Vous êtes sur la page 1sur 16

NUMERICAL MODEL VALIDATION FOR LARGE CONCRETE GRAVITY DAMS Florian Scheulen1 Nicolas Von Gersdorff2 Ziyad Duron,3

Ph.D. Mike Knarr,4 P.E., S.E. ABSTRACT The interaction effects between a large concrete gravity dam, its foundation, and impounded reservoir have been investigated using 2-D and 3-D finite element representations validated against field measurements. Field tests conducted on the dam provided acceleration profiles across the dam crest and along the dam-foundation interface, and were used to assess material properties in the dam and in the foundation. Field measurements of hydrodynamic pressures acquired along the upstream dam face were used to evaluate the relative accuracy of various techniques for representing the reservoir water in the numerical model of the dam-foundation-reservoir system. Techniques for representing the reservoir water included the Westergaard added mass approach, the RSVR2 approach, fluid elements modeled as plane strain elements with the properties of water and acoustic elements. Comparisons of measured and computed acceleration and hydrodynamic pressure frequency responses were obtained that illustrate the advantages and limitations associated with each modeling approach. The paper provides relevant details associated with the field tests on the dam, presents an overview of the numerical models, and discusses the assumptions used in developing the reservoir representations. Frequency responses are compared to highlight model performance. These results will be used to validate numerical models of the dam-foundation-reservoir system in support of an ongoing risk-based performance evaluation for this dam. INTRODUCTION An accurate finite element model is crucial for most risk-based dam analyses. This paper demonstrates how field data can be used to validate the accuracy of a model. Various methods of modeling the reservoir were applied to 2-D and 3-D models of a large concrete gravity dam, Big Creek Dam No. 7, and compared to hydrodynamic pressure values acquired using hydrophones. Acceleration profiles along the crest and at the damfoundation interface were used to enhance the models ability to accurately reproduce observed behavior, and the models ability to predict observed hydrodynamic pressure response was ultimately used as the basis for validating model response behavior. An
De Pietro Research Fellow, Department of Engineering, Harvey Mudd College, Claremont, CA 91711, Florian_Scheulen@hmc.edu 2 Structural Engineer, Southern California Edison Company, San Dimas, CA 91773, Nicolas.von@sce.com 3 Professor of Engineering, Department of Engineering, Harvey Mudd College, Claremont, CA 91711, Ziyad.Duron@hmc.edu 4 Principal Structural Engineer, Southern California Edison Company, San Dimas, CA 91773, Mike.Knarr@sce.com
1

Numerical Model Validation

233

overview of the field investigation is presented. Big Creek Dam No. 7 was completed in 1951. It is located on the San Joaquin River in Fresno County, California and impounds Redinger Lake which has a surface area of 465 acres. Figure 1 shows an upstream view of the dam. The dam is 250 feet high, with a compound downstream slope of 0.75 horizontal to 1.0 vertical on the upper portion, flattening to 0.83 horizontal to 1.0 vertical near the bottom. The dam is 875 feet long and consists of 19 monoliths.

Figure 1. Big Creek Dam No.7 profile EXPERIMENTAL TESTING Forced vibration tests were conducted on Big Creek Dam No. 7 in the summer of 2008. A large eccentric mass vibrator, or shaker, was employed to induce steady-state responses in the dam over a 20 Hz range at force levels approaching 20,000 lbf. The frequency spacing was selected to be 0.05 Hz. The mass vibrator and the associated force curves are shown in Figure 2. The blue line indicates the force curve for the large weights and the red line for the small weights.

