基于LSQR法与相关系数自约束的磁异常三维反演  

3D inversion of magnetic data based on LSQR method and correlation coefficientself constrained

作　　者：梁生贤[1] 焦彦杰[1] 樊文鑫 杨博志 LIANG Sheng-xian;JIAO Yan-jie;FAN Wen-xin;YANG Bo-zhi(Chengdu Center,China Geological Survey,Chengdu 610081,China)

机构地区：[1]中国地质调查局成都地质调查中心

出　　处：《地球物理学进展》2019年第4期1475-1480,共6页Progress in Geophysics

基　　金：国家重点研发计划(2016YFC060308);地质调查项目(121201010000150007,121201010000150014)联合资助

摘　　要：本文从节省磁力反演的计算成本和提升反演结果的可靠性两个方面着手.其一,利用迭代正则化方法LSQR法求解反问题,结合等效几何格架技术将核矩阵分为若干个子矩阵进行存储与运算,以节省计算成本;其二,将快速扫描的互相关系数作为先验信息,通过处理与深度加权函数同时引入到核矩阵约束中,以提高反演结果的可靠性和分辨率.合成模型试验结果表明,这种方法求解的结果稳定,与真实模型更加接近.并将其应用于西藏驱龙甲玛铜多金属矿集区磁化率三维反演成像,取得了良好的效果.This paper focuses on saving the computational cost of magnetic inversion and improving the reliability of inversion results.Firstly,the iterative regularization method(LSQR method)is used to solve the inverse problem,and the matrix is divided into several sub-matrices for storage and calculation with the equivalent geometry grid technology to save the computational cost.Secondly,using the cross-correlation coefficient of the fast scan method as the prior information,and introduced depth weighting function into the constraint of the kernel matrix,which can improve the reliability and resolution of the inversion.The results of the synthetic model test show that the result obtained by this method is stable and is closer to the true model.This method was applied to the 3-D inversion imaging of magnetic susceptibility of the Qulong Jiama polymetallic mining area in Tibet,and it achieved good results.

关 键 词：迭代正则化 LSQR法 磁异常三维反演 相关系数 自约束 

分 类 号：P631[天文地球—地质矿产勘探]