraulic loading also has a critical state. The critical hydraulic load is the amount
of energy required to invoke an initiation of internal erosion by means of seepage flow.
The mechanisms of internal erosion (Concentrated leak, Backward erosion, Contact Erosion
and Suffusion) are a factor of these conditions. Figure 2-4 shows that an unfavourable
coincidence of material susceptibility, hydraulic load, and stress conditions are needed for
internal erosion to occur:
1. Geometric →Material Susceptibility (Inability to resist internal erosion)
2. Hydraulic →Critical Hydraulic load (seepage velocities and hydraulic gradients)
3. Mechanical →Critical Stress Conditions (arching and low stresses)

11
Isaac Cornish
CIVL4560

Geometric
Conditions

Hydraulic Mechanical
Conditions Conditions

Figure 2-4: Interplay of Geometric, Hydraulic and Mechanical Conditions on


Internal Erosion Triggers (Garner and Fanin, 2010)
In relation to filter design, the geometric and hydraulic criterion must be fulfilled together to
limit internal erosion. That is, filter voids must be small enough to block fine materials, but
large enough to allow for adequate drainage of excess pore pressure. See section 2.7.

2.4 Physical Processes of Internal Erosion

2.4.1. Concentrated Leak Mechanism


Water which flows through a soils pore is referred to as seepage. To distinguish this from
water which flows through cracks (including hydraulic fracture, tension crack, differential
settlement, and cracks through bedrock discontinuities), the term “concentrated leak” was
adopted (Sherard, 1973).
Cracks can be formed via differential settlement, in openings of plastic soils (cohesive enough
they can ‘hold a roof’ below which an opening is sustained), along interfaces 2 such as
between dam fill and spillway wall and hydraulic fracture along low stress zones. Where there
is a crack or ‘opening’ through which concentrated leaks occur, the wall of the leak will be
eroded away when the shear stress of the soil is less than the applied force from the flowing
water.

2 Bedrock discontinuities that erode adjacent embankment materials, are also common initiators for openings
or crack formation.
12
Isaac Cornish
CIVL4560

Figure 2-5: Concentrated Leak Example (Foster, 2000)


In Figure 2-5: Concentrated Leak Example the crack will undergo particle detachment
because of scouring along the crack wall.

Figure 2-6: Hydraulic Fracture Example (Foster, 2000)


Historically many dams have failed by concentrated leaks for no apparent reason (Jun-Jie,
2014), in many of these cases a concentrated leak occurred abruptly at the downstream side
of dam as the reservoir was first filled. This water pressure applied to the upstream face
slowly deformed the dam via differential settlement. This occurs when the minor principal
stress (σ3 ) in the zone of cracking approaches zero, and as a result the major principal stress
(σ1 ) is less than σ3 . As a result, the stress acting on the walls of the crack change to a
compressive stress, leading to an increased crack width. Obviously, water from the reservoir
enters this crack (since the compressive stress would approach the reservoir), or the water
pressure may form a new opening. This is referred to as hydraulic fracture (Jun-Jie, 2014).

13
Isaac Cornish
CIVL4560

2.4.2. Suffusion Mechanism


Suffusion is like contact erosion, but it occurs within a single soil mass (as opposed to along
an interface). When water flows through gap-graded, non-plastic soils such as gravel,
suffusion may occur if fine particle are transported through the pore spaces of the gap-graded
material. The fine particles are eventually removed from the coarse material, leaving behind
a soil skeleton formed by coarser particles. This results in more permeability and higher
hydraulic gradients and usually leads to the initiation of other internal erosion mechanisms.

Figure 2-7: Suffusion Process Example (Ziems, 1969)


2.4.3. Contact Erosion Mechanism
Contact erosion can be thought of as erosion that occurs at the interface of a coarse and fine
soil. When flow parallel to the contact erodes the fine soil into the gravel, contact erosion is
said to be occurring. This process is essentially a concentrated leak since the interface
creates a ‘crack’. It is considered separately however, because there are methods that can
be used to assess the likelihood of contact erosion.

Figure 2-8: Contact Erosion Process (Foster, 2000)

14
Isaac Cornish
CIVL4560

2.4.4. Backward Erosion Mechanism


When a pipe progresses from the downstream side to the upstream pipe, backwards erosion
piping (BEP) is said to have occurred. If the pipe can’t sustain a roof (material above it
collapses), the global backward erosions is said to have occurred. The mechanisms remain
the same but are distinct because of the different failure each of these sub-modes would
inherently form.

Figure 2-9: Backwards Erosion Piping Model (Van Beek et al. 2011)
BEP usually initiates through heave failure at the downstream side (heaves occur when the
hydraulic gradient is equal to the void unit weight divided by the weight of water, which is
roughly equal to 1). In other words, when your hydraulic gradient is equal to one at the toe, a
heave will form. This initiates the event-tree described by Figure 2-9.
Seepage →Pipe-formation → Progressive-erosion →Instability →Breakthrough
Pipe-formation can be observed via the ‘sand-volcanoes’ that form from the displacement of
material due to the pipe-formation.

15
Isaac Cornish
CIVL4560

2.5 Initiation Criteria of Internal Erosion


This section does not include all approaches to determine initiation that are within the
literature of ‘internal erosion’. Instead, each mechanism of internal erosion has been
described by a modern and relatively well researched methodology3.

2.5.1. Concentrated Leaks


The procedure to determine whether erosion will initiate in a crack or flaw involves:
1. Estimating the hydraulic shear stress in the crack for a given applied external force.
2. Comparing the hydraulic shear calculated in (1) to the critical shear stress of the soil
τc .
This process must consider the chemical properties of the seepage flow, and the dispersion
properties of the soil in question. To date, there exists analytical expressions to estimate the
hydraulic shear stress on the surface of a cylindrical pipe and a vertical transverse crack.
Table 2-1: Analytical Expressions for Hydraulic Shear Stress (Wan, 2006)
Cylindrical Pipe g × Hf × d
τ = ρwater
4L
Vertical Transverse Crack g × Hf2 × W
τ = ρwater
2(Hf + W)L

Where:
τ = Hydraulic Shear Stress in N/m3
ρw = Density of Water in kg/m3
g = gravity = 9.81m/s
Hf = Head loss in pipe or crack due to driction in meters
L = Length of pipe or crack base in meters
d = diameter of pipe in meters
W = Width of crack in meters

3 Backwards erosion initiation criterion has been excluded since it was experimentally
impossible to occur within the conditions of this project.

