Hybrid Simulation of a Frame Equipped with MR Damper by Utilizing Least Square Support Vector Machine

In hybrid simulation, the structure is divided into numerical and physical substructures to achieve more accurate responses in comparison to a full computational analysis. As a consequence of the lack of test facilities and actuators, and the budget limitation, only a few substructures can be modeled experimentally, whereas the others have to be modeled numerically. In this paper, a new hybrid simulation has been introduced utilizing Least Square Support Vector Machine (LS-SVM) instead of physical substructures. With the concept of overcoming the hybrid simulation constraints, the LS-SVM is utilized as an alternative to the rate-dependent physical substructure. A set of reference data is extracted from appropriate test (neumerical test) as the input-output data for training LS-SVM. Subsequently, the trained LS-SVM performs the role of experimental substructures in the proposed hybrid simulation. One-story steel frame equipped with Magneto-Rheological (MR) dampers is analyzed to examine the ability of LS-SVM model. The proposed hybrid simulation verified by some numerical examples and  results demonstrate the capability and accuracy of  this new hybrid simulation.