Identification Experiments and Data Pre-Treatment
Identification is modelling based on experiments. If the experiments are not performed properly then no identification algorithm however sophisticated can arrive at a relevant model of the process. On the other hand, if the process input/ output data are collected from a well designed experiment, the simplest least-squares method can often deliver a good model. Therefore, experiments are the most important part in the total identification procedure. In the identification literature, much more attention has been paid to estimation algorithms than to the design of identification experiments. This often leads people to think that identification is just a set of algorithms and it is simply data in-and-model out.
KeywordsProcess Input Final Experiment Test Input Nyquist Frequency Pseudo Random Binary Sequence
Unable to display preview. Download preview PDF.