Microwave Detection Of Concrete Structures Simulation And Inversion

Microwave non-destructive testing technology is emerging to develop a high-frequency electromagnetic non-destructive testing technology, has broad applicationprospects. This technique is based on measuring the surface high-frequencyelectromagnetic field to detect internal defects or structural distribution.This paper first describes the principle of microwave nondestructive testingtechniques, characteristics, research status and development trends. Microwavepropagation in concrete structures on the basis of electromagnetic theory, based on thedielectric properties of steel and concrete, reinforcing steel and the laws that governthe reflection and scattering. Relatively high frequency for solving the inverseproblem of a variety of numerical calculation of the characteristics of the method,final choice of the finite element method, the inversion algorithm of the method ofmoments and neural network research issues in the subject.Based on the finite element algorithm theory, through the establishment ofsuitable for reinforced concrete structure high frequency analysis of3D model of thedesignated area, the applied load and boundary conditions is solved, and the resultsvisually shows the steel bar in the concrete level distribution, subsequent analysis ofreinforced position and the scattering field distribution relationship. At the same timefor different position in steel simulation, obtained the corresponding signalcharacteristics, established the application of neural network to the inversion database.Subsequently, the subject design high-frequency electromagnetic field simulationanalysis based on the method of moments test, model and grid divided according tothe graphical user interface, use the card system of the FEKO software to set thenecessary parameters and view the results through the post processor and show thatthe scattering field distribution. Ultimately the results with ANSYS simulation results,the curve trend is more consistent, realistic testing needs.The neural network has highly nonlinear parallel processing of the informationability, full consideration of the reinforced concrete structure, position and thescattering field of complicated nonlinear relation between, the introduction of used forstructure inversion BP neural network and generalized regression neural networkalgorithm, and realize the depth of the reinforced accurate prediction. The BP neuralnetwork select the gradient descent momentum method,the Levenberg-Marquadt backpropagation algorithm and quantify the conjugate gradient algorithm for training, it modeling simple convergence speed in a known samples under the circumstances ofless able to show strong adaptability, and the accuracy can reach more than80%.generalized regression neural network parameters for less regulation, approximationhas obvious effect, retrieval accuracy up to90%. Both in reinforced locationquantitative prediction has the very good extension.