针对三维地电场正演计算过程中形成的超大规模稀疏线性方程组,采用不完全Cholesky分解方法进行预条件处理,经过条件数改善后形成的新线性方程组的系数矩阵变为一个近似的单位矩阵,再应用Lanczos算法将会提高数值计算的稳定性,加快迭代收敛的速度,通常在迭代次数远小于系数矩阵阶数时就能得到较好精确解的近似值,为下一步的电阻率三维反演计算打下了非常好的基础.
An incomplete Cholesky factorizafion is used to preconditioning large sparse coefficient matrix in 3-D geoelectrie field forward modelling. Coefficient matrix of the new system is an approximate unit matrix. Preconditioned Lanczos algorithm improves stability and convergence of iterative procedures. In the algorithm, as iterative numbers are smaller than dimensions of coefficient matrix, approximation of the exact solution is obtained. Preconditioned Lanczos algorithm is effective in 3-D geoelectric field forward modeling. It makes a foundaiton for 3-D geoelectric inversion.