Sampling based numerical seismic assessment of continuous span RC bridges

• A computationally non-expensive simulation scheme is investigated for RC bridges.
• Multiple bridges and real ground motions are employed for reliable conclusions.
• Statistical characterization of pier cross sections requires 50–100 realizations.
• Failure probability of RC bridge piers requires at least 200 realizations.
• Latin Hypercube is resilient to high dispersion levels and independent repetitions.

The seismic response of RC structures is dictated by a number of aleatory and epistemic uncertainties, related to material properties and seismic hazard. Typically, when following a probabilistic approach, the different variables are statistically defined using a random simulation tool that will take into account the different sources of uncertainty. This manuscript addresses the seismic reliability of RC bridge structures, particularly through the assessment of a random sampling scheme based on the Latin Hypercube algorithm for numerical simulation of material and ground motion properties. An optimal use of such sampling technique is thus sought so as to ensure higher accurateness in the calculation of the failure probability of a set of different existing bridges, through a relatively simple framework. The study demonstrates the robustness of the Latin Hypercube algorithm, which secured accuracy of the estimated probability of failure for both regular and irregular configurations. Recommendations to researchers and practitioners are made accordingly, particularly with respect to the minimal sampling size, which does not need to exceed one to two hundred realisations.