Improving parameter estimation using minimal analytically redundant subsystems

This work presents a novel parameter estimation approach for system modelling based on model decomposition. This approach uses Possible Conflicts to decompose the system model into minimal submodels that are used to obtain minimal parameter estimators for non-faulty situations. A laboratory plant was used to test the approach. The results obtained were compared against two classical parameter estimation techniques, the SQP optimization method and a curve-fitting approach using non-linear least squares. Both classical approaches use the global simulation model of the plant to carry out the optimization. The properties of the three techniques are presented and discussed. The developed parameter estimation approach improves the results obtained with the cited classical approaches.