You are on page 1of 52

ICOLD

CFRD - New
Bulletin

May 29, 2022


1. Introduction
The construction of Concrete Face Rockfill Dams (hereafter referred to as CFRD) has increased
in the last decades. This enthusiasm for CFRDs has its origin in both its inherent stability
characteristics and its construction and schedule features. In addition, its foundation
requirements and treatments that are less strict and more straightforward to carry out,
particularly when compared to gravity or arch dams, have made this type of dam a very attractive
solution. Therefore, the CFRD is currently a very common type of dam with several projects
recently finished or under construction, with heights above 180 m. Figure 1 illustrates the trend
in height of CFRDs. The design and development of CFRD dams have been based primarily on
precedent and empiricism, however, recent incidents have shown that the extrapolation of
precedents with the current procedures can have serious consequences. Recent events have
shown the need to improve the current design approaches and evolve beyond empiricism,
including the development of analytical methodologies to analyze the behavior of this type of
dams. This paper illustrates the evolution of the CFRD from its early years in California, through
its development in the mid of the last century. The recent problems that have occurred in some
dams in Brazil, Lesotho and China where the face has suffered considerable cracking are also
presented to finish with current successful experiences that illustrated some effective design and
construction procedures

The confidence of engineers in a certain class of structures is largely based on the satisfactory
behavior of a large number of specimens of the same type, designed and built with the same
technique. The reasoning implicit in this concept is valid if the data covers the whole range of
stresses and structural characteristics of interest, combined.

Regarding earth and rockfill dams, current experience is based on many dozens of cases of failure
or damage and thousands of well-behaved cases, covering almost all possible combinations of
structural characteristics and stresses. There have been very few instances where an earthquake
has damaged an embankment dam enough to result in the uncontrolled release of reservoir
water. Many embankment dams are exposed to earthquake shaking each year, but either the
damage caused by the earthquake has not been extensive enough, or in the rare cases where
damage was extensive, many of the reservoirs were, by chance, low at the time of the
earthquake, so uncontrolled releases did not happen. The failure probability estimation
procedures described below are built upon standard analysis techniques used to predict responses
of soil to dynamic loading and upon observations from case histories of embankments that have
been exposed to earthquakes.

The description of damage in earth dams and rockfill dams subjected to earthquakes has been
the subject of several publications in recent years (USSD, 2014, USBR 2015, Swaisgood 2014,
Pells and Fell 2002). The high relative frequency of damage attributable to landslides or shear

1
distortions in the dam or foundation is remarkable. When evaluating the information included in
the three volume report from USSD on Observed Performance of Dams During Earthquakes
(USSD, 2014) with the statistics of non-seismic failures (ICOLD, 2017) it appears that landslide
failure is much more frequent among seismic damage than among non-seismic dam failures.
Therefore, experience indicates that in the analysis of the seismic reliability of a dam, special
attention should be given to the slip failure mechanism. However, it is evident that not all other
mechanisms have negligible probability as implicitly seems to be sometimes considered. In
addition to sliding or shear distortion, at least the following damage mechanisms can be explicitly
considered: transverse cracking, both due to differential settlement and other causes, loss of free
edge due to densification, rupture of buried conduits, damage due to displacements in geological
faults and slope collapse.

It is important to differentiate between the seismic response of a rockfill and earth embankment
dams, considering that compacted rockfill dams have an inherently high resistance to seismic
loading. With modern construction techniques and gradation control the materials that make up
a rockfill dam have a high shear strength, high deformation modulus, as well as high permeability
and therefore this type of dam is less deformable and the possibility of generating excess pore
pressure in an earthquake is very low, so the capacity of a rockfill dam to resist seismic excitation
is in general terms greater compared to an embankment dam.

The design of large dams in high seismic hazard zones requires an estimation of seismic design
parameters calculated taking into consideration the project´s tectonic framework and its
seismicity. Definition of seismic design parameters for large dams needs to fulfill the most strict
requirements because the consequences of a dam failure can be catastrophic, in particular for
dams with large storage capacity (>120 hm3), dams higher than 45 m, dams that in case of an
emergency could require to evacuate more than 1000 people and/or if high potential downstream
damage (economical and/or environmental) as a result of the dam´s failure is assessed (ICOLD,
2010).

ICOLD, (2016) defines the design scenario for a dam [Safety Evaluation Earthquake (SEE)] as the
level of shaking for which damage can be accepted but for which there should be no uncontrolled
release of water from the reservoir. The SEE replaces the terms Maximum Design Earthquake
(MDE) used in the first edition of the bulletin and Design Basis Earthquake (DBE) used in (ICOLD,
1983) Bulletin 46 (Seismicity and Dam Design). The Safety Evaluation Earthquake (SEE) is the
maximum level of ground motion for which the dam should be designed or analyzed. For dams
whose failure would present a great social hazard the SEE will normally be characterized by a

2
level of motion equal to that expected at the dam site from the occurrence of a deterministically-
evaluated maximum credible earthquake or of the probabilistically evaluated earthquake ground
motion with a very long return period, for example 10,000 years.

The Maximum Credible Earthquake (MCE) is defined by ICOLD (2016) as the largest reasonably
conceivable earthquake magnitude that is considered possible along a recognized fault or within
a geographically defined tectonic province, under the presently known or presumed tectonic
framework. The most severe ground motion affecting a dam site due to an MCE scenario is
referred to as the MCE ground motion.

Deterministically evaluated earthquakes may be more appropriate in locations with relatively


frequent earthquakes that occur on well- identified sources, for example near plate boundaries.
It will be required at least that there is no uncontrolled release of water when the dam is subjected
to the seismic load imposed by the SEE. Depending on the circumstances (e.g. the importance of
the dam, the consequences of a dam failure) it is recommended to design safety-critical elements
such as the bottom outlet and/or spillway gates for the SEE (ICOLD, 2016).

Design and verification of a project involving a large dam also requires that no damage or loss of
service must happen in the event expected to occur in the lifetime of the dam. This scenario is
referred to by (ICOLD, 2010) as the Operating Basis Earthquake or OBE, and has been defined
(ICOLD, 2010) as a 145 years return period event. Therefore the estimation of the OBE design
intensities (PGA, PGV, PGD and spectral accelerations) are determined using probabilistic seismic
hazard assessment (PSHA) results.

In addition to SEE and OBE design scenarios the design of a large dam project needs to consider
a construction earthquake scenario or CE. This scenario is used for the design of temporary
structures such as coffer dams and takes into account the service life of the temporary structures.
This design scenario is usually determined using PSHA results and assuming a return period to
that of the design flood of the river diversion.

Historical Perspective
Concrete face dams have their origins in the mountains of California, used not only to provide
reservoirs for mining exploitation but also for the hydroelectric developments required for the

3
economic expansion that took place in the region during the beginning of the twenty century.
The presence of numerous rock outcrops of granitic rocks in the heart of the Sierra favored
embankment dams constructed from dump rockfill and covered with timber facing that provided
a reasonable barrier to contain the water. As higher dams were built, the practice evolved
empirically, towards a concrete slab to cover the rockfill. The slab was built in sections separated
by joints placed over a hand-made masonry bed.

In the 1930s taller dams were built, like Salt Springs dam in California (101 m) [1]. For this dam,
that was the tallest in the world for 23 years, vertical joints with openings of one inch were
included and horizontal joints were built as construction joints. Eight of the central vertical joints
closed completely, and in three, the concrete failed by crushing. Some of the horizontal joints
also experienced concrete crushing. At the moment, it was believed that the major source of
leakages were the cracks near the union with the abutment, where no perimeter joint was built.
The leakages in Salt Springs reached 450 l/s. The empirical solution adopted consisted in
increasing the number of joints and in introducing compressible materials, as done in Bear Creek
Dam [1]. In addition, a perimeter joint and a hinges slab parallel to the canyon were included to
decrease the demands on the slab near the abutments. In the Bear Creek Dam, the leakages
were reduced considerably (112 l/s) and also the cracking was limited to some cases were
superficial spalling occurred in the walls of the joint.

The Paradela Dam (120 m) was the tallest structure built with this technology known until the
1950s. In reference [2] the behavior of the dam and the concepts considered for its design are
presented. The design was absolutely empirical and based on additional elements that included
the creation of new perimeter joints (two perimeter slabs were built, like the ones used in Bear
Creek Dam), opening of vertical joints with compressible elements (7 cm) and of 3 cm in the
horizontal joints. During the repair, after the first reservoir impounding, they were further
increased by cutting the border of the joints to widen the space for movements trying to relief
the compression stresses in the concrete face. The materials used in the fill were so compressible
that, even before the reservoir impounding, it was observed that the joints between the slabs
were closing as a result of the deformations in the filling by their own weight. With the reservoir
load, the deformations reached a value of 2.04 m in the direction normal to the face. Excessive
stresses generated by movements in the first and second impounding produced some cracks in
the borders of the slabs. However, the largest cracks were reported in the perimeter slabs. The
leakage reported in the first years of operation exceeded 3 m3/s; confirming the capacity of the
fill to manage high levels of leakage without the risk of failure of the structure. The leakages
registered in Paraderal Dam were also greater than the ones registered in the higher dams that
experienced severe cracking recently, illustrating the relevance of the introduction of a modern
support zone, with special gradation, a maximum size of 4 inches and a maximum of 7% of fine
material, in controlling leakage. This level of gradation allows having an average permeability in

4
the order 10-4 cm/s that creates an element able to limit the magnitude of leakages as
anticipated by Sherard [6]. However, the high deformations and the leakage associated resulted
in neglecting this type of dam for some years. As a result, only until the development of modern
construction equipment, robust enough to allow the placement of compacted rockfills, a new
stage started in the development of this type of dams.

