Covariance Subspace Identification: Numerical Analysis of Spurious Mode Stability

We recall briefly the covariance driven subspace identification method with some illustrative results. One of the well-known problems in most of identification methods is that we get spurious (computational) modes together with the true ones. In this paper we present an example of such a situation and we do different simulations with Scilab to compare the behavior of 2 modes and define what is a possible procedure to avoid the choice of a spurious mode in the modal signature of the structure. We compare different techniques for the perturbation of the steps of the algorithm and we follow the evolution of the singular values, the stability of the eigenfrequencies, the stability of the damping factors and the behavior of the eigenvectors of the transition matrix of the system. We conclude by a possible procedure on data perturbation and monitoring on sliding time periods to discard the spurious modes. The last illustrative result concerns the auto validation of the identification result with diagnosis test for checking the influence of computational modes.