Data-driven development of a soft sensor for the flow rate monitoring in polyvinyl chloride tube extrusion affected by wall slip

Abstract In the manufacturing process of polyvinyl chloride (PVC) tubes, required thickness and weight depend on extruder flow rate. The setup can be very time-consuming inefficient since it requires adjusting screw rotational speed by trial error, as relation between rate is not known a priori. Furthermore, also affected material properties, melt temperature, pressure drop in die. Direct measuring or tube would require expensive gravimetric dosers X-ray systems, respectively. Therefore, soft sensor was developed to monitor its per unit length. Two alternative approaches are proposed predict under wall slip conditions: one based analytical model data-driven algorithms. Results show that machine learning regression models achieve high predictive performance (a relative error 1.2% using support vector regressor).