Soft Sensors for Predicting Sheet Quality Properties During Paper Manufacture, 2001 Engineering / Finishing & Converting Conference Proceedings
S. Budaraju--Critical sheet quality properties are generally available only after a paper reel is made and a sample is taken for
analysis. Virtual sensors, based on artificial neural networks, can predict these properties in real time during the
build up of the reel. This would help the operators make the correct decisions, without waiting for the lab results
and thus reduces the amount of rejects. Building a virtual sensor or a property predictor requires extraction of the
historical process data, careful data preprocessing, model training and validation, and finally on-line
implementation. A step-by-step approach to building property predictors will be presented here with examples.
Soft Sensors for Predicting Sheet Quality Properties During Paper Manufacture, 2001 Engineering / Fi
