Açıklık kuplajlı mikroşerit yama antenler için yapay sinir ağ modeli

Recently, in parallel with the development of technology, attention on wireless comunication provides rapid progress in antenna technology. One of the wireless communication tools is microstrip patch antennas (MPA) that is used in mobile applications and space vehicles in general. Heavy demand on personal portable device further increases the significance of MPA. MPAs, due to its small size and sharing the same dielectric layer with circuit members are harmonized easily with integrated circuits. Neverthless, narrow bandwidth, high loss in feeder circuit, low cross polarization and low power control capacity are the main weaknesses of the basic MPAs. Previous researches and studies show that, most of the aforementioned disadvantages might be removed or reduced by means of making various extensions on the basic MPA units. MPAs are utilized efficently on various system applications including wireless and satellite communication, biomedical irradiator, environmental instruments and remote sensing systems. Number of these applications will be raised with the parallel in development of technology. Defining resonant frequency is crucial issue since MPAs operate at narrower bandwidth than other antennas. Parameters that influence the resonant frequency of MPA as follows; thickness of used dielectric material, dielectric constant, size of ground surface thickness and width of conductive patch. In this study, Aperture Coupled Microstrip Antennas (ACMA) extended from MPA class are investigated. Note that, ACMA is fed microstrip line which has relatively higher bandwidth compared to other microstrip patch antennas. ACMA prototype is used throughout the study. It is prepared via High Frequency Structure Simulator (HFSS) software. HFSS software is a high performance full-wave electromagnetic simulator and has an effective graphical user interface. HFSS software is basically providing reference data. Simulation of the HFSS software package can remove the excessive cost during the fabrication and provides positive contribution on the results during the produciton stage. On the other hand, this model has high learning curve and fairly low physical anlaysis capability. In order to obtain desired parameters of antenna, simulation programs generates results in a long period of time due to its heavy computation load and complex analytical algorithm behind. Therefore as an alternative to the HFSS, new computer aided methods should be investigated. One of these methods is the Artificial Neural Network (ANN). Learning ability, rapid applicability on various problems, generalization capability, requiring less information, fast and easily processing power make ANN popular in recent years for this particular problem. According to many studies, ANN can address the chalenging problems particularly resonant frequency which are actually quite complex and time consuming processes. In this study, producing desired parameters in the range of 1 - 3.5 GHz for ACMA, eligible ANN model was developed. Outputs of the developed ANN model were evaluated and then compared to HFSS simulation software results. It was observed that our proposed method is more efficient (100 times faster than HFSS software) and has acceptable accuracy rate (96.5 %) with respect to the reference HFSS model.