20,000 16,000 12,000 8,000 4,000 0

Shaker Force Curves

Force (lbf)

8 12 16 Frequency (Hz)

20

Figure 2. The shaker system and force curves used for vibration testing. Hydrodynamic pressure responses in the reservoir along the upstream face of Monolith 6 were acquired using a hydrophone instrument array. The array consists of 8 solid-state

234

Collaborative Management of Integrated Watersheds

hydrophones spaced 50 ft apart along an 800 ft long Kevlar reinforced cable. The hydrophones are Halliburton Geophysical Services Model 12 and contain a piezoelectric crystal which produces a voltage proportional to hydrodynamic pressure. Hydrophone sensitivity averaged 0.7 V/psi. The reservoir depth at the measurement location was less than 150 ft, so the hydrophones were looped such that the spacing was reduced to 25 ft, allowing for a total of 6 pressure measurements to be made. Accelerations on the crest of the dam and along the foundation interface were acquired using Q-Flex Model QA-700 and QA-750 accelerometers manufactured by Honeywell. These accelerometers incorporate a servo-force-balance design to produce a current output which is proportional to surface acceleration, which allows extended cable lengths to be used throughout the dam without loss of signal quality. Accelerometer sensitivities ranged from 10.7 V/g to 12.3V/g for the tests at Dam No.7. Analog signals were band-pass filtered using a 2-pole high-pass Butterworth with cutoff at 1 Hz and a 4-pole low-pass Butterworth with cutoff at 30 Hz. Amplification gains were varied depending upon signal strength between 10 and 1000 to fill the dynamic range of the A/D converter. These signals were digitized at a rate of 1000 samples per second to achieve satisfactory signal quality while guarding against unwanted aliasing effects. A sample steady-state time history of hydrodynamic pressure data is shown in Figure 3. This is representative of the signal quality observed during testing, for both hydrodynamic and acceleration measurements.

0.02 Pressure (psi) 0.01 0 -0.01 -0.02

Sample Steady-State Time History Depth 67.5 ft

0.2

0.4 0.6 Time (sec)

0.8

Figure 3. Sample steady-state time history of hydrodynamic pressure at 67.5 ft below the water surface.

Numerical Model Validation

235

FINITE ELEMENT MODELING 2-D Single Monolith Finite Element Model A 2-D finite element model of the 7th monolith, the tallest non-spillway monolith, which is representative of non-spillway blocks 5 to 7 and 11 to 13, was developed. It was modeled using SAP2000, which is a commercially available general purpose finite element package (Ref 1). The model includes the dam and portions of the impounded reservoir and underlying foundation. A 2-D mesh is shown in Figure 4.

Figure 4. Mesh of 2-D model of the 7th monolith. The model of the gravity dam consists of four-node plane strain elements, which are restrained out-of-plane, thus supporting two translational degrees of freedom and one rotational degree of freedom. The foundation model consists of the same elements and extends approximately one dam height upstream, downstream, and below the dam. The bottom and side boundaries of the foundation were modeled using a fixed condition at the base and translations fixed laterally but free to move vertically at the sides. The foundation model was assumed as massless. The reservoir was modeled using various methods as will be detailed in later sections.

236

Collaborative Management of Integrated Watersheds

3D Full Dam Finite Element Model A 3-D model of the dam-foundation-reservoir system was constructed and analyzed using Abaqus (Ref 2), a commercially available general purpose finite element package. The model was constructed for the purpose of evaluating its ability to reproduce observed behavior. It incorporates some major features of the dam system while others, such as the spillway gates, are accounted for by lumped masses or parameter variation. Features not modeled include the spillway radial gates, the deck over the spillway section, the penstocks or outlets, and the gallery in the dam. The model of the dam consists of eightnode brick elements whose material properties were based on previous core test results. The model of the foundation was contoured and shaped using the outline of the dam to extend away from the dam in the stream direction a distance equal to the dam height. The model of the reservoir was placed at an elevation 15 ft below the crest, the approximate level of the reservoir during the actual forced vibration test period. The dam model incorporates 17 monoliths which are separated by vertical joints. The joints are not keyed, but lower portions of the tallest monoliths are grouted. To account for the condition of the joints in the dam, and to provide capability within the model to evaluate the influence of joint closure or opening on dam response, the joints in the dam were modeled as a thin (1 in wide) column of elements between monoliths. The column allowed material property parameters to be adjusted so that fully closed (high modulus, nominal concrete mass density) or open (low modulus, low mass density) effects of joint behavior on dam response could be evaluated. Modeling actual joint behavior in a large dam, particularly under strong motion loading conditions typically requires sophisticated, non-linear modeling techniques. However, for the studies and comparisons of the lowlevel response behavior of Dam No. 7, a linear elastic model capable of adjusting relative joint stiffness and mass to account for open or closed joint behavior was deemed satisfactory. The reservoir geometry included in the model is shown in Figure 5. Water only:

