Tube-Based Model Predictive Position Control of Machine Tools (tMPC-WZM)

• Applicant: Professor Dr.-Ing. Alexander Verl
• Subject Area: Production Automation and Assembly Technology
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 562673097

In addition to the mechanical design, the position control of feed axes is a key factor in influencing the dynamic accuracy and therefore the quality of manufactured workpieces. Improvements in control are therefore a constantly topical field of research. Model predictive control (MPC), which solves an optimization problem about the future predicted dynamic behavior of the machine at each time step, shows great potential for improving dynamic accuracy. As the optimization problem must be calculated within the short cycle times of the CNC (less than 1 ms), a greatly simplified dynamics model of the machine must be used for this. Using MPC, the robust stability against model simplifications or even milling forces cannot be formally proven, which is however indispensable for the industrial use of the process, because it cannot be ensured that the position controller remains stable during machining. The Tube-Based MPC is therefore to be investigated in this application. This stabilizes the deviation between the simplified model and the actual system behavior via an additional feedback law and can be specially designed for robust behavior. The additional effort for the robust design occurs offline, so that the online calculation can continue to take place in the real-time cycle of the controller. In our own preliminary work, a procedure for robust parameterization of the controller is presented, which reduces the complexity to three setting parameters for which setting guidelines have been developed. Furthermore, the functionality and robustness of the controller will be demonstrated by simulation. Building on the simulative findings, the practical validation of the controller on three feed axes of a machine tool will take place in the proposed work program. The integration of the optimisers into the real-time system will be investigated, the determination of model uncertainties from measurements will be automated and the control and disturbance behavior will be compared with other controllers (including the standard MPC and regular cascade control). Milling experiments and further robustness tests are also part of the investigations. At the end of the proposed project phase, the synthesis of the controller will be automated, except for the specification of the three setting parameters, and both the performance and robustness of the controller will have been experimentally proven and analysed in detail.

• DFG Programme: Research Grants
• Co-Investigator: Dr.-Ing. Armin Lechler