Nonlinear process modeling based on just-in-time learning and angle measure

Abstract

A new just-in-time learning methodology by incorporating both distance measure and angle measure is developed. This result enhances the existing methods, where the available information of angular relationships between two data samples is not exploited. In addition, a new procedure of selecting the relevant data set is proposed. The proposed methodology is illustrated by a case study of modeling a polymerization reactor. The adaptive ability of the just-in-time learning is also evaluated.