二维直流电阻率与音频大地电磁自适应渐进联合反演

Study of 2-D Joint Inversion of Direct Current Resistivity and Audio Magnetotelluric Data Using Adaptive Progressive Mesh Refinement Strategy

摘要:

开展了非结构化三角网格条件下的二维直流电阻率与音频大地电磁自适应渐进正则化联合反演研究。通过研发以模型灵敏度信息为依据的反演网格自适应优化技术，构建由粗网格到细网格逐步反演的自适应渐进反演策略，减少了反问题对稳定因子的依赖，进而降低了正则化因子搜索的计算量；通过最小二乘算法求解非结构化三角网格的模型粗糙度，构建了非结构化网格条件下的最小结构稳定因子；采用高斯?牛顿法优化求解正则化反演目标函数，通过双共轭梯度稳定算法求解高斯?牛顿方程，确保反演稳定性，同时有效减少了反问题对内存的需求。理论模型与实测数据的联合反演试算表明，直流电阻率与音频大地电磁数据联合反演能够有效减少反演多解性,提高反演效果，实现音频大地电磁静态效应异常体的直接反演。

Abstract:

The joint inversion of direct currency resistivity data and audio magnetotelluric data under the condition of unstructured triangle mesh is studied. A step-by-step inversion technique is developed,and an adoptive mesh refinement method is studied based on the model sensitivity information. The inversion from coarse mesh to fine mesh reduces the dependence on the stabilizer. Least computational cost of searching regularization factors is required compared with traditional regularization method. The model gradient is calculated by using the least square method.Therefore,the minimum structure stabilizer is established based on unstructured triangle mesh. The cost function is optimized by using the Gauss-Newton method, and the Gauss-Newton system is solved by using the bi-conjugate gradient stable method, aiming to reduce the memory requirement and ensure the stability of inversion. Both the synthetic and field data are inverted.The inversion resolutions show that the joint inversion of the two data sets improve resolutions, and reduce the non-uniqueness of the regularization inversion problem. Joint inversion can get shallow anomalous bodies, which can cause static shift of the audio magnetelluric method.