Q. 98 – R.
COMMISSION INTERNATIONALE
DES GRANDS BARRAGES
-------
VINGT-CINQUIÈME CONGRÈS
DES GRANDS BARRAGES
Stavanger, Juin 2015
-------
THE UNIFIED PLOT APPROACH FOR ASSESSING INTERNAL EROSION – A
CASE STUDY OF THE GRUNDSJÖN DAM SINKHOLE EVENT (*)
Hans RÖNNQVIST
Lic. Eng., Luleå University of Technology
Sweden
R. JONATHAN FANNIN
Ph.D, Professor, University of British Columbia
Canada
Peter VIKLANDER
Ph.D, Adjunct Professor, Luleå University of Technology
SWEDEN
1. INTRODUCTION
An embankment dam normally comprises several zoned fill materials,
including a core, one or more filters and support fills, all of which are necessary
for the satisfactory performance of the dam. Although nearly impervious, the core
permits some seepage from the dissipation of the reservoir’s hydraulic head
acting on the upstream face of the dam. In adverse circumstances, the seepage
concentrates into leakage paths which may cause erosion within the zones of the
dam. Internal erosion occurs when soil particles migrate through the constrictions
of a zoned soil or into a neighbouring zone. Surpassed only by overtopping,
internal erosion is a major cause for the failure of embankment dams [1]. Four
mechanisms associated with the initiation of internal erosion are recognised by
ICOLD [1], namely i) concentrated leak erosion, ii) backward erosion, iii) contact
erosion and iv) suffusion. Suffusion, due to internal instability, erodes free fines
inside of a soil and may cause a property change; in contrast, backward erosion
(*) Approche par scénario visant à évaluer l’érosion interne – Étude de cas du trou
d’affaissement du barrage de Grundsjön
66
Q. 98 – R. 5
usually originates from an incompatible base-filter interface eroding progressively
backwards. The core is protected from internal erosion, albeit indirectly, by a
filter. An ineffective filter is unable to arrest an initiated erosion process, thus the
continuation of internal erosion is generally dependent on the filter action of the
zone located downstream of its base.
The literature offers several methods to evaluate the susceptibility of a
material to initiation and continuation of internal erosion [1]. In general, the
methods are empirically developed from laboratory studies with little or no
systematic comparison to field experience. However, in a study of empirical
methods for the assessment of filters in dams, Rönnqvist et al [2] assessed the
filter gradations compiled in a database of 80 existing embankment dams that
includes 23 dams which are reported to have experienced some form of internal
erosion. The methods examined were those of Kenney and Lau [3, 4] with its
adaption by Li and Fannin [5], and that of Burenkova [6] with its adaptation by
Wan and Fell [7], as well as an alternative method by Wan and Fell [8].
Furthermore, potential core-filter incompatibility was assessed by using the
method of Foster and Fell [9]. The comparative analysis of Rönnqvist et al [2]
found a relation between attributes of the filter gradation and the probable
occurrence of internal erosion in a dam. The finding suggests a unified plot of
internal instability and a filter’s capacity for soil retention may be useful as a
screening tool for internal erosion susceptibility in engineering practice.
This paper describes the unified plot approach for assessing internal
erosion susceptibility of existing dams. Thereafter, it reports a case study of the
Swedish Grundsjön dam and an evaluation of its sinkhole event in 1990 using the
unified plot approach.