The relevant lessons, from the first generation of CRFD dams, referred to the effects that the
excessive movements of the fill generated on the face slabs. The deformations were associated
with the low deformation modules that were obtained from the placement process. The
experience in these dams suggest that in the process of trying to eliminate the cracks by
compression, with the creation of more deformable joints, the cracks were reduced but the
leakage increased. The leakage increased as a result of the greater number of joints and the
opening of those that did not closed. This is a very relevant assessment for modern dams that
will be further discuss later in this paper.

SECOND GENERATION OF CFRD: MIXED EXPERIENCES

The construction of concrete face rockfill dams (CFRD) with compacted rockfill started in the 60s
with mixed experiences. The New Exchequer (150m) dam in California, which consisted in the
heightening of a concrete dam by placing CRFD on the older dam, had an undesirable behavior
given that the perimeter joint in the support allowed very high leakage. In Australia, with the
dams of Kangaroo (65 m) and Cethana (115 m) the situation was different because favorable
behavior was observed. Further, in the Cethana Dam an extensive instrumentation program
allowed for a better understanding of the behavior of the slab during operation. Is in this dam,
that the first values of strain in the concrete were recorded. These values were used to estimate
the induced stresses on the concrete slabs and have been useful to evaluate modern numerical
analysis capabilities to reproduce the observed behavior of that dam and other more recent dams.

The Alto Anchicaya Dam (140 m), built at the beginning of the 70s, where horizontal joints were
eliminated given the considerable problems that caused in Paradela dam, had in general terms
low deformations, corresponding to rockfills of high quality, with a modulus of deformation of
more than 100 MPa. However, this dam experienced, during the first impoundment, high leakage
associated with the zone closed to the abutments (perimeter joint). The cracking observed during
the repairs, that required lowering the reservoir level, was associated to the particular placement
of the slabs, close to the abutment, (similar to what was observed in Paradela), and to the
deformations associated to the way the material was placed. The material was placed close to
the abutments, after erosion of original material took place during an intensive rain that
concentrated great flows toward the abutments. In the repair, plastic material (IGAS) was used
encapsulated inside steel angles, as a second line of defense to the waterstop located in the

5
middle of the joint. This solution, of including the IGAS, was later adopted for the first time in a
design in the Foz de Areia Dam. .

The Foz de Areia Dam (160 m) set a precedent by being the tallest dam for more than 10 years
with very good behavior in spite of the big deformations observed. This dam was built in a wide
valley with basalt rockfill with a low deformation modulus (around 40 MPa) that translated in
deformations by water load of more than 1.0 m of the slab in the central part of the dam. These
deformations did not translated into cracking or significant leakage and hence Foz de Areia
became a reference for future dams. Today, it is clear, that having a very wide valley allowed
for a favorable behavior of the face given that no excessive horizontal stresses were induced.

After the construction of the Foz de Areia Dam, the construction of several concrete face dams,
with heights below 130 m, and rockfills basically composed of basalts, started in Brazil. The
process of placement and compaction was relatively homogenous in these dams with materials
lifts of 1 m in the upstream shell and 2 m in the downstream shell. All of these dams were
considered successful experiences, even though, leakages between 300 and 500 l/s were
recorded. Some of the dams, like in the Xingo case, experienced cracking in specific places that
illustrated the relevance of precluding sudden changes in the deformation pattern of the slab,
avoiding hard points behind the face (Marulanda and Pinto [4]).

During these years, in general, from a leakage point of view, the dams with low compressible
rockfills behaved well. The Golillas Dam (125 m) was an exception, because even though it was
constructed with gravels, (the fill was associated with a deformation modulus of around 120 MPa),
large leakage occurred during the first impoundment that required lowering of the reservoir to
perform repairs. However, leakage was not associated with movements in the fill but with defects
in the foundation of the rock abutments. There, the soft fills in the relief joints of the foundation
rock mass (very steep abutments) were eroded by leakage through the left abutment.

The other exception was the Aguamilpa Dam (187 m) in Mexico that for more than 15 years was
the tallest dam of its type in the world. Even though the shell upstream of the dam was built
with gravels of high deformation modulus, the greater compressibility of the rockfill used in the
downstream shell generated an unusual situation by introducing a non-uniform pattern of
deformations in the upper part of the face. Other cases where large leakage occurred were
associated with particular defects in the dimensions of the plinths, like in the case of Shiroro dam.
In this case, , the structural instability of the plinth, built in a particular sector of the foundation
that required the construction of a base to maintain the level of the plinth, generated the facture
of the slab near the perimeter joint.

Towards the final years of the last century, there was a clear tendency to consider that concrete
face dams could be viable for heights of more than 200 m, without major changes in the

6
configuration and design procedures commonly used. In some cases, it was considered the ideal
dam, and therefore, to some people there was no limit to its height.

7
Major Incidents - unfavorable behavior
Towards the end of the last century, a new stage in the evolution of CRFD dams started.
Considering that the height limitations for a dam of this type were overcome, the more or less
simultaneous construction of very high dams started in China (TQ1), Brasil (Barra Grande and
Campos Novos) and Lesotho (Mohale). All of these dams experienced intense cracking in the
slabs. Marulanda [5] presented an evaluation of these cases, making emphasis in the fact that
during these new developments of high CFRDs, the precedent seemed to be ignored in relation
to the consequences of using highly deformable rockfills. Figure 2 shows an empirical graph that
correlates the deformations experienced by CFRD dams due to water load with the shape factor
of the canyon. This indicates that it was possible to empirically predict in a roughly manner the
possibility of excessive stresses in the slabs of the dams. Figure 2 describes the observed
behavior of several dams in terms of a non-dimensional shape factor A/H² against the normalized
deflection of the face, where A is the area of the concrete face, and H is the height of the dam.
The factor A/H² will be smaller for narrow canyons. A shape factor A/H² of 3.5 can be considered
a narrow canyon, such as in Campos Novos and Barra Grande Dams. The dashed line presented
in Figure 2 is based on the observed behavior of the dams that have experienced significant
vertical cracking in the concrete face due to high stresses.

Figure 2 includes two cases in China that were recently reported. One of those cases is Sanbanxi
dam, a 185m high dam, built between 2002 and 2006, where an intensive cracking was reported
by Xu et al [13]. Xu et al reported the calculated stresses based on the records of strain gauges
installed in the face before cracking of the face, which was defined at the moment were a sudden
increase in leakage was reported. Also, the numerical analysis performed in the design stage was
presented to reveal if it was possible to predict this adverse behavior. Xu et al [13] suggested
that by using the representative parameters of the rockfill materials derived from the laboratory
tests it was impossible to estimate the observed behavior. However, once the values obtain during
construction were used the analysis it was possible to estimate stress values similar to the ones
obtained with the instrumentation. The stresses estimated indicate higher values than the ones
registered in the Mohale Dam (18.5 MPa), reaching 24 MPa before fracture. The value to be
considered as the maximum acceptable in the analysis, is still a subject of discussion since a
consensus is needed regarding the safety factors that are considered acceptable. The other
Chinese case included in Figure 2, is the Buxi dam, 136 m high, a crest length of 271 m. The dam
was constructed between March 2008 and February 2011. After a sudden increase in water level,
the leakage increase sharply from 0.1 m³/s to a value near two (2) m³/s. When the water level
was lowered, serious damages to the concrete face were observed.

All of these cases make evident the need to have a deeper analytical capability to understand the
behavior of this type of dam, therefore, converting the engineering for these dams in a more

8
rational process that allows sensitivity analysis of different variables to optimize the real
requirements.

Figure 4. Extent of face failure at (a) Campos Novos (b) Barra Grande and (c) Mohale

9
Successful experiences after the Bulletin 141 was
published
When the three serious incidents of the dams in Brasil and Lesotho took place, two very high
dams were being built, one in Mexico (Cajón, 188m) and one in Island (Karajnukar, 200 m).
The rockfill materials used in these dams were different from the ones used in the problematic
cases. In those cases with undesirable behavior, the rockfill was essentially basalt and the
deformation modulus of the fill was in the order of 50 MPa. In the Cajon dam, where the owner
(CFE) maintained the tendency, initiated several years ago, with the experience of the high
central core rockfill dams, of requiring high levels of compaction, and therefore modulus of
more than 120 MPa for the fill were achieved. As a result, the deformations in the face were
three times smaller than the deformations experienced in the dams where the face cracked.
The Cajon dam, 188 m high built in a narrow valley, has exhibited an outstanding behavior.
Figure 2 compares the gradation of the rockfill of Cajon dam with the ones built with basalts.
The absence of fine material and lower compaction efforts are the main reasons for the
differences in the obtained modules. In the Karhanjukar dam in Iceland, and as a reaction to
the events in the other dams, additional elements were incorporated during construction to the
design of the face including adding a geomembrane to cover the lower part of the slab, the
increase of the compaction of the materials in the upper part of the fill and the introduction of
compressible material in some of the compression joints. The dam had an adequate behavior,
however, the authors’ beliefs that it is not the result of the additional measures taken during
construction but mainly to the differences in the characteristics of the basalt of the rockfills. In
Iceland a Moberg is a basalt that is cooled below layers of ice and when blasted converted into
a material that has a great content of fine materials. With this material, a very well gradated
rockfill was obtained and, therefore, allowed to achieve higher modules. In fact, the
instrumentation records in this dam demonstrate that the representative modules of these
rockfills range between 100 and 120 MPa[8] and for that reason, the deformation of this dam,
with a height of more than 200 m, was considerably lower than the ones in the Brazilian dams
and Mohale in Lesotho.

In several parts of the world and opposition towards this type of dams was generated, given
the incidents reported. These cases had greater impact than the cases of dams with clearly
satisfactory behavior. The authors observed how, by maintaining the criteria of building this
dams by precedent, even in dams with less heights and with very good materials, additional
elements were being introduced like compressible materials in the joints, slabs with less width,
greater requirements for the fillings, etc., that increased unnecessarily the costs of this type of
dams.

