An ANN-ANN Iterative Optimization Algorithm for Pattern Synthesis of Arbitrary Conformal Arrays

In this paper, an iterative algorithm based on artificial neural network (ANN) is proposed for the accurate and efficient far-field synthesis of conformal arrays. Inspired by the iterative fast Fourier transform (IFFT) technique, two ANNs with different structures are trained and integrated in an iterative process. The former ANN is designed to predict the amplitude and phase of the excitation by inputting the pattern of conformal array. The latter ANN is designed to calculate the far-field pattern by inputting the excitation. Starting with a random excitation, the real and imaginary parts of the far-field pattern are calculated by the latter ANN. The actual pattern predicted by the latter ANN is then corrected under the given constraints and utilized as input to the former ANN to predict the reconstructed excitation. After several iterations of this process, the synthesis of the pattern with arbitrary constained can be rapidly realized. Two numerical examples, including the synthesis of a shaped beam and a flat-top beam of a circular-ring array, have been successfully conducted. It is worthwhile to be noted that the proposed algorithm can also take into consideration for the the mutual coupling and the edge effects of arbitrary real conformal arrays.