Experimental Validation of Prediction Models of Pressure‐Throughput Behavior for Melt‐Conveying Elements in Fully Intermeshing Co‐Rotating Twin‐Screw Extruders

ABSTRACT

Modeling and simulation of the melt conveying of co‐rotating twin screw extruders, one of the important manufacturing machines in polymer processing, is commonly done by means of generalized, dimensionless models. Recently, novel general hybrid models have been published, which combine physics‐based approaches with extensive simulation and machine learning. This article experimentally validates the pressure‐throughput behavior prediction of these models for four different polymer grades, three screw element geometries and a large range of extruder setpoints. An automated data processing evaluates the data, resulting here in an average prediction error of 7.3%, and showing a close agreement between the models and the measurements. However, subsequent uncertainty analysis shows that large measurement noise raises the mean uncertainty of the processed experimental data to 14.5%. Furthermore, we show that the experimental results exhibit large sensitivity to fluctuations in viscosity, which is further very sensitive to temperature and assumes a constant representative shear rate. Most operating points suffer from having a wide‐range temperature distribution across the melt channel, which is shown by numerically finding representative cup‐mixing temperature using the Monte Carlo approach. Despite these limitations, the results demonstrate the hybrid models to agree well with experimental data and outperform existing analytical modeling frameworks.