基于并行预处理算法的三维重力快速反演

随着地球物理设备和探测技术的不断发展，快速处理大规模地球物理数据的需求也随之增长。为了解决三维重力数据密度反演的耗时问题，提出一种并行的预处理共轭梯度算法来提高计算效率。本文分别采用两种不同的预处理算子通过组合模型数据反演进行测试比较，并利用迭代残差和计算用时共同评价其加速效果。结果表明：对称逐次超松弛预处理方法比对角预处理方法反演计算速度快，密度结果更贴近实际模型；与传统串行的共轭梯度算法相比，本文并行预处理快速算法可以获得近19倍的加速比。将该算法应用于美国Vinton盐丘的实测重力数据中，反演结果能够很好地圈定出岩体的位置，验证了本文并行预处理共轭梯度法在三维重力数据快速反演中的高效性和可行性。

Three-Dimensional Fast Gravity Inversion Using Parallel Preconditioned Algorithm

With the dramatic development of the geophysical equipment and exploration technology, the demand of rapid processing of large-scale geophysical data has been increasing.In order to solve the time-consuming problem of three dimensional gravity data inversion,a parallel implementation of preconditioned conjugate gradient algorithm is proposed to improve the computation efficiency.We adopted and tested two different preconditioned strategies with synthetic data,and evaluated their acceleration quality through observing the residual error of iterations and the running time.The results indicate that the symmetric successive over-relaxation(SSOR)preconditioner has a better performance with density results closer to the true model.Compared with the conventional conjugate gradient algorithm in serial code,our parallel preconditioned conjugated gradient algorithm can result in up to approximately 19x speedup.We used it to the real gravity data from the Vinton salt dome in USA,and the inversion results well delineated the location of the cap-rock,which verified the high efficiency and feasibility of this parallel preconditioned conjugate gradient algorithm in 3D fast gravity inversion.