Nachdiplomkurs in angewandten Erdwissenschaften: Naturgefahren - Erdbebenrisiko ETH Zrich und Volkshochschule im Schwarzwald, 15. - 19.

Mai 2000

STATE OF PRACTICE IN EARTHQUAKE ANALYSIS OF DAMS

Dams, Embankment dams, Gravity dams, Arch dams, Earthquakes, Earthquake analysis, Earthquake safety G.R. Darbre
Federal Office for Water and Geology, Safety of Dams, P.O.Box 957, CH-2501 Bienne, Switzerland

1 INTRODUCTION 1.1 The role of the analyst

The complexity of the problems that the engineer faces during the design, the construction, the maintenance and the surveillance of a structure as well as the requirements regarding the reliability of the proposed solutions require a systematic and precise study of all the elements that can contribute to these problems and to their solution. The analyst, together with the designer, has to define the corresponding comprehension needs while taking due account of the characteristics of the structure and of the environment. This leads to a set of elements that are to be modeled by the analyst (loads, geometry, materials, goals of analysis). The results of the analytical and of the numerical calculations are interpreted by the analyst, together with the designer. The understanding sought follows. This process is depicted in Figure 1.

designer

environment

structure

comprehension needs

selection of elements to model

sollicitations

geometry

materials

goals of analysis

modeling

analyst

load cases

geometric model

numerical model

material laws

calculation

results

interpretation

designer

comprehension of behavior

structural consequences

Figure 1 - Analysis process.

-1-

The quality of an analysis depends largely on the ability of the analyst at grasping the physical phenomenon at play, at recognizing those that are important, at translating them into mathematical terms, at making a correct and appropriate analysis and, at the end of the analysis process, at interpreting the analytical and the numerical results in physical terms. 1.2 Remarks on the content of the paper

Only those aspects that relate to the calculation of the behavior of dams during earthquakes are presented here. It is assumed that the comprehension needs and the elements to be modeled have been adequately selected, that the state of static equilibrium is known and that the design (or safety evaluation) earthquake is defined. It is further assumed that the rupturing fault does not cross the close foundation of the dam. The indirect effects of the earthquake on the dam (e.g. following a rock slide in the reservoir) are further not considered. The methods and calculation models available are presented for embankment, gravity and arch dams in Sections 2 to 4. It is distinguished between the methods and models that are established in practice (established practice), those enjoying limited use but whose basis is recognized to be physically and mathematically sound (consolidated research), and those presently under development or whose physical or mathematical basis is still under discussion (open research). The paper builds on the experience gained by the author over the years. It becomes then difficult to point to specific references, and this is not done here. Still, several of the concepts presented stem from the work performed at the University of California at Berkeley. A selected literature list is also provided in Section 6. 1.3 Loading conditions

An earthquake analysis is never disconnected from a static analysis. The ultimate objective being to asses the performance of a dam during and after an earthquake, the static conditions that prevail at that time must be duly accounted for. An advanced earthquake analysis is ultimately of little use if the static state of equilibrium (stresses and strains) has not been determined with similar care.

hydrodynamic pressure (due to earthquakes)

water temperature

variation of water level

abutment movements ice pressure frost and thermic cycles solar radiations

hydrostatic pressure winter summer dead weight

creep and relaxation expansion : thermic, hydric, hydratation chemical ( alkali-aggregate reactions) summer air temperature

pore pressures Silt

winter snow and foundation temperature

winter

summer

underpressure earthquake

Figure 2 - Structural solicitations of an arch dam (adapted from P. Lger & R. Tinawi, A bibliography on structural analysis, behaviour, and safety of concrete dams, Report No. EPM/GCS1997-01, Department of Civil Engineering, Ecole Polytechnique, Montral, 1997).

-2-

The loading conditions usually encountered in the structural analysis of an arch dam are depicted in Figure 2 as an example. They include the dead weight of the dam concrete (the corresponding stresses and strains depend on the construction and grouting sequences, i.e. when and what portion of the arch action is mobilized), sediment (silt) pressure, water-related pressures (hydrostatic, uplift, internal, ice - all depending on the water level), temperature-induced solicitations (water, air, foundation, snow deposits, solar radiation), material-related ones (from frost and thermic cycles, creep and relaxation, expansion of various origins), abutment movements, and finally earthquake excitation including the associated hydrodynamic pressure. Quite often though, only a few of these solicitations will significantly affect the overall integrity of the dam. For the associated verifications, it is then sufficient and customary to consider only the following load combinations: 1. Construction stages 2. Usual load combinations: Gravity load of dam concrete + a. Hydrostatic loading for maximum operating water level b. Hydrostatic loading for extreme operating water levels + Associated temperature variations + Sediment loading c. Extreme temperature variations + Associated hydrostatic loading(s) + Sediment loading 3. Unusual load combination: Gravity load of dam concrete + Flood loading + Associated temperature variation + Sediment loading 4. Extreme load combination: Usual load combinations + Earthquake loading (maximum credible earthquake - MCE) Different performance requirements are set for the various load combinations.