16
Isaac Cornish
CIVL4560

2.5.2. Contact Erosion - Den Adel (1994) and Beguin (2011)


Contact erosion as discussed occurs at the interface of an embankment’s schematic zones.
Its initiation is a function of both a geometric and hydraulic condition. It was found through
DFilter
15
experimentation that less than 8.1, geometric filtration occurs no matter the hydraulic
dbase
85
loading4. The upper limit from Den Adel implies that for the same ratio, if the value is above
11.7, hydraulic loading controls erosion and there is no filtration effect. A value of D15 /d85
between these limits will result in a geometric and hydraulic factor determining the onset of
internal erosion for this mechanism.
Similarly, the hydraulic condition is derived from experimental results. These experimental
results are based on the assumptions of Darcy’s law. That is, the definition of hydraulic
loading is based on Darcy flow velocities and critical gradients (The author assumes these to
be horizontal and vertical gradients, see ICOLD page 59).
Experimentation by Beguin in 2011, on critical hydraulic conditions of contact erosion showed
that Ucrit had a lower limit of 0.01m/s and an upper limit of 1m/s. Results of this critical Darcy
flow vary within this range based on the PSD of the soil in question.

Figure 2-10: Hydraulic Criteria of Contact Erosion (ICOLD, 2015)


An expression that was found to fit the experimental curve most accurately is:
−1
m
Fj
dH = fine soil effective diameter = (∑ )
dj
j=1

Fj = percentage of the fraction of diameter dj in the graing curve of soil


This formula must be used with caution, since this formula is based on PSD of silts and clay
(< 75 μm), thus only valid for the soils tested by the authors. Geotechnical Engineers, may
also find it better to use d50 from Brauns (1985) for preliminary results (ICOLD, 2015, p61).

4 Filtration is one of two filter criterion that refers to the prevention of erosion of the base material.

17
Isaac Cornish
CIVL4560

2.5.3. Suffusion – Witt Method


In applying a method to determine the initiation of suffusion, the method used should be for
the soil most closely matching the soil being assessed.

Probably the easiest method to understand conceptually is the Witt method (2012). Imagine
that in a well-graded soil sequentially larger particles will trap smaller and smaller particles
until the seepage is blocked or throat size becomes impracticably small in a filter.
D 0
Then, the self-infiltration constant between the coarse fraction and the fine fraction can
Dmax D
be taken as:
coarse
D15
ISF = Iself−Infiltration = Base
D85

For any diameter D determined by the mean plus two standard deviations of a PDF derived
from the fine’s particle size distribution.

Then the criterion dictates that instability occurs when:


ISF > 9
Using this method, you can compare the slope of the grading curve to the boundary of
instability (15.7%). Soils with a slope of less than 15.7% are unstable (hence susceptible to
suffusion).

Figure 2-11: Witt Method for Suffusion Initiation (ICOLD, 2015)

18
Isaac Cornish
CIVL4560

2.6 Historic Filter Design

2.6.1. Basic Theory of Terzaghi Geometric Criterion


In an embankment dam’s zoning schematic, it is commonplace for a filter to be installed. A
filter is an earthfill or rockfill that provides sufficient permeability. This means that it can
provide filtration and drainage (commonly called the piping criterion and the permeability
criterion respectively). Filtration is a criterion that refers to the prevention of erosion of the
base material whilst drainage refers to the allowable seepage to prevent extreme lateral earth
pressures and undesirable water content. The main design criteria that a soil filter must meet
are:
Filtration/Piping (Criterion I):
The constriction size of a filter material must be small enough to prevent the base material
from passing through the filter. For uniformly graded soils the constriction or throat sizes
should be 0.1 to 0.2 times the base soil grain sizes.
Drainage/Permeability (Criterion II):
The filter should be permeable enough so that the water that is passing out of the soil can
pass freely. A general design rule is that a filter material should contain no more than 5%
fines under 60 μm and should be cohesionless.
Segregation is Avoided (Criterion III):
In a gap-graded soil, smaller particles may become separated from coarser particles, this
creates non-uniform filtering and permeation properties. The filter should not be gap-graded,
and the uniformity coefficient should be less than 3.
Derivation of Permeability Criterion:
Filter design can be a purely geometric process and requires determining the particle size
distribution (PSD) of the filter material with respect to the base. The PSD of a material will
have a variety of ‘assemblies’ that determine the average opening-size (OE ). This opening
size/throat/constriction shall be smaller than the base particle it is protecting such that
filtration and drainage are provided simultaneously.
The representative void size of the filter is controlled by the finer particle size of the filter
material, and is given by:
Filter
D15
OE =
9
Filter
Where D15 is the particle size of the filter for which 15% is finer.
There are two assemblies that are considered in filter design. The loosest assembly of
particles is deemed the simple cubic case whilst the densest arrangement is deemed the
hexagonal case.

