Comparison of different methods for identification of industrial processes

摘要: Plants have been modelled using different identification methods. Some results from the of dynamics a nuclear reactor, distillation column, superheater and paper machine are presented. Models for processes, obtained by maximum likelihood method, compared to models, other methods: least squares tally principle correlation analysis, spectral analysis. Comparisons made between model parameters, transient frequency responses models. It is pointed out that such comparisons may not be relevant performance control strategies, synthesized However, give insight into properties The problem choosing order parametric methods discussed. Simulations also used in compare on data known processes. method turns superior, especially low signal noise ratios.