An adaptive-extended modeling to accelerate electromagnetic study and data generation in superconducting magnet applications

Superconducting magnets possess unique electromagnetic properties, making them applicable in fields such as nuclear magnetic resonance, maglev, and fusion. These applications generally involve diverse environments featuring AC or DC conditions, where superconducting properties are influenced by various factors. Specifically, the most concerning properties in high temperature superconducting (HTS) magnets include critical current, AC loss, screening current effects, and so on. Finite element method is widely used in reliable numerical studies for these properties. Several popular models have been proposed and developed to get higher precision and less calculation time. However, constrained by computational resources, they still have challenges in supporting high-throughput analysis. In various studies on electromagnetic characteristics of magnets, a substantial amount of data is often required to facilitate the introduction of artificial intelligence (AI) methods or optimization approaches. This paper proposes an adaptive-extended J-model to compute superconducting properties, further enhancing the efficiency of electromagnetic study and also serving to generate dataset for AI methods. It reduces computation time to only 20%–30% of that of the existing fastest model while maintaining similar levels of accuracy. By using this method as a data-generative tool, the dataset of a series of HTS solenoids including 2000 turns is expeditiously obtained and employed to predict the screening current induced field. The predictive performance is reliable under the dataset calculation time of mere minutes. This study significantly shortens the time to realize big dataset demands, accelerating electromagnetic study of superconducting magnets in various scenarios.