Predicting Transient Electromagnetic Interference (EMI) in Aircraft High-Voltage Systems via Hybrid Finite-Element Method (FEM) and Spatiotemporal Neural Network (STNN) Coupling with Adaptive Filter Synthesis

**Abstract:** This paper introduces a novel hybrid methodology for accurately predicting and mitigating transient electromagnetic interference (EMI) in aircraft high-voltage (HV) systems. Traditional FEM simulations are computationally expensive for real-time applications, while empirical methods lack the predictive accuracy required for comprehensive EMI design. Addressing this, we couple a reduced-order FEM model, pre-trained on a comprehensive dataset of HV system geometries and operating conditions, with a Spatiotemporal Neural Network (STNN) utilizing a recurrent convolutional architecture. This STNN dynamically refines FEM predictions based on transient EMI events, coupled with an adaptive filter synthesis algorithm to design robust EMI filters. The resulting hybrid approach achieves a 10x reduction in simulation time compared to full FEM while maintaining accuracy within 5% of experimental validation, paving the way for rapid HV system EMI design and optimization.