On Identification of the DMC Models: Asymptotic and Simulation Results

The parameter identification problem for the Dynamic Matrix Control (DMC) model is considered in this work. A recursive algorithm for parameter estimation is proposed. Some asymptotic properties of such an algorithm is obtained. It is demonstrated that the algorithm is strongly convergent and a suitably scaled error sequence has normal limiting distribution. Consequently, the asymptotic normality is used to build up interval (confidence region) estimates. A stopping rule is developed. Numerical experiments have been conducted. The simulation results conform the limit theorems. The proposed stopping rule is proved to be practically useful. Moreover, the procedure is quite easy to implement.