Analysis of Electromagnetic Systems Using the Antenna Current Green’s function (ACGF) and Machine Learning

Abstract

Mutual coupling in antenna arrays has been a problem in the development and operation of electromagnetic systems. The increasing demand for integrating more radiating systems in small packages has motivated researchers to investigate advanced means to mitigate the effect of mutual coupling by inspecting the primary cause of this phenomenon. It was observed that the recently developed formalism of the antenna current Green’s function (ACGF) can provide a characterization of the electromagnetic radiation and the system’s physical dimensions. It represents radiators using a spatial transfer function due to a spatial impulse excitation. The ACGF is being used in this thesis to model radiating structures namely antennas in order to form their transfer function either as a single radiator or in array configurations. Moreover, a new mutual coupling interaction function MC-ACGF is developed to characterize mutual coupling.

A natural extension of the approach is to extend it to radar target identification, which is important for military and remote sensing applications. The challenge is in obtaining the unique features of specific targets. Although many solutions were suggested for target identification, they still exhibit some limitations that are affecting the performance of radars. Therefore, this motivated us to use the above modeling approach and to establish a novel method to deal with a newly obtained RCS data of given objects. That is, the field data of an object are formulated in terms of its physical spatial features. This is done by expressing the spatial properties of targets via a newly derived spatial singularity expansion method (S-SEM), in which the surface current of radiators is going to be represented using spatial SEM data. Moreover, it turns out that the spatial-SEM leads naturally to the discovery of a new set of far-field basis functions, which we call here the spatial-SEM radiation modes. Explicit expressions for these modes are derived for the case of wire antennas with arbitrary length and orientation. The summation of the S-SEM modes appearing on an antenna surface results in forming the overall transfer function of the antenna due to a special impulse response.

In the above-described investigations, a machine learning solution is devised to enhance the performance of compensation systems in mutual coupling problems and in predicting target features in radar target identification. A multilayer perceptron artificial neural network (MLP-ANN) has been carried out to form an intelligent mitigation system for antenna arrays. On the other hand, an electromagnetic based genetic algorithm (EM-GA) is developed to search for targets’ singularities to identify their physical geometry.

The developed methodology on mutual coupling is verified by application to estimating the direction of arrival (DOA) of signals impinging on an antenna array. The evaluation is based on extracting the frequency content of the incoming signal using multiple signal classification method (MUSIC) and detects the power of the signal in the desired direction.

In inverse problems, a verification of a radar scenario is simulated and an experiment is conducted in an anechoic chamber and a good agreement with simulation is obtained.

The newly developed methodology will be useful in developing and improving future designs of antenna systems, MIMO, target identification for radars and remote sensing applications.