Identification of linear dynamic systems by short parts of transients with additive measuring disturbances (Q1386951)

The author considers identification of a parameter vector contained in a model of an object that is given by a system of equations of autoregression and moving average (ARMA), which has vectorial input and output signals and for which the coefficient matrices are affine functions of the parameter vector. He uses a variational method to identify the parameter vector by use of measurements of short parts of the transients taking account of additive measuring noise. A software implementation for systems of ARMA with vectorial input and output signals is described. Moreover, under certainty assumptions, he proves that the estimate of the variational method is strongly consistent, asymptotically normal and unbiased as \(L\) goes to infinity, where \(L\) is the number of measurements. Finally, the results of a numerical simulation are presented.