Latest revision as of 10:18, 4 May 2026

Abstract

Eddy current repulsion forces are essential to the operation of magnetic levitation, electromagnetic braking, and wireless power transfer systems, yet their accurate simulation remains computationally intensive due to the complexity of Maxwell’s equations and the need for fine-resolution meshing in transient, high-frequency domains. This paper presents a hybrid AI-Quantum computational framework that accelerates eddy current simulation by combining adaptive meshing via Physics-Informed Neural Networks for adaptive meshing, quantum-assisted solvers (HarrowHassidim-Lloyd algorithm and Variational Quantum Eigensolver) for selected finite element subdomains, and GPU-accelerated finite element methods, and reduced-order modeling based on Proper Orthogonal Decomposition and Quantum Autoencoders. The framework selectively routes well-conditioned sparse subblocks to quantum solvers while retaining classical GPU-based methods elsewhere. Validation across three representative applications—electromagnetic braking, magnetic levitation, and wireless power transfer—demonstrates that the hybrid solver maintains accuracy within 5%–8% of full FEM results while reducing computational cost by 1.3×–2× speedup in the tested scenarios. These results confirm the feasibility of integrating AI and near-term quantum computing into electromagnetic simulation workflows and provide guidance on complexity, resource requirements, and scalability.OPEN ACCESS Received: 14/10/2025 Accepted: 22/12/2025

Document