A robust Adam-SIMP framework for 3D topology optimization of wireless power transfer systems

Abstract The magnetic core is a critical component in wireless power transmission (WPT) systems. However, existing core topology optimization (TO) methods often rely on constraints, resulting in simplified but bulky structures that limit the ability to extract the fine details required to maximize magnetic coupling. The development of 3D printing technology has made it possible to manufacture such intricate structures. This paper proposes a 3D topology optimization framework based on the direct density method, employing an adaptive moment estimation (Adam) optimizer. By introducing a dynamic Heaviside projection strategy, this method effectively handles high-dimensional design spaces and mitigates the gradient vanishing problem. Simulation results demonstrate good convergence stability under different initial density distributions. Furthermore, by tracking the baseline performance of a solid core under volume penalties, the framework successfully identifies critical flux paths and removes electromagnetically redundant material, resulting in a branched structural layout. The optimized design reduces the core weight by approximately 57% while maintaining comparable mutual inductance to a solid core, providing a practical method for designing lightweight WPT magnetic couplers.