基于混合蛙跳算法的三维重力密度异常快速反演

项目名称： 基于混合蛙跳算法的三维重力密度异常快速反演

项目编号： No.41504081

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 天文学、地球科学

项目作者： 蒋甫玉

作者单位： 河海大学

项目金额： 21万元

中文摘要： 三维密度反演是重力勘探工作，特别是资料处理与解释的主要环节之一。通过反演得到的岩石密度信息，对地质填图、地质构造研究以及矿产资源调查具有重要意义。然而，地球物理反问题的多解性及反问题计算的低效率严重地阻碍了既往反演方法的应用。针对这两个瓶颈问题，本课题在对启发式非线性反演研究的基础上，创新地利用混合蛙跳算法以实现对三维密度异常的快速反演。拟结合算法的内在机理，在密度更新过程中加入密度、深度约束条件以减少多解性；拟根据重力异常边缘增强成果，限定种群规模总数、采用格架分离技术和格架等效计算方案、将算法本身具有的并发性与图形处理单元具有多个流处理单元的硬件优势结合起来、将只对最差蛙进行多次迭代更新的策略改进为对子群内所有蛙进行一次更新，以大幅度提高计算效能；拟以梅山铁矿区的重力数据为基础，将反演结果与已有的物探、地质资料作对比，探讨反演结果的可靠性，以期重力勘探在社会发展中发挥更加积极的作用。

中文关键词： 混合蛙跳算法；三维密度反演；多解性；计算效率；并行计算

英文摘要： Three-dimensional density inversion is one of the key processes in gravity exploration work, especially in data processing and interpretation. The obtained density information of rock through inversion has important significance for geological mapping, geological structure research and mineral resources survey. However, multiple solutions and low calculation efficiency of the geophysical inverse problem seriously hampered the application of the previous inversion methods. For these two bottlenecks, we innovatively plan to use Shuffled Leap-Frog Algorithm (SLFA) to implement fast inversion of three-dimensional density anomaly based on the study of heuristic nonlinear inversion. According to the inner mechanism of the SFLA, we intend to add density and depth conditions during the density renewal process to reduce the multiple solutions of inversion. In order to greatly improve the calculation efficiency of inversion, we intend to limit the total frog size according to the result of gravity anomaly edge enhancement firstly, use the grid separation technology and grid equivalent computing solutions, combine parallelism of SFLA with the hardware advantages of graphics processing unit with multiple streams processing unit, improve the renew strategy from updating the worst frog repeatedly by multiple iterations to update all frogs in subgroup only once. We also intend to study the reliability of SFLA by comparing the inversion results of Meishan Iron gravity data with the known geophysical and geological data of Meishan Iron, and expect that gravity exploration can play a more active role in social development.

英文关键词： Shuffled Frog Leaping Algorithm;three dimensional density inversion;multiple solutions;calculation efficiency;parallel computing