Identification of ‘ moving average’ plants under unobservable disturbances

From MaRDI portal
Publication:3691561





The author considers the identification problem for the moving average plant described by the equation \(Z(n)=\xi (n)+\sum^{m}_{k=1}d_ k\xi (n-k)\); \(\xi (s)=0\), \(S<0\), where \(\{\) \(\xi\) (n)\(\}\) is a sequence of independent random values with \(M\{\xi (n)\}=0\); \(M\{\xi^ 2(n)\}=\sigma^ 2\); \(\xi\) (n)\(\in R\); \(d_ k\) \((k=1,...,m)\) are the unknown plant parameters to be estimated. Only output values z(k) \((k<n)\) are assumed to be observable; the input values of \(\xi\) (k) are unobservable. This problem features all the complications of identifying a plant exposed to closured noise. Unlike the conventional methods a method for identification of such plants without a priori information as to whether the plant is minimum or non- minimum phase is presented. The author introduces an equation set for the plant parameter vector. The set differs from the conventional case in using both pairwise correlation and a higher (fourth) order correlation. The asymptotic properties of the algorithms obtained by this method are examined. Classes of plants are specified for which these algorithms lead either to strongly consistent estimates of the plant parameters or to mean-square convergence to the true parameters. The rate of convergence is analysed.











This page was built for publication: Identification of ‘ moving average’ plants under unobservable disturbances

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3691561)