نوع مقاله : مقاله پژوهشی
نویسندگان
1 گروه مهندسی برق، واحد بیضا، دانشگاه آزاد اسلامی، بیضا، ایران
2 گروه مهندسی کامپیوتر، واحد بیضا، دانشگاه آزاد اسلامی، بیضا، ایران.
3 گروه مهندسی کامپیوتر، واحد شیراز، دانشگاه آزاد اسلامی، شیراز، ایران.
4 گروه مهندسی عمران، واحد بیضا، دانشگاه آزاد اسلامی، بیضا، ایران.
کلیدواژهها
عنوان مقاله English
نویسندگان English
In this study, a novel inverse design method for microstrip antenna design is proposed that combines convolutional neural network (CNN) pixel-based antenna modeling and binary particle swarm optimization (BPSO) algorithm to automate the process of generating antenna structures based on desired performance specifications. The framework follows a simple and systematic procedure. First, the antenna transmitter section is converted into a 10×10 binary matrix to enable searching of the combined design sections. Then, a convolutional neural network is trained using 150,000 simulated data to predict the scattering parameters as a fast alternative to time-consuming electromagnetic simulations. Next, using a pixel state fitness function, the BPSO algorithm is optimized with high confidence and the reflection coefficient at the target frequencies is minimized. Representing the antenna transmitter as binary pixels expands the design space and overcomes the weaknesses of manual trial-and-error design. The results of comparison with the genetic algorithm GA and the simulated feedback algorithm SA show that the BPSO-CNN framework has a faster convergence speed and less error in the loss coefficient value at target frequencies. In addition to developing intelligent antenna design methods, this research provides a scalable model for automatic optimization of electromagnetic equipment.
کلیدواژهها English
