You are on page 1of 284

7,

BEHAVIOUR OF EMBANKMENT DAMS

by

EDWARD GLEN TRUSCOTT

A Thesis submitted to the


University of London
for the degree of
Doctor of Philosophy in the Faculty of. Engineering

Department of . Civil Engineering


Imperial College of Science and Technology
London SW7 2BZ

June 1977
BEHAVIOUR OF EMBANKMENT DAMS

PART II — CRACKING OF CORES OF EMBANKMENT DAMS VOLUME 2

CONTENTS. Pare

SECTION A INTRODUCTION

A.1. GENERAL

A.2. HISTORY

A.3. METHOD 'OF PRESENTATION

SECTION B MECHANISM OF CRACK FORMATION

ILL GENERAL 11

B.2. FAILURE ENVELOPE AND BASIC FAILURE MECHANISMS 13

(a)General Failure. Envelope 13

(b)Basic. Failure Mechanisms 18

(c)Cracking Potential 19

(d)Factors Affecting Cracking Potential 20.

(e)Summary 24

R.3. TRANSVERSE CRACKING DURING CONSTRUCTION AND

PRIOR TO IMPOUNDING

(a)Stress Relationships 26

(b)Causes of Cracking

(1)General 29

(2)Differential Foundation Settlement 29


(3)Irregularities in a Rigid Foundation 3,0

(4)Arching of the Core 30


(5)Differences in Properties of Core

Material 31

(6)Construction Procedures 32

(7)Shrinkage 33
ii

Page,

Freezing of Pore Water 35


(8)
Hydraulic Fracture 35
(9)
F.4. TRANSVERSE CRACKING DURING RESERVOIR IMPOUNDING

(a)General 37
(b)Saturation Effects 38

(c)Hydraulic Fracture
(1)General 41

(2)Undrained Hydraulic Fracture 43


(3)Drained Hydraulic. Fracture 44
Propogation of Cracks 48
(4)
(5)Finite Element Analyses 49
(6)Laboratory Modelling of Hydraulic
Fracture 50

(7)Conclusions 51
B.5. TRANSVERSE CRACKING DURING RESERVOIR OPERATION

(a)General 52

(b)Cyclic Reservoir Operation 53


(c)Long Periods of Reservoir Empty 54
B.6. DEFORMATION PROPERTIES OF:CORE MATERIAL

(a)General 55
(b)Tensile Failure Strain 55
(c)Volume Change Characteristics 57
Shear Movements
(d) 59
B.7. NUMERICAL METHODS OF- CRACK PREDICTION

(a)General 60
(b)Development of Finite Element Method in
Crack Determination 61

(c) Use of Finite Element Method in Design-. 63


iii

Page
P.8. CRACK PREDICTION USING MODELS

(a) General 65
(b) Centrifugal Testing 65
(c) Modelling of Mechanisms 68
B.9. OTHER MECHANISMS CAUSING THROUGHGOING
SEEPAGE PATHS
(a)General 70
(b)
Poor Construction Techniques 70
(0) Water Passages in the Foundation 71
(d) Failure of Structures Passing Through
the Core 72
(e) Piping of the Core 72
(f) Solubility of the Core Material 73
B.10. CRACKING DUE TO EARTHQUAKES
(a) General 74
(b) Causes of Cracks
(1)Changes to the Foundation 75
(2)Changes to Embankments 78
(3)Changes in Reservoir 78
(c) Design Requirements 79
B.11. LONGITUDINAL CRACKS

(a)General 82
(b)Causes of Longitudinal Cracks
(1)Differential Settlement Between
Embankment Zones 82
(2)Differential Foundation Settlement 83
(3)
Spreading of the Foundation 83

(4)Influence of Structures Projecting


into the Embankment 84
Page
B.11. (c) Problems Resulting from Longitudinal
Cracks
(1)Causes of Transverse Cracks 84

(2)Connecting Fon—throughgoing
Water Passages 84

(3)Loss of Freeboard 85
B.12. GOOD DESIGN PRACTICE TO MINIMIZE CRACKING
(a) General 85
(b) Geometric Considerations 87
(c) Foundation Treatment 88

(d) Properties of Materials


(1)'General 90
(2)Selection of Borrow Area 93.
(3)
Placement Considerations 98

(4)Use of Admixtures 103


(5)
Properties of Shell Materials 103
(e) Construction Procedures 105

(f) Other Considerations 107


B.13. GENERAL CONCLUSIONS 107

SECTION C THE BEHAVIOUR OF CRACKS


C.1. STABILITY 135

C.2. SIZE OF CRACKS 137

C.3. BEHAVIOUR OF CRACKS ON IMPOUNDING


138
(a)Cohesionless Materials
(b)Cohesive Materials 139
(c)Homogeneous Embankments 141
(d)
Zoned Embankments 143
Pate
C.4. DISPERSIVE SOILS

(a)General 144

Causes of Dispersion 146


(b)
Tests for Dispersive Soils 148
(c)
(d)Identification of a Dispersive Soil 149
(e)Examples of Problems with Dispersive Soils 150

Protective Measures 153


(f)

SECTION D DESIGN OF DOWNSTREAM FILTERS AND DRAINS

D.1. GENERAL 162

D.2. STABILITY OF THE DOWNSTREAM SHELL OF THE DAM

(a)Strength of Materials 162


(b)Drainage Capacity of the Filter System 163

(c)High Water Pressures at the Downstream


Face of the Core 164

D.3. GRADATION OF FILTERS

(a)General 165

(b)Gradation of Base Material 166


Filters for Cohesionless Core Materials 168
(c)
Filters for Cohesive, Flocculated Soils 169
(d)
(e)Dispersive Core Materials 172
Range of Filter Gradations 175
(f)
D.4. FACTORS AFFECTING COHESIONLESS NATURE OF SOILS

(a)General 176

(b)Testing 177

Fines Content
(c) 180

Formation of Bonds
(d) 183

Particle Shape
(e) 184

(f)Conclusions 184
vi

Page
D.5. MATERIALS FOR FILTERS AND DRAINS

(a) Requirements 185


(b) Natural Deposits 186
(c) Artificial Granular Materials 188
(d) Filter Fabrics 188
D.6. GEOMETRY OF FILTERS AND DRAINS 190
D.7. OTHER POINTS
(a) Segregation 192
(b) Construction Control 193
(c) Gradation of Filters and Drains Relative
to Downstream Zones 193

SECTION E UPSTREAM TRANSITIONS 202

SECTION F MONITORING TO DETERMINE FORMATION OF CRACKS OR PIPES.


F.1. GENERAL 206
F.2. VISUAL OBSERVATIONS 207
F.3. SEEPAGE MEASUREMENTS 208
F.4. PORE PRESSURE MEASUREMENTS
(a) General 211

(b) General Piezometric Levels in the Core 211

(c) Piezometric Levels in Low Stressed Areas 213


(d) Piezometric Levels in Upstream Zones 214
F.5. DISPLACEMENT MEASUREMENTS
(a) Settlements 215
(b) Strains 216
F.6. STRESS MEASUREMENTS 217
F.7. USE OF BOREHOLES 219

F.8. USE OF TRACERS 221


vii
Pale

SECTION G REMEDIAL TREATMENT


Ga. 'DECISION ON NEED FOR REMEDIAL WORK

• (a) General 224.

(b)Decisions Before Impounding the Reservoir 225


Decisions During or Subsequent to
(c)
_ImpOunding 225

G.2. METHODS OF REPAIR


(a) General 228

'(b) Minor Repairs 229

(c) Major Repairs 230

SECTION H RISK OF CRACKING IN OLD DAMS

H.1. GENERAL 234

H.2. POSSIBLE LONG TERM PROBLEMS


Foundation Settlement 235
(a)
(b)Embankment Settlement 236
Cohesive Bonds
(c) 237

(d)Other Effects of Chemical Action 238


(e)Inadequate Maintenance 238

—H.3. INSPECTIONS
(a) Office Work 239

(b) Field Inspection 240

H.4. MINOR WORKS REQUIRED


Repair Works
(a) 241

(b)Additional Work 241

Installation of Monitoring Devices 242


(c)
viii
Page
H.5. EFFECTS OF CHANGES TO EXISTING PROJECTS
(a) General 243
(b) Raising of Dam or Reservoir. Level 243
(c) Lower Reservoir Levels 244
(d) Change of Chemistry of Reservoir Water 244
(e) Repair Works 245
H.6. EARTHQUAKE EFFECTS 245

SECTION I CONCLUSIONS
I.1. GENERAL 248
1.2. BEST DESIGN PRACTICE TO MINIMIZE CRACKING 249
1.3. METHODS OF ENSURING SAFETY OF DAM SHOULD

CRACKING OCCUR 251


1.4. AREAS REQUIRING FURTHER INVESTIGATIONS 252

APPENDIX A REFERENCES 256

APPENDIX B SUMMARY OF TENSILE TEST RESULTS 270


SECTION A

INTRODUCTION.

A.1. GENERAL

This part of the thesis discusses cracking in


earth and rockfill dams, the way in which cracking may
be minimized and design methods that should be used to
ensure that the embankment is safe even if cracking

occurs. In this discussion a crack is considered to be


an unplanned opening in the fill, however it may be caused.
A transverse crack is one which runs in an upstream-
downstream direction and therefore provides a potential
water passage through the core. A longitudinal crack is
one at right angles to a transverse crack. Shear planes
are not considered to be cracks. Openings in the
foundation have not been considered except where they
interact with cracks in the dam.

The discussion is based primarily on data


obtained from papers by others. In order to complete the
thesis on schedule the extensive data review was terminated
in June 1976. Where relevant data has been noted in
more recent articles it has, however, been incorporated
into the discussion.

A comprehensive list of dams that have cracked


and methods of repair is shown on Table A.-1. A very
thorough discussion of cracking in embankment dams was
2

written by Sherard in 1972. This article has been used


as a basis of reference for examples of cracks in dams

where possible.

A.2. HISTORY

It is not possible to include a comprehensive


history of the development of design against cracking in

earth and rockfill dams. However, a few papers that are


considered to have provided new thoughts on the matter
have been listed in this section.

Prior to the introduction of modern compaction


methods and modern construction control techniques
cracking had been listed as the cause of failure of many

dams. Some examples are:


Holmfirth Dam in 1852 (Britain) Ref. Binnie (1902)

Dale Dyke in 1864 (Britain) Ref. Wegmann(1922)


Lyman, Antlers, Jumbo, Horse Creek and Apishapa Dams
in the period 1880 to 1923 (U.S.A) Ref. Field (1923)

The danger of cracking of the cores of embankments


placed using modern methods was first brought to the notice
of the profession in 1950 by Dr. Casagrande. He
presented two case histories where problems were experienced
due to cracking of dams and sounded a warning to the
profession concerning cracking.,

Lofquist (1951 and 1955) pointed out the problems


due to arching of a thin, relatively soft dam core between
stiff rockfill shells. Reduced stresses were. measured
in one dam and in another repair was necessary because a
crack was deduced at the junction of the softer lower
part of the core and stiffer upper part of the core.

Sherard (1953) carried out an extensive review


of cracking due primarily to foundation problems and
differential settlements in earth and rockfill dams.

Terzaghi (1953) made the following statement,


"The danger of erosion in cracks across homogeneous earth
dams can be eliminated by providing the dam with a narrow
vertical core made out of filter material".

Lane (1955) described various measures taken


to minimize cracking of Garrison Dam.

Leonards and Narain (1963) obtained tensile


failure strains from beam tests in the laboratory and
compared these with strains measured on the crests of dams
that had failed. They developed a simple analysis to
check for the possibility of cracking.

One chapter of the book published by Sherard


et al (1963) deals exclusively with cracking in dams.
Earthquake was included as one cause of cracking.

Aitchison, Ingles and Wood (1964) proposed that


post construction deflocculation of clay soil caused the
failure of several Australian Dams.

Kjaernsli and Torblaa (1968) investigated the


failure of Hyttejuvet Dam and advanced the theory of cracks
being caused during reservoir impounding by hydraulic
fracture.

Covarrubias (1969) investigated the effect of


dam shape and abutment conditions on the location of
tensile zones in embankments using the finite element
method.

In 1970 Question 36 of the 10th International


Congress on. Large Dams dealt with cracking of dam cores.
Many papers were presented and much discussion was held
on this subject.

Vutsel et al (1973) described the use of


centrifugal testing of Nurek Dam to check for cracking.

A.3. METHOD OF PRESENTATION

The thesis commences by discussing how cracks


are formed and how cracking can be minimized. The first
section concludes that even with modern design techniques
it is not possible to ensure that cracking will not take
place in an earth and rockfill embankment. The next
section deals with the behaviour of a crack when the
reservoir is impounded, primarily the conditions required
to ensure that the crack does not enlarge and thus lead to
a dangerous situation.

Sections D and E discuss the purpose and methods


of design of the granular zones adjacent of the core.
These zones have to prevent erosion and.enlargement of the
crack and thUs ensure the safety of the embankment should

cracking occur.
-

Use of instrumentation and other methods of


investigating cracks or zones of potential cracking are
discussed in Section F. The next two sections discuss
methods of repairing cracks and possible cracking in old
dams. As both these subjects depend very much on
individual circumstances these sections are relatively

brief.

The discussion concludes by summarizing the


findings of the study and 'preseriting rules to (a) minimize
cracking in dam cores and (b) ensure the safety of the
dam even if cracking occurs. A list of areas where
additional knowledge is required is also included.
NAME OF DAM REFERENCE REASON FOR REPAIR CAUSE OF PROBLEM METHOD OF REPAIR

BALDERHEAD Vaughan et al (1970) Increasing, muddy seepage Filter inadequate, hydraulic Grouting and concrete wall
fracture
CHATSWORTH Sherard et al (1963) Dirty seepage Earthquake Healed itself in 30 days
CHEROKEE Sherard (1971) Washed out Crack in horn. dam; possibly Unknown
SANDY dispersive clay
COUGAR Pope (1967) Crack on crest Differential settlement Backfilled 5Ft. deep trench
CUAUHTEMOC Marsal (1960) Cracks on sides of closure - — Backfilled 10m. deep
section (no seepage) - trenches with clay .
CULLALOE Hamiliton (1973) Leakage, probably at Crack after raising old dam None'
previous crest
DIR Dayte (1968) Water losses in boreholes HydrauliC fracture Grouting
DUNCAN Gordon and Duguid Cracks on crest during Foundation settlement Clay in remaining fill and
(1970) construction preloading
EAST BRANCH Sherard (1972) Muddy seepage. Filter inadequate Grouting
EUCLIDES DA Vargas and Hsu (1970) Seepage at abutment Differential settlement Filter over seepage area
CUNHA
FLAGSTAFF Ingles et al (1968) Dirty seepage and rapid Crack in dispersive clay Rebuilt dam very densely
GULLY piping •
GIRNA Dayte (1968) Water losses in boreholes Hydraulic fracture Grouting
GRAND TETON Civil Engineering, Washed out on first filling Cracks in core at foundation Not yet decided
A.S.C.E. (1977) contact; no filter in area

TABLE A71 EXAMPLES OF TRANSVERSE CRACKING INiDAMS — SHEET 1


NAME OF DAM REFERENCE REASON FOR REPAIR CAUSE OF PROBLEM METHOD OF REPAIR.

GUMA Griffiths (1973) Sudden increase in seepage, Differential settlement Grouting and wait for clay
filter sand in seepage to swell
HARRISON Sherard (1972) Cracks on crest Differential settlement Backfilled trenches and
STREET upstream asphalt blanket
HEBGEN Sherard et al (1974) Cracking and slumping of Earthquake Not stated
crest (no seepage)
HILLS CREEK Jenkins and Increasing seepage Foundation settlement Grouting
Bankofier (1972)
HYTTEJUVET Kjaernsli et al (1968) Sudden increase in seepage Hydraulic fracture Grouting
INFIERNILLO Marsal and Arellano Cracks in crest (no seepage) Settlement of shell on Slurry grouted cracks
(1967) saturation
KING GEORGE VI Bishop (1946) Excessive seepage, cracking Prolonged low reservoir many Reduced top water level
near top of core years after construction
LAGUNA Marsal and Pohlez Excessive seepage and Piping through foundation Rebuilt using bentonite
(1972) breaching after 60 years cement slurry cut-off
LEOBARDO- Marsal (160) Cracks in crest (no seepage) Differential settlement. Not stated
REYNOSO
LLUEST WEN Little (1971) Eroaion of puddle clay core Pipe cracked due to Temporary repairs using
through 6inch.cracked pipe differential settlement clay/cement grout
MAST Engineering News Dam condemned due to poor Plastic sheet as internal
Record (1963) construction water barrier
MARTE R GOMEZ Marsal (1960) Cracks in crest (no seepage) Foundation settlement on Not stated
saturation

TABLE Ai-1 EXAMPLES OF TRANSVERSE CRACKING IN DAMS SHEET 2


NAME OF DAM REFERENCE REASON FOR REPAIR . CAUSE OF PROBLEM METHOD OF REPAIR

MATAHINA Galloway (1967) Collapse of part of crest; Downstream filter washed Hole backfilled with clay/
brief large seepage into downstream rockfill sand mixture; crack plugged
by upstream transition
MATTMARK Gilg (1970) Cracks during construction Differential settlement Backfilled trenches
pause
MIGUEL ALEMAN Marsal (1960) Cracks on abutments.(no Differential settlement Not stated
seepage)
MISSION Terzaghi and Lacroix Excessive seepage, sink holes Holes at contact of sheet- Backfilled holes and
(1964) and crack on drawdown pile wall and rock trenches, actually healed
itself
MORWELL No.2 Sherard (1972) Piping at abutment Dispersion in hom. dam Rebuilt hole; upstream
, gypsum blanket
ONO Sherard et al (1963) Severe cracks on crest Earthquake Clay/sand slurry in cracks
PANAMA CANAL Sherard et al (1974) Offsetting on fault during Earthquake Rebuilt
construction
PORTLAND Leonards and Narain Seepage drained reservoir Saturation of foundation Abandoned
(1963)
RECTOR CREEK Leonards end Narain Cracks on crest Saturation of embankment Grouting
(1963) materials
RHODESWOOD Harrison (1972) Crest subsidence and seepage Erosion of foundation Not stated
'ROUND BUTTE Patrick (1967) Cracks on crest Saturation of shell Backfilled with sand/
bentonite mixture

TABLE A-1 EXAMPLES OF TRANSVERSE CRACKING IN DAMS - SHEET 3


NAME OF DAM REFERENCE REASON FOR REPAIR CAUSE OF PROBLEM METHOD of REPAIR

SHEK PIK Carlyle (1965) Water losses in bore-holes HydraUlic fracture Grouting
SHELL OIL Leonards and Narain Large seepage and cracks Saturation of foundation . Cracks filled with slurry
(1963)
STOCKTON CREEK Sherard (1972) Dam breached Differential settlement Rebuilt
TILLIETUDLEM Plessis(1974) Piping of core and partial Spreading of poorly compact- Pipe encased in concrete;,
collapse of downstream shell ed dam; failure of steel filter provided
pipe thrOUgh dam
TORSIDE Harrison (1972) Failure of discharge pipe Spreading of dam on found. Unknown
through dab layer sofiened by saturation
UNKNOWN. Anagosti (1970) Large increase in seepage Laterite core dissolving Sheet-pile wall through core
UNKNOWN Casagrande (1950) Excessive seepage Cracks on side of closure Soil-cement grouting
section
UNKNOWN ' Clevenger (1974) Cracks on left side-of dam; Gypsum in foundation: Upstream remedial blanket
excessive leakage dissolved in seepage water
UNKNOWN Kudlik et al (1973) Excessive seepage Unsatisfactory filter of Plastic sheet on upstream
glass slag face connected to existing
blanket .
UNKNOWN Little (1971) Cracks, excessive leakage Subsidence of foundation Grouting
due to coal mining -.
UNKNOWN Lofquiat (1955) Crest settlement lower than Soft bottom of core settles, Sheet-pile wall through core
expected stiff upper core arches and
cap occurs at join

TABLE A-1 EXAMPLES OF TRANSVERSE CRACKING IN DAMS - SHEET 4


NAME OF DAM REFERENCE REASON FOR REPAIR CAUSE OF PROBLEM METHOD. OF REPAIR

U.S. CONSERV- Sherard (1972) Cracks on crest - Clay/sand slurry in cracks


ATION SERVICE
VIDDALSVATN Vestad (1976) Excessive seepage Hydraulic fracture,filter Grouting
inadequate
VIGARIO Vargas and Hsu (1970) Crack on crest Differential settlement Backfilled trench
VIRGINIA Sherard (1972) Crack on crest Differential settlement due Backfilled trench
RANCH to closure section
WISTER Bertram(1967) Muddy seepage Cracks following river in. Grouting plus sheet-piles
hom. dam
WOODBURN Hamilton (1973) Leakage at previous crest Crack after raising old dam None
WOODCREST Leonards and Narain Large crack through entire Dam left dry for many years Abandoned
(1963) dam
YARDS CREEK Sherard (1972) Seepage, muddy water Differential settlement , Grouting
filter failed

TABLE A-1 EXAMPLTIR OF TRANSVERSE CRACKING IN DAMS - SHEET 5


SECTION B

MECHANISM OF CRACK FORMATION

B.1. GENERAL

This section discusses the stresses and strains


that lead to cracking in dam cores and reviews the effects
of variables,such as fill properties and embankment geometry,

on the crack resistance of dams.

A crack is considered to be an opening in the

fill. For the purpose of this thesis a transverse crack


is defined as one which, extended as necessary, passes
through the dam core. A through going crack is one which
passes completely through the core and a non -through
going crack one which does not pass through the core but
would do so if extended. Transverse cracks may vary
from vertical to horizontal. A longitudinal crack is one
approximately parallel to the dam axis which, even if
extended, does not pass through the core.

A crack is formed when the tensile strength of

the soil is exceeded. It remains open until a compressive


stress is developed across the crack. The formation of a
crack results in a redistribution of stresses and strains
in the vicinity of the crack. The sum of the reduction
in tensile strain is equal to the opening of the crack.
This positive and significant amount of opening occurs
virtually instantaneously. Because of the stress and
12

strain redistribution around a crack a. return to the


external stress and strain state existing immediately
prior to the formation of the crack will not generally
eliminate the crack. In other words once a crack has
formed the damage cannot be undone by returning the loads
on the structure to the state they were in before the
crack occurred.

Although a crack does not form until the tensile


strength of the soil is exceeded this tensile strength is
generally very small and for quantitative purposes can
be ignored. However, the magnitude of the tensile
strength is important in determining the failure envelope.
The shape of the failure envelope determines whether
tensile failure (i.e. a crack) or shear failure (i.e.
closed failure plane) will occur.

In many cases cracks are in fact due to the


application of excessive tensile strain to the embankment
hence the tensile failure strain is important in avoiding
cracking. Similarly the stress strain properties in
tension and compression determine both the stresses and
strains developed in the structure and must be studied.

Cracks may be visible (e.g. on a dam crest) or


hidden (e.g. covered by other zones). The majority of
work carried out by others refers to visible cracks,
although in recent years the possibility and study of
hidden cracks has received more attention. The hidden
13

cracks, particularly those that occur below water level,


are the most important as there is generally no evidence
of their existence until a major problem occurs. • As they
are not readily accessible investigation of them is
difficult and expensive; Their presence has been confirmed
by visual examination in test pits and recovery of
infilling materials (generally. grout or coarse material)

by drilling. In many cases it is suspected that excessive


seepage and even some dam failures have occurred due to
such crackS, however, in most cases it has not been

posSible to verify this by explorations.

B.2 FAILURE ENVELOPE AND BASIC FAILURE MECHANISMS

Before proceeding with specific failure


mechanisms associated with specific loading conditions it
is considered desirable to study the basic failure
mechanisms and the effect of the shape of the failure
envelope on these mechanisms. This matter is fundamental
to the understanding of crack formation and yet appears

to have been ignored by most investigators. Hence it has


been necessary to develop much of the following work from

first principles.

(a) General Failure Envelope

In order to understand the mechanism of cracking

it is necessary to develop the complete failure envelope

for soils. Soils can only fail in shear or tension.


14

The Mohr-Coulomb failure envelope used to


describe the shear strength of a soil under applied
compressive stresses is well known. Three cases are
considered as follows:

(1) the effective strength envelope, which is the basic


relationship between shear strength and the effective
stress
i.e. 7f = f (a')
where j = shear stress at failure
a' = effective stress
This envelope has a small to zero cohesion intercept
(c) a d relatively large friction angle (.1 )

(2) a total strength envelope where pore pressure is


constant (e.g. a constant negative pore pressure due
to capillary action in a partially saturated soil).

= a-u where u is pore pressure and is constant

.". f = f(al) = f(a)

This envelope has the same friction angle as the


effective stress envelope but is displaced to the right
or left depending on the sign of the pore pressure.

(3) a total strength envelope where pore pressure is a


function of the applied stress (as in undrained loading)

a-u and u = f,(a)

• = f(al) = f(a)
15

This envelope has a larger c intercept and smaller


friction angle than the effective stress envelope.

The three envelopes are shown on Figure B-1. Each


envelope consists of a c intercept and a friction angle
and a generalized version will be used in subsequent

discussions.

Tensile failure conditions are represented on


b
the Mohr-Coulomb plot by a circle of radius t/2 etween
the shear axis and tensile strength (t) as shown on

Figure B-2.

A brief review has been carried out of papers


discussing tensile testing of soils and data is summarized
in Appendix B. .Beam tests are not considered realistic
models of field behaviour because of high stress. gradients
in the area of cracking (Ajaz and Parry (1975)) and due-to
doubt as to the actual tensile stress at the boundary when

the crack occurs. If beam test results are excluded the


2
maximum undrained tensile strength measured is 90 KN/m

•and values-are generally much lower.

According to the theory of effective stress the


total stress equals the effective stress plus the negative

pore pressure. However, negative pore pressures can be


2
much higher than 90 KN/m , therefore tensile strengths
2
much higher than 90 KN/m would be expected. Hence it
appears that either the effective stress theory breaks
16

down under tensile loading conditions or the mechanism is


such that average sample pore pressures and stresses do not
represent the stress state at the point where the crack is
initiated. It is tentatively suggested that small
irregularities in the surface of the test piece act as
stress concentrations and crack initiators. At such a
point the total tensile stress. is locally equal to the
effective strength plus the pore pressure, hence a crack
commences and, as, the position is then unstable, tensile
failure occurs. If this theory is correct then tensile
strength depends on the nature and magnitude of surface
irregularities. Under construction conditions
irregularities are considerably more significant than in a
laboratory specimen and even lower tensile strengths would
occur in the field. As the following discussion refers
to the general case the tensile strength has been assigned
a .value. In practice this strength will be low and
indeterminable and, cannot be relied upon to provide any
resistance to cracking of a dam core.

In order to define the complete failure envelope


it is necessary to connect the tensile and shear failure
envelopes. The two limits are shown on Figure B-2.

Either tensile failure will continue until the circle


representing the stress state touches the Mohr-Coulomb
envelope ((g) on Figure B-2) or shear failure occurs
immediately the major principle compressive stress rises
3
above zero ((p) on Figure B-2). Bishop and Garga (1969)
17

carried out triaxial tensile tests (drained) on saturated


London clay and showed that tensile failure occurred with

lateral pressures above zero. Hence the lower line (4)


is not correct. Their results indicated that the upper line

(3) should be used. This is the worst interpretation


from the point of view of cracking (i.e. cracking occurs
instead of shear failure) and has arbitrarily been
selected as the boundary for the purpose of discussions.

In this report stress paths used in investigating


cracking will be drawn on the plot developed by Lambe

(1967). The failure envelope has to be modified somewhat


to correctly represent failure conditions on this plot.
The manner in which this is done is shown on Figure B-3.
The tensile failure envelope is represented by a line at
o
45 to the average stress axis which passes through the
tensile strength.

The failure envelope discussed above defines

tensile failure in terms of stresses. Leonards and


Narain (1963) and some subsequent investigators have used
tensile failure strain to define conditions for cracking on
dam crests. The stress state at the dam crest is
relatively simple and the tensile strain will be primarily
due to the tensile stress. However, at a point remote
from the crest tensile strain will also result from
compressive stresses at right angles to the tensile plane
and, if the shear strength is exceeded, by shear movements.
18

Thus tensile strain can occur when the stress on the plane
is compressive and there is no risk of cracking. Hence,
although the magnitude of the tensile failure strain is
important in assessing the probability of cracking (see
Section B-6) the use of tensile stresses to define the
failure conditions is preferred.

(b) Basic Failure Mechanisms

There are three possible stress paths leading to


tensile failure (i.e. cracking). These are shown as.
paths ACD, AB and AEF on Figure B-4. Initially the soil
is at stress state A.

In path ACD the stresses move to the Mohr Coulomb


failure envelope (AC), shearing occurs with a reduction in
average principle stress to point D where cracking takes place.
It is evident that if the minimum value of the major
principle stress is greater than twice the shear strength
at D cracking cannot occur.

In path AB the stress path moves directly to the


tensile failure line at point B and cracking occurs.

In path AEF the principle stress difference


reduces to zero (AE) and cracking occurs by reducing the
average principle stress to the tensile strength of the
soil (at point F). Path AEF is unlikely to occur in
practice but is theoretically possible.
19

(c) Cracking Potential.

Tensile failure results in a crack and a possible


water passage through the core. Local shear failure
results in an area of disturbed material that is still

relatively impervious. Thus local shear failure


(provided it is not part of a through going shear failure
surface) does not endanger the structure but a tensile

failure may do so. In additictn shear failure introduces


stress redistribution which must increase low stresses
and reduce high ones. Thus the low stress condition that

leads to cracking is reduced. Therefore, if shearing


occurs before tensile failure the probability of cracking
is reduced and the greater the amount of shearing the less

likely cracks are to form.

The boundary between shear and tensile failure •


is represented by point "D" on Figure B-4. This point is
defined by an average principle stress and average principle

stress difference. As either of these increases the


probability of shear failure reduces. If cracking
potential is defined as the probability of a crack occurring
then, as a general statement, it may be said that the
cracking potential will increase as the point D moves away
from the origin in a positive direction.

The cracking potential depends on the tensile


strength, the cohesion intercept and the angle of internal
friction. The shear strength at a given average principle
20

stress depends on the cohesion intercept and friction


angle. Hence the shear strength may either be due to a

high cohesion (c) and small $ or small cohesion and high cp.
These two alternatives are shown in Figure B-5 and it is
apparent that the more cohesionless the soil the closer D
is to the origin and the lower the cracking potential, until
in the limit,a completely cohesionless soil cannot crack.
It should be noted that a so called "cohesionless" soil is
only cohesionless when there is no negative pore pressure
present. Cracking of such a soil can occur if negative
pore pressures exist.

(d) Factors Affecting. Cracking Potential

Discussion under this heading will be based


primarily on theoretical considerations confirmed by
experimental data. Tensile test results are summarized in

Appendix B and the source of shear strength data is given


where it appears. In some cases only unconfined compression

results are available and in such cases (1) has been assumed
equal to zero. This will lead to only small errors.

As mentioned above laboratory tensile strengths


are low and field strengths may be even lower. Hence
variation in tensile strength only has a small and doubtful
effect on cracking potential. In general factors which

cause the cohesion to reduce also cause the tensile strength


to reduce but by a much smaller amount. (refer Figures B-6
to B-12). Hence the location of point D (i.e. cracking
21

potential) is mainly dependent. on - the shear strength of


the soil (primarily the cohesion intercept).

The three basic shear strength envelopes are

shown on Figure B-1. The effective stress envelope has


the smallest cohesion intercept and:hence, should have the
lowest cracking potential. Comparison of undrained and
effective cracking potential are provided from data for
the Toulnustouc Landslide clay (Conlon (1966)) and for Mica

Core Material on Figures B-6 and B-7. Data for


Balderhead Clay from Kennard et al (1967) shows the
effective cohesion equal to zero, that is cracking is not
possible under effective stress conditions, however, the
clay is similar to that used at Selset Reservoir where
one test gave a small effective cohesion intercept

(Bishop and Vaughan (1962)). The location of point "D"


for this case is shown on Figure B-8. it is also of
interest to note that unpublished attempts to measure the
effective tensile strength of remoulded London Clay at
Imperial College using the method described by Bishop and
Garga (1969) have not been able to obtain tensile failure
because shear failure occurs first even under very small

confining pressures. In all cases point. D is much


closer to the origin for the effective stress case, hence
the cracking potential is much lower and for some cases

reduces almost to zero.

The magnitude of the negative pore pressure


22

effect is determined by the rate of loading and the water

content. Several investigators have presented the results


of undrained tests on soils at different water contents.
Data from two papers has been used to show the effect
of increasing water content on cracking potential in
Figures B-8 and B-9. The cracking potential of all soils
reduces as the water content increases that is as the
negative pore pressures reduce.

Shear strength parameters normally quoted are


determined from tests in which failure occurs under
increasing average principle stresses. Cracking occurs
under reducing average principle stress conditions. That
is, provided some consolidation has taken place, the soil
in the core is overconsolidated relative to the stresses
at which cracking will occur. The undrained strength
envelope of an overconsolidated, cohesive soil has a
higher cohesion intercept than the normally consolidated
case. Hence, the undrained cracking potential of the core
material will be higher than indicated by standard
laboratory tests. It is well known that the effective
cohesion intercept (c') of a geologically consolidated
clay is higher than that of a normally consolidated one.
This increase is due to a combination of the following
effects:
(1) Overconsolidation pressure - latest data (Bishop
(1966) and Wesely (1975)) indicates that c' is not
significantly affected by primary consolidation under
higher pressures.
23

(2) Secondary consolidation or time effects - it is


obvious that geological time effects are much more
significant than the construction period of a dam.
However, significant secondary consolidation may occur
in this period and increase c'. No laboratory data

has been found on this matter.

(3) Chemical effects or cementation - this matter is


discussed in detail in section B-12(d). There is
evidence, however, that with some core materials
chemical bonding can lead to a significant increase

in c'.

The increase in c' leads to increased cracking potential.


Data from Bishop and Garga (1969) is plotted on Figure
B-10 and shows how cracking potential is reduced when
London clay is remoulded, that is bonds due to over
.consolidation are broken.

Data leading to an assessment of the effect of the


plasticity of a soil on its cracking potential is

difficult to obtain. Under effective stress conditions


the cracking potential is small and plasticity has little

effect. In partially saturated soils or soils loaded in


the undrained condition the problem is to select a water
content at which strengths can be compared. Common
practice is to select the liquidity index as the base.
Tschebotarioff et al(1953) carried out tensile and
unconfined compression tests on clays of differing plasticity
24

indices at the same liquidity index. Results (see


Figure B-11) are not as consistent as previous data but do
indicate an increase in cracking potential with reducing
plasticity. Thus is consistent with other investigators
who show an increase in undrained compressive strength
with increasing plasticity under the same conditions
(Skempton and Northey (1953) and Youssef et al (1965)).
In cores of earth dams the parameter usually used to control
fill water content is the Proctor, Modified Proctor
Optimum Moisture Content or some variation of this. Data
at 0.M.C. has been taken from Narain and Rawat (1970) and
is plotted on Figure B-12. There is no apparent relation-
ship between plasticity and cracking potential revealed in
this plot.

The effect of compaction effort on cracking


potential would be of interest but no direct data has
been found. As an increase in compaction energy tends
to increase the shear strength it is probable that it
would increase cracking potential.

(e) Summary

(1) In reviewing cracking of soils it is


essential to consider the complete failure envelope of the
soil.
(2) There is insufficient experimental or
theoretical data to accurately define the failure envelope
of soils in tension. The limited.data available suggests
that the tensile strength remains constant under increasing
average principle stress until the Mohr's circle becomes
tangent to the shear failure envelope.

(3) If the tensile failure envelope is defined


as described in (2) above the probability of tensile
failure occurring (defined as cracking potential) increases
as the point where the shear and tensile failure envelopes
meet moves away from the origin in a positive direction.

(4) Laboratory tensile strengths are generally


small compared with the apparent cohesion. It is also
probable that the tensile strength depends on surface
irregularities. Hence at the upstream face of the core
the tensile strength will be even lower and probably
negligible. Therefore cracking potential depends
primarily on the shear strength characteristics of the soil.

(5) The cracking potential reduces as the


apparent cohesion intercept reduces until at zero cohesion
there is no possibility of cracking.

(6) In dam cores all soils are remoulded,therefore


the effective cohesion intercept will be small and
effective cracking potential small.

(7) Negative pore pressures due either to capillary


effects or undrained loading conditions - significantly
increase the cracking potential.

(8) As factors that tend to increase the tensile


strength of the soil also tend to increase the cohesion
26

intercept efforts to increase the tensile strength of the


soil frequently increase the probability of cracking.

(9) Embankment core materials are usually

overconsolidated relative to the stresses at which cracking


occurs. Hence the cohesion intercept, and thus cracking

potential, will be larger than determined from standard


laboratory tests. The magnitude of this effect is not
known.

(10)Cracking potential reduces with increasing.


water content.

(11)Cracking potential is increased by


cementation bonds.

(12)The effect of plasticity index and energy

of compaction on cracking potential is not clear.

B.3 TRANVERSE CRACKING DURING CONSTRUCTION AND

PRIOR TO IMPOUNDING.

(a) Stress Relationships.

During this period stresses are due to the dead


load of the structure, the dissipation of pore pressures
and drying of the soil due to evaporation. Hence the
rate of loading depends on the rate of construction, the
permeability of the soil, the geometry of the structure
and foundation and the compressibility of the various
materials making up the structure and foundation. Unless
27

special provisions have been made in zones of potential


cracking, generally by incorporating wetter material in
such zones, negative pore pressures will be associated with
the cracking and failure will occur under conditions between
the drained and undrained state, depending on the factors

mentioned above.

