A Physics-Embedded Deep Learning Method for Pattern Recovery of Phased Array Antenna With Deformation

A physics-embedded deep learning method is proposed for radiation pattern recovery of phased array antennas with deformation. The proposed method primarily consists of two components: a displacement prediction network and a multi-angle pattern recovery network. The displacement prediction network takes data from pressure sensors surrounding the phased array antenna elements as input and yields the displacements around each geometric center. These displacement values are then employed to calculate the deflection angle of each antenna element. Subsequently, the multi-angle pattern recovery network takes the deflection angles as inputs and outputs the amplitude and phase of the excitation for each element. Following that, these amplitudes and phases are utilized to excite the deformed phased array antennas for the purpose of recovering the radiation pattern. Experimental results validate the effectiveness of the proposed method. It achieves a beam pointing error on the order of 10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-3</sup> degrees and an average response time of approximately 2.784 milliseconds, which demonstrates its efficiency and potential for real-time applications.