Gate Recurrent Unit to Accelerate Radar Cross Section Calculations for Electrically Large-sized Targets

Calculating the Radar Cross Section (RCS) of an electrically large-sized target is an important step in EMC design. In this process, an open-source ship scaling model was selected as the target for low RCS optimization and design. A dataset was established to train a Gate Recurrent Unit (GRU) to perform low RCS optimization and design for this electrically large-sized target. The simulated annealing method was chosen to update the physical model of the ship scaling model during the low RCS optimal design process. The physical model and corresponding RCS values were accumulated during the updating process to establish a dataset. During the low RCS optimization and design process of this electric large-size target, a multilayer GRU neural network is used to learn the dataset of the target object over time. This allows the trained neural network to calculate the RCS of the target object instead of relying on traditional electromagnetic simulation computation. The multilayer GRU neural network's properties are used to decrease computational requirements and increase speed.