Automated Deconvolution of Polyolefin MWDs: A Heuristic Approach Based on Flory Most Probable Distributions

Despite the popularity of molecular weight distribution (MWD) deconvolution methods in polyolefin reaction engineering, no systematic procedure has been devised to generate initial guesses for the deconvolution parameters. MWD deconvolution identifies the minimum number of Flory most probable distributions (MPDs) needed to describe the MWD of polyolefins made with multiple-site-type catalysts by fitting two or more Flory MPDs to the experimentally measured data. Current deconvolution algorithms rely on nonlinear least-squares (NLLS) methods and require first guesses for the number-average molecular weight (Mnj) and mass fraction (mj) of the polymer populations. This article introduces two heuristic rules that provide systematic first guesses for Mnj and mj, an essential step toward the automation of MWD deconvolution. We also propose an adaptive weighting strategy that incorporates molecular weight averages into the objective function without compromising convergence. Together, these developments demonstrate that MWD deconvolution can be made not only reliable but also predictable, opening the way for its use in catalyst screening, process optimization, and high-throughput experimentation.