The rate of construction determines the extent

to which excess pore pressures are relieved during


construction. Minimum cracking will result if all excess
pore pressures are relieved as construction proceeds,
provided the core material is at a sufficiently high water

content. If this occurs most of the settlement will take


place as the fill is raised thus causing minimum tensile
strains. Also undrained negative pore pressures in areas
of potential cracking will be relieved thus allowing
shearing and preventing cracking.

Pauses in construction can lead to cracking in


areas where there would otherwise not be a crack. This is
undesirable. Figure B-l3 shows the stress changes that
occur if a crack is allowed to form and then close up in

such an area. The drained (effective strength) envelope


is shown,, however, the undrained stress paths lead to

similar results.

Assume a simple stress path OA due to construction

as shown. If there is no pause in construction it will


proceed unaltered as shown. When construction stops
28

temporarily the stress state will move towards the tensile


failure condition along line ABC. If the pause is small
no cracking will occur and an increase in average stress
difference will result as shown in Figure B-13. However,
if the tensile loading condition continues the stress path
will eventually intersect the tensile failure line and
cracking will result. Immediately the minor principle
stress becomes zero, a positive displacement occurs at the
crack, and the, stress path moves to D (Figure B-13). When
construction recommences the major principle stress will
increase but the minor principle stress will remain at
zero until the crack closes. This can occur either due
to lateral expansion under increased major principle stress
(at point F) or due to shearing (point G to H). Once
the crack is closed the stress path will move away from
the failure line as shown. If the crack is repaired a
small minor principle stress will be induced (point E).
When construction recommences the stress path will move
away from the failure line. These plots show that in the
case of no crack or a repaired crack there is little
change to the initial stress path, but unrepaired cracks
result in a large increase in principle stress difference,
hence cracking can occur at previous cracks even though
the major principle stress is reasonably high.
29

(b) Causes Of Cracking

(1) Generall,.

Although the causes of embankment cracks are


dealt with separately in this section cracks will generally
be due to a combination of causes. This is particularly
so with variation of fill properties as cracks due to other
causes will always occur in the weakest fill.

Hidden cracks undoubtedly occur during construction,


however, they are not usually detected until the reservoir
is impounded. Hence it is not possible to distinguish
between these cracks and those caused by the impounding.

(2) Differential Foundation Settlement:

Differential foundation settlement can occur with


an alluvial foundation due to variation in alluvium depth,
variation in alluvium properties or due to large diff6rences
in settlement betWeen the alluvium and rock abutments.
Examples are shown on Figure B-14.

A typical example is Duncan Dam (Gordon and


Duguid (1970)). This dam was built on alluvium up to 380
metres deep and a diSh shaped settlement pattern with
maximum settlements from 3 to4.3 metres was predicted.
Cracking was anticipated and the embankment designed
accordingly. Actual maximum settlements were of the
estimated magnitude but occurred very close to the east
abutment, resulting in large differential settlements
relative to the abutment rock. Cracking occurred in
this area and test shafts at the two largest cracks
showed that they varied in width from 2.5 to 8 cms and
were 11 and 12 meters deep.

(3) Irregularities in a Rigid Foundation

Irregularities beneath a dam core can be due


to rapid variation in foundation, level, local projections
on abutments and concrete or other structures that pass
through the dam. Such irregularities lead to stress
concentrations that lead to low stressed areas and possible

cracking. Examples are shown on Figure B-14.

Sherard (1972) describes several examples where


a projection,such as a left in place construction road,has
caused cracking. Examples are Stockton Creek, Cougar,
Round Butte and Matahina Dams. It should be noted
that such a crack may be hidden and at Matahina Dam it was
only discovered when water flowed through it on impounding.

(4) Arching of the Core

The core will generally be of a much more


compressible material than either the abutments or the

shells. Hence at the junction of the abutments and core


and shells and core where the loading on the stiff material
and the core are the same the core would normally settle
considerably more than the stiff material. This cannot
occur unless shear failure occurs at the junction. Hence
31

the vertical stresses in the core are reduced and those in


the shells or abutments are increased until the difference
reaches the shear strength of the weaker material. This
reduction in stress in the core is known as arching and
should the stress path be such that the tensile failure
envelope is intersected prior to the shear failure envelope
cracking will occur. Examples are shown in Figure B-14.
As this is a hidden crack no examples have been found of
cracks being caused in this way.

(5) Differences in Properties of the Core Material.

Such differences may be local due to local changes


in borrow area properties or placement procedures or may
be major due to a change in core material. Cracks due to
local changes in material properties are impossible to
identify,but in theory a stiff area in the fill attracts
stresses and a soft spot rejects them. Hence the stress in
a soft pocket between two stiff zones could have stresses
reduced below zero - that is arching within the core. An
example of the second possibility (i.e. major changes in
material properties) is given by Lofquist (1955). He
describes a dam where the bottom part of the core was of
relatively soft clay and the upper part of much stiffer
material. A crest-settlement of 2 feet had been'estimated
but only 2 inches actually occurred. It was decided
that the bottom part had settled but that the upper part
had arched leaving a crack between the two parts. A sheet
32

pile wall was driven through. the core to ensure water


tightness.

(6) Construction Procedures:

Careful attention to construction procedures


can reduce the possibility of cracking as follows:-
(i) Some effects of rate of construction and pauses in
construction have been discussed in part (a).

(ii)where cracks are due to differential settlement effects


strategic pauses in construction can allow settlements of
one area to proceed before additional fill is placed.
For example where the foundation consists of two
relatively flat areas at different levels with a near
vertical face in between a pause in construction at
the level of the higher area will reduce differential
settlement across the vertical face. Such pauses in
construction must be carefully planned to avoid
cracking at intermediate fill height or the requirement
to repair such cracks included in the planning.

(iii)differential settlements can be reduced by preloading


areas in which maximum settlements are expected.

(iv)trenches required, for example for instrument


installations, should be properly backfilled.

(v) in many earth embankments one part is raised separately


from the rest. In such a case the closure section
should be placed so as to minimize cracking. That
33

is it should be built as slowly as practical and be well

compacted. Casagrande (1950) gives an example where


large flows of water issued through cracks in the closure
section of an 80 feet high embankment. These cracks were

not observed on the crest of the dam. Sherard (1972)


refers to cracking on the crest of Virginia Creek Dam
at the closure section between the concrete spillway and
the dam. He noted that placement of the closure section
was very well controlled with detailed attention paid to
moisture content and density control. Compaction of the
closure section was by means of hand compactors and he
suggested that in order to achiev_e proper compaction use
of heavy construction compactors is required and closure
sections should be made large enough to accomodate these.

(7) Shrinkage

Shrinkage cracks occur on the crest and faces


of dams where the surface is of a fine grained material
and is subjected to long periods of hot dry weather.
Hidden cracks can also occur in the clay cores of dams
where the cores are placed wet and then allowed to dry
out because of lack of storage over large periods.

This problem has been discussed by several

investigators. Leonards and Narain (1963) refer to a


large crack passing completely through Woodcrest Dam after
it had been left dry for several years. Sherard (1972)
pointed out that shrinkage due to drying is a real problem
34

with flood control dams in dry climates. Mackellar (1970)


noted that in South Africa throughgoing cracks could
extend below water level and provide a water passage when

the reservoir, was filled. He recommended a layer of


granular material at least 0.5 metres thick over such dams

to prevent cracking.

Walbancke (1975) measured pore pressures near

the surface of several dams in Britain and found that a


layer of gravel significantly reduced negative pore pressures
due to dry weather and thus reduced the probability of

cracking. The effect of the gravel was to stop capillary


rise of water from the fine grained material and thus
reduce transpiration and evaporation losses. The gravel
also tended to retain some water thus reducing minimum
pore pressures. On the other hand she found that
vegetation cover increased negative pore pressures due to
water losses by transpiration and thus increased the

possibility of cracking.

Bishop (1946) described a case of seepage through


cracks in a puddle clay core due to drying of the core when
the reservoir level was lowered during the war for safety
reasons. Vegetation was considered to have contributed
to these cracks by lowering, the pore pressures. There was
concern that similar cracks could form in the cores of
dams during the 1976 drought. Appleton (1976) described
plans for ponding water on the crest of Alton Dam to
35

prevent cracking due to excessive drying when it was found


that water would not be impounded behind the dam as early

as expected.

The clay in a puddle clay core has to be placed


at a high water content to enable the puddling action.
Bishop (1946) pointed out that a reduction in water content
leads to permanent shrinkage. Hence the cracks in a
puddle clay core will not heal when the reservoir level is
raised and the core becomes saturated again.

Where the core material is compacted at about


optimum moisture content the cracks will heal by swelling
provided the reservoir is filled slowly enough. However,
an unfavourably stressed area will be left as discussed

in part (a). In the case of drought or sudden floods,


filling tends to be too rapid for significant swelling to
take place. Hence a through-going crack will not heal and
a non-throughgoing crack may act as an initiator for
hydraulic fracture as discussed in Section B-4(c).

(8) Freezing of Pore Water:

Wegmann (1922) notes that in some hydraulic fill


dams in U.S.A. the pore water in the fill would freeze in
cold weather. When this water thawed the embankment
settled and cracks resulted.

(9) Hydraulic Fracture:

Hydraulic fracture is the tensile failure of the


36

soil due to applied external fluid pressure. Discussion


of mechanisms is presented in Section B-4(c). During
construction fluid pressure may be applied due to
construction procedures such as drilling through the
embankment or grout leaking from operations in the

abutment. In such cases the pressure is applied rapidly


at one location in the fill hence there is no time for
seepage to occur and undrained failure probably occurs.
Undrained hydraulic fracture will occur when the applied
fluid pressure exceeds the minor principle stress plus the

tensile strength. Hydraulic fracture under these conditions


has been studied by Bjerrum et al (1972), Vaughan (1970)
and Bjerrum and Anderson (1972).

In many dams it has been found that water or mud


losses occur when holes are drilled into the core. Water
or mud continues to flow into the hole so long as the fluid
level in the hole is maintained. However, when the
level is allowed to fall it stabilizes at some lower level
and flow reduces to almost zero. That is the crack
closes at the lower pressure and the permeability of the
fill returns to its previous low value. Hydraulic fracture
of this nature was observed at Djatiluhur and Shek Pik

Dams (Sherard (1972)). In many cases large flows occur


into the cracks hence they must extend to either the
upstream or downstream face of the core.

Hydraulic fracture occurred at Djatiluhur Dam


37

by grout forcing itself into the fill during grouting


operations on the rock abutments (Sherard (1972)).
Inspection of the cracks showed that the grout penetrated
up to 40 metres into the fill. Cracks were not completely
filled with grout and a small unfilled crack was left at

the end. Hence grouting caused an open crack in the core.


It also provided layers of cemented material through the
core which would act as a roof over any crack that could
form in the dam core due to other causes and so prevent
any self healing action of the core.

Vaughan (1970) and Bjerrum and Anderson (1972)


discuss the use of hydraulic fracture techniques to measure
the minor principle stress under field conditions. As
these tests are closely controlled they are unlikely to
cause a through going crack in the core. However, they
do introduce a "precracked" area into the core as discussed
at the beginning of the section and tests carried out by
Vaughan (1970) at Cow Green Embankment (See Table B-1)
showed that the crack was reopened by considerably less
pressure (up t 14 metres of water) than was required to
initiate it.

B.4 TRANSVERSE CRACKING DURING RESERVOIR IMPOUNDING

(a) General:

Cracks which occur during reservoir impounding


are due to a continuation of processes discussed in
Section B-3, settlements due to increased load on the dam
38

or foundation, volume changes in dam or foundation due to

saturation, or hydraulic fracture. In many cases a


sudden increase in seepage through the dam is the first
sign of a crack through the impervious zone. This increase
in seepage may be due to the water level in the reservoir
reaching an existing open crack or a reservoir induced crack

opening up at a lower level. If it is an existing crack


a slight lowering of the reservoir will reduce seepage, if
the crack is caused by hydraulic fracture then it will close
when reservoir pressure is reduced below the minimum stress
at the crack and if the crack is due to other causes
seepage will reduce when the reservoir level falls below

the level of the crack.

Cracking due to a continuation of processes


causing cracking during construction and that due to
increased loads on the dam and foundation are fully discussed
in Section 3 and no further discussion is carried out in

this section.

(b) Saturation Effects:

Saturation of many soils and most rockfills


bring about a reduction in volume. In an embankment dam
saturation can affect the foundation, upstream shell and
the core. A detailed study of the effects of saturation
on the foundation and upstream shell has been carried out
by Nobari and Duncan (1972). Details of these cracks are
shown in Figure B-15.
39

Saturation of the foundation leads to increased


settlements and cracking associated with this as discussed

in Section 3. Where the foundation is semi-pervious


saturation of the foundation proceeds from upstream to
downstream over a significant time period. Hence settle-
ment of the foundation occurs first at the upstream toe of
the dam and proceeds downstream. At the early stages
of saturation settlement has occurred upstream of the dam
axis while none has occurred further downstream. Hence
the upstream face of the dam core is tilted in an upstream
direction initially and returns to the vertical when
saturation has proceeded far enough downstream. The upstream
movement leads to tensile strains at the upstream toe of
the core and cracking could occur in this area. Also the
upstream tilting of the core leads to tensile strains at
the upstream face of the core, particularly at the crest
and cracking is therefore possible. Saturation of the
foundation caused cracks in the Marte R. Gomez Dam.
(Marsal (1960)) and the Portland and Shell Oil Dams (Leonards

and Narain (1963)).

Saturation of the upstream shell of the dam and


the resulting reduction in volume causes transverse cracks
in two ways. As the shell settles it moves in an upstream
direction and the top of the core is bent upstream under
the loading due to the downstream shell. Tensile strains
and cracks can occur on the downstream face of the core
40

near. the crest. Cracking of this type was observed in


test pits at Djatiluhur Dam (Sherard (1972)). Settlement
of the upstream shell also transfers load from the shell
to the core, especially where arching of the core has

occurred. This leads to additional settlements of the core


and increases the possibility of cracking due to differential

settlement effects. Transverse cracks were observed at


El Infiernillo Dam (Marsal and Ramirez (1967)) and Round Butte
Dam (Patrick (1967)) due to saturation of the upstream shell
The core of a dam constructed with proper

controls on moisture content and compaction should not

reduce in volume on saturation. In addition saturation


of the core will take a reasonably long time allowing stress
and strain redistribution to occur thus reducing the
probability of a crack occurring. Also cracking of
the saturated core will take place under effective stress
conditions, the cracking potential will be small and
shear failure will reduce the probability of open cracks..
It is therefore probable that if there is a volume reduction
on saturation of the core it will lead to a crack developing

in an area that is not saturated.

Where saturation of the core leads to volume


reduction the following causes of cracking pertain:

(1) Additional settlement of the core will occur leading


to increased possibility of cracking.

(2) Once settlement of the shells is complete the


41

additional settlement of the core will lead to


additional arching between the shells (or abutments)
with associated problems.

(3) As saturation of the core proceeds the, saturated


part will settle and the stiffer unsaturated part
may not, leading to a crack just above the phreatic'
surface. Should the reservoir be left at one level
for sufficient time to develop such an opening through
the core and the level then raised rapidly such a crack
may not have time to heal. A water passage through
the core will result.

(c) Hydraulic Fracture:

(1) General

Hydraulic fracture in boreholes (particularly


in rock) has long been recognized as a crack forming
mechanism (Haimson (1968)). The concept of water pressure
causing a crack in the upstream face of a dam core appears
to be a relatively new idea, first formulated by Kjaernsli
and Torblaa (1968).

Hydraulic fracture is said to have occurred when


a crack is formed in the core because the reservoir pressure
exceeds the stress on a transverse plane in the core plus
the tensile strength of the soil. Hydraulic fracture will
only occur in areas where the initial stresses are low.
These are the areas where cracks due to other causes are also
most likely to occur. Thus measures to reduce' cracking .
in general will also reduce the possibility of hydraulic

fracture.

When .a crack-is formed , by hydraulic fracture


there will be a sudden increase in.seepage when the
reservoir reaches the critical level. Hydraulic fracture
is identified when investigations reveal that the crack
is below the critical reservoir level.

Cracks due to hydraulic fracture are only detected


when the seepage through the dam shows a marked increase.
The'seepage through an uneroded crack will usually be
small and, provided the dam is properly designed, the crack
should heal rapidly and seepage reduce to a negligible

value. Hence cracks due to hydraulic fracture will not be


detected unless significant erosion occurs. As discussed
in detail in Sections C .and.D erosion of the crack will
only occur if the core material is erodible and there is a
breakdown of filter action of the downstream filter. Thus
before cracks due to hydraulic fracture are significant
enough to be reported in the literature the following factors

are necessary:

(1) Stresses in the core must be low enough to permit


hydraulic fracture.

(2) The core material must erode when water flows


through the crack
(3) The downstream filter must break down.
43

In particular detectable cracks are not only dependent on


low stresses. In a core where large areas are at low
stresses cracks may only be detected at a few locations.
These locations will be determined almost entirely by filter

performance. Hence, provided initial stresses are low


enough for hydraulic fracture, there is no, reason to expect
any correlation between crack location and the lowest stressed
areas. For example it is considered that cracks were
detected at only two or three locations at Viddalsvatn Dam
(Vestad (1976)) because these were the only locations where
the filter failed not because there were only two or three
locations where stresses were low enough to permit hydraulic

fracture.

(2) Undrained Hydraulic Fracture:

Kjaernsli and Torblaa (1968) postulated undrained


hydraulic fracture due the presence of irregularities in
the upstream face of the core. Cracks could also be
initiated by water penetrating into existing cracks or a
layer of more pervious core material. As the undrained
cracking potential is large hydraulic fracture is quite
likely, provided undrained conditions exist.

Undrained cracking can only occur if the seepage


or swelling face has only penetrated a small distance into
the core at the time the reservoir pressure reaches the
value required for hydraulic fracture. Thus the reservoir
level must rise very rapidly relative to the rate of advance
44

of the seepage or swelling front. This usually does not


occur in an embankment under normal operating conditions.
Hence the only examples where undrained hydraulic fracture
is considered a serious possibilityare in boreholes (as
discussed in section B-3(b)) and the small, rapidly filled
flood control dams described by Sherard et al (1972b) where
both filling and through going cracks occurred in a matter

of hours.

(3) Drained. Hydraulic Fracture:

Vaughan (1976) postulated that drained hydraulic


fracture would occur via path ACD (Figure B-4) as the
increasing pore pressures due to the rising reservoir level
reduced the effective stresses in the core. There appears
to be no reason why paths AB or AEF (Figure B-4) should

not also be followed.

For hydraulic fracture to be drained the seepage


or swelling face must have advanced a significant distance
into a low stressed area of the core before the reservoir
reaches the level at which undrained fracture would have
occurred. As the effective cracking potential is low the
the probability of drained cracking is less than undrained
cracking for a given stress state. If filling is sufficiently
slow that the seepage or swelling face advances far enough
into the core to prevent undrained fracture a crack is less

likely to occur. It is believed that this is usually the


case.
Drained hydraulic fracture is considered to have
occurred at Hyttejuvet Dam (Kjaernsli and Torblaa (1968)),
Balderhead Dam (Vaughan et al (1970) and Viddalsvatn Dam

(Vestad (1976)). At Balderhead the low stresses necessary


for hydraulic fracture may have been associated with the
angle in the upstream face of the core. However, there is
no apparent reason that the stresses at the cracks in.
Hyttejuvet or Viddalsvatn Dams should have been lower than
elsewhere in the core (Wood, Kjaernsli and Hoeg. (1976)).

Drained hydraulic fracture is complicated by the


difficulty in predicting the stress changes that will occur
during shear failure and due to volume change effects.
The following general conditions are of interest:
(a) Vaughan (1976), msing the simple elastic model shown
in Figure B-16 demonstrated that if no swelling of
the soil occurred the reservoir pressure had to be
larger than the average total stress before drained
hydraulic fracture would occur. If swelling took
place ,an even higher reservoir level was required.
Hence the initial stresses in the core have to be lower
than for undrained hydraulic fracture.

(b) As discussed in section B-2, when the cracking potential


is small shearing will probably occur prior to cracking,
stresses will be increased in low stressed areas and
the probability of cracking will be reduced. The
larger the area of shearing the lower the likelihood of
cracking. Conversely, the smaller the area the
higher the probability of cracking. Small areas are
likely to be involved when:
46

(1) initial stresses in a small area are so much lower


than in the surrounding material that cracking occurs
in the small area before the shear strength is

exceeded elsewhere.

(2) pore pressures in a more pervious zone increase more


rapidly than in the surrounding material thus lowering
the effective stresses and causing cracking locally
before the shear strength is exceeded elsewhere. In
this case initiation is drained but propogation would

be undrained (see part (.c) (4) following).

Hence drained hydraulic fracture is considered more likely


to occur when there are local variations in geometry or
fill properties that will lead to effective stresses being

much lower locally at the time of cracking.

(c) In previous discussion it has been assumed that if the


cracking potential is low shearing will occur prior to
cracking (i.e. path ACD on Figure B-4 will be followed)

and the probability of cracking reduced. Conversely,


however, if shearing does not occur the probability

increases.

Figure B-17 shows stress changes measured in the

core of Tarbela Dam (Truscott (1977)1. In both


cases the effective stress path moves away from the
shear failure envelope on impounding. This is
because the vertical effective stresses (it) are
initially higher than the horizontal (ah') and a ' is
47

reduced more rapidly than cy'll on impounding. As the


pore pressures increase the stress difference becomes
zero. A further increase in reservoir pressure will

lead to h becoming larger than a'V and the stress

difference starting to increase again. If it is


assumed that the slope of the stress path remains
constant it can be seen (Figure B-17) that for the
upper location at Tarbela the stress path will intersect
the failure envelope at about the same point as in
path ACD, hence the amount of shearing will be the
same as previously postulated. However for the lower
location the envelope is intersected much nearer the
origin, less shearing will occur and the probability

of cracking is increased. It is quite probable that


in a dam there will be some locations where stress
paths will be such that shearing will not occur before

cracking. If the initial stresses are sufficiently


low at such points then drained hydraulic fracture is
more likely to occur there than elsewhere.

(d) As mentioned in section B2(b) cracking cannot occur


if the major principle stress is greater than twice
the stress difference at D. Under effective stress
conditions the stress difference at D is small. In
a dam with a free draining upstream shell the minimum
lateral pressure on the core (provided the shell is
stiffer than the core) is the active pressure. Hence
there should be a limiting depth below which shear
48

failure will always take place and drained hydraulic

fracture will not occur. It is interesting to note


that at both Hyttejuvet and Viddalsvatn dams
cracking was only reported at the upper levels and
repair of these cracks eliminated the problem.

(4) Propogation of Cracks:

The discussion to date has referred primarily to


crack initiation. Cracks only present a risk to the
embankment if they propogate completely through the core.

Undrained cracks propogate by undrained mechanisms.

Drained cracks may propogate by drained mechanisms


but in this case the movement of the crack through the core

will be relatively. slow. Drained cracks may also


propogate by undrained mechanisms under the following

conditions-'

(a) when the crack moves downstream of the seepage or


swelling face

(b) if the speed of propogation is too fast for pore


pressures to equilibrate.

As the formation of a crack is rapid there will be a tendency


for undrained conditions to occur at the root. Hence
cracking potential will be high and undrained propogation

becomes more likely. Also experience in most cases is


that seepage is observed on the downstream face of the dam
49'

shortly after the crack is thought to have initiated which

indicates a rapid opening of the crack. It is therefore


considered that propogation of cracks is usually undrained.

(5) Finite Element Analyses:

The use of finite element analyses is discussed

in detail in Section B.7. These analyses are valuable


in showing areas in the core where stresses are low and for
investigating the effect of possible remedial measures.

Nobari and Duncan (1972) carried out undrained


analyses that showed that the stress on a horizontal plane
through the core at the end of construction was not
significantly affected by impounding. They then postulated
that hydraulic fracture would occur if the reservoir
pressure exceeded this stress. As the stress on a
horizontal plane is approximately equal to the major
principle stress in the core, this requires even lower
stresses in the core than those postulated by Vaughan (1976)
for drained hydraulic fracture under conditions of no

swelling.

Kulhawy and Gurtowski (1976) determined the end

of construction stresses by undrained analyses of plane

strain sections through zoned dams. They then calculated


the out of plane stress (intermediate principle stress),
assumed this would not change on impounding and postulated
hydraulic fracture when the reservoir pressure exceeded
50

this stress. There was no detailed discussion of hydraulic


fracture.

As mentioned in (2) above, undrained hydraulic


fracture occurs due to the water penetrating preferrentially
into the core via some irregularity such as a crack or
more pervious layer. No studies of this mechanism are known.

No analyses have been found in which the core is


drained. As significant drainage will have occurred in
most embankment dam cores prior to hydraulic fracture this
is the pertinent case.

(6) Laboratory Modelling of Hydraulic Fracture:

Modelling of mechanisms leading to cracking is


discussed in Section B-8. Undrained hydraulic fracture
in boreholes has been modelled in the laboratory by Bjerrum
et al (1972). Nobari et al (1973) used a hollow cylinder
aparatus to model drained hydraulic fracture under conditions
of very low shear stress.

Both these models demonstrate that hydraulic


fracture will occur in soils subject to excessive water
pressure. Unfortunately neither test correctly models
conditions at the upstream face of the core during
reservoir loading.
(7) Conclusions:

(a) Hydraulic fracture will only occur in areas where the


end of construction stresses are low.
(b) Undrained initiation of cracks by hydraulic fracture
will only occur under very rapid reservoir loading
conditions.
(c) Drained initiation is most probable under usual
reservoir loading conditions.
(d) Propogation of cracks can occur under either drained
or undrained conditions. The speed of reaction suggests
undrained is more frequent.
(e) Higher reservoir pressures are required for drained
cracking than for undrained cracking, especially if
swelling occurs.
(f) As the effective cracking potential is small drained
hydraulic fracture will probably not occur without
prior shear failure which will increase stresses in
low stressed areas.
(g) The increase of stress due to shear failure is likely
to increase with the size of the zone affected by shear
failure. Drained hydraulic fracture is thus more
likely where the zone of initial low stress is small, or
where the zone affected by the seepage pressure is
small.
(h) There is some evidence that there are locations in a
core where the stress path followed does not intersect
the shear envelope prior to cracking. Provided the
initial stresses are low hydraulic fracture is more
likely to occur at such locations.

(i) In an embankment with a freedraining upstream shell


there is a limiting depth below which drained hydraulic
fracture is not likely to occur.

(j) Although finite element analyses are valuable in


indicating low stressed areas in dam cores, they have
not yet been used to correctly model either drained or
undrained hydraulic fracture.

(k) Although hydraulic fracture has occurred in the


laboratory, the stress conditions under which it
occurred were different from those at the upstream face
of a core during reservoir loading.
(1) Hydraulic fracture is a complicated mechanism which is
not yet fully understood. Additional studies are
needed on this mechanism.

B.5 TRANSVERSE CRACKING DURING RESERVOIR OPERATION

(a) General

This section discusses cracking during reservoir


operation, primarily to provide a base for discussions
under Section H where the likelihood of cracking in old dams

is discussed.

Most cracks that have occurred during reservoir


operation have been due to the continuation of processes
which started during construction or first impounding of
the reservoir as follows,
53

(1) Continuing settlement of dam or foundation - e.


Virginia Ranch Dam (Sherard (1972)).

(2) Advance of seepage in foundation or dam core leading


to additional settlements or hydraulic fracture.

These factors ,are not discussed further under this heading.

(b) Cyclic Reservoir Operation:

Cyclic loading of soils at stresses below the

failure line leads to a reduction in volume. In general


the amount of reduction reduces with each cycle so that
after several cycles volume changes are very small and the

soil behaves almost elastically. However, the volume


reduction will lead to additional settlements and these

could cause cracking. Because of the tend to elastic


behaviour cracks due to this cause should not occur after
the first few years of operation in a properly compacted

dam.

In most reservoirs loading cycles are once or .


twice a year. ' In recent years pumped storage hydroelectric
projects have developed and dams for such projects will be
subject to many more loading cycles during their lifetimes.
Although loading is virtually elastic after a few cycles
small volume changes do continue to occur. Hence, these
dams may continue to settle fora long time and cracking

could eventually result.

The settlements due to cyclic loading are difficult


54

to separate from those due to creep of the embankment

materialb. Hence identification of cracks in dams due to


this cause is difficult. One possible example is the Hills
Creek Dam-(Jenkins and Bankofier (1972)). where a crack
developed over a disContinuity due to a haul road on the
abutment after 6'yearS of satisfaCtory operation.

(c) Long Periods of Reservoir Empty:

The first effect of a period when the reservoir


is left empty is to allow drying out and shrinkage cracks.
This-matter is fully discussed in Section B,3(b)(7).

If the core material is placed wet so that high


construction pore pressures occur and the reservoir is
filled before these pore pressures drop below the steady
seepage levels then the effective stresses in the core
will always have been low. In awet climate or if the
reservoir level is high for most of the year, pore pressures
may remain high and effective stresses low for many years.
A long period of empty reservoir will allow the pore
pressures to dissipate, effective stresses to rise and
settlements to occur. These settlements could lead to
cracking of the core.
55

B.6 DEFORMATION PROPERTIES OF CORE MATERIAL.

(a) General:

Although the formation of a crack depends on


stress changes in the soil these stress changes are
influenced by the deformation characteristics of the soil.
This section describes the effect of various deformation
properties on the likelihood of cracking and how these
properties vary with changes to other soil properties.

Table B-2 summarizes this data.

The optimum core material depends on its


cracking potential, erosion resistance, location in the
core and availability as well as its deformation properties.
All these matters are discussed in section B.12.

(b) Tensile Failure Strain:

The tensile failure strain must be exceeded before


cracking can occur. Near the dam crest the vertical
stresses are small and the tensile failure strain will be
close to that measured in the direct tensile tests. At
depth in a dam the situation is more complicated, however,
it would appear that the tensile failure strain will be
approximately equal to that due to compressive stresses
parallel to the potential crack plus that measured in the

direct tensile test. Hence the likelihood of cracking


reduces as the tensile failure strain as measured in the

direct tensile test increases.


56

Results of tensile testing of soils are summarized


in Appendix B. The relevant tests range from drained to
undrained on partially saturated materials. There appears
to be reasonable agreement between the results. As there
is no recognized standard for tensile testing of soils and
as the behaviour of soils in tension is not fully under-
stood the following discussion must be regarded as tentative
only.

Several investigators (Leonards and Narain (1963),


Nonveiller (1973) and Gopalakrishnayya (1973)) have
investigated the effect of water content on failure strains
and all agree that the failure strain increases up to
Proctor optimum moisture content (O.M.C.). Leonards
and Narain found it reduced above O.M.C. but the others found
it increased, with Nonveiller obtaining a maximum failure
strain at O.M.C. plus 8 per cent.

Nonveiller (1973) and Gopalakrishnayya (1973) also


investigated the effect of compactive effort and found a
slight increase in tensile failure strain with increased
compactive effort.

The effects of increasing plasticity are not clear.


Gopalakrishnayya (1973) found that the addition of 6 per
cent bentonite to Mica Dam material caused a reduction in
failure strain. Tschebotarioff et al (1953) found that
increased plasticity resulted in increased failure strains.
It is tentatively concluded that where the percentage clay is
57

small it has little effect on the failure strain but where


it is large it will increase the tensile failure strain.

(c) Volume Change Characteristics:

As discussed in section B.4(c) the end of


construction stresses at which drained hydraulic fracture
can occur decrease as the amount of swelling increases.
Hence any increase in swelling reduces the likelihood of

hydraulic fracture. Swelling will also tend to seal


cracks formed in the core during construction. Chaudhary
(1955) investigated the effect of moisture content and clay
percentage (i.e. plasticity) on swelling and found that
swelling increased with reducing water content and

increasing plasticity. Lambe and Whitman (1969) also


report increased swell index with increased plasticity.

In a dam core volume reductions are caused by


applied compressive stresses, saturation and shrinkage on
drying. Volume reductions lead to large settlements and
increased arching between the shells, both of which
contribute to lower stresses. Hence, volume reductions
should be minimized.

The amount of compression that occurs under


applied average compressive stresses is defined by the
bulk modulus of the soil (Average stress divided by
Volumetric strain ) . Under drained conditions this is
increased by the compactive effort applied. Chaudhary
(1955) and Lambe and Whitman (1969) report that the bulk
58

modulus reduces with increased plasticity. Lambe and


Whitman (Table 34-1) report that soil is more compressible
on the wet side of O.M.C. at low stresses but more
compressible on the dry side at high stresses. As maximum
bulk modulus would be expected at maximum density compaction
at O.M.C. is the most desirable.

The undrained bulk modulus depends primarily on


the degree of saturation of the voids and increases with
increased saturation until at one hundred percent saturation
the modulus approaches that of water.

The Volume change of eMbankment material on


saturation haS been investigated by Chaudhary (1955),
Leonards and Altschaeffl (1971) and Nobari and Duncan
(1972). All investigators found that the amount of volume
reduction increased as the moisture content was reduced
below 0.M.C. and that volume reduction at O.M.C. was small
to negligible. Chaudhary (1955) investigated the effect
of clay content on volume reduction due to saturation. He
found that at small stresses swelling predominated. For
dry soil (at O.M.C.- 4 percent) under higher stresses
(4 Tons/ sq.ft.) he found that at small clay contents
(3 percent) only a small volume change occurred, from 3 to
about 20 percent clay a large volume reduction occurred
and above 20 percent there was no further increase in
compression due to saturation.

Soil will not shrink on drying if its moisture


59

content is at or below the shrinkage limit. Hence the


amount of shrinkage will reduce as the moisture content is
reduced. The shrinkage limit increases with reduced
plasticity (Lambe and Whitman (1969)). Hence shrinkage
will increase with increasing plasticity.

(d) Shear Movements:

The magnitude of shear movements is defined by


the shear modulus (shear stress divided by shear strain).

In areas of potential cracking such as dam


abutments it is desirable that shear movements be large
(i.e. shear modulus low) in order to promote stress
averaging and thus reduce the possibility of cracking.

In the bulk of the dam shear displacements should


be minimized so that settlements and arching, which cause
low stresses, are minimized.

.The effect of soil properties on the shear


modulus have been obtained from Lambe and Whitman (1969)
as follows:

(1) Modulus reduces with increasing water content


(2) Modulus increases with increasing density
(3) Modulus reduces with increasing plasticity.
60

B.7. NUMERICAL METHODS OF CRACK PREDICTION


(a) General:

It is not intended to carry out a detailed review


of all analyses that have been carried out. The advent
of the finite element method of analysis has resulted in
a relatively easy tool for analysing dam shapes, sections
and abutment geometries under idealized conditions to check
for the possibility of tensile or reduced compressive zones.
As with all analyses the results are no better than the
material properties used in the analyses. Also cracking
depends to a large extent on local variations in material
properties and modellithis effect is difficult. It is
considered that at the present level of the art (particularly
in determination of stress-strain properties of materials)
the results are not quantitatively accurate. Such
analyses do provide a very valuable tool for assessing the
qualitative effect of modifications to material properties
or structure geometry. Eisenstein (1974) presents a
summary of the present state of the art of finite element
analysis as applied to earth dams.

Some investigators have proposed simple methods


of determining the likelihood of cracking. Leonards and
Narain (1963) developed a method in which the dam was
replaced by a simple beam subjected to foundation
displacements. Tensile strains calculated from this beam
analogy were compared with tensile failure strains from
beam tests. Some correlation was obtained between their
calculated tensile strains, the beam test results and
tensile failure strains observed in practice. Such methods
of calculation have today been superseded by the finite

element analysis.

(b) Development of Finite Element Method in


Crack Determination:

Analyses have been carried out by many investigators


and only selected examples are presented here. Other
investigators have also looked into the problems noted.

Covarrubias (1969) appears to have carried out

the first analyses. He investigated the effects of


abutment shape, foundation compressibility, embankment
compressibility and embankment. zoning on the development of
tensile strains on the crests of dams under embankment

loading. He compared.his results with measurements on


dam crests and found that calculated tension zones
correlated well with those observed in practice even with
a two dimensional, linear elastic analysis.

Laing (1971) investigated the effects of core


shape and fill properties on the arching of the core of a
dam between more rigid shells. He showed that arching
under dead load conditions could reduce the minor principle
stress in the core to below the water pressure that would
be applied to that point under reservoir loading. A
reduction of core compressibility relative to the shells
62

reduced stresses in the core and a change in slope of the


core face also reduced stresses at that point.

Nobari and Duncan (1972) developed a two


dimensional, non linear program in which the effects of
water load (including collapse of the upstream shell on

saturation) were modelled. They compared calculated and


measured values for Oroville Dam and obtained reasonable

agreement.

Gopalakrishnayya (1973) developed a three


dimensional program into which he put the foundation
settlements measured at Duncan Dam and from which he
obtained tensile zones in the same area as cracking was

observed at Duncan Dam.

Maksimovic (1973) studied the effect of inclining

the core upstream to reduce the effect of arching between

the shells and the core. He showed that some improvement


resulted from the inclined core but it was relatively small
and is probably less significant than most other factors

determining the position of the core.

Nobari et al (1973) carried out two


finite element analyses in which they used the total stress
on :horizontal plane as an indication of the possibility
of hydraulic fracture and concluded that "the likelihood
of hydraulic fracture may be reduced by making the core
wider, compacting the core at a lower water content, using
transition zones which are less stiff than the shells and
63

using a sloping core rather than a central core." Their


analyses did not take into account any effect of increased
pore pressure in the core due to seepage.

