Skip to Main Content
Coefficient matrix of linear equations from finite element method (FEM) is sparse and symmetrical. For the sake of saving CPU operational time and reducing storing requirement to computer, we introduce fully sparse strategy that stores only nonzero elements of symmetrical part by chain pattern. In order to save computational time to accesses data during LU factorization, we develop a symbol LU factorization method. It can minimize fill-in elements and reduce computational quantity of LU factorization. By an address index system and minimum full-in elements algorithm, efficiency of LU factorization can be improved significantly. Numerical experiments show that combination of the symbol LU factorization method and fully sparse storage structure can improve the algorithmic efficiency for FEM solution of large scaled sparse linear equations.