10
Figure 2. Gradation of the rockfill.

However, all these incidents had some positive effect. Intensive cracking did not translate into
uncontrolled water leakage. The support zone of the slab in the modern dams has
demonstrated its capacity to control the magnitude of the leakage, as postulated by Sherard
[6]. The argument presented by some, that includes the great risk of this dams in seismic
zones due to the possibility of severe cracking of the face, has been losing importance given
that it is difficult to imagine more severe cracks than the ones observed in the Brazilian and
Lesotho dams. In those events the stability of the structure was never in doubt. The recent
experience recorded in The Zipingpu dam in China, where the dam was located only 17 km
from the epicenter of the Sichuan earthquake (2009) and reports indicated that amplifications
of more than 2g in the crest of the dam was reported [9] reinforced the theory that this highly
permeable dams are extremely save event in the occurrence of a sever seismic event and very
high leakage.

Considering the above, it was evident the need to have a deeper analytical capability to
understand the behavior of this type of dam; therefore, converting the engineering for these
dams in a more rational process that allows sensitivity analysis of different variables to optimize
the real requirements.

Recent experiences

11
In the last years several, high CRFD have been and are being built for which detailed analysis
have been performed with the purpose described above. Porce III, Quimbo and Sogamoso
(Colombia), La Yesca (Mexico) have all used numerical analysis as an important element in the
design process including the selection of rockfill material selection (fill zoning) and in the
introduction of additional elements in the face. This aspect, of material selection, is very
important and good practice in the case of tall dams, where the rockfill deformation is essential
for the good behavior of the concrete face, and should be based in thorough investigations of
material characterization and real experiences. It should not be further believed that a good
rockfill is the only one composed of very hard particles. The research of Marsal corroborated
the idea that to reduce the deformations of the rockfill, including creep, there should be a
special emphasis on the gradation. A rockfill with not very hard particles, but properly
gradated, results in lower deformations than a rockfill with very resistant particles but poorly
gradated. The Karajnukar dam, previously describe, clearly illustrates this point.

Porce III is the first dam that demonstrated that is possible to build a high dam with materials
that in previous circumstances could have been considered as inadequate for a rockfill. The
location of the spillway at the left margin required the excavation of a quantity of material
somewhat greater than the volume of the dam. As a result, using this material represented an
economic incentive for the project. However, the material consisted in a sequence of schists of
different characteristics, varying from graphitic schist of low mechanical properties to more
competent granitic schists. Weathering of the material near to the surface also reduced its
mechanical characteristics of the material. As a result, the first excavations of the spillway
produced a highly weathered material, impossible to use in the upstream shell. Trial fills were
constructed with different materials derived from the spill excavation, including rockfill mix from
material derived from different areas of the spillway (graphitic or granitic schist and different
levels of weathering). Odometer tests and plate load tests were performed taking the load to
relatively high values creating reaction with jacks against two 35 ton trucks and with plates
having a radius of 0.75 m. The field tests were complemented with laboratory results such as
LA abrasion and absorption indicated variable results. The graphitic schists produce results to
resistance to abrasion Angeles as high as 75 per cent. The representative values for the
materials are presented in Table 1. Based on the tests and the results of the trial fills, expected
ranges for the deformation modules were established for the different materials.

This information was the base to perform sensitivity three dimensional finite element analyses
[10], [11] to establish the effects of using different types of material mixes in the behavior of
the dam and especially of the concrete face. At the same time, it was necessary to perform a
matrix of origin and destination of the excavation materials to determine the capacity of
producing the amount of materials that the adopted zoning required. From these analyses, it
was determined that the material for the upstream zone could not have less than 85 MPa, while

12
materials with greater deformation were allowed downstream of the fill. Figure 3 illustrates the
adopted section. Given that the use of materials with less mechanical characteristics implied
also a decrease in the permeability, a vertical drain was connected to a horizontal drain to
prevent a high water level in the downstream shell.

During construction of the dam, plate load tests were performed systematically to check the
characteristics of the mix of materials used. The results were also correlated with the modules
that could be calculated from the instrumentation installed within the fill. Figure 3 shows the
main section of the Porce III dam with the instrumentation.

The instrumentation and the derived modulus were used to calibrate the numerical analyses to
further estimate the deformations and stresses in the concrete face and its joints. These
analyses indicated that the horizontal compression stresses in the central part of the dam could
reach values closed to 20 MPa. From the dams that have suffered failures by compression in
the concrete face, the Mohale [11] dam was the only one that had records of strain gauges that
could be used to estimate the stresses on the face. These records indicated that the failure by
compression, caused by both the horizontal stresses and by the ones parallel to the slope, the
values reached before failure were around 18.5 MPa. Based on this information, it was defined
that for the Porce III project the criteria adopted consisted in not accepting configurations
where the estimated stress could surpass 15 MPa. As a basic measure to reduce the
development of high compressive stresses in the face a compressible element was introduced in
the vertical compressible joints (compressive joints fill with a compressible material to dissipate
stresses, see Figure 4). The decision to incorporate these compressible joints was based on
detailed numerical analyses, were their effect in reducing in an effective way the levels of
stresses was clearly demonstrated. The analyses also helped defined the dimensions (width)
that could be used and the number of compressible joints to be implemented.

The construction of Concrete Face Rockfill Dams (hereafter referred to as CFRD) has increased
in the last decades. This enthusiasm for CFRDs has its origin in both its inherent stability
characteristics and its construction and schedule features. Also, its foundation requirements and
treatments that are less strict and more straightforward to carry out, particularly when
compared to gravity or arch dams, have made this type of dam a very attractive solution. As a
consequence, the CFRD is currently a very common type of dam with several projects recently
finished or under construction, with heights above 180 m.

Figure 6 is a typical section of a CFRD of compacted rockfill founded on sound rock. Outer
slopes can be as steep as 1.3H:1V and for the case of weaker rockfill and foundation in high
seismic areas, slopes and zoning are adapted to accommodate the specific conditions.
Depending on the characteristics of the rockfill, a vertical drainage chimney should be included
to avoid saturation of the downstream shell as shown in Figure 6.

13
The zone designations of 1, 2 and 3 have become common practice where Zones 1A and 1B are
concrete face upstream protection in increasing order of maximum particle size. Zones 2A and
2B are processed granular downstream supporting zones for the concrete face, in increasing
order of maximum particle size. Zones 3A and 3B are the rockfill zones for the dam shells, in
increasing order of maximum particle size and starting with more compacting effort [Bulletin
141, (ICOLD, 2010)].

The design and development of CFRD dams have been, until recently based primarily on
precedent and empiricism, however, incidents that occurred within the last fifteen years have
shown that the extrapolation of precedents with common procedures can have serious
consequences. The undesirable behavior of some CFRDs have shown the need to improve the
design approaches that were considered state of practice a few years back and evolve beyond
empiricism, including the development of analytical methodologies to analyze the behavior of
this type of dams.

Towards the end of the last century, a new stage in the evolution of CFRD dams started.
Considering that the height limitations for a dam of this type were overcome, the more or less
simultaneous construction of very high dams started in China (TQ1), Brasil (Barra Grande and
Campos Novos) and Lesotho (Mohale). All of these dams experienced intense cracking in the
slabs (See Figure 4). Marulanda (2006) presented an evaluation of these cases, making
emphasis on the fact that during the evolution process, the precedent seemed to be ignored in
relation to the consequences of using highly deformable rockfills.

Figure 5 describes the observed behavior of several dams in terms of a non-dimensional shape
factor A/H² against the normalized deflection of the face, where A is the area of the concrete
face, and H is the height of the dam. The factor A/H² will be smaller for narrow canyons. A
shape factor A/H² of 3.5 can be considered a narrow canyon, such as in Campos Novos and
Barra Grande Dams. The red line presented in Figure 5 is based on the observed behavior of
the dams that have experienced significant vertical cracking in the concrete face due to high
horizontal stresses. The governing parameters for the building of high horizontal stresses in the
face have been identified as being: 1) the shape of the canyon and 2) the rockfill deformation
modulus, which controls the deformations of the face.

14
Figure 5. Empirical graph correlating the deformations experienced by CFRD dams

15
Design Criteria for design of CFRDs
In a CFRD, no portion of the dam is saturated because the impervious element is located in the
upstream face, the dam body is essentially dry and therefore the earthquake cannot generate
pore water pressures. With the absence of pore water pressures and the high shear strength
than can be achieved in a rockfill following modern construction practices, a CFRD has
inherently a very high resistance to seismic loading. Nevertheless, the design of a CFRD should
incorporate defensive features to secure acceptable behavior in the event of an earthquake.
The following criteria are the main criteria for the seismic design of a CFRD

2.2.1. Filter criteria

Zoning of a CFRD and the functions of the various zones has gradually evolved over the years.
In addition to the behavior and targeted characteristics of the main rockfill of a CFRD dam
explained in the previous section, that focused in obtaining a well compacted, adequately
graded rockfill that generates a high modulus rockfill with less susceptibility of deforming in the
event of an earthquake, the filter zone (2A) and support zone of the concrete face (2B) are of
paramount importance with respect to the seismic behavior.