Results of preliminary analyses at Imperial


College (Davachi (1978)) indicate that in the core of a dam
the total stresses are increased when pore pressures are
increased due to seepage. If this is proved correct it
will confirm the general statements made in this report that
saturation reduces the shear strength and thus the cracking
potential of the soil with the result that effective shear
strength of the soil is exceeded prior to the tensile
strength. Thus shearing occurs with a corresponding
sti-ess
increase in average total s4e4cigth.

(c) Use of Finite Element Method in Design:

Articles referred to in part (b) above refer to


cases where the finite. element method was used either to
obtain behaviour observed in practice or to check the
effect of varying factors on the probability of cracking.
It has only been possible to find one case where the
analysis was performed before the dam was built, the
results used in the design and checked by instrumentation
readings during construction. This was the case of Lower.
Notch Dam (Sigvaldasori and Tawil (1972)).

Lower Notch Dam is located over a very narrow


gorge with vertical sides as shown in Figure B-18. The
64

problems investigated were the possibility of cracking due

to the differential settlements at the edge of the steep


gorge and the possibility that arching between the rock
abutments would reduce the minor principle stresses to
very low values in the gorge.

The first series of analyses were undertaken to


investigate the degree of flattening required to the side
of the gorge where it met the horizontal ground surface
to provide significant relief from cracking. It was found
that, within practical limits, flattening of the slope
effect but that a suitable transition from the flat
surface to the vertical gorge wall would provide a

significant benefit. It is interesting to compare this


mathematically obtained behaviour with the assessment based

on experience by Sherard (1972) where he said " The


steepness of the rock abutment is considerably less important
from the standpoint of cracking potential than is the
existence of pronounced surfade irregularities".

The arching of the core between the gorge walls


was carried out using a' finite element program in which a
shearing term was introduced such that shearing took place
at the abutments when the shear strength of, the soil was

exceeded. It was found that because of the almost vertical


abutments the shear strength was exceeded when the vertical
stresses in thefill fell below about 10,000 pounds/square
2
foot (480 MN/m ). Hence the minimum vertical stress in
the core would not drop below 10,000 pounds/square foot
which was considered acceptable. Stress cells were
65

installed in the core and stresses measured (see Figure

B-19). It was found that the vertical stress reduced


rapidly to half the overburden pressure and then reduced
very slowly at least partially confirming the finite

element results.

B.8. CRACK PREDICTION USING MODELS.

(a) General

The use of physical models'to investigate cracking


in earth cores is difficult because in all cases the zone
of low or tensile stress where cracking initiates is due
to a gravity loading condition. Very little work has
therefore been done before the advent of centrifugal testing.

Taylor (1969) refers to construction of a model


abutment to investigate the best shape and slope to
minimize cracking in W.A.C. Bennett Dam. However, no
details are given.

Leonards and Narain (1963) used a simple physical


model when they proposed that the failure strain at the
crest of an embankment could be related to the failure
strain measured in beams in the laboratory. This matter has
been discussed more fully in Section B7(a).

(b) Centrifugal Testing:

In centrifugal testing prototype gravity loads are


66

applied to a small model of the prototype by subjecting


the model to accelerations many times that of gravity,
using a centrifuge. The problems with this method are
correct modelling of the soil conditions and the size and
expense of the centrifuge. The test method is still in

the developmental stages.

Vutsel et al (1973) used centrifugal testing in


checking for cracks in the design of Nurek Dam. They
found that cracks up to 10 metres deep could be expected
near the abutments. In order to avoid problems more
flexible material was placed in these zones and a surcharge
was applied to the embankment in these areas. Bassett
(1973) carried out centrifugal testing of an. embankment
on a soft foundation to check its stability. During
testing a crack occurred through the upstream sloping core
for the end of construction case. It was decided that
this crack was caused by excessive settlements of the shell
on which it was supported. The excessive settlements were
due to the sand being placed in the saturated condition
and the construction procedure was modified to ensure that
the sand drained before it was placed. The embankment
has now been built and no cracking has occurred.

Bassett's paper also indicates clearly the


problems associated with all modelling whether physical or
mathematical. The dam mentioned above was checked for

stability and found satisfactory. On being constructed a


slide occurred through the foundation. Subsequent
investigations showed that a hitherto undiscovered weak

layer existed in the foundation. On modelling the correct


foundation conditions the field behaviour was reproduced in
the centrifuge. Hence it is considered that no matter
how accurate modelling techniques become the physical
limitation of correctly defining foundation and embankment
properties will still place some doubt on the results.

It is considered that centrifugal testing can


probably play a significant part both in checking the
likelihood of a crack occurring in a dam and in studying the
mechanisms that cause cracking.. The main advantages are
that the soils themselves are used hence there is no need
to depend on mathematical relationships to define material
behaviour and that three dimensional effects can be

modelled directly. Also,cracking is a strain problem


which scales directly in a centrifugal model. There are

also significant disadvantages as follows:

(1) Difficulty in obtaining information. - it may be


relatively simple to observe if a crack has formed
but may be much more difficult to determine why it

occurs.
Difficulty in modelling large sized material e.g.

rockfill.
Hydraulic fracture is thought to be more likely to
occur in low stressed areas due to local variations in

geometry or, material properties. Reproducing these


variations in a small model would be difficult.
68

(4) Although strain scales directly on a centrifugal


model other factors such as time effects and stress
gradients do not. It has already been noted (Section
B2 (d)) that beam tests give higher tensile strengths
than direct tensile tests due apparently to higher
stress gradients inthe beam tests.

(5) Hydraulic gradients through the core are much higher


in the model than in the prototype. This relates
more to erosion of cracks than formation of cracks.

Despite the above disadvantages it is considered that


centrifugal testing is one of the most promising methods
available for learning more about the cracking of dam
cores and, in particular, of the mechanisms leading to
cracking in dam cores.

(c) Modelling of Mechanisms:

One area where physical models can help in the


understanding of cracking of dams is in the modelling of
the mechanisms of cracking. Bjerrum et al (1972) and
Bjerrum and Anderson (1972) describe two methods in which
hydraulic fracture was induced in bore holes in models in

the laboratory. In these methods the crack was physically


observed and the pressures required to cause hydraulic
fracture compared with theory.

A similar modelling arrangement could be set up


to investigate hydraulic fracture (under both drained and
69

undrained conditions) due to reservoir loading on the


upstream face of a dam. Vaughan in an unreported series
of tests has carried out a simple test of this nature on
non-cohesive material and found that a soft spot formed

under high applied water pressures. Nobari et al (1973)


carried out drained tests in which cracking was induced in
a soil under very small shear stresses thus successfully

modelling drained hydraulic fracture. It would appear


that the method used by them could relatively easily be
modified to give drained results under shear loading and
undrained results with typical discontinuities introduced
into the upstream face. Such information would be very

valuable.

As mentioned in Section B2(a) the complete failure


envelope for both undrained and effective stress conditions

has not been defined. Bishop and Garga (1969) describe


a test method whereby the complete range of stresses,
including tensile stresses, can be applied to a soil
sample. . Wesely (1975) reports on a triaxial cell in
which any stress path can be followed to failure. The
combination of these two methods should enable the
determination of the complete failure envelope in the

laboratory for any soil. This information would be valuable


in assessing cracking potential.
B.9. OTHER MECHANISMS CAUSING THROUGHGOING
SEEPAGE PATHS.

(a) General:

This section deals with throughgoing seepage paths


that are caused by mechanisms other than cracking. These
may be completely separate from cracking, act in conjunction
with cracking or may lead to transverse cracking.

(b) Poor Construction Techniques:

These will not generally lead to throughgoing


seepage paths but will act in conjunction with cracks in

providing a water passage. The following construction


deficiencies could lead to preferred seepage paths through.

the core;

(1) Zones in the core of the dam which are poorly compacted.
(2) Local use of dry material. This material could
collapse on saturation or be so lumpy that either voids
or very loose material are left between the lumps after

compaction.
(3) Contamination of the dam core by material from granular,
adjacent zones.
(4) Poor borrow area operations resulting in unsuitable
material in the core,

(5) Segregation of a well graded or gap graded core material.


(6) Inadequate care in construction of trenches through
the core for instrumentation.
71

(c) Water Passages in the Foundation:

Foundations of embankment dams can consist of bedrock


or soil. Control of seepage through the foundation is
usually by means of a grout curtain, grout blanket,
impervious fill or other blanket or some form, of cut-off

wall. It is not intended to discuss the construction or


relative benefits of these but to discuss the effect of
inadequate seepage controls on development of water passages

through the core.

An uncontrolled seepage path through the foundation


can provide water at almost full reservoir head to locations
downstream of the upstream face of the core. Where
seepage control is by means of a vertical cut off the
hydraulic gradient through the core may be greatly increased
as it will consist of the full reservoir head divided by,
the width of the vertical cut off at its junction with the

core. Should an inadequate filter occur between the core


and the foundation (such as when there is an open joint in
the rock) downstream of the cut off the core material can
pipe into the foundation and a throughgoing crack result.
In relatively compressible foundations, sealing of cracks
with grout or shotcrete prior to placing fill may not be
satisfactory as the brittle grout or shotcrete could
easily crack under the load of the fill.

Where the grout curtain is ineffective or the


alluvium is quite pervious and there is no satisfactory
72

filter between the core and foundation the velocity of


water in a more pervious part of the foundation may wash
the core material into the pervious zone thus leading to
a throughgoing hole at the bottom of the core.

A more pervious zone, either a crack in a bedrock


foundation or open gravel layer in a soil foundation, may
exist beneath the downstream part of the core and the well
designed filter zone downstream of the core. The core
material may then wash into this open zone and wash out
downstream of the filter layer thus leading to a through-

going water passage. It is suggested that this, in


conjunction with a crack in the core,' lead to the problems
at Yard's Creek Upper Reservoir Dam (Sherard (1972)).

(d) Failure of Structures Passing Through the Core:

Diversion works or permanent water supply structures


sometimes consist of a steel. or concrete structure which
passes through the embankMent. Should a break occur in
this structure the core may be washed into it and a
throughgoing water passage develops. This was the cause
of the trouble at Lluest Wen Dam (Little (1971)).

(e) Piping of the Core:

Piping of the core is the mechanism whereby the


water seeping through the core washes particles out of the
downstream side of the core and the hole so formed pipes
back to the upstream face of the core. Piping can be
prevented by use of a properly designed filter layer down-
stream of the core. This matter is discussed in detail
in Sections C and D.
73

(f) Solubility of the Core Material:

Some naturally occurring soils are soluble to


differing degrees in water. Where a dam core is
constructed of such a material then a throughgoing water

pasage could result. Anagnosti (1970) describes a dam


with a laterite core where seepage through the core became
so large that it was necessary to construct a sheet pile
wall through it to ensure a watertight barrier. Chemical
analysis of the laterite showed that the aluminium and
iron salts dissolved in the seepage water and the remaining
silica went into colloidal suspension. Therefore the
downstream filter could not retain any of the core material.
This is a specific case of a laterite being unsuitable as
a core material. Many dams have been built successfully
using laterite as core material (Little (1967)).

In most dams constructed using modern compaction


techniques seepage through the core is very small and even
if the core is soluble it should not be a problem. In any
event monitoring of the chemistry and quantity of seepage
water should reveal the problem long before it becomes
serious. A soluble core material may cause serious problems
when it occurs in conjunction with the much higher water
flows associated with a crack through the core, or a
pervious foundation. Under these conditions the rate at
which the core could be dissolved could be increased markedly.
As filter zones provide no protection against this mechanism.
74

the water passage would continue to enlarge until the flow


of water was stopped either by lowering the reservoir or
some form of remedial measure.

B.10. CRACKING DUE TO EARTHQUAKES.

(a) General:

This section discusses briefly the way in which


earthquakes can cause cracks through the cores of
embankment dams and design methods that should be adopted
to minimize cracking and eliminate the risk•of embankment
failure due to such cracks. It does not pretend to deal
exhaustively with the subject and a more detailed review
is presented by Sher and (1967)and Sherard et al (1974).

Although several embankments have been subjected


to quite severe earthquakes there is no record of any failing
due to cracking of the core although severe damage has
frequently resulted.

It has recently become apparent that reservoirs


themselves may be the cause of earthquakes, even in areas

that have been previously seismically inactive, due to


the large additional load imposed on the earth's surface or
due to increased pore water pressure at depth beneath the
reservoir. The mechanism of reservoir induced earthquakes
and the probability of their occurrence is still incompletely
75

understood. Hence it must be assumed that, even in areas


that are seismically inactive, the possibility of a
significant earthquake exists and, if the dam is located
on a fault, movement may occur on that fault due to reservoir

loading.

(b) Causes of Cracks:

(1) Changes to the Foundation

One way in which earthquakes can cause cracks


in dam cores is by modification to the foundation.

Changes can occur either due to movement on a


fault or associated feature or can occur due to the
behaviour of a soil foundation under the vibrational loading
imposed by the earthquake. Movements associated with
faults appear to fall to low values within 3 Kilometres of
the fault (Bonilla (1970)) but changes to soil foundations
can occur in any area when subject to large enough

earthquakes.

Movements on faults can occur in three directions


as shown in Figure B-20. Tensile forces can be applied
to the core when the movement of the fault increases the

length of the dam axis. In addition to movements on the


fault itself compression and tension zones are set up in
the rock adjacent to the fault which can result in open

cracks.

In literature some people talk of faults opening


76

up by large amounts and others claim faults themselves

cannot open up. In any event cracks can open up either


at the fault or in the disturbed zone adjacent to the

fault. Bonilla (1970) talks of fractures that opened


up to 10 feet wide in conjunction with the Nevada Fault,
and another that opened wide enough for a cow to drop into

and then closed again. Should such openings occur beneath an


earth dam it would be very difficult to ensure the
safety of the structure and dams should not be built in
the vicinity of faults where such movements could occur.
Fortunately openings of this magnitude are very infrequent
and openings generally observed are less thah 1 foot.
These should not present any risk to a properly designed

embankment.

Sliding on faults can cause extensions of the


dam axis and displacement of the zones of the embankment.
Sherard (1974) indicates that the maximum horizontal
and vertical displacements that have been observed are
8 metres and 14 metres respectively. Hence protection
against dam failure can be assured by making the zones
large enough that this amount of movement does not result
in an unsafe embankment. Details of how this is achieved
are shown in Figure B-21.

Tensile forces are applied to an embankment when


the block on one side of the fault tilts relative to the

other. No quantitative data has been found on this aspect


of cracking.
77

In a soil foundation the effect of the earthquake


depends on the nature of the soil. Where a saturated
sand layer exists beneath a stiff surface zone the
earthquake can cause excessive pore pressures in the soft
saturated layer which can result in water forcing its way
through the stiffer layer and causing cracks in that layer.
Where soil is soft and cohesive the earthquake forces will
cause transient compression and tension zones in the soil

which could result in cracks. Cracks caused by these


two mechanisms should not be large and should not endanger

the dam.

In a granular foundation the earthquake could


result in additional settlements with the increased
possibility of differential settlement cracks as discussed

in Section B.3(b).

Movements at faults or existing weaknesses in


the foundation do not only occur during large earthquakes.
Problems due to a fault beneath an embankment occurred at

Baldwin Hills Reservoir. In this case the cumulative


effect of several small earthquakes plus erosion of fine
material from the fault lead to the failure of the
embankment (Casagrande et al (1972)). Tomasson (1976)
reports that the pressure due to the reservoir water forced
open cracks in the foundation of a small dam. Nilsson
(1964) reported sink holes on the crest of Messaure Dam
which were thought to have formed by stoping upwards
78

from a fault in the dam foundation thus indicating problems


associated with the fault. Hence care must be taken in
the design of embankments on faults even when there is.no
reason to expect any future activity on the fault.

'(2) Changes to Embankments:

The effect of earthquakes on embankments is


discussed in detail by Sherard (1967).

Earthquakes can cause additional settlements in


both the shells and cores of dams thus increasing the
possibility of cracks.

In a saturated zone within the embankment the


shaking by the earthquake can set up excessive pore pressures
If the dam core is on the limit of cracking such local
excessive pore pressures could result in cracking. The
stress and strain redistribution that accompanies cracking
could then result in the crack remaining open after the
earthquake had passed.

(3) Changes in Reservoir:

Earthquakes can. result in large waves occurring


in the reservoir. The increased water preSsure associated
with such waves could cause undrained hydraulic fracture
Of the core.
79

(c) Design Requirements:

In order to design an embankment to resist


cracking under seismic loading it is necessary to carry
out a thorough seismic investigation to define the
following factors:

1. The location and extent of all faults, crushed zones


and earthquake sensitive soils that could threaten the
safety of the dam.
2. The activity of the faults and an. estimate of movements
that could occur on all faults or crushed zones.

3. The possibility of reservoir induced earthquakes.


4. The magnitude and probability of seismic accelerations
and velocities that could occur in the embankment and

foundation.
5. The location of any areas in the reservoir area where
inundation or earthquakes could cause large slides thus
causing large waves or possible damage to the dam.

It would appear that major problems could occur


if transverse joints or faults open in the foundation and
form water passages beneath the dam as reported by Tomasson

(1976). No discussion of this problem has been found in


the literature. If such openings are large (such as the
10 feet mentioned by Bonilla (1970))then safety cannot be
guaranteed and the site should be abandoned. If the
opening is smaller and considered acceptable, cracks may
be formed in the core. Erosion of these can be prevented
by the use of a properly designed downstream filter as

described in section D. The main problem is the erosion


of the core material into the opening in the foundation
from where it could be washed out of the dam. Three
methods of dealing with this problem are:

1. Use of a tough clay as core material. This will arch


over the opening and resist erosion by the water

flowing in the crack. It is considered that the long


term resistance of the clay to erosion, especially under
turbulent flow conditions, is doubtful. Hence a
cohesionless downstream filter should.be provided that
will control erosion of the clay (see Section D) and
will flow into any significant opening in the foundation.

2. A low plasticity core material will not collapse into


the opening if its grading is such that the opening acts
as a filter relative to the core (see Section D).
Hence the opening will not enlarge as a result of
erosion of the core.
3. If the core material is non cohesive it will collapse
into the crack provided it is sufficiently fine. Hence
the opening will be sealed. However, if normal filter
design criteria are followed,there will be a range of
crack openings for which the core will collapse into
the crack but the filter will bridge over the crack.
Therefore the core may pipe at the downstream side in

the crack and be washed away. It therefore appears


81

desirable to provide a supplementary filter downstream

of the core in such. areas. This filter would flow


into the same size cracks as the core material but
would be more pervious so that the exit gradients would
be small and downstream erosion would be resisted.

The problems of excessive seepage through the


foundation and erosion of fine material in such openings
are also important but are outside the scope of this report.

Where vertical or horizontal offsetting of the


core and other zones is expected the dam zones must be of
sufficient dimensions that even after the offsetting
has occurred there is still adequate overlapping of each

zone and the safety of the dam is ensured as shown in


Figure B-21. Sherard (1967) suggests that for dams where
offsetting is likely to occur a wide filter zone of broader
grading is more desirable than a series of filter and
drainage zones of increasingly coarser gradation.

Dynamic loading due to earthquakes leads primarily


to increased settlements in the dam and foundation. The
measures required to reduce the probability of cracks due
to settlements and the way in which the safety of the dam
can be ensured in the presence of these cracks is discussed

in sections B.12 and D.

Where faults exist beneath dams movements across


them should be fully monitored and an instrumentation
program set up to carry this out for the life of the project.
82

B.11. LONGITUDINAL CRACKS.

(a) General:

Longitudinal cracks are not discussed in the same


detail as transverse cracks as they do not provide a water

passage through the core and are not so important. They


tend to occur more frequently than transverse cracks and
generally are not the cause of undue concern. However,
they can have adverse effects and these are discussed in

this section.

In many cases longitudinal cracks occur at the


upstream face of the core and are accompanied by diagonal
surface cracking which makes them difficult to distinguish

from a slide. In fact slicqing can occur if the depth of


the crack would otherwise expose a vertical surface that was

unstable. In such cases it is usual to investigate the


feature thoroughly to ensure that it is not a failure
surface on the upstream face of the core (Sherard (1972)).

(b) Causes of Longitudinal Cracks:

(1) Differential Settlement between Embankment Zones.

Longitudinal cracks can occur between the shells


and core of a zoned embankment when the shells settle more
than the core because they are more compressible than the
core (for example in a dumped rockfill embankment) or where
the upstream shell settles substantially on saturation.
Where a homogeneous embankment is built of material that
83

settles on saturation longitudinal cracks can occur as


saturation progresses from the upstream face in a downstream

direction. Examples of longitudinal cracks caused by


saturation of shell materials are given by Nobari and Duncan

(1972). Details are shown on Figure B-22.

(2) Differential Foundation Settlement:

Where a dam is to be built on an alluvial

foundation a cutoff to rock is frequently constructed


beneath the core to control seepage beneath the dam. In
such a case the core will be founded on material that is
stiffer and will settle less than foundation beneath the
shells. A similar condition exists as seepage through
the foundation progresses in an upstream direction. The
upstream part becomes saturated first and settles thus
allowing the shell to settle and longitudinal cracks

form. This mechanism is discussed in Section BA(b).


Examples of cracking due to saturation of foundation materials

are given by Nobari and Duncan (1972). Details are shown

in Figure B-22.

(3) Spreading of Foundation:

In a dam founded on a weak foundation large


horizontal movements can occur and tensile strains develop

at approximately the centre of the dam. Cracks can


therefore occur at this point. Examples are given by
Bishop (1966) and details are shown in Figure B-22.
84

(4) Influence of Structures Projecting into


the Embankment:

Where an embankment joins a rigid structure such


as a spillway one way of building the junction is to wrap
the embankment around the end of the structure. A
longitudinal crack can develop at the upstream face of the
concrete due to the large difference in settlement properties
of the concrete and fill material. This crack can then
project into the embankment for some distance. An example
of this occurred at Tarbela Dam (Truscott (1977)) and
details are shown in Figure B-22.

(c) Problems Resulting from Longitudinal Cracks:

(1) Causes of Transverse Cracks.

Where longitudinal cracks are due to settlement


of the upstram shell the stresses in the core will be

increased. Hence additional settlement will occur in the


core. This could lead to transverse cracking near the
abutments for the reasons given previously. Transverse
cracks apparently occurred at Cougar and Round Butte Dams
(Sherard (1972)) for this reason.

(2) Connecting Non Throughgoing Water Passages.

A longitudinal crack may provide a water passage


between two otherwise unconnected non-throUghgoing water

passages. Sherard et al (1972a) give an example of a dam


where the water passage through the dam went along a crack
85

in the foundation through a longitudinal crack on the


upstream face of the core and along a transverse crack
through the core adjacent to a concrete outlet pipe.

(3) Loss of Freeboard.

Settlement of the upstream shell on saturation


accomPanied by the development of a longitudinal crack at
the upstream face of the core exposes the upper part of the

core. If there is little freeboard and the reservoir is


full this exposes the core to the erosive effects of wave
action and could eventually result in overtopping the dam.
In most dams this situation is not likely to occur because
settlements will not be large enough to expose the core
and the problem can be readily observed and repaired before
any deterioration of the core can occur.

B.12 GOOD DESIGN PRACTICE TO MINIMIZE CRACKING.

.(a) General:

It is not considered that existing methods of


field investigations, knowledge of cracking mechanisms or
methods of analysis can ensure that a transverse crack will
not occur through an embankment. There may be an
occasion when very favourable conditions exist and the risk
of cracking appears negligible. However, in such cases
the safety of the dam depends on complete compliance to
design requirements, perfect workmanship and that no changes

are found in the foundation or borrow areas. This appears


86

inadvisable, hence it is considered that all designs should


assume cracking and provide adequate filters to prevent
erosion of cracks (see Section D). However, the methods
outlined in this section should be adopted to reduce the
probability of cracking to a minimum.

Before deciding on the best methods it is necessary


to define the mechanisms which could cause cracking of a
particular dam and adopt measures to minimize the risk of
cracking due to that mechanism. The correct measures for
one dam are not necessarily those for another.

In some cases the best measures to resist cracking


are a matter of opinion and cannot be confirmed by analyses,

laboratory testing or field data. Table B-3 has been


prepared to summarize methods adopted for various embankments.

In at least two cases (Section B.8(b)) centrifugal


testing has been used to check for cracking. If there is
considerable concern that cracking may occur in any
embankment this appears to be the best method available for
checking the possibility. Experience with the method is
limited and even if the modelling indicates cracking will
not occur the dam must be designed to ensure that cracking
of the core will not result in failure.

Cracking due to earthquakes is, discussed in


Section 410 and will not be discussed further in this section.
87

(b) Geometric Considerations:

Cracking of the dam core can be minimized by


paying attention to the dam location, geometry and layout

of internal zones.

One common recommendation is that the dam axis be


arched upstream and located upstream of any restriction in
the river (see Figure B-23) so that the application of
water load will induce compression in the core. Sherard
et al (1963) note that the effect of this is probably small
but the curve cannot harm performance. As discussed

previously the movement of the core .on impounding may


initially be upstream due to collapse in the upstream shell
on saturation. Where this occurs the core which is
arched upstream or wedged into the abutments will have
tensile strains induced due to reservoir loading and thus
the probability of cracking will be increased. No data
has been found where longitudinal strains were measured on
such a curve during impounding but at Tarbela the reverse
occurred - i.e. compressive strains were measured. on a
section arched downstream and these were marginally higher
than those measured on the adjacent straight length

(Truscott (1977)). Experience is that after the initial


upstream movement subsequent movements are downstream.
Hence if cracking does not occur during the initial
impounding behind such an embankment it is less likely to

do so subsequently.
88

In a large number of cases cracking of dam cores


has been associated with projections or rapid changes of
slope in rock abutments. Therefore one of the most
important requirements for avoiding cracking is to properly
shape the abutment by excavation or backfilling with
concrete as shown on Figure B-23. Similarly the slopes
of all structures in or through the core must be properly
designed to minimize cracking. If possible such struqres
should be blended into the abutment by shaping and
excavation as required. Another factor to be considered
is that steep or very fllt abutments lead to less probability
of cracking than those of intermediate slope ( De Mello

(1977)). Abutment and structure shapes can be checked


using the finite element method of analysis as described

in Section B7.

The shape and side slopes of the core of the dam


influence the amount of stress reduction that occurs in the
core due to arching between stiffer shells. Finite
element analyses, as discussed in Section B7, indicate that
in order to minimize cracking the core should:

1. Be as wide as practical
2. Be inclined upstream
3. Not have abrupt changes in slope of either face.

(c) Foundation Treatment:

Shaping of the abutment to remove irregularities


that could lead to cracking is discussed in (b) above.
89

A major source of cracking is a compressible


foundation. Where the embankment is to be founded on any
soil the properties of the soil and bedrock profile beneath
the soft foundation must be fully investigated. The best
way of avoiding cracking is to remove the soil beneath the
entire dam and shape the rock to avoid cracking as discussed

above. This is frequently not economically justifiable.


Where the soil is left in place efforts should be made to
preconsolidate it at a time when there is no risk of
cracking the dam. This can be done by preloading or by
constructing the parts of the dam over the deeper parts of
the soft foundation first. Other possible methods include
densification by grouting, compaction, blasting or
vibroflotation. In cases of less pervious soft foundations
it may be desirable to increase the rate of consolidation
by use of drainage such as sand drains.

Another problem with compressible foundations is

the collapse on saturation. Sherard et al (1963) note


that with fine grained soils prewetting by ponding does
not produce any significant settlement due to dead load of
the layer. However, it does allow the settlement to take
place during construction leading to smaller differential
settlements occurring over.a longer time period which is
less severe than all the settlement occurring rapidly on
initial impounding. It appears that for coarser grained
soils this does not necessarily occur because Nobari and
Duncan (.1972) report that ponding on a somewhat coarser
90

foundation material had little effect in reducing collapse


when it saturated.

Throughgoing water passages can also occur due to


erosion of the core into open joints in a rock foundation
or open gravel deposits in an alluvial deposit.
Foundations should always be treated to eliminate such open
areas by grouting and shotcreting in the case of the rock
and use of a suitable filter layer against the open gravel.
Nevertheless if either condition could occur attention
should be paid to the nature of the core material placed on
the foundation and it should preferrably meet the filter
criteria relative to the openings as discussed in Section D.

(d) Properties of Materials:

(1) General.

In this section it is assumed that the core will


be placed using modern construction methods and the
properties of the material checked regularly in a. field
laboratory. Hence the compaction and moisture content
will be in the vicinity of the Proctor maximum density
and optimum moisture content respectively and material
within a given zone will be reasonably uniform.

In determining the best material for a dam core


other factors must be considered as well as cracking. The
main one of these is the shear strength. A material of low
shear strength would resist cracking because of its low
cracking potential but the embankment slopes would have to
91

be made very flat to ensure stability, thus would not be


economical. Hence the low shear strength material can
only be used where it will not interfere with the overall

stability of the embankment.

The best core material for a dam depends on its

location within the core. Where tensile stresses are


expected to be high the material must be crack resistant
and in the main body of the .dam the material must settle

as little as possible.

Tensile stress areas occur primarily at the dam

crest and along the abutments. In these areas the core


material should have low cracking potential and large

tensile failure strains. Fortunately there are areas where


low shear strengths can be tolerated from the viewpoint of

overall embankment stability. Wilson (1972) recommended


that weaker material be used above that level in the dam
where the uncomfined compressive strength of the material
used for the main body of the core is equal to. the weight
of a column of fill above that level. This would ensure
that shear failure occurs at the top of the dam instead of
cracking. (Note: Care must be taken in the way the

strength is reduced. Increased water content may increase


the probability of shrinkage cracks). Gopalakrishnayya
(1973) carried out a finite element analysis in which he
replaced the material in the tensile zone at the abutment

near the crest with more flexible material. He found that


92

the tensile zone was considerably reduced and the maximum


tensile strain was also reduced.

In many cases a zone of more flexible material


is placed against the abutment contact with the core. The
width of this zone is determined by abutment geometry and
placement conditions but is generally less than 1 metre

(Sherard et al (1963)). This zone is intended primarily


as a seal between the core and the abutment; There are
now methods available to determine areas where this zone
could be increased to resist cracking. Finite element
analyses of a longitudinal section through the dam axis,
centre of core or upstream face of the core will give an
indication of the extent of the tensile zones (Covarrubiat
(1969))and these may be replaced with_more flexible
material and, checked by finite element analysis

(Gopalakrishnayya (1973)) The layout so obtained must
also be checked at critical sections by finite element
analyses on transverse sections as a thick layer of more
flexible material underlying the stiffer core material
could lead to stress reductions at the bottom of the core
due to arching (Lofquist (1955) and Laing (1971)).
Sherard (1972) reports that more flexible material was used
on the crest of Sklope Dam in zones where cracking was most

likely. Vutsel et al (1973) used centrifugal testing


to locate zones of potential' cracking and used more flexible
material in these zones.
93

In the body of the dam the core material must


give minimum settlements and minimum arching. Both these
requirements are met by using the least compressible core
material. Londe (1970) presented a plot of the ratio of
maximum crest settlement to crest length against water
content divided by plasticity index for dams that have or
have not cracked. This data'is presented in Figure B-24'.
This plot shows that where the maximum crest settlement is
less than one thousandth of the crest length no cracking
has been observed on the dam crest. This indicates the
importance of minimizing crest settlement.

(2) Selection of Borrow Area:

(a) General:

In most projects the borrow area will be selected


on the basis of economics. However, in some cases there
may be two or three areas that are suitable and this section
discusses the manner in which the best material can be
selected on the basis of crack resistance. This particularly
applies to the material for use in tensile zones as this
will normally be a small volume relative to the total core
material required and therefore more sources may exist.
The requirements for low cracking potential are discussed
in Section B2 and requirements for maximum tensile strain
are shown in Table B-2. The optimum properties for borrow
area materials to reduce cracking are summarized in Table
B-4.
94

(b) Plasticity:

Plasticity is the most difficult property to

assess relative to cracking of the dam core. This is


partially because of inadequate data and partially because

of conflicting requirements. If it is possible to ensure


that the core material will be cohesionless under all
loading conditions then this would be the best core material

as the cracking potential is zero. Such a material would


have zero plasticity index. However, the core material
must have a low permeability and such a material will
probably develop negative pore pressures under capillarity

effects or during undrained loading. Hence the failure


envelope in terms of total stress may have a significant

cohesion intercept. If the material is not substantially


cohesionless under all loading conditions then the core
material to be used in tensile zones must have the maximum
tensile failure strain, and maximum swelling on effective
stress reduction. This material has a high plasticity

index. However, a material with a low, but positive,


plasticity index is the worst material to resist cracking
as it has a significant cracking potential, and no
resistance to cracking due to its deformation characteristics.

In the body of the dam minimum compressibility is


provided by lowest plasticity, and high placement densities.
In addition low plasticity material will generally ensure
higher permeability and greater relief of construction pore

pressures (as discussed in (e) following). Thus low


95

plasticity material should be used in the body of the dam.

The more plastic a material is the more it


shrinks on drying. Therefore it should either not be used
in areas where drying is possible or should be protected
against drying out. This is discussed further in part (3)

(c) Gradation:

Sherard et al (1963) presented a range of


gradations which were considered most susceptible to cracking.
Since 1963 a large number of additional dams have cracked
and the range of gradations is now much larger.

The soils most likely to crack, as discussed


above, are those of small but positive plasticity index.
These are soils with a low but positive clay percentage.

Such soils should be avoided.

Low plasticity material will best resist cracking


if it is well graded for the following reasons:

(1) it will have the lowest compressibility.


(2) it will have minimum shrinkage on saturation (Nobari
and Duncan (1972)).

(3) if a crack does occur it will tend to be self healing


(see Sections C and D).
(4) it provides the best resistance to erosion
(see Sections C and D).
96

(d) Cementation:

As discussed in Section B2 cementation increases


the cohesion intercept and therefore increases the cracking
potential. Differential cementation leads to the
situation where a crack will not heal because the cemented
material forms a bridge over the crack. Hence any form
of cementation should be avoided.

Cementation can occur due to chemical action.


Water that is rich in dissolved salts or dissolves material
from the upstream shells, core or abutments deposits these

salts in the core of the dam.

Knill and Best (1970) analysed the weathered


rhyolite core material used at Shek Pik Dam and suggested
that cracking of the core could have been due to the
temporary cementation resulting from the presence of
goethite (a crystalline variety of ferric oxide hydrate).
Sherard (1972) refers to areas of cemented material, too
stiff to be removed by scrapers, occurring in borrow areas
inspected during his investigations into cracking of dams.
He suggested that cementation effects could have
contributed to ..the cracking. Such effects are particularly
disturbing as the downstream filter zone could also be
affected and the prime safety factor against damage due

to cracking would be eliminated. Hence soils which may


contain cementing agents should be thoroughly investigated
and discarded unless their lack of cementing potential can

be proven.
97

Cementation bonds occur in clays and probably


other soils due to time effects and loading history. Both
Conlon (1966) and Bishop and Garga (1969) investigated
and found .significant bonds in undisturbed clay samples.

(e) Permeability:

Permeability of the dam core is not generally


considered in discussions on crack resistance. However,
most dam cores are more impervious than is necessary to
reduce seepage to acceptable levels and provided the
-7
permeability of the core is less than 10 m/sec it will be
adequately impervious.

Using the most pervious material acceptable for

the core has the following advantages:

(1) construction settlements will occur quickly thus


reducing differential settlements at the top of the
dam and reducing post construction settlements
eiredide Siee-$.5es
(2) negative pore preccurec due to capillarity will be a
minimum hence cracking potential will be low.

(3) on impounding of the reservoir swelling will occur


rapidly thus eliminating negative pore pressures and
reducing cracking potential. At Tarbela Dam the

saturation- of the core occurred almost as quickly as


the filling (core permeability 10-8 m/sec) thus
reducing cracking potential substantially (Truscott

(1977)).
98

(4) maximum drainage will occur during application of load


tending to cause cracking thus forcing cracking to
occur under effective stress conditions and reducing

cracking potential to a minimum.

(5) self sealing by collapse as discussed subsequently


such material will tend to collapse into cracks and
thus be self sealing as discussed in Section C.

(3) Placement Considerations.

:(a) Moisture. Content:

The correct moisture content to avoid'Cracking would


appear to be a subject of some confusion. To illu.strate
this the following three statements are quoted from the

conclusions of Sherard (1972):

"Cracking has been common in dams that were compacted


at an average water content several percentage points

below Standard AASHO optimum. Important cracking has also


developed in a considerable number of dams where the
foundation settlement was not large and the embankments
were well constructed, with water content at or above
Standard AASHO optimum water content".

"In none of the central core dams that were damaged


by piping believed to have originated in cracks was the
core compacted at a low water content".

"Much experience shows that the rigidity of compacted


99

embankments, and thelikelihood of cracking, increases


rapidly with decreasing water content below standard AASHO
.optimum. For this reason, water contents close to modified
AASHO should not be generally used (or specified)".

It is suggested that the probl,em of which moisture


content to use can best be resolved by considering the
following placement conditions:

(a) Tensile Zones - No Drying Out

These are tensile zones below top water level


and those protected from drying out above that level.
these zones material should be placed wet of O.M.C. to
minimize cracking potential by minimizing negative pore
pressures and to optimize deformation characteristics.

(b) Tensile Zones - Drying Out.

The relates to zones on the dam crest. In such


areas the moisture content must be kept low (not more than
0-1l.C) to avoid shrinkage cracks. It would appear that the
use of a cohesionless material with a relatively high
permeability in such locations would present advantages.