2 EMBANKMENT DAMS

Figure 3 - Gschenen Dam (picture www.strom.ch).

-3-

2.1

Established practice

2.1.1 Two-dimensional stability analysis The earthquake analysis of an embankment dam such as Gschenen Dam, Figure 3, is typically performed for a two-dimensional model whose geometry and characteristics are those of the central section of the dam (Figure 4).
Dam 2D model

Figure 4 Two-dimensional model of an embankment dam.

The analysis consists of verifying the stability of individual portions of the dam (blocs), whereby the resultant of the inertia forces acting on a bloc is compared to the resisting force along the bloc-dam interface (Figure 5a) as follows: 1. Determine the static state of equilibrium that prevails in the dam (for example from a finite-element calculation, accounting for the construction stages and consolidation); 2. Determine the nonlinear dynamic mechanical characteristics of the dam materials from laboratory tests (modulus of elasticity, Poissons ratio and damping, all depending on the level and rate of loading); 3. Determine the earthquake motions and stresses in the dam by an equivalent, linear finite-element calculation (horizontal base excitation only). The mechanical properties obtained under point 2 are used to this end, introducing in every point a reference strain equal, say, to 2/3 of the maximum calculated strain (the latter is obtained by iteration until convergence is reached); 4. Submit the test specimen of dam materials to the total stresses (static + cyclic dynamic) calculated under point 3 in various parts of the dam. Determine the loss of shear resistance, the increase in pore pressure and the resulting strain; 5. Determine the zones of the dam susceptible of liquefying (loss of shear resistance due to the rise in pore pressure); 6. Perform one or several sliding bloc analyses whereby the zones susceptible of liquefying (as determined under point 5) are properly accounted for and the accelerations in the dam body (as calculated under point 3) are used. An incremental displacement of a bloc occurs when the weighted, horizontal acceleration becomes larger than the critical horizontal acceleration corresponding to sliding inception. Sliding stops as soon as the weighted acceleration returns below the critical acceleration (Figure 5b). If no sliding occurs, the permanent, remaining strains obtained under point 4 are transformed semi-empirically in dam permanent, remaining deformation.

-4-

Dam response

Sliding

Comparison between dynamic action and modified static resistance Base excitation Action > Resistance => Sliding

Zone susceptible of liquefying

a(t): Weighted acceleration of sliding bloc a(t) > acr => Sliding acr: Critical acceleration (sliding inception) for bloc

Figure 5 Bloc sliding analysis: a) Principle; b) Implementation.

The procedure above is typically used to analyze dams constructed of materials prone to pore pressure build-up and to the ensuing possible loss in shear resistance (saturated sands of low and medium density). For dams built of materials whose characteristics change only very little under cyclic loading (compacted cohesive clays, dry sands and very dense saturated sands), only steps 3 and 6 are usually performed. Further simplifications are introduced for smaller dams and preliminary studies, in particular replacing the finite-element calculation (step 3) by empirical expressions for the fundamental period of vibration and the associated mode shape (the peak acceleration response is then obtained from a response spectrum) as well as by introducing expressions relating the typical number of cycles of motions as a function of earthquake magnitude in step 6. 2.1.2 Major limitations This approach is based on clear physical concepts. Two main limitations are however recognized: 1. The acceleration used in the sliding calculation stems from the dynamic analysis of the monolithic dam. The modification of the oscillations of the main part of the dam due to sliding of a bloc and the modification of the response of the latter are not taken into account (interaction); 2. Only the safety against sliding is addressed directly. The calculated deformation is very approximate due to the use of an equivalent linear calculation. 2.2 Consolidated research

2.2.1 Hydraulic-fill dams It is recognized that hydraulic-fill dams are extremely vulnerable to earthquakes because they are prone to liquefaction (this is also true of dams constructed on loose sands).

-5-

2.2.1 Important aspects The initial static conditions have a great influence on the mechanical dynamic characteristics of the dam materials. This is to be remembered when applying the approach of Section 2.1. It is also important to recognize that the most unfavorable state of equilibrium may be the one prevailing after the earthquake because of the redistribution of the pore pressures. 2.2.2 Stresses and deformations Knowledge of the constraints in the dam body and of the permanent deformations are essential to the assessment of the earthquake behavior of an embankment dam that includes the issues of (Figure 6): crest settlement (with resulting loss of freeboard), longitudinal cracks (associated with large lateral oscillations), transverse cracks (associated with large longitudinal oscillations or transverse asynchrone excitations) and cracks within the dam body (piping initiation). All cracks can lead to erosion.
Settlement Crack

Cracks Transverse excitation Crack

Longitudinal excitation

Figure 6 Permanent deformation and cracks in an embankment dam: a) Settlements (loss of freeboard); b) Longitudinal cracks (associated with large lateral oscillations); c) Transverse cracks (associated with large longitudinal oscillations or transverse asynchrone excitation); d) Cracks in dam body (piping initiation).