The zone 2A (ICOLD, 2010), which corresponds to a sand and gravel filter located within two
meters of the perimeter joint has a major importance in the event of damage of the waterstop
at the perimeter joint in case of a seismic shaking. This filter will prevent the movement of silt
size particles through the zone and thus serves as a secondary line of defense against leakage.
This zone 2A consists of manufactured material very similar to a concrete aggregate and its
gradation has to follow both the retention and permeability criteria (Sherard, 1985; ICOLD,
2017; USSD, 2021). It should not segregate or change in gradation during the handling, placing
or compaction and should be internally stable and meet the retention criterion. The
recommended gradation should limit the percentage of more than ¾ inch material to 0 to 15
%, should include a generous percentage of sand size particles between 50 and 75% and
percentage of fines passing the No. 200 sieve below 5% in addition to an average uniformity

16
coefficient of 20. This recommended gradation will not segregate during placement, the fine
content will provide some binder for stability during placement and the permeability of around
10-² cm/s will be the same as for typical fine filters for ECRD and will not exhibit cohesion.

The zone 2B, provides support to the concrete face and consists of sand and gravel sized
particles placed typically in 40 cm horizontal layers with 4 or less passes. Compaction of filters
should be minimal. Excessive compaction, especially of crushed rock, can generate sufficient
fines in the filter to make them susceptible to cracking (Milligan, 2003). The horizontal width of
the zone varies from 2 to 4 m depending on the height of the dam. Similar to zone 2A, the
grading of zone 2B has evolved with experience. Placing a brittle material, with high fines
content that can hold a crack open should be avoided. With the objective of avoiding cracks,
the zone must be completely non-cohesive so that the zone can accommodate deformations
that occurred during construction and reservoir impounding without cracking. A crushed,
processed material with a maximum size of 80 mm, 40 to 50% passing No. 4 sieve and a
maximum of 5% non-cohesive fine is recommended. It is important to protect the surface of
zone 2B against erosion during heavy rains and to provide a firm base for the construction of
the concrete face. Currently, the most common practice, initiated years ago during the
construction of ITA dam in Brazil, is the use of a concrete curb at the upstream face after every
layer. This construction method generates various construction benefits in addition to the
erosion protection. These include the reduction of segregation, and a clean surface for form and
reinforcement placement and concrete face construction. It is highly recommended to use a low
cement content, in the order of 70 kg/cm³, to avoid having a stiff material that will crack as the
dam body deforms during construction.

2.2.2. Protection of the foundation

In general terms, the requirements of the dam foundation treatment of a CFRD are less strict
than for an ECRD previously described. The foundation treatment and protection of a CFRD
should be focused on removing any unstable or inappropriate foundation material and
controlling seepage through the plinth. If the removal of unsuitable material beneath the plinth
and dam body becomes impractical, defense measures that prevent erosion of the material
must be implemented. In the case that any deposit is left in the abutments or riverbed, it
should be carefully evaluated if a loss of strength during an earthquake is possible. In that case,

17
a detailed evaluation should be performed to determine the reduction in stability conditions due
to a strength reduction and the requirements of removal of the deposits. Erodible material that
is left in place may need to be protected with filters in order to prevent internal erosion and
fines being washed into the rockfill. Also, the material left in place, especially upstream of the
dam axis, should have a modulus similar to that of the rockfill to avoid excessive face
movements.

2.2.3. Stability considerations

Earthquakes impose additional loads to a dam over those under static conditions. In general,
strong ground shaking can result in the instability of the dam and loss of strength at the
foundations However, rockfill dams, that are well compacted according to the specification and
on a good foundation are suitable type of dam for regions having high seismic activity and
can withstand strong earthquake shaking with no detrimental effect. In high seismic hazard
zones one of the key aspects to seismically analyze the dam is the definition of design seismic
hazard intensities such as peak ground acceleration (PGA) or spectral accelerations (Sa)
calculated by developing probabilistic and deterministic seismic hazard assessments (PSHA and
DSHA) [(Cornell, 1968), (Hanks, 1994) and (ICOLD, 2010), (ICOLD, 2016)]. In the initial design
stages (conceptual design and feasibility design) pseudo-static analyses are developed in order
to assess the stability of the main components of the project, to minimize inadequate behavior
due to acceleration, velocities and displacements induced by an earthquake event. Pseudo-
static analysis required as an input to define the PGA and SA values and a ratio or an equation
to calculate the pseudo-static coefficient.

The definition of seismic intensities at a dam location is usually based on the results of seismic
scenarios associated with the Maximum Credible Earthquake (MCE) for regional and local
seismic sources or by developing a seismic hazard disaggregation analysis with the PSHA
results.

For CFRD as well as ECRD dams the pseudo-static coefficient is typically calculated as a
percentage of the PGA [e.g. Hynes-Griffin and Franklin (1984), Stewart et al., (2003) and Bray
and Travasarou (2009)] or by estimating the contribution of the spectral accelerations for the

18
fundamental structural modes of movement (e.g. Makdisi and Seed (1978)), therefore, a
seismic hazard assessment is required even for the initial design stages of a project involving a
large dam or a spillway structure. The design return period, the distance to active seismic
sources and the attenuation law and its standard deviation are some of the key parameters that
have been identified to be very relevant in defining the seismic design parameters for design
projects where the authors have been involved in high seismic areas. The simplified method of
implementing a pseudo-static coefficient in limit equilibrium analysis should use a coefficient
calculated using displacement criteria methodology in order to assure limiting the displacements
that an earth dam can experience during an earthquake event. The use of a kh = 0.5 is
considered conservative and appropriate for conceptual and feasibility design analyses. For
more advanced stages of design and for dams with significant seismic loads, high risk, or those
where simple analyses identified seismic concerns, a more advanced deformation analysis
should be performed, including a non-linear dynamic analysis as described in the following
section

2.2.4. Seismic induced deformations

For a rockfill dam, permanent deformations can result from densification of the foundation or
rockfill material, but the dominant cause is likely to be shear strains. If deformation is large
enough to allow overtopping of the dam, a rapid breach of the dam would generally be
expected. In the case of a CFRD, deformations without overtopping could also cause cracking
of the concrete face nevertheless, as previously described, should not generate an imminent
threat to the stability of the dam. Even if a rockfill dam and its foundation soil would not liquefy
sufficiently to allow gravity-driven instability, permanent deformations may still occur. A
sufficient freeboard above the normal maximum reservoir level should be provided to mitigate
against deformations induced by a major earthquake. Considering the uncertainty in the
estimation of permanent deformation due to seismic events, the freeboard above the maximum
operational reservoir elevation should be several times larger than the estimated value (should
not be less than three to four times the maximum expected deformations (ICOLD, 2016).

2.2.5. Plinth design under seismic considerations

19
When designing the reinforced concrete plinth, which is supported directly on the abutment
rock and is tied to its foundation with anchor rods that support an upward pressure equal to the
pressure of the grouting pressure, no structural dynamic analysis has to be performed, mainly
because it works as an extension of the foundation rock. In the event that the plinth is not
founded on rock, due to changes in alignment or rock conditions, the structure where it is
founded must be analyzed as a gravity dam, in terms of stability and stresses to ensure that no
deformations occur that could threaten the water tightness of the plane of the concrete face.

2.2.6. Seismic considerations for the design of the concrete face and parapet wall.

A clear understanding of the behavior of a face slabs of the concrete face is essential for a
proper design of a CFRD. As previously mentioned, the concrete face of modern CFRDs are
supported on well compacted and well graded rockfill that should provide a continuous support
under reservoir loading. The slabs of the concrete face of a rockfill dam, due to their geometry
(e.g. thickness, length and slenderness) behave like a membrane. The tensile, compressive and
bending stresses that develop in the concrete face are essentially due to the deformation of the
slabs as a reflection of the deformation of the rockfill under the weight of the rockfill itself and
the hydrostatic load of the reservoir. In addition, stresses develop due to setting shrinkage and
temperature changes. The design of a concrete face must consider all the mechanisms that
contribute to the development of these stresses. These mechanisms include the development of
friction between the rockfill and the concrete face, the three-dimensional deformation that the
concrete face undergoes during construction and filling of the reservoir. However, the difficulty
in obtaining analytical solutions that include all these mechanisms and the absence of
computational tools to calculate the joint effect of the mechanisms acting on the slab
contributed to the fact that the design of the concrete face has been based, until today, on a
large empirical component. A reflection of this trend is the determination of the thickness of
concrete slabs, which is essentially based on precedent. The thickness of the concrete face has
been determined as a function of the reservoir pressure, initially with the formula e=0.30 +
0.003H; in some cases e=0.30 + 0.002H has been used and even dams with heights less than
120 m have been designed with the formula e=0.30 + 0.001H, where e is the thickness of the
slabs and H is the height of the reservoir in meters. Additionally, in the design of concrete-faced
dams, reinforcement amounts in the order of 0.5% have been used, but over time this has
decreased to 0.4% and even to 0.3%, as in the case of the Aguamilpa dam. Even before the

20
recent incidents mentioned in the previous section, it was considered that due to the absence of
vertical cracks, even with amounts in the order of 0.3%, horizontal reinforcement could be
further reduced.

Recent trends to reduce costs in dams of considerable height (> 150 m) include the elimination
of anti-spalling reinforcement at compression joints, further reductions in concrete slab
thickness, and in areas of low seismicity slopes have been increased to 1.30H:1.0V and
1.35H:1.0V, even in high head dams. It is essential that the designs of rockfill dams with a
concrete face, and especially in cases where the precedent is extrapolated, be complemented
with numerical tools that allow the appropriate estimation of the behavior of the different
components of a concrete-faced rockfill dam, especially the prediction of face stress
development

It is important to note that in terms of displacements, the concrete face, being a very slender
structural member, behaves as a membrane and its displacements are controlled by the
deformation of the rockfill. However, this does not mean that there is a compatibility of unit
deformations between the two bodies. The load transfer mechanism from the rockfill to the slab
is due to the development of friction at the interface, which depends on the contact
characteristics. The maximum load that can be transferred by friction depends on the normal
force acting on the contact and the friction angle that characterizes the roughness of the
contact. When the relative deformation that causes the maximum shear stress is exceeded,
slippage occurs between the two bodies and no further load is transmitted.

