Quality monitoring of the rubber mixing process by using the discounted-measurement RPLS algorithm

Mooney-viscosity is the most important quality index of the compounds. It\´s stability and uniformity are the right items the rubber industry wants to improve urgently. It is very economical and practical to build the viscosity prediction model of the mixture to realize the automatic quality control of the mixing process by using multivariate statistical techniques. A discounted-measurement recursive PLS (DRPLS) algorithm is proposed to extract prediction model with enough freedom degree and to overcome the "data-over fitted" phenomena. Furthermore this improved RPLS model could be on-line updated itself with new data without increasing computer memory usage.