Study On Some Precursory Theorises And Their Applications To Forceidentification

It is generally very difficult to measure excitation in operating environment in the case of the complexity of the structure or the variant forms of the excitations, sometimes it is even immeasurable for the impact location is usually inaccessible. Nevertheless, certain theories and techniques were developed for force identification or so-called force reconstruction. Aiming at the nonstationary characteristic of the dynamic loads and the relation between the responses and the frequency response functions, by use of the responses only, some novel theories are conducted to develop force identification and impact location approaches in this dissertation.The modal-model-based method for force identification and the error propagation of the method are analyzed, and an approach is presented based on the modal model within a general domain. By exciting at the other sites of the structure rather than the actual input points which are usually inaccessible, the modal parameters of the system and the dynamic loads are estimated through computations under both modal and physical coordinates. The errors of identification caused by the usual modification of the structure or boundary conditions are consequently evitable. The kernel for the development of impact identification techniques and the means for error elimination are proposed.On the basis of the modal model theory and the conception of spatial filtering which is derived from the field of control, a more precise force identification model is developed based on the discrete modal filter. The error accumulation caused by the errors of modal vectors is banished and an emendation to the classical modal model is conducted, due to the extraction of the independent contribution of each mode by means of the orthogonality of the modal filter. The computation size is reduced with the technique as well.In the light of the applications of time-frequency analysis hi many cases of the signal processing, the possibility of constructing a force identification model by wavelet transfer is discussed. By means of the wavelet transformation the deconvolution is done in the time domain. The dispersive properties of the elastic wave within the impacts acted are analyzed, and with the nonlinear optimization method concerned the impact location is identified. On the other hand, by the p-norm based fuzzy inference neural network as proposed in this dissertation, a wheel-rail coupled nonlinear system identification is addressed.Utilizing the force identification approaches above mentioned, experiments on the model of a supersonic fighter and on a wheel-rail coupled system are performed respectively, in order to verify the validity of the proposed theories in this dissertation. The results show that unknown inputs as well as their locations can be identified feasibly by these developed methods.This project is sponsored by the Aeronautics Prospective Research Fund of the National "Ninth Five-year Plan" of China.