2.2.3 Three-dimensional effects Three-dimensional effects may be important in narrow valleys (significant contribution of the lateral abutments to the dam stiffness) and in the presence of vertical, longitudinal and transverse asynchrone excitations. 2.2.4 Geomechanical formulation A geomechanical, nonlinear inelastic formulation combined with a finite-element modelisation permits to address some of the issues above. The availability of such formulations is however still largely limited to simple, single-phase models. 2.3 Open research 2.3.1 Constitutive laws The development of constitutive laws for a two-phase, fluid-solid material (fully and partially saturated) is a major research objective (plane-strain condition). The emphasis is put on the modeling of the nonlinear inelastic solid matrix. The coupling between the solid and the liquid phases is done based on an extension of Biots theory (mixture the-

-6-

ory). Rapid cyclic loading, loss of resistance and generation of pore overpressures are considered. 2.3.2 Wave propagation It is necessary to consider the variation of the excitation along the base of the dam when the significant apparent wave lengths of the earthquake excitation are of the same order of magnitude as the width of the dam base. This can be done by introducing a multiplesupport excitation in a finite-element model, provided an appropriate representation of the influence of the foundation is available (see foundation-dam interaction below). 2.3.3 Foundation-dam interaction Foundation-dam interaction describes the reciprocal influence exerted by the foundation and the dam during an earthquake. In its simplest formulation, this interaction is totally ignored (rigid-base formulation) or represented by a series of discrete springs and dashpots that represent the flexibility of the foundation and its capacity to dissipate energy. The infinite extent of the foundation is accounted for in the most rigorous representation, what leads to interaction coefficients that depend on the frequency content of the movements at the dam-foundation interface (for a visco-elastic foundation). The simplest formulation is adequate in most cases because the flexibility of the dam itself is larger than that of the foundation and because large amounts of energy is being dissipated in the dam body during large earthquakes. 2.3.4 Reservoir-dam interaction Similarly, dam-reservoir interaction describes the reciprocal influence exerted by the reservoir and the dam during an earthquake. Water compressibility (and associated radiation damping when the reservoir is of quasi-infinite extent in the upstream direction) and energy dissipation at the reservoir bottom and sides are considered in a rigorous formulation. This results in hydro-dynamic added-mass coefficients whose values depend on the frequency content of the motion of the upstream (U/S) face of the dam. A simplification consists of assuming that the water is incompressible. The value of the resulting entrained mass of water depends primarily on the slope of the U/S face. It reaches its maximum for a vertical face and is nil for an horizontal one. Because of the inclination of the U/S face of an embankment dam, the entrained mass is limited and the simplest formulation is usually adequate. 2.3.5 Three-dimensional effects The three-dimensional effects mentioned in Section 2.2.4 can be incorporated in a finiteelement model provided a constitutive law for a three-dimensional state of stresses is available and the option of multiple-support excitation is implemented. Alternatively, an analytical decomposition can be performed in the longitudinal direction; the finiteelement discretization then collapses to a two-dimensional one. Studies tend to demonstrate that the maximum acceleration values of a dam built in a narrow valley are larger than those at a long dam. This is because of three-dimensional effects (excitation stemming from the lateral abutments and increased contribution of the higher modes of vibration to the response). This tendency is however compensated

-7-

to an extent that is still unknown by the reduction in response associated with increased inelastic effects (more energy being dissipated internally and resonance being suppressed) and from the inhomogeneous excitation at the dam base (averaging of the earthquake excitation).

3 GRAVITY DAMS

Figure 7 - Gravity dam of Grande Dixence (Picture Les barrages et les digues, UCS-INFEL).

3.1

Established practice

3.1.1 Dam
incompressible fluid finiteelements > 3H H

entrained water rigid base

Figure 8 - Model of a gravity dam.

The analysis of a gravity dam as the one of Grande Dixence (Figure 7) is performed for a two-dimensional slice (usually the central or the highest section), Figure 8. The section is discretized in plane-strain finite-elements (no deformation perpendicular to the plane of the section). The finite-element grid is selected fine enough to permit an adequate reproduction of the stress distribution in the dam body. Distorted elements and elements with an aspect ratio (ratio of side lengths) departing too much from unity are avoided. A linear visco-elastic constitutive law is used in the calculations, with the modulus of elasticity of concrete adapted to the dynamic loading conditions (of the order of 25% to 30% higher than the static modulus). A damping equal to 4% to 5% of critical is introduced.