19
Isaac Cornish
CIVL4560

Table 2-2: Extreme Filter Particle Assemblies


Particle Arrangement Diagram

Simple Cubic Case


DFilter
OE =
2.4

Hexagonal Case
DFilter
OE =
6.5

In each of these cases a limit can be derived. The three equal diameter spheres in the
densest arrangement show that if a filter material has uniform grain size, a base material
would have to have a diameter 6.5 times less than the diameter of the filter material to pass
through. Similarly, for the loosest assemblies of uniform filter particles, a base soil would have
to a diameter 2.42 times less than the diameter of the filter material. By taking the arithmetic
average of the two extreme cases a value of 4.4 is given. Terzaghi, conservatively rounded
this value to 4 and 5 in the Terzaghi model, as shown in the following equations:
Terzaghi Filtration Criteria - (Piping Criterion):
Filter
D15 ≤ 4 or 5 × DBase
85

Terzaghi Drainage Criteria- (Permeability Criterion):


Filter Base
D15 ≥ 4 or 5 × D15

20
Isaac Cornish
CIVL4560

Terzaghi Permeability Criterion (with conservative consideration):


Filter Filter
D15 D15
< 4 < Base
DBase
85 D15
Figure 2-12: Boundary Conditions Between Soil and Filter, shows that from equation 4 DB85
F
85% passing of the base material is trapped by the D15 . This prevents sequentially larger
B
particles (than D15 ) entering the filter past an artificially created screen. Equation 4 states that
the diameter of the 15% size of the filter must be 4 to 5 times as large as the diameter of the
15% size of the base.
Limitations to the criteria exist. The Terzaghi criterion does not consider the uniformity
coefficient of the soil (Cu ) or variations in the PSD. The higher the value of Cu the wider the
gradation is, leading to a soil susceptible to suffusion and segregation (A filter material, may
be neither of these).

Figure 2-12: Boundary Conditions Between Soil and Filter (Douglas, 1968)

21
Isaac Cornish
CIVL4560

2.6.2. Segregation Criterion


The U.S Army Corps of Engineers (1953) attempted to correlate the results of over 100 filter
tests to derive empirical criteria for the design of filters. This is shown in the following table.
Filter
Further to this, plots of the D15 /dBase
85 vs the uniformity coefficient (Cu ) were made. This plot
showed that the maximum allowable ratio of D15 /d85 decreases as the uniformity coefficient
D
increases. When the uniformity coefficient reached 3 the piping ratio (or stability ratio, d 15 ) is
85
not sufficient as a filter criterion. The report from 1953, concluded that an increase in the Cu
value has little to no effect on the overall filter performance (Patrick Avant Douglas, 1968).
To design for segregation (Criterion III5) the soil should be uniformly graded. To achieve this
from a PSD plot, the curve of the filter should be parallel to the base material curve.
Segregation Criteria (Criterion III):
𝐹
𝐷15 D𝐹50
𝐵 < 20 and 𝐵 < 25
𝑑15 𝑑50
Besides this criterion, the report also attempted to derive an empirical criterion for the design
of filters. One plot that was made was the uniformity coefficient versus the Terzaghi
permeability, which showed that the maximum allowable permeability ratio appears to
decrease as the coefficient of uniformity decreased. This is shown in Figure 2-13. The second
criteria stipulated was and improvement to the one above, it is:
𝐹
D15
For uniform Base Material → Cu < 1.5 and 𝐵 ≤ 6
𝑑85
𝐹
D15
For widely graded Base Material → Cu > 4 and 𝐵 ≤ 40
𝑑15

Figure 2-13: Coefficient of Uniformity VS Permeability Criterion (Segregation


Limit)

5 Based on the January 1948 Waterways Experiment Station, the U.S. Army Corps in 1948 investigated
drainage filters beneath riprap for the Enid and Grenada Dams. They stipulated two criteria for segregation and
noted that the deficiency of intermediate-size grains in a soil PSD will increase the tendency of segregation.
22
Isaac Cornish
CIVL4560

2.6.3. Basic Theory of Hydraulic Criterion


In relation to the Terzaghi criterion derived above, if the assumption that seepage water in
the base and filter are roughly equal, then Darcy’s law 6 can be used to describe the flow
(Douglas, 1968).
q = KiA
Where:
K = Permeability of the soil
h1 − h2
i = Hydraulic Gradient =
L
A = Cross/Sectional Area of Flow
The hydraulic gradient i is the difference in head between the point of entry of water and its
discharge point over the length of flow between these two points.
For the sake of explanation, if Darcy’s law is accepted as a description of seepage flow, then
in combination with Hazen’s formula, Terzaghi permeability ratio and the seepage force
2
expression, can be used to derive that the permeability is proportional to D15 . This is a
significant result as in Terzaghi filter design it indicates that the seepage force will be 16 (or
25 if 5D15 is used) times smaller per unit volume than those in the base (Douglas, 1968).

Figure 2-14: Hazen's Relationship (Barnes, 2016)


Hazen Formula = K = CD2e
Where:
K = permeability of the soil

6 If Darcy’s law is used to describe the flow of seepage water, then it must apply only to soils which are not
coarse-grained. This is because Darcy’s law is limited to flow which is laminar.
23
Isaac Cornish
CIVL4560

100cm
C=
sec
De = D15 = The effective grain size, satisfying Terzaghi (hence D15 )
Seepage Force = Fsp = i × γwater
Fsp = Force exerted by a unit volume of water
Derivation:
d15 = 4D15
K base = C(1)2
2
K filter = C × 4D15
q
i= (From Darcy)
KA
(7) → (8)
q soil = q filter
1 1
Fsp = or
K soil K filter
1 1
= 2
or
C(1) C(4)2
1 1
The result is that Fsp in the filter equal to 16 (to 25 if 5 is used) those in the soil for Terzaghi’s
permeability criterion.

2.6.4. Vertical Critical Gradient (Buoyancy)


Using the aforementioned information, the hydraulic criterion of filters could be defined by the
critical hydraulic gradient ic . As we already know, the mechanics of erosion are related to the
seepage forces, and if these forces are acting vertically then the buoyant weight of the soil
particle must be greater than the seepage force to maintain stability.
By imagining this scenario, the Terzaghi critical hydraulic gradient can be derived using the
1
fact that Fsp in the filter was roughly 16 that of the base soil. Since. Fsp is a function of the
hydraulic gradient, i and the submerged unit weight of soil is equal to i in an equilibrium
system, then i can express the stability of the soil mass subject to a seepage force:
Fsp = i × γw
γsub
icriticalTerghazi =iCTV =
vertical γwater
If iCTV exceeds one, the pressure between the soil particles is zero and piping (segregation
erosion) can occur. This derivation concerns the critical vertical hydraulic gradient. This
γ
critical gradient varies depending on particle packing. This is because γ sub is the expression
water
Gs −1
and the void ratio depends on the packing, so must the critical gradient. For lightweight
1+e

24
Isaac Cornish
CIVL4560

particles in a loose condition iCTV can be as low as 0.6, whilst for a densely packed quartz
grain a value of 1.0 could occur.

