AC Loss Prediction of HTS Propulsion Motors Using Convolutional Neural Network-Based Surrogate Model for System-Level Modelling of Electric Aircraft

System-level modelling, crucial for designing a superconducting propulsion by simulating the interactive behavior of components, faces a trade-off: simple models like look-up tables are fast but inaccurate, while detailed Finite-Element (FE) models are accurate but too slow for system-wide studies. This work addresses the challenge of AC loss analysis of superconducting propulsion motors by proposing a novel deep-learning surrogate framework that delivers extremely fast and highly accurate solutions. This enables more efficient design, monitoring, and optimization of superconducting motors within large-scale system models. The proposed artificial intelligence-based model benefits from integrated feature extraction powered by Convolutional Networks, followed by a deep interconnected neural network to grasp dynamic AC loss waveforms. This work advances real-world applicability by enhancing time-dependent surrogate models using data from an FE machine model. The results demonstrate that the proposed deep neural network captures time-dependent AC loss behavior with above 99.92% accuracy, dynamically predicting a full AC loss waveform over a cycle in just 3.8 milliseconds. These findings will enable advanced system-level aircraft simulations to incorporate dynamic, highly accurate, and ultra-fast AC loss predictions, allowing exploration of various operating conditions encountered during flight.