Identification And Response Prediction Of Dynamic Modeling For Car Body With SCLD

Smart structure has a broad prospect of application to active control of structuralvibration and noise because of its advantages of fast reaction self-adaptingself-diagnosing and self-recovery. And using smart constrained layer damping (SCLD)technology to achieve structural vibration control has become one of the hot points athome and abroad in recent years. The establishment of accurate dynamic model is thebasis for the active vibration control of structures with SCLD treatment. At present,there’s modeling research for simple SCLD structure with regular shape by analyticalmethod or numerical method that based on structural assumptions and simplification.But there still exists considerable problem for the modeling of more complex SCLDstructure. Experimental method based on system identification provides an efficient wayto modeling of the complex systems that without clear internal structure andcharacteristics. Considering the excellent performance of dynamic neural network interms of model identification for nonlinear complex systems model, the thesis exploresthe use of this method to achieve dynamic model identification of the SCLD car body,and thus lay the foundation for structure active vibration and noise controller design ofSCLD car body.First of all，for the identification model based on NARX neural network, the mainfactors that influence the neural network generalization capability and neural networktopology design method are discussed. On the basis of analyzing theoretical basis anddesign criteria of neural network architecture, correlation-based pruning algorithm ispresented for the topology optimization of neural network. And numerical examplevalidates the superiority of the algorithm.Then regarding the steel plate with SCLD treatment as the object of study, theNARX network is selected as the identification model, and a combination ofseries-parallel and parallel method is used for training network. According to theexperimental data, the modeling and response prediction of external disturbance channeland control channel that characterizing system dynamic performance is developed. Thechannel modeling is established in the case of single sine, complex periodic sine andrandom signal excitation respectively. And the results prove that the NARX networkcan establish the dynamic model of the SCLD plate quickly and accurately. Modelspectrum and prediction analysis indicates that the identification model can characterize the dynamic performance of SCLD plate, so the model can be directly used forcontroller design.Finally, according to the above conclusions, the NARX network is used for thedynamic model identification and response prediction of the SCLD body. The dynamicmodel of external disturbance channel and control channel is identified under differentexcitation conditions, and achieves satisfactory results. The model identification of theactual SCLD car body structure under specific engine speed is studied, and alsoachieves satisfactory modeling results. The studies above show that the identificationmodel of SCLD body can obtain the main system dynamic characteristics and providereference to further study for active control of SCLD body vibration and noise.