CONTROL-RELEVANT DISTURBANCE MODELING OF PAPER-MACHINE FULL PROFILE PROPERTIES USING ADAPTIVE PCA

In this paper we use the Karhunen-Loeve expansion to identify on-line the significant modes of a paper machine foil property profile. We als o show how the corresponding temporal eigenvectors can be modeled usin g low order linear autoregressive (AR) processes. Several parameters a ffecting the profile predictions, such as the order of the AR model an d the window size of the data used in the KL expansion, are discussed through a simulation example.