2.6.5. Horizontal Critical Gradient


Critical Horizontal gradients (iCTH ) occur along the seepage or leakage pathways within an
embankment dam. They control (or are a key mechanism) in the likelihood that internal
erosion will occur, whether by concentrated leak, BEP, or suffusion. Gravity does not act as
a resisting force in horizontal critical gradients (as it would in the vertical condition).
Above it was derived that a ICTV of one is needed for internal erosion to initiate. Past
engineering experience and dam failures have shown that iCTH as low as 0.02, can be the
initiating cause of internal erosion.

2.6.6. Effective Filter Criterion based on Vertical Hydraulic Gradient


The Terzaghi permeability criterion can be used to determine the permeability values of the
filter that is effective. Parcher and Means (1968) start with suggesting the filter should be 16
times more the permeability of base soil.
K filter C(Dfilter )2
= = 16
K base C(Dbase )2
16
Obviously, to obtain a permeability ratio of 1 , De must be equal to 4De of the filter (Terzaghi
permeability criterion). As was previously explained, the flow must equal in the application of
Darcy’s law, between soil and filter.
Hence,
q filter = q base
Vfilter = Vbase
Kifilter = kibase
By increasing the permeability of a filter, a reduction in the gradient will occur, thereby
reducing the seepage force. As K increases, i must decrease (Douglas, 1968).

25
Isaac Cornish
CIVL4560

2.7 Modern Filter Design Criteria

2.7.1. Sherard and Dunnigan - Geometric


The geometric filter criterion first developed by Terzaghi was refined by Sherard and
Filter
Dunnigan in 19897. This method splits soil into various groups based on the void size D15
and dbase
85 , the representative size of the base. This method is a considerably better method
than the Terzaghi geometric criterion, as it incorporates a range of PSD for assessing filter
criterion.
Table 2-3: Sherard and Dunningan - Geometric Criterion (ICOLD, 2015 p82)
Group Criterion
Impervious Soil Group 1 (Silt and Clay) When more than 85% of base particles are
finer than 0.075mm:
Filter
D15 ≤ 9 × DBase
85

Impervious Soil Group 2 (Sandy Silts and When 35% to 85% of base particles are finer
clays and silty and clayey sands) than 0.075mm, the allowable filter should be:
Filter
D15 ≤ 0.7mm
Impervious Soil Group 3 (sands and sandy When less than 15% of base particles are
gravels with small fines content) less than 0.075mm, then the allowable filter
should be:
Filter
D15 ≤ 4 × DBase
85

Impervious Soil Group 4 (Coarse soils When 15% to 35% of base particles are finer
between groups 2 and 3) than 0.075mm, then the allowable filter is
linearly interpolated between:
Filter
D15 = 0.7mm (Group 2)
Filter
D15 = 4 × DBase
85 (Group 3)

Table 2-4: Early Filter Design Criterion


Investigator and Date8 Criterion

Terzaghi, 1922 D15 D15


< 4 or 5 <
d85 d15
Bertram, 1939 D15 D15
< 6 and <9
d85 d15
Newton, 1940 D15 D15
< 32 and < 15
d15 d50

7 This method is determined on a 4.75 mm sieve.


8 Various soil types were used in each study, with various gradations. These gradations will influence the validity
of the results.
26
Isaac Cornish
CIVL4560

3. Experimental Methodology
3.1 Experimental Apparatus
List of Apparatus
1. 2x Bucket
2. 2-3.6 mm Sieve
3. Tools to assist in moving soil (scoop/plasterer’s hawk)
4. Plastic Enclosure 1m by 0.2m by 0.3m (LxWxH)
5. Readily Available water supply
6. 2x Plastic Tubes (20mm Diameter, 0.5m long)
7. Valves or flow control mechanism of choice
8. Protractor (15cm, 180°)
9. Ruler (30cm)
10. Drill (or similar tool to create hole in enclosure for installation of 20mm plastic tubing)
11. Abundance of desired testing soils
12. Whiteboard Markers and Eraser
13. Dye (Optional)

Figure 3-1: Experimental Setup

27
Isaac Cornish
CIVL4560

A 1m long by 0.2m wide by 0.3m high rectangular plastic box was used as the enclosure for
laboratory embankment dams. This glass would act as an impervious bed for the constructed
dams. The enclosure consisted of an upstream and downstream drainage pipe, created by
drilling two holes through the bottom-centre of the enclosure. A typical valve was installed
through each of these drilled holes. Two buckets were placed upstream and downstream of
the enclosure and plastic tubes connected to the valves were fed into both buckets via a tap
(upstream) and directly into the bucket (downstream). A 10mm Diameter plastic tube was
connected to a regular tap, leading into the bucket upstream. When the tap was engaged,
the pressure head of the bucket upstream would rise steadily. At H = 0.2cm the upstream
plastic tube was engaged to steadily apply an external force onto the embankment dam in
question. If equilibrium was reached the flow from this outlet would stop, to prevent this the
downstream valve was opened, then closed when the downstream surface water level was
as close as possible to 0.