(c) Bulk of Dam - High Reservoir Levels

In dams with cores of low permeability 'pore


pressures may be high if the core is placed at optimum
moisture content or above. These pore pressures will
reduce slowly with consolidation. If the dam stores water
for most of the year and is in a high rainfall area the
100

pore _pressures will probably not drop much below the steady

seepage levels for full reservoir level. Total settlements


will occur under effective stresses that are low compared

with those in a fully drained embankment. Therefore


settlements will be less and arching and cracking effects

will be less. However, if there is then a drought or the


operation characteristics of the reservoir are changed so
that the water level is low for long periods then additional
drainage occurs and the core will reduce in volume. This
will result in additional differential settlements, possible
cracking due to shrinkage and reduced stresses due to

arching. Such behaviour was feared in the clay cores of


British dams in 1976 (Appleton (1976)). Such behaviour
is most undesirable because:

(1) Cracking would not be expected after years of


satisfactory dam performance. Reaction to the
emergency might be slow and substantial delays might
occur in investigating the fault and carrying out the

repairs.

(2) differential settlements between the core and shells


or abutments would be larger than otherwise as all
shell settlement would have already occurred and a
substantial amount of settlement of the lower layers
of the core would normally occur before the upper

layers are placed. The increased differential


settlements would lead to additional arching effects.
101

(3) total settlements at the crest will be larger thus


increasing the probability of cracking.

(4) when the drought breaks the reservoir will quite


probably be filled rapidly thus loading the dam

rapidly. Sherard (1972) cites rapid dam loading


as a major contributor to cracking in dams. This is
to be expected as it leads to undrained loading and
high cracking potential and does not allow time for the
core to swell and develop higher compressive stresses
in this way.

(d) Bulk of Dam - Low Reservoir Levels.

Most.dams fall into this category. Unless the


designer can be sure that the pore pressures in the cores of
the dams in category (c) above will remain high it would
be desirable to include them as well.

The basic requirements are that total settlement


be a minimum and that it occurs as quickly as possible.
Total settlement is minimized by placing the material at

the maximum density. This will be achieved at about O.M.C.

Construction settlement will occur quickly if the


construction pore pressures are low and the rate of
consolidation is high. Therefore settlement will occur
more quickly if the fill is placed drier. However, it is
also necessary to ensure that the core does not collapse
on saturation and, as discussed in Section B6 and shown on
102

Table B-2, collapse tends to occur below O.M.C.. It


therefore is considered that the best placement moisture
content will be determined by the properties of each
individual core material. It appears that in most cases
it will be slightly dry of 0.M.'C.

Kulhawy and Gurtowski (1976) recommend that the


core be placed dry of optimum to obtain lower compressibility
and thus less arching during construction. This may be
considered, however, there is a risk that volume reduction
of the core on saturation will eliminate the advantages.
The volume reduction will increase arching effects and
will cause additional settlements of the core. Such
settlements will result in larger differential settlements
than if they occurred during construction of the dam and

may therefore cause cracking.

(b) Degree of Compaction:

Table B-2 shows that in order to obtain the best


deformation characteristics the maximum amount of compaction

should be used. However, it is almost certain that


increasing the amount of compaction also increases the
cracking potential of a soil by increasing its cohesion

intercept.

In tensile zones it is essential that cracking


potential be kept low hence heavy compaction may be counter-
productive. However, such fill is usually compacted by
hand which automatically results in lower densities than
103

obtained elsewhere. (Sherard (1972)). In the body of


the dam settlements and arching . should be minimized and as
much compaction as practical should be carried out.

Sherard (1953) and Marsal (1960) both found that


most of the dams in the Western United States and Mexico
respectively which developed severe cracks were poorly
compacted usually at low moisture contents

(4) Use of Admixtures:

Two types of admixtures are sometimes considered.


Calcium salts may be incorporated to eliminate

dispersion. This is discussed in Sections C and D.

Bentonite or other clay is sometimes mixed with


non plastic material to obtain a more flexible material

for use in tensile zones. This practice should be very


thoroughly investigated before being adopted. As mentioned
in B.12d (2) above a soil with a small amount of plasticity
is the most likely to crack. Hence, the addition of a
small amount of clay will worsen the situation and adding
enough clay to result in significant improvement would
probably not be economical.

(5) Properties of Shell Materials:

The optimum material for the shells should have


approximately the same compressibility as the core and
should not reduce in volume on saturation. Kulhawy and
104

Gurtowski (1976) show that the presence of a relative thin


transition or filter layer between the core and the shells
does not significantly affect the degree of arching.
HenCe a thin zone need not be considered in assessing
arching potential but should the transition or filter zones
be large their properties must also be considered in
assessing the overall properties of the shell.

Where the core and shells are of basically the


same materials 7 that is both clay or.both granular -
it -is easy to achieve the above requirements.

Where the core is stiff relative to the shells


load transfer will increase the stresses in the core and
cracking due to .hydraulic fracture should not be a problem.

Where the core is soft relative to the shells


arching will reduce the stresses in the core thus increasing
the probability of cracking. In such a case a balance
must be struck between increasing the compressibility of the
core and reducing the compressibility of the shells by
reduced compaction without increasing the settlement on
saturation significantly. In many cases shell materials
have larger particles which make testing of compressibility
and volume reduction on saturation impractical in existing
laboratory equipment. In such cases the amount of
compaction for the shell must be based on data available for
other projects where field measurements of compressibility
have been made on similar materials constructed in various
105

ways. Such data is available for the construction period


(e.g. Wilson (1972) and Truscott (1977)) but little is
available for the saturation effects on shells. The best
that can be done therefore is to compact granular shells
using a substantial amount of water.

(e) Construction Procedures:

As mentioned in (c). above parts of the dam on a


compressible foundation should be built before constructing
that part on a more rigid foundation.

Where closure sections are required in embankments


the slope of the fill should be made as flat as possible.
and there .should be no irregularities, such as construction
roads left in place, on the surface. The space left for
construction of the closure section should preferrably be
wide enough for compaction to be carried - out using
machines instead of relying on hand compaction. Sherard
(1972) points out that even properly controlled hand
compaction methods tend to result in a more compressible
fill than is achieved using machine compaction. The
closure section should be constructed as slowly as possible
to enable settlement to occur during construction and thus
reduce differential settlement between the two parts of
the embankment. In critical cases this should be clearly
designated in any contract documents relating to construction
of the dam.
106

Areas of probable cracking, for example the


abutments near the crest, may be placed last to avoid
cracking in these areas. This procedure requires careful
organization as should the delay be too large cracks will
occur at the abutment at the temporary top of the fill
where there would otherwise not be any cracks with the
disadvantages discussed in B3. It is desirable that such
areas be inspected prior to recommencing fill placement
and any cracks be repaired.

An alternative to the above is to overbuild the


crest in areas where cracking is expected. This forces
shear failure to occur in preference to cracking. The
extent of overbuild is thus determined by the shear
strength of the core material.

Care must be taken in grouting, water pressure


testing or drilling using water or mud in the abutments
after the dam is built or in the embankment itself.
Hydraulic pressures developed in such operations can cause
hydraulic fracture of the core and a surface which is less
resistant to future cracking. Exploratory drilling

through the core of the Yards Creek Upper Reservoir Dam


while water was impounded behind the dam almost certainly
was the cause of one of the transverse cracks through that
dam (Sherard (1972)). Abutment grouting can result in
layers of grout penetrating the core through such cracks.
These layers than provide a surface along which cracking
107

can occur and provide a bridge of cemented material below


which a crack can enlarge. This behaviour was observed at
Djatiluhur Dam (Sherard (1972)).

(f) Other Considerations:

Sherard (1972) observed that cracking is much


more likely to occur if the reservoir is filled rapidly.
This is to be expected as the slower the reservoir is
filled the more closely the drained loading condition is
approached and the lower the cracking potential.

Lowe (1970) suggests the use of tensile


reinforcement such as steel or fibreglass to distribute
cracks in areas of possible tensile stress. This may help
in areas of low confining pressures such as the dam crest
but the variation in moduli between fill and reinforcement
would render its use of doubtful value elsewhere.

B.13. GENERAL CONCLUSIONS.

(a) Although there are obvious reasons for


transverse cracks through the cores of some dams there are
also cases where the reason is not obvious. Furthermore
there are some mechanisms by which cracks are formed, e.g.
hydraulic fracture, that have not been properly confirmed
either by mathematical analyses or by laboratory tests.
Until there is a better understanding of cracking mechanisms
prediction of cracks will remain uncertain.
108

(b) Many methods of reducing the probability of


cracking are known and agreed among the profession.
However, there are other areas where there is not agreement
and some cases where changes to improve crack resistance
based on one mechanism would increase the probability of
cracking from other causes.

(c) Although the probability of cracking can


be reduced substantially by the use of suitable design
procedures the state of the art is such that it is not
possible to be absolutely sure that .a transverse crack will
not develop through a dam core.
TEST NUMBER PRESSURE AT WHICH CRACK OPENED PRESSURE AT WHICH CRACK CLOSED REOPENING PRESSURE

METRES WATER KN/m2 METRES WATER KN/m2 METRES WATER KN/m2

COW GREEN
38 1960 - ' 27 1400 32 1650
C1
32 1650 28 1450 28 1450
C2
34 1760 28 1450 31 1600
C3
34 1760 28 1450 4o 2060
C4
_ 38 ' 1960 23 1190 25 1290
c5 '
BALDERHEAD
C22 54.4 2810 32 1650 MMMI.M. •••/..,

C21 31.2 1610 10.9 560 — —


C23 28.5 1470 12.4 650 — —
C24 12.3 64o 6.5 3141 ..._. _.....

Data from Vaughan (1972).

TABLE B-1 RESULTS OF HYDRAULIC FRACTURE TESTS


VALUE REQUIRED HOW TO ACHIEVE REQUIREMENT
MATERIAL PROPERTY
TO REDUCE
CRACKING MOISTURE CONTENT COMPACTIVE EFFORT PLASTICITY

Tensile Failure Strain Maximum Increase (Note 1 ) Increase Increase (Note 2)


Bulk Modulus Undrained Maximum Increase Increase Reduce
In Compression Drained Maximum O.M.C. Increase Reduce
Swelling on Reduction in
Effective Stress Maximum Reduce Unknown Increase

Tensile Areas
Shear Modulus (Note 3) •Minimum Increase Reduce Increase
. Rest of Dam Maximum Reduce Increase Reduce
Volume Decrease on Saturation Minimum O.M.C. or Above. Increase Reduce
Volume Decrease on Drying Minimum Reduce Increase Reduce

Notes (1) Maximum tensile failure strain at O.M.C. or above.


(2)Small increase in plasticity do not appear to affect tensile failure strain.
(3)Tensile areas are adjacent to abutments and crest of dam where gravity loading tends to give low stresses.

TABLE B-2 EFFECT OF DEFORMATION PROPERTIES ON CRACKING


NAME REFERENCE ABUTMENT SHAPING FILL FILTERS CONSTRUCTION METHODS
Cow Green Vaughan et al Desighed to block
(1975) finest floc in
core.
Bennett' Taylor and Shaped to 1 to 1 Wide core of strong material Wide transitions.
Morgan (1970) , Modulus of all fill approx.
equal.
Duncan Gordon and Added 6% bentonite increased 're oa.e• area.
Duguid (1970) P.I. from 4 to 20 %).
Lower Notch Sigvaldason Undercuts filled Shearing promoted at abutment Stress cells install-
and Tawil(1972) in,corners round- contacts. ed and vertical
ed on basis of :, . stresses monitored.
f.e. analysis.
Mica Webster (1970) Shaped at top to Flaired core at abutments, Wide gravel zones Heavily compact core,
avoid downstream upstream inclined core,placed to avoid arching. remove soft found.
widening, dam wet of 0.M.C.. materials.
arched upstream.
Mission Terzaghi and Fill adjacent to sheet pile cut-1 Careful design of Main dam kept off
Lacroix (1964) off carefully zoned. filters. clay zones and placed
on noncohesive zones
with impervious
blanket to clay zones

TABLE B-3 EXAMPLES OF DAMS DESIGNED TO RESIST CRACKING SHEET 1


NAME REFERENCE ABUTMENT SHAPING FILL FILTERS CONSTRUCTION METHODS
Nurek Vutsel et al Studied using Discarded loess because of arch- esigned for Larger freeboard to
(1973) centrifugal ing and erosion, used well-grade racking using provide preloading.
models. rocky clay as self-healing, used odel studies.
more plastic soil at abutments.
'1Outardes k Dumas et al Concrete fill to Thicker core at abutment,consid-.'kicker zones at Reservoir filled
(1970) smooth abutments, ered adding bentonite but too butments. slowly.
dam arched costly,considered Ungitudinal
upstream. trench of bentonite mixture in
centre of core to relieve arching
but not done.
Sklope Nonveiller 'Dam arched Placed clay of high moisture
(1973) upstream. content and increased compaction
in tensile zones (on basis of
beam tests ).

Summerville Bertram (1967) Shaped to 1 to 1


by excavation.

1arbela Abutments smoothed, Well-graded, non-cohesive core Sequence arranged to


densely compacted. preload area of deep
alluvium.

TABLE B-3 EXAMPLES OF DAMS DESIGNED TO RESIST CRACKING ---SHEET 2


NAME REFERENCE ABUTMENT SHAPING FILL FILTERS. CONSTRUCTION METHODS

Talbingo Cooke (1973) Designed upstream


transition for
clogging.
Garrison Lane(1955) Compacted at above 0.M.C. Inclined chimney Flat slopes of 1 ins
drain in area of in closure section.
expected cracking

TABLE B-3 EXAMPLES OF DAMS DESIGNED TO RESIST CRACKING --- SHEET 3


MATERIAL REQUIREMENT REASON AND COMMENTS
PROPERTY
TENSILE ZONES BODY OF DAM
Plasticity Index Nil or High. Nil (1)
No plasticity means no cohesion and no cracking
potential.
(2)High plasticity means high tensile failure
strain, high erosion resistance and swelling
on saturation.
(3)No plasticity gives minimum compressibility.
Gradation Well-graded Well-graded (1)Minimum collapse on saturation.
(2)Maximum resistance to erosion.
. (3)Allows build up of graded filters.
Permeability High High (1)Effective stress condition more likely,
therefore lower cracking potential.
(2)Rapid consolidation, therefore less
differential and post-construction settlements.
Cementation Nil Nil (1)
Cementation increases cracking potential.
(2)
Preferrential cementation may provide roof for
crack and prevent healing.
Solubility Insoluble Insoluble
Dispersion Non-dispersive Non-dispersive (1)Dispersed particles cannot be filtered.

TABLE B-4 SELECTION. OF BORROW AREA:-DESIRABLE PROPERTIES FOR CORE MATERIAL


115

Total Stress Envelopel,..--


Effective Stress
Envelope

Effective Stress
Circle (typ.)

u (constant negative)
Gi‘e3 & 614.e,3 (pore pressure
2
z

(a) EFFECTIVE STRESS AND TOTAL STRESS WITH CONSTANT NEGATIVE


PORE PRESSURE.

Effective Stress Envelope 0,


4. Total Stress
Envelope

\/!0-`Total Stress
Circle (typ.)

Negative Pore
14—C4° "Positive Pore
Pressure Pressuro

(b) UNDRAINED CONDITIONS -- EFFECTIVE AND TOTAL STRESS

FIGURE B-1 MOHR-COULOMB FAILURE ENVELOPES


116

(4)

V
A
Tension t Compression
e+G) e"
c 3

(1) Mohr-Coulomb failure surface (generalized).


(2)Mohr's circle for direct tensile test.
(3)Minimum possible failure surface in tension (any confining pressure
results in shear failure).
(4)Maximum failure surface in tension (an increase in 'A' must result
in shear failure).

FIGURE B-2 FAILURE SURFACES IN TENSION

Mohr-Coulomb Envelope
(generalized) 4c1Shear Failure Envelope
for Lambe Stress Plot
Tensile Shear
Failure Failure

Tensile Failure Envelop. Maximum Tensile


for Lambe Stress Plot' "\f" Failure Circle

Tension Compression
6., ÷
1

FIGURE B-3 FAILURE ENVELOPES FOR LAMBE STRESS PLOT


117

Shear
Failure

F
Tension Compression

FIGURE B-4 STRESS PATHS RESULTING IN CRACKING

,r- Failure Envelope (High c


(Small

Common Shear
Strength
Failure Envelope (Small c
(High iS

45°

FIGURE B-5 EFFECT OF c AND /5 ON CRACKING POTENTIAL FOR A GIVEN SHEAR STRENGTH
118
I I
, , , .
GlyOttAllileb jrRcriVe."7111 Kl71-fit;i7 .
1'

400

RANGE oriocimews
OR . a/aim/vet)

'VORA/1Jc!) ;,y
PqE111;71/5
1:

/00

4rA- _,AoAit 0 en ON 9 66

R 8-6 - EFFECWE _A#D _UAORN/VED !CRAchar .


17C OE
POTZTNT/AL-÷:TaiL/WS7-OUG LANOSZ/DE CLIV

r 1 --

tiNDRA/NL-o SHEilR

/0 .W 30 40
i --'. _,,I.,,,,...,
f--1 -

_1VorEs•—0_)1 _SO/Z. 47-__Odit C


' " . ' CO SHEAR STRENGTH D4TA FROM /A/SLEY AND /1/LLIS (49G5)
(3) UI/DRA///E40 TE /YSILE ST'RE'NGTH CR DA/ CORA L. 4A1Z/SHAMYY4 0973)

- +.) E64-ECT/VE TEA,Sai STRE/1/GTH ASSUMED ZERO, _- 1 '

/76Y/RE: 877 : EFFECT/%E A4/01 UNDRAMED CRI9C111/4/6


t'l :POTE/v7-7,4L.-!" MICA DAV CORE MOTE /AL
11- ,
-• -t rd -
• t
"7-1:

] 1 LL1 .

PTIA141.11
CoNrcar.
/519
:
d5 ..

: ZOCAr/oN o
"-- Po/An
(7L y/6.)

41111AXMIYAl /1)55/6LE OdYCKIA:14


r1r1L- -a1062 EFFTC7714.' I

0 50 1: /00

' .

BALDERHEAD CLAY GAULT CLAY

Data from Ajaz and Parry (1975).

FIGURE B-8 EFFECT OF WATER CONTENT ON CRACKING POTENTIAL BALDERHEAD AND GAULT CLAYS
. , t
120
t

r
_.:._OPTAWN 1/0/STURE — ------4-----
CO,/TENT
200 ' 204 '200
4-2 I - INCREASING _ /VCREAS///6
INCREASING
//10/S Tt/R4_;7 _
1............2___ /1/40/577/RE: . liOISTeiRe:
f
'CovrENT: .1:22
--r :Catre.,ir ' —4'J.-. /00
1. - i- 1.---/00 CONTENT
100 ,
i
.7--='' 5:3Z,--- _Son.:8, . ."-
_ ca.-A7:) Z ;L•
----,-f- ---r
f c
" -'= I441
C, ....1
::_1

200 20o

INCREAS/NG 410.15-77/RE
CONTENT -
00 '
1300
livuericsiac, itto/srwz
COVrENT
200 ' --

/00

Fa/RE 67-3 Erfrcer OF 144TER CONTEIVT O/1. CRACIOV6'


POTE/1/77AL Z24TA 1.4:54).0/1_NARA/MacalWA71:497.0)

(/NO/STUR 6E0

,, 2 ' •.
MOTES' -0-1)47.A--Ftecill8/.57/Crie-11/15-6.14Gn-096

9)
. i ..L., A '
!..1 ALL . S Ti e 6 ss ZS . . EPA-ECM/E. -
/ ,
_ ' 4. . /S. 4 01/D 0,./ 1 .: 4q01, Y. . _.
). bill 7E/1/4

FIGURE B-10 tnECT OF CEMENTATION BONDS ON CRACKING POTENTIAL


- .

Cmy PR --
(1630imen-e)
1:- 7Z

.8z/v ram rg-


4717'a

1 2 ---- L-
Notes: 1) Data from Tschebotarioff et al (1953)..
2) All materials at same Liquidity Index.
3) Tests were undrained.

FIGURE B-11 EF.PECT OF PLASTICITY INDEX ON CRACKING POTENTIAL -CONSTANT L.I.




••••• eN.
•. .4.- . 1 . . . .1 i !II 1444 4i4 -ree . , , . I. 4 1 4 , .i , ; 1 i , 1 i : .., I., ;

.- 1 ' — t • , '' ' ' ' — 1 ! • 1 i ! '


i I 4.1 4 ! 4..
' .
, , : '
I 1 ' 'H , HI'
.:;1 I Ht
. .
... . . . . . ' ' '
. . •• • • • • • . •
....... . .. . . .
.. _........ ..._. . . . .
2: .:. 16 • S P.T= 4 7
, . . OIL1 . . , • .
et
— : : _L.__ . 4
. t........,.... ... ..7. .......... ... . : ._ .... .0 .__ .Soii...F pi „ 757. . ,. .. . .
7
, . . r
.......... .. . .. ... . i + -
... . ..... ..:,....e , . .... . . . '. . .. . 0
3072., z9 ,P.I=. 8-87 : •
...
..... ....._..... .._• _ . . ...
:
v. . • . . ._,.,.. .
.-,-;.,,.; ... . ... , •
—. .4 4 .

• . . . . ' na r = 4-.. :71 ....
. . • ..._40. __,...)on.,_ .„.„.1.-...A. • ul , ...-Aa
ftco
•d — - • _. .. . L.
PI.-- /4; 41_ ..L._.i _

___ ... • ° SiOli .0 ..•• PI;r


• T .. .20 cr ..
, . --,
. .
. •
1 ,_ ----f-- LI
•-,
.,_
• . 1_1_ 1_
'- . . • '

. • '
... •
.
• - ,. - -.. , ,•4-,L-..-__•,_•_,.
-;_ =_._.4;--._•_•__ • ,_i__

. _.. _._.
.. _
---,--,------ ,
—. .. ,
.
-r- '

---1-•÷.------.--4.--H-"-•--,, _..._
..- -•
—....___.... . _ ... ..
.
•0±,-1-. --___-_-25.o.'...--i--
...... .__.
-,- . r SO - -7
, -.- .- /00-7 * -- . :- --/50 . - .
, . L 1... .4 ..
r . 1 ' l
• • .._...
. _. 1.-.. , .
• ' 1- , , - 1 . : : , , , r . : . • . ' G4 IL ,
•,..,1-4
., , . . I . .
' : '-- ' • •...._,_' [: HA/Pt . ..,
. , . • ,_ i 71, i , • . • , i • , ;. _12 I i----- •7-1- .-r-
. , . •
• "
- -7-7-7-f- : •
'-±
. Mori 5.71) &TA - Aeo,44W47qiirAt --._4 /./ a ....11-Awar 4.177?). I--
/I r /1' ,
..-i--- —2•) 4L.Le P011,175 ' '- : 4 r___. .v.prtiyao itifors.r
6-ze.cr4a.:,' . qRc .J...,01).rewri
LI:-..
._ ,
__' ---4-.......- peak, ' - E/G41/..'r - ,. • 1i • • • -, •i• . .. • ..t,- • • :... . • .-
• . t 1 ! ,•• t
iFH) .Verrir r/OAT....:z. .
.
747 ..._.e.--/Tor7-r4K -Eitir a. __ ACC0c1A17
,. . , : 1 - - .: .i : :•:. , j •---: •
,.... Lc,.. ._-',7-e- Sr $7.117.Cit,0 PnIlle•D' - -.. -- v-T" . 7-:- . ...•
1-
.._i-. . . _Li . 71 • . . :' -.
. .- !.. , is —

1CURE 40-12- ___17/1/Dg.X.• .011(...L. _


. 7'4E17/Ter -01-PLAST/C/TY
__;_____,_
. . _-4----- .474"—i-- .-;--. CRACK/A/C - POTENT/4L --.CONST/ INT a iw.c:::-.
.. .._..,
. . . . .
: ., • i . ' I .
„._,.._ ,.,_._ _
. . „. "
,
• . ., ..._.,_. ._.. .___
...
.. _. ._ . .

_1 • • • , . i ...t.•. . ...... ..... ...


.. _ ..
...•..... , ...... .....:.. . _ .
_ . . . : ..
. .. .• •;• : _ 4__ ___
• .. .
. . .... . . . . . . _._. .. , , ... _ .. ........
.... .. _.: ...:
... 4. ....... .._ .. ..
. ... .. . . . .... .. . ... ... . _. ...
. . .. _ • • - • • .
- •• . ".- • • • _ ..
_ . : .. ....... .._ ...
I. ... ..... ....... ... ._...... . . . ..
i • ; .... .... ..._..
..
...
... .. , •.• - :I! •-• • - :: • : 1 - . • • •, ...
• •: • . , .
-i- -
4 '1 • '
.
• • , • • i ' • • :—...
—4.
123

WIDE CRACIC

sktgacR4cx
REPAIRED CIZACX
PAUSE BUT
No CRAG
NO PAUSE \
coNsmuc7no

FIGURE B-13 EFFECTIVE STRESS PATHS - CRACKS DUE TO PAUSE IN CONSTRUCTION



124

VAR/47/ail/ /1/ VAR/AT/ON ITUA/CT/ON OF


ALIVV/I/M DEPTH ALLYV/viti pRoley-Rngs ALLUMMAAID ROCK
DIFFER LAI774L FOUNDATON SET 77 ElEl EN T

RAPID VARIA770/1/ IN PROTECT/OA/8 ON CONCRETE


F0 NO LEVEL Asamiz-NTs STRUCTURES
IRRECULAHT/ES /A/ R/C/D FON/DAT/OA/


ARCIIING OF CORE' D/A-FERE/1/CE //V CORE CLOSURE
BETH/EP/ Sli'LLS 47-EXIL RPOPERTYS SECT/ON

F/CUREB-4. CAUSES OF TRANSVERSE CRACIeS - NO


RESERVO/R LOA zO/NG.
125

SATURATION OF FOUNDATION NORZONTAL CRACKS

UPSTREAM MOVEMENT
OF CREST

SATURATION OF roYNDATIN - VERTICAL CRACIeS

UPSTREAM MOyemEN I OF
TOP OF CORE

SATURATION OF UPSTREAM SNELL—DOWNSTREAM CRACKS

F/CURE 8-15 CAUSES OF TRANSVERSE CRACKS —


RESERVOIR LOADINC
126

//y/ K.LE
AO

AOF-7 K. AE..
3 3

13; SOIL PROPERTIES


SOIL beforefailure
pore pressure /
failure
/
E 3 / CV-
E f101)1
73 =sn0
cr+ 01;
Ay A (17-1' A —
Cr"-1-111 —
v Ef

V=0.3
2 A ko;‘\
total stress . -ffective stress

E':: K (swelling)
E'rz co (no swelling)
rr4

u to cause
0-54
cracking
2 2

Data from Vaughan (1976).

FIGURE B-16 DRAINED HYDRAULIC FRACTURE - SIMPLE ELASTIC ANALOGY


127

/56 FLEvATIOA/ /A/ FEET (rYI?)


UPSTREAM'
M2G S CELLS 2/ rots
1)OWNSTREA/V
/29e
CELLS C ro /0
/1/0!

(hos TREAisf CORE DOWNSTREAM' 11ELL.


SHEZZ.

LOC/1778N PLAN

CELLS 2/ To 26"
RAA/Ca- Or FA/L1 E
LAVEs
/0
/
i

PRO TECTloN OF STATss


PATH DIE To [575 I/WAX

6,-,',../32)
-..__________0. 7 -..17-o I
CELLS C ro /0
20
RA/(/GE or FA/ LURE
LINES 7
-VI'
1 iti
/
V / I (CIWO
/0 seiRilic coilisTge/cRov.
- x PIG, Of caysTReicriai!
4 8Efofg 4.974crIZZ/N6.
STRESS PROTECTION OF STRESS 0 8giapr /97517azva.
PATH . PATH DUE TO 1975 F/LLI AS. 0 RAI. REsSRvorx - 497g-
- TOTAL STR Ess Es
I - — EFFECrivE 57,fess(s.
0 SO
/0 20 30 40 60
e,

• FIGURE g-17 TARBEZA ,0114, - STRESS PATHS


850 850
CA.FiTTTFTpipitil
ric$114
RESERVOIR EL. 8/6'
800 800
n i.,,,,,,r, 0 -burs ',......1“,,i.11:..•,:,:.?....
1...

750 \ _,,i44,1,::A4s"
,....0
50
.41,,,w,...4 //1 ,:',.511C'tif_ k 7
N;.-44\j...7111.:
..,,..4 ? 1 Ct.,'n

ORIGINAL GROUND .'1'4

700
:
.'
/It' SURFACE •r al

•- 700
:ill.
:' ;\‘‘,

_,,;,, i•vi..
EXCAVATED .,.. ,,,,...
4...
p,..,.
,
.iiiit.,W,iNc..ifi
.- ,11',• =11CW
/ -h
DIVERSIO •
BEDROCK SURFACE
.:,:.):.).
TUNNEL
650 .4., ,..,.., .14.31k 650
ORIGINAL -"6. i '..),Z .
co Lu •
BEDROCK SURFACE---- e.• la' --GLACIAL TILL AND TALUS 1.1.3
c‘, w 600 600 u..
1_i ' ----"77Pt.i ,,.... STRATIFIED SILTS
--"-7-fiP. AND FINE SANDS
ELEVATION IN

550 550 z
22
0.
500 :1,-;:,
,?. - AQUIFER" ZONE
1,p 500 8
:°Q64.16
. , k-.7.v
-0. (SILT, SAND, GRAVEL AND BOULDERS)
450 ..,:-
/ 450

400 1 400 w

F/CURE 8-18 LOWER NOTCH B,44( - SECT/ON ALO/VC DA/if AX/S

711/Z.((X972)
DATA /'"ROM SAW/110430N AND 7:

129

TOP OF DAM - 830'

80 GAUGED STRESSES CORRESPOND TO


RECORDED VALUES AT ABOUT ONE MONTH
AFTER COMPLETION OF IMPOUNDING

+4\ •
. A
70 ) + EFFECTIVE WEIGHT OF OVERBURDEN
GORGE SHAPE 1 ---,: °
BUOYANT WEIGHT = 87.5 POUNDS/CU. FT.
(COMPUTED STRESSEI

1 .,
if, 22 27 2021 23 GAUGES 20, 21,22,23- NEAR UPSTREAM
i I I I I LIMIT OF CORE
L. • k 4.
LLI • 27-NEAR DOWNSTREAM
14-kt
. LI- STRESSES .LIMIT OF CORE
•k. OBTAINED FROM FIELD
- MEASUREMENTS
-....
(TYPICAL)
60
0 • e b
%-..
• 4
Q V—GORGE SHAPE 11
+44. (COMPUTED STRESSES)
L. 13 _• 4-4. 12 1614 18 17
;.....4 lui.0 1 J I II i I GAUGES 9,10-NEAR UPSTREAM
I •
• , LIMIT OF CORE
c 12,13,14 -CENTRE OF CORE
c 1+ 16,17,18 -NEAR DOWNSTREAM
+ + LIMIT OF CORE
500 o y +11+
111

o 0 + 4.
GAUGES 0,1 NEAR UPSTREAM LIMIT OF CORE
I 6 2 7 0 2 - IN CENTRE OF CORE
t Hi 5„ 6,7,8 - NEAR DOWNSTREAM LIMIT OF CORE
• / -1- 8
BOTTOM OF GORGE - 430'

400
10,000 20,000 30,000
VERTICAL PRESSURE POUNDS/SQ. FOOT

F/CURE 11-1.9 LO//V&? MOTC1/01,19- Cal/PAR/SON Of


VERT/CAL STRESSES ON DAV CENTREL/NE

'fizoo SictMIDfisoiv Anetik-ai 072)


130

DAM CREST"?

FxrE„,,,0A, oc/E To /M;vemswr 0AJA-ntizr

On1/11114 FACTURE

Ca)OPEN/A/C OF FRACTURES ASSOC/ATED WITH FAULT/,VC

CATST

47, TO Alati(iewr ON glIJLT


AL
I /BEFORE VEMEA/L?
i61.reR Milicia-arq

/541LT

h) v4VEA/1gAIT ON FAULT /NCL /NED TO VERT/CAL.


(

Exremsiom Deis- TO llOVE MENT 041 /271/47:

(c)410VEMEA/T OA/ FAULT OBLIQUE TO DAM AX/S

F/CURE 8-20 EXTENSIONS CAUSED BY MOVE/JENKS'


FAU Lis
131

Original Dam Location


on ‘
Transiti ittivik
Location After Movement
On Fault
Upstream Downstream
Reduced Width of
Filter Zone
duced
dth
gillii1111111„(

_
Vertical Movement-
Fault On Fault

(a)Vertical Fault Displacement

Downstream
orizontal Movement On Fault
Fault f
Shell
Reduced Width
of Filter
Ait

Reduced
Core width Core
Reduced Width
of Transition

Shell

Upstream

(b)Horizontal Section Through Dam

FIGURE B-21 REDUCED ZONE WIDTHS DUE TO DISPLACEMENTS ON FAULTS


132

Shells more compressible than core Collapse of upstream shell on


saturation

Reser,::;ii1 1111,11116,

Collapse of foundation soils on Downstream deflection of core


saturation under reservoir loading

Spreading of a soft foundation

FIGURE B-22 CAUSES OF LONGITUDINAL CRACKS


133

AbutMent Abutment

Locate dam upstream


of place where valley
starts widening on the
downstream side.

Upstream
Note: Locating the dam as shown above presumes downstream movement of
tore under reservoir loading. In many cases the core moves upstream
on initial impounding. For further discussion see text.

(a) Plan View

Dam Crest

Transition by fill and excavation

Original Abutment Profile

Abutment Rock
-y,....Remove ledges
Core Contact
Backfill overhangs with concrete

Preferrably remove
alluvium otherwise preload
/
%Noma flow
'.... ••••• MP..

(b) Longitudinal Section

FIGURE B-23 GEOMETRIC CONSIDERATIONS IN AVOIDING CRACKING


134

0 Miguel Hidalgo
Ip

0 Los Lajas '

0 Estreito

Sarre - Poncon

0 Sonolono Mihoro
0Solis 0
0Shek Pik 0 Cougar
Hyttejuvet
OGuodalupe 0 Lnis Lion
Victoria era
Netzahualtoyalt
A.R.Cortines::
to Miguel Aleman
°Ewan 0 0 Gepotsch
P . Ale ma
n& Et InternIllo
B -Ware z 0 •i
A. Lopez
Djatiluhur

10-* lo''
(Settlement)
Tossernent
Longueur
( Length )
0 Barrage non fissuri trartivirsalament
(No transverse crack)
0 Barrage tissuri fronsversalement
(Transverse crack)

I/CURE 8-24 DATA FROM CRACI(ED AND al/CRACKED


DAA1S (LOA/E (/970))
135

SECTION C

THE BEHAVIOUR OF CRACKS

C.1. STABILITY
A crack will stay open as long as the pressure
on the face of the crack is equal to or greater than the
total stress in the soil at the face of the crack and
provided the stress parallel to the face of the crack is
not high enough to cause shear failure. For a dry crack
this means that the crack will remain open as long as the
stress normal to the face of the crack is zero and stress
parallel to the face is below the unconfined compressive
.strength with the existing negative pore pressure. For
a crack filled with water under pressure this means that
the crack will remain open as long as the water pressure
is equal to or greater than the stress normal to the face
of the crack and the unconfined effective strength is' not

exceeded in the soil around the crack.

The formation of a crack results in stress

redistribution around the crack. Hence once formed a


crack is unlikely to close up if the stress conditions are
returned to those existing immediately prior to cracking.
The stress redistribution is due to a combination of the
effect of tensile strength and the effect of strain
redistribution. The magnitude can be quite large,
Hydraulic fracture tests in the core at Balderhead Dam
136

(Vaughan (1972)) showed that under undrained conditions


the close up pressure of a crack could be up to 22 metres
of water less than the initiation pressure (refer Table
B-1). At Hyttejuvet Dam (Kjaernsli and Torblaa (1968))
seepage through the core increased rapidly at reservoir
level 738 metres and didn't reduce again until the
reservoir level reduced to 716 metres. In this case it is
probable that erosion of the crack was a major contributor
to the difference.

Dry cracks will only close up if the stresses


across them are increased by increased loading, (e.g.
increasing the height of the embankment).. However, they
can also be sealed by construction procedures such as
backfilling with slurry, grout or soil.

A crack full of flowing water will only close up


if the rate of closure is greater than the rate of erosion.
Closure can occur due to the following reasons:
(a) Increases in load as discussed above
(b) Shear failure due to reduced effective stresses
(c) Swelling on saturation
(d) Reduced reservoir level below that required to maintain
crack.
(e) Infilling with material from other zones
(f) Sealing by construction procedures.
137

C.2 SIZE OF CRACKS.

The width of a crack depends on the deformation


characteristics of the core material, the stresses developed
in the embankment and the number of cracks. It is not
possible to estimate crack widths accurately. analytically
and they can only be estimated on the basis of experience.
The depth of a crack at the dam crest depends on the
unconfined compressive strength of the core material.
If the core is dry this can be quite large due to negative
pore pressure effects.

Sherard (1972) refers to transverse cracks on


the crests of small dams in the United States commonly
being 1/8 to 3/4 inch wide. In a few extreme cases
cracks as wide. as 6 inches at the crest have developed.
Cracks have occasionally extended completely through the
dams into the foundation to a depth of 10 feet or more.
Sherard (1972) also discusses many higher dams. The
maximum width of transverse crack on the crest he noted
was 3 inches at the Virginia Ranch Dam which reduced to
zero at a depth of 9 feet. The most spectacular
longitudinal crack Sherard states he has seen was on the
downstream slope of El Isiro Dam. This crack was 90 metres
long and between 40 and 60 cm wide. However, it was only
2 metres deep.

Cracks that occur below water level are difficult


to inspect and their size is therefore difficult to asertain.
138

Some idea may be gained from cracks formed in El Isiro Dam


during abutment grouting (Sherard (1972)). One crack
Was 100 mm wide but the majority were only a few millimetres
wide with a number in the range 20 to 30 mm. Hence it
appears that although such cracks can be quite wide locally,
the openings will be small on the average anddt provided no
erosion occurs, will not by themselves endanger the structure.

