A Hybrid Approach for Modelling Polymer Processing Problems
Sprache des Vortragstitels:
Englisch
Original Tagungtitel:
Polymer Meeting 14
Sprache des Tagungstitel:
Englisch
Original Kurzfassung:
Modelling and simulation are essential for predicting and understanding polymer processes. Models are used for troubleshooting, as well as for process improvement and optimization. Traditionally, modelling polymer processing includes analytical and numerical methods. Both require various assumptions to create a simplified representation of the reality. Analytical methods provide closed-form mathematical expressions of the problem, but are limited to a few idealized problems. Numerical methods can be applied to much more complex problems yielding approximate solutions. Thus, reducing the assumptions, the reality is represented more accurate. With the explosive growth of available data, the importance of data-based modelling has grown enourmously over the last few years, and it offers completely new potentials in polymer processing.
We present a new hybrid approach, that incorporates analytical, numerical, and data-based techinques into one integrated method. Hence the advantages are combined and limitations of the single methods omitted. Our proposed strategy covers following steps: first, a fundamental analysis of the governing equations and a dimensional analysis are performed; second, based on these findings, a parametric design study is set up, which design points are solved numerically; third, an analytical model is developed by applying symbolic regression based on genetic programming; and, finally, the model is validated. We originally applied this hybrid approach to model the flow of polymer melts in single-screw extruders. Currently, we are investigating twin-screw extrusion and co-extrusion processes. This approach, however, is not limited to polymer melt flows. In fact, it can be used in any engineering discipline (e.g., solid-mechanics, and electro-dynamics).