Using analytical and data-driven methods to develop a soft-sensor for flow rate monitoring in tube extrusion

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