Towards flexible production of quality-critical injection molding parts via iterative learning control

In this work, we address the practically relevant challenge of transferring an already commissioned injection molding process from one machine to another. Traditionally, such transfers are performed manually by highly skilled personnel under significant time pressure. We propose a two-stage learning strategy that facilitates automated process transfer by exploiting measured reference trajectories of antechamber volume and pressure. In the first stage, an iterative learning controller adapts dosage volume and decompression stroke to reproduce the initial antechamber melt state. In the second stage, a model-based norm-optimal iterative learning controller adjusts the feedforward input of the existing control system to accurately track the reference trajectories. A blended error formulation enables a smooth and tunable transition between volume and pressure tracking. The method requires no machine- or mold-specific experiments, relying only on commonly available parameters. Experiments with a challenging test mold demonstrate fast convergence within a few cycles and accurate trajectory tracking during the transfer from a hydraulic to an electric injection molding machine, resulting in molded parts of high quality.