EFFECTIVE MODELING OF DAM-RESERVOIR INTERACTION EFFECTS USING ACOUSTIC FINITE ELEMENTS Matthew Muto, Ph.D.

1 Nicolas von Gersdorff, P.E.2 Zee Duron, Ph.D.3 Mike Knarr, P.E., S.E.4 ABSTRACT Southern California Edison (SCE) is currently evaluating the seismic stability of two large concrete dams. Extensive finite-element analyses of these two dams are being performed to evaluate their response to large earthquakes. The earthquake analysis of these dams requires an accurate representation of dam-reservoir interactions to capture appropriate hydrodynamic loading and radiation damping effects. Traditional modeling considerations for the reservoir include the use of lumped added masses or the use of finite fluid elements to represent a portion of the reservoir domain that attaches directly to the dam model. However, in a seismic analysis the elements can become excessively distorted, increasing the computational cost and potentially affecting the accuracy of the results. SCE has recently completed a series of evaluations that suggest modeling the reservoir domain as acoustic medium using finite elements that track only pressure, not deformation, can provide reliable representations of damreservoir interactions. These evaluations included comparisons between computed hydrodynamic pressure responses from numerical models with acoustic reservoir elements and measured hydrodynamic pressures acquired during forced vibration tests on a large concrete dam. Additional and enhanced confidence in the use of acoustic elements to model dam-reservoir interactions is gained by demonstrating that these elements can be used to obtain satisfactory comparisons in problems with known closedform solutions. The results presented in this paper show that the use of acoustic finite elements avoids some of the potential problems with traditional reservoir modeling techniques while providing comparable accuracy. INTRODUCTION Validation of numerical models using observed data can be a valuable aid in determining the suitability of a given model for predicting the response of a structure. A growing database of measured hydrodynamic pressure responses acquired during forced vibration testing of large concrete dams provides observed evidence of dam-reservoir interaction effects that can be used for this purpose.
1 2

Southern California Edison, 300 N. Lone Hill Ave., San Dimas, CA 91773. [email protected] Southern California Edison, 300 N. Lone Hill Ave., San Dimas, CA 91773. [email protected] 3 Harvey Mudd College, 1200 Platt Blvd, Claremont, CA 91711, [email protected] 4 Southern California Edison, 300 N. Lone Hill Ave., San Dimas, CA 91773, [email protected]

Modeling of Dam Reservoir Interaction Effects

1161

These interactions are often modeled with lumped added mass on the upstream surface of the dam (Westegaard 1933), or by using solid elements with fluid material properties in the reservoir domain. While these methods have been shown to capture overall trends in pressure response and interaction effects, improved comparisons have been demonstrated through the use of acoustic elements. The ABAQUS acoustic finite elements are used to model problems ranging from sound radiation in automobile engineers (Simulia 2007) to the effects of underwater explosions (Simulia 2007b). Schuelen et al. (2010) showed that pressures predicted using acoustic finite elements in the ABAQUS program compared well with hydrophone measurements taken during a forced vibration test at SCEs Big Creek No. 7 Dam. Acoustic finite elements have subsequently been used by SCE in modeling several concrete dams VALIDATION PROBLEM Dr. George Housner examined the problem of obtaining closed-form expressions for the dynamic pressures exerted on a container by impounded fluid (Housner 1954). The following sample problem and Housners analytical solution were used to evaluate the appropriate application of acoustic elements for modeling dam-reservoir interactions. It is one of a class of problems in which dynamic fluid pressures are computed for containers under earthquake loading. The particular problem is of an accelerating tank into a water domain, and a sketch of the configuration is shown in Figure 1.

Figure 1. Diagram of the simple example problem of an accelerating tank of water. From Housner (1954). Analytical solution Housner developed closed-form solutions by using an approximation of the fluid flow, which is an approach he considered to be analogous to the Rayleigh-Ritz method used to solve elastic deformation problems. This method was applied to several types of containers and the results are comparable to the solutions of Westegaard (1933) and

1162

Innovative Dam and Levee Design and Construction

Graham and Rodriguez (1952), with the advantage of having a mathematically convenient solution form. Consider a tank of length 2l and height h, filled with a fluid moving with horizontal velocity u and vertical velocity v, as illustrated in Figure 1. The tank is rigid and moves with a specified horizontal velocity u0. Housner showed that the pressure pwall on the wall of the tank is given by p wall y 1 y 2 l 0 tanh 3 = hu h 2 h h (1)

0 is the horizontal acceleration of the tank. If we consider the dam-reservoir where u system as a very long tank where l is much greater than h, this simplifies to

y 1 y 2 0 p wall = hu h 2h Numerical Models

(2)

We now compare this solution to the results given by two numerical models. In the first model, a rectangular section of a reservoir measuring 150 feet wide, 200 feet long and 50 ft deep is modeled with a mesh of 8-node acoustic brick elements, as shown in Figure 2(a). As described in the ABAQUS manual (Simulia 2010), these elements model an acoustic medium as an elastic material in which stress is purely hydrostatic (no shear stresses) and the pressure is proportional to volumetric strain. Acoustic finite elements only track pressures, nodal displacements are not calculated.

(a) (b) Figure 2. Finite-element meshes used in the example problem to model (a) the rectangular reservoir and (b) the rigid dam-foundation system.

