Comparison of First-Order Second-Moment and Latin Hypercube Sampling Methods on Probabilistic Seismic Hazard Analysis of Dam-Reservoir-Foundation Systems

Simplifying the assumptions of a problem may lead to significant design and analysis cost reduction, but increase the risk of hazardous consequences, especially when safety-critical structures such as large dams, skyscrapers, etc. are involved. These simplifying assumptions usually come from considering a homogeneous foundation, while inhomogeneity in rock foundations is more rational in modeling of these mega structures. On the other hand, considering a unique geometry of inhomogeneity and mechanical properties for all site foundations around the world is impossible. The purpose of this study is to perform probabilistic analysis of a critical geometrical situation for concrete gravity dam-reservoir-foundation system, using two approaches of Latin hypercube sampling (LHS) and first-order second-moment (FOSM). Pine Flat Dam is considered as a reference case and tensile strength, sliding of dam base and opening of dam heel are considered as controlling performance functions. The results illustrate that exceedance probability (EP) curves which are resulted from FOSM method are not reliable, whereas the shape of curves are similar to the curves resulted from LHS approach. Furthermore, the results show that the importance measure vector can be derived from FOSM method, while the LHS approach is incapable of providing this information.