Uncertainty quantification in data-driven stochastic subspace identification

A crucial aspect in system identification is the assessment of accuracy identified matrices. Stochastic Subspace Identification (SSI) a widely used approach for linear systems from output-only data because it combines high computational robustness and efficiency with estimation accuracy. Practical approaches estimating (co)variance matrices that are using SSI exist case where obtained shift-invariant structure extended observability matrix. However, data-driven SSI, often different way, state sequences. This treated present work, three common types weighting. First, shown estimated depend entirely on sample output correlation estimates, covariance which can be straightforwardly estimated. Subsequently, sensitivity analysis algorithm performed, such also computed. memory efficient implementation by computing related Jacobian only implicitly. An extensive numerical validation, covering range parameter choices, demonstrates variance description. Finally, practical use method context operational modal demonstrated an experimental study.