EXPERIMENTAL STUDY ON NONLINEAR AUTOREGRESSIVE WITH EXOGENOUS INPUT FOR REAL-TIME HYBRID SIMULATION

In traditional real-time hybrid simulation (RTHS), numerical substructures are usually calculated by finite element method; as a result, it is difficult to complete the calculation for complex structures within specified integral step time. In order to improve computational efficiency, a RTHS method based on nonlinear autoregressive with exogenous input (NARX) is proposed. The NARX model is trained through numerical simulation to replace numerical substructure. RTHS of a nonlinear single-degree-of-freedom structure with a self-centering damper is conducted for proof-of-concept. RTHS tests show that results of NARX based RTHS are almost the same as those of traditional RTHS. Thus it is indicated that NARX based RTHS is a potential alternative for traditional RTHS.