In injection moulding, controlling process variables instead of machine variables can improve productivity and quality. In the ‘Thermoset control’ project, this concept was applied to thermoset processing for the first time – in collaboration with the IRT and the IKV.
© IKV In injection moulding, process control can increase productivity and part quality by controlling process variables instead of machine variables. In the thermoset control project, such control via process variables, which previously only existed for thermoplastics, was also implemented for thermoset processing. The control system was developed with the Institute for Control Engineering at RWTH Aachen University and was implemented and analyzed in close cooperation at IKV. First, the cavity pressure curve was derived from injection moulding tests as a suitable control variable for real-time control of the process. To this end, the component quality was evaluated in terms of weight, mechanical properties and surface quality and correlated with the process variables.
Model predictive control (MPC) of the cavity pressure according to a specified reference pressure curve was implemented by adjusting the screw speed via an external, analog voltage signal on the servo inverter of the injection molding machine. The reference pressure specified by the operator could be regulated precisely and without overshooting by controlling the screw speed by specifying the control voltage. The injection molding process was described with the special material properties of the thermosets using a physical model, taking into account the volume flows and reaction kinetics, in order to incorporate process knowledge into the control system.
By determining the component mass at three different pressure levels, a Gaussian process regression model (GPR) could be trained without initial process knowledge. The Gaussian process regression model identified a correlation between the cavity pressure reference and the measured component mass. Various phenolic and epoxy resins were tested as materials. Process fluctuations in the injection molding process were simulated by varying the machine settings. The process model was continuously improved through machine learning during production. The MPC achieved good results in the tests and is superior to a simple proportional-integral-differential controller (PID). Measured by the cross-cycle consistency of the cavity pressure integral and the component mass, higher process stability was achieved by controlling the cavity pressure.
Funding and project partners
We would like to thank the BMWK for funding the IGF project (funding code 22193 N) and the project partners for their cooperation.
Project duration: 1.3.2022 to 30.11.2024