-8-

3.1.2 Foundation The foundation rock is considered to be rigid. This crude simplification stems from the fact that the static stiffness of a foundation resting on a half-plane of infinite extent is nil according to the theory of elasticity. The impossibility to treat such a situation simply in a calculation has lead to the assumption of rigid foundation. 3.1.3 Reservoir A water mass attached to the upstream face of the dam and entrained by the movements of the latter reproduces the influence of the reservoir on the dam. It is obtained analytically (incompressible fluid contained in a reservoir of infinite length according to the formulation of Westergaard) or by discretization (incompressible fluid finite elements comprising a reservoir length of at least three times the reservoir depth). This additional mass is obtained for an uniform movement of the upstream face of the dam in the stream direction. 3.1.4 Excitation The earthquake is introduced as an uniform motion at the base of the dam (generally as an accelerogram compatible with a design response spectrum). The vertical component of excitation is regularly omitted as the frequencies of the natural modes of vibration of the system that have a large participation factor under vertical excitation are generally higher than the energy-carrying frequencies of the excitation. 3.1.5 Time integration Integration over time is performed directly, progressing step by step while enforcing equilibrium at each step (explicit, implicit or mixed integration). The integration is performed for all degrees of freedom directly (direct integration) or for a limited number of generalized degrees of freedom obtained by modal decomposition. In the latter case, only those few modes of vibration are retained whose participation factor for the direction considered are not negligible as well as those whose natural frequencies are of the same order of magnitude as the predominant frequencies of the earthquake. An equivalent static correction can be introduced to account for the contribution of the higher modes not retained in the analysis. An analysis based on the response-spectrum method, in which the peak response of each mode of vibration is read directly from a diagram, is used solely in the analysis of smaller dams and for comparison purposes. The disadvantage of such an approach lies in its approximate prediction of the maximum dam response (according to the square root of the sum of the squares of the individual modal responses) and in its lack of information on the duration of large solicitations. 3.1.6 Global stability The dynamic stresses are combined with the static ones and compared to the dynamic ultimate strength of the concrete (higher than the static strength). The global stability is evaluated in an empirical way, interpreting the stresses obtained in terms of value, spatial distribution and duration. A simple assessment of the overall stability of the dam against sliding and against overturning is also performed.

-9-

The concrete strength on which these assessments are based is the apparent strength rather than the effective one. The former is obtained by linear extrapolation of the initial stress-strain relation up to the strain that corresponds to the effective strength (Figure 9). This is motivated by the fact that the process is deformation-driven rather than forcedriven.

stress

apparent strength effective strength

strain
Figure 9 - Stress-strain relation.

3.1.7 Modes of vibration Determination of the normal modes of vibration and of the natural frequencies is an essential step of any dynamic analysis. Their study allows an experienced analyst to recognize possible modelization errors, as well as gain an understanding of the way the structure behaves. 3.1.8 Major limitations The approach described above has three main limitations: 1. Nonlinear effects are not considered directly. The analysis thus applies only to cases where the behavior of the dam is essentially linear (no or few irreversible deformations or cracks). 2. The influence of the flexibility of the foundation rock on the dynamic behavior of the dam as well as the ability of the foundation to dissipate energy internally or by way of radiation are ignored (no dam-foundation interaction). 3. The water mass attached to the dam is obtained for an incompressible fluid assuming that the motion of the upstream face is uniform in the stream direction. The motion of the dam is however not solely in the stream direction. The entrained water mass associated with a vertical motion of the upstream face is nil in reality while an important mass is considered in the calculation (this limitation can be relaxed by introducing nodal mass matrices). Further, the vertical component of excitation generates an hydrodynamic pressure on the upstream face of the dam because of the compressibility of the water (that is ignored in the calculation). 3.2 Consolidated research

3.2.1 Dam-foundation interaction The flexibility of the foundation and its capacity to dissipate energy affect the dynamic behavior of the dam. It is possible to obtain the corresponding interaction coefficients (dynamic-stiffness matrix) by application of a boundary-element method. These coeffi-

- 10 -

cients are different for each harmonic component and their introduction implies that the analysis be performed in the frequency domain (see Section 3.4). Alternatively, the coefficients that prevail at a specific frequency (for example natural frequency of the dam on rigid foundation) can be used as discrete springs and dampers. Introducing this approximation permits to avoid a calculation in the frequency domain. 3.2.2 Reservoir-dam interaction The pressure exerted on the upstream face of the dam during an earthquake is affected by the compressibility of the water. In particular, the entrained mass of water is different for each harmonic component of motion and energy is dissipated by waves propagating upstream and by absorption at the reservoir bottom (also frequency dependent). This is summarized under the term of hydrodynamic mass. The variation of the hydrodynamic mass with frequency can affect the dynamic behavior of the dam as soon as the fundamental natural frequency of the dam fD is larger than 2/3 the fundamental natural frequency of the reservoir (fD > 2/3 fw), as well as when the frequencies of the energy-carrying waves of the earthquake are in a frequency range in which water compressibility affects significantly the hydrodynamic mass coefficients. It is then necessary to make an analysis in the frequency domain (see Section 3.4). fw is equal to cw/4H [Hz] for a prismatic basin of infinite length and depth H, cw being the pressure wave velocity in water equal to 1451 [m/sec]. It can be assumed that a reservoir extends to infinity when the time needed by a water pressure wave created at the dam-fluid interface to reach the back of the reservoir and come back (2L/cw, L being the length of the reservoir) is longer than the duration of strong response. 3.2.3 Contact dam-foundation Stability against base sliding can be evaluated with the help of a friction law of the Coulomb type. 3.2.4 Nonlinear inelastic behavior Cracks can occur during severe oscillations, in particular at the dam heel and in the vicinity of abrupt changes in geometry. Their consideration requires the application of nonlinear inelastic calculation models. In their simplest formulation, they are elastoideal plastic with a strength-based crack-initiation criterion. The development of such models is largely of the area of open research. 3.2.5 Three-dimensional effects Three-dimensional effects can have a non-negligible influence on the behavior of a dam during an asynchron transverse excitation (in particular at a dam with shear keys), during a longitudinal excitation and in the presence of a narrow canyon. Three-dimensional modeling permits to consider these effects.

