A Fast Pattern Synthesis Method for Conformal Antenna Array Based on Conbined Convex Optimization and Neural Network

Different to the beam scanning of planar phased antenna array, the beam shape and sidelobes of conformal antenna array will be distorted due to the nonuniform coupling and anisotropic radiation patterns of each antenna unit. The simple linear phase shifting strategy for the planar array is invalid. The amplitude and phase distribution of the conformal array need to be re-optimized at every scanning angle without analytical formula, greatly increasing the calculation complexity and time cost, and it is not suitable for the real-time beam controlling. In this paper, a fast pattern synthesis method for the conformal antenna array based on the combinated convex optimization and neural network is proposed. Firstly, the problem of maximizing the directivity coefficient with sidelobe level(SLL) constraint is expressed as a standard convex model. The amplitude and phase distribution of the array in a specified scanning angle can be obtained by convex optimization method, as the training dataset for the following neural network training. Secondly, using scanning angles as input and amplitude/phase distributions as output, the trained neural network can grasp the implicit scanning rules for the conformal array, and it can quickly export the amplitude/phase distribution at arbitrary scanning angle. The efficiency of the proposed pattern synthesis method is also verified by typical sample simulations and the comparison of other different methods.