2. THE UNIFIED PLOT APPROACH
A database was compiled by Rönnqvist [10] comprising 80 embankment
dams commissioned in the period 1941 to 1993, with heights in the general range
10 to 125 m. The database includes information on the core and filter zone of
each dam, and it comprises dams in Sweden (64), Canada (4), Norway (4), the
USA (3), Finland (2), Australia (2) and New Zealand (1). Common to all of the
dams is a core material of glacial till, with non-plastic or low-plasticity fines, in
contact with a granular filter. By the use of performance monitoring data from
published and, in some cases, unpublished reports, 23 of the dams were
identified as earth structures in which a probable occurrence of internal erosion
has been documented. For these dams, the performance records reveal one or
more incidents believed indicative of internal erosion (e.g., high piezometric water
levels in the downstream filter, increased seepage flow, muddy discharge, and in
many cases sinkhole activity). They are categorised as dams with “probable
occurrence of internal erosion” (denoted by capital letters in Fig. 1 and described
67
Q. 98 – R. 5
in detail in [2]). Thus, 57 are dams with no record of any deficiency, else have
exhibited a deficiency that cannot be attributed with any reasonable confidence to
seepage-related erosion phenomena (categorised as dams with “no indications
to-date of internal erosion” in Fig. 1).
Kenney and Lau [3, 4] proposed that the potential for internal stability be
evaluated from a shape analysis of the gradation curve over a designated portion
of its finer end. The finer grains in such a soil are smaller than available void
space in the coarse fraction allowing for erodible free fines in the under-filled
voids of the coarser fraction. The comparative analysis of Rönnqvist et al [2] of
available empirical criteria for internal stability assessment indicated that the
approach by Kenney-Lau was relatively more successful in identifying the dams
with probable occurrence of internal erosion (Fig. 1). Furthermore, the
examination of the database of 80 dams revealed that the greater the value of
D15max, the more likely a dam in the database has a probable occurrence of
internal erosion [2].
Fig. 1
Dimensionless unified plot of the relation between stability index of the filter
gradation and dam-specific ratios for base soil retention D15max/D15EE [2]
Scénario adimensionnel de la relation entre l'indice de stabilité correspondant à
la granulométrie du filtre et les coefficients - spécifiques aux barrages - de
rétention du sol de base
Analysis of the soil retention capacity of a filter using the method of Foster
and Fell [9] indicates that the Foster-Fell empirical boundaries for base soil
68
Q. 98 – R. 5
retention by a filter (i.e., a No Erosion (NE) boundary, an Excessive Erosion (EE)
boundary; and, a Continuing Erosion (CE) boundary) correlate well with the
inferred probable occurrence of internal erosion (Fig. 1). The range between the
NE and EE boundaries is termed, herein, the Some Erosion zone of base soil
retention. By taking all the available gradation curves for the filter of the dams into
account, gradation analysis established the most vulnerable gradation curve in
terms of stability indices H/Fmin and D15max, yielding one data point per dam.
Furthermore, the dimensionless plot is established by using the dam-specific EE
boundary to calculate the ratio D15max/D15EE (Fig. 1). Thus, combining the stability
index (H/F)min < 1 of Kenney and Lau [3, 4] with the soil retention capacity of a
filter (by using the filter boundaries proposed by Foster and Fell [9]), yields a
significant improvement in the identification of dams that exhibit deficiencies from
internal erosion, and thereby encourages use of the unified plot (Fig. 1) as a
screening tool for internal erosion susceptibility in engineering practice. The dam
denoted by “J” in Rönnqvist et al [2], as illustrated in Fig. 1, is the most vulnerable
data point for the Grundsjön dam, which will be studied in detail in the following.
