Experimental validation ofnon‐Newtonianstratifiedco‐extrusionprediction models using a digital process twin

Abstract

In the co‐extrusion of plastics, pressure‐throughput behavior, layer distribution, and residence time are crucial parameters, modeling of which contributes to manufacturing high‐quality products at optimized process efficiency and significantly shortens development times for die systems. In previous work, we have presented symbolic regression models to predicting the (i) pressure‐throughput behavior, (ii) position of the interface, (iii) interfacial shear stress, (iv) ratio of volume flow rates, and (v) interfacial flow velocity for isothermal two‐layer co‐extrusion flows through rectangular ducts. These regression models are mathematically simple and capable of capturing the shear‐thinning nature of polymer melts without the need for numerical methods. Here, we present an experimental study validating the proposed models against co‐extrusion process data and comparing them to existing theories. To this end, a two‐layer co‐extrusion demonstration die instrumented with an optical coherence tomography sensor for detecting the interfacial position was used. To accurately set up and evaluate the die flows, the overall co‐extrusion process was represented by means of a digital process twin. Industrially relevant combinations of materials were tested under a wide range of processing conditions. Comparisons of pressure losses and interfacial positions to the predictions showed excellent agreement and the results outperformed the concept of representative viscosity.

Investigation of interdiffusion between compatible polymers under static and co‐extrusion processing conditions

“…Since these process parameters are difficult to measure directly, we used a soft‐sensor approach to accurately adjust and evaluate them. These soft sensors were based on symbolic regression models integrated in a digital process twin to fully digitalize the overall co‐extrusion process as described thoroughly in our previous work 20,21 . This means that various process quantities that correlate with the target quantities are measured continuously (such as pressures and melt temperatures) and are subsequently used as input to the symbolic regression models.…”

“…This study used conventional single-screw extruders with a smooth intake zone, which means that the throughput mainly depends on screw rotational speed and back pressure. Throughput tests with an adjustable back-pressure valve, as described in detail in 21 , were carried out. To determine this relationship (i.e., the screw characteristic line) for SAN-19 and HI-PMMA, screw rotational speed and back pressure were varied systematically, and for each set of operating conditions the mass throughput was measured.…”

“…These soft sensors were based on symbolic regression models integrated in a digital process twin to fully digitalize the overall coextrusion process as described thoroughly in our previous work. 20,21 This means that various process quantities that correlate with the target quantities are measured continuously (such as pressures and melt temperatures) and are subsequently used as input to the symbolic regression models. Further, the viscosity models and the Tait equation were evaluated (considering the mass throughputs, melt pressures and melt temperatures) and also used as inputs to the symbolic regression models.…”

Multi‐source data‐driven approach for prediction of melt density during polymer compounding

“…Extrusion processing involves the use of specialized machinery to transform polymer materials into a variety of shapes and forms. This stage of fabrication is critical for producing high‐quality polymer products that meet the desired specifications and performance requirements 1 . The extrusion process requires careful monitoring of various parameters, such as temperature, pressure, and flow rate, 2 to ensure that the material is processed correctly and consistently.…”