Research On Mode Decomposition Based Methods For Structure Parameter Identification Of Time-varying Mode And Nonlinearity

Hypersonic vehicles will show obvious time-varying dynamic characteristics when flying at high speed,so the control system needs to use real-time dynamic characteristic parameters to achieve optimal control.However,this kind of vehicle has complex structures,so it is not easy to provide its accurate time-varying dynamic characteristic parameters through a fine modelling method.Consequently,the online identification method has more engineering application value.Although many time-domain methods can be used for online modal parameter identification,their development is restricted by the problems of difficult model order determination,large amounts of computation,and limited tracking efficiency.Conversely,although frequency-domain methods have higher computing efficiency,few of them are suitable for online identification,especially recursive methods.Hypersonic vehicles also need to bear large loads when flying at high speed.At this time,local assembly gaps will induce significant non-linear characteristics,which will seriously affect the flutter characteristics of the hypersonic vehicle structure.Therefore,accurate local non-linear parameter identification is of great significance to study the flutter characteristics of these structures.Among the methods suitable for local non-linear parameter identification,almost all the methods that can provide parametric models need to measure the relevant response of the non-linear description function,limiting their practical application.To sum up,this paper researches the identification of time-varying structural modal and local non-linear parameters.The main research contents are as follows:(1)Aiming at the problem that the original univariate(single-channel)variational mode decomposition needs channel-by-channel decomposition and manual interaction to realize mode alignment when analyzing multi-sensor vibration signals,a non-parametric frequency-domain modal parameter identification method is developed based on the latest multivariate variational mode decomposition.Moreover,another parameterized frequency-domain method is also established based on the modal superposition principle and the variational mode decomposition framework.The simulation results show that the two new methods inherit the ability of the original univariate method in dense mode decomposition and time-varying structure modal parameter identification,and the identification effect is better than the univariate method.The practical application also demonstrates the effectiveness of the two new methods.(2)Aiming at the problem that variational mode decomposition methods need to predefine the number of modes to be decomposed and pre-estimate the mode center frequency,two novel methods that can successively extract modal parameters from multisensor vibration signals are developed based on the multichannel signal variational mode decomposition methods in(1).Same simulation examples as in(1)show that the two successive improvements inherit the ability of the original method in dense mode decomposition and time-varying structure modal parameter identification and improve the convergence as well.In addition,the practical application shows that the two successive improvements can use output-only response data or frequency-response data to identify modal parameters of time-invariant structures.(3)Aiming at the problem that the aforementioned variational mode decomposition methods are all based on the narrow-band mode assumption,not suitable to analyze long-term non-stationary vibration signals with wideband or overlapped modes,and only suitable for the off-line identification,two online recursive identification methods of timevarying structures are developed based on the short-term stationary assumption and the multichannel signal variational mode decomposition methods in(1).Simulation and experimental examples show that the two recursive improvements have the ability of online identification and can continuously track the modal parameters of time-varying structures.(4)Aiming at the problem that the existing local non-linear identification methods,which can provide parametric models,almost all of them require the relevant responses of the non-linear description function are measurable,a backbone curve fitting and adaptive chirp mode decomposition based method for the clearance non-linear parameter identification of the folding rudder structure is proposed,and this method is applicable when the relevant responses of the local non-linear description function are unmeasurable.The proposed method is verified by a numerical folding rudder structure and an actual folding rudder structure.