MS13.2 - Hybrid analyses, experimental tests and numerical modeling in earthquake engineering

In this paper, the results of a bidirectional hybrid test performed on a reinforced concrete column are described. The specimen is loaded at its end by two hydraulic actuators applying the horizontal displacements at the top of the column. The numerical substructure is modelled using nonlinear multifibre beam elements to consider the decrease of stiffness due to the appearance of damage during the strong motion phase of the earthquake. A bilinear elastic-plastic law is used for the steel rebars, and a unilateral damage law with frictional sliding is used for concrete. During the hybrid test, the FEM analysis related to the numerical substructure is here directly run on a real time controller board. To reduce its computational cost, snapshots are firstly computed by performing fully numerical FEM (finite element method) analyses and used to build a POD (proper orthogonal decomposition) projection modal basis. In addition, a DEIM algorithm (discrete empirical interpolation method) is used to build a reduced integration domain (RID) where the nonlinearities are the highest. As a result, material properties are only updated at the elements belonging to the RID, which further decreases the CPU time. The delay of the hydraulic actuators is corrected in real time using NPS (normalized passivity control), guaranteeing the passivity of the transfer system, and ILC (iterative learning control), improving the actuator tracking performance iteratively in order to increase the fidelity of the test out-come. The experimental set up is firstly described (actuators, specimen, equipment, boundary conditions…). To assess the reliability of the controlling procedures in the case of earthquake engineering hybrid tests, the results of the real-time (RT) and pseudo-dynamic (PsD) experiments are then compared, including the force/displacement responses of the actuators, the experimental crack patterns and the experimental strain fields (built by performing field measurements).

A recent large campaign of vibration measurements has been conducted on a hospital building in Italy to perform operational modal analysis of the structure in order to update a representative numerical model. Subsequently, this latter will be used to carry out a vulnerability analysis of the whole structure. The building is composed by 41 structural subsystems separated by technical joints. At the same time, several subsystems are interconnected in different manner especially through a roof sustained by large-span wooden beams. Long acceleration measurements have been acquired by wireless sensors located at the top level of the building in two days testing. During the tests conducted in operational conditions, in different locations of the building air conditioning systems and elevators were functioning affecting the measures in the high frequency content. This work aims to investigate on the structural identification of the complex structure, on the effectiveness of technical joints and on the influence of the wooden roof connection on the dynamic structural behavior. In pursuing these goals, the main challenge is considering the influence of air conditioning and elevator systems on the natural frequency identification of the structure. Therefore, an Artificial Neural Network algorithm trained on numerical model results is implemented to reconstruct the unknown dynamic load which can be considered as white noise smeared in some intervals by external disturbances (colored noise). By this approach, the network should be able to recognize the time ranges in which are operative the external machines.