- 11 -

3.3

Open research

3.3.1 Materials laws Emphasis is put on the nonlinear inelastic dynamic modeling of mass concrete and in particular on fracture mechanics. It is generally admitted that a crack initiates when the first principal tensile stress reaches the concrete strength. In the smeared crack models, the isotropic stress-strain relation valid before crack initiation is replaced by an orthotropic relation aligned on the direction of the cracks. In the discrete crack models, crack propagation is followed by application of a model based on strength of materials or on fracture mechanics. In a strength-of-materials based model, cracks propagate as the stress or the strain at the crack tip reaches a resistance value. A fracture-mechanicsbased model relies on the energy dissipated during the fracturing progress. Aspects such as penetration of water in the cracks, friction along the crack lips and impact occurring while the cracks close need to be treated. On the numerical side, perturbing high frequency oscillations can occur during opening and closing of a crack and the spatial discretization may loose its validity after a crack has appeared. The latter case is illustrated in Figure 10 where the zone on the right of the cracked region is modeled by a single column of finite elements (hatched zone); it is then not possible to adequately reproduce the true dynamic response of that region and thus crack development itself (modeling was appropriate before cracking inception).
Crack

Cracked region

Figure 10 - Inappropriate finite-element discretization of a gravity dam after cracking inception (after B. El-Aidi, Nonlinear earthquake response of concrete gravity dam systems, California Institute of Technology, Passadena 1988, Report No. EERL 88-02).

The constitutive models presently available must be used with great care, as they are often not sufficiently validated. Further, their use may lead to a modeling that is not objective (different results for different finite-element grids). 3.3.2 Criterion of dynamic instability A crack propagating across the section of a dam does not necessarily lead to an instability because of the dynamic nature of the phenomenon involved (a calculation program

- 12 -

must be able to handle this situation without generating a numerical instability). Criterions to assess such situations, including after the earthquake has occurred, are still largely missing. 3.3.3 Foundation The foundation influences the response of a dam and can be the weak element of the structural system. An important research effort is necessary in rock mechanics so as to obtain formulations that are consistent with those used in the modeling of the dam itself. 3.3.4 Infinite domains and nonlinear effects Frequency-dependent coefficients appear when considering infinite domains (dynamic stiffness of the foundation and hydrodynamic mass of the reservoir). This requires the use of a solution algorithm working in the frequency domain (see Section 3.4). At the same time, modeling of nonlinear effects requires calculating the behavior of the dam progressively with time (time-domain analysis). Analytical problems occur as soon as it is necessary to account simultaneously for the two effects because of the apparent incompatibility between the two types of analysis. Three approaches can be investigated in such situations: 1. Analysis with time-domain boundary elements; 2. Use of impulsive dynamic-interaction coefficients (either directly or by analogy with discrete systems); 3. Use of the hybrid frequency-time-domain method. 3.3.5 Three-dimensional effects The three-dimensional effects mentioned in Section 3.2.5 can be incorporated in a finiteelement formulation, provided that a constitutive law for a three-dimensional stress state be available and that the option of a multiple-support excitation be implemented. The size of the resulting system of equations can however be substantial. Alternatively, an analytical decomposition in the longitudinal, cross direction of the dam can be performed (the finite-element discretization remaining two-dimensional). The three-dimensional behavior of a dam located in a narrow canyon tends to lead to peak acceleration values that are higher than those calculated by assuming an ideal twodimensional behavior (excitation through lateral abutments, larger contribution of the higher modes of vibration). This tendency is however compensated to an extent that is still largely unknown by a reduction in response stemming from the consideration of inelastic effects (increase of internal energy dissipation and suppression of possible resonance) and from inhomogeneities of the excitation at the dam base (averaging of the excitation). 3.3.6 Cavitation It is assumed that cavitation is localized and of limited duration. It has thus not received much attention up to now.

- 13 -

3.3.7 Wave propagation It is necessary to consider the variation of the excitation at the base of the dam when the predominant apparent wave lengths of the earthquake are of the same order of magnitude as the width of the dam at the base. 3.4 Time- and frequency domains
Response r(t)

Analysis

Excitation a(t) Time domain

Frequency domain

Figure 11 Principle of analyses in time and frequency domains.

