Professional Documents
Culture Documents
541]
The engineering behaviour of fill materials: the use, misuse and disuse
of case histories
J. A . C H A R L E S *
Un certain nombre de developpements survenant dans
notre monde moderne, y compris laugmentation de
lurbanisation, la presence dimportants plans damenagement du territoire ainsi que lelimination de quantites
considerables de dechets solides, produits par les exploitations minie`res et les activites industrielles, devraient
assurer que, dans un avenir previsible, les terrains rapportes auront une importance toujours majeure en geotechnique. Un terrain rapporte specialement amenage,
fortement compacte en couches de faible epaisseur et
dans des conditions controlees de tre`s pre`s, devrait etre
un materiau relativement uniforme, presentant un comportement facilement previsible sur la base de proprietes
moyennes. Par contraste, un terrain rapporte mal compacte, rempli de facon peu controlee avec des couches
profondes est susceptible detre peu compacte et de presenter une grande diversite de proprietes geotechniques :
le comportement de ces terrains rapportes a tre`s peu en
commun avec ces proprietes moyennes, et sera determine
en grande partie par des zones de remplissage a` letat
metastable et au comportement imprevisible. On examine
des etudes de cas de comportement de terrains rapportes
dans quatre dinteret pratique pour le geotechnicien :
(a) le comportement geotechnique de remblais de mises a`
ciel ouvert ; (b) le comportement de barrages en enrochement ; (c) lefficacite du traitement du sol ; et (d) evaluation de letat de barrages en remblai. Dans chacune de
ces categories, la communication se penche sur les resultats de controles a` long terme sur le terrain dans un
certain nombre de sites, et a` partir desquels elle tire des
conclusions generales. Lexamen detudes de cas necessite
un certain discernement, mais en depit de ses insuffisances, il fournit une compensation bien necessaire a` une
tendance excessive aux grandes theories.
A number of developments in our modern world, including increasing urbanisation, major land reclamation
schemes and the disposal of vast quantities of solid waste
generated by mining and industrial activities, should
ensure that, for the foreseeable future, fills will be of
increasing significance in geotechnical engineering. An
engineered fill, which has been heavily compacted in thin
layers under closely controlled conditions, should be a
relatively uniform material and have behaviour that is
easily predicted on the basis of average properties. In
contrast, poorly compacted fill dumped with little control
in deep lifts is likely to be in a loose state and exhibit
great diversity in its geotechnical properties: the behaviour of such heterogeneous fill will bear little relation to
average properties, and will be controlled largely by
zones of fill in a metastable state with unpredictable
behaviour. Case histories that include field measurementsthat is, quantitative dataare of particular value
in gaining an understanding of the performance of fill
materials. Case histories of fill behaviour are examined
in four areas of practical interest to the geotechnical
engineer: (a) the geotechnical behaviour of opencast
mining backfills; (b) the performance of rockfill dams;
(c) the effectiveness of ground treatment; and (d) the
condition assessment of embankment dams. In each of
these areas the lecture focuses on the results of long-term
field monitoring at a number of sites, from which some
general conclusions are drawn. Discernment is required
in the study of case histories, but despite shortcomings,
they provide a much needed counterweight to excessive
theorisation.
KEYWORDS: case history; compressibility; dams; deformation;
embankments; field instrumentation; ground improvement;
monitoring; rockfill; settlement; time dependence
INTRODUCTION
Importance of case histories
Quotations from great geotechnical men of the past have
often been included in Rankine Lectures. In a deviation from
this precedent, the following quotation is from a nineteenthcentury prime minister. Benjamin Disraeli wrote:
exaggeration can establish an important point, and biographiesthat is, accounts of the lives of real peoplecan give
an insight into the human condition at a particular time and
place that a general historical narrative, which is likely to be
heavily biased by the preconceived ideas of the historian,
may fail to do.
Substituting case history for biography, an analogous
recommendation to a geotechnical engineer would be: Study
no theory: read nothing but case histories, for that is actual
ground behaviour undistorted by preconceived theoretical
concepts. Such a recommendation would be unwise, for
several reasons.
541
542
CHARLES
Significance of fill
Mankind has been creating fill throughout recorded history. Some 4000 years ago, over a considerable period and
for purposes that we cannot now determine, the 40 m high
Silbury Hill in Wiltshire was carefully engineered in a series
of six stepped horizontal layers. Its complex internal structure was created by concentric rings of chalk block walls,
which together with radial walls, formed compartments that
were infilled with chalk rubble. Cross-sections of some
engineered fills constructed during the last 4000 years are
shown in Fig. 1, and details are provided in Table 1.
