Abstract

We present a methodology for inferring the molecular weight distribution (MWD) of polydisperse linear polymers from their linear rheology using machine learning techniques. Inferring the MWD from rheology would traditionally be prohibited by the complexity of state-of-the-art tube models, but our technique bypasses this by use of a machine learning methodology. Our approach utilises the model of Das and Read (2023) to generate large datasets of artificially produced rheology data. These data are used to train neural networks (NNs) to make MWD predictions from rheology measurements. We present a method for transforming the rheology data, using only the entanglement time and the plateau modulus, into a ‘universal’ rheology space. Here, any polymer chemistry can be readily analysed using the same approach. The NN is trained to predict the distribution in the number of entanglements in a polymer melt, and this is subsequently converted into the molecular weight distribution by knowledge of the mean entanglement molecular weight. Since we are interested in predicting the properties of industrially relevant polymers, we focus on broad MWDs. We present predictions for a range of polymer chemistries, including commercial samples of polyethylene (HDPE and LLDPE ). Good agreement with experimental gel permeation chromatography data is obtained.