In order to handle increasingly complex engineering applications with highly nonlinear behaviors, various advanced system identification algorithms have been developed for control and diagnostic purposes. Since the performance of these algorithms depends significantly on the selection of input variables, a systematic input selection methodology is needed to identify the nonlinear relation between the input variables and system outputs, even in the presence of high correlation among the candidate input variables. The methodology proposed in this paper converts the problem of selecting appropriate input variables for the identification of a nonlinear dynamic system into one of a set of properly linearized models. In order to enable the approximation of the nonlinear system behavior with a set of linear models, a growing self-organizing map is employed to appropriately partition the system operating region into sub-regions via unsupervised learning. Evaluated based on the minimum description length principle, a model with its most related input variables is then selected using a genetic algorithm so that the computational burden can be reduced. The effectiveness of this methodology has been demonstrated with two simulation examples and a real-world example of modeling the air mass flow rate and intake manifold pressure in a diesel engine airflow system.

This content is only available via PDF.
You do not currently have access to this content.