The eigen problems based on the finite element method have been solved numerically by an enhanced subspace iteration method. Bathe's basic algorithm of subspace iteration is improved by eliminating recalculation of converged eigenvectors, introducing Krylov sequence as initial vectors and incorporating with shifting techniques. The number of iterations and computational time are considerably reduced when compared with the original one, and reliablilty for the catching second copy of the multiple roots is retained successfully. Further research would be required for the mathematical justification of the present method. And comparison with the Lanczos method is also required.