It is a sad reflection on human progress that, 3000 years
after the building of Silbury Hill, William the Conqueror
was throwing up mounds of earth, in what by then had
become England, on which to build crude fortifications for
military purposes that are only too easy to recognise. In
1069 a 15 m high mound was built on low-lying ground
adjacent to the river in York to facilitate the subjugation of
the north of England. The mound was built in horizontal
layers of fill comprising stones, gravel and clay. Initially the
mound provided a base for a timber structure; the stone
tower known as Cliffords Tower was built on the mound in
the middle of the thirteenth century. Major cracking of the
tower occurred in 131516 during severe floods, which
softened the fill.
It was only in the nineteenth century, in the great age of
embankment dam building, that earthworks as high as Silbury Hill were again constructed in England. In the twen-
Silbury
Hill
Brianne
Dam
50 m
Scale
Nurek Dam
BC
to
AD
2000
BC
to
AD
543
2000
Structure
Location
Purpose
Silbury Hill
Cliffords Tower Mound
Dale Dyke Dam
Brianne dam
Nurek Dam
Chek Lap Kok Airport
Wiltshire, England
York, England
Sheffield, England
Llandovery, Wales
Tajikistan
Hong Kong
Unknown
Military
Retain water
Retain water
Retain water
Platform for airport
Date built
pre-2000
1069
1864
1971
1980
1996
BC
Height: m
Volume:
106 m3
Surface area:
ha
40
15
29
90
300
25
0.25
0.04
0.4
2.0
58
194
2
0.4
4
7
60
1248
the
the
the
the
Since these are all large subjects that cannot be dealt with in
a comprehensive manner, each section focuses on the results
of BRE field monitoring at a number of sites, from which
some general conclusions are drawn.
GEOTECHNICAL BEHAVIOUR OF OPENCAST MINING
BACKFILLS
In 1949 the Building Research Station (now BRE) published Digest No. 9, Building on made-up ground or filling,
which stated that
Suitable sites for new buildings and estates in industrial
areas are becoming more difficult to find and it is more
frequently necessary to build on made-up ground or filling.
The passing of nearly 60 years has not invalidated this
statement, but the qualifying phrase in industrial areas is
now no longer needed. Opencast mining has been a major
producer of deep fills whose geotechnical behaviour is of
critical importance when restored opencast sites are considered for building development. The principal practical interest concerns the potential for long-term settlement of the
backfill. Where use of a restored site for building purposes
is foreseen prior to backfilling, the fill should be placed in
layers and heavily compacted to an appropriate specification
under controlled conditions: such an engineered fill should
be reasonably uniform, with a potential for settlement that is
both limited and predictable. Where such a future use is not
foreseen, or was ignored and backfilling was carried out
with little control and without systematic compaction, the
situation is very different.
Creep settlement soil model
Early work on the settlement of fill was carried out at
BRE by Meyerhof (1951), who, from a literature review,
presented the long-term creep settlement data shown in Fig.
2. Although the settlement of the fill materials varied from
30% for domestic refuse to less than 1% for compacted
CHARLES
544
0
7
2
1. Well-compacted,
well-graded soil
5
2. Medium-compacted
rockfill
10
15
3. Lightly compacted
clay and chalk
20
4. Uncompacted sand
5. Uncompacted clay
25
6. Well-compacted
mixed refuse
30
rockfill, the creep rate of all the fills diminished with time,
leading to the conclusion that, if fill is left long enough, the
rate of settlement of the ground surface will become negligible. This comforting conclusion suggests a simple solution
to developing a filled site: leave the site for long enough
and significant settlement will cease. But is this so, and if it
is so, for how long must the site be left?
For many rockfill dams in the United States, Sowers et al.
(1965) found an approximately linear relationship between
crest settlement and the logarithm of time that had elapsed
since the middle of the construction period, as shown in Fig.
3. The values of the creep compression rate parameter
05
0
35
Lewis Smith
Dix
River
05
Nantahala
10
15
Rockfill
: %
UD
Dumped limestone
11
78
SC
Dumped graywacke
07
97
SC
Compacted sandstone 02
Dam
Date
Height: m
Type
Dix River
1925
84
Nantahala
1942
Lewis Smith
1961
s
H
4:34
t
t
mm=year
(2)
545
Traverse of surface
settlement stations
Borehole settlement gauge
N
C
C11
Lagoon
Pump
B2
B
Overburden heap
D1
D
Oldest
A9
fill
A
D15
Most recent
fill
500 m
Scale
Fill type
Horsley
Blindwells
Tamworth
Silt and
clay: %
rd :
Mg/m3
w: %
rs :
Mg/m3
n: %
Va: %
10
20
45
1.70
1.56
1.78
7
7
9
2.54
2.45
2.62
33
38
32
21
23
16
wP : %
wL : %
cu : kPa
23
17
41
28
150
100
Ilkeston
Corby
w: %
Mean
Range
19
18
1225
728
The table presents typical values of geotechnical properties to give an indication of fill type, but it
should be noted that the most significant property of these non-engineered fills is their heterogeneity.
rd , dry density; rs , particle density; n, porosity; Va , percentage air voids; w, water content; wP, plastic
limit; wL, liquid limit; cu , undrained shear strength.
