Pattern Compensation for Faulty Phased Array Antenna Based on Deep-Learning Technique

Abstract

This study proposes an approach to compensate for pattern distortion in a phased array antenna caused by antenna element failures. The proposed approach utilizes a deep-learning network explicitly trained for a phased array antenna with damaged elements to generate the necessary excitation, producing a new pattern closely resembling the intact phased array antenna. Compared to alternative methods that focus on reducing side-lobe level, this compensation approach offers the advantages of rapid response and minimal computational overhead for the re-synthesis of the desired pattern that is close to the original pattern. This approach makes it particularly suitable for scenarios involving faulty phased array antennas, such as those on satellites or mountain-top antenna towers, where replacement or repair is not readily feasible in a short timeframe. This study demonstrates the pattern compensation for the two phased array antennas with damaged antenna elements. This work analyzes several randomly selected patterns and proposes quantitative indices to evaluate the performance of the approach. The proposed approach produced the compensating excitations of the remaining undamaged elements within 0.1 sec after inputting the desired pattern. The simulated results indicate that the proposed method effectively reduces pattern distortion resulting from antenna element failures and thus regenerates an optimal pattern as close as possible to the original one.