3.2 Methodology
1. Whilst all valves were shut, the water was turned until the upstream bucket was
sufficiently filled (approx. 0.2m).
2. Using a ruler, the position of the upstream toe of the dam was found to be 20cm away
from the upstream valve, so that incoming flow did not scour the upstream face.
3. Using a protractor, the slope angle was determined, and whiteboard markers were
used to sketch an outline of a dam cross-section.
4. Using a combination of available materials, which satisfied the conditions, an
embankment dam was constructed within the outlined cross-section.
5. The valves upstream were opened, releasing water into the enclosure.
6. When the reservoir level downstream was starting to collect enough water to put the
upstream flow in static equilibrium, the downstream vale was opened to continuously
drain the enclosure.
7. The experiment was recorded using a Camera
8. The results were analysed using approaches outlines in literature to qualitatively
assess internal erosion.
9. The materials were removed from the enclosure and oven dried.
10. The excess remaining water was drained from the downstream bucket and into the
sink
11. A sieve is used to separate the available soils.
12. The apparatus is cleaned, left to dry and repeated for a series of desired dams, built
with different combinations of available soils.

28
Isaac Cornish
CIVL4560

4. Results and Discussion


4.1 Results
The various soils used in the experiment were tested to determine whether they would
provide an adequate filter. The following tables show the results of the Terzaghi criterion for
drainage and permeability. No soil combination satisfied both conditions simultaneously.
Table 4-1: Legend
Colour Interpretation
4 × d85 or d15 is within the limit
4 × d85 or d15 is not within the limit

Table 4-2: Terzaghi Filter Criterion Applied to Experimental Soils


Filter Soil Base Filtration Criterion Drainage Criterion
Soil
D15 < 4 × d85 D15 > 4 × d15
Coarse 7C
4 2.8
Gravel
Coarse 18/40
2.8 2
Gravel
Coarse 30/60
2 0.8
Gravel
7C 18/40 2.8 2
7C 30/60 2 0.8
18/40 30/60 2 0.8

The segregation criterion was tested using the coefficient of uniformity and coefficient of
curvature limits. The uniformity coefficient showed that the soils have roughly identical
particles with the expectation of the coarse gravel. This means that the segregation should
not occur as the uniformity coefficient is low and less than 3.
Also, 7C-soil had traces of gravel in it from past experiments at The University of Queensland
(UQ), which are not accounted for in the PSD. Nevertheless, the coefficient of curvature
shows that the soils are well graded, since the value is between 1 and 3, apart from the
coarse gravel.

29
Isaac Cornish
CIVL4560

Table 4-3: Segregation Criterion of Experimental Soils


Filter Soil Base Soil Segregation Criteria (U.S Army Corps,
1948)
D15 < 20 × d15 D50 < 25 × d50
Coarse Gravel 7C 11.42 12.17
Coarse Gravel 18/40 16 17.25
Coarse Gravel 30/60 40 23
7C 18/40 1.4 1.41
7C 30/60 3.5 1.88
18/40 30/60 2.5 1.33
Soil 𝐂𝐮 - Uniformity 𝐂𝐜 - Coefficient of
Coefficient Curvature
Coarse Gravel 37.5 16.6
7C – Coarse Sand 1.64 1.14
18/40 Intermediate Sand 1.34 1.11
30/60 Fine Sand 1.5 1.05

Using the design curve outlined in the literature, this can be further verified. The curve shows
𝐷
that for a plot of 𝐶𝑢 versus 𝑑15 satisfactory filters are within the red outline for steady flow
15
(laminar).

Figure 4-1: Limits for Acceptable Filters (Permeability Criterion)


30
Isaac Cornish
CIVL4560

The base materials and filter materials have roughly parallel PSD curves. The design of
segregation can also be tested using this relationship. The PSD distribution of the
experimental soils is shown below. Note, the Coarse Gravel was an interpolated PSD and
was based off 16 gravel samples. The curve is roughly linear, but smaller gravel particles
were less likely. The maximum gravel size in the sample was 18mm whilst the minimum was
6mm in grain size.

Particle Size Distribution of Soils


100
90
80
Percent Finer [%]

70
60
50 7C - Coarse Sand
18/40 - Intermediate Sand
40
30/60 - Fine Sand
30
Coarse Gravel
20
10
0
100 10 1 0.1 0.01
Grain Size [mm] - Log Scale

Figure 4-2: Particle Size Distribution of Experimental Soils


Table 4-4: Percent of Fines Passing at Different Limits
Soil D15 D85 D50 D30
Coarse 8 18 10.35 10
Gravel
7C 0.7 1 0.85 0.75
18/60 0.5 0.7 0.6 0.55
30/60 0.2 0.5 0.45 0.38

31
Isaac Cornish
CIVL4560

Qualitative Analysis of Experimental Embankment Dams9:

Homogenous Embankment Dams:


Homogenous embankment dams were susceptible to internal erosion, especially in gap-
graded soils that were not adequately compacted. Table 4-4 describes the key observations
of the 7C homogenous embankment, which was the most widely graded homogenous
embankment.

Table 4-5: 7C Homogenous Embankment – Experiment Number 1

Initial Scouring of the upstream


face due to outflow and
reduction in soil shear. The red
area indicating the loss of fines
since the beginning of the
experiment.
A filter material could have
been installed on the upstream
face to prevent the scouring of
this face. Similarly, when rapid
drawdown occurs downstream.

Downstream face when valve


remains open upon first fill.

Sliding Failure, occurring


because of sudden drawdown
after system held in
equilibrium, radically
increasing the force in the 𝑥
direction.

9 (Recordings can be found in Appendix II – Experimental Results)

32
Isaac Cornish
CIVL4560

Similar behaviour was observed in Experiment 2 as the soil properties were the same, and
the only change was the design angle of the embankment face. Due to the adhesion of the
7C material, it was able to hold a roof if compaction was poor. These small irregularities are
held open by the water pressure, albeit slowly eroding. These cracks did not progress to exit.

