Rapid seismic damage assessment of railway bridges using the response-surface statistical model

• A response surface model (RSM) is developed for seismic damage assessment of bridges.
• PGA, pier height, and number of spans are defined as the input variables for the model.
• The output of the model is the mean and standard deviation of seismic damage.
• The accuracy of the established RSM model was statistically validated.
• The RSM approach could be effectively used for seismic risk analysis of bridges.

In the conventional approach for analytical fragility analysis of structures, a single set of seismic fragility curves is developed and utilized for risk assessment of structures having the same classification. This approach implicitly assumes that all structures corresponding to the same classification undergo the same level of damage under a given seismic excitation. While this approach is appropriate for assessment of the average seismic risk to a large population of structures, seismic upgrading of arbitrarily selected structures for risk reduction should not be based on the average structure risk because the physical configuration details differ among individual structures having the same classification. This paper proposes a new method for rapid estimation of the seismic damage to track-on steel-plate-girder (TOSPG) bridges so that a seismic risk analysis of a TOSPG bridge with an arbitrary physical configuration can be effectively performed without significant loss of time and effort. The response surface modeling (RSM) technique is utilized for probabilistic estimation of seismic damage to a TOSPG bridge without the need to repeat a large number of time–history analyses. First, the variables that describe the physical configuration of the bridge are identified. Among the variables, the ones that significantly affect the seismic damage of the bridges are selected as the input variables for the response surface model. The response surface model is then developed to create second-degree polynomial equations for estimation of the anticipated values for the median and variation of the seismic damage due to a specified level of earthquake loading. The accuracy of the established RSM model was statistically validated. The approach developed in this study can be effectively applied for making macro-level decisions on seismic retrofit through flexible estimation of the seismic damage and fragility of arbitrarily selected structures in a given class because the simulation is performed not with a number of time–history nonlinear dynamic analyses but with simple numerical equations.