IFA Design Simplified: an MLP-Based Geometrical Parameter Prediction from Expected Performance

Designing compact and efficient antennas for Internet of Things (IoT) devices remains a significant challenge, particularly when integrating antennas into small terminals. This study introduces a machine learning (ML)-based approach to predict the geometric parameters of an Inverted-F Antenna (IFA) based on its performance metrics, specifically fractional bandwidth (FBW) and total efficiency (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\eta$</tex>). A custom dataset was created to train a Deep Neural Network (DNN) employing a Multi-Layer Perceptron (MLP) architecture. This dataset includes terminal dimensions, antenna clearance areas, and other relevant design parameters. The proposed method seeks to establish a predictive model that enables the estimation of terminal dimensions and antenna geometries directly from performance requirements, thereby reducing design time and optimizing antenna integration for IoT applications. In this framework, the user specifies desired performance characteristics, such as target <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$F B W$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\eta$</tex> values, and the model predicts the corresponding antenna geometry and terminal dimensions. Validation of the predictions was carried out using simulation data, showing a close match between the predicted and simulated results. The discrepancies were minimal, with differences of less than <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 0 ~ M H z} {@}-\mathbf{6 ~ d B}$</tex> for FBW and less than 2% for total efficiency (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\eta$</tex>).