CHARLES
546
Ground level:
mAOD
98.6
101.8
94.9
108.1
119.2
115.8
A9
B2
C11
D1
D15
E12
Rockhead:
mAOD
38.0
38.7
49.2
52.6
72.7
68
Fill depth:
m
Fill date
60.6
63.1
45.7
55.5
46.5
48
1961
1964
1965
1966
1970
1966
Inundated
depth: m
Settlement: m
46
45
35
31
11
17
19741977
19731992
0.31
0.33
0.06
0.10
0.15
0.35
0.40
0.50
0.11
0.10
0.31
0.79
Fill condition
Oldest
Deepest
Lagoon
Preloaded
Most recent
Intermediate age
1974 1976 1978 1980 1982 1984 1986 1988 1990 1992
0
30
100
C11
D15
300
Settlement: mm
A9
400
(a)
0
100
D1
200
300
400
C11
200
05
B2
10
15
500
E12
B2
600
700
E12
800
(b)
1974 1976 1978 1980 1982 1984 1986 1988 1990 1992
4 (491 m)
6 (370 m)
Settlement: mm
100
8 (258 m)
200
9 (197 m)
300
10 (138 m)
400
13 (0 m)
500
(a)
13
10
10
Groundwater level
9
Depth: mm
20
8
30
6
40
4
50
60
Fill
1 Bedrock
(b)
1975
1976
1977
1978
1979
(a)
67
05
10
13 GL
12
11
10
9
8
7
Fill
6
5
4
3
2
1 Bedrock
78
56
63 m
Vertical compression: %
547
15
20
4 5
89
910
(c)
(b)
CHARLES
548
0
Settlement: m
02
03
01
04
Vertical compression: %
0
1
2
05
20
Rise in
groundwater
level
40
60
Fig. 9. Settlement and vertical compression against depth at gauge B2, Horsley
10
100
11
10
200
D1
10
800
7
20
7
6
C11
30
5
5
4
40
3
2
50 1
1000
Settlement: mm
0
200
100
600
400
3
2
1
200
10
Collapse compression: %
20
collapse compression of opencast backfills that are predominantly clay has been investigated at sites at Ilkeston and
Corby, and some typical properties are given in Table 2(b).
The stiff clay backfill at Ilkeston was placed by scrapers
without any additional systematic compaction in 1959. There
was no water table within the backfill.
In 1973 a newly constructed block of eight two-storey
houses suffered some settlement when excavation for drains
began close to the north gable end and, following heavy
rain, further movement took place in the centre of the row
of houses. Soon all the houses in the block were affected,
and movements continued, although underpinning and pres-
2002
2003
1999
2000
2001
1996
1997
1998
1994
1995
1991
1992
1993
1988
1989
1990
1985
1986
1987
Settlement: mm
200
C
E
400
G
600
I
K
800
0
5
10
15
20
25
30
35
40
45
50
55
60
K
I
E
New
inundation
Backfill
Natural ground
1994
1995
1993
1991
1992
B
C
200
D
400
E
600
F
H
800
1990
1988
1989
1987
1986
1985
1984
1982
1983
1981
1980
1978
1979
1977
Settlement: mm
H
F
10
E
D
New
inundation
20
C
Backfill
30
in 24 h of filling the trenches with water, additional settlements of up to 50 mm had been recorded, confirming that
water penetrating into the opencast backfill via surface
trenches could cause significant collapse compression. Fig.
14 shows the settlement measured at various depths within
the fill at a magnet extensometer installed 1.5 m from two of
the trenches. Settlement occurred immediately water was put
into the trenches, and the compression was located between
magnet marker g, 0.5 m below ground level, and magnet
marker e, 7 m below ground level. Six days after the start
of the test 40 mm settlement had occurred, and the trenches
were backfilled. However, settlement continued at a significant rate. A temporary increase in the rate of settlement in
late 1976 followed a period of heavy rainfall. The average
vertical compression between magnet markers g and e
over the whole monitoring period was 2.6%.
549
B
A
Natural ground
sure grouting were carried out. A year after the houses were
completed, floor levels showed a maximum differential settlement of 0.14 m across the 9 m wide block, and the east
wall was 0.065 m out of plumb. The houses were never
occupied, and the block was demolished in 1982, by which
time there was an estimated total settlement of 0.3 m.