Figure 5. Reservoir model geometry (images not drawn to scale). The reservoir model includes a sediment layer along the floor, known to be present in the reservoir at Dam No. 7, and indicated in brown (left image, Figure 5). The sediment layer begins 40 ft upstream from the dam and slopes upward to the measured elevation within 15 feet of the bottom of the original reservoir. This sedimentation profile was

Numerical Model Validation

237

determined to give the best results. Sediment thickness varies from 4 ft to 27 ft within the reservoir. A complete mesh of the model is shown in Figure 6.

Figure 6. Meshed 3-D finite element model of Dam No. 7. The dam and foundation elements are eight-node linear solid bricks, and the reservoir elements are eight-node linear acoustic elements. Interactions along the dam-reservoir and foundation-reservoir interfaces are selected to ensure that only normal motions of the dam and foundation result in pressure changes in the reservoir. The acoustic element was chosen to facilitate comparisons against measured steady-state hydrodynamic pressures, and incorporates the assumption of compressible, inviscid fluid flow. The foundation boundaries are fixed everywhere except along the foundation surface at the dam crest elevation. Reservoir water boundary conditions consist of setting the surface acoustic element nodes to zero pressure, consistent with actual atmospheric conditions in the reservoir. The boundary at the far upstream face of the reservoir model was evaluated under both prescribed and unprescribed conditions. Using the acoustic element formulation, a nonreflecting boundary can be specified in terms of an acoustic impedance to define a termination. The impedance can be user defined or associated with a range of geometries available in Abaqus. For the comparisons shown here, an acoustic termination associated with a planar geometry was selected.

238

Collaborative Management of Integrated Watersheds

PARAMETER STUDY A parameter study was conducted on the 3-D model. Material property values included in the dam-foundation-reservoir model are listed in Table 1. Included are values for the concrete in the dam, the rock in the foundation, and the water and sediment in the reservoir. The joint properties listed in Table 1 were arrived at after a series of eigenanalyses in which model behavior, characterized in terms of resonant frequencies and response shapes, was compared against measured behavior from the forced vibration tests. Table 1. Material properties used to evaluate dynamic response behavior of Dam No. 7. Property Measured Model Concrete (Dam) Density 150 pcf 150 pcf Youngs Modulus 3.3 Mpsi 4.2 Mpsi Poissons Ratio 0.2 0.2 Foundation (Rock) Density 150 pcf 0.150 pcf Youngs Modulus 3 Mpsi 2.5 Mpsi Poissons Ratio 0.25 0.25 Monolith Joints Density N/A 0.15 pcf Youngs Modulus N/A 7.3 Kpsi Poissons Ratio N/A 0.2 Water (Reservoir) Density 62.4 pcf 62.4 pcf Bulk Modulus 315 Kpsi 181 Kpsi Water (Sediment) Density 125 pcf 125 pcf Bulk Modulus N/A 218 Kpsi To arrive at these values, an iterative approach was used in which the material properties were varied and the numerical frequency response computed and compared to the experimental data. The two aspects that were targeted for this comparison were the location and magnitude of the resonant peaks along the dam. Computed frequency response functions were obtained from a steady-state harmonic analysis using the model in which both a direct integration and modal solution approach were evaluated. The direct integration approach is assumed to provide added accuracy in terms of evaluating the full response model behavior, as compared to results from a modal approach which truncates the models frequency characteristics included in the response calculations. To assess the relative accuracy of the direct integration and modal solutions for Dam No. 7, analyses using both approaches were completed and results from each are included in the comparisons of acceleration responses. Damping associated with each approach differed slightly Rayleigh damping was used to ensure 5% of critical at the 1st and 3rd modes, and damping in the amount of 5% of critical was Numerical Model Validation 239

used for all modes included in the modal solution. Fifty modes were used in the modal solution approach representing behavior between 3-20 Hz, chosen to match the frequency range during test. The magnitude and phase of the frequency response function for Monolith 7 is shown in Figure 7.

