Lanczos eigenvalue algorithm for large structures on a minicomputer

This paper considers the problem of determining the partial eigensolution of linearized symmetric eigenvalue equations on a minicomputer arising from the finite element analysis of structural vibration problems. The Lanczos algorithm for transforming the problem to a symmetric tridiagonal eigensystem is discussed. The algorithm is implemented using an out of core skyline storage technique for the stiffness and mass matrices. This storage method is efficient when the matrices are sparse and of large order. Numerical results are given comparing the Lanczos algorithm with the Subspace Iteration and the Determinant Search methods.