Cracks because
of poor
compaction

Cracks and gaps from poor compaction also generated dominant flow paths, that is, flow
paths that water preferred to take the path of. This occurred when the downstream valve was
released. The sudden release of pressure forces water to take on the easiest flow path, which
occurs along this impervious interface (the enclosure artificially creates an interface that is
impervious). In a homogenous embankment this interface is a result of two particle matrices
being vastly different10, namely plastic and soil. Furthermore, in all experiments that involved
7C soil as part of the downstream zoning schematic, concentrated leaks would form as
channels created by coarse gravel and sand fines. Under the influence of gravity and a
reduction in shear, gravel particles of 7C soil would ravel downwards towards the toe, whilst
finer particles were able to be washed downstream. This inherently created artificial gravel
filters (See the next image). It is unclear whether the flow paths created by these downstream
concentrated leaks are a result of internal concentrated leaks that cannot be observed. But
the most likely explanation is that since the flow velocity in these regions (red arrows) is
higher, then there must be an arrangement of particles within the embankment that dictated
this flow path.

10In experiments that involve more than one material, the addition of this sidewall can create an interface at a
point in space with a maximum of 4 different particle matrices
33
Isaac Cornish
CIVL4560

Zoned Embankment Dams:


Zoned embankments resisted internal erosion for much longer periods of time. Experiment 6
did not have a failure, since the 30/60 particles would not be able to travel through the 18/60
and 7C soil without having higher applied forces and longer experimental time frame. This is
a good example of the drainage criterion not being satisfied. The enclosure would build
equilibrium and external pressures would increase until overtopping occurs. Although
overtopping is not available in the video file, the homogenous embankment built of 30/60
fines demonstrates this.
Experiment 6 Filtration Drainage
Fine Sand15 Fine Sand15
Intermediate Sand85 Intermediate Sand15
= 0.28 = 0.4
∴ Okay ∴ Not Okay

Overtopping – Experiment 3 Filtration Drainage


0.4 1
∴ Okay ∴ Not Okay

Internal erosion was not observed in either of these experiments, but it implicitly would have
occurred given an extended time frame. Probably the most likely failure mechanism though,
would have been overtopping.
Experiment 4 was designed so that internal erosion could be recorded through the turbidity
of the downstream water. In doing so, instead of placing gravel upstream it was placed
downstream so that suffusion could be observed. At a time of 4:25 in this video file, Dye is
added to the downstream drainage filter at this time. Shortly after, the impervious enclosure,
alongside the gap-graded and coarse gravel create a toe drain. This creates a critical
horizontal hydraulic gradient that is roughly near the base of the embankment. This critical
flow theoretically should be normal to the filter, but it is possible that since the enclosure is
impervious the critical flow was slightly altered, and stronger along the walls.

34
Isaac Cornish
CIVL4560

Contact Erosion by critical horizontal


gradient. The pores of the coarse gravel
are sufficiently large and allow 7C soil to
be transported within its own matrix.
∴ Contact Erosion
Suffusion is not observed. This is because
there are more than enough fine particles
such that the coarse particles would
eventually (given a longer experimental
time frame) be ‘floating’ in the matrix of the
fines, due to clogging. The coarse gravel
would not form its own matrix. Therefore,
two criteria for suffusion are agreeable,
but the latter is not agreed. Clogging
∴ No suffusion

The turbidity of the water increased as the toe-drain filtered 7C soil (but was already partially
turbid because of failure to adequately drain enclosure during experiments). Some evidence
of fines passing entirely through the gravel matrix are present, but clogging was a more viable
conclusion.
Experiments in which the horizontal critical gradient are governing factors of internal erosion
could have been improved. One way of doing this would be to more frequently place dye on
the upstream face, so that the flow net lines could be used to observe critical locations of
flow.

35
Isaac Cornish
CIVL4560

Central Shell Embankment Dams:


Probably the most crucial observations of all experiments are that the fine soil (30/60) was
susceptible to all observable internal erosion mechanisms. The fine soil (called quartz by the
supplier – Appendix II) is rounded, uniform, non-angular and cohesionless. The combination
of these characteristics means that when saturating the soil, the water flows between
particles that are unable to hold themselves together (Cohesion is the property of a material
to hold itself together) simultaneously the adhesion was also significantly low (when
constructing with this wetted material, it would slip), and the fine sand was unable to attach
itself to surrounding material because of it’s rounded, uniform particle arrangement. Overall,
these properties that make up the capillary forces are inadequate by themselves to resist
very low applied forces, particularly hydraulic forces.
To further exacerbate internal erosion, not only was the finest available material used, but
predefined conduits also were installed to create more opportunity for concentrated leaks to
occur. It should also be made clear from the onset, that coarse materials were used in the
central Shell-fill experiments, but these dams are never constructed in practice. Central core
dams would have better represented the role of a filter. This affected the analysis because
uncontrolled scouring occurs upstream and downstream, affecting turbidity and hence the
observation of internal erosion through the filter is made more difficult.
Concentrated leaks occur almost instantaneously as the phreatic surface encroaches the
shell-core. Simultaneously large amounts of fine ravel downwards towards the centre of the
shell-core. This is where flow would be the highest, and flow would also rebound slightly
before pores are particularly opened by the loss of shear. This occurrence is continuous and
eventually leads to the ‘ponding’ of fines in the centre of the shell-core. If the enclosure had
not been impervious, almost certainly it would have been possible for BEP to have occurred.
The addition of the predefined pipe created a critical area in which 5 interfaces laid in
proximity, namely, the wall, the pipe, the gravel, the bottom of the enclosure, and the
upstream and downstream 30/60 soil matrices. The following shows the rapid transportation
of fines through suffusion, contact erosion and concentrated leaks within this zone.

Video Time 2:50

Video Time 2:10


Segregation between these two soils also occurred, there is no uniformity in the deformation
(transportation) of fines between coarse gravel. This is as expected since these soils are
incompatible in this criterion.

