Based On Modern Signal Processing Of The Structural Modal Parameter Identification And Damage Detection Research

With the development of society, kinds of civil and steel structure has been widely applied in city construction, water conservancy and hydropower, power transmission, new energy and other areas, and showed a large, complex and diversified development trend. The large-scale structure has closely relationship with life, its reliability and safety should be paid more attention, so the structure health assessment and health monitoring can not be ignored. The modal parameter identification and damage recognition are seen as a core technology for structural health monitoring, which are concerned by engineering and academia. Based on modal analysis theory and structural dynamics, the thesis analyzed the feasibility of application of some the modem signal processing methods on the field of the modal parameters identification and the damage recognition, then presented damage identification method and structural modal parameter identification base on a complex modern signal processing technology. The primary work and achievements are as follows:Method of empirical mode decomposition (EMD) is improved. In random excitation case, characteristics of single-mode response signal of different damping ratio and natural frequency indicate that the modal response signals can be simple as the slowly varying envelope of the narrowband AM signal, in the case that structural vibration is approximately linear. A new method for local mean solving is earned out based on this research. The new method weakened the ability of the EMD method to nonlinear processing, while enhancing the ability to decompose a component of the AM signal. The times of interpolation fitting of EMD is reduced by this new method, while improved the decomposition efficiency.And the mode mixing and limitation of EMD method were analyzed from density of maximum.The method of modal parameters identification based on time series analysis and stabilization diagram is presented. According to the characteristic of time series analysis, mode energy ratio was introduced to stabilization diagram as a new criterion. Modal order determination of modal analysis can be avoided in a certain sense by both methods combined. It realized the adaptive modal order determination which considers the number of stable axis of stabilization diagram as the reference, then a lot of false points are rejected, and the computation efficiency of stabilization diagram is improved significantly.According to modal analysis theory, a method of mode response extraction is introduced which based on time series model spectrum, cross-correlation detection and shifting process of EMD. A five degree of freedom linear simulation of vibration systems was used as application object for this mode response extraction method, and mode response was extracted successfully for this simulation model. The modal parameters of simulation model were identified by using analysis of AR model.Damage identification based on model energy feature change and BP neural network is studied. Intrinsic mode functions (IMFs) are got from vibration response by using EMD method, and model energy feature of IMFs is computed. The damage identification method is applied to a element model of communications tower, which based on model energy feature change and BP neural network, and the method is proved effective.