In addition to the stresses in the slab transferred by friction, there are the bending stresses that
develop in the slabs due to the effect of the resistance of the section to the imposition of the
acting loads and deformations. The deformation imposed on the slab by the effect of the
settlement of the backfill and the load transfer due to hydrostatic pressure induces compressive
and tensile stresses in the section according to the arrangement of the slab. Due to the
slenderness of the slab in the slope direction, the bending stresses generated by the
deformation in this plane are considerably higher than the bending stresses that can be
generated in the horizontal direction because in that direction the face is articulated by slabs
every 13 m, which reduces the generation of bending stresses in that plane.

21
On the other hand, the geometric configuration of the slab, supported on an inclined slope and
in some cases on narrow canyons, involves the generation of mainly compressive stresses due
to the effect of load transfer resulting from its support condition on the backfill. In the case of
hydrostatic load application, the face is subjected to deformation involving compression of the
central zone of the face and tension on its extreme perimeter, the latter represented in the
opening of the perimeter joints.

To consider the effect of the development of the aforementioned stresses, a detailed numerical
model has to be performed with the capability of incorporating the different mechanisms that
generate the stresses induced by movements of the concrete face in response to the settlement
of the fill due to self-weight, reservoir loading as well as the seismic loading. It is important to
mention that a numerical analysis of a concrete-faced rockfill dam can give a general idea and a
trend of the behavior by providing approximate values of the different variables (e.g. unit
stresses and deformations). However it would be inappropriate to expect exact results from
these complex analyses.

Unlike ECRD, CFRDs have traditionally used the benefit of using a concrete parapet wall to
reduce the volume of rockfill. Frequently a single upstream parapet wall is used, however
additional volume savings can be achieved using an upstream and downstream parapet wall.
The verification of the dimensioning, stability analysis and structural design of the parapet wall
is carried out taking into account the most critical operating conditions evaluated based on
loading conditions that include usual and unusual operating conditions, extra-ordinary operating
conditions, and extreme conditions during construction and operation. The seismic load
produced by the peak acceleration defined in the seismic hazard study is applied, including the
amplification at the crest of the dam fills. Additionally, the overpressure in the walls due to the
backfill, which occurs at the time of the earthquake, shall be determined using available
methodologies including for example the Mononobe-Okabe method [Mononobe and Matsuo
(1929) and Kramer (1996)].

22
Seismic Design of concrete face rockfill dams
Rockfill dams, both ECRD and CFRD have performed well during earthquakes. One of the major
differences between these dams is that the upstream shell of the ECRD is saturated whereas in
the CFRD the entire dam body is dry (no saturation). There are only a few cases of CFRDs that
have been subjected to strong ground motions. Some of the few records of the performance of
CFRDs include the dams of Minese and Sugesawa in Japan, Cogoti dam in Chile, Cogswell dam
in the United States, Torata dam in Peru and more recently Zipingpu dam in China.

The Zipingpu dam in China is one of the most relevant recent cases. This dam is located only 17
km from the epicenter of the 8.0 Richter Magnitude and 14km deep 2008 Wenchuan
Earthquake. The 156 m high Zipingpu dam was subjected to a recorded acceleration at the dam
crest of 2.0 g. considering the amplification factor the estimated ground motion was larger than
0.50 g, which is far larger than the 0.26 g considered in the design. The maximum settlement
of the dam crest reached 74 cm and the maximum horizontal displacement was 20 cm towards
downstream direction, in addition to cracks of the concrete face and spalling in portions of the
parapet wall. Nevertheless, the dam stability of the dam was never in question and
demonstrated the safety features of a CFRD during earthquake loading. (Chinese National
Committee on Large Dams, 2009).

This experience of a major dam exposed to an earthquake that doubled the design earthquake
is extremely important. Cracking of the concrete slab was significantly smaller from the cracking
observed in very high dams that under static conditions cracked due to compressive stress
larger than expected. (Campos Novos and Barra Grande in Brazil and Mohale in Lesotho (See
Figure 4). As will be described in the next section, analytical tools are now available capable of
predicting this type of failure, caused mainly by inadequate compaction of the fills or lack of
joints capable of absorbing the additional stress.

23
As it has been already mentioned, the intrinsic resistance capacity of a CFRD dam under
earthquake loading relies on the drainage capacity of the downstream shell. It is very difficult to
imagine a larger destruction of an upstream face than the one suffered in the above mentioned
dams (See Figure 4). The fundamental design concept of a CFRD is that the dam zones,
including the face support material, filters, transitions, under drainage and the body of the dam
remain stable even if extremely large leakage rates were to occur. The ability of rockfill to
accept and pass large flows is well known in the literature. Thus large leakage rates are not
necessarily an indication that safety is a problem, but rather that remedial treatment may be
needed to reduce leakage. However, all these incidents had some positive implications as.
Intensive cracking did not translate into uncontrolled water leakage. The support zone of the
slab in the modern dams has demonstrated its capacity to control the magnitude of the
leakage, as postulated by (Sherard J.L., 1987). The argument presented by some, that includes
the great risk of these dams in seismic zones due to the possibility of severe cracking of the
face, has been losing importance given that it is difficult to imagine more severe cracks than the
ones observed in the Brazilian and Lesotho dams. In those events, the stability of the structure
was never compromised.

When the three serious incidents of the dams in Brazil and Lesotho took place, two very high
dams were being built, one in Mexico (Cajon, 188 m) and one in Iceland (Karajnukar, 200 m).
The rockfill characteristics in these dams were different from the ones used in the problematic
cases. In those cases with undesirable behavior, the rockfill was essentially basalt and the
deformation modulus of the fill was in the order of 50 MPa. In the Cajon dam, where the owner
(CFE) maintained the tendency, initiated several years ago, with the experience of the high
central core rockfill dams, of requiring high levels of compaction, a modulus of more than 120
MPa for the fill was achieved. As a result, the deformations in the face were three times smaller
than the deformations experienced in the dams where the face cracked. The Cajon dam, 188 m
high, built in a narrow valley, has exhibited outstanding behavior. When the gradation of the
rockfill of Cajon dam is compared with the ones built with basalts, the absence of fine material
in the latter ones is evident. This, in addition to the lower compaction efforts, constitute the
main reasons for the differences in the obtained modules. In the Karahnjukar dam in Iceland,
and as a reaction to the events in other dams, additional elements were incorporated to the
design of the face. This included adding a geomembrane to cover the lower part of the slab,
increasing the compaction effort of the materials in the upper part of the fill, and introducing
compressible material in some of the compression joints. It is considered that the adequate
behavior of the dam is not the result of the additional measures taken during construction but
mainly of the differences in the characteristics of the basalt of the rockfills. In Iceland, a
Moberg is a basalt that is cooled below layers of ice and when blasted is converted into a

24
material that has a great content of fine materials. With this material, a very well graded
rockfill is obtained, allowing to achieve higher modules. In fact, the instrumentation records in
this dam demonstrate that the representative modules of these rockfills range between 100 and
120 MPa (H. Perez, 2009) and for that reason, the deformation of this dam, was considerably
lower than the deformations in the Brazilian dams and Mohale dam in Lesotho.

In several parts of the world, opposition towards this type of dams was generated, given the
incidents reported. These cases had greater impact than the cases of dams with clearly
satisfactory behavior. It has been observed how, by maintaining the criteria of building these
dams by precedent, even in dams with less heights and with very good materials, additional
elements were being introduced (compressible materials in the joints, slabs with less width,
greater requirements for the fillings, etc.), that increased the dam cost unnecessarily.

Some practitioners have argued that the great risk of CFRD’s in seismic zones is due to the
possibility of severe cracking of the face. It is difficult to imagine more severe cracks than the
ones observed in the Brazilian and Lesotho dams. In those events, the stability of the structure
was never compromised. Based on this experience if any cracking is observed after an
earthquake immediate remedial measures could be applied as they were used in the very old
dams where large leakages were recorded. These consist mainly in dumping sand materials
above the cracked portions. For these measures to be successful it is essential that the face
support zone does not crack precluding a filter effect. This characteristic depends essentially on
the gradation of the material used. Sections 8.2 and 8.3 of ICOLD 141 Bulletin (ICOLD 2010)
are the best reference available on this subject. It contains a description of how the
recommended gradation has varied through time searching for a material that precludes
segregation and prevents crack formation.

The main effort to obtain an adequate behavior should be concentrated in obtaining the best
possible zoning of rockfill with the available material, based on gradation, sluicing and
compaction and providing the best possible concrete face for the predictive behavior.

The current computation power available, allows to perform, during the design and construction
stages, numerical analyses using techniques such as the finite element method that provide a
more adequate and comprehensive framework to evaluate the behavior of CFRDs.

25
In recent years several high CRFD have been and are being built for which detailed analysis
have been performed. Several dams throughout the world have used numerical analysis as an
important element in the design process including the selection of rockfill material (fill zoning)
and in the introduction of additional elements in the face. The issue of material selection is very
important and good practice in the case of high dams, where the rockfill deformation is
essential for the good behavior of the concrete face, and should be based on thorough
investigations of material characterization and real experiences. The idea that a good rockfill is
the one composed of very hard particles is no longer valid. The research of Marsal (Marsal,
1973) corroborated the idea that to reduce the deformations of the rockfill, including creep,
there should be a special emphasis on the gradation. A rockfill with not very hard particles, but
properly graded, results in lower deformations than a rockfill with very resistant particles but
poorly graded. The Karahnjukar dam, previously described, clearly illustrates this point.

