Anomaly Correction for Radiation Patterns of Active Phased Array Antenna Based on a Neural Network

Anomaly interference in the radiation pattern of an active phased array antenna (APAA) is an intractable problem, that affects the overall system performance. To overcome this problem, this paper proposes an antenna radiation pattern anomaly correction model (ARPACM) based on a neural network to compensate for the shortcomings of the antenna radiation pattern (ARP) under anomalous conditions. In addition, to overcome the problem of scarce data, this study uses simulated anomalous ARPs that are consistent with practical experimental conditions to construct and train a pre-trained model, named simulated ARPACM (SARPACM). The weights of the last hidden layer of the SARPACM are optimized using experimental data by introducing transfer learning, yielding an optimized ARPACM. Extensive experiments are conducted to verify the proposed ARPACM. The results indicate that the proposed model can outperform widely used correction algorithms and other training strategies, in terms of a mean coefficient of determination ($\text{R}^{2}$), a mean square error (MSE), a mean relative error (MRE), achieving improvements in 99.76%, 0.08 dB, and 1.29%, respectively. Finally, the experimental results demonstrate that the ARPACM can effectively correct the anomalies of ARPs caused by various interfering factors.