It was suspected that water penetrating into the fill
through drain trenches had initiated collapse compression
within the backfill, and a field inundation test was carried
out (Charles & Burford, 1987). When 3 m deep trenches
were filled with water to a depth of 1.8 m, the rates at which
water levels fell in the different trenches varied from as little
as 0.04 m/h (0.08 m3 /h) to as much as 1 m/h (2 m3 /h). With-
Settlement: mm
f
100
h
g
12 m
f
e
Backfill
d
gh
c
Natural
b ground
a
200 Inundation
test
CHARLES
550
100
f
d
c
b Natural
a ground
200
300
House
construction
Trenches
backfilled
Inundation test
commenced
gh
20
Settlement: mm
Settlement: mm
1975
1977
1979
1981
1983
1985
1974
1976
1978
1980
1982
1984
1986
0
h
g
f
24 m e Backfill
Settlement
due to
deep
mining
40
60
80
100
120
140
1980
1981
1982
1983
and, unless it can be established that the fill does not have
significant collapse potential (e.g. a previous inundation or
wet placement of the fill should have greatly reduced, if not
eliminated the potential for further collapse), there will be a
risk that collapse settlement could occur during or subsequent to building on the site and, consequently, either deep
foundations or ground treatment are likely to be required.
Collapse compression can result from a rising groundwater level, but may also occur above the water table from
downward infiltration of surface water or groundwater infiltration into the backfill through the high wall of an opencast
mine. There is likely to be some time dependence in the
response to wetting, particularly where inundation is due to
downward percolation of water or in a clay fill. The field
experiments have demonstrated that water infiltration from
the ground surface via trenches can cause collapse compression in clay fill that continues for many years, and can be as
large as 6%. The non-uniform response to inundation at
different locations within clay fill indicated that it was not a
practical form of ground treatment.
The engineering behaviour of poorly compacted heterogeneous fills is difficult to predict, and may bear little relationship to average values of geotechnical properties. Settlement
that damages buildings will be largely a function of the most
adverse properties encountered within the fill. Well-documented case histories, with long-term field measurements, provide
a helpful basis for assessing likely field performance.
Significant creep movements can occur in poorly compacted fill, and for mudstone/sandstone fill values of the
order of 1% are typically observed. Creep movements will
generally be relatively small, unless the fill is very deep or
has been placed quite recently. Where settlement predictions
are based, either explicitly or implicitly, on the simple creep
settlement model, building developments may appear to be
successful simply because inundation of the fill has not yet
occurred. However, at some restored opencast sites and other
types of filled site large collapse settlements have occurred
subsequent to building development, with unpleasant consequences for the buildings (Charles & Watts, 2001).
PERFORMANCE OF ROCKFILL DAMS
There is an instructive contrast between the performance
of poorly compacted opencast backfills and the behaviour of
551
(a)
(b)
(c)
(d)
Scammonden
Brianne
Winscar
Marchlyn
Megget
Roadford
Date built
1969
1971
1974
1979
1981
1989
Height: m
73
90
53
47
56
41
Rockfill
Vibrating roller
Layer depth: m
Number of passes
Weight: t
0.9
0.9
1.7
1.0
0.4
0.45
5
4
4
4
4
8
11.5
13.5
13.5
13.5
5.5
9.1
Sandstone/mudstone
Slatey mudstone
Sandstone
Slate
Gravel
Sandstone/mudstone
rd : Mg/m3
w: %
rs : Mg/m3
n: %
Va: %
2.02
2.35
2.03
2.25
2.07
7
3
6
4
4
2.69
2.75
2.60
2.81
2.74
25
15
22
20
24
11
7
10
11
15
rd , dry density of fill; rs , particle density; n, porosity; Va , percentage air voids; w, water
content.
CHARLES
Construction deformations
The compressibility of the rockfill materials was measured
in tests on samples heavily compacted in layers in 0.6 m and
1.0 m diameter oedometers. Rockfill with a maximum particle size of 125 mm was tested in the 1.0 m diameter
oedometer. Fig. 18 shows that the compressibilities of the
rockfills from Brianne, Scammonden and Winscar dams
were quite similar, but the gravel fill from Megget was much
less compressible.
Settlement occurs during rockfill placement owing to the
self-weight of the fill. The vertical strains measured in the
downstream rockfill shoulders at Brianne, Scammonden and
Winscar were quite similar, since the compressibility of the
rockfills measured in oedometer tests were not very different. The deformations at Megget were much smaller, because the gravel fill was much less compressible. This
difference is illustrated in Fig. 19, which shows the settlement during embankment construction on the centreline of
the Scammonden, Winscar and Megget dams. Brianne is not
included in the figure because the maximum construction
settlement of 2.3 m was principally a function of the behaviour of the soft central clay core, not of the rockfill
shoulders (Carlyle, 1973); the constructional horizontal
movements at Brianne induced by the lateral pressure from
the soft clay core at the interface between the clay core and
the granular filter are shown in Fig. 20.
