Design and optimization of a novel solenoid with high magnetic uniformity

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Окт. 20, 2024

Currently, solenoids are extensively utilized in various research fields due to their flexibility of fabrication and high magnetic field strength. However, the internal solenoid itself exhibits some non-uniformity defects, which limits its application domains. This article proposes a novel single winding tightly wound structure with an improved uniformity. To optimize near aperture conventional solenoid, auxiliary gradually changing diameter is included at each end solenoid. By adjusting different parameters edge effect was reduced, overall uniformity improved. The optimization achieved through GA-KAN network techniques. Finite element simulation results optimized show that proportion uniform areas can be by more than five times. reference for high-precision electromagnetic sensing applications based on solenoids.

Язык: Английский

---

Enhancing Artillery System Analysis through Innovative Neural Networks: A Comparative Study of Kolmogorov–Arnold Networks in Fourier Operators

Journal of Physics Conference Series, Год журнала: 2024, Номер 2891(10), С. 102017 - 102017

Опубликована: Дек. 1, 2024

Abstract Many problems in artillery systems can be described using partial differential equations (PDEs), and engineers need to repeatedly adjust the design object meet requirements of phase. Therefore, an efficient PDEs solver is needed during solvers based on deep learning, especially neural operators, this requirement. However, operators use multi-layer perceptrons (MLP) project data features onto output dimension, MLP lack interpretability, often face overfitting gradient vanishing, scalability. Kolmogorov–Arnold Networks (KAN) has recently been introduced considered a potential alternative MLP. Based this, KAN are used construct Fourier Neural Operators (FKANO) for solving forward inverse engineering. Especially three tasks approximation, equation solving, building surrogate models, proposed FKANO FNO were compared. It was found that although robustness training process lacking FKANO, performance comparable or even surpassing still achieved. The new network believed have advance development engineering analysis.

Язык: Английский

---

A Kolmogorov-Arnold Networks-Based Model for Forecasting of Natural Gas Consumption

Опубликована: Янв. 1, 2024

Язык: Английский

---