36
Isaac Cornish
CIVL4560

4.2 Limitations
1. Excess water in between experiments creates a saturated enclosure, reducing shear
strength along the interface of soil and wall.
2. Repetition of experiments in a short time frame (less than two days), would result in
inadequate drying of base materials.
3. Rarely, initial scouring of upstream toe would occur from outlet flow. This is
subsequently prevented after experiment 3 in which the embankment is moved back
20cm.
4. Since the tank sometimes had remnants from previous experiments, the turbidity
downstream was already such that it was hard to observe the onset of a concentrated
leak or internal erosion of fines through coarse filter.
5. No triaxial tests were performed on the tested soils, preventing the determination of
shear strength in an unsaturated condition. Furthermore, the intermediate stress of the
material could be predicted if a true triaxial test is performed.
6. Flow of water through most soils is laminar, however through large-pore gravels it may
be turbulent. Darcy’s law does not consider this variability.
7. The permeability of the material is largely unknown, but even so, the temperature
conditions where not recorded to consider the affect this would have on the coefficient
of permeability K.

4.3 Improvements
1. Oven drying of materials between new experiments
2. Drying of tank entirely using rag
3. A venturi meter to calculate the outflow rate could be used in conjunction with literature
criteria to assist in the determination of a horizontal critical gradient.
4. A clear table could be used to record the internal erosion at the foundation, and to see
sediment transport downwards in the vertical direction.
5. Multiple camera angles to film simultaneously all portions of the embankment dams.
6. The duration of each experiment could be extended, particularly when a stable
embankment was constructed, since internal erosion is a time-dependant event.
7. Simulation programs could be used to better understand the particle behaviour
between sand and water, and perhaps provide information on what happens internally
within a concentrated leak.

37
Isaac Cornish
CIVL4560

5. Conclusion
This project presented a study of grain-size and hydraulic criterion for soil-based filters, in
doing this, the aim was to observe the initiation of internal erosion in various embankment
dam zoning schematics. Laboratory embankment dams failed largely through translational
mechanisms related to the capillary forces of fine sands being reduced by internal erosion.
Grain-size criteria should be used with caution, as no criterion satisfied all the conditions
needed to design a satisfactory filter. Furthermore, the internal erosion mechanisms were
able to be explained by the inadequacy of the base/filter criterion not being met, as was
defined in literature. However further research is desirable, as this project has numerous
shortfalls and the literature that it reviewed were based on little data or concerned studies
with only uniformly graded materials. More research is required on non-uniform graded soils.
Regarding the hydraulic criterion, this report failed to make initial calculations that would have
been useful guides in determining the initiation internal erosion. It is highly recommended
that all filter design should aim to quantify seepage, as the horizontal hydraulic gradient
appeared to be the driving force of most suffusion or contact erosion mechanisms.

38
Isaac Cornish
CIVL4560

6. Bibliography
ANCOLD. (2012, October 1). Guidelines on the Consequence Categories for Dams. Perth,
Western Australia, Australia.
Douglas, P. A. (1968). Filter Design Criteria and Their Application. Auburn: Auburn University.
Fannin, J. (2008). Karl Terzaghi: From Theory to Practice in Geotechnical. JOURNAL OF
GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING , 267-270.
Fell, R., MacGregor, P., Stapledon, D., Bell, G., & Foster, M. (2015). Geotechnical
Engineering of Dams 2nd Edition. London: CRC Press Taylor and Francis Group.
Foster, M., Fell, R., & Spannagle, M. (2000). The statistics of embankment dam failures and.
Can. Geotech. J. Vol. 37, 2000, 1000-1023.
Garner, S., & RJ, F. (2010). Understanding internal erosion: a decade of research following
a sinkhole. International Journal on Hydropower and dams, 93-98.
ICOLD. (2015). Bulletin 164 - Volume 1 - Internal Erosion Processes and Engineering
Assessment. Internal Erosion of Existing Dams, Levees and Dikes, and their
Foundations, 0-174.
Jun-Jie, W. (2014). Hydraulic Fracture in Earth-Rock Fill Dams. Singapore: China Water and
Power Press.
Sufian, A. (2020, August 1). Dam and Embankment Engineering - Lecture 08: Internal
Erosion (University of Queensland). Brisbane, QLD, Australia.
Sufian, A. (2020, August 1). Dam and Embankment Engineering - Lecture 09: Filter Design.
Brisbane, QLD, Australia.
U.S. Department of the Interior Bureau of Reclamation. (2014). Design Standards No. 13
Embankment Dams. United States.
United States Bureau of Reclamation. (2019). Internal Erosion Risks For Embankment and
Foundations. NA, 0-178.

39
Isaac Cornish
CIVL4560

7. Appendix II – Experimental Results


Table 7-1: Links to Experiment Videos

Experiment Embankment Details Link to


No. Recording
Type Homogenous Part 1
Photo Part 2
Part 3

Material 100% Sand (7C)


Composition
Slope (Upstream and Downstream) 50° and 40°
Type Homogenous Part 1
Photo Part 2
Part 3

Material 100% Sand (7C)


Composition
Slope (Upstream and Downstream) 30°
Type Homogenous Part 1
Photo Part 2

Material 100% Fine Sand (30/60)


Composition
Slope (Upstream and Downstream) 30°

40
Isaac Cornish
CIVL4560

Type Zoned Part 1


Photo

4
Material 60% Intermediate Sand (18/40)
Composition 40% Gravel
Slope 50° upstream slope (sand)
35° measured from crest as downstream slope
(gravel)
Type Zoned Earthfill Part 1
Photo Part 2
Part 3
Part 4
Image 1

Material 80% Sand (7C)


Composition 80% Sand (30/60)
Slope Upstream and Downstream 40°
Slope of Zoned Material
135° from crest-centre
Type Multiple Zoned Earthfill FULL
Photo Part 1
Part 2

Material 25% Sand (30/60)


Composition 25% Sand (18/40)
50% Sand (7C)

