The application of system identification to an electrohydraulic servo system is described. Identified models are used to design pole-placement controllers for the system. A variety of parameter estimators and model structure selection techniques are compared by assessing the performance of the controllers designed from the estimated models. The least squares method processing filtered data is found to yield reliable models, and an appropriate model structure can be selected successfully by comparing prediction errors for models of different structure.
Model structure selection for switched NARX system identification: A randomized approach