张量积型Said-Ball曲面的预处理渐近迭代逼近法

为加快张量积型 Said-Ball曲面渐近迭代逼近法的收敛速度，探讨了张量积型Said-Ball曲面渐近迭代逼近法的预处理技术。首先利用对角补偿约化技术构造了预处理子，然后结合矩阵Kronecker积性质，采取预处理渐近迭代逼近法求解张量积型Said-Ball曲面。为进一步降低计算量并提高算法的稳定性，利用广义极小残差法求解预处理方程，得到预处理渐近迭代逼近法的非精确求解方法。分析了预处理渐近迭代逼近法及非精确求解方法的收敛性。最后用数值实例说明预处理子能大大减小迭代矩阵的谱半径，令预处理技术及其非精确求解方法的计算效率明显提高。此外，由于对角补偿预处理子能改善配置矩阵的谱分布，因此也可用于对广义极小残差法的预处理，以改善其收敛性。

Preconditioned progressive iterative approximation for tensor product Said-Ball patches

In order to accelerate the convergence of progressive iterative approximation (PIA) for tensor product Said-Ball patches, the preconditioning technique for PIA for tensor product Said-Ball patches is studied. Firstly, the diagonally compensate reduction is used to construct the preconditioner, then combined with the properties of Kronecker, the preconditioned progressive iterative approximation (PPIA) for tensor product Said-Ball patches is exploited. To further reduce the amount of calculation and improve the stability of the algorithm, the generalized minimal residual method is employed to inexactly solve the preconditioned equations and yield the inexact version of PPIA. The convergence of the PPIA and its inexact version are analyzed. Finally, numerical results illustrate that the proposed preconditioner can greatly reduce the spectral radii of PPIA's iteration matrices, hence accelerating the convergence rate of PPIA and its inexact version. Besides, since the preconditioner can improve the spectral distribution of the collocation matrix, it can also be used in the preprocess of the generalized minimal residual method to improve its convergence.