Evaluation of nonlinear response biases of arch dams due to amplitude scaling via three-component record selection based on conditional spectra

The nonlinear response of ultrahigh concrete arch dams under strong or even extreme earthquakes is particularly important for assessing their safety performance, which is highly dependent on the input ground motions (GMs). Based on the current China probabilistic seismic hazard analysis (CPSHA) scheme, a multicomponent hazard consistency conditional spectrum (CS) was first constructed to meet the specific hazard level at the Baihetan site; the existing issues in evaluating the seismic performance of ultrahigh arch dams via CS-based tridirectional records were then investigated. The results show that neglecting the vertical (V) component hazard consistency may result in significant biases in estimating the response dispersion of arch dams under extreme earthquakes as constraints on the V component are removed, but it has little effect on the mean response. In the CS-based tridirectional record, an excessively large scaling factor (SF) can lead to a significant overestimation of the dam's nonlinear responses because the selected record set, which is mainly composed of far-field low-amplitude records with long duration, tends to possess more energy content when scaled up than the true records. By selecting appropriate SF values (e.g., 2-5), the hazard characteristics of the selected records converge to those of the unscaled records, so the response bias of arch dams caused by neglecting the V component hazard consistency can be diminished, which provides a reference for assessing the seismic performance of structures (e.g., high arch dams) that are significantly influenced by the V component.