Enhancing computational efficiency of Bayesian Inference by identifying the intensity measure range to update seismic fragility curves

Abstract

The United States Nuclear Regulatory Commission (US NRC) has established stringent criteria for the acceptance of new Standard Plant designs. These criteria require that the fragility curves have a high degree of confidence, especially in the low-probability regions. However, conventional methods of developing seismic fragility curves require a substantial number of dynamic analyses, which can be computationally intensive. To address this challenge, the Bayesian framework offers a more efficient and effective solution. Bayesian Inference enables the integration of prior knowledge with newly acquired data to refine the data-generating process. This manuscript presents a systematic Bayesian framework to update the seismic fragility curve of structures, systems, and components (SSCs). The efficiency of the framework is illustrated using an application to the seismic fragility of a concrete shear wall. Concrete Damage Plasticity Model (CDPM) is used to characterize the nonlinear behavior of concrete. The seismic fragility curve developed with our proposed approach aligns closely with those generated through conventional nonlinear simulations while significantly reducing the computational cost.