It is current practice to specify an earthquake by a set of accelerograms as represented schematically in Figure 11. The response of the dam is then obtained by following the action of the earthquake step by step (time-domain analysis). An other method consists of expressing the earthquake as the superposition of a large number of harmonic excitations of different frequencies. The behavior of the dam is then established for each frequency of excitation, the individual results being combined to obtain the total dam response (frequency-domain analysis). The final result is identical to that obtained by a time-domain analysis. The advantage of a frequency-domain analysis lies in the fact that the response of a linear visco-elastic structure to a harmonic excitation is also harmonic and of same frequency (only amplitudes and phase differ). This permits direct consideration of frequency-dependent parameters (dynamic stiffness or hydrodynamic mass). The frequency-domain analysis is however based on the principle of superposition and can thus be applied only to linear visco-elastic systems and not to nonlinear systems.

- 14 -

4 ARCH DAMS

Figure 12 - McKays arch dam.

4.1

Established practice

4.1.1 Dam The three-dimensional geometry of an arch dam as McKays Dam, Figure 12, is modeled by finite-elements of the thin- or thick-shell type or of the volume type (1 to 3 elements across the thickness), Figure 13. In the former case, advantage is taken of the fact that the state of stresses is essentially two-dimensional, the stresses perpendicular to the dams plane being close to zero. The finite-element grid is rectangular with vertical and horizontal sides (except at the abutments). This configuration is taken from that used in a static analysis in which it is necessary to consider the effect of the weight of the concrete that acts on the independent cantilevers before the vertical joints are grouted.

Figure 13 - Model of McKays Dam (thick-shell elements).

- 15 -

The shape of the elements deserves special attention (no distorted elements and elements whose ratio of the sides is close to unity). In the example of Figure 14, the stresses calculated in the hatched elements will not be representative of the true dam response.

Figure 14 - Inappropriate spatial discretization of an arch dam (hatched elements).

When the dam body is modeled by shell elements (5 degrees of freedom per node), it is necessary to introduce transition elements along the abutments so as to be able to connect the grid of the dam body to that of the foundation rock that consists of volume elements (6 degrees of freedom per node).

Figure 15 - Model of the foundation region at McKays Dam.

4.1.2 Foundation The foundation rock is modeled by finite elements of the volume type with linear-elastic characteristics. The size of the region modeled depends on the ratio of the modulus of elasticity that is representative of the rock Ef to that representative of the concrete Ec . Referring to Figure 15, Rf (distance to the outside of the region modeled) must satisfy: Rf >H for Ef / Ec=1, Rf > 1.5 H for Ef / Ec=0.5 and Rf > 2 for Ef / Ec=0.25 (the model of the foundation rock is fixed along its outside boundary and H is the maximum height of the dam). These values of Rf are obtained from parametric studies that indicate that an increase in Rf beyond these values modifies only marginally the modal characteristics calculated. The same values of Rf apply also to the static analysis, the deformation under hydrostatic pressure being calculated with sufficient accuracy when these values are used (no change in calculated deformations when augmenting the size of the region modeled).

- 16 -

Only the foundation flexibility is considered in the analysis. Introducing inertia would lead to waves remaining trapped within the region modeled, what does not occur in reality (transfer of the waves towards the outside of the region). It is presently impossible to model this radiation (energy transfer in the region extending to infinity) in a simple way. 4.1.3 Reservoir The considerations of Section 3.1.3 also apply. The analytical formulation for an incompressible fluid according to Westergaard is adapted to the double curvature of the upstream face of the dam. 4.1.4 Excitation The earthquake is introduced as an uniform motion of the outside boundary of the rock region. The three components (cross-stream, stream and vertical) of excitation are usually defined as statistically independent from one another and compatible with a design response spectrum. 4.1.5 Time integration Similar to Section 3.1.5. 4.1.6 Global stability Similar to Section 3.1.6. 4.1.7 Modes of vibration Determining the modes of vibration and the natural frequencies is also an important step of the analysis of arch dams, for the reasons indicated in Section 3.1.7. In dams with an essentially symmetric shape with respect to a vertical central axis, the mode shapes are classified according to their deformation and participation factors in symmetric, antisymmetric and vertical modes. 4.1.8 Major limitations The following major limitations are associated with the approach described above: 1. Nonlinear effects are not considered directly. The analysis thus applies only to cases where the behavior of the dam is essentially linear (no or few cracks or irreversible deformations); 2. Dam-foundation interaction is considered only approximately (disregard of the inertia and energy dissipation, introduction of an uniform seismic excitation); 3. The incompressible water mass attached to the dam is obtained under the assumption of a uniform motion of the upstream face (stream direction). Thus the limitations of Section 3.1.8 also apply (further enhanced by the presence of the cross-stream excitation).

