Pattern Synthesis of A 2 Element Patch Antenna Arrray Using Neural Networks with Tensorflow

This The growing need for highly directive and reconfigurable antennas in contemporary communication systems has fueled significant research in antenna array synthesis. This study introduces a new approach to antenna array synthesis employing a neural network model built in TensorFlow. This model predicts optimal parameter combinations to achieve desired radiation patterns at an operating frequency of 2.4 GHz. Utilizing a 3-layer architecture and incorporating ReLU activation functions in the hidden layers, the model effectively manages non-linear relationships within the data. Training is conducted on simulation data acquired from a 2xl patch antenna array designed in ANSYS HFSS software. The antenna elements are positioned at a distance of 0.28λ to examine the influence of deep learning on closely coupled arrays. Results confirm the model's accuracy in predicting optimal phase combinations for both validation and custom radiation patterns, emphasizing its potential to simplify antenna design and optimization. This research offers valuable insights into the use of machine learning in antenna array synthesis, advancing the development of high-performance antenna systems for future communication technologies.