A Novel Kernel method for Dynamic Process Identification

Abstract

This paper proposes a new kernel method for dynamic process modelling. The developed algorithm is titled Adaptive Reduced Kernel Partial Least Squares (ARKPLS). The developed ARKPLS uses the Reduced KPLS technique in an offline scenario in order to build a model which have a small parameter number after that, the number of the retained parameters are update in an online scenario. The suggested technique has been used to identify a nonlinear process.