仿晶体生长快速算法用于加速2D 静磁场设计中的参量扫描

    Crystal Growth Rapid Algorithm for Accelerating Parameter-Sweep of 2D Magnetostatic Design Problem FAN

    • 摘要: 为了加速二维静磁场设计中参量扫描的速度,本文提出仿晶体生长快速算法:网格空间不变,结构在网格空间中微动,利用前一次计算出来的结果,构造新的初解,以提高下一次有限差分法计算矢量磁位的迭代速率。该方法可以将扫描时间平均减少一半以上,并且计算结果与普通计算结果偏离很小。与同类软件相比,采用该方法的参量扫描和优化过程速度更快,也有潜力应用于其他复杂结构的优化过程。

       

      Abstract: This paper presents a new rapid algorithm to accelerate parameter-sweep of 2D magnetostatic design problem: All designed structure parameters were jiggled in a fixed mesh space. To speed up the iteration process of finite difference method calculation for vector magnetic potential after jiggle, new initial potential distribution was constructed according to the former one in the mesh space. At least 1/2 parameter-sweep time can be reduced with this algorithm and the calculated results are good agreement with those of normal calculations.

       

    /

    返回文章
    返回