A Fast Radiation Performance Prediction Method for Reconfigurable Transmitarray Based on BP Neural Network

Traditional full-wave simulation method and electromagnetic calculation method based on ray tracing method (RTM) have the huge defect of consuming computing time and hardware resources. In this paper, a fast radiation performance prediction method for reconfigurable transmitarray (RTA) based on BP neural network (BPNN) is proposed. The discrete phase distribution of the RTA is used as the input of the dataset, and the gain and sidelobe level of the corresponding E-plane and H-plane of the pattern as the output. Randomized phase quantization method-appropriate two probable value (A2PV) method and pattern calculation tool based on RTM are used to generate dataset for training. Reference to calculated results, the trained network outputs precise gain and sidelobe results of E-and H- planes of the RTA under different random phase distributions. The proposed method has a huge speed advantage, which can realize the real-time output of the pattern parameters.