研究目的
To synthesize a cylindrical-rectangular ring microstrip conformal antenna using a PSO-based approach with SVR models to compute the resonant frequency, employing RBF and wavelet kernels for improved accuracy and efficiency.
研究成果
The PSO-based synthesis using SVR models with wavelet kernels (Morlet and Mexican-Hat) achieves faster convergence and higher accuracy compared to RBF kernels. HFSS validations confirm that the optimized dimensions yield resonant frequencies close to the desired values, demonstrating the effectiveness of the approach for antenna design.
研究不足
The method relies on HFSS simulations for training data, which are computationally intensive and time-consuming. The accuracy and generalization of SVR models depend on the quality and range of the training data. The approach is specific to cylindrical-rectangular ring microstrip antennas and may not generalize to other antenna types without retraining.
1：Experimental Design and Method Selection:
The study uses a PSO algorithm to optimize antenna dimensions (slotL and slotW) for a desired resonant frequency, with SVR models trained using different kernel functions (RBF, Morlet, Mexican-Hat) to predict resonant frequencies. The SVR models are trained on data generated from HFSS simulations.
2：Sample Selection and Data Sources:
A dataset of 225 samples is created via HFSS simulations, with slotL and slotW ranging from 2mm to 9mm. Training uses 190 samples, and testing uses 35 samples, randomly selected.
3：List of Experimental Equipment and Materials:
A computer with a 4.0 GHz CPU and 16GB RAM is used for simulations. Software includes ANSYS HFSS for electromagnetic simulations and custom implementations for SVR and PSO algorithms.
4：0 GHz CPU and 16GB RAM is used for simulations. Software includes ANSYS HFSS for electromagnetic simulations and custom implementations for SVR and PSO algorithms.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: First, HFSS generates training patterns. SVR models are trained with different kernels. PSO is initialized with a swarm of particles (e.g., 50 particles) representing slotL and slotW values. The cost function minimizes the difference between desired and computed resonant frequencies. Iterations update particle positions and velocities until convergence.
5：Data Analysis Methods:
Accuracy is measured as mean percentage accuracy from test results. Optimization performance is evaluated based on convergence time and iteration count. HFSS simulations validate the optimized dimensions.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
