Data-driven modeling and optimization of the thermoforming heating phase

Thermoforming is a crucial technique in plastics manufacturing that relies on the expertise of skilled operators for effective process control. Changing operating conditions require frequent tuning of the thermoforming machine parameters to maintain the quality of the resulting product. However, this tuning often results in a waste of resources that should be minimized. Our work collects input-output data from experiments and investigates how to optimize machine parameters to maintain product quality. Specifically, we perform model identification using neural networks and solve an optimization problem for the heating phase. The results validate our approach, demonstrating its ability to obtain optimized recipe parameters that accurately replicate the desired thermal characteristics.