Research Of Pyramid Space Truss Damage Detection Method Based On Wavelet Packet Energy-Neural Network

With increment of national economic and rapid development of science and technology, the proportion of the infrastructure construction is much than before in the national production, a large number of long-span spatial structures, large hydro-structures and nuclear power station are emerging. The trend of structure towards large-scale and complication, and theirs'life cycle could be for hundreds years. As one of commonly used spatial structure, truss structure is widely used in civil engineering fields. Because of the impact of environment erosion and other factors, the problem such as accumulated fatigue, resistance degradation and material aging must arise in the truss structure. The structure damage appears in the usage period and structure performance declines and it maybe result in global failure in the structure when the damage has been accumulated to some degrees. So the security of the truss structure should be taken seriously. With regard to security of the truss structure, its detection and diagnosis has become the recent problem to be resolved and future research direction. Structure health detection is the leading direction in civil engineering fields. Health monitoring on major buildings has been the future leading direction of the civil engineering fields. Structure damage detection is the foundation and key content of the structure health detection. Therefore, considering the importance of the truss structure, it is necessary for the detection and diagnosis on truss structure to prevent accident catastrophic accident and economic losses.In short, the findings and conclusion in this paper go as follows:1) Analyze the steel cantilever structure damage with finite element software ANSYS. First, using transient incentive and extract useful information, build index parameters of structural damage (wavelet packet nodal energy and wavelet packet energy changing rate). Then, choose the wavelet packet function, make the transformation of wavelet packet on damage parameters, and finally identify the damage location of the structure. To guarantee the data veracity of simulations, dynamic modal test to the full scale model of the steel cantilever beam is made in the lab, and correspondent data should be compared with the simulations data to test the accuracy of the numerical simulations.2) During the detection on the actual project, useful signals from spatial truss dynamic detection could always be polluted by noise. During the service period of spatial truss, because of its complex boundary conditions, large weight and damping, and insufficiency of inspiration response, the quality of the dynamic detection signal decreases. This paper on the basis of the research achievements of other people in this field, makes the study on the method of combination between wavelet threshold and MATLAB software to do simulation research of the signal de-noising, for the sake of the accuracy of the extracted signals.3) Analyze the orthogonal pyramid spatial truss damage with finite element software ANSYS. Use transient excitation to conduct dynamical modal numerical simulation about intact structure and damage structural, and extract the acceleration signal of every measuring point to construct wavelet packet energy changing rate as damage index. An artificial neural network method compared with wavelet packet energy will be applied for two steps damage identification of spatial truss structure. Firstly, damaged location will be determined by the wavelet packet energy changing rate; secondly, damage degree will be determined by radial basis function neural network with inputting of wavelet packet energy changing rate. Using two steps damage identification method to discern the damage with neural network can effectively reduce calculated quantity of inputting specimen, and reduced the network training time, and then the network recognition precision can be improved.