厚导体涡流扩散稳态损耗的结构化精细积分法

    A Structured Precise Integration Method for Steady Eddy-current Loss in Thick Conductors

    • 摘要: 厚导体屏蔽板在工频及谐波磁场下存在显著趋肤效应。为分辨导体表层边界层,厚度方向需要采用细网格离散,导致扩散算子谱半径增大和系统刚性增强。显式方法受稳定步长限制,隐式方法在细网格条件下每步求解代价较高。针对这些问题,本文提出一种用于二维准静态厚导体涡流稳态损耗评估的结构化精细积分方法。该模型仅在导体域内建立,导体上表面的正弦激励通过Neumann边界输入引入,不再显式离散空气区域及其界面。时间推进采用扩散型矩阵指数更新,并通过 \phi 函数卷积处理非齐次边界输入;针对离散算子的Kronecker可分离结构,构造基于Sylvester型算子作用的低存储实现流程。结果表明,本文方法在周期平均焦耳损耗指标上与参考结果一致;在统一稳态提取、后处理和主循环计时前提下,其误差-CPU时间关系优于Crank-Nicolson方法,并在高频细网格工况下保持较低的内存需求。

       

      Abstract: Thick conducting shielding plates exhibit strong skin effect under power-frequency and harmonic magnetic fields. Resolving the boundary layer near the conductor surface requires a fine mesh in the thickness direction. This enlarges the spectral radius of the diffusion operator and makes the semi-discrete system stiff. Explicit schemes are then limited by the stability step size, while implicit schemes become expensive under fine meshes because a large linear system must be solved at each step. For this problem, a structured precise integration method is developed for steady-state eddy-current loss evaluation in two-dimensional quasi-static thick conductors. The model is solved only in the conductor domain. The sinusoidal excitation on the upper surface is introduced through a Neumann boundary input, and the air region is not discretized explicitly. Time marching is carried out by a diffusion-type matrix-exponential update. The nonhomogeneous boundary input is treated by \phi -function convolution. For the Kronecker-separable discrete operator, the matrix-function action is implemented through Sylvester-type operator actions. Dense propagation matrices are therefore avoided, and the memory requirement is reduced. Numerical results show that the proposed method agrees with the reference results in cycle-averaged Joule loss. Under the same steady-state extraction, post-processing, and main-loop timing criteria, it gives a better error-versus-CPU-time relation than the Crank-Nicolson method and requires less memory under high-frequency fine-mesh conditions.

       

    /

    返回文章
    返回