CONFIDENCE BOUNDS FOR FREQUENCY RESPONSE FUNCTIONS FROM TIME SERIES MODELS

Abstract The harmonic probing algorithm allows the generation of Frequency Response Functions (FRF) from discrete-time system models. If the model is non-linear, the higher-order FRFs or Volterra kernel transforms can be obtained. The object of this paper is to supplement the algorithm, with a procedure for estimating confidence bounds for the FRFs. A Monte Carlo approach is used, and the procedure is illustrated on models produced from simulation and from an experimental data set.