【摘要】 针对大地电磁法有限元模拟中计算量大的特点,采用按行压缩存储方式的不完全LDLT预处理共轭梯度法快速求解大型复系数方程组。引入不完全LDLT预处理,提出快速求解(LDLT)-1r的方法,以加快预处理共轭梯度法的收敛速度。研究结果表明:当网格节点自由度超过1万时,压缩率达到99.9%,求解方程组时间在1 s以内,为进一步快速反演奠定了基础。更多还原

【Abstract】 Based on the characteristics of large computing cost in finite element modeling about magnetotelluric method,preconditioning conjugate gradient method based on compressed storage was proposed to solve large and complex systems.Incomplete preconditioner LDLT was introduced and fast solving(LDLT)-1r was proposed to accelerate the convergence rate of preconditioning conjugate gradient method.The results show that the compression ratio reaches 99.9%,and it completes in less than 1 s for the equations more than 10 000 degrees of freedom,which further lays the foundation for rapid inverse.更多还原