Porce III (154 m high) is to the authors knowledge the first high dam (above 150m) that
demonstrated that it is possible to build a high CFRD dam with materials that in previous
circumstances could have been considered as inadequate for a rockfill. The location of the
spillway at the left margin required the excavation of a quantity of material somewhat greater
than the volume of the dam. As a result, using this material represented an economic incentive
for the project. However, the material consisted of a sequence of schists of different
characteristics, varying from graphitic schists of low mechanical properties to more competent
granitic schists. Weathering of the material near the surface also reduced the mechanical
characteristics of the material. As a result, the initial spillway excavations produced a highly
weathered material, unacceptable in the upstream shell.

Trial fills were constructed with different materials derived from the spillway excavation,
including rockfill mixes with materials derived from different areas of the spillway (graphitic or
granitic schist and different levels of weathering). Odometer tests and plate load tests were
performed, taking the load to relatively high values and creating reactions with jacks against
two 35 ton trucks, with plates having a radius of 0.75 m. The field tests, complemented with
laboratory results such as LA abrasion and absorption, indicated variable results. The graphitic
schists produce results to abrasion as high as 75 per cent. Based on the tests and the results of
the trial fills, expected ranges for the deformation modules were established for the different
materials. This information was the basis to perform sensitivity three dimensional finite element
analyses (A. Marulanda, 2007), (A. Marulanda, 2011) to establish the effects of using different
types of material mixes in the behavior of the dam and especially of the concrete face. At the

26
same time, it was necessary to perform a matrix of origin and destination of the excavation
materials to determine the capacity of producing the amount of materials that the adopted
zoning required. From these analyses, it was determined that the material for the upstream
zone could not have less than 85 MPa, while materials with greater deformation were allowed
downstream of the fill. Given that, the use of materials with less mechanical characteristics
result also in a decrease in the permeability, a vertical drain was connected to a horizontal drain
to prevent a high water level in the downstream shell. Figure 6 shows the main section of Porce
III dam with the instrumentation of the fill.

Figure 6. Porce III Dam Main section

The instrumentation data were used to calibrate the numerical analyses to further estimate the
deformations and stresses in the concrete face and its joints. These analyses indicated that the
horizontal compression stresses in the central part of the dam could reach values closed to 20
MPa. From the dams that have suffered failures by compression in the concrete face, Mohale
dam was the only one that had records of strain gauges that could be used to estimate the
stresses on the face. These records indicated that the failure by compression, caused by both
the horizontal stresses and the ones parallel to the slope, reached values before failure of
around 18.5 MPa (P. Johannesson, 2007). Based on this information, it was defined that for the
Porce III dam the criteria adopted consisted in rejecting dam configurations in which the

27
estimated stress could surpass 15 MPa. As a basic measure to reduce the development of high
compressive stresses in the face, a compressible element was introduced in the vertical
compressible joints (compressive joints fill with a compressible material to dissipate stresses).
The decision to incorporate these compressible joints was based on detailed numerical
analyses, where their effect in reducing in an effective way the levels of stresses was clearly
demonstrated. The analyses also helped define the dimensions (width) that could be used and
the number of compressible joints to be implemented.

One important aspect related to the incorporation of these elements was the selection of the
material and the placement method. It was decided not to use rubber tape, considering that it
will leave spaces in the opening of the joint with the obvious adverse consequences. As a result,
it was decided that the material should be placed in liquid form to ensure that all the joint space
was filled and to have complete adherence with the wall of the joint. Another criterion for
selecting the material was its consistency with time, to guarantee a high enough modulus to
prevent closing with low stress levels. The selected material was a polyurethane and a joint
opening of 3 cm was adopted. Laboratory tests were performed looking for the adequate
mixture of the chemical elements to produce a material with a modulus in the order of 10 MPa
under the predicted stresses levels. An additional line of defense was incorporated, and
consisted of a PVC cover filled with ash-type material, similar to the design of typical extension
joints (see Figure 9). When these analyses were being completed, there were suggestions to
increase the compaction in the upper part of the dam as applied in the Karahnjukar Dam.
Consequently, analyses were performed to simulate the effect of this measure in the behavior
of the concrete face. These analyses indicated that the effect was negative since an excessive
curvature of the face could be generated, introducing undesirable stress levels in the face.

Figure 7 illustrates the final configuration of the concrete face of Porce III dam, including the
implemented instrumentation and compressible joints. The reservoir impounding started in
October 2010 and was completed two months later. The behavior of the face, and the levels of
stresses measured, confirmed a good behavior of the structure. The maximum deformation of
the face was 29.8 cm in the maximum section.

28
Figure 7. Porce III dam concrete face with compressible joints

Another example to illustrate the positive use of compressible joints is the Yedigoze dam in
Turkey. The dam is 130 m high, and was built with limestone rockfill and a relatively low
modulus fill (45 MPa) was obtained. The numerical analysis indicated the need to introduce
compressible material in some of the central compression joints. The same technology for those
compressible materials described previously for Porce III was used; however, it was decided to
increase to five the number of central joints with compressible material. In this case, in spite of
the low modulus of deformation obtained in the rockfill, the maximum stresses by compression
were reduced to values below 10 MPa.

Regarding the issue of introducing compressible material to some joints in the face, it is worth
emphasizing that the decision to use this type of element in the face requires a careful

29
evaluation of its implications. It is important not to forget that the joints of a CFRD, constitute a
weak point to the impervious element that represents the face. Therefore, by adding the
compressible material, its installation procedure has to be very carefully performed looking for
an adequate bonding between the concrete and the compressible material to avoid having
unrestricted water paths. As previously described, it is in the authors’ opinion that if the
compressible joints are to be implemented, they should be protected with a PVC cover filled
with ash-type material, similar to the design of a typical extension joint (Figure 8).

Figure 8. Yedigoze Dam - a) Aerial view of the dam, b) cover of compressible joint and c) stress
measurements in the concrete face

The Yesca Dam in Mexico, 205 m high, immediately upstream of the Cajon Dam was finished in
2013. In this dam, compressible material was also used in 12 vertical compression joints with a
material with a modulus of 100 MPa, however, the main approach consisted in minimizing the
deformations of the face using materials of very high modulus (low deformation). The upstream
shell of the dam is composed of clean gravels, followed by a transition zone and downstream
material, both composed of rockfill. The rockfill materials are truly exceptional given that the
tests before construction and the data from the instrumentation revealed modules higher than

30
150 MPa for the rockfill and higher than 250 MPa for the gravels. In the case of the rockfill,
these values are the result of a material with very hard particles but also with a relatively low
maximum size and with very good gradation. The materials were placed in maximum
thicknesses of 60 cm and compacted with a vibrating roller of 18 tons. The three dimensional
finite element analyses indicated that with these values, both the deformations of the face and
the stresses were inside the acceptable ranges. After impoundment, the concrete face
deflection was in the order of 35 cm.

Figure 9. Details of a typical compressible joint

Sogamoso dam in Colombia, represents a good example of the modern approach that should be
applied in the design and construction of CFRD, where experience, precedent and empiricism
are complemented with thorough numerical analyses to define an adequate engineering
solution. The Sogamoso Dam, 192 m high, is located in a narrow canyon with an A/H2 of 2.32
(see Figure 10). The dam construction followed the same procedures previously described used
in Porce III. During the design stage, a detailed evaluation of the materials foreseen for the
dam was performed, the design was based not only in experience and precedent but was
complemented with numerical analyses that help improve the zoning of the fill and the

31
excavations for the plinth. The final configuration of the dam is shown in Figure 10. The initial
sensitivity analyses performed during the design, indicated that if the fill material could result in
a modulus of deformation higher than 120 MPa, no compressible joints were needed. The tests
performed during construction, complemented with the instrumentation, indicated that the
upstream fill, consisting of gravels, could be associated with a modulus of 150 MPa and the
more sandy material from the downstream shell to a modulus of 90 MPa. During construction,
systematic monitoring of the quality of the materials was conducted through field and
laboratory tests. In addition, the deformation of the fill was recorded using magnetic
settlement plates and vibrating wire settlement cells. With the deformability records and the
numerical analyses, it was determined that there was no need to install any compressible
material in the central joints of the concrete face. The stresses that have been calculated from
the strain gauges installed in the face, have shown that the largest stress is around 5 MPa.

The largest event that the Sogamoso dam has experienced occurred on March 10 2015 when
6.2 moment magnitude (Mw) earthquake took place. Colombian Geological Survey (SGC) using
Colombian National Seismological Network (RSNC) records defined that the event epicenter was
located at a depth of 161 km. l and 42 km to the dam site.

32
Figure 10. Sogamoso dam section, zones and aerial view

In the case of the March 10th (2015) seismic event, maximum peak ground accelerations of
0.098 g, 0.077 g and 0.020g where measured at the crest, middle of the dam and rock outcrop
for the horizontal component perpendicular to the dam crest axis (y axis on Figure 11. A

33
significant amplification occurred. The PGA was amplified by a factor close to five between the
rock outcrop and the crest) due to wave propagation throughout the dam body. Figure 12
presents the seismic response spectra for y direction of motion for the three earthquake
records.

Figure 11. Accelerometer location at Sogamoso Dam site

The analysis of the earthquake records allowed to determine a fundamental frequency for the
dam of 0.6 Hz (period 1.67 sec). Based on the period calculated and using Makdisi and Seed
(1978) method it was possible to estimate an effective damping of 1.6% for the dam for this
low intensity event and an average shear wave velocity of 330 m/s for the dam rockfill body.
Makdisi and Seed's simplified procedure allowed to estimate a PGA value at the crest of 0.094 g
that compares reasonably well with the 0.098 g measured value (4% difference). Seismic
behavior of the dam confirms the construction premise that rockfill was properly compacted,
achieving high construction moduli for the dam body.

