Seismic Performance Assessment Of Highway Bridges And Bridge Networks Based On Data-driven Methods

Many post-earthquake reconnaissances inticated that bridges may suffer catastrophic damages under earthquakes,posing a severe threat to the operation safty of bridge networks.It is meaningful to evaluate the seismic performance of highway bridges and bridge networks in a rapidly way for disaster relief works.This study focuses on the rapid seismic performance assessment of highway bridges and bridge networks based on data-driven methods,the main contents include:(1).An effective finite element simulation method for RC columns based on machine learning methods is proposed.To this aim,a large number of historical tests on RC columns are collected and screened for a test database,which can provide a comprehensive and well-distributed database for learning the mechanisms of RC columns.Then machine learning methods are implemented to relate the column charecteristics and their critical properties.By comparing with test results,existing formulas and other structure tests,the proposed model is validated to be able to provide more reliable estimates for seismic design and finite element simulation of RC structures.(2).A multiple level of detail finite element(LODFE)model with corresponding model parameter calibration methods are presented.The critical characteristics influencing the seismic performance of bridges are first identified.Then,bridge models of different level of detail,including FE models of high,moderate and low levels of detail can be generated for regional bridges according to engineering demands.The effects of the classes and characteristics of key components(decks,piers and abutments,etc.)are considered through the developed machine learning model for the critical properties of RC columns and the parameter determination methods for other components.They can be used to rapidly estimate the seismic damages of regional bridges by conducting timehistory analysis on the corresponding LODFE models.(3).A data-driven fragility model(DFM)is developed for the seismic assessment of bridges.By referring to the knowledge about seismic design and fragility analysis of bridges,the influencing characteristics and fragility parameters of bridges are identified.Besides,a procedure to build the detailed finite element model of bridges according to the influencing characteristics is proposed.Then,a well-distributed bridge database is designed,while the incremental dynamic analysis(IDA),probabilistic seismic demand analysis(PSDA)and fragility analysis are conducted on them with a series of selected ground motions.After that,the ANN-based fragility model is trained and generated by learning the relationship among the influencing characteristics and fragility parameters,which is demonstrated by comparing with the fragility curves by the conventional IDAbased method.(4).A seismic performance assessment method for regional bridge networks is proposed.The functionalities of bridge networks are summarized and represented by holistic system funcationality resilience,important subsystem resilience and functionality resilience for the response of secondary disasters or emergencies.Based on the LODFE and DFM methods,the damage states of regional bridges can be estimated in different cases.Along with other influencing factors,including the seismic hazard and traffic flow,etc.,the functionalities of bridge networks can be analyzed.Finally,for comprehensive pre-disaster management and post-disaster recovery and response,analytic hierarchy process(AHP)is used to integrate the functionality resilience vectors according to the importance of functionalities.It can provide a useful reference for shareholders to the decision-making in both pre-and post-earthquake phases.