case 2: Tunnel bore stable, tunnel heading not stable
11.4 Dam foundation design and construction in recent years: case histories
11.4.1 Geomechanical models of dam abutments: design and
The technique of testing dams, mainly arch dams, on models was first developed in the United States (Bureau of Reclamation) and in Italy and was rapidly adopted by other countries. The first models were plaster, hard light concrete, or rubber, built on rigid hard concrete abutments. Then the tech- nique developed and the rock abutments were represented by a material with a convenient modulus of elasticity. The latest stage is to represent rock
348 Rock mechanics and dam foundations
stratifications, fissures and faults on the model and test them until rupture occurs. Several typical examples are instructive.
(1) The Vajont dam. The 261-6-m-high Vajont dam (fig. 11.26) is built on highly jointed limestone which can best be described by its modulus of
test gailery
level 682 v test'gallery
test galleries at
level 555 test galleries
at level 555
0 20 40 60m y_y_yj
Fig. 11.26 Vajont dam, measurement of rock displacement.
elasticity E which varies from 787 000 to 852 000 kg/cm2, measured on labora- tory samples, the higher values of E being those for rock at the bottom of the valley. Inside test gallery measurements of E of from 100 000 to 300 000 kg/cm3 were obtained. A model of the dam was built and these varying E values carefully reproduced, so that stresses and strains measured on them could be compared with those obtained on more conventional models of homogeneous abutments (Pancini, 1961a, b).
The main problem at Vajont was not the modulus of elasticity of the rock, which varied between known and reasonable limits, but the joints and fissures in the rock. A model was therefore built in which the abutments were formed by precast parallelepiped blocks of plaster with barium sulphate piled one on top of the other so that the actual contour lines and the three main directions of the joints were correctly reproduced. The joints were filled with different types of glue to represent different types of roughness coefficients. In addition, several rock fractures, some of them parallel to the main direction of the valley were correctly reproduced by sawing the blocks in the appropriate places.
The deflection of the crown was about five times larger on the model with the lowest friction factor than on the model with massive, homogeneous, non-fissured rock abutments. Measurements made on the prototype dam showed the actual deformation to be larger than on the more optimistic model, but still far less than on the worst model. The models were tested to
breaking point. One collapsed when, due to the accidental rupture of a watertight rubber sheet protecting the abutment, water seeped through the pile of blocks, simulating the fissured abutments. This may be taken as proof of the very real danger from the seepage of water and full uplift pressure in rock fissures and joints.
In addition to these tests, a considerable number of observations and measurements were made on the site to ascertain the rock properties and rock displacements. Vibrations inside the rock mass were measured during rock blasting and concreting. The wave velocities in rock and dynamic modulus of elasticity were measured before and after grouting. Five test galleries were driven. The modulus E was measured by applying compression forces either with hydraulic jacks or by filling the galleries with water under pressure. Tests were made in galleries to determine the cohesion and friction along weak rock seams. Displacements of the rock abutments were measured both in the direction of the dam thrust and perpendicular to it. Geodetic measurement of any displacement can be supplemented by observation of pendulums hung in vertical shafts in the rock. Displacements of the embank- ments of the reservoir are also checked from the geodetic net. Cables were used to prestress the rock in the upper part of both abutments, and their length and strain were measured and checked. (The Vajont rock slide will be analysed in chapter 14.)
(2) Kurobe IV dam, Japan. Model tests carried out for the Kurobe dam (Oberti, 1960) included those where the correct ratios Econcrete/Erock were used on the model and the main rock faults represented. In addition, detailed tests and calculations (with an analogue computer) were carried out to determine the piezometric pressure lines of water percolating through the rock abutments (Yokota, 1962; Pacher, 1963). Particular attention was paid to the grout curtain and to the drainage system, both of which differ in many ways from conventional designs. The grout curtain is displaced in the upstream direction.
An elaborate and costly campaign was launched (John, 1961) to:
(a) record all the rock fissures and their characteristics (direction, dip, thickness, etc.),
(b) test rock under compression and under shear inside galleries,
(c) measure the physical rock properties and its strength near the founda- tion at the surface of the rock,
(d) measure the modulus of elasticity in the rock.
In fig. 11.27a the fissures on the right abutment of the dam are shown and in fig. 11.276 the arrangements for inclined shear tests on rock in situ, isolating a rock block between two parallel galleries. Figure 11.27c shows similar tests carried out at the rock surface. Arrangements for triaxial tests
350 Rock mechanics and dam foundations downstream face vertical cantilevers
abutment
open rock / faults filled with concrete
vertical joints / /
abutment poor rock replaced with concrete
concrete block
rock
Fig. 11.27 The Kurobe IV dam: (a) the dam; (b) shear test on a concrete block;
(c) shear test at the surface of a rock mass.
on filling material inside a wide rock fracture are similar to those shown in fig. 11.276.
(3) Tang-e-Soleyman dam {Northern Iran). The dam has a height of 100 m with a central angle of 73°. The radii of the reference surface are 124 m for the central zone and 211 m for the lateral zones (fig. 11.28). Comparative
Fig. 11.28 Schematic representation of the failure of the model of the Tang-e- Soleyman dam (Iran): (a) fissure due to bending; (b) shear failure in the concrete, along the dam foundation; (c) horizontal shear fracture (after Lane & Serafim,
1962).
tests were carried out on homogeneous and on heterogeneous models. The values of E shown in table 11.1 were measured in kilograms per square centimetre by the Lisbon laboratory (Lane & Serafim, 1962).