Although the vertical strains at Brianne, Scammonden and
Winscar were quite similar, with a maximum of 34%
compression (Fig. 21), the pattern of horizontal strain at the
dams was very different. The horizontal strains shown in
Fig. 22 are for the upstream-downstream direction; horizontal strain along the axis of the dams was not measured. The
clay cores at Brianne and Scammonden were placed wet of
optimum water content and had high pore water pressures
and low effective stresses during construction. At the end of
1500
Vertical strain: %
552
4
W
Fig. 18. Compressibility of rockfills measured in large oedometer: B, Brianne slatey mudstone; M, Megget gravel; S,
Scammonden sandstone; W, Winscar sandstone
553
10
h/H
4
(a)
05
S
Me
W
1
2
3
3
0
05
Settlement: m
(b)
10
1
2
4
(c)
Top of
embankment
strain
(%):
80
70
60
50
40
03
Downstream movement
Settlement
04
05
06
03 02
01
30
(a)
20
10
Base of
0
embankment 0
05
10
Movement: m
(a)
15
05
10
Movement: m
(b)
01
01
02
02
0
(b)
02
04
0
(c)
The stressstrain properties of rockfills measured in onedimensional compression tests are generally non-linear, but
the internal distribution of settlement during embankment
construction can be predicted with little error using a
constant equivalent compressibility, and the maximum settlement occurring during construction can be related to the
constrained modulus (D). Assuming that a large-diameter
CHARLES
554
280
U
E
For D a constant DH :
!
2
H
smax 0:25
D H
(3)
260
smax
H2
0:195
D H
!
(4)
Post-construction deformations
Subsequent to the construction of a rockfill dam, movements will occur because of
10
S
smax: m
W
Eqn (4)
R
M
0
3
H 2/DH: m
B
220
200
100
200
Movements monitored during reservoir impounding illustrate the determining influence of the position and nature of
the watertight element within the embankment. At Brianne
the continuing settlement of the crest was largely a function
of the behaviour of the central clay corethat is, primary
consolidation due to dissipation of excess pore pressures
followed by secondary compression. The upstream fill was
submerged during impounding, thus reducing the effective
stresses within the fill. While in theory this should cause the
rockfill to undergo a slight expansion or heave, in practice
any small creep movements or collapse compression will
negate such a tendency. Fig. 24(a) shows that the points
labelled U on the upstream slope settled a very similar
amount to monuments E and F on the downstream slope
during reservoir impounding, indicating that the rockfill had
Eqn (3)
C
240
05
Elevation: m AOD
For D k(v )0 5 :
F
U
E
D
C
B
A
(b)
Settlement: mm
100
h/H
200
300
01
025
064
100
1
Time since end of construction: years
10
head at the location where the horizontal stress was measured. However, as shown in Fig. 26, the downstream deflection of the asphaltic diaphragm was very small owing to the
stiffness of the gravel fill.
At dams with upstream membranes the reservoir water
applies a loading normal to the membrane. First filling of
the reservoir causes major stress increases and embankment
deformations. Subsequent fluctuations in reservoir level affect the stresses in the rockfill in a similar way, but have a
smaller effect on embankment deformations, as the rockfill
is much stiffer under these reloading and unloading stresses.
When the water load is transmitted to the rockfill immediately beneath an upstream membrane the mean effective
C4
555
B5
18
0
16
120
100
80
140
Movement
scale
60
40
20
(a)
0
50 mm
A6
02
04
06
10
08
02
(b)
CHARLES
556
Rockfill
14 m
18 box sections housing E-Ls
8 at 05 m long
10 at 10 m long
Drainage layer
Sand waste
Dis
tan
ce
fro
:m
10
12
oe
mt
14
1
4
2
0
0
20
Predicted deflection
m
:m
40
60
D
c
efle
tion
Date
Res head: m
1 November 1989
2 April 1991
3 July 1994
80
146
33 (full)
32
100
120
02
W
nmax: m
01
The settlement of the crests of Brianne, Megget, Scammonden and Winscar dams has been monitored for long
periods following the end of construction. In Fig. 30 the
data are plotted as vertical strain. The vertical strain at
Brianne is twice as large as that at Scammonden and
Winscar, reflecting the fact that at Brianne the consolidation
of the clay core is the dominant effect. At Winscar the extra
loading during reservoir impounding was a significant factor:
the increased rate of settlement that occurred between six
and a half years and nine years after the end of construction
was associated with the first full filling of the reservoir in
19811982 after it had been emptied in 1980 for membrane
repairs. At Scammonden the consolidation of the clay core
had only a minor effect on crest settlement.