- 17 -

4.2 Consolidated research 4.2.1 Dam-reservoir interaction The considerations made for gravity dams remain valid (Section 3.2.2). Here, the natural frequency of the first symmetric mode (rather than the first anti-symmetric one) has to be compared to that of the reservoir. A frequency-domain analysis permits consideration of water compressibility, energy absorption at the bottom and along the sides of the reservoir and the quasi-infinite extent of the reservoir in the stream direction. In the analysis, a region of finite dimension is generally introduced. It is adapted to the geometry of the upstream face of the dam and to the topography of the site (near zone, Figure 16). A prismatic region extending to infinity in the stream direction is attached to it.

Figure 16 - Model of the near region of McKays reservoir.

4.2.2 Cracks In general, large tensile stresses (that can induce cracking) tend to develop in the arch direction in the upper central part of the dam, in the vertical direction at the base and along the abutments. Lift and grout joints also represent potential weak zones. No generally validated model that can handle these aspects is however available. 4.2.3 Foundation-dam interaction Because of the three-dimensional nature of the foundation-dam system, different aspects of foundation-dam interaction can contribute in a non-negligible way to the dynamic response of an arch dam.
a) b)

c)

d)

Figure 17 - Dam-foundation interaction: a) Canyon effects; b) Kinematic interaction; c) Inertial interaction; d) Incoherent excitation.

- 18 -

First, the topography of the valley and the inertia and energy dissipation properties of the foundation rock leads to a non-uniform earthquake excitation, and this also in the absence of a dam (canyon effects, Figure 17a). Then, assuming that the dam has no mass (flexibility only), the motion along the abutment of the dam is affected by the static resistance that the dam offers to a deformation (kinematic interaction, Figure 17b). Canyon effects and kinematic interaction are more pronounced when the significant wave lengths of the excitation are short compared to a characteristic dimension of the site and of the dam. Finally, the motion along the abutment is modified by the inertial response of the dam (inertial interaction, Figure 17c). All these elements are considered approximately or not at all in the established methods of analysis due to the fact that only the flexibility of the foundation is considered in the calculation and that the excitation is introduced as uniform along the external boundary of the model of the foundation rock. 4.2.4 Stability of the rock foundation A way of assessing the stability of the foundation consists of performing a two-step analysis. In a first step, the foundation is modeled in a relatively coarse grid and a dynamic analysis performed as described in section 4.1. In a second step, the resulting forces along the abutments are introduced as external loads to the foundation rock that is modeled by a fine finite-element grid. 4.3 Open research

4.3.1 Nonlinear inelastic dynamic modeling of concrete The constitutive models of the concrete that are developed are identical to those described in section 3.3.1. A three dimensional state of stresses is considered. 4.3.2 Criterion of dynamic instability Similar to Section 3.3.2. 4.3.3 Foundation Similar to Section 3.3.3. 4.3.4 Infinite domain and nonlinear effects Similar to Section 3.3.4, the dynamic-stiffness matrix of the foundation and the hydrodynamic mass of the reservoir being however more difficult to obtain than in the twodimensional case. 4.3.5 Dam-foundation interaction The aspects described in Section 4.2.3 are being studied. The foundation rock is treated as homogeneous, linear visco-elastic and of infinite extent. Ways to account for the dynamic-stiffness coefficients are sought, either with the help of the boundary-element method or by introducing absorbing boundaries at the outside boundary of the rock region.

- 19 -

4.3.6 Incoherent excitations The seismic excitation is composed of various wave trains coming from different directions that arrive in an incoherent way (Figure 17d). Only preliminary studies on this topic have been performed in relation with dams. 4.3.7 Cavitation Similar to Section 3.3.6.

FINAL REMARKS 5.1 River dams

Older river dams such as Eglisau, Figure 18, are often made-up of a water retaining structure at river level and of a machinery deck located high above water and supported by piers. Having large masses (machinery) located on top of slender piers is detrimental to a good earthquake behavior. In newer river dams, machinery is usually integrated in the lower part of the structure.

Figure 18 - River dam of Eglisau (Picture NOK).

5.2 Fault crossing An aspect which has not been addressed in the paper is that of a fault crossing the dam site. A vivid example of what can happen in such a situation is brought by the damages caused at Shih Kang Dam during the Chi Chi Earthquake of September 21, 1999 in Taiwan (Figure 19).

- 20 -

Figure 19 - Bays 16 to 18 at Shih Kang Dam after Chi Chi Earthquake (picture and report R. Charlwood, Acres International).