3. THE GRUNDSJÖN DAM AND SINKHOLE EVENT
3.1. THE SINKHOLE EVENT AND INVESTIGATIONS
In September 1990, 18 years after commissioning, a crater-like depression
was discovered on the upstream edge of the crest of the Grundsjön Connecting
dam (Fig. 2). The depression measured 0.4 x 0.5 m on the surface, and closer
examination revealed it to extend approximately 1.6 m vertically into the dam
body. It was aligned with the upstream face of the core, and thus possibly a
sinkhole due to internal erosion [11]. In a first phase of examination in the fall of
1990, immediately after the discovery of the sinkhole, the crest of the dam was
removed exposing the crest of the core in the area of the sinkhole. Although
these initial test pits revealed relatively little about the extent of the damage, the
exploratory penetration tests conducted from the crest of the core, as well as
falling-head tests carried out in the piezometers, provided more information; the
core soil was loose locally and exhibited a high take of water during the infiltration
tests. Furthermore, the piezometric readings from the standpipes in the core
showed pore pressure levels higher than theoretically expected. These findings
indicated that the core was, indeed, damaged, possibly by internal erosion, but
additional investigations were needed to clarify the cause of the sinkhole and to
ensure a safe resumed operation of the reservoir. Thus, a second phase of
examination was undertaken in the spring of 1991 which involved a complete
uncovering of the core and filter over a length of 60 m. Excavation was carried
out in stages of 1 m lifts and at the lowest it reached down to +644m, 13 m below
the dam crest, completely exposing the filter and core (Fig. 3). Visual
observations are shown in Fig. 3. Several signs believed indicative of internal
69
Q. 98 – R. 5
erosion were observed, most notably the actual sinkhole, cavities in the core, wet
zones, core soils devoid of fines and the presence of a leakage path through the
core (observation 5, Fig. 3). Furthermore, there were observations indicative of
the continuation of erosion, e.g., filters washed out of fines (Fig. 3) as well as
probable contributing causes to the process (pockets of coarse filter of cobbles
possibly a consequence of segregation during placement). The dam was
carefully rebuilt in 1991, and was provided with an improved drainage system as
well as the implementation of an extensive monitoring program of instrumentation
[11, 12].
Fig. 2
The Grundsjön dam (in plan view) and connecting dam (in cross-section)
Le barrage de Grundsjön (vue en plan) et le barrage de liaison (vue en coupe)
1) Main dam 1) Barrage principal
2) Connecting dam 2) Barrage de liaison
3) Sinkhole location 3) Emplacement du trou d’affaissement
4) Core of glacial till 4) Noyau en matériau morainique
5) Filter of sand and gravel 5) Filtre de sable et gravier
6) Shell of gravel 6) Coque de gravier
7) Transition zone of gravel 7) Zone de transition en gravier
8) Shoulder of cobbles and stones 8) Recharge en galets et pierres
9) Transition zone of gravel 9) Zone de transition en gravier
10) Foundation of glacial till 10) Fondation en matériau morainique
70
Q. 98 – R. 5
a)
b)
Fig. 3
The excavation of the core in 1991 of the Grundsjön dam with a)
observations and b) photo evidence (adapted after [11, 12])
Excavation du noyau du barrage de Grundsjön en 1991 avec a) observations
et b) preuves photographiques (selon [11, 12])
1) Sinkhole 1) Trou d’affaissement
2) Large cavity in the core 2) Grande cavité dans le noyau
3) Fines washed out of filter with 3) Matériaux fins du filtre lessivés avec
evidence of core fines in fill, and présence de matériaux fins du noyau
pockets of cobbles dans le remblai, et galets
4) Fines washed out of filter 4) Matériaux fins du filtre lessivés
5) Fines washed out of core and leakage 5) Matériaux fins du noyau lessivés et
path chemins d’infiltration
6) Wet zones 6) Zones humides
7) Small cavities in the core 7) Petites cavités dans le noyau
8) Excavation limit 8) Limite d'excavation
71
Q. 98 – R. 5
3.2. DAM DESIGN AND CONSTRUCTION
The Grundsjön dam is located on the Ljusnan River in Sweden and put into
service in 1972. The embankment dams comprise the 1200 m long and 43 m
high Main dam, and the 600 m long and 15 m high Connecting dam, as shown on
the layout in Fig. 2, together with location of the sinkhole and a cross-section of
the Connecting dam which suffered the sinkhole. The core zone of the
embankment dams was founded on bedrock through a fully penetrating cut-off
trench, whereas the shoulders were placed on natural soils of morainic deposits
(Fig. 2). The bedrock was deep-grouted in one centrally-aligned to the core
injection row, with additional rows of surface grouting added where needed.
