Identification of Nonlinear Nondynamic Systems with Application to a Hot Steel Rolling Mill

Abstract The problem of obtaining information about the structure of an unknown nonlinear system is considered and an algorithm is presented for identifying the structure of a nonlinear nondynamic system from a known set of nonlinear functions. The algorithm selects one or more of these nonlinear functions for the final model. It is shown that for certain categories of functions the algorithm may select a suboptimal model. An analysis of this selection error is presented along with a method for assigning a probability to its occurrence. In addition, the method can be used to reduce the order of an existing nonlinear model. This identification algorithm is used to derive roll force setup models for a hot steel rolling mill.