Case Study Investigating Probabilistic and Deterministic Sampling Methods for Developing In-Structure Response Spectra: Phase II

The goal of this ongoing, multi-phase research project is to assess the accuracy of simplified deterministic sampling methods used in practice to estimate seismic structural response compared to more rigorous probabilistic sampling approaches. Phase I of this research (Cappa, et al. (2017)) compared probabilistic and deterministic seismic responses of a relatively simple, fixed-base, steel structure founded on rock in terms of median and 84% non-exceedance probability (NEP) in-structure response spectra (ISRS) at selected nodes. The scope of Phase I was limited to assessing the validity of a simplified deterministic sampling method associated only with structure frequency uncertainty. Other sources of structure response variability (soil properties, horizontal direction peak response, structure damping, etc.) were omitted accordingly. This paper presents Phase II, which expands the investigation by analyzing a soil-founded concrete reactor building (RB) and including soil-structure interaction (SSI) effects. A base case study is evaluated, and results are then investigated with three sensitivity cases. Base case deterministic and probabilistic approaches are selected to comply with the ASCE/SEI 4-16 (2017) and EPRI 3002012994 (2018) guidelines for SSI analysis. In the base case, soil properties, structure properties, and horizontal direction peak response variability (HDPRV) are included the probabilistic simulations. The deterministic approach develops ISRS based on five samples in which soil damping and soil and structure stiffness are varied to plus or minus one standard deviation values. To realize a meaningful comparison between the two approaches, deterministic 84% NEP ISRS are adjusted to incorporate variabilities included in the probabilistic simulation (i.e., structure damping, HDPRV, and time history randomness). Results confirm the Phase I findings: deterministic median and 84% NEP estimates generally provide a reasonable approximation to probabilistic responses. Probabilistic and deterministic responses are compared for one horizontal direction at each of two representative locations in the RB. Maximum ISRS discrepancies are less than 30% at all frequencies and less than 20% at dominant peaks. Three sensitivity studies investigate the influence of 1) an alternative method for calculating deterministic median and 84% NEP spectra, 2) a different set of deterministic ground motions, and 3) exclusion of structural damping and HDRPV variabilities from both probabilistic and deterministic analyses. The sensitivity studies provide insights on possible causes of mismatch between probabilistic and deterministic results for this case study.

Read the article.