Quite large crest settlement continued at a reasonably
constant rate for some time after reservoir impounding at
Roadford, with no sign of the steady reduction in settlement
rate observed at the other four dams. This additional settlement has been attributed to an increase in water content in
the embankment fill (Evans & Wilson, 1992). The crest
settlement rate eventually reduced with time (Hopkins et al.,
2002); nevertheless, by May 2001 the crest had settled
535 mm since the end of construction, of which 425 mm
had occurred since impounding began, corresponding to
vertical strains of 1.3% and 1.0% respectively, which is
surprisingly large. Vaughan (1994a) estimated that a little
more than 1% collapse compression had occurred at a slow
rate.
Figure 31 shows the compressibility of two samples of
Roadford rockfill measured in tests in a 1 m diameter
oedometer. The properties of the two samples are given in
Table 6. Sample a was heavily compacted at a relatively
high water content, and sample b was less heavily compacted at a lower water content. Until sample b was
inundated the compressibility of the two samples was not
dissimilar, but when sample b was inundated a collapse
compression of 2.6% occurred. Sample a was not inundated, but as the initial air voids were only about 4% there
would not have been any collapse potential. The two dashed
lines for sample b on the figure give estimates of the
behaviour of sample b if it was not inundated at all and if
it was inundated before any load was applied. The initial
density and water content of sample b were close to the
average field values (Table 6).
In Fig. 32 the crest settlement data for the five dams are
plotted using a logarithmic timescale. There is generally a
uam
cf
Ma
20
0
05
smax: m
tion went only part way up the upstream slope: the hollow
circle in Fig. 29 represents the maximum measured deflection, and the solid circle is the estimated maximum deflection on the assumption that the deflected shape of the
membrane at Roadford was similar to that measured at
Winscar. The dashed line on the graph corresponds to a ratio
of nmax /smax 0.25, which would be expected if the constrained modulus controlling the deformation due to reservoir impounding was about twice as large as the modulus
controlling construction settlement.
02
s/H: %
S
04
06
B
R
08
15
Vertical strain: %
a
5
b
Inundation
10
Test a
Test b
Field
rd : Mg/m3
w: %
n: %
Va: %
2.16
2.05
2.07
7.9
4.7
4.4
21
25
24
4
15
15
05
0
20
M
557
Conclusions
Field measurements have confirmed that carefully controlled placement and compaction can produce rockfills with
relatively uniform geotechnical properties and predictable
behaviour. The case histories go some way towards establishing benchmarks for normal behaviour of different types of
rockfill dam. The monitored movements give a good indication of the deformations likely to occur during successive
stages in the life of a rockfill dam, and substantial departures from such behaviour in a dam could indicate the onset
of unsatisfactory behaviour.
Although rockfill behaviour is not elastic and not linear,
movements during embankment construction can be predicted using simple linear elastic models. The limiting factor
in making such predictions is not the sophistication of the
soil model used in the calculations, but rather the difficulty
of establishing representative soil parameters for a fill material containing large rock fragments. Movements during
10
02
Scammonden
Winscar
04
s/H: %
06
08
W
Uniform creep rate
05
10
Creep rate
proportional to
vertical stress
12
14
0
0
05
Long-term settlement
Long-term crest settlement
10
CHARLES
558
Settlement of
Settlement of houses (mm)
ground surface
induced by
During house Total during and after house
treatment: m
construction
construction to 1999
Mean
Preloading
Dynamic compactiony
Inundation{
No treatment
0.41
0.24
0.10
1.4
7.0
6.1
2.7
11
52
54
33
Maximum Minimum
25
74
149
53
5
23
30
14
300
10
559
10
5
0
0
May
June
July
August
10 m depth
30
40
50
Settlement: mm
20
200
43 m depth
400
22 m depth
Surface
600
Settlement: mm
200
400
600
Height of fill: m
10
15
20
25
Fill
Bedrock
CHARLES
560
0
Settlement: mm
200
400
3m
57 m
Vertical compression: %
4
2
9m
3m
57 m
9m
10
15
20
25
003
10
005
Hs/B
01
02
03
04 05
30
100
Settlement: mm
ze/Hs
3
Mean settlement of 5 m deep
old refuse fill subsequent to
construction of 3 m high
embankment
20
05
40
Settlement: mm
0
100
200
300
Dynamic
compaction
5
57 m surcharge
10
561
ground treatment can be an effective risk mitigation measure, there may be some residual risk, for example inundation of untreated fill at depth in a deep fill due to a rising
groundwater level, which can occur when pumping associated with deep mining ceases.
Natural ground
Puddle-clay-filled cut-off
trench
CHARLES
562
In his reply Rankine stressed the importance of the foundation, which should be sound rock, if practicable, and should
a rock foundation be unobtainable, firm impervious earth.
