Deep learning techniques developed in the fourth industrial revolution are used in various research fields. In addition, electrical resistivity exploration is a technique that can achieve ground engineering property values and is utilized in combinati...
Deep learning techniques developed in the fourth industrial revolution are used in various research fields. In addition, electrical resistivity exploration is a technique that can achieve ground engineering property values and is utilized in combination with the two fields. The objectives of this study are Deep Neural Network (DNN), Long-Short Term Memory (LSTM), and Gated Recurrent Unit (GRU), which utilize the combinatorial structures LSTM-DNN and GRU-DNN to predict electrical resistivity and ground behavior. For data collection, a test bed was installed at the top of the mountain on the slope for the exploration of electrical non-resistance, and the total length of the side lines and electrode spacing are 200m and 2m respectively. Electrode arrangements utilize a winner arrangement suitable for vertical resolution. The electrical resistivity values were measured for a total of 16 months from January 2019 to April 2020, with 52,800 electrical non-resistance data values. The data preprocessing process utilized min-max scaling techniques, and hyper-parameters were set in model generation. Finally, reliability was evaluated using RMSE, MAE, and R2 values in model verification. Finally, an electrical resistivity prediction model was developed, and the ground engineering property behavior of the slope was evaluated by converting the predicted electrical resistivity into the gap rate and the pitching coefficient.