This report presents a general probability-based approach for assessment of roadway safety hardware (RSH). It was achieved using a reliability analysis method and computational techniques. With the development of high-fidelity finite element (FE) models, numerical crash simulations can be performed to evaluate various RSH systems, in addition to crash tests. For highly nonlinear and implicit impact responses, metamodeling techniques provide a rational approach to replace the expensive numerical simulations. In this study, radial basis functions (RBFs) were employed to create approximation functions of limit state/performance functions using a relatively small number of sample points. Once the RBF metamodels were created, the failure probabilities were estimated using simulation methods such as Monte Carlo simulations (MCS).