He added that It may be doubted whether any earthen
foundation is thoroughly to be relied on where the depth of
water exceeds 100 or 120 feet. Rankine also warned Tulloch
that It is not advisable to build a masonry dam on an
earthen foundation (Tulloch, 1872).
The response of the legislature to the 1864 disaster was
not rapid, and it took further loss of life occasioned by two
dam failures in 1925 to renew concerns about reservoir
safety (Charles, 2002). The 1930 Reservoirs (Safety Provisions) Act imposed a safety regime requiring an inspection
of large raised reservoirs by a qualified civil engineer,
known as the inspecting engineer, at intervals not exceeding
10 years. The 1975 Reservoirs Act introduced the additional
role of the supervising engineer, who, unlike the inspecting
engineer, has a continuous responsibility for the reservoir.
The term inspection was retained in the 1975 legislation,
but condition assessment or safety evaluation would have
given a better indication of what is required.
Surveillance and monitoring form important elements in
the safety management of old embankment dams. Careful
visual observations by personnel familiar with an embankment dam are a critical element now that most British dams
no longer have the benefit of a resident reservoir keeper.
Leakage measurement and settlement of the crest of the
embankment are the most common forms of routine monitoring. A sudden large increase in leakage is likely to give
a clear indication of the hazardous development of internal
erosion, but the significance of an increase in the rate of
settlement of the crest of the embankment is less easy to
assess.
Surface settlement and horizontal displacement can be
measured using precise surveying techniques, but in most
cases optical levelling is the simplest and most reliable way
of monitoring crest settlement. Table 8 provides information
on several old embankment dams where BRE has carried
out long-term measurements of crest settlement. The measurement of internal displacements in existing dams requires
the installation of relatively complex and expensive instrumentation in vertical boreholes: magnet extensometers have
been used for measuring vertical displacements, and inclinometers for measuring horizontal displacements. The monitoring of piezometric pressures in the fill forms an
important part of a detailed investigation, and requires
intrusive borings.
Crest settlement
A critical question relates to whether or not movements,
measured in some cases more than 100 years after the
construction of a dam, are due to an incipient malfunction
that could lead to failure if remedial action is not taken.
s
H log t2 =t1
(5)
Date
built
H: m
Watertight
element
Holmestyes
Ogden
Yateholme
Ramsden
Cwmwernderi
Walshaw Dean Lower
Challacombe
1840
1858
1872
1892
1901
1907
1945
25
25
17
25
22
22
15
ucb
cpcc
cpcc
cpcc
cpcc
cpcc
cpcc
Foundation cut-off
H, height of embankment; ucb, upstream clay blanket; cpcc, central puddle clay core.
563
Date
built
H: m
Measurement
period
SI : %
Megget
Scammonden
Brianne
Challacombe
Cwmwernderi
Holmestyes
1981
1969
1971
1945
1901
1840
56
73
90
15
22
25
cac
urcc
crcc
cpcc
cpcc
ucb
19811996
19691989
19711992
19811989
19811989
19922001
0.02
0.17
0.36
0.8
0.9
2.7
cac, central asphaltic core; cpcc, central puddle clay core; crcc, central rolled clay core;
ucb, upstream clay blanket; urcc, upstream rolled clay core; H, height of embankment; SI ,
settlement index.
1988
1989
1990
Reservoir level:
m below crest
0
5
10
15
Empty
20
185 m
Settlement: mm
20
126 m
40
65 m
60
Crest
80
Settlement units ( )H 2
028
05
D
0 013
(0%)
TWL
(50%)
50%
100%
25 m
(100%)
(a)
Heave units wH
D
(a)
0
(0%)
019 0
025
(50%)
100%
019 0
TWL
25 m
50%
(100%)
(b)
(b)
CHARLES
Drawdown depth: m
564
Jan
1990
0
Jan
1994
Jan
1998
Jan
2002
Nov
1991
Nov
1995
Nov
1999
10
20
Crest settlement: mm
25
50
75
(a)
(b)
Fig. 45. Monitored crest settlement during reservoir drawdown of dams with watertight element (after Tedd et al.,
2002): (a) Walshaw Dean Lower, central puddle clay core; (b) Holmestyles, upstream clay blanket
06
05
Strain: %
04
sm /H
32
50
150
200
03
Initial
loading
02
34
sp /H
01
00
10
15
Drawdown depth: m
20
25
Fig. 46. Mean vertical strain as a function of depth of drawdown for old puddle clay core embankment dams in BRE
Yorkshire Water study (after Tedd et al., 1997b): sm , maximum
settlement; sp , permanent settlement; sr , recovered settlement;
H, embankment height
Vertical strain: %
sr /H
B
36
Initial
unloading
D
38
40
Permanent
strain due to
4 major
reloading
unloading
cycles
A
First reloading
unloading cycle
C
Permanent
strain due to
4 minor
reloading
unloading
cycles
42
565
566
CHARLES
567
Owner
Consulting engineers
Brianne
Marchlyn
Megget
Welsh Water
Central Electricity Generating
Board
Lothian Regional Council
Roadford
Scammonden
Winscar
Challacombe
Cwmwernderi
Holmestyes
Ogden
Ramsden
Walshaw Dean Lower
Yateholme
failure, will be only too ready to point out the failure of the
designer to heed lessons from the past.
