Quantification d'incertitudes d'un modèle à temps continu dans une procédure d'identification indirecte

Abstract

This paper deals with the problem of indirect identification approach. In this framework, a new method for estimating the parameter covariance matrix (natural frequency, damping factor, static gain, time constant) of a continuous-time model is proposed. The indirect identification approach involves two steps: at first, a discrete-time model is obtained by applying discrete-time model estimation methods to the available sampled data; then, the discrete-time model is transformed into the required continuous-time model. The method proposed focuses on first and second orders transfer functions: uncertainties on static gain, time constants, natural frequency and damping factor are estimated directly, according to the model structure. Furthermore, this procedure is generalized to the case of transfer functions with an unspecified order. The resulting algorithm is then successfully applied to three examples via Monte Carlo simulations and numerical studies further illustrate the performances of the method.