The Bathe subspace iteration method enriched by turning vectors

We present a novel extension of the Bathe subspace iteration method for the solution of the generalized eigenvalue problem in structural dynamics. The key idea is to enrich the subspace by using some turning vectors to replace current iteration vectors. The turning vectors are evaluated from the subspace of the current iteration. The scheme is simple and a considerable improvement in computational efficiency is achieved. A simplified convergence analysis is given and the results of some example solutions show the effectiveness of the method.