• Journal of Institute of Control, Robotics and Systems
• /
• v.12 no.7
• /
• pp.706-712
• /
• 2006

Recursive Partial Least Squares(RPLS) method has been used for processing the on-line available multivariate chemical process data and modeling adaptive prediction model for process changes. However, RPLS method is unstable in PLS model updating because RPLS method updates PLS model by merging past PLS model and new data. In this study, Adaptive Predictive Partial Least Squres(APPLS) method is suggested for more sensitive adaptation to process changes. By expanding APPLS method, block-wise Adaptive Predictive Partial Least Squares(block-wise APPLS) method is suggested for a lager scale data of chemical processes. APPLS method has been applied to predict the reactor properties and the product quality of a direct esterification reactor for polyethylene terephthalate(PTT), and block-wise APPLS method has been applied to predict the cetane number using NIR Diesel Spectra data. APPLS and block-wise APPLS methods show better prediction and updating performance than RPLS method.

• 제어로봇시스템학회:학술대회논문집
• /
• 1988.10a
• /
• pp.225-230
• /
• 1988

In this paper, a new method for determining the order of AR, ARMA processes based on PLS (predictive least square) principle is proposed, This method using modified lattice algorithm which has additional step is amenable to on-line or adaptive operation and is more accurate than any other mpthod. Some computer simulations are presented to show the efficiency of the proposed algorithms.