Abstract: According to the problem of the positioning accuracy and lack of inversion iterative efficiency in the microwave detection of reinforced concrete structures,BP neural network and generalized regression neural network were introduced into the microwave detection technology.Simulating analysis was based on ANSYS software,and the signal characteristics for inversing the reinforced concrete structure were extracted.Neural network models of BP neural network end GRNN neural network were built,which were suitable for quantitative analysis of reinforced concrete structure.By using the results normalized as a test sample of neural network,the information storage was completed.The intrinsic link between the output and input was found and the prediction of the reinforcing bars was completed.The results indicate that the two models can reflect the position of steel bar better.However the precision and predominance of GRNN neural network is better than that of BP model.