论文部分内容阅读
汪荣江提出一个简单的正交归一化技术来克服经典的Thomson-Haskell传播矩阵方法中存在的数值不稳定问题.为了进一步提高计算效率,给出该方法的2种改进.一种改进方法是将传播矩阵中与频率无关的部分分离出来,对于某一固定的水平慢度,这些矩阵只需计算一次;另一个改进是利用Langer块对角化的技术,将传播矩阵分解为几个稀疏矩阵的乘积.我们将改进之后的算法应用于计算水平分层模型中的广义反射系数.较之原有方案,提出的改进能节省一半计算时间.
Wang Rongjiang proposes a simple orthogonal normalization technique to overcome the numerical instability problem in the classical Thomson-Haskell propagation matrix method.In order to further improve the computational efficiency, two improvements of this method are given.An improved method is to use In the propagation matrix, the frequency-independent parts are separated. For a fixed horizontal slowness, these matrices are calculated only once. Another improvement is to use Langer block diagonalization technique to decompose the propagation matrix into several sparse matrices Product.We apply the improved algorithm to the generalized reflection coefficient in the calculation of horizontal hierarchical model.Compared with the original scheme, the proposed improvement can save half the computation time.