System identification is a simple and effective method for nonlinear modeling of the smart constrained layer damping (SCLD) structures. NARX network is presented to characterize the model that to be identified, combining series parallel and parallel structures for the training of the network. Then the dynamic model of nonlinear system is achieved according to the experimental data. The SCLD plate structure is studied as the example, and the models are identified for the disturbance channel and control channel respectively. The results prove good identification performance of NARX network and that the approach employed is correct. To further validate the feasibility and effectiveness of the method, NARX network is used to model the complex car body with SCLD treatments and satisfactory results are obtained.