Abstract: With the advancement of safety monitoring technology, new technologies, such as the flexible intelligent displacement meter and pipeline robot, have been increasingly employed in the safety monitoring of rockfill dams. Over the lifespan of a high rockfill dam, an extensive amount of monitoring data has been accumulated. Using these data to conduct parameter inversion analysis can help enhance the accuracy of numerical simulations of rockfill dams and improve assessment of dam safety. This study develops a new method for using time series clustering to identify the representative and diverse sets of measurement points from the vast monitoring data. This method extracts time series features to construct an objective function that captures the spatiotemporal evolution characteristics of deformation in rockfill dams, and adopts a multi-objective optimization algorithm to perform material parameter inversion for each zone of the rockfill dam. Compared to the conventional methods of parameter inversion, it enables systematic use of a substantial amount of rockfill dam monitoring data, effectively capturing the deformation development and spatial distribution characteristics in the filling, impounding, and operational stages. The calculations of dam settlement obtained through the inversion analysis exhibit a favorable agreement with the measurements, demonstrating a significant improvement in the accuracy of parameter inversion.

Key words: rockfill dam, monitoring data, time series feature, parameter inversion, measuring point selection, multi-objective optimization