Application of surrogate modeling method in hybrid simulation for uncertainty propagation

Hybrid simulation combines physically tested experimental substructures and numerically modeled numerical substructures that works as an effective solution for structural response prediction under earthquake excitations. The uncertainties from numerical and experimental substructures can lead to large variance on the system response, and it raise the challenge of using full potential of this advanced simulation technique. In this study, Particle Swarm Optimization, Markov Chain Monte Carlo simulation and surrogate modeling techniques are applied for uncertainty propagation. Also, a procedure of experiment design o f hybrid simulation is proposed. With the presentation of proposed procedure, good estimation on the uncertainty o f system response can be reached with a relatively small number of experiments.