IJCAS

Many control techniques rely on the mathematical models of the systems to be controlled. This paper copes with the modelling problem of dynamical systems aiming to develop highly accurate modelling approaches. By an impulse response identification experiment, the dynamical observations with increasing data length are designed for the purpose of capturing the real-time information of systems and serving for on-line identification. According to the different features of the parameters of the systems to be identified, two separable identification models are constructed through the parameter decomposition and the model decomposition for simplifying the structure of the original identification model. On basis of the separable identification models, a separable Newton recursive parameter estimation approach is developed by means of the Newton search for acquiring highly accurate parameter estimates. In terms of the coupled terms in the separated sub-algorithms, a joint estimation algorithm is presented for removing the coupled terms. The experimental results through the Monte-Carlo tests show that the obtained parameter estimates through the separable algorithm are more accurate than those obtained by the Newton recursive estimation method without the model separation.

Keywords: Impulse response, Newton search, parameter estimation, recursive identification.