Research on the identification and simulation of the earth-rock dam leakage inversion based on a high-density electrical method

In this study, the earth electricity model of the earth-rock dam with leakage hazards is established, and the scale, quantity and spacing of leakage hazards are changed to simulate various leakage conditions. A dipole-dipole device was selected to detect the model, and Res2Dmod and Res2Dinv software were used to perform forward calculation and inverse imaging respectively. The law was summarized according to the inversion results, and the detection ability of the device for leakage regions of different sizes, quantities and distances was analyzed. Finally, convolutional neural network is introduced in the Matlab to train and test the inversion samples, and the discrimination model of leakage hidden danger is obtained, which improves the research and judgment ability of inversion results. The main conclusions of this study are as follows: With the increase of the size of single leakage channel, the inversion results are closer to the preset model. As the distance between the two leakage channels decreases, the low resistance region in the inversion results tends to be coupled. The convolutional neural network can accurately identify and classify the inversion results, and the prediction accuracy is 94.4%.