The core material (zone 4, Fig. 2) consists of glacial till which, according to
specifications, should have a maximum of 85 % < 2 mm and 15 to 50 % < 0.075 mm
(fines content) when regraded to a Dmax = 64 mm [11]. Although not tested, the
fines are assumed non-plastic. Larger stones were removed directly at the
borrow area, and further sorting was done by sieve buckets on site. The core soil
was spread by blading out piles of glacial till dumped in the vicinity of its intended
final location, and compacted was done at optimum water content of 7 % [11].
Achieved dry densities in-situ averaged at 94 % relative the maximum dry
densities from Modified Proctor tests (i.e., 2.1 to 2.2 ton/m 3); however, reports
show occasional relative densities as low as 88 % [11].
Against bedrock, a 0.5 m thick layer of contact glacial till (Dmax < 64 mm)
was placed in lifts of 0.2 m that were compacted with vibratory plates. Above the
contact layer, the glacial till was initially placed in 0.4 m lifts (with Dmax < 250 mm)
and compacted with six passes by the 9.0 T rollers. Subsequently, field tests
showed that satisfactory compaction was obtained at higher lifts, thus the lifts
were increased to 0.75 m (Dmax < 400 mm) and the compaction effort was
increased to eight passes by the 9.0 T rollers [11].
The downstream filter (zone 5, Fig. 2) is composed of sand and gravel,
which according to specifications, should have Dmax < 50 mm, a maximum of 10 %
< 0.25 mm and 30 % < 4 mm < 70 % [11]. The filter was sorted with sieve
buckets to Dmax = 64 mm; processing and handling made it segregate readily,
requiring special attention during placing. The filter was spread, placed and
compacted similarly to the core specifications. The shell of gravel (zone 6, Fig. 2)
was spread by blading out to 1 m thick layers, and compacted with eight passes
by the 9.0 T rollers, with a Dmax = 400 mm permitted. The erosion control below
the core is provided by the transition zone of gravel between the core and the
downstream face of the trench (zone 7, Fig. 2). This material was dumped from
the trench crest without any particular compaction, with a Dmax = 300 mm.
72
Q. 98 – R. 5
Fig. 4
Envelope of core gradations (< 4.75 mm) and as-placed filter gradations of the
Grundsjön dam
Enveloppe de granulométrie du noyau (< 4,75 mm) et du filtre en place sur le
barrage de Grundsjön
3.3. ASSUMED CAUSE FOR THE SINKHOLE AS PER 1991
The probable cause of the sinkhole as postulated at the time was the
combined effect of a series of events over a long period of time as follows: i)
wetting settlement of the lower part of the core during the first impoundment that
creates weak-zones at the interface to the arched upper part of the core, ii)
hydraulic fracturing and leakage through weak-zones and ineffective downstream
filter, iii) the formation of transverse zones through the core from backward
erosion causing significant leakage, and iv) the collapse of the core and
subsequent sinkhole formation [11, 13].
4. APPLICATION OF THE UNIFIED PLOT APPROACH
The gradations of the core and filter gradations are presented in Fig. 4.
They comprise 43 filter gradations and 27 core gradations taken from
construction records [11]. The core soil is a glacial till and when regraded on the
#4 sieve (i.e., 4.75 mm), as shown in Fig. 4, it has about 30 to 50 % < 0.075 mm
(i.e., fines content); less than 2 % < 0.002 mm (i.e., clay-size); and a d85 of 0.3 to
1.0 mm. The filter is a sandy gravel with 0.3 ≤ D15 ≤ 3.5 mm and Dmax = 64 mm.
73
Q. 98 – R. 5
There are no records as to whether larger Dmax were permitted in the filter, and
although the specifications stipulated 50 mm, forensic photo evidence indicates
that coarser fractions than this were used (Fig. 3).
