Probabilistic assessment of the seismic performance of two earth dams in Southern Italy using simplified and advanced constitutive models

Abstract

The large majority of existing earth dams were designed with old standards, which often accounted for the effects of earthquakes in a simplified manner. Nowadays, safety assessment of these structures is becoming of great importance, particularly for dams suffering the effects of ageing. This study presents a fully probabilistic approach to evaluate the seismic performance of two critical earth dams in the Calabria region, a seismically active area in Southern Italy. One of them (the Farneto del Principe dam) is not susceptible to liquefaction, whereas the other dam (the Angitola dam) is founded on potentially liquefiable soils. Seismic input motions are derived from site-specific probabilistic approaches. Non-ergodic ground response is implemented within a probabilistic seismic hazard analysis (PSHA) framework for one of the two dam sites. This non-ergodic PSHA is derived from numerical amplification functions based on one-dimensional simulations. Such well-documented early application of non-ergodic PSHA for earth dams in Italy may encourage a transformational shift from years of past practices based on deterministic amplification functions merged with PSHA results by means of hybrid approaches. Simplified (i.e., using the Mohr–Coulomb failure criterion coupled with a simplified hysteretic procedure) and advanced (i.e., PM4Sand and PM4Silt) constitutive models are used to perform a comprehensive numerical simulation program for both dams. Field and laboratory geotechnical characterization data are used to calibrate these models. This calibration process is fully documented and potential issues discussed. Such fully-documented calibration process will enable future studies on similar infrastructure systems when advanced constitutive models are necessary. Shear strain and deformation patterns are analyzed and discussed, showing that for the Farneto del Principe dam (comprising non-liquefiable materials) both constitutive models provide similar results. However, when potentially liquefiable soils are involved, advanced constitutive models are necessary to capture the complexity and nuances of such materials. This effect is evident for the Angitola dam. For both dams, seismic vulnerability is analyzed by means of analytical fragility functions for various damage mechanisms and intensity measures. Such fragility functions are based on nonlinear deformation analyses within the multiple stripe analysis framework. All fragility functions derived in this study are shown and main outcomes are illustrated by summary tables reporting mean and standard deviation values of these curves. Finally, the efficiency and predictability of various ground motion intensity measures to predict different damage levels and mechanisms are calculated for both dams. Predictability of recent semi-empirical ground motion models is also calculated for all analyzed intensity measures. Overall, results from this analysis indicate that velocity-based ground motion properties, such as Peak Ground Velocity, Arias Intensity, Cumulative Absolute Velocity, and Cumulative Absolute Velocity after application of a 0.05 𝑚𝑠2 threshold acceleration provide good efficiencies in predicting damage. These intensity measures are the best in predicting damage states for both dams and all damage mechanisms. However, some of them are more predictable than others. After merging efficiency and predictability information, the best intensity measure to predict damage is the Cumulative Absolute Velocity, followed by the Arias intensity.