Using Machine Learning to reduce injection molding's sensitivity to variability in recycled materials

The use of post-industrial and post-consumer recycled feedstocks in injection molding causes variations in material properties such as viscosity, solidification temperature, density and thermo-mechanical properties such as the thermal expansion coefficient. These in turn cause variations in part quality (shrinkage and warpage) and process variables (e.g. maximum injection pressure).

In this study, we present a methodology for predicting the effect of material variations on part quality. This methodology is then also used to identify the process conditions where the material variation can be tolerated while the variation of part quality is reduced. In addition, the effect of modest geometry changes (e.g. flow leaders) to reduce the part quality sensitivity to material variation is also predicted.

The methodology employed includes an analysis of part quality over the range of expected material properties variations, and the use of machine learning to identify the optimum process conditions which minimize part quality sensitivity.