41
Isaac Cornish
CIVL4560

Slope Upstream and Downstream 40°

Type Centre Shell Fill Part 1


Photo See Video. Part 2
7 Material 90% Fine Sand (30/60) Part 3
Composition 10% Coarse Gravel (16-18mm)
Slope (Upstream and Downstream) 30° and 40°
Type Centre Core Fill Part 1
Photo See Video. Part 2
8 Material 90% 7C Sand Part 3
Composition Backward
10% Coarse Gravel (16-18mm)
Erosion
Slope (Upstream and Downstream) 30°
Type Centre Shell Fill Part 1
Photo Part 2
Image 1
Image 2
Image 3

Material 90% Intermediate Sand (18/40)


Composition 10% Coarse Gravel (16-18mm)
Slope (Upstream and Downstream) 40°
Note: Homogeneous with predefined pipe Part 1
Photo See Video. Part 2
10 Material 100% Sand (7C) Image 1
Composition
Slope Upstream and Downstream 30°

42
Isaac Cornish
CIVL4560

Type Homogeneous with predefined pipe and Shell Core Part 1


Photo Part 2
Part 3
Part 4
Image 1

11

Material 90% Fine Sand (30/60)


Composition 10% Coarse Gravel
Slope Upstream and Downstream 40°

Table 7-2: Percent Finer - Limits


Orange and Blue lines Particle Size Distribution of Tested Soils

𝐷85 and 𝐷15

43
Isaac Cornish
CIVL4560

𝐷60 and 𝐷10

𝐷50 and 𝐷15

44
Isaac Cornish
CIVL4560

Figure 7-1: Supplier Soil Information (18/40 and 30/60)


45
Isaac Cornish
CIVL4560

8. Appendix III - (Slope Stability, Geo5 Iterations)


Soil Image Soil Properties (Geo5 Results)

SM

MS

G-F

GW

46
Isaac Cornish
CIVL4560

Rigid

47
Isaac Cornish
CIVL4560

Iteration Embankment Geo5 Input FoS (Against


Number Type Bishop
Analysis)
1 Homogenous - 1.07
30° Upstream
and
30°
Downstream

2 Homogenous - 0.96
30° Upstream
and
30°
Downstream

3 Homogenous - 1.23
30° Upstream
and
30°
Downstream
Isaac Cornish
CIVL4560

4 Homogenous - 0.72
40° Upstream
and
60°
Downstream

5 Homogenous - 0.64
40° Upstream
and
60°
Downstream

6 Homogenous - 0.82
40° Upstream
and
60°
Downstream

1
Isaac Cornish
CIVL4560

7 Homogenous 0.38

8 Homogenous 0.33

9 Homogenous 0.43

2
Isaac Cornish
CIVL4560

11 Central Corefill 3.27


45° slope
upstream and
45° slope
downstream
Core Slope:
60°
12 Zoned Earthfill 13.31

13 Zoned Earthfill 13.69

3
Isaac Cornish
CIVL4560

14 Central Corefill 33

15 Central Corefill 2.1

16 Central Corefill 1.68

4
Isaac Cornish
CIVL4560

17 Central Corefill 3.98

18 Zoned Earthfill 1.59

19 Zoned Earthfill 1.68

5
Isaac Cornish
CIVL4560

20 Zoned Earthfill 1.29

6
Isaac Cornish
CIVL4560

8.1 Rotational Failure Analysis - The Simplified Bishop Method


Embankment Dams are those which are constructed of natural materials, such as coarse and
fine soils. The stability of a dam’s slope is its ability to hold its design geometry. The factors
governing stability are dependent on the properties of the material and the overall geometry
of the Dam. The material properties affect the available shear strength (τf ) of the soil, such
as the cohesion and friction angle of the soil.
Commonly a Factor of Safety (FoS) method has been adopted within the industry as a
measure of a dam’s stability. The FoS is the ratio of the available shear strength and the
mobilised shear strength (τm ) of the natural materials within an embankment dam. The
mobilised shear strength will increase from various natural factors, such as wind, rain, and
snow loading. Obviously, as the saturation of a soil increases, its ability to hold its shape in
shear with the soil particle around it is decreased. Furthermore, the mobilised shear strength
is dominated by the mass of soils tendency to move downward and outward due to the
influence of gravity.
The following equation shows the FoS of any slope.
τf (7-1)
FOS = i
τ mi
Where the index i shows that several different external and internal forces are responsible
for the overall shear stress acting on a soil mass.
The Simplified Bishop’s Method can be used to analyse the stability of a slope and is widely
used in practice to give an estimation to the stability of a slope. A slice (Figure 7-1) is as an
increment across a slopes cross section that is governed by the vertical equilibrium. When
all the forces acting on a slice are considered, the following expression is derived from basic
physics. The weight component (Wi ), is a summation of the normal (N) and uplifts (U) reaction
force’s vertial component and the tangential force resolved in the vertical component (T).
Wi = Ni cos cos (αi ) + Ti sin sin (αi ) + Ui cos (αi ) (7-2)
The tangential reaction force is equal to the mobilized shear strength on the area if acts on:
τf li (7-3)
Ti =
FOS
The available shear strength of a slope is based on the Mohr-Coulomb failure criteria, so the
expression above can be expressed as:
cli + Ni tan tan (ϕ) (7-4)
Ti =
FOS
Substituting equation 2-3 into equation 2-2 and solving for N whilst assuming that the
b
porespressure U is Ui = ui li and cos cos (α) = l then;
cbi tan tan (αi ) (7-5)
Wi − ui bi −
Ni = FOS
ma
Where, b is the with.
Isaac Cornish
CIVL4560

Further algebraic manipulation leads to;

tan tan (αi ) tan tan (ϕ) (7-6)


ma = (1 + ) (αi )
FOS

The resisting and disturbing forces can be equated on the slice such that the moment
resisting is equal to the moment disturbing Mr = Md . This equated slice has the following
derivation for FOS;

cli + Ni tan (ϕ) (7-7)


F=
Wi sin (αi )
Since the expression for Ni is given in equation (7-5), a substitution presents;

{cbi + (Wi − ui bi ) tan tan ϕ }/mα (7-8)


FOS =
Wi sin (αi )

Figure 7-1: Bishop's Simplified Method

You might also like