Development of a data-driven model for the evaluation and improvement of process robustness in the die face design of deep-drawing dies

• Stochastic and long-term parameter fluctuations lead to failure or undesirable deviations from product specifications in the thermoforming process
• Stochastic parameter fluctuations (or process noise) are insufficiently taken into account in the design process
• Robust design of the active surfaces requires consideration of the process noise
• Increasing process robustness by modelling the process noise

The aim of the project is to develop a standardised explanatory model for the accelerated design of the active surfaces of deep-drawing tools while at the same time increasing the robustness of the manufacturing process with the help of artificial intelligence. Firstly, the process limits are determined using experiments and FE simulations, taking into account the process noise, and a database is created. The drawing edge length is selected as a quality feature. To increase process robustness, the metamodel is supplemented by an optimisation algorithm for the design of the active surfaces and process parameters. At the same time, data from clocked production processes is continuously collected and analysed during the project. The knowledge about the noise is used to make the deep-drawing process more robust against unpredictable influences and to be able to design the deep-drawing tool based on parameter predictions for the industrial application. The aim is to find out at which positions the drawing edge length should be recorded in order to achieve the most precise possible statement about the component quality. This data acquisition is used as the basis for transferring the model based on the simple geometry of a cross mould to complex industrial geometries.

The expected result of this project is a standardised explanatory model of process noise in deep drawing, which can be transferred to industrial applications. The basis for this is the use of the drawing edge as a quality feature, which is handled as an output variable when creating a database from experiments and FE simulations. The model created is then to be actively integrated into the process of mould surface design.