Seismic responses of ordinary highway bridges that feature stiff superstructures have been shown to be strongly affected by abutment-backfill interactions. Seismic risk assessment of these bridges at the regional scale faces the added challenge of having to deal with the large uncertainty in backfill properties. In this regard, the study develops probabilistic backfill models to better quantify the uncertainties in modeling the abutment. First, efforts to establish the validity of the numerical method in predicting the pushover response of backfills are described. Advanced plasticity materials are used in finite element models (FEMs) to simulate sandy and clayey soils that yield consistent backfill force-displacement relationships against full-scale test results. Second, a probabilistic analysis framework is constructed to incorporate soil uncertainties that are identified from field investigations. Statistical moments are extracted from the resultant pushover curves to fully define the probabilistic backfill models, which are verified to bear appropriate uncertainty treatment and reasonable height adjustment factors. Further, statistical analysis tools are used to investigate the influences of different backfill models on the bridge demand estimates of two common bridge classes. The study reveals that backfill models will affect the response estimates of different bridge components in both diaphragm and seat-type abutment bridges. However, probabilistic models shall be especially considered on backfills for the bridge components that are expected to have dominant responses in the longitudinal direction. The proposed backfill models appear to outperform previous deterministic models in predicting realistic bridge responses. The models can be employed in the task of regional seismic assessment of various bridge classes.
Read full abstract