C.3. BEHAVIOUR OF CRACKS ON IMPOUNDING.

(a) Cohesionless Materials:

A dry crack may exist in a cohesionless soil due


to negative pore pressure effects. In theory .a crack can
be formed in a low permeability cohesionless soil due to
undrained hydraulic fracture although this possibility is
considered remote. When water flows through such cracks
negative pore pressures are eliminated and the crack
collapses and seals itself. As the upstream face becomes
wetted first a seal may form at this location. This was
observed at Round Butte Dam where no water was found in a
crack in the core that extended from the crest to below
water level. Apparently the non-cohesive core material
collapsed and formed a seal at the upstream end of the
crack when it was wetted (Sherard (1972)). The collapse
of material into a crack may leave a void into which other
material collapses. The hole stapes upwards until a sink
hole appears on the dam surface. Sink holes were observed
139

in the crest of Messaure Dam above a fault zone in the


foundation (100 m.below). Probing indicated the
disturbed material was at least 20 metres deep and it was
believed that it extended to the base of the dam, presumably
to fill holes associated with the fault zone (Nilsson (1964)).
Hence a crack in a cohesionless soil will normally heal
itself and not present any risk to the safety of the dam.

It should be noted that a material can only be


regarded as non cohesive if it retains its non-cohesive
characteristics under all conditions existing in the
embankment. Conditions under which apparently non-
cohesive materials can develop cohesion are discussed in

Section D.

(b) Cohesive Materials:

For the purpose of this discussion a cohesive


soil is considered to be one that can sustain a crack when
flooded and with pore pressure equilibrium near its walls.

The collapse mechanism in the crack determines


if the dam will be safe when a crack occurs. In a
homogeneous embankment the crack must seal itself to ensure
safety. In an embankment with a chimney of filter
material enough of the core material must be retained on
the filter to ensure that the crack seals itself.

Where the material is erosion resistant and the


flow through the crack relatively small the clay may swell
more rapidly than erosion increases the size of the crack.
140

In this case the crack will heal itself. As low erodibility


and high swelling are associated with higher clay contents,
soils likely to seal by swelling normally have a high
plasticity index. They will also tend to be placed at
relatively low moisture contents (see Table B-2).
Exceptions to'this are the dispersive soils. These are
discussed in part C4 which follows..

As water flows through a crack the soil becomes


softened and is likely to be eroded from the sides of the
crack. The particles are then subject to gravity and
viscous drag of the water. The heavier particles will
tend to sink to the bottom and smaller particles to be
carried along by the water, that is segregation will tend
to occur. At Balderhead Dam samples from drill holes
showed that segregation had occurred at cracks (Vaughan
et al (1970)). In time the sides of the, crack may
become unstable, either due to relief of negative pore
pressures by swelling or because erosion has exposed a large
unsupported face and chunks of core material may fall into
the crack. Excavations along eroded cracks in the Tarbela
Blanket showed large pieces of unchanged blanket material
in a maitrix of the coarse remains of the blanket material.
(See Figure C-1). The collapse of large chunks into cracks
leads to sudden changes in flows through the cracks. Such
changes are usually reported for cases where dams have
cracked. Examples are Balderhead and Hyttejuvet Dams
'(Sherard (1972)).
141

As discussed above the finer material may be


removed from a crack in cohesive material,and the crack
tends to increase in size. As the zones adjacent to
the core are cohesionless they will tend to collapse or be
washed into this crack as erosion occurs. Hence stoping
and formation of surface sink holes will occur in these
zones if there is enough erosion. Sink holes due to this
behaviour were observed atMatahina Balderhead and Yards
Creek Dams(Sherard (1972)). Sink, holes are more likely
to occur under these conditions than when the core is
non-cohesive because a crack in a non-cohesive core should
seal before significant erosion occurs.

(c) Homogeneous Embankments:

In a homogeneous embankment there is no downstream


filter and particles in suspension will be washed out of
the dam. Unless erosion can be stopped quickly the crack
will enlarge and the dam fail. Cracks may heal in a
homogeneous embankment by the following mechanisms:
(1) If the core material is non-cohesive the crack will
seal when the material collapses into it on wetting.
Hence the dam will be safe provided erosion of the
non-cohesive material does not occur at the downstream
face. This is usually ensured by providing a more
pervious downstream toe so that exit gradients will be
low and not capable of eroding the material.
142

(2) If the flow is slow, the crack small and the core
material erosion resistant and capable of swelling then
the crack may seal by swelling.

(3) As discussed in (b) above the sides of a crack will


usually become unstable either by relief of negative
pore pressures or by erosion and a large chunk of
material may fall into the crack. If this chunk is
large enough to block the crack or resist erosion
long enough for a graded filter to build up on the
upstream side the crack will be sealed.

If the flow through the crack is slow only the fine


material will be washed out and the - material remaining
may bulk and seal the crack. This will only occur if
the embankment is densely placed and consists of well
graded material in which there is only just enough
fine material to fill the voids. This is not usually
the case as it is generally necessary to use material
in which the fine fraction controls the density to
ensure adequately low permeability.

The only mechanism that can be relied on to


prevent failure is the collapse of non.cohesive soil
((1) above). Hence unless the non-cohesive nature of the
soil can be guaranteed cracks should be protected against
erosion by the use of an internal filter. Terzaghi (1953)
noted that this was the way in which to eliminate the
It
problem. (See Section A). Sherard (1972) stated in his
143

conclusions,- "It is believed that even low, homogeneous


dams should include at least a thin vertical chimney of
filter material as a "crack stopper" except in those cases
where the consequences of a failure are not. great':

(d) Zoned Embankments:

A zoned embankment is one that contains an


internal filter that extends the full height of the dam.
The filter material will beron-cohesive and collapse on
wetting thus providing a zone on which the material eroded
from the core will be retained.
The internal filter will prevent failure of the
embankment if it retains all material coarser than that
below which bulking will occur. As discussed in (c) above,
this will usually mean retaining almost all the core
material.

Another way in which the internal filter can


prevent failure is to provide a surface on which a graded
filter can buildup. In this mechanism successively finer
layers build up on the downstream filter until the finest
particles in the core are retained. The exact mechanism
is not clear. No examples have been found of either
laboratory investigations into this matter or positive
reports of it occurring in the field. It would appear more
likely to occur if the velocity of water in the crack is
high enough to carry the coarsest particle to the downstream
filter. However, even in this case the heavier particles
144

will fall and tend to arrive at the filter near the bottom

of the crack. Hence finer particles may still be washed


through the filter over the top of them. Thus the
mechanism by which graded filters are built up is complicated
and, unless a large percentage of the core material is
retained, does not appear to be very reliable.
The presence of an internal crack will usually be
indicated by an increase in seepage and the seepage water
becoming muddy. If healing occurs the seepage water will
become clear and the seepage reduce to about its original

value. As discussed in (b) above, erosion can lead to loss


of material from the non-cohesive zones adjacent to the
core and the formation of a sink hole on the surface. As
the upper parts of the embankment are not saturated negative
pore pressure effects may delay the stoping action and the
sink hole may not appear on the dam surface until after
the crack is healed and the embankment is safe. Hence
such a sink hole should not be a reason for undue concern.
Thorough investigations should be carried out, however, it
is believed that in the majority of cases the only repair
required will be to fill in the hole.

C.4. DISPERSIVE SOILS:

(a) General:

It has recently been recognized that in some areas


of the world problems with embankment dams have been due to
145

piping of dispersive clays. The first papers were from


Australia (summarized in Aitchison et al(1964) and
Aitchison and Wood (1965)) and additional areas were
identified in Israel (Kassiff and Henkin (1967)) and the

U.S.A. (Sherard et al (1972b)). Although the identification


of these soils has been studied in detail by Sherard et. al
(1976a and 1976b) it is still not possible to be completely
sure whether a soil is dispersive or not. Although
failure has occurred in several dams constructed of
dispersive clays there have been many other dams constructed
of the same soil under apparently the same conditions which

have not failed.

Usually clays occur in nature in the flocculated


state, that is the individual particles are combined
together to form flocs. These flocs are about the size

of fine to medium silt. However, in some cases chemical


effects result in the flocs breaking down into individual

particles. Such a soil is known as a deflocculated or


dispersed soil.. The individual particles are very small.
Because of their negligible interparticle attractive forces
they are easily eroded and their small mass enables them
to be transported in water moving very slowly. As they
are very small they can pass through very small openings
and in a poorly compacted dam core it appears that they can
pass through the macro pore spaces even if no crack is

present.
146

Dispersion of clay occurs due to the mineralogy


of the clay and chemistry of the pore water. Dispersive
soils therefore tend to occur in specific areas. Sherard
et al (1972b) have identified areas in the U.S.A. where
such soils occur. Preparation of similar maps for other
countries would appear desirable. No areas are known to
exist in the U.K., but a more thorough review is required
to be sure.

A locality in which dispersive soils occur can


be identified by the presence of certain peculiarities in
the landscape. Ingles and Wood (1963) point out that
such areas are characterized by notable gully erosion,
field tunnelling and excessive turbidity of reservoir
water. Sherard et al (1972b) note that such soils form
spectacular erosion gulleys. They also note they have
been known to soil scientists for years as "nonsaline-
alkali" soils. They cause problems in agriculture either
because the particles form a dense impermeable layer
making irrigation and drainage difficult or, where a more
open deposit exists below the surface, sink holes and
erosion tunnels are formed as the soil is washed away.

(b) Causes of dispersion:

Dispersion of a clay occurs when the repulsive


forces (electrical surface forces) between individual clay
particles exceed the attractive forces (Van der Waals forces).
The governing factor is the quantity of monovalent cations
147

(primarily sodium) relative to the quantity of higher


valence cations (primarily calcium and magnesium) in
the exchange complex (the pore water). The sodium acts
to increase the thickness of the diffused double water
layer surrounding the individual clay particles and so
decrease the attractive forces between particles.

Soils can be dispersed in two ways. The clay


can exist in a chemical state such that it is susceptible
to dispersion but has not dispersed as it is dry. When
placed in a dam seepage water saturates the core and
dispersion takes place with the resulting problems.
Secondly the soil may be in a stable flocculated state
as placed in the dam but the water stored in the reservoir
is of the chemistry required to cause dispersion. In
such a case the soil will disperse when the seepage water
changes the pore water chemistry to that required to cause
dispersion.

The effect of clay mineralogy on the dispersive


behaviour of soil is not fully understood. Where
investigated it has been found that all soils with dispersive
characteristics contain a significant amount of
montmorillonite. Aitchison and Wood (1965) believe that
montmorillonite must be present for dispersion to occur.
They investigated several small dams that had failed and
found that the soil from no dams fell into the chemical
state where montmorillonite would be flocculated (see
148

Figure C-2). A recent report by Olson (1974) could help


explain the behaviour. He compared various properties of
kaolinite,illite and montmorillonite and found that for
kaolinite there is no reduction in effective strength on
dispersion, for illite there is a slight reduction and for
montmorillonite the effective strength reduces to 50 percent

of the flocculated strength. In addition the particles


of kaolinite are relatively large, those of illite
intermediate and of montmorillonite the smallest.
Furthermore data from Aitchison and Wood (1965) (Figure C-2)
shows that montmorillonite is dispersed for a larger range
of pore water chemistries than illite. Hence a soil with
significant amounts of montmorillonite will be more likely
to disperse, will provide smaller particles and suffers
a considerable loss of strength on dispersion. Under
these circumstances it would be expected that soils with
a significant proportion of montmorillonite would account
for the majority of cases where problems occur due to
dispersion.

A point worth noting is the loss of effective


strength of montmorillonite on dispersion. This could
be significant in stability determinations, where the
reservoir water causes dispersion of the soil.

(c) Tests for Dispersive Soils:

Tests consist either of determining the chemical


state of the soil and comparing the quantity of monovalent
149

cations with the total number of cations and

relating this to.a dispersion boundary or physical tests


Which check if dispersion occurs either by particle
measurement or determination of erosion resistance of

the soil. Table C-1 summarizes these tests.

The problem with all tests is that they do not


distinguish between the two causes of dispersion. The
chemical tests and pinhole test refer to the
condition of the soil. Tests in which distilled water
or chemical solutions other than reservoir water are used
do not relate directly to the problem but probably provide

a conservative answer. Tests in which a large amount


of reservoir water is used relate to the condition where
the seepage water has flushed the natural pore water from

the voids.

(d) Identification of a Dispersive Soil:

It is still not possible to be absolutely sure


that a soil will disperse or not, however, the latest
papers by Sherard et al (1976 a and b) help considerably.

In their papers Sherard et al (1976 a and b) have


described tests to determine if a soil is naturally in a

dispersive condition. They do this by the combination of


physical testing (the pinhole test) and chemical testing.
In general the pinhole test will define whether the clay is
dispersive or not but chemical tests should also be carried

out. They have defined zones in which soil may be


150

dispersive, doubtful or non dispersive. Figure C-3 shows


the boundaries between these zones.

Tests to determine whether dispersion will occur


sometime in the future when the porewater is replaced
with reservoir water are not so clear. The best test to
date appears to be the hydrometer test described by Vaughan
et al (1975) which determines the critical grain size for
filter design. Another simple method would appear to be
to flush the soil in reservoir water until the existing
pore water is replaced and then carry out the pinhole test.

(e) Examples of Problems with Dispersive Soils:

In order to illustrate the real problems that can


occur due to dispersion of soils and possible remedial
,measures the following examples are given:
(1) From Aitchison and Wood (1965)

The authors studied the behaviour of twenty small


dams in Australia. The following examples are selected
as being of particular interest:

At Lakes Entrance a dam which had stored ground-


water relatively high in salt content had only slight
leakage for several years. Sea water was then stored with
no problems. However, when the sea water was replaced
by river water low in dissolved salts the dam failed within
3 days. Thus the introduction of water low in dissolved
salts lead to dispersion and failure in a short period. The
151

change in chemical, state is shown on Figure C-2.

Two dams constructed of soil in an initially


flocculated state were used to store saline bore water.
As seepage took place the pore water was replaced with
water high in sodium cations. Dispersion of the soil
occurred and the dam failed by piping. Repair with local
soil was ineffective but repair with calcium rich soil

has been successful.

Nambour Dam had not suffered any distress even


though the clay should have dispersed. The authors
attributed the safety of the dam to the fact that the core
was placed wet of optimum moisture content and has low
permeability. Therefore the water can only seep into the
core very slowly and the clay has time to swell and reduce
the possibility of erosion.

(2) From Ingles Lang and Richards (1968)

This paper describes the repair work on a 50 feet


high rockfill dam which had failed by piping through the

core. The soil was found to be highly dispersive and this


caused the failure. In order to overcome future dispersion
problems the core of the repaired dam was placed so that
it had a very low permeability and the advance of seepage
through the core would be very slow. Piezometric measure-
ments and chemical analyses of water in the core show that
the phreatic surface is advancing very slowly and it will
152

take at least 15 years for the seepage water to reach the


downstream face of the core. The authors concluded that
swelling of the core material prevails over the dispersive
effect of the replacement water thus leading to a
condition of long term stability in the core.

(3) From James and Wickham (1970)

This paper describes remedial measures carried


out on Morwell Dam. This is a 30 feet high homogeneous
dam that developed a pipe on first filling. As part of
the remedial measures a 1.inch thick layer of gypsum
(calcium sulphate) was spread on the upstream slope of the
dam. This layer was intended to ensure a large percentage
of bivalent calcium cations in the seepage water and so
ensure no dispersion until swelling of the core had occurred.

(4) Speedie (1970)

The discussor mentioned a case where 1 to 2


percent of lime was mixed with the soil used in the down-
stream 20 feet of the core of the dam in order to prevent
dispersion in the zone adjacent to the filter and thus
ensure that the filter would be able to retain the core
material.

(5Y Sherard et al (1972b)

The authors investigated a large number of dams


which had failed in the U.S.A.and the Zulia River Flood
Control Dyke in Venezuela.
153

In most of the dams investigated rainfall erosion


on the downstream face of these dams consisted of vertical
tunnels into the fill joined to horizontal pipes exiting
downstream. These tunnels occurred even where the slope
was protected by thick layers of grass. The authors
considered that these tunnels were formed by rainwater
entering desiccation cracks on the surfacel percolating
downwards 6 to 10 feet and then flowing out horizontally.
The rainwater would be very pure thus tending to accentuate
dispersion problems.. The width of the cracks increased
by erosion more rapidly than they could close by swelling
and the result was these peculiar erosion tunnels.

On the Zulia River Dyke the erosion was very


variable despite no apparent variation in soil properties
as determined by standard classification tests. Chemical
tests conducted from various locations showed that the
soil was very variable in its dispersive characteristics.

(f). Protective Measures:

Dams constructed of dispersive soils can be


protected against failure by one of the following procedures:
(1) Use of a downstream filter layer that will retain the
smallest possible particle when deflocculation occurs.
-This is generally not possible using conventional
filter criteria as the percentage of material less than
the number 200 sieve is greater than would usually be

tolerated. Further discussion is included in Section D


Design of Filters
154

(2) Use of a calcium rich admixture such as lime or


gypsum in all or part of the core or in an upstream
zone to ensure the water seeping through the core is
rich in bivalent ions. This method appears
satisfactory provided there is an adequate supply of
calcium ions to last the life of the projector until
sufficient swelling of the core material occurs to
prevent dispersion.

(3) Placement of the core in such an impervious state that


the front of new seepage water advances very slowly
and the core can swell more quickly than the water
causes dispersion or erosion. This system has
aparently been used successfully on several dams in
Australia (as described in (e) above) and Aitchison
and Wood (1965) propose that the maximum permeability
that should be permitted for a core of dispersive clay
is 10-5 cm/sec. This is based on a very simplified
calculation and must therefore be regarded as a high
value. It is also not clear what happens when the
reservoir water reaches the downstream face of the
core and dispersion of the soil immediately adjacent
to the filter occurs. There is no apparent reason
why this should not erode and over a period of years
form a pipe back through the core. This matter
requires additional investigation.
155

In describing the three methods of protecting


the core of dispersive soil,no consideration has been so
far given to the effect of a crack through the core. Where
the filter is designed to retain the finest possible particle
there is no problem provided the filter is non-cohesive
and the crack cannot project through this zone. Where
calcium is added to the entire core there should not be
any problem in the'core resisting dispersion as the
particles should build up on the downstream filter and
the flow through the crack be reduced quite quickly.
Where part of the core is treated with calcium the situation
is not so clear and part of the soil may be lost before
the crack is healed. " Where the calcium is placed in an
upstream zone the water may pass through this zone too
quickly to dissolve a significant amount of the calcium
salt with the result that the core will disperse and
failure occur. Where low permeability is relied upon for
resistance to dispersion a crack immediately destroys the
low permeability effect and failure can occur. An example
of this is almost certainly the Mashabei Sade Dam in

Israel (Kassiff and Henkin (1967)). This dam was


rebuilt twice with very good compaction control but failed
both times. However, in both cases failure occurred by
piping along an outlet structure which probably caused a
local tensile zone and hence a transverse crack.

In summary protection against failure of a dam


with a core likely to be dispersive can apparently best be
156

obtained by use of a suitable downstream filter (as


described in. Section D) or by the use of admixtures in the
core that will ensure that the soil remains in a
flocculated state for the life of the project.
TYPE OF TEST BRIEF DESCRIPTION REFERENCE VARIATION DISCUSSION

Chemical Quantity and type of cations Aitchison and Developed relationship Plotted boundaries between
Analysis in soil are measured. Wood (1965) - between sodium absorption flocculated and deflocculated
Relationships developed (1) ratio (S.A.R)•and_concent. illite and montmorillonite
between sodium ions and of salts in reservoir water, (Fig. C-2). Showed soil from
, S.A.R.= , (Na) „ all failed dams plotted in
total ions.
40.5(Ca-1-1-4- Mg-)
zone of defloc. montmorill.
Kassiff and Plotted S.A.R.against concent Much higher concent. than
Henkin (1967) of salts in pore water sol. others. Modified above curve
(2) obtained by soaking a given to fit results. As sol. is
quantity of soil in a given arbitrary differences are to
quantity of water. be expected.
Sherard et al Plotted percent sodium Plotted failed dam data on
(1972b) against total soluble salts curve and found they fell.
(3) in the saturation extract, into a small area. Agreed
% N a+= Na with (1) above; disagreed
. Na++Mg" +Ca44.4.K+
with (2). No comparison with
Saturation extract is a
reservoir water.
given amount of water mixed
with soluble salts in soil.
Sherard et al As (3) above, see Fig.C-3. Related results to pinhole
(1976b) test. Modified (1) above at

(4) high and low ends of range.

TABLE C-1 . TESTS FOR DISPERSION IN SOILS --- SHEET 1


TYPE OF TEST BRIEF DESCRIPTION REFERENCE VARIATION DISCUSSION

Crumb Test Air dry crumbs of soil are Railings (1966) Sol.: 1m.eq. sodium . Obtained positive results
placed in a beaker of sol.. (5) hydroxide. for soil from 25 out of 26
If, after 1 hour,a colloidal dams that failed. (4) above
cloud of dispersed clay found no correlation between
exists then soil is this test and dispersive
considered dispersive. soils.
Speedie (1970) Sol.: Reservoir water Should check effect of
:
(6) reservoir water. No comments
by (4) above.
Sherard et al So14: Distilled water If soil disperses in test
(1976b) will disperse in field, if
(7) no dispersion in test may or
may not disperse in field.
Pinhole Pass distilled water Sherard et al - (4) above showed results
Test through a 1mm. diameter hole (1976a) agreed with chemical tests
in compacted soil. If water (8) and failed dam data.
colours and hole erodes then
soil is considered
dispersive.

TABLE C-1 TESTS FOR DISPERSION IN SOILS --- SHEET 2


TYPE OF TEST BRIEF DESCRIPTION REFERENCE VARIATION DISCUSSION

Grain Size Carry out two hydrometer Sherard et al Standard test on minus No.100 .(4) above found reasonable
Distributions analyses, one with and one (1972b) material, compute % less - agreement with pinhole test.
without dispersant. • Soil than 0.005mm.(a). Dispersion>50% Soil
Comparison of clay percent. Conservation Repeat test with no mech. dispersive,
is measure of dispersion. Service (S.C.S.) breakdown or dispersant, Dispersion <30% Soil
(9) compute % less than , nondispersive,
0.005mm.(b). 540>Dispersioi>30%
a
Dispersion-v. - Soil intermediate.
b
Vaughan et al Design filter using minimum Effect of high Nal- soil,and
..,
(1975) particle size obtained in high Ca++ reservoir unknown.
(10) hydrometer test using
reservoir water without
dispersant.

TABLE C-1 TESTS FOR DISPERSION IN SOILS --- SHEET 3.


160

Blanket. Surface

Sink Hole
Silt

Pocket of B2(1) + Silt


Pockets

Washed B2(1) + Silt


Pockets
, . 0 • 6
-- , •,• , • •
as
0% f).•
1% ,4 • • •,„7:4 '0 -
' 0 • ,s 1 a"
. *4'4 *, „

• 0' B2(1)
0 .0.,
.„
. - • •••e;•. -11.0 '''''..' • (Original Blanket)
0 •
.... _ ,- 4

o

20= ( ' ,, ; - 4 4 • _ ” ,
. o u . -:,.•
• -
0 ..r.' ' 4' D s.P1 —0 ' ':%:' t)..:* Ops
-0 • t';'° . • a . 0 ,. ti - •
.a. ''' ' . Ds, as. , a .4 • 4* 0
..
•a.' * * I, • •
-4 .-. 1. (.1 • • .ri '
Pocket of 04) d.. f• •
.
kf, '',. 0 ' ) 0 ' ''//'' ' a-
, 4) . 0. ,
2 0.
1 .,
... 0
. t I S •

••,-•,•0 .'0
. ...s ti).
° .
0 0 02
• • O .° °••
.•0
„. . 0 * c; ‘0 "
*., '0 • • '. 6 %
0
..,j
• 0 4,, .0 Natural Filter
-... .=:. °I a P • oc• ....
• t'1.2:...42.. , ...—
( 0 '4.4„
3o - iREG Open-work Gandaf
Gandaf Silt Coated RBG +Washed B2(1)1 Silt Coated
)
33 RBG Open-work, Gandaf
Silt Coated

FIGURE C-1 -RESULT OF EROSION OF WATER PASSAGE IN TARBELA DAM BLANKET


3.61

120

o rated
100 . • U..1
0 //A LAKES ENTRANCE DAII" SOUND
oc 20
z Deflocculated
I/B LAKES Ei-t/TRaNCE DAV - FIVZI .
0
60 0 21
cc 11A
11B la

0 IA
in .1

PecsOCCutatEd
1.142/M,Jkl.1.0 NI TS DArog "Raw A/TGIMSON AND tt/ODD•51
v. 0 It .

coip- 0.
-20
to
d stT • IA A ! 2A
Flocculated
tcac"E
0
103 . 10' • 10'
TOTAL CATION CONCENTRATION me / l

FIGURE C-2 BOUNDARIES BETWEEN FLOCCULATED AND DEFLOCCULATED ILLITE AND


MONTMORILLONITE

Legend:
...Percent Sodium e.1-1P ze.41(.1n, 001sperslue in P;nhole Test
(oil measured in nutliequivolents per liter • Non—dispersive in Pinhole Test
of soturation extract)
1 1 '• • l O t lrttlI

•06,
efA O CT) co
01
0 0 0,/ 06P
:
7ona A 0
Previousie
° proposed 0
0 0• criteria
f50

-.--0—asi
. ---
•—.---G;----6.- . o 0-' O

,--
. ,--- Zone B •

• • • , ,!,
• • • . •
• 0
•• •• • ••
, •• • •I , • •
• 0° • %Q% i4:6 4 •
/„.„t . . ,I,„,[ , ,i.•.,1
1 02 . 0.5 1.0 50 10 50 100 200 390

Total Dissolved Salts in Saturation Estract in Milliequivalents per Liter


(TOS • CosMo+IVa.K)

ZONE 4 - ALL $01I,6 ISPERS/Ve


ZONE a- ALAeosr ALL SO/LS //0414/5.04-gs/v4-
ZON6 C' RANGe AA7041 NSPERSIVC TO NOND/SAXRS/VE

1247-4 FRO,/ SHERille0 C71" 41 (ien 6),

FIGURE C-3 BOUNDARIES BETWEEN FLOCCULATED AND DEFLOCCULATED SOIL


162

SECTION D

DESIGN OF DOWNSTREAM FILTERS AND DRAINS

D.1. GENERAL:

This section deals with the filter and drainage


zones located downstream of the core in a zoned embankment.
When cracking occurs in the core of an embankment dam the
only way to guarantee the safety of the structure is by
the correct performance of these zones. Thus filter
design is of the greatest importance.
The geometry of zones in various embankments is
shown in Figure D-l. The filter system consists of one
or more zones of granular material with the finest zone
adjacent to the core. An arrangement commonly used is a
finer zone adjacent to the core to prevent migration of
fines from the core and a more free draining zone downstream
of this to drain seepage from the filter layer with no
significant build up of pressure.

D.2. STABILITY OF THE DOWNSTREAM SHELL OF THE DAM:

(a) Strength of Materials:

As the filter chimney and blanket could form a


failure surface through the downstream part of the dam they
must have sufficient strength to ensure stability.
Generally this will not be a problem as the well placed
163

granular materials will exhibit higher shear strengths


than the core and even adjacent shell materials. The
only time a problem is likely to arise is when the filter
blanket consists of reasonably uniform sand and is under-
water for a considerable part of the time. Under these
conditions liquefaction under seismic loading is possible
and must be considered'.• Special compaction requirements
or modification to the drainage may be required to eliminate
this possibility.

(b) Drainage Capacity of the Filter System:

The drainage system must be capable of ensuring


that flows through the core do not result in excessive water
pressures which could cause instability of the downstream
shell. The amount of water that seeps through a properly
compacted core will be small and readily handled by any
practical thickness of drain. The problem therefore is to
ensure adequate drainage in the event of cracks. Typical
values of flows that have been measured at other embankments
are shown on Table D-1. The filter and drain should be able
to handle such flows without overloading the downstream shell.
The capacity required can be calculated from the permeability
of the materials and suitable flow - nets. The zones should
have a substantial reserve to allow for possible deterigration
with time.
The width of the chimney and thickness of the
blanket filter or drain will be influenced by many factors
164

such as construction, tolerances, construction methods, possible


allowances for earthquake movements and variation in material
properties. The width is therefore usually chosen on the
basis of previous experience and may be arbitrarily thickened
up in possible problem areas. However, the dimensions
of all such zones must be such that adequate drainage is
available in the event of a crack occurring.

(c) High Water Pressures at the Downstream


Face of the Core:

With a properly designed filter system a crack


will seal itself by building up successively finer layers
of material on the upstream face of the downstream filter
zone. Hence in a relatively short time the flow will be
reduced to a negligible amount and the full reservoir
pressure will be applied virtually at the downstream face
of the core. If a number of cracks develop the seepage
pressure conditions for which the stability of the embank-
ment is normally calculated would be changed and, in the
limit, the stability of the downstream shell should be
checked for the case of a seepage pressure equivalent to
full reservoir pressure upstream of the face of the filter.
Generally the stability of the downstream shell is based
on the steepest slope for the material in the downstream
shell and not on internal stability hence this will not be
critical. However, where the core has a reasonably flat
downstream slope or where the dam is founded on a weak
165

foundation it could lead to the selection of a flatter


section. Where modifications to the section are indi-
cated by such calculations the probability of cracking and
frequency of cracking should be carefully reviewed and if,
for part of the length of the dam or below a certain level
in the dam cracking is not probable then the increase in
section may not be required in these areas.

D.3. GRADATION OF FILTERS:

(a) General:

Filter zones must ensure that no significant


damage occurs.to the core of an embankment dam whether or.
not a crack.occurs through the core. In order to do this
the'filter zone must fulfil the following requirements:

(1) It must be sufficiently fine to prevent


migration of fines from the core when there is no crack
and to prevent enlargement of any crack by erosion.

(2) It must be coarse enough to be free-draining


relative to the base (core) material to ensure relief of
excess water pressure or (if the filter zone daylights on
the surface) to ensure exit gradients are low enough to
prevent erosion of fines from the filter.

(3) It should be as coarse as possible to


ensure that the number of filter zones in an embankment
are kept to a minimum.
166

(4) The amounts of fine material should be


limited and the fine material must contain no clay to
ensure that the filter is cohesionless. This matter is
discussed in detail in Part 4 of this section.

Where a filter system consists of several layers


of successively coarser materials each zone must meet
the filter requirement with respect to the adjacent zone.
Filter requirements have, in the main, been
derived from laboratory tests on cohesionless materials of
maximum grain size about one inch. Commonly used rules
and their sources are summarized in Table D-2. Although
these appear to be similar there are significant differences
which illustrates the problem of correct yet economical
design of a filter system. It is •therefore strongly
recommended that for any large embankment dam filter design
should be checked and the gradation modified as required by
laboratory testing. Such testing should investigate both
the condition of no cracking and of a crack occurring.
Laboratory testing of filters for important projects is a
common recommendation (:e.g. U.S. Corps of Engineers (see
Table D72.), Terzaghi and Peck (1967), Sherard et al (1963)).

(b) Gradation of Base Material:

The amount and nature of fines in the base material


affects the cohesionless nature of the base material. This
matter is discussed fully in Part 4 of this section.
The U.S. Corps of Engineers (1953) compiled data
167

on filter tests from many sources. They related various


parameters to the uniformity coefficient (Cu) of the base
soil and found that the best relationship existed between
Cu and the ratio between the fifteen percent sizes of the
filter and base (PD1sE ). as shown on Figure D-2. Above a
153
Cu of four the boundary between stable and unstable con-
D
ditions was constant at a 15f ratio of about forty and
D15b
below a Cu of four the ratio at which conditions became
unstable reduced rapidly. Hence when the base soil is
of a relatively uniform grain size the filter must be
finer.
In many dams where cracking and subsequent erosion
has occurred the grading of the core material is very flat
(e.g. Balderhead Cu : 60, Hyttejuret C 140, Yards
u
Greek : 1000 from Sherard (1972)). The flattest grading
studied by the U.S. Corps of Engineers (1953) had a
Cu of thirty. No work is known in which filters for
materials of Cu larger than thirty have been investigated.
The erosion problems cited above indicate that in some
cases filters for core materials of higher Cu have not per-
formed satisfactorily even though they have been designed
on the basis of commonly accepted filter criteria.
One method of reducing the problem is to design
the filter to retain that part of the core material below
a certain size. The U.S.B.R. and Sherard et al (see
Table D-2.) specify that only material below the No. 4
sieve and one inch size respectively should be considered

168

when designing filters.


Another method sometimes discussed (e.g. De
Mello (1977)) is that if the core material can be split
into two gradations which do not meet the filter criteria
relative to each other, then the filter should be designed
to retain the finer gradation. This method may also be
used to design filters for gap graded core materials.
No references have been found in which this method is dis-
cussed in detail or has been studied either theoretically
or experimentally. However, in the cases cited above the
core material can be split into two gradings where the
1SE 1SE
ratio (see Table D-2.) exceeds five and the D855
D855
ratio between the filter and finer fraction also exceeds
five which does not meet the majority of the filter
criteria (refer Table D-2.)
Vaughan et al (1975), suggest that this problem
can be overcome by designing the filter to retain the
finest particle in the core using laboratory tests. This
matter relates particularly to core materials containing
a significant quantity of clay and is discussed in detail
in Part 3(d) following.

(c) Filters for Cohesionless Core Materials:

A cohesionless core material is one which, when


wetted, remains cohesionless at all times. Factors
affecting the cohesionless nature of a soil,are discussed
in Part 4 of this section.
169

In a cohesionless core a crack will not remain


open when flooded, hence the filter is only required to
prevent migration of fines from the core under seepage
conditions. Laboratory tests have been used to obtain
simple relationships between the gradations of the core
and filter materials which can be used to design a filter
to retain the core material. Initial work was carried
out by Terzaghi (1922) and has been supplemented by Bertram
(1940), U.S.C.E. (1953), Karpoff (1955) and Davidenkoff
(1955). Commonly used rules are shown on Table D-2.
In designing a suitable filter the effects of
core gradation discussed in Part (b) should be taken into
account.
The ability of a filter to retain the core material
depends on its pore size. This, in turn, depends on the
particle shape of the filter grains with spherical grains
giving small pores and angular grains larger pores. The
only filter criteria to take this into account is that
proposed by the U.S.B.R. (see Table D-2.). It is therefore
considered that where angular material is to be used for
the filter the design should be conservative, and, preferably,
based on laboratory tests.

(d) Filters for Cohesive, Flocculated Soils:

Where cracking has not occurred, the filter is


requiked to prevent seepage through the core eroding
particles from the downstream face of the core. In a
170

cohesive material. an arch is formed between adjacent


particles, of the downstream filter and the maximum,spacing
of the grains is determined bythe maximum distance over
which this arch can form. Davidenkoff (1955) and Kassiff
et al (1965) investigated this matter both theoretically
and experimentally. They showed that for a clay core
this distance was much larger than would be obtained in
a normally used filter gradation. This confirmed the view
held by many engineers that piping of a clay core is not a
problem because erosion does not occur. In such a case the
filter.gradation will generally be determined by the
requirements to protect a cracked core.

In.a core of slightly cohesive material, properties


will vary and pockets of non-cohesive material may be
located adjacent to the filter. In this case a stable arch
will not. form in the core material due to cohesion. - There-

fore the filter should be designed to ensure that no erosion


occurs due to seepage, as described in (c) above.
Where the core of a dam is cohesive, or may be-
come cohesive in the future (See Part 4): the downstream
filter must be designed to ensure that erosion of the core
will not occur if the core cracks. This matter• is not
generally considered and therefore needs additional study.
As discussed in Section C3 the filter must re-
tain enough of the core material to ensure bulking and
sealing of the crack. In practice this usually means

retaining the majority of the core material and the


171

conservative approach is to design the filter to retain


the finest particle in the core. Such an approach was
adopted for the Cow Green Embankment (Vaughan, Lovenbury
and'Horswill (1975)).
The minimum grain size for the core material was
determined by means of a hydrometer analysis of the soil
using reservoir water instead of the usual distilled water
plus deflocculating agent. This determined the minimum
grain size that was likely to occur in practice. From
this curve (Figure D-3.) it is apparent that the finest
material to be retained is that in the: fine silt range
(clay flocs.). This material was separated from the rest
by sedimentation. The filter gradation was designed using
laboratory tests in which the clay flocs were suspended in
water and passed through various gradations of filter material.
The coarsest one to retain the flocs was selected as the
filter for the embankment. It is interesting that the
filter material which was washed to remove all material
smaller than the number. 200 sieve did not retain the silt
but that the unwashed material with about five percent
passing the number 200 sieve worked satisfactorily.
It should be noted that the above method does not
ensure that the core material will not disperse under its
initial chemical constitution and tests outlined in Section
C4 are needed to determine this.
Vaughan (1976) notes from his investigations
that there appears to be a simple, linear relationship
172

between the permeability of a filter and the minimum particle


size it will retain. This is shown as Figure D-4. Should
further investigations indicate that this relationship is
valid it could simplify filter design. The minimum design
particle would be determined as described above, the design
permeability read off the chart and a soil with this per-
meability selected. Care would still be needed to ensure
the filter was non-cohesive..

