Investigation and Experimental Validation of Kriging-NARX Modeling for Real-Time Hybrid Simulation

This study proposes the use of nonlinear autoregressive with exogenous input (NARX) to emulate dynamic behavior of analytical substructures in real-time hybrid simulation (RTHS). The training of the NARX model can be conducted numerically, and its implementation presents a good representation of structural dynamics with minimum computational demands on the RTHS equipment. A Kriging metamodel is further introduced to surrogate the NARX model coefficients to account for ground motion uncertainties. RTHS of a single-degree-of-freedom (SDOF) structure with a self-centering viscous damper are conducted as proof of concept to experimentally demonstrate the effectiveness of proposed Kriging-NARX metamodeling approach.