Determination of Dynamic Characteristics of Double Curvature Concrete Dez Dam by Using Ambient Vibration Test with Three Different Signal Processing Methods

Considering the human and financial costs resulting from the destruction of dams, studying the seismic behavior of these structures and controlling their stability against events like earthquakes is crucially important. Numerical models are utilized to carry out this vital process. However, these models possess many assumptions that differ between mathematic models and reality. As a result, experts use system identification methods to ease such problems. Generally, system identification consists of two parts; experimental tests and signal processing. The first part refers to field experiments like ambient vibration tests, and the latter section means analyzing records gathered from the field tests. From the estimated dynamic properties, the mathematical models can be calibrated to evaluate the possible responses of the building under study to future earthquakes.
The ambient vibration test, used in this paper, is one of the dynamic structural tests which records natural vibrations such as wind and human activities on the structures. These oscillations stimulate the structure’s dynamic properties on a marginal scale which can be seen and analyzed in seismic records. Recently, an ambient vibration experiment was conducted on the double curvature concrete Dez dam, which is one of the highest dams in the country. The dam is currently 203 meters high, and a project is being undertaken to add around 8 meters to its current height. For this reason, evaluating the dynamic characteristics of its body and calibrating the existing dam’s numerical models is vital. In this regard, the ambient vibration test with 19 seismometers and accelerometers was conducted when the height of the dam reservoir was at its maximum water level (MWL). Sensors were placed on the dam in four different layouts to obtain dynamic mode shapes from the plan and the dam’s height.
In this paper, some stationary records have been selected from data recorded within several days, and the data are analyzed with three different methods. These methods are 4-Spectrum, Frequency Domain Decomposition (FDD), and Enhanced Frequency Domain Decomposition (EFDD). Each of the techniques used in this research has advantages and disadvantages. The 4-spectrum method is a reliable method for obtaining the modal frequencies owing to the simultaneous control of the four spectral with each other. However, since all sensors' spectrums must be separately controlled by different source points, some possible errors lie in this approach. In the FDD, the power spectrum density and correlation spectrum are automatically controlled by the computer, and a powerful mathematical tool called SVD is used to analyze the data matrix. The EFDD has a similar structure to the FDD method, while it uses a special algorithm to select the peaks in the spectrums. This makes it slightly more accurate than FDD. Furthermore, to calculate modal damping, the Half Power method was used in all these methods.
After analyzing all the records with different system identification methods, natural frequencies in six modes were found in the frequency range of 2.63 Hz to 13.92 Hz and mode shapes in five modes were drawn. Likewise, the damping ratio was estimated at 1.8% to 0.6%, which shows the half-power method is not an accurate way to calculate the damping ratio in concrete dams. Finally, the process of system identification, using field experiments, is an experimental process, for obtaining more accuracy, the results extracted from the Dez dam were compared with the outcome of Birkeh Dam's essay, which has the same features as the Dez dam. The comparison of the system identification results shows that the extracted outcomes are relatively precise, which is suitable for calibrating the existing computer models.