(e) Dispersive Core Materials:

This is an area where there are still a consider-


able number of unknowns. AdditiOnal work is urgently
required.
Where chemicals are used to ensure that the core
material will always be flocculated, as discussed in
Section C4, the filter should be designed using the criteria
for cohesive, flocculated soils as presented in qd) above.
As mentioned in (d) above work by Davidenkoff
(1955) and Kassiff et al (1965) has shown that cohesive
soils form an arch at the core filter interface that is
large compared with the maximum pore size of a normal filter
material. Olson (1974) carried out drained shear tests
on clay minerals including both calcium and sodium mont-
morillonite. He showed that the shear strength of the
sodium montmorillonite was about one half that of the cal-
cium compound. In both cases he has drawn a curved
envelope passing through the origin, however, a straight
line relationship would result in cohesive intercepts for
both montmorillonites. Hence even when dispersed there
appears to be a positive cohesion intercept for the most
susceptible clay mineral. Under these conditions a large
enough arch may form in a dispersed soil to span the
maximum pore size in a normal filter material.
Aichison and Wood (1965) carried out a simple
analysis relating the erosion resistance (due to particle
weight) to the exit velocity of the seepage water from the
core and showed that provided the permeability of the core
was less than 10-5 cm/sec erosion was unlikely to occur.
This analysis appears to disregard the effects of gravity
on a particle at the edge of the-core. However, several
Australian dams which had failed were rebuilt more densely
and have performed satisfactorily to date.
Despite the above evidence there appears to be
no theoretical reason why, in-the limit, repulsive- forces
should not develop between dispersed particles. If this
is the case then arching will not occur and the soil will
erode through the filter. The only way to ensure satis-
factory filter operation is to carry out laboratory tests
using the same core material, filter material and all
variations of reservoir water that will be used in the em-
bankment. No examples have been found where this has been
done. Additional work is required on this subject.
When there is a crack through the core the con-
servative design procedure is to ensure that the filter
can retain the smallest particle in the core (see Part (d)
174

above). However, with a deflocculated soil this means


retaining an individual clay particle which is not possible
using a non-cohesive filter material.
It is possible that a well graded core material
containing dispersed clay could build up a graded filter
on the design filter that eventually retained the individual
clay particles. As noted in Section C3(d) this mechanism
appears rather unreliable unless a large proportion of the
core material is retained by the filter. In dispersed
soils it is even less likely to work as the individual clay
particles will be eroded in very slow moving water and, even
with .a very fine design filter, a significant amount may
be lost before the graded filter starts to build up.
Also it is doubtful if a well graded core material could
form a satisfactory filter for individual clay particles
in suspension as there quite probably will be a large gap
in the gradation curve from the individual clay size to the
next smallest particle. Hence .a cohesionless_filter is
unlikely to control erosion of the dispersed clay and, unless
the amount of dispersed clay in the core is very small,
should not be relied upon to ensure the safety of the dam.
In summary

(1) A non-cohesive filter cannot guarantee


sealing of cracks in a core of dispersed clay.
(2) As discussed in Section C4 the alternative

of using a high density, low permeability core material


assumes no cracking and thus provides no protection against
175

erosion of cracks and subsequent failure.


(3) If a dispersive soil must be used, the dam
core must be stabilized by the use of a chemical containing
multivalent ions, such as calcium. Conservative practice
could be to treat the entire core material but safety will
probably be ensured by having a thick, treated zone at the
downstream side of the core.
(f) Range of Filter Gradations:

In determining the range of gradation to be


specified for a filter material consideration must be given
to the means used to obtain it. If it is intended to be
borrowed direct from a natural borrow area then the range
must be such that there is an adequate amount of material
available with a reasonable reserve. If the material is
to be manufactured the range must be wide enough for it to
be able to be obtained from reasonable operations at

reasonable cost.
In theory a filter should be designed so that the
coarsest material in that gradation will satisfy the filter
criteria laid out in (c), (d) and (e) relative to the core
material and the finest material should satisfy the free
draining criteria laid out in Table D.2. In practice these
criteria will usually be conservative. Hence the best
method is to carry out suitable tests in which the finest
core material is matched against the coarsest filter gradation.
As discussed in (d) above it is considered that the only
conservative way in which to design a filter to ensure
176

sealing of cracks in a cohesive core is by suitable tests.

D.4. FACTORS AFFECTING COHESIONLESS NATURE OF SOILS:

(a) General:

As mentioned previously the filter material must


remain cohesionless at all times so that it will collapse
on saturation, not be able to sustain a crack and ensure
that any crack projected through the core st,cps at the
filter zone.
When the filter is not saturated it may exhibit
some cohesion because of negative pore pressures due to
capillary action. When the filter is saturated these
will disappear and the soil become non-cohesive. However,
there will be a boundary above which saturation may not
occur and in which arching may therefore develop. Should
the saturated filter reduce in volume a gap could occur
below the arch through which water and core material could
flow. In order to ensure that this does not occur the
filter should be placed at a density and water content such
that it does not reduce in volume on saturation.
In most cases it is apparent if a soil is cohesive or
non-cohesive. However, there appears to have been little
work done in defining the conditions under which a non-
cohesive soil acquires cohesion. The fine filter of a dam
may approach the limit at which a soil retains its non-
cohesive characteristics, especially if it is designed to
177

prevent erosion of a crack as described in. Part 3(d).


Hence, it is considered that further work on this subject
is urgently required.

(b) Testing:

The existing methods of measuring the strength


and plasticity of soils are not sensitive enough to detect
the presence of the small amount of cohesion which could
allow a crack to remain open in a filter material. There-
fore it is necessary to use new tests. It is considered
that tests are required for the following conditions:

(1) Short Term Behaviour:


Vaughan (1976) suggested a test in which a cone
of dry. material (i.e. a "sand castle") is formed in a
tray of water. If the soil is cohesionless it collapses
into the water, if not it retains its form. This test
is very simple and disregards the effects of compaction and
consolidation under embankment loads. However, it could
provide the basis for a test in which these factors are
taken into account. It would appear that this.test is
the least that should be done to ensure a filter is non-
cohesive.
Manta (1975) discusses two tests for checking
if fine particles are cohesive or not. He notes that
these tests must be carried out in water containing the
same salts in the same concentration as would occur in
practice. In both tests the method is to split the
178

material into groups of uniform size particles and test


each size separately. By comparing the results it should
be possible to determine the size at which the soil b
comes cohesive.
The first consists of sedimenting the fraction
into still water and measuring the packing coefficient
(ratio:.of volume of solids to total volume). The higher
the value the less cohesive the material. Mantz obtained
a value of 0.50 for 16.5 micron particles and considered
these cohesionless.
The other method, which appears to have been
preferred by Mantz, is to measure the angle of repose of
the material. The smaller the angle of repose the less
cohesive the material. The value obtained is very depen-
dent on the test method. Mantz sedimented the material
into water in a 70 cc cylinder of 2.5 cm inside diameter.
The cylinder was stoppered to exclude air rotated through
90o and the angle of repose measured. The following results
were obtained for flaky particles using a slightly
different method. It was considered that the 25.5 micron
particle was cohesive but above 37.5 microns the material
was non-cohesive in the water used during the test.
179

PARTICLE SIZE
ANGLE OF REPOSE
(MICRONS).

25.5 65 ± 5°
37.5. 38 ± 3o
±
46 39 30
56 42 ± 4o
68 41 4°
79 42 + 40

(2) Long Term. Behaviour:


As mentioned in Section B 2(d) there appears
to be a possibility of secondary consolidation leading
to the development of a cohesion intercept. Suitable
long term tests are required to investigate this possi-
bility, in particular at the low stresses which are
necessary for cracking.

(3) Chemical Effects:


A non-cohesive material may acquire cohesion by
cementation. Chemical analyses of all water that may
pass through the filter zone i.e. from the reservoir,
seepage through the core and abutment seepage. A thorough
investigation is then required to check if interaction
between the water and the filter could lead to cementation
of the filter.
Chemical dffects can also cause weathering or
180

breakdown of the filter material. The fines resulting


.from such weathering should be small in quantity and non-
cohesive in nature. Standard weathering tests are avail-
able to check the quantity lost. However, it is not
usual to check the nature of the fines. In some cases
special tests will be required to investigate weathering
due to special conditions existing at the site.

(c) Fines Content:


One cause of cohesion in filter materials may be
the presence of excessive fines of the wrong nature.
Fines are generally considered to be material passing the
No. 200 sieve and may result from the following causes:
(1) Excessive Fines in the Borrow Area:
Sherard et al (1963) note that various washing
methods have been devised to remove excess fines from
borrow material.
(2) Breakdown during Placement:
Sherard et al (1963) note that on some jobs it
has been necessary to adopt special methods of placing
and compaction to avoid creating too many fines and
that the grousers on the treads of tractors, in par-
ticular, exert very high unit pressures which crush
soft or brittle filter materials.
(3) Washing of Fines into the Filter:
Sherard et al (1963) note that this occurred
during construction of Briones Dam necessitating
removal of a large amount of filter material.
181

In the final embankment the filter should be


completely enclosed with suitable materials to prevent
fines being washed in.. Surface or abutment drainage
should not be lead into the filter for the same reason.
(4) Weathering:
Material used in filter zones should be durable
and resist breakdown by weathering.

There appears to be considerable disagreement as


to the quantity of fines that can be tolerated in a filter
zone. lerFaghi and Peck (1967) note that it should contain
no silt as even a slight admixture of silt can increase
cohesion greatly. Common practice is to limit the quantity
of material smaller than the No. 200 sieve to five percent
(e.g. U.S.B.R. and Sherard et al in Table D.2.). Vaughan
et al (1975) found that a minimum of five percent was neces-
sary to ensure no erosion of a crack in the Crow Green
Embankment and allowed up to ten percent passing the No. 200
sieve. Sherard (1972) reports that a crack at Round Butte
Dam sealed at the upstream face due to collapse of the well
graded, cohesionless silty sand core material on saturation.
Thirty to forty percent of this material passed the No. 200
sieve. It therefore appears that the amount of fines that
can be tolerated depends very much on the nature of the fines.
Common practice is to specify that the fines
should be cohesionless (e.g. Sherard et al (1963)). This
is usually taken to mean that the filter material should
contain no clay or organic material. Mantz (1975) notes
182

that with small particles the forces acting between the


particles are the mutually attractive London-Van der Waals
force a force between electrical double layers which form
at each solid surface and a steric repulsive force due to
the solution of water at the mineral surface which is usually
neglected. In the case of dispersion of solids which have
the same charge (such as silica grains in a water of normal
pH) the second force is usually of repelling nature. If
this is of such a magnitude as to overcome the cohesive
London-Van der Waals force the solids behave as cohesionless.
The repelling action depends on:

(1) the diameter of the particles


(increases with increasing diameter)
(2) the particle shape of the grains
(3) the mineralogy of the particles
(4) the spacing of the particles
(5) the concentration of the electrolyte
(6) the types of salts in the electrolyte

Hence the development of cohesive bonds is com-


plicated and depends on'.'a large number of variables. As
shown in Part (b) above cohesion can be present in particles
of size 25 microns (i.e. a coarse silt). Hence the require-
ment that the soil be cohesionless does not only mean removing
clay and organic material.
183

(d) Formation of Bonds:

Bonds may be formed between soil particles for


a variety of reasons as follows:

(1) The material itself may contain a bonding


agent. Knill and Best (1970) attribute cementation of the
core of Shek Pik Dam to the presence of Goethite in the
core material.
(2) Water seeping through the soil may deposit
salts which cement the particles together. Water seeping
into the filter will come through either the core or the
abutment. Problems are most likely if the abutment contains
soluble material such as limestone or iron salts. If
possible, abutment seepage should be prevented from entering
the filter in such. cases.
(3) Abutment grouting operations can cause
hydraulic fracture and seams of grouted material in the dam.
Sherard (1972) reports that grout seams were found in the
downstream filter of Matahina Dam due to this cause.
(4) Compaction may lead to cold welding of
particles. Tbr.zaghi and Peck (1967) note that slight
cohesion is greatly increased by intense compaction.
(5) Soft materials may develop cohesive bonds
under embankment. loading. Sherard (1972) made the follow-
ing comment about the downstream filter of Matahina Dam,
"Although it was a completely cohesionless soil, the com-
pacted embankment formed had some cementation and unconfined
compressive strength. This was observed particularly in
184

the drainage gallery in this zone which was installed by


tunnelling through the completed embankment. In this
work it was noted that the soft crushed-rock transition
material appeared to revert to its original state (i.e. a
homogeneous soft rock) under pressure. This gave the
material sufficient cohesion that there was no trouble with
roof support or lost ground during the tunnelling."
(6) The possibility of the development of
cohesive bonds due to secondary consolidation has been dis-
dussed in Section B 2(d).
(7) Mantz (1975) found that bacteria and proto-
zoa produced mucilage which formed cohesive bonds between
particles.

(e) Particle Shape:

At Viddalsvatn :Dam (Vestad (1976)) the down-


stream filter was designed using laboratory tests in which
a crack through the dam core was simulated. However, on
impounding cracking and erosion occurred, indicating that,
at least in some places, the filter did not function correctly.
The author attributed the breakdown of the filter to arching
because of the angular particle shape of the filter material.

(f) Conclusions:

Although it has long been recognized that a filter


zone must remain cohesionless for the life of the embank-
ment there appears to have been little work done to define
the conditions under which the material may become cohesive.
185

Therefore it is considered that investigations are urgently


required to define the limits beyond which a soil becomes
cohesive.
Until more data becomes available the following
rules are suggested as a method of obtaining a cohesionless:
filter zone:
(1) Use hard, durable, chemical resistant
material.
(2) Keep the amount of fines to a minimum.
(3) Make sure the fines contain no clay or
organic material.
(4) Test both the fines and filter using
the "sand castle" test.
(5) Carry out tests for and guard against
cementation.
(6) Do not overcompact, if material breaks
down on compaction develop alternative
methods of compaction.
(7) Prevent fines from entering the filter
zone for the lifetime of the project.

D.5. MATERIALS FOR FILTERS AND DRAINS:

(a) Requirements:

Materials for filters should be sound, non


erodible materials that will not deteriorate for the life
of the dam under any climatic conditions or chemical
environments to which they will be subjected (see Part 4(f)
186

above). Laboratory tests to determine that materials are


satisfactory consist of the routine tests for soundness of
concrete aggregate or riprap supplemented by special tests
developed to test the behaviour of the material under special
conditions pertaining to the job. Routine tests consist
of the sodium sulphate soundness test, Los Angeles abrasion
test, absorption, freeze thaw cycles (for cold climates),
heating cooling cycles (for hot climates) etc. It is mot
the intention of this report to go into details of these
or other tests. Details are given in the testing standards
of various countries.
The particle shape should be approximately
cubical. A commonly used criteria is to require that the
maximum length of a particle be less than three times the
minimum dimension. However, in specifying any require-
ments, the properties of the material that is expected to
be used in the job must be taken into account.
Testing of alternative sources to discover
those satisfactory for use as filter and drain zone materials
should be carried out early in the investigation stage of
the job. These are critical zones in dam design and the
discovery when the design is complete that a source expected
to be used as a borrow area does not meet the requirements
will lead to either considerably increased costs or the use
of unsatisfactory material in the filter zones.

(b) Natural Deposits:

Natural deposits can consist of granular deposits


or quarried rock.
Material from granular borrow areas may be used
in unprocessed or processed form. The most common processing
is washing to remove excess fines. Additional data on
processing may be obtained from Sherard et al (1963).
Vaughan (1976) notes that a given percentage of non-cohesive
fines may be readily obtained by passing a free draining
material through a ball mill or similar. As the gradations
of the filter zones are very important processing of the
filter material will frequently be required. In such cases
this, together with the amount of processing envisaged,
should be clearly stated in any contract documents. A.
contractual position where a low price has been submitted
for an item that should have been much higher will lead to
arguments with the contractor and the likelihood of sub-
standard work on one of the most critical parts of the job.
Where no granular deposits exist within economic
hauling distance of the job it may be necessary to produce
filter material from quarried rock. In such cases con-
siderable processing will be required. It is most
desirable that if quarried rock is to be used a test quarry
be opened up early in the investigations to check the nature
of the rock and the amount of wastage. Generally crushing
will be required and trials should therefore be carried out
using a crusher as close as possible as that expected to be
used on the job. Crushing will give information as to the
particle shape of the crushed rock and the cuantity and
nature of fines to be expected.. One problem with crushed.
188

rock is that the fines may tend to breakdown into cohesive


material. This matter should be carefully checked.

(c) Artificial Granular Materials:

Where no satisfactory granular deposits occur


close to the dam site there may be artificially produced
materials within a more economic hauling distance. Such
materials will generally be produced as waste products to
an industrial operation, for example slag from blast furnaces
or coal fired power stations. There is no reason why such
material cannot be used satisfactorily in filter or drainage
zones, however, caution is recommended. Particles tend to
be smoother and rounder than those in natural deposits.
Gradations may also vary in unexpected ways. Kudlik et al
(1973) report of a dam which had to be repaired because a
filter of glass slag did not function correctly. There
may also be peculiar chemical reactions. Therefore all
such materials should be thoroughly tested. In addition
it is suggested that artificially produced material should
not be used without testing of it as a filter under both
uncracked and cracked conditions in the laboratory.

(d) Filter Fabrics:

In the last ten years filter fabrics have become


reasonably popular in engineering works, primarily road
construction, where they are used to provide a filter
between the natural soil and the sub-base of the road or to
prevent fine material washing into a coarse drainage zone.
189

The fabrics consist of plastic meshes formed to give a


reasonably constant mesh opening by weaving or felting.
As there is a maximum of ten years experience available
and considerably less for most meshes they cannot yet be
considered for the internal filter of a zoned embankment.
They may be used for minor repairs where their performance
can be observed. The U.S. Corps of Engineers has carried
out tests on some fabrics (Calhoun (1972)) and found that,
in general, the woven fabrics perform satisfactorily. The
felted fabrics tested by Calhoun were not as satisfactory
but recent reports (Seemel (1976) and. Kallaire (1976))
suggest that these also perform satisfactorily.
The, two main' problems with the fabrics is that
they tend to breakdown slowly when exposed to sunlight and
they are difficult to join satisfactorily. In an embank-
ment the first is not too important as a filter zone will
be protected from sunlight. However, a satisfactory joining
or lapping of adjacent sheets is essential. All authors
emphasize that the fabric must be properly tested and
matched to the job required of it.
Filter fabrics have been used successfully
between granular materials and fine materials varying from
coarse silt to highly plastic clays.
The main drawback in the use of filter fabrics
in dam construction is that they are not cohesionless.
Hence they will not prevent a crack progressing downstream
of the filter and if they are broken or overstretched they
190

will not prevent enlargement of a crack by erosion. Per-


haps some method could be developed in which the fabric
was place in a loose condition or several, overlapping layers
were used to ensure that no breakage occurs. Such methods
are conjecture at present, hence filter fabrics cannot be
used until reasonable, proven methods are developed to
ensure that they will function correctly if the core cracks.
A further problem with an internal filter of
filter fabric is one of. construction. Placing a near ver-
tical or vertical sheet across the valley could lead to sig-
nificant construction problems.
In summary it is considered that filter fabrics
should not be used as internal filters in embankment dams
until:
(1) more experience is available
(2) a satisfactory method of joining is
available'
(3) a fullproof method of placement can be
developed and proven to ensure that tensile
failure or excessive enlargement of
openings does not occur under the tensile
strains and loads applied inside the
embankment.

D.6. GEOMETRY OF FILTERS AND DRAINS:

Typical layouts for filter and drainage zones


are shown in Figure D-l.
The downstream filter and drainage zones must
191

extend over the entire longitudinal section of the dam at


least to top water level in order to intercept any crack
that may form through the core. Experience shows (Sherard
(1972)) that cracks are more likely to occur near the abut-
ments, hence special care should be taken in this area.
As mentioned in Section D2 the thickness of the
filter and drainage zones is usually determined by the method
of construction envisaged. Details may be found in

Sherard et al (1963). However, thickness must be checked


for adequate drainage capacity as discussed in Section D2.
The other requirement is that the zones be thick enough to
maintain an adequate overlap after any possible movements
due to an earthquake, particularly movement on a fault
beneath the dam. This matter is discussed in Section B10.
The downstream filter zone must be located
immediately downstream of the core or fill zone to be pro-
tected and made self healing, otherwise erosion through a
crack in the core will not be prevented by the filter. In
order to be self healing the filter must be able to flow
downwards readily. Where it is approximately vertical or
inclined downstream this presents no problem. However, in
a dam with an upstream inclined core the filter will tend
to flow away from the face it is protecting and frequently
the filter is made *ider at the bottom to account for this.
Where the filter material is not expensive the downstream
face of the filter can be made vertical. This enables the
drainage zone to be made vertical and water from a crack to
192

be carried vertically downwards thus avoiding erosion of


downstream zones. In an otherwise homogeneous embank7
ment it is common practice to place the chimney filter
vertically. This ensures positive downward movement of
the filter if a crack occurs and also is the easiest orien-
tation for construction purposes. De Mello (1977) notes,
however, that in a high embankment (more than forty metres
high) a vertical, central, filter zone leads to unfavourable
stress condi.tions and causes upstream cracks. He therefore
recommends that a downstream inclined filter be used in
such cases.

Blanket filters and drains should be located


as low as possible in an embankment but preferrably above
tailwater so that seepage from the blanket can be carefully
monitored.

D.7. OTHER POINTS•

(a) Segregation:.

Segregation of a filter material may occur during


placement operations. This is discussed by Sherard et al
(1963) and they recommend that the maximum size of a filter
zone be limited to three inches to reduce this problem to
a minimum.

Segregation may also occur during vibratory


compaction when the fine material collects on the top of
the layer (Bertram (1972)). This is another reason for
limiting compaction of filter zones (see Part 4 above).
193

(b) Construction Control:

As emphasized throughout this section the filter


and drainage zones are critical to the safety of the embank-
ment. It is therefore essential that they be correctly
constructed. The first requirement is that they be properly
specified. Should this be done construction control
should be relatively easy.
The following factors should be properly controlldd

(1) Gradation should be checked regularly at


source, on the fill and when placed
(2) no contamination of filter zones by material
from other zones or any other materials
(3) correct compaction by checking passes of
compaction equipment, density tests and
flooding of filters and observing settle-
ments
(4) no segregation of materials in the zone
(5) correct geometry of zones is maintained.

(c) Gradation of Filters and Drains Relative


to Downstream Zones:

The gradation of the downstream zone relative to


the zone adjacent to it, that is either filter or drainage
chimney, must meet the filter requirements as laid out in
this section.
This applies to the case where large voids
194

could occur in the downstream shell, for example where the


downstream shell consists of rockfill. Piping occurred
at Matahina Dam (Sherard (1972)) where a large void in the
rockfill immediately downstream of the filter zones allowed
the filter to be washed out.
If the downstream shell consists of fine grained
material cracks may form downstream of the filter or drain-
age zone. In this case the filter or drainage material
must be capable of bridging such cracks without significant
loss of material.
195

DAM HEIGHT SEEPAGE FLOWS


(Metres) (Litres/Second)
Wister 27 570
East Branch 56 300
Shek Pik 55 , 5
Euclides da Cunha 60 0.3 (through crack)
Djatiluhur 112 5
Matahina 61 570 (420 estimated for crack)
Balderhead 48 60
Yards Creek Upper Reservoir 24 80 (25 to 30 in one area)
Hyttejuvet 90 63
Abiquiri (2) 97 106 (through abutment)
Viddalsvatn(5) , 75 210

Note: (1) All data except (2) & (3) from Sherard (1972).
(2)Data for Abiquiri from Bertram (1967).
(3)Data for Viddalsvatn from Vestad (1976).

TABLE D-1 SEEPAGE FLOWS THROUGH DAMS


196
1. Terzaghi's Rules:
Ref.: Terzaghi (1922).
(a) D15f
D
85bkfinest grading
(b)D15f >4
15b(coarsest grading)
2. U.S.C.E. (1953)
(a)Cu <:1.5
D
15f e/' 21E <Z/N,?5 15f 20
D D D
85b 50b 15b
(b)1.5 <Cu < 4
D
1515f <5 0f- 15f 20
D8 D
$ 50b 15b
(c)Cu>. 4

2221L.e'25
D15f D
85b 5010 15b

(d) Gap graded materials and materials so widely graded that they tend
to segregate during placement are not recommended.
(e) If there is any doubt, laboratory tests recommended for final
design.
(f) Suggests use of Figure D-2 for design purposes.
3. U.S.C.E, (1955)
(a) D
15f D50f <::,:a5 D15f >
85bs 50b $ D15b •
(b) For medium to highly plastic clays D15 = 0.4mm. and
requirement is disregarded. 50b

(c) The filter must be well-graded.


(d) For important jobs laboratory testing is recommended.
4. U.S.B.R. (1968)
(a) For uniform filters,
5‹ D50f.,(10
D
50b
(b) For graded filters with subrounded particles,
D D
12< 50f ,./58 12 15f < 40
D
50b 15b

TABLE D-2 SUMMARY OF FILTER RULES --- SHEET 1


197
4. U.S.B.R. (1968), Cont/d,
(c)For graded filters with crushed stone particles,
9c D 50f4(: 30 6‹ 15f < 18
D
D50b 15b
(d)Not more than 5% of filter material to pass the No. 200 sieve.
(e)Maximum stone size of filter material =3 inches.
(f)Only base soil smaller than No. 4 sieve to be used in design.
(g)The gradation curves to be roughly parallel in finer regions.
5. Sherard et al (1963),
15f
(a) ! . 5 1:1 <
D D 4
15b 85b
(b)Not more than 5% of filter material to pass the No. 200 sieve,
all fines to be cohesionless.
(c)Where base soil contains a large percentage of gravel, the filter
should be designed for base soil smaller than 1 inch size.
(d)Filter and base gradings to be approximately parallel.
(e)Maximum filter stone size to be 3 inches.'

TABLE D-2 SUMMARY OF FILTER RUTRS --- SHEET 2


198

Sand Filter/Drain

Suitable Granular
Toe

(a) Homogeneous Dam with Central Filter

Upstream Transition Downstream Filter

Chimney Drain

Downstream Drainage Blanket


Core Shell

(b) Central Core Dam

Downstream Filter
Upstream Transition
Chimney Drain
Upstream
Shell Drainage Blanket

(c) Dam with Upstream Inclined Core

FIGURE D-1 TYPICAL LAYOUTS - FILTERS AND DRAINS


199

40 40.

.. ..

NOTE: • •

20
oaa a -UNIFORMITY COEFFICIENT OF BASE'
IN
W B
010
0 STABLE
- • FAILED CURVE 1.74-"'" -
44.0 A STABLE WHEN VIBRATED .....5.-e uovt 7
0 FAILED WHEN VIBRATED
o

10
0 — —

2
— • a a a °a o ° go 0 • • • • -,
0
— a . —
•4 0, A A
-
ttO0 •
• ' • • •
2 •
u
4 .
% 04 •
o .
A
I - • . -

. / • •
rt a e .
e, • 2
2
/ • II •
/ a ' • • • •
. oe eye 8.
• ° • / • •
i.o 041
I I 1111 I 1 1111 1
4 10 20 40 100 200 400

RATIO "
-IL
0,5B

Notes: (1) Data from U.S.C.E. (1953).


(2)Curve I refers to static conditions.
(3)Curve II refers to dynamic conditions i.e. soil vibrated

-24=AGAINST C (BASE)
FIGURE D-2 PLOT OF FILTER RATIO D
u
15b
100

r- 50
1

CLAY F M ,I F M C
SILT . SAND GRAVEL

(a) Grading of fill when chemically dispersed.


(b) Grading of fill when mechanically dispersed in river water.
(c) Grading of floculated clay suspension used in filter tests.
(d) Grading of uniform filter which retained clay - k = 7 x 105m/s.
(e).Grading of well graded natural sand which retained clay - k = 2 x 10-5' •

(f) Grading of uniform filter which passed clay - k = 1 x. 10 4m/s.


(g) Grading of well_ graded clean sand which ..Dasped_ clay — k = 2 x 104m/s.
Data from Vaughan (1976).

FIGURE D-3 FILTER TESTS - COW GREEN DAM


201

10

E
to

LLI

tr,
LONDON CLAY I F-1 DI S LLED WATER-
U3

<

10-5 io-4
Permeability of Filter - k m/s
Data from Vaughan (1976).

FIGURE D-4 FILTER PERMEABILITY VERSUS AVERAGE FLOC SIZE WHICH PASSES
202

SECTION E

UPSTREAM TRANSITIONS

.The upstream transition of a zoned embankment


fullfills the following functions.

(1) It prevents fines washing from the core into a more


pervious zone under wave action or due to water seeping
from the core under drawdown conditions. This is usually
a minor consideration as waves are damped to very small
heights by, the upstream zones and in most dams flows during
drawdown will be small and occasional. However, with the
advent of pump storage schemes drawdown will be occurring
frequently in some dams. A further problem is the
performance of an embankment with a dispersive core under
drawdown conditions. As there will be no crack and
reservoir to provide large quantities of flow deterioration
would be slow but some deterioration could occur if the
dispersed material did not arch between the transition
grains. As requirements for filters have been fully
discussed in Section D no further discussion is included
here.

(2) The permeability and thickness of the upstream


transition must be such that it reduces the flow into a
crack to allow time for the clay to swell to seal the crack.
In a potentially dispersive core flocculated by mixing a

chemical with the natural material the flow through a crack


should also be minimized to avoid leaching of the chemical
from the soil. It is common to allow a larger percentage
of fines (that is minus 200 sieve material) in the
upstream transition to reduce permeability. Such fines
must be non-cohesive.

(3) The upstream transition forms one part of a possible


failure plane. The strength of the material must therefore
be adequate to prevent instability - in particular it
should be at least as strong as the core material. In
order to achieve this some compaction will normally be
required and as the situation is similar to the downstream
filter similar compaction to that described in Section
D,6 (a) should be used.
In addition the upstream transition will normally be
in a fully saturated condition. Hence due precautions
must be taken against liquefaction under earthquake forces.
The susceptibility of the material to liquefaction under
seismic forces can be checked using vibrational testing
techniques but as a general rule it should be well graded
and adequately compacted rather than have a uniform
gradation.

(4) The material in the upstream transition may flow


into any crack and effectively seal it. This behaviour
has been credited with sealing cracks on Yards Creek and
Matahina Dam (Sherard (1972)).
204

Terzaghi and Peck (1967) make the following


comment about the upstream transition (layer b):
"Ideally it should consist of a moderately compacted sand
that will migrate into a crack in the core and fill it.
Even if layer b contains some silt and therefore possesses
slight cohesion, the cracks in b will not remain open
because of seepage pressures. However, if b is well
graded and thoroughly compacted a natural filter will
develop and prevent the desirable migration of finer
particles into the crack in the core".

It is considered that the use of a moderately


compacted fine sand in such a location could lead to
stability problems in the case of earthquake, especially
with an upstream inclined core. One possibility would
appear to be to use a gap graded material consisting of
a mixture of coarse material to provide the strength and
a fine, non-cohesive material to provide the mobility and
low permeability. The fine material would just fill the
voids of the coarse and be washed into any cracks that may
form in the core. No evidence has been found of such
an idea being considered previously and laboratory testing
would be required to check the concept and obtain a
suitable mixture.

In summary the ideal upstream transition should


meet the following requirements:
205

(1) the gradation should meet the requirements of a


filter relative to the gradation of the core.

(2) it should be cohesionless.

(3) the permeability should be relatively low - in the


-3 to 10-4cm/sec.
range 10

(4) It should contain more fines than is usual for the


downstream filter - up to 10 percent minus the
number 200 sieve. Such fines should be non-
cohesive.

(5) Maximum size should be limited.

(6) The material must have shear strength at least as


high as the core material.

(7) The zone should not be overcompacted.

(8) The material should not liquefy under earthquake


forces.

(9) The finest material in the transition must be


retained by the downstream filter.
206

SECTION F

MONITORING TO DETERMINE FORMATION OF CRACKS OR PIPES

GENERAL

It is desirable that a program of monitoring of the

embankment be set up to investigate the possibility of transverse,

throughgoing cracks occurring in the dam core. This program will

function as part of any overall instrumentation program.

The program should start when construction of the dam

starts. During construction positive identification of cracks can

only be by visual means, although measurements of stresses and strains

may indicate possible problem areas. Where cracks occur on the surface

of the fill due to delayed construction they should be backfilled

as described in Section B.3. Where cracking is indicated as a possib—

ility from stress or strain measurements and the dam has been built

assuming that cracking can occur then there is generally no need to

rush into an expensive repair program, however, careful observations

should be kept during impounding.

When impounding commences the presence of cracks is able

to be determined by seepage measurements and pore pressure behaviour,

as well as the methods described above. Once again the presence of a

crack should not lead to immediate lowering of the reservoir where the

dam has been designed for cracking. A reasonable amount of time should

be allowed for healing to occur.

In monitoring the performance of an embankment one isolated

reading on one instrument is of little value. Hence a program must be


207

drawn up in which there are an adequate number of instruments or

measuring points and readings are taken at regular intervals the

magnitude of the intervals being dependent on the operation Of the

reservoir and performance of the embankment. In this way a possible

problem will be revealed by several instruments and a peculiar reading

by only one instrument will probably be due to a fault in the instrument

(it should not be completely disregarded, however.) In addition it

will be possible from a plot of measurements against time to see if the

situation is deteriorating, static or improving. If there is

deterioration the rate of deterioration will enable determination of

the time at which the reservoir should be lowered to prevent problems.

F.2. VISUAL OBSERVATIONS


Visual observations may be direct observations of dam

performance based on regular inspections, direct observations in test

pits or of cleared surfaces or indirect observations by means of

television or photographic cameras in bore holes.


Direct observations for surface cracking should be carried

out on a regular basis. During construction all surfaces of the core

left for some time should be examined before restarting construction

at that location. Once construction has been completed, during

impounding and during reservoir operation regular inspection should

be carried out in which specific areas are examined. The frequency

of these inspections will be highest for the period from the end of

construction to a few weeks after the reservoir is filled for the

first time. Inspections should be carried out according to a fixed

program. The main items to be included in the program are,

(1) Cracks of any sort in any location. Areas where cracking is


more probable e.g. abutments at crest and in areas of odd geometry
208

should be carefully'exaMined
(2) any areas of subsidence.. Upstream shell should be examined for

these.
(3) Seepage on the downstream side of dam from either shell, abutment

or drainage layer.

(4) Sediment in seepag •


(5) Any sign of surface disturbances or other changes to existing

conditions that cannot be explained by construction procedures

or other reasons.

In areas of possible surface cracking special preparations

may be carried out to make observations for cracking easier. Such

preparations may include cleaning up and fencing. off the area or

placing a bitumen layer on the area. Once cracking is observed in any

area photographs should be taken so that there is a permanent record

should the cracks be disturbed. Sherard (1972) notes that in many

cases cracks may soon be covered up (due to traffic for example).

Where cracking is observed either directly or by noting

seepage from abutments investigations may be put in hand in which

trenches will be dug and horeholes may be drilled.. All ob-ervations

should be carefully recorded by photographs and field logging.

F.3. SEEPAGE MEASUREMENTS

Seepage measurements are the most positive way of

determining if a significant crack has formed through the core and

whether the situation is improving or deteriorating. In monitoring

seepage it is desirable that the embankment be divided into a number

of areas so that the area in which seepage increases is Imown. In

order to do this it will be necessary to divide the drainage blanket


209

into separate channels using impervious material. In addition

seepage can be more thoroughly monitored if the drainage blanket exits

above tailwater level. Because of these reasons seepage measuring

methods should be included in the basic design of the dam.

The presence of .a crack is indicated by excessive seepage

and erosion of a crack by the presence of soil particles from the

core in the seepage water. As discussed previously there are three

possible mechanisms that could cause cracks as follows:

(a) Cracks that are present before impounding. No flow will occur

through these until the reservoir level reaches the bottom of the

crack. If no erosion occurs flow will increase rapidly but regularly

with increasing reservoir level. If erosion occurs the flow will

increase rapidly with no change in reservoir level unless the crack

heals itself. However, once the reservoir level drops below the level

of the crack the flow will virtually cease.

(b) Cracks due to increased tensile strain caused by impounding of

the reservoir (see Section B.4.). When the crack opens seepage flow

will immediately jump and will not reduce until the reservoir level

drops below the level of the crack or the crack is healed.

(c) Cracks caused by hydraulic fracture if the water pressure on the

upstream face exceeds the minor principle stress as explained in

Section B.4. There will be an immediate jump in flow when the crack

opens. If excessive erosion of the crack does not occur the flow

will cease when the reservoir level drops below that required to

keep the crack open (as discussed in Section B.4.) This behaviour

was observed at Balderhead Dam (Vaughan et al (1970) ). Flow

measurements are shown on Figure F-1. The jump in flow did not occur

until reservoir level 332 metres but subsequent investigations showed

cracking occurred between elevations ,307 and 317 metres approximately.


210

Where a dam is founded on an impervious foundation such

as rock or low permeability alluvium collection of seepage can be

carried out for the entire longitudinal section of the dam provided

tailwater levels are sufficiently low. Where the dam is founded

on pervious alluvium or normal tailwater levels are high, monitoring

of seepage in the riverbed section is not possible. However, as

most cracking occurs on dam abutments special efforts should be

made to collect the seepage near the abutments at the top of the

alluvium or at maximum tailwater level and measure it on both abutments.

Where possible seepage thrOUgh the dam should be

separated from other sources of water such•as abutment drainage

water or construction water. Where this cannot be done measurements

of the quantity of water coming from other sources should be carried

out. In addition daily records of rainfall at the site should be

carried out so that the effect of this on seepage measurements can be

estimated.

