Data-Based Estimation and Control of a Multistage Press Hardening Process

In multistage press hardening processes, where a sheet material undergoes rapid austenitization, tempering, stretch-forming (SF), and die bending (DB), the resulting product properties are influenced by the thermo-mechanical history. This work aims at controlling the product properties of the formed blanks by making use of the model-based estimation and control of the spatial-temporal temperature distribution in the sheet. A data-driven dynamical model is constructed using dynamic mode decomposition (DMD) based on finite element (FE) simulation data. This model is further extended by means of parametric DMD to accommodate changes in process parameters like stroke rate, blank holder force, and austenitization temperature. The approach is validated and the model accuracy is improved through experimental analysis. The dynamics of the available temperature sensors are identified, whereupon a Kalman filter is developed based on the parametric DMD model to estimate the spatial-temporal temperature distribution. A time-varying, stage-dependent output matrix is employed to account for different numbers and locations of thermocouples in the three stages. Additionally, an optimal control strategy is implemented to achieve desired temperature trajectories, allowing targeted manipulation of the blank’s geometry and properties. Experimental validation of this system design and control strategy is carried out under real-time constraints.