34
Figure 12. Seismic response spectra of the March 10th, 2015 earthquake event. Horizontal
component perpendicular to the dam axis (y axis)

It should be noted that the design PGA value for Sogamoso dam was 0.26 g wich is 13 time
larger than the measure PGA value at rock outcrop for the March 10th seismic event which
explains the fact that no permanent displacements (settlements) were measured for this
seismic event and also explains the low value of damping associated with the overall dam
response.

It is expected that for the design earthquake event, significant shear modulus degradation
should occur inducing an increase of the dam fundamental response period and the overall
effective damping for the wave propagation throughout the dam. Although the intensity of this
event was low, it allowed to validate design premises such as the rockfill stiffness and a low
fundamental period for a 190 m high CFRD.

The experiences described demonstrate that it is possible to build high CFRD and obtain
excellent behaviors. In addition, assuming that all CFRD require the same details in the
joints/face or in the rockfill is not always necessary and in many cases could create problems

35
that otherwise would not be presented, e.g., increase the number of slabs or leave too many
compressible joints with excessive spaces.

Methods of analysis for seismic response of Rockfill dams

The methods of analysis currently used in practice to evaluate the seismic response of rockfill
dams follows similar practice to embankment dams and vary widely, ranging from empirical
correlations to estimate seismic induced deformations, simple limit equilibrium analysis to highly
advanced numerical analysis methods. There are intrinsic differences between ECRDs and CFRDs
which are essential to understand to adequately analyze the seismic response of each type of
dam and the implications of the methods used for a particular seismic evaluation.

Performance of rockfill dams during earthquakes


There are several studies that have compiled the performance of embankment and rockfill
dams. From a detailed review of past experience, it can be concluded that rockfill dams have
performed satisfactorily, nevertheless for the evaluation under seismic loading the possible
failure mechanisms of each type of rockfill dam (ECRD or CFRD) have to be understood.

The first failure mechanism to consider is overtopping of the dam. The evaluation of the seismic
response of all rockfill dams depends on the magnitude of the expected seismic induced
deformations. If the deformations exceed the freeboard of the dam, erosion from overtopping
can cause the failure of the dam. This failure mechanism applies to all rockfill dams and the
methods presented in the following section are of imperative importance to adequately estimate
earthquake induced deformations of the dam.

A second failure mode is cracking and internal erosion. If the seismically induced deformations
generate cracks or distortion of the filters the dam may fail due to erosion. The amount of
deformation a dam can withstand without generating cracks highly depends on the materials of
the dam and its foundation, the details of material zoning, especially filters and transition zones.
This mode of failure apply more specifically to ECRD, although there is the reported case of the

36
internal erosion failure of a CFRD Gouhou dam in China constructed mainly of sand and that was
not permeable enough to allow free drainage and failed due to erosion when water came into the
dam through a damaged concrete slab. This highlights the importance of having an adequate
well-compacted, free draining rockfill as has been emphasized previously in this chapter

Several studies have published performance of rockfill dams during earthquakes over the years.
Some of the most relevant work on the topic includes the compilation of behavior reported by
USSD in three separate volumes of Observed performance of dams during earthquakes, started
in 1992 and with its latest volume in 2014. Other pertinent references include (Swaisgood, 2003),
(Ishihara, 2010), (Pells, 2002). The response of several rockfill dams that have been subjected
to considerable ground shaking has been registered in the form of crest settlement against the
intensity of the shaking. Figure 14 is based on the data gathered first by Swaisgood (2003) and
includes more recent records reported on rockfill dam behavior. The band presented in Figure 13
represents a first order estimate of possible deformations due to earthquake loading and does
not include any case where liquefaction or strain softening has occurred. For dams where the risk
of liquefaction exists, the methods presented in the following section should be considered to
estimate the response of the foundation under ground shaking.

Seismic Loading Characterization

The state of practice for dam safety evaluation has evolved from a conventional approach, in
which dams are required to withstand the most severe loadings possible at the site without failure,
to the use of probabilistic risk analysis. In the design of a new dam or in the vulnerability
assessment of existing dams, it is necessary to evaluate the structure performance and its
probability of failure under both extreme earthquakes and smaller ones. It is not unusual in a risk
analysis that the greatest contributions to the risk result from smaller earthquakes because of
their much higher probability of occurrence, even if the probability of failure is much smaller in
the smaller earthquakes (USBR, 2015).

For detailed analyses, the development of Probabilistic Seismic Hazard Analysis (PSHA), and
earthquake ground motions for selected return periods is necessary. The PSHA results provide
curves indicating the annual probability of exceedance for PGA and spectral acceleration for

37
selected oscillation periods (ICOLD, 2016). For simpler seismic geotechnical analyses, this may
be all the information that is required, but for more detailed analyses, the hazard curves are used
to select ground motions associated with a particular return period. This allows the estimation in
a risk analysis of the probability of dam failure resulting from loadings with various annual
probabilities, so that an annual probability of dam failure can be calculated.

For a given earthquake scenario (a combination of fault geometry and area of fault rupture) the
values of PGA and spectral acceleration, and the ground motions vary with location relative to
the dam and foundation. Estimates of peak ground acceleration and/or ground motions are most
often provided for a level rock outcropping or the surface of stiff soil. However, the widely used
Boulanger and Idriss (2014) simplified method for assessing liquefaction potential (described in
later sections) requires as input the peak acceleration of the soil surface, at free-field soil surface,
which can be significantly different than at level ground rock outcrop. The acceleration of the
embankment crest can be even higher because the shape of the embankment cross section
affects the response. The simplified method was developed from 1D response analyses for level
sites, so it is not strictly correct to apply it under a dam embankment. As needed, hazard curves
and earthquake records to represent motion at a soft soil surface can be developed, at the contact
between soil and bedrock, or at a greater depth in the foundation. One-dimensional response
analysis using (Hashash Y.M.A, 2020) or a similar program can use a record representing each of
the previously mentioned locations. FEM or FDM usually require the ground motion to be put in
at the base of the model (deconvolved motion within bedrock). To obtain a record for that
location, usually, a rock-outcrop record is numerically “deconvolved” to account for the response
of the material between the base elevation and the surface.

Methods of analysis for seismic induced deformations

This section does not intend to include thorough description of the methods of analysis to evaluate
the stability conditions of a rockfill dam, rather, a description of the main procedures available as
a state of practice in dam engineering and its applicability and limitations.

Under large seismic loads the main possible manifestation of a rockfill dam is permanent
deformations. However that does not mean the dam has failed provided the deformations are

38
tolerable. Hence the emphasis shall be on deformations, not factors of safety. It is only dams
which experience liquefaction, or are founded on strain weakening soils resulting in post-
earthquake factors of safety below 1.0 that have a stability issue.

For seismic analysis of dams, neither theory nor precedent is adequate on its own. Numerical
analyses may be based on sophisticated theory and constitutive models based on laboratory
testing, but the constitutive models are merely idealizations of nonhomogeneous materials,
generally treating them as continua. Analytical results are necessarily based on a number of
simplifications and assumptions that cannot be verified. Therefore, back analysis of historic
performance of dams in earthquakes is needed to validate the models. The number of well
documented and instrumented case histories is limited, and embankments and foundations are
all different, so it is very unlikely that there would be directly relevant case histories that could
provide a clear, purely empirical indication of the behavior of a particular dam in a particular
earthquake. Theory and back analysis are needed to understand the case histories so that the
lessons can be extended to the dam under study through forward analysis. This requires
knowledge of both the underlying theory, to understand historic performance, and the historic
performance to validate and calibrate theory and analytical methods.

Empirical methods

If liquefaction does not occur, settlements within the dam body without distinct signs of shearing
displacement can lead to deformation that exceeds the available freeboard. Swaisgood (1998,
2003, and 2014) examined case histories of seismic-induced settlement and mass deformation
where the earthquake shaking causes embankments to settle downward and sideward, toward
the deepest center portion of the valley, and then spread upstream and downstream away from
the dam axis. In the Swaisgood empirical methodology, the crest settlement is expressed as a
percentage of the total embankment height and foundation thickness, as shown on Figure 14.
The crest settlement (given that no liquefaction occurs) is given as a function of PGA and surface
wave magnitude (Ms) as shown on Figure 15. The incident database does not contain any cases
with PGA greater than 0.7g or normalized settlements greater than 5 percent. USBR included
some incidents involving liquefaction in the cases shown on Figure 19 (USE D-8 USBR Report),
Hebgen Dam (1959), Upper San Fernando Dam (1971 and 1994), and Masiway Dam (1990).
Austrian Dam (1989) did not experience liquefaction but had other issues like poor compaction

39
and an existing slide left in place in one abutment. If these cases are excluded, the incident
database does not contain any cases with normalized settlements greater than 1 percent.

Pells and Fell (2002, 2003) gathered data from 305 dams, 95 of which reported cracking, and
classified these for damage according to the system shown in Table 1.

Table 1. Damage classification system (Pells and Fell 2003)

Figure 13 shows plots of damage contours versus earthquake magnitude and peak ground
acceleration for earth and rockfill dams with respect to crest settlement and cracking. These
figures can be used to estimate the magnitude of crest settlement based on the range of relative
crest settlements given for each damage class given in Table 1.

40
Figure 13. Contours of damage class versus earthquake magnitude and peak ground acceleration. After Pells and Fell (2003)

41
Figure 14. Earthquake damage as a function of crest settlement after Swaisgood (2014)

42
Figure 15. Settlement as a function of the PGA

43
Figure 16. Damage as a function of crest settlement after Swaisgood (2014)

Simplified methods for deformation analysis

One of the most common geotechnical analyses for a dam involving seismic loads is the use of a
pseudo-static coefficient (kh) in a limit equilibrium analysis. (Hynes-Griffin, 1984) proposed the
use of kh=0.5 PGA limiting the maximum displacement to 1 ft. This recommendation has been
used widely for rockfill dam design. More recently, (Stewart, 2003) and Bray and Travasarou
(2009) have proposed methodologies to determine the pseudo-static coefficient as a function of
the PGA, the earthquake magnitude and the threshold displacement. In the Stewart et al., (2003)
screen methodology for a high magnitude earthquake event (Mw=8.0), the pseudo-static
coefficient varies from 0.50 to 0.62 times PGA for a displacement threshold of 15 cm and 5
respectively. (Bray, 2009) method provides similar ratios between the pseudo-static coefficient
and PGA than (Stewart, 2003) and (Hynes-Griffin, 1984) recommendations.

