Extruder analysis, modeling and dynamic matrix control

The Modern Plastics extruder of the Material Science and Engineering Department is used to extrude high density polyethylene (Alathon 7040). Detailed Step test transfer function models are developed where the input variables are take-up speed, screw speed, gear pump speed, and die temperature set point. The output variables are die temperature, die pressure, end of the barrel temperature and pressure. A total of 18 transfer functions are obtained. Diameter sensor characterization is done due to the sensitivity of the sensor (Zimmer) and diameter variations observed in the sensor signal. This characterization reveal that the fiber diameter variation is mostly real variation. A typical fiber of 80 μm always shows a periodic variation with 4.1 second period and an amplitude of ±3 to ±6 μm at high sampling rates up to 400 samples per second. Nearly all of this variation, ±2 to ±5 μm, is real change in the fiber diameter. The relationships between known periodic inputs and the variables in the outputs are analyzed by the power spectrum analysis. Correlation established are between the gear pump rotation and die pressure variations, the screw rotation and barrel pressure variations, the zone 2 barrel heater on-off cycle and barrel temperature variations, and the die heater on-off cycle and tension. The periodic fiber diameter variations does not correlate to any input variation. A limited draw resonance study is undertaken and shows that the fiber diameter variation is perhaps from draw resonance, and thus is called in this work a draw resonance like phenomenon. Due to the dominance of the draw resonance like variation, the main control objective is to eliminate this variation. Classical control methods, proportional-integral (PI), integral (I), and the Smith predictor for dead time compensation with PI are the controllers, applied to control fiber diameter. In addition, the modern control method called dynamic matrix control (DMC) is also implemented. The results show that all the controllers can handle set point tracking but achieve only a limited reduction in the amplitude of the fiber diameter variation. A technique of using high and low-pass filters to characterize these two frequency components in the diameter variations helps in evaluating the controller performance. Further development of this technique has merit. It now seems that the controller adjusting the take-up speed at the frequency of the diameter variation to eliminate the variation also interacts with the draw resonance like phenomena. The control action is correct but the amplitude of the variation grows and is not eliminated. Suggestions are made for future work to address this interesting phenomenon.