Abstract

When hybrid simulation (HS) with substructures is employed for assessing the seismic behavior of a large complex structure, it is unrealistic to test all the components that may exhibit strong nonlinearity. Hence, the accuracy of the numerical substructure (NS) faces an increased challenge. To this end, this paper will emphasize on improving the accuracy of the NS in hybrid simulation based on the model updating approach. Most hybrid simulations with model updating (UHS) focus on updating the parameters of the component constitutive model (story shear model) leading to large modeling errors and the unknown detail responses. Moreover, the most extensively used component constitutive models, such as the Bouc-Wen model, are the models in a narrow sense because they are different for various RC members when the size of the component, boundary conditions, axial compression ratio, and the volumetric stirrup ratio are different from one component to the other. Thus, numerous parameters are needed to describe the displacement-force relations of different components, which leads to the huge computational burden. With this respect, this paper proposes a novel hybrid simulation approach based on identifying and updating the parameters of the material constitutive model. The main work and results are concluded as follows, 1. The unified constitutive model of unconfined and confined concrete is derived from the existing uniaxial concrete constitutive models by introducing the volumetric stirrup ratio. 2. To solve the problem that the relation of the measurements (force of the specimen of a RC member) and the identified parameters (concrete constitutive parameters) are difficult to analytical expressed, an OpenSees embedded unscented Kalman filter is proposed for parameter identification. To this end, several parts of the OpenSees source codes are developed and modified. 3. The proposed identification method and hybrid simulation based on updating the concrete constitutive parameters are respectively validated through a monotonic loading test on a RC column and a UHS on a RC frame. The results show that the convergence values of each parameter under various experimental cases are close to each other with a small variance, which indicates that the proposed identification method is robust and reliable. Comparing to the standard HS, the accuracy of the NS, hence the UHS, is significantly improved. 4. Apply the proposed UHS to a RC continuous rigid bridge with tall thin-walled piers. It concludes that the performance of the identification method is still quite good. Contrasting to the simpler structure, the improved accuracy of a large complex structure is even greater. Moreover, the accuracy of the NS can be greatly increased even though the model error is increased. By tuning the initial values of constitutive parameters, the negative influence of model error is decreased to further improve the precision of the NS. By observing the specimen, it is found that the thin-walled hollow section specimen is a bending-shear type damage mode and finally damaged due to the tensile rupture of the stirrup.