Antenna Optimization Method Based on Multi-Knowledge Embedded Artificial Neural Network for Ultra-wideband Phased Array

In this paper, an antenna optimization method for ultra-wideband phased array antennas, based on multi-knowledge embedded artificial neural network (ANN), is presented. The array is based on the tightly coupled dipole array (TCDA) antenna, which has a large number of antenna parameters requiring optimized. When considering the beam scanning requirement, it turns into a multi-parameter and multi-objective optimization problem. To solve this problem, the optimization method based on the multi-knowledge embedded ANN is proposed. It transforms the complex optimization model into two simple sub-models embedded with the prior knowledge models. In Sub-Model-1, the S-parameter transfer function of the feedline is regarded as the prior knowledge, while in Sub-Model-2, the equivalent circuit model of the tightly coupled dipole and wide-angle impedance matching layers is regarded as the prior knowledge. An error matrix is introduced to correct the inaccuracies arising from the cascading of sub-models. This is equivalent to transforming a high-dimensional problem into two lower-dimensional problems, which reduces the sample space size by several orders of magnitude and enhances the effectiveness of the full frequency band and key frequency band optimization of the structurally complex TCDA antenna.