Table 11.1 (in 1000 kg/cm2)
Transitional Sandstone Mudstone Limestone limestone Oven-dried
Saturated Jacking tests
in a gallery 185 115 70 to 570
238 188 160 to 500
798 715 90 to 420
541 535 50 to 60
Tests were first carried out in Lisbon by Rocha and Serafim on two models of a mixture of plaster of Paris (P), diatomite (D) and water (W) in a pro- portion of WjP = 2 and PjD = 2. The modulus of elasticity of the founda- tion rock was assumed uniform and equal to that of the dam. The models were loaded with mercury.
For a second stage of the study two other models were built, the ratios Er0C)xilEconcrete being 1/2 for limestone and 1/4 for sandstone and mudstone.
The mixtures used were:
W\P = 2, DjP = 0-5, for the dam.
W\P = 2-9, D/P = 1, for the limestone.
W\P = 1-2, D\P = 5, for sandstone and mudstone.
Due to the increased deformability of the foundation, the arch effect became more marked, especially in the upper part of the dam, and the canti- lever effect decreased. It is important to note that in such circumstances the maximum compressive stresses in the entire dam decreased, the results showing much smaller bending moments. The compression at the down- stream face on the crown of the arches increased, particularly in the upper arches. This was certainly caused by the fact that the upper region of the foundations was more rigid than the region below.
Tests were carried out until rupture of the model dam occurred. Rupture took place either by vertical compression fissures starting in the area of weak foundations when the rock (model) was crushed, and/or by sliding of the foundations in the same area. (This observation caused Serafim to suggest that a similar process might have been involved in the failure of Malpasset.) It is likely that sudden local weaknesses of the Eroc}c modulus are more dangerous than a uniformly low Erock value. Very thin arch dams may adapt themselves to substantial uniform rock displacements, but less to local
352 Rock mechanics and dam foundations
overstraining caused by sudden changes in the Eroc^ value along the peri- phery of the dam foundation. The central angle of the arch in plan is only 73° compared with 110° to 125° which is more usual. The shape was specially studied on models and adapted to the contour lines; the thrust from the arches of the dam is well directed towards the rock mass.
Miiller (Thirteenth Congress, Austrian Society for Rock Mechanics, Salzburg) emphasized that he considers the usual central angle value of about 120° often to be too wide, for the arch thrust pushes in a direction unfavour- able to the rock abutment. A smaller angle would suit some sites better and increase the stability of the rock abutment. This was done at Tang-e-Soleyman.
(4) Tests in the ISMES Laboratory, Bergamo {Italy). Workers at the laboratory have concentrated on further advanced research into arch dam design and the stability of rock abutments. In a few cases they were able to observe the failure of the rock abutment by overloading the model. The friction factor along faults observed in situ was as low as <f> = 25° (Ca Selva dam) or even </> = 13° (Gran Carevo dam). This low <f> value has been correctly represented on the dam abutment models, together with the stratification and faults of the rock mass as observed in situ.
Displacements of the rock on the downstream side of the dam were care- fully observed and progressive fissuration of the rock abutments recorded.
At first (Ca Selva dam) dislocation and surface cracking may occur in the upper elevations of the arch dam abutment. Surface brittleness is the greater danger because it may cause a general collapse. In such a case, failure is not always preceded by premonitory signs and it may occur abruptly. The rock excavations for dam foundations should always extend sufficiently deep into the abutment rock to avoid such failure (Fumagalli, 1966, 1967, 1968).
The second phase, leading to the ultimate collapse of the model, was determined by the sliding of the lower beds, which occurred only in the final stage of the testing. This plasto-viscous type of failure of the entire abutment rock mass usually presents less unforeseen contingencies. Moreover, its occurrence is presaged by large and slow deformation processes.
After completion of the tests of the Ca Selva dam, the rock buttress was dismantled and analysed layer by layer. It was thus possible to detect the cracking at different elevations. It was confirmed that, at the upper eleva- tions, the dislocation of the blocks was accompanied by a series of surface separation cracks. In a lower zone, a plastic compression process appeared to have occurred near the upstream concrete abutment of the arches, corre- sponding to a particularly high triaxial state of stresses existing in that area.
The influence of the faults was slight. Only two faults moved, which had beddings consistent with the cracks.
It is important to note that the Ca Selva test was carried out by applying the entire water pressure against the vertical grout curtain inside the rock.
It is worth while to compare these tests with the comments on the Malpasset
dam failure (chapter 13) when the failure was attributed to water pressure on deep rock layers. At Malpasset the rupture of the rock appeared to be by surface brittleness.
Tests on the Place Moulin arch-gravity dam carried out on a 1:70 scale three-dimensional model, showed that the first horizontal cracks to appear on the downstream face of the dam occurred at approximately 4 to 6 times the working load. The general collapse took place at a slightly higher loading (5-1 times the working load) and may be attributed to yielding of the rock along the entire dam foundation (see Malpasset dam case history, chapter 13) (Oberti; Fumagalli, 1967).
In addition some two-dimensional tests were carried out. Plane pumice- cement mortar models were made which corresponded to the most charac- teristic and heavily loaded arches, in order to investigate the static con- ditions of the rock in the downstream vicinity of the abutments, where the yieldings brought about the collapse of the three-dimensional model.
Four arches, located at four different levels, from top to bottom of the dams, were tested, first under the 'equivalent hydrostatic' load. On the completion of this test, the loading was increased until the model collapsed:
this was done so that the corresponding safety factor could be evaluated.
The hydrostatic loading was applied not only to the extrados of the arches, but also to the track of the waterproof grout curtain in the foundation rock.
The support given by the rock portion upstream of that curtain was neglected.
Strains and stresses in the 'rock' were measured in four directions, and principal stresses were calculated.