Long-term Dynamic Monitoring of a Concrete Block of the Itaipu Hydroelectric Dam

Dissertação apresentada ao Programa de Pós-Graduação em Engenharia Civil da Universidade Federal da Integração Latino-Americana, como parte integrante dos requisitos para obtenção do título de Mestre em Engenharia Civil.Dams are structures that offer multiple benefits: water for irrigation, energy, flood control, among others. The structural integrity of these structures can be affected over time by agerelated deterioration, flooding, earthquakes and other factors, representing a potential risk to the population and neighboring ecosystems. Therefore, it is necessary to implement monitoring systems that allow the assessment of the structural integrity of these structures by obtaining and observing the evolution of the main static and dynamic properties. Dynamic monitoring involves measuring the structural response in the form of acceleration due to ambient excitations and then identifying modal parameters: natural frequencies, damping ratios and mode shapes. As these dynamic characteristics depend on physical properties such as mass or stiffness, the detection of significant changes in the modal parameters indicates the presence of structural anomalies. However, many studies have shown that environmental and operational conditions can induce changes in modal parameters to the same degree as the damage. Therefore, dynamic monitoring should include the normalization of information to separate the changes in modal parameters produced by these environmental and operational variables from those produced by damage. In this work, continuous monitoring of the modal parameters (natural frequencies and damping ratios) of a hollow-gravity concrete block of the Itaipu Hydroelectric Dam is performed for three years of operation (January/2018 - December/2020). This block is instrumented with two triaxial sensors that permanently record acceleration data with a sampling frequency of 200 [Hz]. For this application, a methodology was configured to estimate modal properties automatically. The procedure starts from the identification results using the SSI-COV algorithm to 30-minute long accelerations time series. Subsequently, the obtained stability diagram goes through a cleaning phase, where the spurious modes are discarded, a grouping phase, where the modes with similar modal characteristics are clustered, and a selection phase, where the identified physical modes are extracted. These modes are compared with reference modes, allowing automatic tracking of natural frequencies and damping ratios. The process is repeated for each set of 30-minute long time series, producing a new set of modal parameters in each analysis. Once the identification and tracking process is completed, the effect of ambient temperature and reservoir water level on the temporal evolution of the identified parameters is analyzed. Autoregressive moving average models with exogenous inputs (ARMAX) are fitted to separate the effects of environmental variables from the modal parameters. As a result, nine modes were identified and tracked during the three years of operation. Correlation analysis between the identified modal parameters and temperature and reservoir water level showed that these environmental variables influence natural frequencies. ARMAX models were able to mitigate these effects and reduce the variability in natural frequencies by more than 10 [%] for the modes associated with the frequencies most influenced by these environmental conditions