44
(Saragoni, 1993) proposed an equation (Equation 1) to calculate the pseudo-static coefficient as
a function of the PGA for gravitational walls. This equation has been used for developing pseudo-
static analysis for dams and tailing dams in Chile, however, it provides values that can be as low
as kh=0.2 PGA for a PGA of 1g or kh=0.3 PGA for PGA values lower than 0.67 g. This means if
the Saragoni (1993) equation is used this can provide a pseudo-static coefficient 1.67 to 2.50
times lower than the value calculated using Hynes-Griffin and Franklin (1984), Stewart et al.,
(2003) and Bray and Travasarou (2009) recommendations.

Equation 1.

Based on analyses performed by the authors and results from the screen methodologies
developed by Stewart et al., (2003) and Bray and Travasarou (2009) it is recommended to use a
pseudo-static coefficient (kh) between 0.4 to 0.5 times PGA for the simplified calculations
developed as part of the conceptual and feasibility design stages

Macedo et al., (2017) have shown that the calculation of the pseudo-static seismic coefficient
needs to take into account the earthquake event mechanism. The difference on frequency
content, duration, magnitude scaling that subduction earthquake records show compared with
records of shallow crustal earthquakes result in variations on the estimated displacements and
therefore in the coefficients that should be utilized in the simplified assessment.

As a first order estimate of earthquake-induced displacements and before any advance numerical
analysis is carried out simplified procedures should be performed. The Newmark (1975) model
could be used, using the approximate procedure developed by Makdisi and Seed (1978). The
Newmark model indicates that when the induced acceleration is greater than the critical or creep

45
acceleration, a potential sliding mass experiences a displacement. For the simplified iterative
procedure of Makdisi and Seed, the following input data are required: the PGA of the site, the
height of the dam, the maximum shear modulus of the material along with its shear wave velocity
and density. The critical or creep acceleration, Ky, is determined from two-dimensional pseudo-
static limit equilibrium stability analyses performed taking into account the critical strength
condition. An iterative process should be carried out to calculate the maximum acceleration at
the crest and the natural period of vibration of the dam. This process requires the use of the
shear modulus and damping variation curves and the elastic response spectra of the signals
selected as representative of the dam’s SEE safety verification earthquake. Since the (Makdisi F
and Seed H.B., 1978) methodology is used for a simplified analysis, only a single shear modulus
and damping variation curve with cyclic shear deformation can be used. Therefore, in these
analyses, the curve proposed by (Rollins, 2020) or Zhou et. al., (2017) for rockfill can be
considered.

A more recent simplified procedure to estimate earthquake induced permanent deformations is


the one proposed by (Bray, 2009) which uses a Newmark type model and incorporates the
flexibility of the dam using a nonlinear coupled stick-slip deformable sliding model.

Advanced numerical methods

Advanced numerical methods including nonlinear deformation analyses using finite element or
finite differences methods are frequently used not only for the design of new dams but also for
evaluating the response of existing dams during earthquakes. These evaluations include the
estimate of permanent deformations due to ground shaking and also the effects of liquefaction
on the dam. The fundamental differences between the ECRD and a CFRD dam described
previously have important implications on the requirements for an adequate numerical analysis
that satisfactorily predicts the dynamic response of each type of structure.

For an ECRD, a numerical analysis should be a coupled analysis to determine the stress, strain
and pore pressure fields over time in response to boundary conditions and loads applied to the
simulation domain, including the ground motion. The analysis is called coupled since it requires
the solution of two three-dimensional fields: the displacement field and the pore pressure field.

46
Figure 17. Materials and interfaces of a typical earth core rockfill dam (ECRD).

For the dynamic analysis not only of new dams but existing dams as well, the modeling of the
construction process of the dam allows to adequately estimate the state of stresses and
deformations reached throughout its construction, which constitutes an adequate tool to
represent the loading history of the materials and to establish a state of pore pressures and state
of stresses closer to reality. Therefore, it is necessary to consider the construction process of the
dam in an adequate way to perform the calibration of the constitutive models from the
instrumentation records as well as to start the dynamic analysis in a stress state closer to the real
state of the dam. Figure 17 illustrates the materials and interfaces of a ECRD numerical analysis

A numerical model of an ECRD shall consider the body force associated with the self-weight of
each of the construction stages and the hydrostatic force of the reservoir, which is applied
progressively according to the reservoir filling curve. In addition, the dynamic analysis shall
consider the hydrodynamic effect of the reservoir and a Rayleigh-type viscous damping, unless
explicit hysteretic damping is included in the material constitutive model.

In the case of an analysis for a CFRD, its behavior is controlled by the three dimensional nature
of the problem. To obtain accurate enough results from the analysis, it is essential to develop a
three-dimensional numerical model. The model should include the topographic features of the
site and incorporate the different stages of construction, including rockfill placement, reservoir
filling and the construction sequence of the concrete face. Nevertheless, in comparison with the
analysis of an ECRD no coupled analysis is required considering that the rockfill is highly
permeable and non-pore pressure will be generated. Figure 19 illustrates a typical geometry of a
finite element model developed based on the topography of a particular project site. Some other
important aspects of a finite element analysis for CFRD include:

● The importance of modeling the detailed construction sequence adequately. Simulation


of such activities in a numerical analysis should be performed with caution. The code
must be able to consider several issues, including: deactivation prior to the analysis of
the elements representing the material to be constructed, and the addition of weight to
the added constructed element by applying self-weight body forces. This procedure
should be checked so that no unrealistic deformations are imposed to the model.

47
● The type of stress-strain relationship used for modeling materials. The choice of the
stress-strain constitutive model used in the analysis is a balance between simplicity and
accuracy, and the choice of the constitutive model will depend on the purpose of the
modeling. If the purpose of the analysis is to analyze stresses and/or deformations, then
more simplistic models could be suitable, depending on the relative stiffness between
different fill materials (Marulanda E, 2008). More accurate modeling of deformations
requires a more complex stress-strain model that more realistically approximates the real
behavior of the rockfill. Elasto-plastic models that model pre-peak plasticity are more
suited for modeling the deformation behavior of the upstream face of CFRD during
reservoir impounding due to the ability of these constitutive models to more realistically
account for the rockfill deformation. Considering that the displacements observed in this
type of dams have led to the conclusion that the rockfill present an anisotropic behavior
in which the moduli in the horizontal direction are greater than those measured vertically,
the ratio is defined as the ratio between the deformation moduli in the horizontal and
vertical direction (Figure 18). Typical values of this ratio vary between 1.5 for gravels
and 4.0 for rockfill.

Figure 18. Variation of the ratio of the tangential and vertical modulus of the rockfill to the geometry of the dam. Data taken from:
ICOLD (2010)

48
● The importance of adequately modeling the concrete face. The simulation of the
concrete face involves: (1) modeling of the physical concrete face; (2) modeling of the
concrete face construction; and (3) modeling of the concrete face-rockfill interaction.
Bending behavior of the concrete slabs in the vertical direction, as well as the transferring
of compressive stresses in horizontal and vertical direction, are the most important
aspects that need to be captured correctly in the model. For the analysis of the concrete
face it is important to model each slab independently with an interface element between
slabs to simulate the joint behavior. To accurately model the bending action and
transferring of horizontal stresses, the real slab thickness and the concrete face weight
should be included in the analysis

● Incorporation of different interfaces. In a soil-structure interaction situation, relative


movement of the structure with respect to the soil can occur. Interface elements, or joint
elements, as sometimes referred to, can be used to model the soil-structure boundary
such as the sides of the concrete face. For an analysis of a CFRD, the following interfaces
should be considered: bedrock-rockfill interface, curb-rockfill interface, concrete face-
curb interface, and interface between slabs (See Figure 17)

Figure 19. CFRD Three Dimensional model include the recommended interfases to considering in a numerical model

49
In order to characterize the behavior of the rockfill under dynamic loading, several models exist
to represent the stress-strain and damping behavior of geomaterials. To describe the response
of the rockfill to dynamic loads a non-linear cyclic model shall be used in which the stress-strain
curve for monotonic loads follows a particular shape (often hyperbolic) that describes a
degradation of the shear modulus as the shear deformations increase (skeleton-curve) in Figure
20. To represent the load-discharge behavior, the model proposed by Masing (Numanoglu, 2018)
is often adopted, which defines the rules of behavior of the hysteresis cycles to reproduce a
damping mechanism that is independent of frequency. Characterization information on the
dynamic properties of rockfill is limited; shear wave velocity values are commonly presented,
which allow estimating the maximum shear modulus, but are insufficient to characterize the shear
modulus behavior for different levels of deformation. One of the compendiums of shear modulus
degradation curves and damping ratios is the one reported for rockfill and gravels by Rollins et
al. (1998 and 2020) (Kyle M. Rollins, 2020), (see Figure 20)

50
Figure 20. Degradation of shear modulus and damping with strain rate

In addition to the deformations induced by the reservoir under static conditions, the dynamic
response of the reservoir during an earthquake induced displacements in the dam that must be
considered in the analysis. Hydrodynamic forces exerted on a dam due to a seismic event are
equivalent to the inertial forces of a volume of water attached to the dam moving while the rest
of the reservoir remains inactive (Pelecanos, 2020).

51

You might also like