Stochastic subspace projection schemes for dynamic analysis of uncertain systems

We present stochastic subspace projection schemes for dynamic response analysis of linear stochastic structural systems. The underlying idea of the numerical methods presented here is to approximate the response process using a set of stochastic basis vectors with undetermined coefficients that are estimated via orthogonal/oblique stochastic projection. We present a preconditioned stochastic conjugate gradient method based on the conjugate orthogonality condition for approximating the frequency response statistics of stochastic structural systems. We also outline a new stochastic projection scheme for solving the generalized algebraic random eigenvalue problem. Some preliminary results are presented for a model problem to illustrate the performance of the proposed methods.