Modeling of Ground Penetrating Radar System based on Recurrent Neural Network Architecture

The finite-difference time-domain (FDTD) method is the most commonly used method in electromagnetic simulations due to its advantages of easy implementation and stabilization. However, in the ground penetrating radar (GPR) simulation, it requires performing multiple FDTD calculations on the same physical model according to the number of GPR emitters, which brings heavy calculation burden, making it difficult to implement in practical application. This paper proposes a GPR simulation method based on the recurrent neural network (RNN) architecture. The new method reconstructs the FDTD algorithm through deep learning theory and operators without training. The proposed method can maintain the same accuracy as FDTD while speeding up the simulation.