ISSN 1004-4140
    CN 11-3017/P

    分块多级优化的时空双变正演模拟算法

    Optimized Dual-variable Grid Forward Modeling Algorithms using a Multi-block and Multi-stage Method

    • 摘要: 近年来,兼顾模拟精度和模拟效率的变网格方法越来越受到重视。在传统变网格算法的基础上,发展了分块、多级优化的时空双变正演模拟算法。主要的改进之处有:首先,本文从理论角度推导变网格算法在网格变动区域产生的虚假反射误差的数学公式,通过引入Lanczos滤波算子增强算法的稳定性,有效减弱了由网格变动引发的误差。其次,针对传统变网格算法模拟微小裂缝,超高倍网格变化时引入的误差和不稳定性,在交错变网格算法中引入多级变网格思想,可以高效精确的实现了超高倍变网格模拟。最后,当存在多个分散目标区,且单个目标区变网格格式存在差异时,提出分块变网格思想,可在多个区域进行不同倍数的变网格模拟,最大化的提高模拟计算效率。

       

      Abstract: In recent years, owing to its trade-off simulation accuracy and simulation efficient, the variable-grid method has garnered increasing attention. Based on the traditional variable-grid algorithm, this study proposed a multi-block and multi-stage method. First, we derived the mathematical formula for the false reflection error generated by the variable grid algorithm in the grid change area. By introducing the Lanczos filtering operator to enhance the stability of the algorithm, the error caused by grid changes can be reduced. Second, to simulate small cracks using traditional variable-grid algorithms, a large error and instability are often introduced under ultra-high grid ratio variations. However, by introducing a multi-stage condition to the staggered variable-grid method, we can achieve efficient and precise simulations with ultra-high grid ratios. Finally, to improve the simulation efficiency for an existing number of discrete target areas with different variable grid formats, we presented a multi-block variable grid to conduct different multiples of variable-grid simulations in each region.

       

    /

    返回文章
    返回