Data-informed assessment of SSI effects on peak floor acceleration: a case study on the Fukushima nuclear power plants

Peak Floor Acceleration (PFA) is a key parameter for assessing seismic demand on nonstructural components and systems. Critical infrastructure such as nuclear power plants (NPPs) is particularly vulnerable to the failure of these secondary systems during earthquakes. However, NPP structures are often deeply embedded in soil, resulting in significant soil–structure interaction (SSI) effects that are not considered in simplified code-based PFA prediction approaches. Moreover, the combined influence of SSI and the frequency content of ground excitation on PFA predictions has not been systematically investigated. This study examines the impact of SSI and high-frequency excitation on PFA predictions for deeply embedded structures using ground and structural acceleration records from the Fukushima NPPs during the 2011 Tohoku earthquake. Simplified code-based approaches are evaluated against the recorded data. It is found that FEMA P-58 and Eurocode 8 overpredict horizontal PFAs by about 20% on average, whereas NEHRP 2020/ASCE 7–22 underpredict them by about 12%, with overall biases within ± 16%. Then, random vibration analysis is applied using analytical and empirical (i.e. data-derived) frequency-domain SSI as well as structural transfer functions calibrated from observations and their potential for useful PFA predictions is compared vis-a-vis. It is seen that the empirical SSI transfer functions are reliable up to about 10 Hz, whereas the use of analytical SSI transfer functions alone lead to significant PFA overprediction. Thus, new formulations for SSI kinematic interaction are proposed to adjust analytical SSI transfer functions based on observational data, which achieve superior PFA prediction potential, highlighting the essential role of observations in accurately reproducing the PFAs. Further, a parametric study is undertaken revealing that SSI can differentiate PFAs by factors between 0.2 and 0.8, depending on earthquake magnitude and regional κ, which establishes the existence of strong interaction between SSI and high-frequency amplitude decay. Overall, the herein reported findings emphasize the importance of observationally constrained SSI modelling and site-specific excitation characteristics for reliable PFA prediction in critical infrastructure.