In helping to acquire a measurement of the characteristics of one or more polymers online NMR systems, the use of squares regression technique partial least Neural Networks

摘要

Related methods (57) [Abstract] online and nuclear magnetic resonance (NMR) systems are useful in predicting the properties of one or more of the problems of the polymer. In one embodiment, by using a neural network, also take the correlation of process variables in addition to the NMR output was operated to expand the model to predict the properties of the polymer in question. In another embodiment, deploying the model using a partial least squares regression techniques, with improved accuracy. Can be used in combination with optimal model selection method or multi-model described according to the present invention, one of the partial least squares regression techniques or neural network techniques. Polymer may be a rubber such as ethylene-propylene rubber or plastic, such as polyethylene, polypropylene or polystyrene.