Applying the Kenney and Lau criterion [3, 4], and the Li and Fannin
adaptation [5], Fig. 5 shows the H:F shape curves for the 43 filter gradations.
Extracting the stability index ((H/F)min) from the H:F shape curves, i.e., the
smallest value of H/F in the evaluation range 0 < F ≤ 20 %, reveals that 16 of the
43 gradations (37 %) are potentially unstable (i.e., (H/F)min < 1) with a lowest
stability index (H/F)min of 0.43. Many of the gradations are unstable on their finer
end (Fig. 6). The Grundsjön dam data for all 43 filter gradations are thus
superimposed over the Rönnqvist et al [2] database which comprises the single-
most vulnerable data point for each dam. Similarly, the method of Wan and Fell
[7], which evaluates internal instability from a characteristic value for the slope of
the coarser fraction (D90/D60) and the overall slope of the gradation curve
(D90/D15), establishes that 18 of the gradations are unstable, with the highest
probability of 90 % for one gradation (Fig. 7). The Foster and Fell filter
boundaries [9] analysis shows, moreover, that the dam-specific NE boundary is
0.7 mm for the Grundsjön dam (i.e., Group 2A base soil for which the fines
content is 35 to 85 % of base soil passing the No. 4 (4.75 mm) sieve), the EE
boundary is approximately 4.5 mm and the CE boundary is 18 mm
(i.e., D15 > 9 d95 after [14]).
Fig. 5
H:F shape curves of Grundsjön dam filter gradations superimposed over
Rönnqvist et al [2] database
Courbes H:F de granulométrie du filtre du barrage de Grundsjön superposées
à la base de données Rönnqvist et al [2]
74
Q. 98 – R. 5
Fig. 6
Stability indexes of Grundsjön dam filter gradations superimposed over
Rönnqvist et al [2] database
Indices de stabilité des types de granulométrie du filtre du barrage de
Grundsjön superposés à la base de données Rönnqvist et al [2]
Fig. 7
Probability of internal instability of the filter gradations of the Grundsjön dam
superimposed over Rönnqvist et al [2] database
Probabilité d'instabilité interne des types de granulométrie du filtre du barrage
de Grundsjön superposés à la base de données Rönnqvist et al [2]
75
Q. 98 – R. 5
Analysis of the unified plot (Fig. 8a) shows that a considerable number of
gradations locate in the No Erosion and internally stable zone, but there are
gradations on the coarser end of the envelope that clearly plot in the Some
erosion and internally unstable zone. This zone is where many dams with
probable occurrence of internal erosion are found to plot. In the dimensionless
unified plot (Fig. 8b), on a dam-specific basis, the data of the Grundsjön dam are
plotted in relation to its EE boundary of 4.5 mm: the majority of gradations plot in
the Some erosion zone (which is located in the range between the NE and EE
boundaries), which according to the method of Foster and Fell [9] would make it
equally likely that the filter will seal with Some erosion as with Large erosion in
the case of a concentrated leak through the core.
4.1. DISCUSSION
Strictly, from an assessment of the filter gradations reported in the
construction records, the filter satisfies the EE boundary (ergo the CE boundary).
According to Foster and Fell [9] that would make a scenario with no sealing of the
filter unlikely and a continuing erosion process highly unlikely as the filter appears
finer-grained than one which would allow unrestricted erosion (Fig. 8a and
Fig. 8b). However, taking into account that the filter is potentially unstable (Fig. 6
and Fig. 7), it may be coarser in-situ than expected from the reported as-placed
gradations. In an internally unstable soil, the finer fraction is susceptible to
erosion.
According to ICOLD [1] the amount of finer fraction can be located by
identifying the point of inflection of a gradation where there is a change in slope
of the sieve curve occurring at the transition from the coarse fraction to the finer
fraction. In addition, Wan and Fell [8] proposed a way of estimating the shape of
the suffused gradation by assuming that 50 % of the finer fraction of the unstable
soil is lost and, in a conservative assessment, a 100 % loss.
