Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Executive Summary Concerns about the seismic safety of concrete dams have been growing during recent years, partly because the population at risk in locations downstream of major dams continues to expand and also because it is increasingly evi- dent that the seismic design concepts in use at the time most existing dams were built were inadequate. To respond to these concerns, the Panel on Earthquake Engineering for Concrete Dams was appointed to serve under the NBC Committee on Earthquake Engineering. The mandate of the panel was to evaluate the present status of knowledge about the earthquake per- formance of concrete dams, including procedures for investigating the seis- mic safety of such structures, and to summarize its findings in a report. The report was intended specifically to inform research workers about the state of the art of earthquake analysis of concrete dams and to identify subject areas where additional knowledge is needed. Questions about the safety of concrete dams were first brought into focus in this country by the failure of St. Francis Dam in California in 1928, which caused extensive property damage and the loss of over 400 lives. This disaster led to the formation of an agency of the State of California that would be responsible for dam safety. Now known as the Division of Safety of Dams of the California Department of Water Resources, this agency has jurisdiction over nearly 1,200 dams, and its operation during the past 60 years has served as a model for similar agencies in many other states and several other countries. The hazard posed by large dams has been demonstrated since 1928 by the failure of many dams of all types and in many parts of the world. However, no failure of a concrete dam has resulted from earthquake excita- tion; in fact the only complete collapses of concrete dams have been due to failures in the foundation rock supporting the dams. On the other hand, two

2 significant instances of earthquake damage to concrete dams occurred in the 1960s: Hsinfengkiang in China and Koyna in India. The damage was severe enough in both cases to require major repairs and strengthening, but the reservoirs were not released, so there was no flooding damage. This excellent safety record, however, is not sufficient cause for complacency about the seismic safety of concrete dams, because no such dam has yet been subjected to maximum conceivable earthquake shaking while retaining a maximum reservoir. For this reason it is essential that all existing con- crete dams in seismic regions, as well as new dams planned for such regions, be checked to determine that they will perform satisfactorily during the greatest earthquake shaking to which they might be subjected. Major factors that must be considered in verifying the seismic safety of existing or proposed concrete dams include definition of the expected seis- mic excitation and evaluation of the response of the structure to this input. Usually, the structural response is first calculated assuming that the dam is a linear system in which the displacements are directly proportional to the input excitation. To establish the ultimate resisting capacity, however, damage mechanisms and the resulting nonlinearities must also be considered. Observa- tional evidence provides the best indication of the true performance of a dam in the nonlinear response range; hence, one important part of earth- quake engineering for concrete dams is the evaluation of data from dynamic tests real earthquake response information as well as laboratory and field test data. Finally, to ensure that proposed or existing concrete dams pro- vide adequate safety, suitable performance criteria must be established against which the predicted earthquake performance of a dam can be judged. Dis- cussions of all of these topics are presented in Chapters 2 through 6 of this report; a brief overview of each is contained in the following paragraphs. EARTHQUAKE INPUT To evaluate the expected earthquake performance of a concrete dam, it first is necessary to establish the intensity and frequency characteristics of the earthquake motions to which the dam might be subjected. Of course, these properties depend directly on the seismicity of the dam site that is, upon the magnitude and epicenter location of the expected earthquake as well as its probable recurrence interval; however, these topics are not considered in this report. Instead, it is assumed here that the ground motion to be expected at the site is known, in the form of a strong-motion seismograph record, for example; thus, this report is concerned only with the expected performance of a specified dam when it is subjected to these given motions. The first factor to be considered in a response analysis is how the prescribed earthquake motions are applied to the dam. In the early days of earthquake engineering it was always assumed that the structure was supported on a

3 rigid base and that the earthquake input was applied as specified motions of this base. For lightweight relatively small structures supported on hard foundation rock, this input assumption was acceptable. However, for massive stiff concrete dam structures that are supported by very broad foundation rock surfaces, it is not reasonable to assume that the base support system is rigid. Clearly this assumption would not be consistent with the fact that an earthquake is actually a vibration wave being propagated through the earth's crustal structure. Thus, one of the principal questions to be answered with regard to the earthquake input is how to express the excitation in the context r · ~ or seismic wave propagation. The typical strong-motion earthquake record portrays three components of the acceleration history observed at a single point, usually at the ground surface in a free-field location (i.e., at a point where the motion is free of the vibratory influences of adjacent structures). Analytical procedures have been developed that account for the differences between the specified free- field earthquake motions and the motions of the points at which the structure is supported; however, a major difficulty in the analysis of dam response is the fact that there may be significant variations between the motions to be expected in the free-field at widely spaced points on the dam-foundation interface. Many theoretical studies have been made of the spatial variation of seismic waves propagated to highly idealized dam-canyon interfaces, but there is almost no information on the seismic motions that have actually occurred at the boundaries of a real dam. Hence, the planning and installa- tion of strong-motion accelerograph networks at the canyon sites of actual concrete dams and also at free-field sites in canyons where concrete dams might be built are among the urgent research needs in the general area of defining earthquake input for concrete dams. LINEAR RESPONSE ANALYSIS In the earliest efforts to represent the effects of earthquakes on dams, the dams were considered to be rigid systems supported on a rigid foundation. Thus, the earthquake forces acting in the structure could be expressed as the product of the earthquake acceleration and the mass of the corresponding part of the structure. More recently the elastic properties as well as the mass of the dam materials have been incorporated into the mathematical models, and as a result the vibration properties of dams (the mode shapes and frequencies) have been recognized to have a controlling influence on earthquake response behavior. Because of the ability of the finite element method to define mathemati- cal models with arbitrary geometries and variations of material properties, this method is usually adopted in formulating a mathematical model of a concrete dam. In this sense the seismic response analysis of a concrete dam

