Data Driven Surrogate Modelling of Microstrip Frequency Selective Surface for 5.8GHz WLAN Application

Abstract

This study offers a thorough examination of data-driven surrogate modeling approaches used in the development of microstrip frequency selective surfaces in microwave systems. The research centers on the application of surrogate models to improve the effectiveness and productivity of FSS systems. The M2LP model, which is an enhanced version of the Multi-layer Perceptron neural network architecture, exhibited enhanced performance by addressing the constraints associated with the conventional MLP. The system integrated sophisticated activation functions, such as rectified linear units (ReLU) or exponential linear units (ELU), in order to enhance the smooth transmission of gradients and expedite the process of convergence during training. The M2LP methodology presents a succinct yet efficient strategy, rendering it particularly well-suited for the optimization of FSS design. In order to enhance the design of the FSS unit, the integration of the M2LP surrogate model with the Honey Bee Mating Optimization (HBMO) algorithm was undertaken, drawing inspiration from the mating strategy employed by honey bees. The successful implementation of modern artificial intelligence approaches and optimization algorithms has demonstrated significant efficacy in attaining superior levels of performance and efficiency. The validation of the outcomes obtained from the selection of optimal parameters using the M2LP model was conducted by comparing them with the results obtained from the full-wave simulation model. This comparison aimed to ensure that there was a precise agreement between the response of the electromagnetic solver and the proposed data-driven surrogate. The research findings indicate that the surrogate model developed in this study demonstrates effective optimization of FSS designs across a range of applications within the specified domain. In general, this study makes a significant contribution to the development of design methodologies for Frequency Selective Surfaces (FSS) and provides vital insights for future research in the domain of microwave antennas and circuits. Specifically, it focuses on the application of data-driven surrogate modeling and optimization algorithms.