Nonlinear time series model identification and its application in structural damage detection

The similarities between the general expression for the linear and nonlinear auto-regressive model with exogenous inputs (GNARX) and Volterra series model is compared. The frequency domain of GNARX model is deduced. On the basis, as nonlinear output frequency response function and modified Akaike information criterion are combined to determine the model order and memory steps, a novel method of structure identification for GNARX model is proposed. With data simulation, the feasibility and effectiveness of the method is verified. Finally, GNARX model together with the proposed structure identification method is applied to structural damage detection for steel plate. The results show that the effect of structural damage detection of GNARX model is better than those of AR, ARX, GNAR models, which indicates the superiority of GNARX model applied to structural damage detection.