Comprehensive Surrogate Models for Predicting the Melt Conveying Characteristics of Channel Segments in High‐Performance Single‐Screw Extruders

ABSTRACT Plastics converters are increasingly considering high‐performance screws in single‐screw extrusion to meet demanding production requirements. In this context, reliable prediction models for melt conveying support the search for optimum designs without exhaustive experimental trials. Although full‐scale fluid dynamics simulations would closely represent the process, their high computational effort is rarely affordable in an industrial setting. Conversely, fast‐computing extruder calculations using surrogate models fail to capture the influence of channel curvature in multi‐flighted screw segments, thus becoming inaccurate for high‐performance screws. To overcome these limitations, we developed generic dimensionless analytical equations for the local pumping capability, viscous dissipation, and average shear rate in metering channels. First, a comprehensive simulation database for the flow of power‐law fluids through confined curved channel segments was generated that covers a vast range of designs and process configurations in single‐screw extrusion. Subsequently, this database was approximated by continuous functions using knowledge‐guided symbolic regression. The resulting characteristic equations excellently forecast the simulations throughout the entire design space with less than 3.5% average deviation, significantly outperforming existing approximations. When implemented in segmented extrusion calculations, these surrogate models enable quick and reliable statements on melt conveying in both conventional and high‐performance extruders for more efficient screw design and troubleshooting.