Abstract
Filters are critical components in modern communication systems, particularly at high frequency applications. However, conventional filter designs such as electromagnetic (EM) simulations etc. based approach becomes computationally demanding and time-inefficient at higher frequencies. In this study, we propose a hybrid optimization approach that integrates Artificial Neural Networks (ANN) with Grey Wolf Optimizer (GWO) to design a microstrip filter. The developed ANN was trained to predict S-parameters for a given set of microstrip dimensions. Subsequently, the GWO algorithm was employed to optimize filter dimensions based on specific design objectives: (1) minimizing S 21 deviations in the passband, and (2) ensuring S 21 < -20 dB in the stopband. The optimized filter dimensions were validated against advanced design system simulations, showing strong alignment between the GWO-ANN predictions and EM results. This hybrid approach significantly reduces design time while maintaining high accuracy, combining the precision of EM simulations with the efficiency of machine learning and evolutionary optimization. The proposed method provides an effective and fast tool for high-frequency filter design, demonstrating its potential for practical implementation in advanced communication systems.