The Research On Wavelet And Neural Network Method For Damage Identification Of Frame Structure

In the process of using,structure affected by action of the external load and aging of materials,which are able to cause cumulative damage and resistance to decay,even leading to overall structural is destroyed and catastrophic accidents.Therefore,the research of structural damage identification methods not only has great theory significance,but also has important practical significance.Wavelet analysis can represent partial characteristics of signals in time and frequency domain,and the BP neural network algorithm possesses numerous characteristics, for example highly nonlinear mapping ability,the capacity of self-organizing and the capacity of parallel processing etc.According to the damage of structure will change the dynamic characteristics of the structure,which is able to identify the damage by the nonlinear relationship between dynamic damage and structural.The strain parameters with high sensitivity to subtle damage of local structure,which is making it for ideal parameter of structural damage diagnosis.A frame structure is regarded as the object of research in the thesis,and the method is based on wavelet analysis and neural network theory,which is combines the advantages of them,applying wavelet analysis to determined the location of damage of frame structure,and using the neural network method to identify the damage degree. Getting the principle of strain mode parameter identification of damaged frame structure and establishing a wavelet neural method for which is to identify damage structure.This method can effectively identify frame structure which is damaged location and damaged of injury.A simple frame structure is regarded as the object of research in the thesis,using wavelet neural network method is proposed in this thesis,according to the different ratio of fracture length and the height of cross section(d/h) to simulate the damage and finite element model is established in thesis. Through the establishment of damage identification method of vibration mode and strain made based on the position of the crack,for the simple framework with multiple injuries were effectively identified and compared these two kinds of modes of damage location identification.Then the vibration parameters and strain parameter is using continuous wavelet transform which is to obtain the two maximum mode parameters.Applying neural network to identify the degree of structures which is simulate thenonlinear relationship between the maximum of wavelet coefficients of the structure and the damage degree,and it compared the damaged identification of two mode parameters.The numerical analysis indicates that,the wavelet neural network can identify the damage location and degree of structure effectively,and the accuracy of damage identification method which bases on strain mode is better.The thesis regard the three kinds of frame structure with damage as research object,which are one layer two cross,two layer one cross,two layer two cross,establishing the finite element model of frame structure with two place and multiple places damage.Based on strain mode, the work situation of all damages is dealt with wavelet transformation. The damage location of structures can be detected by the coefficient figure of wavelets. the neural network can be formed with input parameter as the maximum of wavelet coefficients of the structure base on strain mode.The method is verified useful by which the degree of the damage can be identified by output parameter.The method in the thesis can be useful for reference in engineering application of identification and diagnosis for the damaged structure.