Inline optimization for injection molding processes for abrupt and gradual process behavior alterations

AIP conference proceedings, vol. 3158, pp. 020003

Abstract

An automized anomaly and process condition drift recognition in injection molding could significantly improve the inline optimization of injection molding processes.Two different algorithms, Adaptive Windowing (ADWIN) and a custom approach based on the Jenson-Shannon-Divergence metric between two probability distributions (IMIDD) are evaluated for drift indication with the process parameters maximum injection pressure, dosing time and remaining melt cushion.A test dataset with abrupt and gradual changes is generated and comprised of data with varied machine settings to induce drifts in the process parameters.For ADWIN, several different hyperparameters are evaluated to examine the predictive quality.While an abrupt change is identified shortly after occurence, a timely reaction to gradual changes and especially both gradual and abrupt changes cannot be assured.IMIDD shows prompt reaction to an abrupt change and acceptable delay for the notification of gradual changes.Challenges remain, inter alia false-positive drift indications or unrecognized drifts.This may be mitigated by definition of an inertia time for the algorithm.In future work, IMIDD will be further explored and used as indicator for necessary injection molding process adjustments.

Authors 2

1. Yannik Lockner corresponding Aachen Institute for Plastics Processing in Industry and Craft

   RWTH Aachen University

   Affiliation as printed

   Institute for Plastics Processing in Industry and Craft at RWTH Aachen University, Aachen, Germany

2. Christian Hopmann Aachen Institute for Plastics Processing in Industry and Craft

   RWTH Aachen University

   Affiliation as printed

   Institute for Plastics Processing in Industry and Craft at RWTH Aachen University, Aachen, Germany