Training times and failure ratios of knowledge-guided symbolic regression models for melt conveying in single-screw extruders

Excel spreadsheet with numeric values and textual description.

Contains the individual training times for symbolic regression models to predict the dimensionless melt flow rate in single-screw extruders, as well as the failure ratios and expected training times across all repetitive runs for obtaining satisfactory solutions. The models were created from numerical simulation data with four different cases of integrated domain knowledge:

• Case 1: theory of similarity only,
• Case 2: additional derived input features for pure pressure flow,
• Case 3: logarithmic scaling of the dimensionless flow rate with channel aspect ratio as derived input feature,
• Case 4: theoretical approximation equation for superimposed drag and pressure flow in dimensionless space.

Solutions were regarded satisfactory in case of smoothness and prediction accuracies within three times the standard deviation from the group mean.