Predicting Mechanical Properties of Polymer Films after Extrusion Coating using Supervised Machine Learning Algorithms

Sammanfattning: Tetra Pak is a world leader in providing innovative packaging solutions and processing technologies and has been so for a very long time. It is well – known that the material used for packaging purposes gets subjected to high temperature and pressure during folding and filling which could compromise the integrity of the material. The polymer layer in the packaging material acts as a barrier, and thus a break in the polymer layer will influence the barrier properties. The process by which the polymer layer is applied on to the paperboard is called extrusion coating. The barrier properties of the polymer are affected by the process conditions in the extruder. It is therefore important to understand how the processing affects the mechanical properties of the polymer and hence the barrier properties. In this thesis, polyethylene films are created by varying the process settings, line speed and melt temperature, in the extruder. The mechanical properties are measured by the use of tensile tests and the effect from processing on the mechanical properties are investigated. A model for predicting the mechanical properties of polymer films is built using the collected data. The primary tool for data processing used in this thesis is Python, and the predictive model is built using machine learning algorithms. In the data, there is clear and visible effect from line speed on the mechanical properties but effect from melt temperature is not as strong. The predictive capacity of the simplest model based on linear regression has been found to predict with highest accuracy. Predictive models built on random forest regression has also been found to predict fairly well. The complex models are overall sensitive and require more data than was collected in this study to provide reliable predictions.