Part-mass control in injection molding of recycled thermoplastics by learning-enabled model predictive cavity-pressure control

Post-Consumer-Recycled (PCR) thermoplastics exhibit inconsistent material properties across batches due to impurities and material degradation. Achieving consistent part-quality attributes in processing different batches of PCR requires continuous adjustment of the state-of-the-art injection molding process. In our work, we present a learning-enabled nonlinear model predictive controller (NMPC) for cavity pressure that updates its model after each injection molding cycle, combined with a learning-enabled part-mass controller that serves as its reference generator. Within the NMPC, we use a physics-based model of ordinary nonlinear differential equations. The model parameters are updated between each injection molding cycle using a sequential quadratic programming approach. We incorporate constraints into the NMPC to prevent issues such as cavity-pressure peaks. The model used inside the part-mass controller is a Gaussian process regression model that leverages a cycle-variant kernel function to account for varying material properties. We test the proposed control algorithm on a plate-mold geometry, processing both virgin polypropylene and multiple batches of PCR material. While transitioning between virgin and two PCR batches over 50 production cycles without interrupting the injection molding process, the NMPC model and the cavity-pressure reference are automatically adjusted, maintaining a mean part-mass deviation of 0.21 % relative to the part-mass reference. The results show strong potential for automated process-model adaptation and part-mass control when transitioning between virgin material and different PCR batches. • Cavity-Pressure NMPC For Processing Post-Consumer-Recycled Thermoplastics. • Automated Process-Model Adaptation when Transitioning between different Materials. • Model-based Part-Mass Control during Batch Transitions of PCR.