Taking the ICOLD [1] and Wan and Fell [8] approach into account, in terms
of finer fraction and its loss through the action of suffusion, the resultant gradation
for two construction gradations (i.e., #20 and #28) deemed representative of the
Grundsjön dam are shown in Fig. 9. Accordingly, assuming a 50 % loss of the
finer fraction yields an effective D15 of approximately 6 mm, and the conservative
approach of assuming a 100 % loss yields a D15 of 7 to 18 mm. Considering the
dam-specific EE boundary of 4.5 mm the two selected construction gradations
would exceed it. In support, forensic photo evidence (Fig. 3) indicates that the
filter was indeed washed out (suffused).
76
Q. 98 – R. 5
a)
b)
Fig. 8
Unified plot approach with data points from the analysis of Grundsjön dam
superimposed over Rönnqvist et al [2] data base
Approche par scénario avec des points à partir de l'analyse du barrage
de Grundsjön superposée à la base de données Rönnqvist et al [2]
a) Unified plot with average values of NE, a) Scénario avec les valeurs moyennes
EE and CE. de NE, EE et CE.
b) Dimensionless unified plot for b) Scénario adimensionnel pour
D15max/D15EE D15max/D15EE
77
Q. 98 – R. 5
Fig. 9
Selected filter gradations of the Grundsjön dam that are potentially unstable
and corresponding filter assuming a 50 % loss of finer fraction
Granulométries sélectionnées du barrage de Grundsjön potentiellement
instables et filtre correspondant supposant une perte de 50 %
de la fraction la plus fine
It has been postulated that contributing factors to the internal erosion
process, both from the design conditions and also from the technical
specifications of materials and procedures, are as follows:
Potential wetting settlement due to dry compaction at or below optimum water
content causing arching and differential settlements.
An increased lift thickness to 0.75 m, larger Dmax of 400 mm, and moderate
compaction.
Segregation issues in terms of the filter during handling as reported at the time
of construction [11].
The theory put forward by Eurenius and Sjödin [11] as to the cause of the
sinkhole at the time of the incident involved the formation of concentrated leaks
(by settlements, arching and hydraulic fracturing), and the presence of an
inadequate filter which allowed backward erosion through the core, causing
unacceptable leakage and eventually leading to the formation of the sinkhole.
This review supports the previous postulated cause of the sinkhole event.
Based on the many observations linking the incident to internal erosion, clearly
the process which ultimately leads up the sinkhole event is the result of several
interplaying internal erosion mechanisms; suffusion, concentrated leak erosion
78
Q. 98 – R. 5
and insufficient capacity of the filter for base soil retention (i.e., backward
erosion). This supports the use of the unified plot that combines the assessment
of two key attributes of a filter; the Grundsjön dam plots where dams with
probable occurrence typically plot, i.e., potential internal instability of the filter and
inability to arrest an initiated internal erosion process.
5. CONCLUSIONS
The unified plot approach [2] combines two empirical methods with
experience from performance of existing dams; internal instability and incapacity
for soil retention of filters correlate well with dams that exhibit deficiencies that
can be reasonably attributed to internal erosion.
Application of the unified plot approach provides a possible explanation to
the 1990 sinkhole event in the case study of the Swedish Grundsjön dam. The
filter is potentially internally unstable, and filter criteria indicate that the base and
filter interface is incompatible. These aspects when combined suggest a
vulnerability to the initiation of internal erosion and its continuation given that the
filter is potentially inconstant in the long-term (due to filter suffusion). Photo
evidence supports this as filter suffusion indeed had occurred and cavities and
wet zones of the core provide evidence for the occurrence of backward erosion,
and furthermore, the sinkhole formation is the result of a late-stage internal
erosion process which probably had reached the continuation phase.