4 may be considered to be similar to any other structural dynamics analysis. However, the concrete dam analysis is greatly complicated by the fact that the structure interacts with its environment during its dynamic response- with the water retained in the reservoir and with the deformable foundation rock that supports it. These interaction mechanisms may be included in the model in a crude way by combining finite element meshes representing a limited extent of the reservoir water and foundation rock together with the model of the concrete dam. However, it is recognized that dynamic pressure waves will be generated by the earthquake in the reservoir water, and similarly there will be stress waves in the foundation rock. The effect of these waves acting in the actual unbounded media will be to transmit energy away from the dam. Thus, a significant energy loss mechanism is not represented by the bounded finite element models of reservoir water and of foundation rock, and the development of mathematical models that properly account for these effects has become a major objective in research on the seismic behavior of concrete dams. Gravity dams and their environment often can be idealized as simple two-dimensional systems, and for this reason interaction response analysis procedures were first developed for gravity dams; the most powerful analytical techniques employed a substructure formulation and a frequency-domain response-analysis procedure. More recently a similar approach has been applied to the analysis of three-dimensional arch dam-reservoir systems in which the arch dam is modeled as a finite element mesh and the reservoir is approximated as a limited form of continuum. In these analyses, however, the foundation rock is still modeled as a bounded finite element mesh, and the elimination of this rock boundary constraint is an important current research objective. Another problem of present concern is the evaluation of the effective stiffness of typical reservoir bottom boundaries, which are generally covered by a layer of silt having indeterminate thickness and mechanical properties. It has been found by analytical experiment that a significant part of the vibratory earthquake response energy of a dam may be transmitted by reservoir pressure waves into the bottom silt layer, but the amount of such energy loss depends directly on the absorption coefficient of the reservoir bottom, and no measurements of this property have been made in actual reservoirs. A field measurement program to obtain this information is urgently needed, because the bottom absorption directly affects the earthquake response stresses that are calculated in the dam. NONLINEAR RESPONSE ANALYSIS Although linear response studies have provided great insight into the earthquake performance of concrete dams, it is evident that a rigorous esti-

s mate of the seismic safety of a dam can be obtained only by a nonlinear analysis if a significant amount of earthquake damage is expected. In fact, minor local damage may have little effect on the global stiffness of a dam, and in such cases it is possible to make a reasonable estimate of the ex- pected degree of damage by proper interpretation of the results of a linear response analysis. But for cases of severe damage, in which a collapse condition may be approached, the dynamic behavior is drastically changed from the linear response mechanism, and the true nonlinear performance must be incorporated into the analysis procedure if a valid estimate of the damage is to be made. A nonlinear response analysis requires considerably greater computational effort than a linear analysis, because a much more detailed description of the performance is required to express the nonlinear response mechanisms. Special techniques are needed to carry out such calculations, and continued efforts must be directed toward improving the efficiency of nonlinear analysis procedures for concrete dams. However, the greatest impediment to effec- tive nonlinear analysis at present is not the computational procedures; it is the lack of knowledge about the nonlinear properties of the mass concrete typically used in dams. It is well known that the mechanical properties of concrete vary with time, temperature, and moisture content; additionally, both the strength and the stiffness vary with the rate at which loads are applied. Thus, the properties associated with earthquake damage may be quite different from those measured by typical slow-speed laboratory tests, and it is essential that major programs of dynamic testing be carried out to determine material properties suitable for use in nonlinear seismic safety evaluations. When efficient analytical procedures have been developed and when the nonlinear mass concrete material behavior can be modeled effectively, it will be necessary to perform extensive numerical parameter studies of con- crete dams in order to understand the factors that control the seismic safety of different types of dam systems. In addition, properly modeled shaking- table tests will be needed to verify the effectiveness of the safety evaluation procedures that have been developed. OBSERVATIONAL EVIDENCE All analytical predictions of earthquake performance of concrete dams are based on numerous assumptions, each of which has a limited range of validity. Even the design of shaking-table experiments employing physical models of concrete dams requires the introduction of limiting assumptions, with regard to both the nature of the simulated earthquake motions to be applied and the properties of the model structure and of the foundation rock in which it is placed. For these reasons the best evidence by far about the

