提出了一种MAINV稀疏近似逆预条件算法,用于改善电磁场边值问题的有限元分析所产生的的线性系统的迭代求解。该预条件子是在基本AINV算法基础上,在分解过程中对可能导致算法崩溃的极小主元进行实时补偿,从而获得高质量的预条件子。数值结果表明,MAINV预条件子对SQMR以及若干经典迭代法的加速效果十分明显;此外,与其他常规预条件子相比较,MAINV具有更好的求解性能。 A MAINV sparse approximation inverse preconditioning algorithm is proposed to solve the linear system iteratively generated by the finite element analysis of electromagnetic field boundary value problems. Based on the basic AINV algorithm, this preconditioner compensates the minimal principal components that may cause the algorithm to collapse during the decomposition, so as to obtain high quality preconditioners. The numerical results show that the MAINV preconditioner has an obvious accelerating effect on SQMR and several classical iterative methods. In addition, MAINV has better solving performance than other conventional preconditioners.