Correlating polymer resin and end-use properties to molecular-weight distribution

The prediction of polymer resin and end-use properties for a dual-reactor solution polyethylene process is investigated. Polymer molecular-weight distribution (ABVD) is highly influential in achieving the desired properties, but the extent of the importance and key areas of distribution to achieve specific properties are less well understood. The best empirical approach for resin and end-use property prediction using the entire MWD along with other influential variables as inputs is investigated. Two modeling methods are considered: partial least squares (PLS) and a novel strategy that uses the weight fraction of polymer in a given molecular-weight range (referred to as a binning technique). Both linear and nonlinear variants of the two algorithms are used. The intention is to develop a model that facilitates the analysis of the simultaneous influences of process operating conditions and resin characteristics such as MWD on a specified set of end-use properties. Results demonstrate that the nonlinear valiant of the binning technique provides the highest accuracy as well as indicating regions of MWD that are of greatest influence. Such information is particularly useful for the control of the polymerization reactors.