Shih Kang Dam is a 18 bay gated spillway with a low concrete rollway. It appears that it is located directly over a branch of the fault that ruptured during the earthquake (the location of the main fault 300 meters away was known, but apparently not that of the branch). The fault rupture caused a differential offset of about 8 meters under bays 16 to 18 on the right hand side of the structure. The left part of the structure was raised by about 10 meters over a large area upstream and downstream. The right side was raised by about 2 meters, also over a large area. There was also a diagonal horizontal offset through the dam of about 7 meters. Peak ground motion accelerations of 56% g were measured close to the structure. Figure 19 shows the situation after the earthquake at Bays 16 to 18 where the fault ruptured the structure. The reservoir drained following the earthquake. 5.3 Basic principles The complexity of the assessment of the behavior of a dam during an earthquake makes the use of sophisticated mathematical and numerical models necessary. So as to be able to master the analysis and thus be able to give reliable answers to the questions posed, the analyst must avoid complicating unduly the model while avoiding not modeling or modeling in a too simplistic way important elements. It is also appropriate to perform simple preliminary analyses that permit to evaluate the importance of the individual effects before embarking in more sophisticated calculations. The analyst must further grasp the advantages and limitations of the methods of analysis at disposal. He must finally recognize that physical and engineering aspects are at the foreground of any analysis. It is also necessary to constantly verify the plausibility and correctness of the results obtained. Comparison analyses must be performed (a linear analysis before a nonlinear analysis, an analysis with a very limited number of generalized degrees of freedom, a response-spectrum analysis, ...), intermediate and partial results obtained and interpreted (mode shapes, resonant frequencies, response to individual components of excitation, ...) and comparisons with results from previous studies or from results published in specialized journals done.

- 21 -

5.4 Other aspects Modeling must be adapted to the physical reality (topography, geometry, loads, materials, ...) and not the problems to the models and calculation programs at disposal. This often implies using several models and calculation programs in the course of an analysis. The aspect of the design earthquake is not addressed here. The time-wise variation of the excitation, its frequency content, the duration of strong-motion and the peak accelerations are all factors that influence in a most important way the dynamic behavior of a dam. Also, no mention is made of the stochastic methods of analysis. They are conceptually very attractive because the seismic load is stochastic. They are applicable without much difficulty to linear analyses, but less to nonlinear analyses. The latter are at the center of todays modeling preoccupations in the dams community, together with interaction problems. 5.5 Observational needs

Gaining a better understanding of the dynamic behavior of dams requires not only the development of tools of analysis, but also the acquisition and interpretation of in-situ observations. A detailed seismic instrumentation of dams, ambient vibration tests and forced vibration tests are necessary to cover these observational needs, complemented by material tests in the lab and possibly in-situ. Quality interpretation of dam earthquake response is based on system-identification techniques. 5.6 Comparison with behavior of buildings

Damages to and collapses of buildings are observed in many earthquakes. They are, as a rule, due to the horizontal components of earthquake excitation. Dams are meant to resist horizontal loads (hydrostatic pressure) and are designed accordingly, whereas buildings are not. Because of this, dams are less prone to collapse during earthquakes. This fundamental difference in structural systems has to be kept in mind when transferring to dams the knowledge gained from the performance of buildings during earthquakes.

SELECTED LITERATURE

A.K. Chopra, Earthquake resistant design of concrete dams, Journal of the Structural Division, ASCE, 1978, Vol. 104, No. 6, 953-971. A.K. Chopra, Earthquake response analysis of concrete dams, in Advanced dam engineering for design, construction and rehabilitation, Ed. R. Jansen, Van Nostrand Reinhold, New York, 1988, 416-465. G.R. Darbre, Strong-motion arrays at four large Swiss dams, The International Journal on Hydropower and Dams, Nov. 1995, 65-70. G.R. Darbre, Strong-motion instrumentation of dams, Earthquake Engineering and Structural Dynamics, 1995, Vol. 245, 1101-1111. G.R. Darbre, Phenomenological two-parameter model for dynamic dam-reservoir interaction, Journal of Earthquake Engineering, 1998, Vol. 2, No. 4, 513-524. G.R. Darbre, State of practice in earthquake analysis of concrete dams, 11th european conf. earthquake eng., Paris, 1998. G. Gazetas, Seismic response of earth dams: Some recent developments, Soil Dynamics and earthquake Engineering, 1987, Vol. 6, No. 1, 2-47. G. Gazetas & P. Dakoulas, Seismic analysis and design of rockfill dams: State-of-the-art, Soil Dynamics and earthquake Engineering, 1992, Vol. 11 No. 1, 27-61.

- 22 -

F. Ghrib, P. Lger, R. Tinawi, R. Lupien & M. Veilleux, Seismic safety evaluation of gravity dams, Hydropower & Dams, 1997, Vol. 4, No. 2, 1997, 126-138. J.F. Hall, The dynamic and earthquake behaviour of concrete dams: review of experimental behaviour and observational evidence, Soil Dynamics and Earthquake Engineering, 1988, Vol. 7, No. 2, 57-121. N.M. Newmark, Effects of earthquakes on dams and embankments, Gotechnique, 1965, Vol. 15, No. 2, 139-160. H.B. Seed, Considerations in the earthquake-resistant design of earth and rockfill dams, Gotechnique, 1979, Vol. 29, No. 3, 215-263. H.M. Westergaard, Water pressures on dams during earthquakes, Transactions ASCE, 1933, Vol. 59, No. 8, Part 3, 418-472.

- 23 -