There are many seepage measuring devices and it is not

intended to describe them here. The most commonly used are V—notch

weirs, rectangular weirs, Parshall flumes, pipes containing flow

meters, velocity measurements across a known cross section (stream

gauging) and, for small flows, the measurements of the time recuired

to fill a container of known volume. Where weirs are used to measure



seepage a continuous record can easily be kept using a simple clock—

work recorder. If a problem occurs this record may be very useful in

determining when the crack occurred and helping to locate the cause.
In addition to monitoring seepage the water should be

sampled to check for the amount and nature of suspended solids and

amount and chemistry of dissolved solids. This information will be

valuable in deducing where the seepage is occurring whether the filter

is functioning properly, whether dispersion of the core is likely and

whether the filter material or core is likely to be cemented or dissolved.

The quantitative assessment related to time gives an indication of

whether the situation is improving or deteriorating.

F.4. PORE PRESSURE MEASUREMENTS:

(a)General

Pore pressure measurement is commonly carried out in dam

cores to compare measured values with design values. It is not

intended to discuss the types of piezometers available, the selection

of piezometer types for a given situation or methods of installation.

This data is discussed fully by Vaughan (1973).

(b)General Piezometric Levels in the Core

As mentioned above there will usually be a network of

piezometers installed in the core to monitor pore pressures for

comparison with design values. These piezometers will generally be

located on typical sections through the dam.

As discussed in Section B,12 (d) construction pore

pressures may remain high, effective stresses low, settlements small

and the amount of arching small. Under these conditions the

probability of cracking during construction or impounding is small.

In fact if pore pressures are higher than the steady seepage levels

cracking cannot occur.


212

The other possibility is that pore pressures at.the end

of construction are low effective stresses high, settlements high

and the amount of arching may be considerable. Thus the probability

of cracking during construction or impounding is greater and care should

be taken during impounding.

If the end of construction pore pressures are predicted

to be high and the measured values,are low then either consolidation

occurred more rapidly than expected or the average total stresses in

the core are lower than expected. Low average stresses indicate

that arching has occurred and cracking is more likely. If pore

pressures are as predicted in one location indicating material

properties in accordance with design predictions, and low in another

then arching is very probable in the area of low pore pressures.

This behaviour was observed at Balderhead Dam (Vaughan et al (1970) )

where low pore pressures due to arching were observed in - the upper

part of the dam only.

If, in the case Where construction pore pressures are

high, subsequent reservoir operation leads to the pore pressures

dropping below the steady seepage values, the increased effective

stresses will cause additional settlements and arching and increased

possibility of cracks. In such cases careful monitoring should be

continued for the lifetime of the project. Special attention should,

be paid during periods of low reservoir so that suitable precautions

can be taken if the pore pressures fall below previous minimum

values.
213

During impounding pore pressure observations will enable

comparison of the rate of impounding and the rate of advance of

the seepage or swelling front. If the relative rate of impounding

is slow then hydraulic fracture must be drained and cracking is

less likely than if the reservoir is impounded rapidly compared

with the swelling of the core.

(c) Piezometric Levels in Low Stressed Areas:

It may be considered desirable to install piezometers

to measure pore pressures in an area of potentially low stresses.

In such an area low pore pressures during construction will

indicate low stresses. As mentioned above, low pore pressures

may also indicate rapid consolidation, hence it is necessary to

measure pore pressures in other areas for comparison. If the area

to be checked is near the abutment then consolidation may be quicker

than elsewhere due to drainage to the rock. In addition core

material placed against the abutment may be wetter than elsewhere,

also influencing pore pressures. Hence piezometers will also be required

near the abutment in areas where low stresses are not expected to obtain

a valid comparison.

Piezometers installed in low stressed areas will enable

stresses to be measured in these areas using controlled hydraulic'

fracture tests as explained in part. 6 following.

When the reservoir is impounded the pore pressures in

the immediate vicinity of a crack will be influenced by the water

pressure in the crack. The distance the influence extends depends

on the permeability of the core and the time the pressure is

applied.
214

If the core is of relatively high permeability (i.e. a silty core)

piezometers located in the area may measure changes in pore pressure.

The water pressures that will occur in the crack are:

Non through going Crack — Full reservoir pressure to the down—

stream end of the crack.

(2) Throughgoing Crack — Depends on size of crack. If the crack

is wide most head drop will occur in the upstream zones and

down—stream filter. Hence the water pressure will be almost

constant at a level below reservoir level.

(3) Sealed Crack — Initially at full reservoir pressure through—


out hopefully dropping to steady seepage values with time.

(4) Hydraulic Fracture — Steady seepage values changing to (2)


above when the critical reservoir level is reached and reverting

to steady seepage values when the reservoir drops below the

critical level (provided the crack is not excessively eroded).

(d) Piezometric Levels in Upstream Zones:

Where there is no crack through the core of a dam the

pore pressures in the upstream zones will be equal to the reservoir

level. When a crack is formed flow through the crack will be

partially controlled by the upstream zones and the piezometric

level in these will drop below reservoir level. This behaviour

was observed at Balderhead Dam (Figure F7,1) and as part of the repair

cracks a network of piezometers was installed in the upstream zones

(Lovenbury (1973), See Figure F-2). Locations of piezometers

must be based on estimates of drops in piezometric levels. It

should be noted that where the upstream shell is more pervious

than the upstream transition all head drop is likely to occur in


215

the transition. The. method of design used for Balderhead Dam is

described by Lovenbury (1973).

F.5. DISPLACEMENT MEASUREMENTS:

(a) Settlements:

Settlements may be measured by level surveys of crest and

surface monuments inclinometer or cross arm devices installed either

vertically or at an inclination in the fill and a variety of devices

using horizontal tubes through the fill. Settlements measured by


all devices should be related to the same bench mark preferrably

. remote from the site,

During construction measured settlements should be

compared with predicted values. If they agree then the embankment

is probably performing as expected. If settlements are less than

expected there are three probable causes

(1) The fill material is less compressible than anticipated.

Where this is thought to be the cause, reasons for the

reduced compressibility should be found and tests' may be

required for confirmation.,

(2) ConsolidatiOn may be progressing more slowly than expected.


In this case construction pore pressures will be higher than

anticipated.

(3) Arching may be occurring. In this case both settlements and

pore pressures will be low. The effects will probably vary

within the core and it should be possible to separate arching

from (1) above. This effect was noticed at Balderhead Dam

(Vaughan et al (1970) ).
216

During impounding suitable settlement measurements will

enable identification of the following mechanisms,

(1) Relief of arching in the core.

(2) Collapse of upstream shell on saturation.

(3) Collapse of foundation on saturation.

In the long term settlement measurements will determine

how much creep is occurring. If this large, cracking is more likely

in the future.

'(b) Strains:

Strains in an embankment may be obtained by measuring

between surface monuments and using inclinometers or extensometers

buried in the fill. Measurement of strains will not generally give

a direct indication of a crack, although a large local increase in

strain between successive readings suggests a probable crack. In

general strain measurements allow the most likely location of a

crack to be determined — particularly on the dam crest.

On the crest strains may be obtained by measuring

between monuments or using buried extensometers. The latter are

more expensive and difficult to install but they have the advantage

that they are more accurate and, once installed, are less likely

to be damaged. Tensile strains at which cracking has occurred

on dam crests range from 0.09% (Leonards and Narain (1963) ) to

0.6% (Nonveiller (1973)). Laboratory tests give tensile failure

strains from 0.05% (Leonards and Narain (1963))to 4.5% (Hasegawa

and Ikeuti (1966) ). Leonards and Narain (1963) comp,red the

actual strains at cracking with those determined by tests on beams


of the core material and found some agreement. Hence this is one

way that strain measurements may be used to predict cracking on

dam crests. Usually beam tests are not carried out and an estimate

must be made of the strain at which cracking is likely.

In the body of the dam tensile strains may occur when

all stresses are compressive. Hence internal strain measurements

are of little value in determining if a crack has occurred. They do

indicate where tensile strains are high and thus cracking may be

more likely.

F. • STRESS 1'.IEASUREMENTS:•

When there. is a crack in the core the minor effective

stress at the surface of the crack will be zero. However cracks

are very localized and are oriented in specific directions hence

it is not practical to install stress cells to act as indicators

of cracks occurring. Stress measurements are undertaken primarily

to measure areas of. loW stress and to compare measured stresses with

those calculated.

Stress measurements may be carried out using conventional

stress cells, hydraulic fracture tests from hydraulic piezometers

or from measurements in boreholes.

The most practical measurement to make using conventional

stress cell installations is the reduction of vertical, stress in

the core due to arching. This has been carried out by Lofquist

(1951) and by Sigvaidason and Tawil (1972). Cracking takes place on

the plane of minor principle stress which may be oriented in any

direction. Hence measurements are required in all directions.


218

Quantitative measurements of stresses in directions other than

vertical using stress cells installed in the fill have not been
particularly successful and are not generally recommended.

Stress cells may also be installed in the face of a concrete

structure in the soil. In thi- case the cell forms part of the

structure and provided the material around it is placed as part

of the overall fill will probably give a reasonable indication

of the stress on the face of the structure. Hence this would give

an indication of low stressed areas and possible risk of cracking

near the structure. Vaughan and Kennard (1972) describe stress

cells used in this way at Cow Green Embankment.

It has recently been suggested that the minor principle

stress in a soil mass can be measured by causing controlled

hydraulic fracture in the soil at a piezometer and measuring the

pressure at which the crack closes (Bjerrum and. Anderson (1972) and

Vaughan (1972) ). As only a small amount of water is injected at

the piezometer only minor changes to soil geometry would occur and

therefore only minor changes to the minor principle stress would be

expected. Results presented by the investigators are reasonable

and therefore suggest the method is satisfactory. However, the large

differences between pressures required to initiate a crack and those

at which cracks close (Vaughan (1972) and Table B-1) are more than

would be expected to overcome tensile strength of the soil. This

indicates that there are still unknowns with this method and further

investigations are required.


219

There has also recently been development of devices for

measuring stresses in boreholes. A comparison of devices is

included in Massarsch and Broms (1976). These devices measure the

total stress at right angles to a borehole hence will measure

horizontal stresses in a dam embankment. As the devices are new

the results are still somewhat suspect, however, these cells could

be quite useful in determining areas of low stress in an investigation

where cracking of dams is thought to have occurred.

When pore pressures are positive, cracking depends on

effective stress considerations. Hence the pore pressure must be

measured at all stress cell locations.

F.7. USE OF BOREHOLES:

In many cases boreholes drilled through dam cores to

carry out grouting operations, install instruments or for other

purposes have shown large losses of water, mud or grout into the

core. This has been a matter of considerable concern and in some

cases extensive repair operations have been carried out because of

such losses ( refer Shek Pik, Dir and Girna Dams in Sherard (1972) ).

On the other hand a large number of dams which have experienced

such losses have not been repaired and have not had excessive

leakage on impounding.

Losses occur in drill holes because the pressure due to

the head of water, grout or mud exceeds the minor principle stress

in the soil. In a core with no arching the minor principle stress

will be approximately Ko times the vertical stress where Ko is the

at rest coefficient. For a reasonable Ko value of 1/2 the average


220

minor principle stress will therefore be about 1/2 the pressure

due to the weight of overlying fill. A small local variation in

pressure will thus reduce the minor principle stress to the pressure

exerted by the head of water in a borehole. Hence hydraulic

fracture under these circumstances is quite possible and should

not be unexpected. It is therefore considered that hydraulic

fracture in boreholes drilled from the dam crest does not necessarily

indicate an adverse condition.

When hydraulic fracture occurs in this way there is

frequently a large flow of water, grout or mud into the dam core.

This may cause a crack that remains open due to erosion. Vaughan

(1977) noted that at Balderhead Dam cracks formed by hydraulic

fracture in bore holes remained. open after only a few. hours of

water flushing through them, due presumably to erosion. Hence such

cracks should be avoided. This is best done by not drilling holes

through the core. Sherard (1972) recommends that, if holes have

to be drilled, then methods should be employed which do not use

drilling fluids in uncased holes, such as hollow stem augers or

by driving and cleaning out casing. Where cracks have been formed

they should be sealed by grouting with flexible grouts as described

by Vaughan et al (1970).

When significant cracking of a dam core has already

been revealed by seepage flow increases or other information and

repair work is necessary drill holes may be used to identify zones

of disturbed material, reworked material or low stresses. In such

cases usual exploratory techniques may be used such as thin wall

sampling, and the standard penetration test. Areas where total

water losses occur should be identified as lower stressed zones.


221

Borehole cameras may be used to identify zones of coarse material

or open cracks and pressure meters (as discussed in 6 above) used

to identify low stressed areas. Soil samples should be tested for

water content and grain size distribution and the salts chemically

analysed to see if dispersion is a possibility.

F.8. USE OF TRACERS:

Once excessive seepage is occurring and cracking is

suspected the location of the crack may sometimes be determined by

introducing a suitable tracer into certain areas and checking for

signs of the tracer at piezometers, drill holes or exit locations.

Tracers must disperse very rapidly and be able to be identified

when present in very small quantities. Examples of the use of

tracers in crack detection are provided by Vestad (1976) and.

Coxon and Crook (1976).


222

Risorvoir loiet ,., _.i


:
E '
/Ni veau reservoir. 2
.332•23,,_
. jam,,
c
1 - 40 A

z
E
330


D4
,
6 -
‘.... D4 .
v, 20 ■ 1 Eli 330

E
z
Q 325 /V U7
. i Ned 325 Fi
ss C6
jC --: 4/
,••;4'
I.Q. CI D4' ci
320 320
50 Man undertiran 50
Drain trneneur prncipal
FLOW L./sec..

DEBITL./sec.
40 40
T South undardrain
30 Drain inforiour sud 30
20 1. 13)---- 20
_ 0 10
10
T w"It-Ailtt
L ----.....k......e=..... 0
f.-..
0
1l,l 11111 1.11.111i. r-r-in—r-r-r-T-T-T—r-rr-r-r1
1965 1966 1967 1968 1969

R6WRE" F-1 BALDER HEAD DAM - RISER VO/R LEVEL,


SEEPAGE FLOWS A, ,D WSTREA P/EZOMETERS.

DATA FA'Ob/ VA I/GRAN ET AG 0970


GROUT TREATMENT
DIAPHRAGM WALL
1
CH. 0 190 200 1" 300 400 .SOO 600 • 700 800 900 m
1 330
330 - 1 1
320 - 320
310 310
300 Top of concrete cutoff 300
290 290
m 0.D. m O.D. 280
280

340-
KEY 610 620 630 640 650 660 670
330- 335 335
Shale Multi-unit
till niter drain piezorneter 330 - 330
ELEVATION m

Diaphragm 325 - 325


wall Perforated
310- standpipe 320 - 320
Boulder clay
core 315 - 315
I Casagrande-type
300- piezorneter
310 -310
Concrete No water loss
290- cutoff WATER TESTS "1 1 11 1111111 111 1 i
Water kiss

280- ADDITIONAL PIEZOMETERS INSTALLED IN 1970


SECTION AT CH. 372

F/CURg F-2 134LDERI/L:AD DAV -LAYMT of P/LZO/WETERS IN UPSTREAM SHELL

Fro LovEweuRY (4970)


224

SECTION .G

REMEDIAL TREATMENT

- G.1 - DECISION .ON NEED FOR REMEDIAL WORK

(a) General

The decision whether or not to drain the reservoir


and proceed with remedial work because of cracking must
be made in the final instance on the basis of engineering

judgement. It is therefore essential that people involved


in the decision be fully advised of all pertinent information.
In making a judgement decision the following
factors must be taken into account:
(1) The risk to people and property downstream of the
dam. If the area downstream of the dam is unpopulated
greater risks can be taken than if it is
(2) Whether the dam has been designed to resist cracking
and whether there are indications that the design
assumptions were not correct

(3) Drainage facilities of the reservoir. If the reservoir


can be drained completely or the top part of the
reservoir drained in a short period decisions can be
delayed somewhat
(4) The amount of instrumentation and other monitoring
that is carried out and the degree of knowledge that
225

is available concerning the crack and changes in


the crack

(b) Decisions before Impounding of Reservoir

During construction of the dam any cracks that


are observed on the fill surface should be repaired as
discussed in previous sections.
Where cracks are formed by hydraulic fracture
in boreholes they should be repaired as discussed in Section
F.7.

Cracks may also occur on the crest of the dam


after construction is complete but before impounding. Such
cracks should be repaired.

As a general rule it is considered that the


occurrence of cracks prior to impounding of the reservoir
is no reason to embark on an extensive program of repairs
provided due consideration of the effects of cracks has
been taken during the design of the dam. Should no account
have been taken of cracking during the design stage then
an immediate review of the capability of the filters to
ensure the safety of the dam is required. However, should
extensive cracking be noted prior to impounding care should
be taken during impounding and frequent observations, as
outlined in Section F, should be carried out so that any
significant crack is noted immediately.
(c) Decision During or Subsequent to Impounding

There will be certain situations such as excessive


water in the drainage blanket causing erosion of the downstream
226

shell or uncontrolled seepage issuing from the downstream


shell when no decision is required. It will be obvious
that the reservoir must be drained under emergency conditions.
In an embankment designed to be safe even if
cracking occurs the above situation is most unlikely.
In general there will be somewhat more seepage than expected
and there may be some fines in the seepage water. A decision
will be required as to whether to drain and repair, hold
the present reservoir level or allow the reservoir level
to continue to rise. The decision must be made on the basis
of available data but, again as a general rule, usually
the best decision will be to remove the variable of changing
reservoir level and study the change in seepage flow under
constant reservoir level. This should stabilize and the
amount of fines reduce. In such a case the source of the
fines should also be identified. If they are not from
the core the downstream filter will be functioning correctly.
In most cases seepage will stabilize or reduce and filling
of the reservoir can continue.
Although all methods mentioned in Section F
will help identify and locate cracks the most important
is the seepage water. Should the flow be constant and
water clear under constant reservoir head or increase in
quantity regularly with increasing reservoir head there is

no problem. If it increases more rapidly than the reservoir


level or if it contains fines from the dam core then the
situation should be watched carefully. A filter can function
correctly by allowing some particles to pass through before
227

a graded filter has built up on its upstream side. Thus


some fines can be tolerated but they should reduCe rapidly.
.

In practice several cracks may form at irregular time


intervals so the quality, and quantity of the seepage may
vary erratically. If the seepage is collected separately

from several areas this procedure will.be more easily


identified. The reservoir level should not be raised until
all cracks at a given level have healed themselves as
indicated by no fines from the core in the seepage water
and constant. or reduced seepage flows. Should excessive
seepage or loss of fines occur at any reservoir level the
reservoir should be drained and repairs carried out. A
very good description of the behaviour of Viddalsvatn Dam
under cracking and self healing conditions is given by
Vestad (1976).
Another source of concern is the possibility
of the core material dissolving in seepage water. This
should be checked by periodic chemical analyses of the seepage
water. Initially this water will contain salts which were
already present in the porewater of the soil. Hence,
sufficient time must be allowed for the porewater to be
replaced by reservoir water before dissolving of the core
material can be postulated. Generally, the salt content
will be reasonably constant until the reservoir water appears
in the seepage water and then the percentage will reduce.
No reduction or a change in chemical nature of the seepage
could indicate a problem.
228

One reason for repair work not so far mentioned


is that the seepage flow may be more than can be tolerated.
As seepage water quantities are usually small and some water
usually has to be released for downstream uses this will
not be a consideration in most cases.
Sometimes sink holes appear in the upstream
face of the dam or cracks appear on or near the crest during
or after impounding. These in themselves should not be
a cause for lowering the reservoir. The cause of these
should be investigated. In the case of sink holes the
cause will probably already have healed itself and all that
need be done is repair the sink hole. Cracks on the crest
will usually only be a cause for concern if they are
accompanied by increased seepage.

G.2. METHODS OF REPAIR

(a) General

It is not intended to cover this subject in


detail. The methods chosen will depend on the nature of
the problem, the available equipment and materials, the
scheduling of the project and economics. Table A-1 summarizes
repair methods carried out on dams that have cracked and
reference should be made to the articles from which these
have been selected should greater detail be required.
229

(b) Minor Repairs

Minor repairs will generally be associated


with cracks that appear on the surface of the dam core
during construction or near the crest of the dam subsequent
to construction.
Sherard (1972) reports that the U.S. Soil
Conservation Service repairs cracks in small dams by filling
the cracks with a slurry consisting primarily of natural
clayey soil and water. Material with a fairly high content
of fine to medium sand is preferred. Also a plasticity
index of less than 10% is desirable in order to reduce
shrinkage 011 drying. The mudfilled cracks are allowed
to dry for about a week. If appreciable shrinkage of
the slurry occurs then the process is repeated. Finally
the crack is sealed at the surface by trenching and recom-
pacting to a depth of 1 to 2 feet. More details are
included in Sherard (1972). He notes that feelings are
mixed on this method of repair but that in most dams
repaired there has been no reappearance of cracks on the
surface or reports of other trouble.
Another method is to excavate a trench through
the cracks and backfill the trench with compacted core
material either placed wet of optimum moisture content
or with some clay added. This method has the advantage
that the extent and nature of the cracks can be observed.
(Sherard (1972) ).
230

Where water is seeping through the crack the


treatment is to place a suitable filter layer over the
crack, ensure adequate drainage of the filter layer and
place a suitably dimensioned weighting zone on the filter
to ensure that it is stable. This method was used on
'Euclides Da Cunha Dam in Brasil (Vargas and Hsu (1970)).

(c) Major Repairs

Where a dam falls down the only repair is to


rebuild it properly. This matter is not considered here.
Where an uncontrollable crack has occurred
through a dam core the two methods of repair that have
been used most extensively are to provide a new impervious
membrane or to grout the existing core material.
Where an impervious membrane is provided this
may be sheet piles, flexible concrete wall or, where the
dam is homogeneous an upstream impervious membrane. Sheet
pile walls have functioned successfully in a number of
cases (e.g. Lofquist (1955) ). The possibility of broken
interlocks, however, makes them suspect as the preferred
seepage path caused by these could endanger the dam (e.g.
.Casagrande (1950) ). Where concrete walls are used they
must be flexible and their modulus of elasticity closely
match that of the soil in the core. This was done at
Balderhead Dam (Vaughan et al (1970) ) refer. Figure G-1.
Where the dam is homogeneous an upstream impervious blanket
of plastic, clay, asphalt or concrete can provide a suitable
231

solution provided they are adequately protected. Several


instances are shown in Table A-i.
Grouting of the core has been used at Balderhead
Dam (Vaughan et al (1970) ) and at Hyttejuvet Dam
(Kjaernsli and Torblaa (1968) ). This method has the
advantage that as well as sealing the cracks a prestress
can be applied to the core thus reducing the possibility
of future cracking. Grouts must be flexible and preferably
of low shear strength or they will increase the cracking
potential of the soil. Cement and clay mixtures are
therefore desirable. Design of .a suitable material is
presented in the paper by Vaughan et al (1970) refer
Figure G-1. During the grouting operation it is essential
that grout does not penetrate the upstream transition or
'downstream filter zones. Laboratory tests should be
carried out to check if penetration will occur and water
should be flushed through holes drilled in the zones adjacent
to the core while grouting is being carried out to check
that no contamination occurs.
Where uncontrolled erosion of the core material
occurs because it becomes deflocculated the only way in
which it can be repaired without rebuilding the embankment
is to return the soil to the flocculated state. At Norwell
No. 2 Dam (Sherard (1972) ) the section lost due to dispersion
was rebuilt and a blanket of gypsum was placed on the

upstream face of the dam to ensure seepage water was rich


232

in calcium ions. This would appear to be almost the


only practical way of ensuring flocculated soil in an
existing embankment where the core has begun to disperse.

233

30 ----17 LT Test Material


rale %
.Age, days teneur kg 0:
cin2
Essal Materiaux Age, fours en EMI

All Grout: 44 200 2.1


Cou/is :
22 A21 52 % cement 44 200 6.3
ciment
BI Grout : (51 140 2.1
20 Coulis :
B2 69 % cei,m
n eennft 51 6.3

. CI ) Boulder clay ( 12.6 2.1


8b
—A2 C25 Argyle 6 blocaux 1 — 12.6 4.2
2.5
8b Plastic concrete 59
22 Beton plastique 57 5.0
10 /i - C2
---
I7LT Cement
Ciment 9.7 %
241/280 2.5

Bentonite
2.0 %
Bentonite
— —C1 Water 19.3 %
Eau
Aggregate
69.0 %
Agregats
0 10 20
STRAIN
DEFORMATION
Fig. 5
Drained triaxial tests on lawlder clay, grout and plastic concrete.
Tests Sb, 22 at 3 % strain per hour.
Test 17LT, 39-day incremental dead-load test.

F/CORE 6-1 SAL DER /1E110 ,EPA /RS — CO4PAT/8/Z/TY 01


REPAIR NAMR/ALS WITH coRE AwATER/AZ

PA7'4 A-Rom. 1444,0#4,7 47. At 0770)


234

SECTION H

RISK OF CRACKING IN OLD DAMS

H.1. GENERAL

This matter is dealt with only briefly in this


report and statements consist mainly of recapitulation of
previous data. A general discussion on performance of dams
with time was presented by Little (1971). Figure H-1- is
taken from his paper and shows that most, dam failures occur
in the first 5 years of use but that failure can occur even
after 60 years of service. Hence, although the first five

years is usually the most critical in the life of a dam and


observations should be carried out carefully and more frequently
during this period, failure is possible at any stage of a dam's
life and a regular program of careful observations is required
for the life of the project.

In many cases records of design, construction


and subsequent performance for older dams are sketchy,
spread throughout the filing systems of many organisations
or lost. Records for all dams should be collected, collated

and readily available to any inspectors. These records may


then be kept up to date with each inspection.

In evaluating the possible risk or need for repair


the factors outlined in Section G.1 (a) must be taken into
account.
235

This section only refers to possible cracking


or piping of a dam core. Other factors such as overtopping,
surface erosion, instability, abutment and foundation
deterioration and deterioration of structures should also
be taken into consideration in checking the safety of old

dams.

H.2. POSSIBLE LONG TERM PROBLEMS

(a) Foundation Settlement

Where a dam is founded on clay foundations


drainage of the clay under loads applied by the embankment
can take many years. Similarly the advance of the seepage
water interface through such material will be very slow.
Hence settlements due to these causes may continue for many

years after completion of the dam.


Leaching of soluble material from the foundation
of embankments can lead to collapse in the foundation and

consequent cracking. Clevenger (1974) refers to an 80 feet


high dam that cracked because gypsum was leached out of
joints in a shale foundation leading to settlement of the

foundation.
In an alluvial foundation consisting of different
materials there could be adjacent materials that do not
meet the filter criteria relative to each other. In rock
foundations there are freqently joints filled with clay or

rock flour. When seepage gradients are increased in these


236

foundations erosion of the fine material is possible.


Over the long term this could lead to local instabilities
and cracking.

In or near to some dam foundations mining


operations are being or have been carried out. In some
cases this can lead to subsidence of the dam foundation.
Little (1971) refers to a dam where coal mining in the
foundation lead to subsidence and cracking of the dam core.

(b) Embankment Settlement

Most of the settlement in a dam will usually


occur during construction and first impounding. Therefore
it is most common for cracks due to embankment settlement
to occur in this period. However, consolidation, creep
or cyclic loading may result in settlements continuing
long after the first impounding. Cracks may occur in a
dam core as long as settlement continues.

Sherard (1972) notes that although cracks in


small dams in the U.S. commonly develop within a year of
dam completion, there are cases where significant cracks
have first developed as much as 5 years after completion.
He also reports that at Virginia Ranch Dam cracks occurred
in the crest at the closure section after the dam had
performed satisfactorily for 3 years. East Branch Dam
also performed satisfactorily for 41/2 years before excessive

seepage lead to the need for repairs. In this case cracking


was associated with an irregularity in the foundation so
was almost certainly due to continuing settlements. Jenkins
and Bankofier (1972) report that although a minor amount of
237

seepage occurred at Hills Creek Dam during the first impounding


it soon disappeared indicating healing of the crack. No
further seepage occurred until 6 years later when repairs
were needed because of the sudden appearance of excessive
seepage.. Again the crack was located near an abutment
irregularity. Hence cracks due to continuing settlement
may occur without any warning after the embankment has been
in service for several years and without any change to the
operating or other characteristics of the project.
As mentioned previously in a low permeability
core pore pressures may never have fallen below steady
seepage values even though the embankment is many years
old. If the reservoir level is kept low for a long period
the pore pressures will drop, effective stresses rise and
additional settlements will result. Cracking due to this
cause was described by Bishop (1946).
(c) Cohesive Bonds
The formation of cohesive bonds has been discussed

in detail in Section D.4. Bonds could form in either the


core or downstream filter. Cohesion in the filter is by
far the most critical.
Bonds may develop due to:
(1) Secondary Consolidation
(2) Cementation - Sherard et al (1963) suggest that
cementation occurred at Vermillion Dam

(3) Weathering of the material in the zone


(4) Washing of fines into the filter

(5) Bacterial action


238

Specific examples have not been found of problems


due to development of cohesion in filters. It is of
interest to note, however, that at Hills Creek Dam the
seepage during initial impounding quickly disappeared,
indicating healing of the crack and 6 years later erosion
and enlargement of a crack occurred in the same location.
(Jenkins and Bankofier (1972) ). One possible explanation
is that the filter functioned satisfactorily during initial
impounding but didn't 6 years later.
(d) Other Effects of Chemical Action

Where the core consists of material soluble in


the seepage water deterioration of the core will continue
as seepage continues. Anagnosti (1970) describes the case
of a lateritic soil that dissolved in the seepage water
and resulted in excessive seepage through the core.
If the chemical nature of the seepage water causes
dispersion of the core material the effect may not become
known for many years. As discussed previously (Sections C.4.
and D.3.) the behaviour of a dispersed soil adjacent to a
granular filter is not known but could lead to piping of the
embankment.

(e) Inadequate Maintenance

Inadequate maintenance of the project could result


in:
(1) Blocked exits to drainage blankets
(2) Blocked or uncleared surface drains
239

(3) Non repair of erosion channels on the downstream face


of a dam - Sherard et al (1972 (a) ) refer to this
as being the cause of failure at Middle Clear Boggy

Creek Dam

(4) Non treatment of small seepage areas on the downstream


face of a dam

( ) Presence of animal burrows in the downstream shell


of the dam

(6) Cracks and sink holes being left untreated


(7) Non repair of damage or local failure to other structures
associated with the dam. Little (1971) refers to
damage to Lluest Wen Dam being due to local failure
of the structure passing beneath the dam.

H.3. INSPECTIONS:

(a) Office Work:

As mentioned previously it is essential that all


available data on the dam to be inspected be compiled and
be available for examination prior to the field inspection.
All data on geometry of the embankment, materials used in
the embankment, modifications to the embankment, previous
problems with the embankment and all previous instrumentation
data and other observations should be available. Should
inadequate data be available an investigation program may
be required to fill in any gaps.

Before proceeding to the field the adequacy of the

design to resist cracking should be checked. This relates


240

mainly to filter design and the capability of the filter


to prevent rapid failure if cracking should occur. A
suitable laboratory testing program may be required to
check this important matter.
A thorough examination of all data is required
to determine possible problem areas. In some cases
adequate data on geometry and zoning of the dam will not
be available. In such cases field investigations are
required to determine these. Prior to carrying out field
work a comprehensive inspection program should be drawn
up based on office work and previous problems.

(b) Field Inspection

The program for the field inspection will


depend on the previous history of the project and
possible problem areas. However, in all inspections the
following items are of interest:
(1) Any seepage from any location. If there is any,
a program of observation and sampling should be
initiated immediately
(2) Any cracks or subsidences on the dam surface

(3) Drainage systems, both internal and surface, should


be functioning correctly
(4) Damage to any part of the project such as erosion
channels in the downstream face, animal burrows in
the downstream face, damage to the slope• protection,
broken pipes or damage to other structures
241

(5) Signs of dispersive material as discussed in


Section C.4.
(6) Any sign of recurrence of previous problems

(7) The effects of any changes to the initial project


layout

H.4. MINOR WORKS REQUIRED

(a) Repair Works

Minor repair works will consist of carrying out


the maintenance work described in Section H.2. (d). This
will be primarily to:

(1) Establish adequate drainage facilities for internal


drains and surface run off for the entire dam

(2) Repair surface protection, cracks, sink holes,


erosion gullies and animal burrows wherever they
occur on the dam

(3) Place filt.ers and suitable weighting zones over any


seepage issuing from the downstream side of the dam
(4) Repair any local damage to structures associated with
the dam

(b) Additional Work

Where there is inadequate data concerning the


the zoning of and material in the embankment an investigation
program consisting of exploratory drilling, trenching and
laboratory testing may be required to adequately define
242

these properties.
If the analysis of seepage water or presence
of deposits at the outlet of the drainage chimney or any
other factor indicates the possibility of cementation of
the downstream filter then samples of the filter are required
to check this. As the most probably location of such
cementation is near the bottom of the dam samples should
be taken for the full depth of the filter, if possible.
Where there is any indication of possible
dispersion of the core material measures must be undertaken
to correct this. As mentioned in Section G this requires
that the seepage water be rich in bivalent ions and this
can be achieved by blanketing the upstream slope with a
suitable material.
(c) Installation of Monitoring Devices
It is considered that wherever possible projects
should include devices to measure seepage through the
embankment and monuments to measure the external movements
of the embankment. Where these do not exist on old structures
suitable devices should be installed. The nature, number
and location of seepage measuring devices will depend on
local conditions. Movement measuring monuments should be
located on the dam crest and downstream shell. Initial
measurements of elevation related to a suitable bench mark,
and distance between adjacent monuments should be taken and
a regular program for subsequent measurements drawn up.
Where there is doubt about the safety of the
structure it may be necessary to install a more elaborate
instrumentation system to monitor the performance. This
243

matter is discussed in SectJon F.

H.5. EFFECTS OF CHANGES TO EXISTING PROJECTS

(a) General

Where a dam has functioned satisfactorily for


many years the most likely cause of cracking or piping is
a change that is made to the project. This change may
be due to additional construction work or may be due to some
less obvious cause such as change in the chemistry of the
reservoir water due to effluent being emitted upstream or
coal mining in the dam foundation.
Prior to modifying the structure in any way it
•is essential that the adequacy of the present design be fully
checked both analytically and by field inspection. All
modifications should then take into account all effects of
that modification relative to cracking.
Where changes are less obvious they can only
be observed by adequate monitoring and attention to general
. industrial and other action in the area. In most cases,

however, there will be some indication of adverse behaviour


in the seepage or movement measurements.

(b) Raising of the Dam or Reservoir Level

Bishop (1946) examined the upper part of a puddle


clay core and found that above top water line it was badly
cracked and discoloured due to drying out, weathering and
bacterial action. Such a zone would be very pervious and
probably allow excessive seepage. Woodburn and Cullaloe
reservoirs experienced seepage in this area when the water
244

level was raised (Hamilton (1973) ). Hence it is considered


that before raising either the embankment or reservoir
level suitable explorations should be carried out to check
the upper part of the core.
Other problems that could occur due to the
raising of an old dam are:

(1) The increased height will lead to additional loading


of the existing dam and foundation which will lead
to additional settlements
(2) The increased water level will lead to saturation
of the core and shells of the old dam above present
maximum water level with possible collapse of these
on saturation

(3) The increased water level leads to increased possi-


bility of hydraulic fracture
(4) The difference in material properties between the
new and old dam could lead to problems at the interface
due to arching or differential movements.

(c) Lower Water Levels

As mentioned on several occasions a reduced


upstream water level allows pore pressures in the core of
an embankment to fall below previous minimum values. This
causes a volume reduction in the core which can result in
cracks forming or increased arching.

(d) Change in Chemistry of Reservoir Water

A change in the reservoir water could lead to

the following:
245

(1) Cause the soil to change from the flocculated to


the deflocculated state. This occurred at the
Lakes. Entrance Dam (Aitchison and Wood (1965) )

(2) Initiate leaching of the core


(3) Cause cementation of the core or filters

(e) Repair Works


Any repair works carried out to other parts of
the project should be carefully examined to ensure that
they do not adversely affect the embankment. Several
examples are given below but many other operations could

also have adverse effects.


Remedial grouting of the abutment could lead

to hydraulic fracture in the dam core thus leading to


layers of grout in the core and possible open cracks

through the core. It could also lead to contamination


of the filter or drainage zones with grout.
Holes drilled into the core of the dam could
lead to hydraulic fracturing with consequent problems as

discussed in Section F.7.


Excavation of drainage tunnels or adits in
the foundation of the dam could lead to additional settlements

which could cause cracking.

H.6. EARTHQUAKE EFFECTS

In any area there are certain loadings or


displacements that can be applied to an embankment due to
246

an earthquake. Investigations should be carried out


for each dam to establish these and the adequacy of
the dam to resist such loadings checked as described
in Section B.10.
247

70

• 50

O

2. 40
' AiIII All &mills Construe Iona d
.
I 4) firs flee of
years
30 4510
life only (70)

410
411 ■
20
Seep ge

r r incidents (54) Seep be


. ncidents
10 4 . (34)
• ' .., f
5
Andir — .." ., /1r/
./. / ••)/ /
Air As. - . ,. • r.
0 5 ID
'teen from coast uctIon

FiCURE 1--/-I RELARONSIIIP gETIVER/ DM1 /WI URG S


Afia) 77/i7R ACE

19,?rq FROld Z/77 ('?7,)


248

SECTION I

SUMMARY AND CONCLUSIONS

GENERAL

(a) All embankment dams should be designed and constructed

so that the probability of cracking is reduced to a


minimum. However, it is not yet possible to ensure
that a dam will never crack.
(b) In order to ensure the safety' of a dam the downstream

filter system of a dam should be designed and constructed


to prevent erosion of any crack that may form.
(c) For dams with cores likely to disperse the only method

of safe design is to chemically ensure no dispersion


occurs. Filter action in these circumstances requires
further examination.
(d) To date no dams have failed due to cracking caused

by earthquakes. Where dams are located on a fault


or in a seismically active area due allowance must
be made for possible displacements that may occur
across or along the dam.
(e) Where cracks are observed during construction of a

dam they should be repaired. However, for a dam


designed to resist cracking, no additional work will
normally be required.
(f) Where problems occur due to cracking of a dam during
impounding of the reservoir and the dam is designed
249

to resist cracking sufficient time must be allowed

for the crack to heal itself. Draining and repair


work is only required if healing does not occur in

a reasonable time.