ACKNOWLEDGEMENTS
The authors extend gratitude to the dam owner Vattenregleringsföretagen
for kind permission to publish the case history of the Grundsjön dam, and also
acknowledged with gratitude is Gunnar Sjödin, Dam Safety Coordinator at
Vattenregleringsföretagen, for valuable discussions and providing data on the
Grundsjön dam.
The research presented was carried out as part of the ”Swedish
Hydropower Centre - SVC”. SVC has been established by the Swedish Energy
Agency, Elforsk and Svenska Kraftnät together with Luleå University of
Technology, the Royal Institute of Technology, Chalmers University of
Technology and Uppsala University.
The authors acknowledge with gratitude the financial support received from
WSP Sweden.
79
Q. 98 – R. 5
NOTATION
d grain size of the base soil (mm)
D grain size of the filter (mm)
F mass passing at grain size D (%)
H mass increment between D and 4D (%)
(H/F)min stability index, defined by the smallest value of H/F,
for 0 < F ≤ 20 % in soil with a widely-graded coarse fraction
REFERENCES
[1] ICOLD, Internal erosion of existing dams, levees and dykes, and their
foundations (ed. Rodney Bridle and Robin Fell), Bulletin 164,
Volume 1: internal erosion processes and engineering assessment, 2013.
[2] RÖNNQVIST, H., FANNIN, J., VIKLANDER, P. On the use of empirical
methods for assessment of filters in embankment dams. Géotechnique
Letters, [Link] 2014.
[3] KENNEY, T.C., LAU, D. Internal stability of granular filters. Canadian
Geotechnical Journal, 22 (2), 215-225, 1985.
[4] KENNEY, T.C, LAU, D. Internal stability of granular filters: Reply.
Canadian Geotechnical Journal, 23 (3), 420-423, 1986.
[5] LI, M., FANNIN, R.J. Comparison of two criteria for internal stability of
granular soil. Canadian Geotechnical Journal. 45 (9), 1303-1309, 2008.
[6] BURENKOVA, V.V. Assessment of suffusion in non-cohesive and
graded soils. Proc. First International Conference “Geo-Filters”.
Filters in geotechnical engineering, Brauns, Heibum & Schuler (eds),
Rotterdam: Balkema, 357-360, 1993
[7] WAN, C.F., FELL, R. Experimental investigation of internal instability
of soils in embankment dams and their foundation, UNICIV report No.
429, The University of New South Wales, 2004.
[8] WAN, C.F., FELL, R. Assessing the potential of internal instability and
suffusion in embankment dams and their foundations. Journal of
Geotechnical and Geoenvironmental Engineering, 134 (3), 401-407, 2008.
[9] FOSTER, M.A., FELL, R. Assessing embankment dam filters that do
not satisfy design criteria. Journal of Geotechnical and
Geoenvironmental Engineering, 127 (4), 398-407, 2001.
[10] RÖNNQVIST, H. Predicting surfacing internal erosion in moraine core
dams. KTH TRITA-LWR Licentiate degree thesis 2050, Royal Institute
of Technology, Stockholm, 2010.
[11] EURENIUS, J., SJÖDIN, G. PM angående undersökning och
reparation av dammskada i Grundsjöns regleringsdamm. Report
P5795 VBB VIAK, In Swedish (Investigation and restoration of
deteriorated zone in the Grundsjön dam), Stockholm, 1991.
80
Q. 98 – R. 5
[12] ICOLD, Rehabilitation of dams and appurtenant works - State of the
art and case histories, Bulletin 119, 2000.
[13] BARTSCH, M. Safety Analysis of Swedish Dams - Dam Performance
and Incident Data Analysis. Licentiate thesis, Royal Institute of
Technology, Stockholm, Sweden, 1995.
[14] SHERARD, J.L, DUNNIGAN, L.P., TALBOT, J.R. Basic properties of
sand and gravel filters. Journal of Geotechnical Engineering, 110 (6),
684-700, 1984.
SUMMARY
The unified plot approach combines two attributes of a filter gradation,
namely its potential for internal instability and its capacity for soil retention.