Response Magnitude (g/Mlbf)

1 0.8 0.6 0.4 0.2 0 0

Numerical and Experimental FRF Comparison Monolith 7 Direct Modal Experimental

10 Frequency (Hz)

15

20

Response Phase (degrees)

200 150 100 50 0

Numerical and Experimental FRF Comparison Monolith 7

Direct Modal Experimental 0 5 10 Frequency (Hz) 15 20

Figure 7. Magnitude (top) and phase (bottom) of the Acceleration Frequency Response at Monolith 7. The comparison shown is considered to be quite good. The models frequency behavior below 6 Hz compares well against observed behavior even though the character of the response is pseudo-dynamic at best, and the agreement suggests that the stiffness-mass ratio of the model in this frequency range is reasonable. The models behavior in the 914 Hz frequency range is characterized by single and closely coupled peak response. The modal solution provides a better match with the experimental result based on the apparent

240

Collaborative Management of Integrated Watersheds

modal coupling between 10-13 Hz. The direct solution response does not indicate the same degree of coupled behavior. Nonetheless, the overall magnitudes in these responses in this frequency range are consistent. Between 14-20 Hz, the comparisons remain fairly consistent in terms of magnitudes and trends, but details associated with individual resonant peaks vary throughout this frequency range. The agreement in the phase response curves is also considered to be quite good, especially when the complex nature of the interaction effects between the damfoundation, dam-reservoir, and foundation-reservoir are considered. Foundation flexibility was measured during the forced vibration tests at Dam No. 7 by measuring vertical acceleration response at the heel and toe of the 7th and 11th monoliths. This test is similar to those previously conducted on large concrete gravity dams including Folsom Dam and San Vicente Dam, where flexibility was characterized by the ratio of heel-to-toe response (magnitude) that exhibited a 180 degree (phase) difference. For those dams, foundation flexibility was reported using a single ratio at the particular frequency where 180 degree response was indicated. The numerical model studies of Dam No. 7 have allowed a broader study of foundation flexibility. Shown in Figure 8 is a comparison of measured and predicted (model) frequency response function defined as the ratio of vertical heel-to-toe response at Monolith 7.

Relative Acceleration Response

4 3 2 1 0 6.5

Numerical and Experimental FRF Comparison Heel-to-Toe Monolith 7 Direct Modal Experimental

7.5 8 Frequency (Hz)

8.5

Numerical Model Validation

241

-100 Relative Phase (degrees) -150 -200 -250 -300 -350 6.5

Numerical and Experimental FRF Comparison Heel-to-Toe Monolith 7

Direct Modal Experimental 7 7.5 8 8.5 Frequency (Hz) 9 9.5

Figure 8. Relative acceleration magnitude (top) and phase (bottom) of the heel-to-toe at Monolith 7. The best match between model and observed relative behavior, where magnitudes are of the same order and phase is near 180 degrees (denoted by a dashed line in Figure 8) , occurs in the vicinity of 8 Hz. Outside this narrow range, however, the comparison of relative magnitude is poor, even though the phase comparison appears to be slightly better (below 8 Hz). A comparison of magnitude and phase heel-to-toe rocking behavior defined at 7.75 Hz, where phase is at or near 180 degrees, is provided in Table 2. These values are taken directly from the predicted and measured vertical responses in 7th and 11th monoliths where near out-of-phase motion (rocking) at each base is observed. Table 2. Foundation flexibility indicative of near out-of phase rocking base response. Monolith 7 Monolith 11 Method Phase(degrees) Magnitude Phase (degrees) Magnitude Model/Direct -205 1.357 -134 1.158 Model/Modal -193 0.842 -146 1.051 Experimental -192 0.467 -173 3.226 These foundation flexibility studies have been used to adjust the elastic modulus used in the foundation finite element models. RESERVOIR MODELING METHODS Westergaard Westergaard added mass was incorporated into the finite element analysis by applying a mass to each upstream node, scaled by the tributary area around that node. The added mass is equal to: 242 Collaborative Management of Integrated Watersheds

