Based On Neural Networks Emi Structural Health Monitoring And Damage Detection Research

Engineering structures serving in complicated environment often suffer different kinds of damages, and the failure will occur in the structure, evenly result in catastrophes because of the accumulation of such damages, thereby, health diagnosis and safety appraisal are most crucial for engineering structures. Therefore, it has important theoretical and realistic meaning and social and economic value to carry out research in the health monitoring and damage diagnosis technology. Supported by the National Natural Science Foundation of China (No.50778179), the thesis has studied structure online health monitoring based on the integration technology of piezoelectric driving and sensing.Firstly, structural damage detection based on vibration test are discussed in this thesis. The new method using the electro-mechanical impedance (EMI) method to detect the structural damage is also introduced. And research progress of structural damage detection based on computational intelligence at home and abroad are summarized. The basic theory of the application of the PZT impedance method in structural damage identification is analyzed which includes the notion and feature of the intelligent material, the application of PZT in the intelligent structure, the basic theory of piezo-electricity vibration and the deduction the relation between PZT and the structure coupling electrical impedance. Then, the effectiveness of the proposed EMI method used in the structural damage detection has been verified by the damage detection experiments of steel beams with fixed support conditions. the damage magnitudes of the steel beam in different damage states are quantitatively identified by using a damage condition factor.The basic theory of neural network, the BP neural network model and designing, the factors,the process and the basic principle of neural networks used in structural damage identification are introduced. A cantilever beam model is established by simulation. And using BP neural network identify the damage location. The results reveal that the BP neural network for single damage identification.Study on damage detection is also presented using electro-mechanical impedance (EMI) signatures and artificial neural networks (ANNs). Piezo-electricity admittance spectrum curves obtained by experiment were used as input parameters of neural networks. Then, structural damage identification are researched by the training of input parameters. The experimental results reveal that the designed BP neural network can identify the damage successfully. The hybrid technique, fully making use of the high-frequency EMI signatures and neural network features, is a potential method of detecting damages in structures.Finally, some results are summarized in this project with prospects for the further research.