Comparison to the performance of 80 existing embankment dams that
includes 23 dams which are reported to have experienced some form of
internal erosion, these attributes of the filter have been found to correlate with
deficiencies related to internal erosion. Thus, in engineering practice, the
unified plot may serve as a screening tool for internal erosion susceptibility.
The Grundsjön dam, located on the river Ljusnan in Sweden, suffered a
sinkhole event in 1990, 18 years after commissioning. In depth examination of
the damaged stretch of the dam conducted in 1990 to 1991 revealed that
internal erosion had indeed occurred; signs e.g., cavities in the core and filter
devoid of fines suggest the process had initiated and been allowed to continue
to the point that a sinkhole formed on the crest.
Analysis of 43 as-placed filter gradations extracted from the Grundsjön
dam filter envelope indicates that the filter is potentially unstable, although, not
excessively coarse in terms of base soil retention. These results, as seen
separately, provide relatively little explanation as to why the filter ultimately was
unable to arrest the internal erosion process. However, combined, in the
framework of the unified plot, it is readily apparent that the Grundsjön dam filter
gradations distribute where dams with probable occurrence of internal erosion
characteristically plot. Thus, the unified plot provides a plausible explanation for
the internal erosion process that occurred. Indeed, reverse engineering the
potential internal instability of the coarse filter gradation indicates that once
changed by suffusion due to internal instability it potentially becomes a filter
permitting excessive erosion.
The Grundsjön dam was carefully rebuilt in 1991 and was provided with an
improved drainage system as well as the implementation of an extensive
monitoring program of instrumentation, and has since operated satisfactorily.
81
Q. 98 – R. 5
RÉSUMÉ
L’approche par scénario combine deux caractéristiques de la granulométrie
du filtre : son instabilité interne potentielle et sa capacité de rétention des sols. En
comparant le comportement de 80 barrages en remblai existants, parmi lesquels
23 barrages ayant subi une forme d’érosion interne, ces caractéristiques du filtre
se sont avérées avoir une corrélation avec les altérations liées à l’érosion interne.
Ainsi, en matière de génie civil, le scénario peut être utilisé comme outil de
dépistage d’une érosion interne potentielle.
Le barrage de Grundsjön, situé sur le fleuve Ljusnan, en Suède, a subi un
trou d’affaissement en 1990, 18 ans après sa mise en service. L’examen en
profondeur des parties endommagées du barrage, effectué de 1990 à 1991, a
révélé une érosion interne ; des signes, tels que des cavités dans le noyau et
l’élimination et matériaux fins dans le filtre, indiquent vraisemblablement que le
processus a été initié et a pu se poursuivre jusqu’à former un trou d’affaissement.
L’analyse de 43 types de granulométrie dans le filtre en place, extraits de
l’enveloppe du filtre du barrage de Grundsjön, indique que le filtre est
potentiellement instable, bien qu’il ne doit pas excessivement grossier au regard
de la rétention du sol de base. Vus séparément, ces résultats expliquent
difficilement pourquoi le filtre a été au final incapable d’arrêter le processus
d’érosion interne. En revanche, combinés dans le cadre d’un scénario global, il
est évident que la distribution de la distribution granulométrique du filtre du
barrage de Grundsjön indique les emplacements où une érosion interne risque
fortement de se produire. Ainsi, le scénario donne une explication plausible du
processus d’érosion interne qui s’est produit. En effet, l’ingénierie inverse
appliquée à l’instabilité interne potentielle de la granulométrie du filtre grossier
montre qu’une fois modifié par suffusion en raison de l’instabilité interne, le filtre
permet une érosion excessive.
Le barrage de Grundsjön a été soigneusement reconstruit en 1991 avec un
système de drainage plus performant et accompagné de la mise en place d’un
large programme de surveillance des instruments, et fonctionne depuis de
manière satisfaisante.
82