Modeling of Dam Reservoir Interaction Effects

1163

In the sample problem, the dam and the sides and floor of the reservoir are modeled with rigid shell elements, shown in Figure 2(b). At the upstream end of the reservoir, a nonreflecting boundary condition is applied. Tie constraints are applied between the reservoir and the dam-foundation model. A 1 Hz sinusoidal ground motion is applied to the foundation in the stream direction for 6 seconds, with a peak acceleration of approximately 1.2 g. At the top of the reservoir, a free surface boundary condition is applied to the nodes of the acoustic elements by constraining the dynamic pressures to be zero. This model is very similar to how acoustic elements are utilized by SCE in the analysis of its dams. The only significant differences are the complexity of the geometry and the use of deformable solid elements in modeling the dam and foundation. For the second model, deformable solid elements are used. While these elements are often used in modeling reservoirs, in a seismic analysis the elements can become excessively distorted, increasing the computational cost and potentially affecting the accuracy of the results. For comparison purposes, a finite-element model of the sample problem using these solid elements was also analyzed. The reservoir mesh geometry is identical to the acoustic finite-element model shown in Figure 2(a). Rather than model the dam and foundation separately, the sinusoidal input motion is specified as a boundary condition on the nodes on the sides and downstream end of the reservoir. At the upstream end, a layer of infinite solid elements provides a nonreflecting boundary. Gravity loading is also included in the solid model. An initial stress state is applied such that the reservoir is in equilibrium under gravity load at the start of the analysis. Material properties for water are based on the values used by the Bureau of Reclamation in the nonlinear analysis of their concrete dams (Mills-Bria et al. 2006). The density for both types of finite elements is 62.4 pcf. For the acoustic finite elements, the bulk modulus is 4.49107 psf. For the deformable solid elements, the modulus of elasticity is 2.73104 psf and Poissons ratio is 0.4999. There is no material damping; the only mechanism for energy dissipation in either finite element model is the nonreflecting boundary at the upstream end of the reservoir. Results Dynamic pressures are calculated directly for the acoustic elements model. Figure 3 shows contours of dynamic pressure for the acoustic finite element on a section cut through the middle of the model. For the solid elements, dynamic pressure is obtained by dividing the reaction force on the nodes at the downstream end of the reservoir by the tributary area of each node, then subtracting the hydrostatic pressure resulting from the weight of the overlying water.

1164

Innovative Dam and Levee Design and Construction

Figure 3. Cut-away view of the example problem model, showing contours of pressure for the acoustic finite elements. Figure 4 shows the first four seconds of the time histories of the dynamic pressures for the acoustic and solid element models for a node on the center of the downstream face of the reservoir compared with the dynamic pressure calculated using Equation 2. For the acoustic elements, the dynamic pressures are relatively close to the closed-form solution (within 1 psi). For the solid elements, we see transient vibrations (which are to be expected in any type of spring-mass system) at the start of the record, which decay after 1-2 cycles, and somewhat larger errors compared to the acoustic elements. Figure 5 shows the maximum dynamic pressure as a function of depth for the analytical solution and the two finite-element models. The difference in the maximum pressure between each finite-element model and the analytical solution is shown in Figure 4. The acoustic finite element models more closely reproduce the behavior predicted by Housner in this sample problem. Quantitatively, the normalized root-mean-squared error between the steady-state pressure using acoustic elements and the closed-form solution is 5%, while the error of the pressures computed with solid elements is about 13%. Note that steady-state was defined as the response of the finite-element models between 3 and 6 seconds to allow for sufficient decay of the transient response of the solid elements.

Modeling of Dam Reservoir Interaction Effects

1165

Figure 4. Comparison of the first four seconds of the time histories of dynamic pressure computed using Housners closed-form solution for a oscillating tank with values computed by the ABAQUS program using acoustic and solid finite elements.

Figure 5. Maximum dynamic pressure as a function of depth for the closed-form solution and the ABAQUS finite element results.

1166

Innovative Dam and Levee Design and Construction

CONCLUSIONS A comparison of measured and computed hydrodynamic pressures for a large concrete dam indicated good agreement was achieved using acoustic elements. To gain better understanding of how these elements can be used, and to provide enhanced confidence in their application to the earthquake analysis of dams, comparisons against the closed-form solution of a sample problem were also pursued. These comparisons verified the agreement observed from the previous studies, and improved agreement with the closedform solution was obtained using acoustic elements as opposed to fluid elements. REFERENCES Graham, W., and Rodriguez, A.M., 1952. Char acteristics of fuel m otion which affect airplane dynamics. Journal of Applied Mechanics, 19. Housner, G.W., 1954. Earthquake pressures on fluid containers. Report EERL-19543, Earthquake Engineering Re search Laboratory, Californi a Institute of Technology, Pasadena, CA. http://resolver.caltech.edu/CaltechEERL:1954.EERL.1954.003 Mills-Bria, B., Nuss, L., OConnel, D., Harris, D., 2006. State-of-practice for the nonlinear analysis of concrete dams at the Bureau of Reclamation. Technical report, Bureau of Reclamation, Denver, CO. Schulen, F., von Gersdorff, N., Duron, Z., Knarr, M., 2010. Numerical model validation for large concrete gravity dams, 2010 United States Society on Dams Conference. Simulia Corp., 2007. Sound radiation analysis of automobile engine covers. Abaqus Technology Brief TB-06-COVER-2. Providence, RI. http://www.simulia.com/solutions/tb_pdf/TB-06-COVER-2.pdf Simulia Corp., 2007b. Shock resonse and acoustic radiation analysis. Abaqus Technology Brief TB-04-SUB-2. Providence, RI. http://www.simulia.com/solutions/tb_pdf/TB-04-SUB-2.pdf Simulia Corp., 2010. ABAQUS manual. Version 6.10. Providence, RI. Westergaard, M., 1933. Water pressures on dams during earthquakes. Transactions of the American Society of Civil Engineering, 98.

Modeling of Dam Reservoir Interaction Effects

1167