Abstract:
Neural network based analysis and design methods of stacked patch antenna have been developed in this dissertation. One of the methods of overcoming the, biggest drawbacks of classical microstrip antennas viz narrow bandwidth is to use a stacked microstrip antenna. With the increase in number of layers and radiating patches, the complexity of the analytical/numerical techniques used to analyze these structures also increases. In order to avoid the same, neural network based black-box modeling approach is developed in the present work.
Besides fast response time, the other benefit of using neural networks is that it produces results that are accurate at par with the data used for its training. Two different neural networks have been developed one for analysis and other for the design of dual patch stacked microstrip antennas in the X/Ku band. The job the analysis network is to locate the operational frequencies of the antenna for specific design parameters and the job of the design network is to find the sizes of the radiating patches for the antenna to resonant at the specific frequencies. The training data for the network are generated using IE3D simulation based on method of moment's techniques. The validity of the networks was tested with test data, different from training data and with experiments as well. The developed neural models can be used as CAD packages for analysis and design of stacked patch microstrip antenna.
