基于瞬变电磁矩变换的快速三维反演方法

瞬变电磁法的严格三维反演计算复杂、占用资源多,在普通计算机上难以实现.本文引入瞬变电磁矩变换的概念,提出一种快速三维反演方法.该方法基于阻性限制(resistive limit)特性,建立包含异常体的三维大地的一阶矩响应正演算法,根据不同约束条件,选择优化的最速下降法实现瞬变电磁快速三维反演.文中通过含异常体的三维大地正演一阶矩与仿真数据一阶矩的对比,验证了快速三维正演算法的有效性,之后在不同约束条件下,利用优化的最速下降法实现了对含噪声的仿真瞬变电磁数据的快速三维反演.结果表明,该方法能够在普通计算机上短时间内较为准确地反演出地下异常体的体积和位置,在瞬变电磁数据的实时解释工作中具有良好的应用前景.

A rapid 3D inversion based on TEM moment transformation

A full EM solution of the 3D inversion problem is difficult to carry out on common computer due to its complex forward calculation and high demand for user resources. These restrictions make rigorous 3D inversion still impractical in the interpretation of TEM data sets. A rapid 3D inversion scheme for interpreting TEM data is presented, utilizing the concept of TEM moment. Approximate forward modeling of the 3D inversion scheme has been accomplished by adopting the first order moment transform which is the resistive limit. The modified steepest decent method is employed for inversion of the underground target response. The inherent non-uniqueness associated with the underdetermined inversion problem is balanced by including constraints in the inversion such as depth weights or conductivity weights. The fully 3D electromagnetic modeling software is employed to calculate the synthetic time-domain data to verify the proposed 3D inversion scheme. A comparison of the forward theoretical results and the moment transformed results of synthetic time-domain data shows that forward calculation gives reliable results. In order to test the inversion algorithm and examine the variability of the inversion results due to non-uniqueness, the synthetic inversion examples are explored based on different constraints. The inversion was successfully completed in dozens of seconds to recover time constant models which approximately indicated the location of the true conductive body. In conclusion, the proposed rapid 3D inversion scheme can be efficiently implemented on the normal computer which indicates a promising application prospect in real-time interpretation of TEM data.