Uncertainty Quantification and Reduction in the Structural Analysis of Existing Concrete Gravity Dams

Abstract

The failure of a large gravity dam might have catastrophic effects putting at risk human lives, not counting the considerable economic consequences. Most of dams are located in natural hazard prone areas so the structural control and the evaluation of the dam fragility (in particular against to flood and earthquake) assume great importance both to apply early warning procedures and to define resilience-enhancing strategies. Numerical models assume great importance to predict the seismic behaviour of the complex dam-soil-reservoir interacting system, nevertheless they are affected by different uncertainties. The effects of uncertainties can be reduced by calibrating finite element models with all available data about the structure. Measurements recorded by monitoring systems and in situ test results take on a major role as important sources of information. This paper investigates the effect of the uncertainties in the static and dynamic analysis of existing concrete gravity dams by means of two case studies. The general Polynomial Chaos Expansion technique is used to propagate the uncertainties through the numerical models of the case studies even without High Performance Computing. The effects of the uncertainties are thus quantified in terms of model output variation. General Polynomial Chaos Expansion-based predictive models are then used for the solution of the inverse problem thus reducing the computational burden.