(g) Methods of monitoring for cracking in dams should


be included from the very beginning of the design

of an embankment.

(h) Old dams which are still in use should be thoroughly


checked to ensure that they will be safe if a crack

occurs.

(i) Many areas exist where additional knowledge is


required. In such cases a conservative design
approach must be adopted.

1.2 BEST DESIGN PRACTICE TO MINIMIZE CRACKING

This matter is discussed fully in Section B.

A summary is included here:

(a) Select core material that is most crack resistant


(b) Select dam location to avoid faults in foundation
(c) Consider arching dam upstream and locating dam
upstream of constriction in river

(d) Smooth out abutments


(e) Design shape of structures to reduce tensile stresses
to a minimum

(f) Consider inclining core upstream


(g) Use wide core with uniform side slopes
(h) If the foundation is soft remove if possible and
consolidate if not

250

(1) Treat rock foundation to ensure no fractures


beneath the core
(j) Where fractures may occur in rock beheath a dam

core due to earthquake or other causes use well


graded core material against rock
.(k) Treat core material as necessary to ensure no

dispersion occurs
(1) Place bulk of core at Proctor optimum moisture
content compacted as densely as possible
(m) Use finite element analyses or centrifugal testing

to define areas of possible cracking and use more


flexible material in . these locations
(n) Use finite element analyses to check for arching

of core between shells and manner in which this can be


minimized.
(o) Ensure no cracking during construction due to

pauses in construction, drying of crest or leaks from


abutment grouting. If cracks do occur repair them
where possible
(p) Do not drill holes into the dam core using methods

employing drilling fluids in uncased holes unless


absolutely essential. Where holes are drilled into

the core care should be taken to ensure no cracking


is caused. If hydraulic fracture does occur then

suitable repairs should be carried out


(q) In dry climates prevent core from drying out for
lifetime of project by suitable protection
(r) Ensure adequate compaction and watering of shells
251

to reduce collapse on saturation to a minimum


but do not overcompact to keep arching to a

minimum

(s) Make closure sections large enough to allow the


use of machines for compaction. Side slopes on
closure sections should be smooth and as flat as

possible

(t) Arrange construction schedule so that parts of the


dam crest that are likely to crack are built last

(u) Alternatively to (t) above overbuild dam in places


where cracking could occur

(v) Build parts of the dam on a more compressible foundation


first and allow settlement to take place before

building closure sections

(w) Where there is a large step in the rock foundation


of the dam pause at the level of the step to enable
settlement to occur before raising the embankment

further
(x) Fill reservoir as slowly as possible

1.3. METHODS OF ENSURING SAFETY OF DAM SHOULD CRACKING


OCCUR

(a) The most important requirement is that the downstream


filter be properly designed to prevent significant
erosion of material from the core should a crack occur.

(b) The capacity of the downstream drains must be such


that large seepage flows associated with cracks will
252

not result in downstream water pressures in excess


of those assumed in stability analyses
(c) The foundation and abutments must be carefully

treated to ensure that no seepage path can develop


which bypasses the downstream filter
(d) Zones downstream of the downstream filter must not

contain voids so large that the filter can be washed


into them
(e) Use a properly designed upstream transition

(f) Make core, filters and transition as wide as possible


(g) Install systems to monitor cracking and establish

a regular program of reading and analysis


(h) In earthquake areas core, filter and transition zones

must be sufficiently wide to ensure an adequate width


of zone remains even if maximum possible earthquake
displacement occurs
) ,The embankment dam must be properly maintained to

ensure that no drainage zones become blocked, all


surface drains function correctly, erosion damage
is repaired, animal burrows are filled in and damage
to structures through the dam is quickly repaired

1.4 AREAS REQUIRING FURTHER INVESTIGATIONS

(a) Tensile strength envelope


(1) Investigations required to determine correct shape
(2) Mechanism of effective stress relative to tensile
failure

(3) Effect of unloading versus loading on shape of tensile


253

failure envelope

(b) Non cohesive material.

(1) Effect of overconsolidation on cohesive intercept


of otherwise cohesionless material
(2) Arching under lateral stresses or due to particle
shape and arrangement in non-cohesive material

(3) Cementation effects due to chemical or bacterial


action
(4) Effect of silt content and nature on cohesive behaviour
of otherwise cohesionless soils

(5) The time required for a cohesionless material that


has cohesive characteristics due to negative pore
pressure effects to return to its non-cohesive state

on saturation
(6) Suitable tests for non-cohesive material for use
as filter material

(c) Hydraulic fracture

(1) Mechanism of drained hydraulic fracture


(2) Finite element analyses to study drained stress path
in core of dam on impounding relative to path postulated
for drained hydraulic fracture

(3) Use of centrifugal testing to model hydraulic fracture


mechanisms
(4) Laboratory testing to model both drained and undrained
hydraulic fracture in the presence of discontinuities
and significant shear stresses
254

(5) Investigations to find if there is a limiting depth


in an embankment below which hydraulic fracture will not
occur
(6) The effect of stress concentrations on drained and
undrained hydraulic fracture

(d) Self Healing Mechanisms

(1) Volume change characteristics of well graded, compacted


soils when they are eroded into cracks with loss of
fine material
(2) Swelling and shrinkage characteristics of soil related
to plasticity, permeability and moisture content

(3) Volume change characteristics of core material when


stress paths leading to cracking are followed
(4) Mechanism by which graded filters are built up on
downstream filters
(5) Study of optimum properties for upstream transitions

(e) Investigations to check on the design of


protective filters on the basis of permeability and minimum
size of core material

(f) Dispersive Material

(1) Behaviour at downstream filter when there is no crack


(2) Mechanism of dispersion due to replacement of pore

water by reservoir water and test to check for this


255

(3) Methods of filtering dispersed material


(4) Study of conditions leading to dispersive soils and
presence of such soils in the U.K.

(g) Optimum density-moisture content


relationship for various places in the core

(h) Optimum compaction for shells, transitions


and filters
256

APPENDIX A - References

Aitchison G.D., Ingles 0.G. and Wood C.C.- (1964), "Post Con-

struction Deflocculation as a Contributory Factor in

the Failure of Earth Dams", Proc. 4th Aust. & N.Z. Conf.

on Soil Mechanics and Foundation Engineering, p275.

Aitchison G.D. and Wood C.C. (1965), "Some Interactions of Com-

paction, Permeability and Post Construction Deflocculation

Affecting the Probability of Piping Failure in Small

Earth Dams", 6th Int. Conf. on Soil Mechanics and Foun-

dation Engineering, Montreal, Vol II, p442.

Ajaz A. and Parry R.H.G. (1975), "Stress Strain Behaviour of

Two Clays in Tension and Compression", Geot. Vol XXV,

No 3, Sept 1975, p495.

Anagnosti P. (1970), "A Lateritic Soil as a Watertight Dam Core",

10th ICOLD, Q36, R14, Vol 1 p225.

Appleton B., "Can Dam-crest Pond Keep Clay Core Damp?", New

Civil Engineer, Sept 1976, p29

A.S.C.E. Committee on Earth and Rockfill Dams (1967), "Defensive

Design Procedures against Cracking", A.S.C.E., Journal

of Soil Mechanics and Foundations Division, May 1967,

p129

Bansi Lal (1968), "A Three Dimensional Tensile Strength for

Cohesive Soil", Journal India National Society Soil

Mechanics and Foundation Engineers, Vol. 7.

Bassett (1973), "Centrifugal Model Tests of Embankments on Soft

Alluvial Foundations", 8th Int. Conf. on Soil Mechanics

and Foundation Engineering, Vol 2.2, p23.

Bertram G.E. (1940), "An Experimental Investigation of Protective

Filters", Publication Graduate School of Engineering,


257

Harvard Publication No 267, Vol 6, 1939-1940, SM No 7.

Bertram G.E. (1967), "Experiences with Seepage Control Measures

in Earth and Rockfill Dams", Transactions, 9th I.C.O.L.D.,

Vol 3, p91.

Bertram G.E. (1972), "Field Tests for Compacted Rockfill",

Embankment-Dam Engineering, Casagrande Volume, John Wiley

& Sons, pl.

Binnie G. (1902), 10th Edition of Encyclopaedia Britannica.

Bishop A.W. (1946), "The Leakage of a Clay Core Wall", Trans-

actions, Institute of Water Engineers, Vol 51, p97-116.

Bishop A.W. (1952), "The. Stability of Earth Dams", Ph.D. Thesis,

Univ. of London.

Bishop A.W. (1966), "The Strength of Soils as Engineering

Materials", Sixth Rankine Lecture, Geot. Vol XVI, No 2,

p89.

Bishop and Garga (1969), "Drained Tension Tests on London Clay",

Geot., Vol 19, No 2, p309.

Bishop A.W. and Vaughan P.R. (1962), "Selset Reservoir - Design

and Performance of Embankment", I.C.E. proceedings,

Vol 21, p305.

Bjerrum and Anderson (1972), "In Situ Measurements of Lateral

Pressures in Clays", Proceedings 5th European Conf. of

Int. Soc. of Soil Mechanics and Foundation Engineers,

Madrid, Vol 1, pll to 20.

Bjerrum L., Nash J.K., Kennard R.M. and Gibson R.E. (1972),

"Hydraulic Fracturing in Field Permeability Testing",

Geot. Vol XXII, No 2, p319.

Bonilla. (1970), "Surface Faulting and Related Effects", Earth-

quake Engineering, Edited by Weigel, p47.


258

Calhoun C.C. (1972), "Development of Design Criteria and

Acceptance Specifications for Plastic Filter Cloths",

U.S. Army Engineers, Waterways Experimental Station,

Technical Report S-72-7.

Carlyle (1965), "Shek Pik Dam", ProceedingsICE (London), Vol 30,

March 1965.

Casagrande A. (1950), "Notes on the Design of Earth Dams",

Boston Society of Civil Engineers, Contributions to Soil

Mechanics 1941-53, p231.

Casagrande A. (1969), Discussion 7th Int. Conf. on Soil Mechanics

and Foundation Engineering, Vol 3, p301.

Casagrande A., Wilson S.D. and. Schwantes E.D. (1972), "The

Baldwin Hills Failure in Retrospect", Proceedings of.

the Speciality Conf. on the Performance of Earth and

Earth Supported Structures, ASCE, Vol 1, Part 1 p551.

Chaudhary (1955), "Compressibility Characteristics of Compacted

Soils" M.Sc. Thesis, University of London.

Civil Engineering (ASCE) (1977), February, Vol 47, No 2, p8.

Clevenger W.A. (1974), "When is Foundation Seepage Unsafe",

Engineering Foundation Conf. on Inspection, Maintenance

and Rehabilitation of Old Dams, p570.

Conlon R.J. (1966), "Landslide on the Toulnustouc River, Quebec",

Canadian Geot. Journal, Vol 3, No 3, p113.

Cooke J.B. (1973), Discussion 11th ICOLD, Vol 6, p389.

Covarrubias (1969), "Cracking of Earth and Rockfill Dams",

Harvard Soil Mechanics Series No 82, April 1969.

Coxon R.E. and Crook D.E. (1976), "Some Simple Approaches to

Leakage Detection in Dams", 12th ICOLD, Q45, R30, Vol 2,

p527.
259

Dallaire G. (1976), "Filter Fabrics: Bright Future in Road

and Highway Construction", Civil Engineering, ASCE,

May, Vol 46, No 5, p61.

Davachi M. (1978), Ph.D. Thesis, London University, in prepa-

ration.

Davidenkoff R. (1955), "De la Composition des Filtres dans des

Barrages en Terre", Trans. 5th ICOLD, Vol 1 p385.

Dayte K. (1968), "Detailing Earth Dams to Minimize Cracking

Hazards", Symp. on Earth and Rockfill Dams, Indian

National Society of Large Dams, Vol 1, p269.

De Mello V. (1977), "Reflections on Design Decisions of Practical

Significance to Embankment Dams", Seventeenth Rankine Lecture

Dumas J.C., Larocque G.S. and Lebel M. (1970), "Mesures Preven-

tatives Contre la Fissuration du Noyau du Barrage Principal

d'Outardes 4", 10th ICOLD, Q36, R28, Vol 1 p445.

Eisenstein Z. (1974), "Application of Finite Element Method to

Analysis of Earth Dams", First Brazilian Conference on

Finite Element Methods Applied to Soil Mechanics, p457.

Engineering News Record (1963), "Dam in Jam gets Plastic Surgery",

E.N.R., Feb. 14 1963, p45.

Field (1923), "Failure of Apishapa Dam in Colorado", Engineering

' News Record, Sept 13, 1923, p418.

Galloway J.H.H. (1970), "Subsidence and Piping in Matahina Dam,

Rangitaiki River, Rotorua, New Zealand", Trans. 10th ICOLD,

Vol 6, p769.

Gilg B. (1970), "Apparition de Fissures dans la Digue de Matt-

mark", 10th ICOLD, Q36, R11, Vol 1, p189.

Gopalakrishnayya A.V. (1973), "Analysis of Cracking of Earth

Dams", PhD Thesis, University of Alberta, Edmonton.


260

Gordon J.L. and Duguid D.R. (1970), "Experiences with Cracking

at Duncan Dam", 10th ICOLD Q36, R29, Vol 1, p469.

Griffiths (1973), Letter to CIRIA describing cracking at Guma

Dam, Sierra Leone.

Haimson B. (1968), "Hydraulic Fracturing in Porous and Non

Porous Rock and its Potential for Determining In Situ

Stresses at Great Depth", PhD Thesis, University of

Minnesota.

Hamilton D.M. (1973), Letter to CIRIA describing seepage at

Woodburn and Cullaloe Reservoirs.

Harrison (1972), Letter to CIRIA describing seepage at Torside

and Rhodeswood Reservoirs.

Hasegawa and Ikeuti (1966), "On the Tensile Strength of Disturbed

Soils", Symp. on Rheology and Soil Mechanics, Grenoble,

Editors Kravtchenko and Sirieys, p405.

Ingles 0.G., Lang J.G. and Richards B.G. (1968), "Pre-equilibrium

Observations on the Reconstructed Flagstaff Gully Dam",

Symp. on Earth and Rockfill Dams, Talwarra-Punjab, India

Vol 1, p162.

Ingles 0.G., and Wood C.C. (1963), "The Recognition of Failure

in Earth Dams by Aerial Survey", Australian Journal of

Science, Vol 26, No 11,.p355.

Insley and Hillis (1965), "Triaxial Shear Characteristics of a

Compacted Glacial Till under Unusually High Confining

Pressures", 6th Int. Conf. on Soil Mechanics and Foun-

dation Engineering, Montreal, Vol 1, p244.

James J.P. and Wickham R.T. (1970), "Failure and Rectification

of the Morwell No 2 Fire Service Reservoir", Bulletin of

the Australian National Committee on Large Dams, July

1970, p20.
261

Jenkins J.D. and Bankofier A.M. (1972), "Hills Creek Dam Seepage

Correction", Purdue Conf. on Performance of Earth and

Earth.Supported Structures, Vol 1, Part 1, p723.

Karpoff K.P. (1955), "The Use of Laboratory Tests to Develop

Design Criteria for Protective Filters", Proc. A.S.T.M.,

Vol 55, p1183.

Kassiff and Henkin (1967), "Engineering and Physic() - Chemical

Properties Affecting Piping Failure of Low, Loess Dams

on the Neger", 3rd Asian Conf. on Soil Mechanics and

Foundation Engineering, Haifa, Vol 1, p13.

Kassiff, Zaslaysky and Zeitlen (1965), "Analysis of Filter

Requirements for Compacted Clays", 6th Int. Conf. on Soil

Mechanics and Foundation Engineering, Montreal, Vol II,

p495.

Kennard M.F., Penman A.D. and Vaughan P.R. (1967), "Stress and

Strain Measurements in the Clay Core at Balderhead Dam",

9th ICOLD, Istamboul, Vol 3, p129.

Kezdi (1973), "Tensile and Flexural Strength of Earth Dam

Materials", 11th ICOLD, Vol III, p165.

Kjaernsli B. (1973), Discussion at 11th ICOLD,' Vol 5, p476.

Kjaernsli B. and Torblaa I. (1968), "Leakage through Horizontal

• Cracks in the Core of Hyttejuvet Dam", Publication No 80,

Norwegian Geotechnlcal Institute, p39.

Knill J.L. and Best R. (1970), "Mineralogical, Chemical and

Geotechnical Studies of Core Materials in the Kainji and

Shek Pik Dams", Proc. of the 1st Int. Congress of Inter-

national Association of Engineering Geologists, p1247.

Krishnayya.G., Eisenstein Zand Morgenstern N. (1974), "Behaviour

of Compacted Soil in Tension", A.S.C.E., Journal of the


262

Soil Mechanics and Foundations Division, Sept. 1974, p1051.

Kudlik, Nosek, Pruska and Stastny (1973), "The Use of a Plastic

Foil for Reconstruction of an Earth Dam", 11th ICOLD,

Q42, R22, Vol 3, p379.

Kulhawy F.H. and Gurtowski T.M. (1976), "Load Transfer and

Hydraulic Fracturing in Zoned Dams", ASCE, Geotechnical

Engineering Division, Vol 102, GT9, p963.

Laing J.M. (1971), "A Finite Element Method for Investigating

Cracking in Embankment Dams", PhD Thesis, University of

London.

Lambe T.W. (1967), "Stress Path Method", ASCE, Journal of Soil

Mechanics and Foundations Division, Vol 93, No SM6,

p309-331.

Lambe T.W. and Whitman R.V. (1969), "Soil Mechanics", John Wiley

& Sons, Inc.

Lane K.S. (1963), Discussion on paper by Leonards and Narain

(1963), ASCE, Journal of. Soil Mechanics and Foundations

Division, Nov 1963, p109.

Leonards G.A. and Altschaeffl A.G. (1971), Discussion, ASCE,

Journal of Soil Mechanics and Foundations Division,

Vol 97, SM1, p269.

Leonards G.A. and Narain J. (1963), "Flexibility of Clay and

Cracking of Earth Dams", ASCE, Journal of Soil Mechanics

and Foundations Division, March 1963, p47.

Little A.L. (1967), "Laterites", Proc. 3rd Asian Conf. on Soil

Mechanics and Foundation Engineering, Sept 1967, p61.

Little A.L. (1971), "Problems of Embankment Dams", News and Views,

May 1971, British National Committee on Large Dams, p5.


263

Lofquist (1951), "Earth Pressure in a Thin Impervious Core",

4th ICOLD, R13, QI3, Vol 1, p99.

Lofquist (1955), Discussion 5th ICOLD, Vol 3, p22.

Londe P. (1970), Discussion, 10th ICOLD, Montreal, Vol 5, p384.

Lovenbury H.T. (1970), "The Detection of Leakage through the

Core of an Existing Dam", Symp. on Field Instrumentation

in Geotechnical Engineering, Butterworths London, p240.

Lowe III J. (1970), "General Report on Q36, Recent Developments

in the Design and Construction of Earth and Rockfill

Dams", 10th ICOLD, Vol 5, pl.

Lowe III J. (1970), Discussion, 10th ICOLD, Vol VI, p365 and 432.

Mac Kellar D.C.B. (1970), Discussion, 10th ICOLD, Vol 6, p783.

Maksimovic M. (1973), "Optimum Position of the Central Clay Core

of a Rockfill Dam in Respect to Arching and Hydraulic

Fracture", 11th ICOLD, Vol 3, p789.-

Mantz P.A. (1975), "Low Transport Stages by Water Streams of

Fine, Cohesionless, Granular and Flakey Sediments",

PhD Thesis, University of London.

Marsal R.J. (1960), "Earth Dams in Mexico", 1st Pan American Conf.

on Soil Mechanics and Foundation Engineering", Vol III,

p1294.

Marsal R.J. and Arellano R. (1967), "Performance of El Infiernillo

Dam", ASCE, Journal of the Soil Mechanics and Foundations

Division, Vol 93, p265.

Marsal R.J. and Pohlenz W. (1972), "The Failure of Laguna Dam",

Purdue Conf. on Performance of Earth and Earth Supported

Structures, Vol 1, Part 1, p489.

.Marsal R.J. and Ramirez L. (1967), Same as Marsal and Arellano

(1967).
264

Massarsch K.R. and Broms B.B. (1976), "Lateral Earth Pressures

at Rest in Soft Clay", ASCE, Geotechnical Division,

Vol 102, GT 10, p1041.

Narain J. and Rawat P.C. (1970), "Tensile Strength of Compacted

Soils", ASCE, Journal of Soil Mechanics and Foundations

Division, Vol 96, SM6., p2185.

Nilsson T. (1964), Discussion, 8th ICOLD, Vol 5, p458.

Nobari E.S. and Duncan J.M. (1972), "Effect of Reservoir Filling

on Stresses and Movements in Earth and Rockfill Dams",

U.S. Corps of Engineers, Report No S-72-2.

Nobari E.S., Lee K.L. and Duncan J.M. (1973), "Hydraulic Frac-

turing in Zoned Earth and Rockfill Dams", U.S. Corps of

Engineers, Report No S-73-2.

Nonveiller (1973), Discussion, 11th ICOLD, Vol 5, p505.

Olson R.E. (1974), "Shearing Strengths of Kaolinite, Illite and

Montmorillonite", ASCE, Journal of Soil Mechanics and

Foundations Division, Vol 100, GT11, p1215.

Parry R.H.G. and Nadarajah V. (1974), "Anisotropy in a Natural

Soft. Clayey Silt", Engineering Geology, 8, p287.

Patrick J.G. (1967), "Post Construction Behaviour of Round Butte

Dam", ASCE, Journal of the Soil Mechanics and Foundations

Division, Vol 93, p251.

Plessis J.G. (1974), "Dam Safety. Laws and Dam Safety in South

Africa", Proceedings on Engineeflng Foundations Conference

on Inspection Maintenance and Rehabilitation of Old

Dams., p241.

Pope R.J. (1967), "Evaluation of Cougar Dam Embankment Performance"

ASCE, Journal Soil Mechanics and Foundations Division,

Vol 93, p231.


265

Rallings R.A. (1966), "An Investigation into the Causes of

Failure of Farm Dams in the Brigalow Belt of Central

Queensland", Water Research Foundation of Australia,

Bulletin No 10.

Ramathan and Raman (1974), "Split Tensile Strength of Cohesive

Soils", Japanese Society of Soil Mechanics and Foundation

Engineers, Soils and Foundations, Vol 14, No 1, p71.

Seemel R.N. (1976), "Plastic Filter Fabrics Challenging the

Conventional Granular Filter", Civil Engineering, ASCE,

March, Vol 46, No 3, p57.

Sherard J.L. (1953), "Influence of Soil Properties and Construction

Methods on the Performance of Homogeneous Earth Dams",

D.Sc. Thesis, Harvard University also Technical Memo 645,

U.S. Bureau of Reclamation.

Sherard J.L. (1967), "Earthquake Considerations in Earth Dam

Design", ASCE, Journal of Soil Mechanics and Foundations

Division, July 1967, p377.

Sherard J.L. (1972), "Embankment Dam Cracking" Embankment Dam

Engineering, Casagrande Volume, John Wiley & Sons, p271.

Sherard, Cluff and Allen (1974), "Potentially Active Faults in

Dam Foundations", Geot., Vol 24, No 3, p367.

Sherard J.L., Decker R.S. and Ryker N.L. (1972a), "Hydraulic

Fracturing in Low Dams of Dispersive Clay", Purdue Conf.

on Performance of Earth and Earth Supported Structures,

Vol 1, Part 1, p653.

Sherard J.L., Decker R.S. and Ryker N.L. (1972b), "Piping in

Earth Dams of Dispersive Clays", Purdue Conf. on Performance

of Earth and Earth Supported Structures, Vol 1, Part 1, p589


266

Sherard J.L., Dunnigan L.P., Decker R.S. and Steel E.F. (1976a),

"Pinhole Test for Identifying Dispersive Soils", ASCE,

Journal of the Geotechnical Engineering Division, Vol 102,

No GT1, p69.

Sherard J.L., Dunnigan L.P. and Decker R.S. (1976b), "Identifi-

cation and Nature of Dispersive Soils", ASCE, Journal

of the Geotechnical Engineering Division, Vol 102, No GT4,

p287.

Sherard J.L., Woodward R.J., Gizienski S.F. and Clevenger W.A.

(1963), "Earth and Earth-Rock Dams, Engineering Problems

of Design and Construction", John Wiley and Sons Inc.

Sigvaldason 0.T. and Tawil A.H. (1972), "Stress Investigations

for the Lower Notch Earth and Rock-Fill Dam", Int. Symp.

on Experimental Mechanics, University of Waterloo, Canada,

June 1972.

Skempton A.W. and Northey R.D. (1953), "The Sensitivity of Clays",

Geot. Vol 3, No 1 p30.

Speedie M.G. (1970), Discussion 10th ICOLD, Vol 6, p408.

Taylor H. (1969), "W.A.C. Bennett Dam", The Engineering Journal,

Vol 52, Part 10, p25.

Taylor H. and Morgan G.C. (1970), "Measures Taken to Limit the

Possible Development of Cracks in a High Earthfill Dam",

10th ICOLD, Q36, R40, p679.

Terzaghi K. (1922), "Der Grundbruch au Stanwerken and seine

Verhiitung", Die Wasserkraft, 17, p445 (Reprinted in "From

Theory to Practice in Soil Mechanics", New York, John

Wiley and Sons (1960) p114).

Terzaghi K. (1953), Discussion, Vol 3, p217, 3rd Int. Conf. on

Soil Mechanics and Foundation Engineering.


267

Terzaghi K. and Lacroix Y. (1964), "Mission Dam, an Earthfill

Dam on a Highly Compressible Foundation", Geot. Vol 14,

No 1,. p13.

Terzaghi K. and Peck R. (1967), "Soil Mechanics in Engineering

Practice", 2nd Ed. John Wiley and Sons, New York.

Tomasson H. (1976), "The Opening of Tectonic Fractures at the

Langalda Dam", 12th ICOLD, Q44, R5, Vol 1, p75.

Truscott E.G. (1977), "Behaviour of Embankment Dams", Part 1,

PhD Thesis, University of London.

Tschebotarioff, Ward and De Phillippe (1953), "The Tensile

Strength of Disturbed and Recompacted Soils", 3rd Int. •

Conf. of Soil Mechanics and Foundation Engineering, Vol 1,

p207.

U.S.B.R. (1968), "Earth Manual", United States Department of

the Interior, Bureau of Reclamation,.U.S. ,Government

Printing Office.

U.S.C.E. (1953), "Filter Experiments and Design Criteria",

Corps of Engineers, U.S. Army, Waterways Experiment

Station, Vicksburg, Technical Memorandum No 3-360.

U.S.C.E. (1955), "Drainage and Erosion Control - Subsurface

Drainage Facilities for Airfields", Part XIII, Chapter 2,

Engineering Manual, Military Construction, Corps of

Engineers, U.S. Army.

Vargas M. and Hsu S.J.C. (1970), "The Use of Vertical Core Drains

in Brazilian Earth Dams", 10th ICOLD, Q36, R36, Vol 1,

p599.

Vaughan P.R. (1970), "Cracking of Clay Cores of Dams", Surveyor

(London), Jan. 30, 1970, p32.


268

Vaughan P.R. (1972), Discussion, 5th European Conf. on Soil

Mechanics and Foundation Engineering, Madrid, Vol 2, p72.

Vaughan P.R. (1973), "The Measurement of Pore Pressure with

Piezometers", Proc. Symp. on Field Instrumentation in

Geotechnical Engineering, Butterworths, London p411.

Vaughan P.R. (1976), "La Fisuracion Y Erosion Interna de Nucleos

de Arcilla de Presus de Tierras Y el Proyecto de Filtros

Para Protegerlos", Boletin de la Sociedad Espanola de

Meca.nica del Suelo Y Ctmentaciones, No 23, Sept/Oct 1976,

p23-34.

Vaughan P.R. (1977), Private Correspondence.

Vaughan P.R., Kluth D.J., Leonard M.W. and Pradoura H.H.M. (1970),

"Cracking and Erosion of the Rolled Clay Core of Balder-

head Dam and the Remedial Works Adopted for its Repair",

• 10th ICOLD Q36, R5, Vol 1, p73.

Vaughan P.R. and Kennard M.F. (1972), "Earth Pressures at a

Junction between an Embankment Dam and a Concrete Dam"

Proc 5th European Conf. on Soil Mechanics, Madrid, Vol 1,

p215.

Vaughan P.R., Lovenbury H.T. and Horswill P. (1975), "The Design,

Construction and Performance of Cow Green Embankment Dam",

Geot. Vol 25,No 3, p555.

Vestad (1976), "Viddalsvatn Dam, A History of Leakages and

Investigation", 12th ICOLD, Q45, R22, Vol 2, p369.

Vutsel V.I., Listrovoy P.P., Malysher M.P., and Shcherbina V.I. (1973),

"Measures Providing Impermeability of the Nurek Dam",

11th ICOLD, Q42, R32, Vol 3, p551.

Walbancke H.J. (1975), "Pore Pressures in Clay Embankments and

Cuttings", PhD Thesis, University of London.


269

Webster J.L. (1970), "Mica Dam, Designed with Special Attention

to the Control of Cracking", 10th ICOLD, Q36, R30, Vol 1,

p487.

Wegmann E. (1922), "The Design and Construction of Dams",

"Failures of Earthern Dams", Chapman and Hall, London,

p246.

Werneck M.L.G. (1974), "Field Observations of the Behaviour of

the Upper Lias Clay at the Empingham Dam Site", PhD Thesis

University of London.

Wesely L.D. (1975), "Influence of Stress Path and Anisotropy

on the Behaviour of Soft Alluvial Clay", PhD Thesis,

University of London.

Wilson D.W. (1972), "Deformation of Earth and Rockfill Dams",

Embankment Dam Engineering, Casagrande Volume John Wiley

& Sons, p365.

Wood D.,Kjaernali B. and Koeg K. (1976), "Thoughts Concerning

the Unusual Behaviour of Hyttejuvet Dam", 12th ICOLD,

Q45, R23, Vol 2, p391.

Yousef M.S., el Ramli A.H. and el Remery M. (1965), "Relation-

ships between Shear. Strength, Consolidation, Liquid Limit

and Plastic Limit for Remoulded Clays", Proc. 6th Int.

Conf. on Soil Mechanics and Foundation Engineering,

Montreal, Vol 1, p126.


270

APPENDIX B - SUMMARY OF TENSILE TEST RESULTS

(1) AJAZ and PARRY (1975)

Type of. Test: Beam and Direct Tensile on Remoulded Samples

Soils Tested: Gault and Balderhead Clays

Test Results:

TENSILE STRENGTH
CLAY FAILURE STRAIN %
L L P.I. (KN/m2 )
SOIL TYPE • • FRACTION
% %
BEAM DIRECT BEAM DIRECT

GAULT 73 39 56 50 to 110 30 to 80 0.5 to 1.6 0.06 to 0.5

BALDERHEAD 34 14 25 100 to 160 20 to 70 0.14 to 1.6 0.1 to 0.5

Testing Speed:

Beam Tests - Stress Controlled, 12 to 22 days per test

Direct Tests - Strain Cohfrolled, 3.6 to 10-4 mm per minute

Stress Controlled, li to 2 days per test.

Discussion:

Largest failure strain with largest moisture content (0.M.C.

+ 4.5 percent for Balderhead clay, 0.M.C. + 7 percent for

Gault clay). Compared results with unconfined compressive

strength.

(2) BISHOP and GARGA (1969)

Type of Test: Triaxial Tests on Hourglass Test Piece of

Saturated, Undisturbed Sample

Soils Tested: London Clay

Tensile Strength: 0.27 to 0.34 Kg/cm2 (26 to 33 KN/m2 )

Failure Strain: 2 to 17%

Testing Speed: 0.00005% per minute (drained test)


271

Discussion:

Tests carried out at confining pressures of 3 to 10 p.s.i.,

all failures tensile, showed modified Griffith theory

relating tensile strength,unconfined compressive strength

and friction coefficient applied.

(3) CONLON (1966)

Type of Test: Direct Tensile on Saturated, Undisturbed

Samples

Soils Tested: Sensitive Clay

Tensile Strength: Undrained 0.17 and 2.5 p.s.i. (1 and


2
17 KN/m ).

Drained 2.5. p.s.i. (17 KN/m2 ).

Failure Strain: About 0.3%

Testing Speed: Undrained 0.04% per minute.

Drained 0.0008% per minute.

Discussion: Testing method considered to be low standard.

(4) HASEGAWA and IKEUTI (1966)

Type of Test: Direct Tensile on Unsaturated, Remoulded

Samples

Soils Tested: Clay, L.L. 98%, P.I. 80%


2
Tensile Strength: From 0.03 to 0.24 Kg/cm2 (3 to 24 KN/m ).

Failure Strain: 0.2 to 4.5%

Testing Speed: Not stated, but probably rather fast.

Discussion:

Found that as moisture content increased from 0.M.C. - 2

to 0.M.C. + 18 tensile strength reduced and failure strain

increased.
272

(5) KRISHNAYYA, EISENSTEIN and MORGENSTERN (1974), and

GOPALAKRISHNAYYA (1973)

Type of Test: Indirect Tensile on Unsaturated, Remoulded

Samples

Soils Tested: Core material, for Mica Dam, L.L.18%, P.I.3.5%

Tensile Strength: 0.04 to 0.25 Kg/cm (4 to 25 KN/m 2 )

Failure Strain: 0.2 to 3.0%

Testing Speed: Most tests at 0.005 inches/minute but some

tests at other speeds to check rate effect.

Discussion:

Found that tensile strength reduced and failure strain

increased with increasing moisture content. An increase

in compactive effort resulted in an increase in failurp

strain, an increase in strength below 0.M.C. and no change

in strength above 0.M.C.. The addition of 6 percent

bentonite resulted in a large increase in tensile strength

and reduction in failure strain at 0.M.C.. Found reason-

able relationships between ratio of strengths, failure

strains and moduli of tensile and unconfined compressive

tests with moisture content.

(6) LEONARDS and NARAIN (1963)

Type of Test: Beam Tests on Unsaturated Remoulded Samples

Soils Tested: 6 types with L.L. from non plastic to 72%

and P.I. as high as 45%

Tensile Strength: 0.3 to 2.7 Kg/cm2 (30 to 265 KN/m2 )

Failure Strain: 0.05 to 0.33%

Testing Speed: Short term, 2 days per test

Long term, 4 weeks per test

Very long term, 6 months per test


273

Discussion:

Found an increase in moisture content from 2 to 3% dry of

O.M.C. to about 0.M.C. substantially increases the flex-

ibility, increases above 0.M.C. (to 3% wet of 0.M.C.)

results in little improvement in flexibility.

Increases in compactive effort substantially decreases

the flexibility.

In general highly plastic clays are more flexible than

clays of low plasticity, but the relative flexibility of

the low plasticity soils cannot be, distinguished on basis

of plasticity.

Found no relationship between unconfined compressive strength

and tensile strength.

(7) NARAIN and RAWAT (1970)

Type of Test: Indirect Tensile on Unsaturated Remoulded

Samples

Soils Tested: 6 types, L.L. from 25 to 105%, P.I. from 7

to 72%

Tensile Strength: 0.35 to 0.70 Kg/cm2 (34 to 69 KN/m2 )

Failure Strain and Testing Speed: Not stated

Discussion:

Plotted tensile strength against moisture content and

showed maximum tensile strength at moisture content less

than 0.M.C.. Showed ratio of unconfined compressive

strength divided by tensile strength increased with

increasing L.L. and P.I., especially at higher plasticities.

(8) NONVEILLER (1973)

Type of Test: Beam Test on Unsaturated Remoulded Samples


274

Soil Tested: Core material for Sklope Dam, L.L. 60%

Tensile Strength: Not calculated

Failure Strain: Maximum of 0.4%

Testing Speed: 1 test per day

Discussion:

Failure strain increases with moisture content to a

maximum at about 8% above 0.M.C.. Failure strain increases

with compactive effort.

(9) RAMANTHAN.and RAMAN (1974)

Type of Test: Indirect Tensile on Unsaturated, Remoulded

Samples

Soils Tested: Kaolinite, Illite and other soils, L.L. 28

to 61%, P.I. 14 to 41%

Tensile Strength: 0.14 to 0.92 Kg/cm2 (14 to 90 KN/m2 )

Failure Strain: Not measured

Testing Speed: 0.2 inches per minute

Discussion:

All samples tested dry of 0.M.C. to ensure brittle failure.

Reduction of tensile strength with increasing moisture

content. Showed the ratio of tensile strength to uncon-

fined compressive strength lay, between 0.2 to 0.4.

(10) TSCHEBOTARIOFF, WARD and DE PHILLIPPE (1953)

Type of Test: Direct Tensile on Unsaturated, Remoulded

Samples

Soils Tested: Bentonite, Illite, Kaolinite and Clay


2 2
Tensile Strength: 0.07 to 0.4 Kg/cm (7 to 39 KN/m )

Failure Strain: 0.15 to 3.4%


275

Testing Speed: Times of tests varied from about 5 to 430

minutes.

Discussion:

Related results to unconfined compressive strengths,

added sand to bentonite, no significant changes until

50% sand, increase in tensile strength until 85% sand

then strength drops to zero; failure strain constant to

about 75% sand then drops linearly to zero with increasing

sand.

You might also like