Model validation in iterative identification and controller design

The paper deals with the validation of model purposivity in iterative identification and controller design. It is known that when closed loop requirements are too high, the model resulting from the iterative procedure might conflict with the prior knowledge about the process. However, in some cases violated plausibility of the identified models does not necessarily imply its violated purposivity. Therefore, it is a matter of practical relevance to have a confident indication of whether the given model will result in a stable closed loop design or not. If not, the iterative identification and controller design should be stopped, i.e. more appropriate model structures should be chosen. In the paper a stochastic robustness measure is proposed which relies on the estimated model error obtainable by the stochastic embedding technique. In the simulated example we consider three models resulting from iterative procedure that have different qualities regarding plausibility and purposivity. It is shown that the stochastic robustness measure provides a reliable estimate of the designed loop stability.