Seismic fragility analysis for prestressed concrete containments of nuclear power plants based on stripe-based approach

Abstract

The Separation of Variables (SOV) approach has been widely used for the seismic fragility analysis of nuclear power plant structures. It begins with linear elastic analysis, and inelastic behavior (e.g., changes in stiffness and damping ratio) is indirectly considered employing the inelastic energy absorption factor. However, recent studies suggest enhancing the accuracy of fragility curves through nonlinear analysis. This study evaluates the efficacy of two-stripe based nonlinear approaches: Method A (fitting a lognormal function) and maximum likelihood estimation (MLE) for assessing the seismic fragility of prestressed concrete containment structures. It also proposes damage measures and failure criteria suitable for considering the effects of uncertainties in structural parameters on failure under high earthquake loads. The uncertainty of demand is assessed utilizing 30 sets of ground motion time histories. The uncertainty of capacity is assessed using 30 random variable sets of three material properties-concrete compressive strength, 1st natural frequency, and damping ratio-generated by Latin hypercube sampling. Pushover and time-history analysis indicate that failure criteria based on lateral drift consistently capture failure more accurately than those based on base shear or strain levels. Both stripe-based approaches yield median capacities within 2 % of the SOV method. When accounting for uncertainties in both demand and capacity, the stripe-based approaches align better with observed failure ratios at high PGA levels, though MLE tends to overestimate failure ratios at lower PGA. These findings suggest that combining method A and MLE can more comprehensively capture nonlinear behavior across the full PGA range, leading to more accurate failure predictions.