6 earthquake behavior of concrete dams is that obtained from real dams that have been subjected to real earthquakes. The number of instances when severe earthquake motions have acted on major concrete dams is quite small, and only in two cases was significant damage caused by the earthquakes: Koyna Dam in India and Hsinfengkiang Dam in China. In neither of these cases did the damage cause release of the reservoir, so no disaster resulted in the downstream region; however, in both cases major repairs and strengthening were done to increase the secu- rity against future earthquakes. Other concrete dams have been subjected to significant earthquake motions and have suffered only minor or no damage, so the seismic safety record of concrete dams is good; however, it must be recognized that no typical major concrete dam has been subjected to maxi- mum credible earthquake excitation, so there is no cause for complacency about the seismic safety of concrete dams. Because real earthquakes provide the best test of earthquake performance, it is essential that many more existing concrete dams be provided with adequate seismic instrumentation so that quantitative evidence can be obtained to be used in verifying safety evaluation procedures. In addition, free-field instrumentation should be installed at potential concrete dam sites in regions of high seismicity so that better information can be obtained about the seismic excitation to which typical dams may be subjected. The lack of adequate earthquake input data is probably the greatest present impediment to progress toward improving the seismic safety evaluation procedures for concrete dams. However, even if the input is well established, the degree of damage to be expected from severe earthquake excitation of a concrete dam cannot be predicted with certainty. Part of the uncertainty lies in the nonlinear physi- cal properties of the concrete, and this information can be obtained only from comprehensive, well-planned material test programs. In addition, many assumptions are made in formulating nonlinear analysis procedures, and these must be validated by experiments that simulate both the excitation and the structural response. Field vibration experiments, using either ambient or forced input excitation, are very useful in verifying the mathematical model of the dam system at low-response amplitudes, but shaking tables provide the only feasible means of testing the system well into the damage range. However, it must be recognized that the difficulty of achieving model similitude increases as the model dimensions are reduced; in particu- lar, for the most commonly used scaling laws the model frequencies must be increased in proportion to the length scale; thus, a 1/100 scale model must be tested at frequencies 100 times greater than the actual earthquake motions. Very few shaking tables capable of testing moderate-size models can apply simulated earthquake motions that meet this requirement. The world's best facility for testing concrete dam models is operated by the

7 Institute of Water Conservancy and Hydroelectric Power Research in Beijing, and the initiation of a cooperative shaking-table research project with this institute would provide an excellent means for U.S. researchers to extend their experience with the nonlinear behavior of concrete dams. SEISMIC PERFORMANCE CRITERIA The design of earthquake-resistant concrete dams must be based on ap- propriate performance criteriacriteria that reflect both the desired level of safety and the nature of the design and analysis procedures. Before computers were used in design calculations, the usual criteria were merely checks on the static stability of the dam; thus, they did not represent the true dynamic earthquake response behavior and did not provide any guarantee against seismic damage or collapse. Now, with modern computer analysis procedures, it is possible to predict with some precision the linear elastic response that will result from a specified earthquake input; however, because of uncertainties in the characteristics of the expected earthquake, in the way the excitation is applied to the dam, in the dynamic strength properties of the concrete, and in the nature of the ultimate failure mechanism, it still is not possible to prescribe exactly the performance criteria to be used in the design of concrete dams. For this reason the criteria presently used in the seismic safety evaluation of existing concrete dams or in the design of new dams generally are simply stress checks in which the predicted elastic stress is compared with the expected concrete strength. The seismic inputs specified in most criteria are the design basis earth- quake (DBE) and the maximum credible earthquake (MCE). The DBE is defined as the greatest earthquake excitation expected to occur at least once during the life of the dam (possibly 100 years). It should cause no signifi- cant damage to the dam; thus, in the response to such an input it is appropriate to limit the maximum concrete stress (static plus dynamic) to the strength of the concrete applied with a factor of safety. The MCE is the greatest earthquake excitation that could ever occur at the dam site; it has a very low probability of occurrence, and therefore significant damage from this earthquake would be acceptable; however, the dam must not rupture and thus threaten life and property downstream. It is evident that the design criterion for response to the MCE must involve more than comparison of the predicted peak stress with the concrete strength. However, some judgment regarding the possibility of collapse can be obtained by studying the extent of the regions within the dam in which possible crack-inducing stresses are expected, especially how much of the structure may be subjected to such stresses concurrently. In addition, the possibility of collapse may be related to the number of times and the total duration that the cracking stress threshold is exceeded.

8 Eventually, it is hoped that nonlinear analysis capabilities will be im- proved to the point where the performance criteria can be stated in terms of acceptable amounts of nonlinear displacement; both the types of displacements and their location in the dam will influence the acceptable amplitude of displacement. However, significant advances will have to be made in the understanding of possible earthquake collapse mechanisms as well as in nonlinear analysis capabilities for concrete dams before such criteria may be proposed. Thus, it is evident that both analytical and experimental research on the nonlinear behavior of concrete dams continues to be of paramount importance.