Steel-concrete composite structures have been widely used in highway and bridge engineering. In a composite girder, efficient shear connectors are the key to ensure the transfer of longitudinal shear forces across the steel-concrete interface. The mechanical behavior of the composite structural members heavily depends on the performance of the shear connections because damage or failure of the connectors will decrease the load carrying capacity of the composite members. In this paper, an acceleration-based shear connector interfacial bonding performance evaluation methodology with neural network where no finite element model of the composite structural member is needed was proposed. Vibration tests on the composite girder before and after loosening or removing some connectors were implemented and the corresponding acceleration responses were acquired. Using acceleration time series at certain location of a substructure of the intact composite girder, a nonparametric base-line model based on neural network were established. The acceleration response measurements of the structure after connectors are loosened or removed were recorded. The nonparametric model was employed to forecast the acceleration response of the damaged composite girder and it was found that the acceleration response of the damaged structural member could not meet the forecasted acceleration response of the damage composite girder. The difference between the output and the measurement provides an index for damage detection of the shear connector. Results show that the proposed nonparametric method is effective to detect the loosening and removal damage of the shear connectors for composite structures solely using acceleration time series from a substructure.

Related Publications