Wireless charging ferrite core design with improved material efficiency for mobile devices by enhancing outer-region...

Wireless power transfer (WPT) systems integrated with machine learning (ML) techniques have emerged as promising solutions for enhancing performance and efficiency. Material efficiency, which involves achieving high performance with minimal material use, is crucial for mobile devices to enhance user satisfaction. This paper proposes a core structure design methodology for WPT systems that enhances mutual inductance (MI) by segmenting and assembling the ferrite core, thereby improving material efficiency within limited space constraints. For genetic algorithm (GA)-based optimization, an ML estimator is implemented to replace iterative finite element method simulations during the GA optimization loop, reducing the average computation time from 203s to 86μs and significantly accelerating the GA optimization process. Furthermore, the designed core structure exhibits improved uniformity of magnetic flux density by enhancing the outer-region magnetic flux density between the coils, thereby contributing to higher MI. Experimental results show that the proposed core structure achieves a 17.44% higher normalized MI compared to conventional structures, using less material through an asymmetrical configuration, which also reduces weight and cost. Furthermore, the optimized asymmetrical structure demonstrates robustness under lateral misalignment conditions ([Formula: see text] mm), consistently maintaining higher material efficiency than the full-piece reference even under spatial offsets. These findings demonstrate the potential of ML-driven approaches for developing high-performance, computationally efficient, and cost-effective WPT systems.