The misuse of case historiesthat is, the failure to draw
correct conclusions from case historiesis a more difficult
matter to rectify. The available case histories of a particular
type of geotechnical structure may not be representative. No
case history will be fully comprehensive: it will not provide
us with all the facts, and each case history will be influenced by the writers technical understanding, or lack of it.
In some cases evidence may have been carefully selected to
make a point that would not be supported by a more
balanced presentation of the evidence.
Discernment is required in the study of case histories, but
despite their shortcomings they provide a much needed
counterweight to excessive theorisation. This study of the
engineering behaviour of fill materials has attempted to
demonstrate the value of case histories, particularly those
that include long-term field measurements.
ACKNOWLEDGEMENTS
In theoretical studies it is quite possible for an individual
to make a major contribution with relatively little help from
others, whereas fieldwork of necessity is a team effort, and
in the work on the engineering behaviour of fill materials at
BRE many people have been involved. Initially I had the
privilege of working for Dr Arthur Penman; at a later stage I
collaborated with Dr Paul Tedd and latterly with Hilary
Skinner. Many of the early field surveying measurements
were made by Don Burford. Not only did Ken Watts carry
out much of the later fieldwork, but he also provided help in
the preparation of this lecture. Much of the work had its
origin in the time when Professor John Burland was Head of
the Geotechnics Division. To all these colleagues, and to
others who have rendered assistance, I acknowledge my debt
and express my gratitude. Thanks are also due to Dr Peter
Bonfield, the Chief Executive of BRE, and to other members
of the senior management for the use of BRE facilities in
preparing this lecture.
Field measurements require the collaboration of the owners of the structures and their consulting engineers, and
these are listed in Table 10 for the research work on the
various dams. Most of these owners and consulting engineers have changed their names during the long periods that
BRE has taken measurements, and it seemed appropriate to
acknowledge their assistance using their earlier rather than
their later names. Much of the work on dams was carried
out in Yorkshire, and the help given by Jim Claydon and
Andrew Robertshaw of Yorkshire Water is gratefully ac-
NOTATION
B width of surcharge
cu undrained shear strength
D constrained modulus
D H secant-constrained modulus for v H
H height of embankment or surcharge
h height of fill
hw reservoir head
n porosity
nmax maximum deflection of upstream membrane during reservoir
filling
ru pore pressure ratio
SI settlement index
s settlement
smax maximum settlement during embankment construction
sm maximum settlement on reservoir drawdown
sp permanent settlement on reservoir drawdown
sr recovered settlement on reservoir drawdown
t time since fill placement or loading
u pore pressure ratio
Va percentage air voids
w water content
wL liquid limit
wP plastic limit
ze depth of effectiveness of ground treatment
creep compression rate parameter
bulk unit weight
9 effective unit weight of submerged fill
d dry unit weight
s unit weight of surcharge fill
w unit weight of water
v vertical strain
r bulk density
rd dry density
rs particle density
v vertical stress
REFERENCES
Atkinson, J. H., Charles, J. A. & Mhach, H. K. (1990). Examination
of erosion resistance of clays in embankment dams. Q. J. Engng
Geol. 23, No. 2, 103108.
Baines, J. A., Newman, V. G., Hannah, I. W., Douglas, T. H.,
Carlyle, W. J., Jones, I. L., Eaton, D. M. & Zeronian, G. (1983).
Dinorwig pumped storage scheme. Proc. Instn Civil Engrs, Part
1 74, No. 4, 635718 (Discussion, 1985, 78, August, 919957).
568
CHARLES
569
VOTE OF THANKS
Professor J. B. BURLAND, Emeritus Professor, Imperial
College of Science, Technology and Medicine, London
Dr Andrew Charles is the fourth Rankine Lecturer from
the Building Research Establishment. If we look a little
more closely at the statistics we find that no less than 10
Rankine Lecturers spent their formative geotechnical years
at BRE. The precise definition of the phrase formative
geotechnical years is deliberately vague. I hope that my
colleagues at Imperial College will forgive me if I say that,
according to my interpretation of the meaning of that phrase,
570
CHARLES