Project Details
Projekt Print View

Determination of the correlation between pressure fluctuations, thickness fluctuations, melting degree and product quality in extrusion

Subject Area Plastics Engineering
Term from 2020 to 2024
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 452432838
 
Final Report Year 2023

Final Report Abstract

Cost efficiency and product quality are key issues in extrusion. Special attention is paid to the design of extruder screws, avoiding the use of prototypes, which are time consuming and costly to produce. Extrusion requires high throughput to ensure economic efficiency, and the quality of the polymer products produced must meet specified standards. The cost efficiency of the extrusion process is influenced by several factors, including process parameters such as screw speed and barrel temperatures, as well as the plasticizing and homogenizing performance of the extruder. The main objective of the project is to identify correlations between process parameters and melt quality. Product quality is directly dependent on melt homogeneity. Different screw concepts and geometries were investigated, including barrier screws and standard three-zone screws with and without different shear and mixing section combinations. The challenge is to find an optimal extruder design without having to rely on costly prototypes. This is where experience and simulation come in. However, current simulations do not provide reliable information on the expected product quality. Complex experimental studies are therefore required after the manufacturing process to determine product quality. Materials such as Lupolen 2420D (PE) and Moplen HP420M (PP) were used in the project. These materials were selected due to their high processing volume in the industry. The Screw Performance Index (SPI) developed by DÖRNER evaluates the performance of the extruder screw through pressure and temperature fluctuations. Correlations between thermal and material homogeneity and the SPI were identified to reduce time-consuming investigations. For the modeling, the investigation plan was simulated in the REX software developed at the KTP. The results are used as input variables for the model. A self-learning algorithm has been developed that outputs a regression model for predicting melt quality. The model takes into account linear and quadratic terms as well as the interactions of the input variables. The results of tenfold cross-validation show a strong linear relationship between the input variables and the SPI. The correlation between predicted and actual values and the mean squared error (MSE) are used to evaluate the accuracy of the model.

Publications

 
 

Additional Information

Textvergrößerung und Kontrastanpassung