Artificial Intelligence-Based Transmission Frequency Prediction in Cable Systems Using Electromagnetic Field Data

This study presents an AI-driven framework for predicting transmission frequency in cables using magnetic field data, offering a modern alternative to conventional analytical methods. A reference model based on transmission line theory analyzes twisted-pair cables via distributed parameters (R, L, C, G), yielding an initial frequency estimate of 750 MHz. Concurrently, an MLP neural network trained on HFSS-simulated magnetic field data predicts 745 MHz, deviating by 5 MHz. To enhance precision, a hybrid neural network—combining transformer, LSTM, and a physics-informed layer grounded in Maxwell’s equations—is proposed, reducing the error to 0.2 MHz (749.8 MHz prediction). The AI approach demonstrates superior computational efficiency while maintaining high accuracy. Key preprocessing steps include data normalization, noise reduction, and magnetic field vectorization. Findings validate AI’s potential to streamline frequency estimation in cable systems, bridging theoretical and data-driven paradigms. This work advances applications in electrical and telecommunications engineering, emphasizing scalable, intelligent solutions for electromagnetic analysis.