(1) where b is added mass, H is the depth of the reservoir at the node, y is the depth of the node below the water, is the density of the water, g is the acceleration due to gravity, and A is the tributary area. John Halls RSVR2 RSVR2 is John Halls upgraded version of RSVR. It calculates added mass similar to the Westergaard method as well as additional hydrodynamic forces on the dam which arise from accelerations of the floor and sides of reservoir. The output of the program was applied to the 2-D model as added mass and force multiplied by earthquake acceleration. Fluid Elements The water was modeled as a fluid using four-node plane strain elements in the 2-D model. For a 3-D model eight-node brick elements can be used in a similar way. The water mass is included to capture hydrodynamic effects, but the weight of the water was not included because the hydrostatic pressure was modeled using pore pressure. Essentially, the gravitational pull of the water was set to zero. The properties assigned to the fluid elements were: Table 3. Fluid element properties Density 62.4 pcf Shear modulus 63.24 psi Modulus of elasticity 189.7 psi Wave speed 4720 ft/sec Poissons ratio 0.4999 Bulk modulus 3.16 Mpsi Using plane strain elements to represent the water adds a large number of extraneous modes to the analysis. To minimize this effect, Ritz vectors were used in lieu of the standard eigenvectors. The first mode of the dam was found to be at the calculated 43rd mode of the analysis. All preceding modes were found to be water modes with no significance to the analysis. Boundary Conditions: The upstream boundary of the water was restrained laterally but free to move vertically. Gap elements were introduced between the foundation and water and between the dam and the water to keep the water in contact with the dam and foundation. The gap elements allowed only compression to be transmitted from the water to the dam and foundation.

Numerical Model Validation

243

Acoustic Elements The following governing relationships apply to acoustic elements, whose single degree of freedom is pressure: (2) where is the dynamic pressure, is the spatial position of the fluid particle, is the fluid particle velocity, is the fluid particle acceleration, is the volumetric drag, and is the density. The constitutive relationship for the acoustic fluid is given by: (3) where is the dynamic pressure, is the bulk modulus, and is the volumetric strain. Additional details concerning the formulation of the acoustic finite element in Abaqus are found in the Abaqus/Standard Users Manual (Ref 2).

COMPARISON OF RESULTS Resonance Frequency Comparison The eccentric mass shaker forces were applied at the crest of the 2-D and 3-D models. Figure 9 - Figure 12 below show a comparison of Westergaard, fluid element, and acoustic element generated resonances to the hydrophone data. Comparisons are made at six different reservoir depths along the upstream face of Monolith 6, where the hydrophones were located. Distances are from the top of the reservoir surface. Using the equation: (4) where C is the wave speed of water and H is the height of the reservoir, produces a rough estimate of the resonance frequency of the reservoir. At the 7th monolith of Big Creek Dam No. 7 the height of the reservoir is 182 feet, making the reservoir resonance estimate 6.5 Hz. The hydrophone data shows reservoir resonances around 5 and 6.8 Hz. Westergaard is close to the first resonance at 4.6 Hz, but completely misses the second resonance. The results of the RSVR2 output are almost identical to Westergaard and have been omitted. The fluid element resonances are at 4.7 and 6.8 Hz. The first and second resonances of the acoustic elements are at 4.8 and 6.9 Hz. The magnitude of the first resonance is around 0.2 (psi/million lbf), which is close to the hydrophone data. The fluid element and Westergaard magnitudes are significantly higher because the shaker forces are being applied to a single monolith instead of being distributed across the entire dam.

