面波频散反演的数值模拟

NUMERICAL SIMULATION FOR INVERSION OF SURFACE WAVE DISPERSION

大量的数值模拟表明 ,SVD和 LSQRD在面波频散网格反演两步法中的应用效果都很好。但 SVD可以用分辨矩阵、信息矩阵和协方差矩阵对解估计进行数学上客观有效的评价 ;而对于大型稀疏方程组的求解 ,LSQRD确是一种内存需求小、计算速度快以及分辨抗噪能力都较强的算法。在现有计算机运算速度较快、内存可以扩充较大的条件下 ,实测数据量不很大时 ,应采用 SVD算法进行线性反演。

A lot of numerical simulations indicate that both SVD and LSQRD are very effectively applied to inversion of surface wave dispersion. Using SVD method through resolution matrix,information matrix and covariance matrix,the solutions can be evaluated mathematically, objectively and effectively, while LSQRD is more useful to solve a large sized sparse linear system because if has high efficiency in computation and strong noise abatement and needs small store spaces, so that SVD algorithm should be adopted in inversion of linear system under fast computations and big store spaces of present computers if the quantity of data is not too large.