244

Collaborative Management of Integrated Watersheds

Pressure/Force (psi/million lbf)

0.4 0.3 0.2 0.1 0 17.5 ft 42.5 ft 67.5 ft 92.5 ft 117.5 ft 142.5 ft

Hydrophone Data

5 Frequency (Hz) Westergaard (added mass)

Figure 9. Hydrophone pressure results.

Pressure/Force (psi/million lbf)

15 17.5 ft 42.5 ft 67.5 ft 92.5 ft 117.5 ft 142.5 ft

10

5 Frequency (Hz)

Figure 10. Westergaard pressure results.

Numerical Model Validation

245

Pressure/Force (psi/million lbf)

12 10 8 6 4 2 0 3 4 17.5 ft 42.5 ft 67.5 ft 92.5 ft 117.5 ft 142.5 ft

Fluid Elements

5 Frequency (Hz) Acoustic Elements

Figure 11. Fluid element pressure results.

Pressure/Force (psi/million lbf)

1 0.8 0.6 0.4 0.2 0 3 4 17.5 ft 42.5 ft 67.5 ft 92.5 ft 117.5 ft 142.5 ft

5 Frequency (Hz)

Figure 12. Acoustic element pressure results from modal solution approach. Stress Comparison The stress results of the 2-D model were used to compare Westergaard, RSVR2, fluid elements, and output from a program developed by Chopra (Ref 3). The time history used in finite element analysis was from the Imperial Valley record. All stress results were taken at the heel of the monolith, where stresses are generally highest in an earthquake. Heel cracking due to high stresses can increase uplift pressures significantly, undermining the structures stability.

246

Collaborative Management of Integrated Watersheds

As seen in Table 4, Westergaard was found to be very conservative with results 15% greater than Chopra, while the fluid element results were within 6% of the Chopra results. RSVR2 results were nearly identical to Westergaard. Chopra (psi) 187 Table 4. Stress results from the 2-D model. Westergaard Fluid Elements Time History (psi) Time History (psi) Response Spectrum (psi) 215 177 137

CONCLUSIONS AND FURTHER STUDY Experimental results can be used to calibrate the parameters of a numerical model to achieve desired performance. Model performance based on the comparisons presented in this paper suggests that the models constructed of Dam No. 7 achieve a good measure of correlation with observed behavior below 16 Hz. This is particularly the case in the modes that correspond to the crest and upper portions in the dam, where reasonably good agreement is achieved. Various reservoir modeling methods have been compared: The Westergaard approach is quick and easy to implement. With fewer elements it also requires the least amount of computing power. It generally produces overly conservative stress results and does not match all the resonances seen in the hydrophone data, but it is still sufficient for a preliminary analysis. The RSVR2 method includes additional forces that may make it more accurate than the Westergaard method but still has the same shortcomings. It also takes considerably more time to implement. The model incorporating fluid elements matches the hydrophone data well and produces good stress results. A large number of water sloshing modes were generated that made it more difficult to analyze the reservoir-dam system since they would crowd out modes with large participation ratios. The acoustic element method produces results that are an excellent match with the hydrophone data. Matching the resonances seen in the hydrophone data is of great benefit in developing confidence in the models accuracy to represent the real behavior of the dam in response to ground motion.Water sloshing modes are not generated because acoustic elements do not translate. More sophisticated reservoir modeling techniques such as coupled Eulerian-Lagrangian systems could yield even better results, but are more difficult to implement and necessitate considerably more computing power.

Numerical Model Validation

247

REFERENCES 1. SAP2000, Version 12.0.2, Computers and Structures, Inc., 2009. 2. Dassault Systemes Simulia Corp. Section 21.3.1 Abaqus/Standard Users Manual. Version 6.8, 2008. 3. Simplified Earthquake Analysis of Concrete Gravity Dams, Fenves and Chopra, Journal of Structural Engineering, Vol. 113, No. 8, August 1987, pp. 1688-1708.

248

Collaborative Management of Integrated Watersheds

Vous aimerez peut-être aussi