Model Predictive Force Control in Milling

Key Info

Basic Information

Duration:
01.11.2012 to 31.12.2018
Organizational Unit:
Chair of Manufacturing Technology, Cutting Technology
Funding:
German Research Foundation DFG, Cluster of Excellence Integrative Production Technology for high-wage countries
Status:
Closed

Research partner

  • Institut of Automatic Control IRT, RWTH Aachen University

 

How can the milling process be operated at its instantaneous maximum productivity while guaranteeing product quality and process safety?

For this purpose, an advanced control method was examined within the excellence cluster "Integrative Production Technology for High-Wage Countries" in cooperation with the Institute of Automatic Control (IRT): model predictive control (MPC). The MPC controls the active force staring as a unique development starting for two-axis roughing milling, now representing a strategic development at the WZL. It represents a paradigm shift in manufacturing away from a control of machine tool set-points (e.g. position) to a control of process parameters (force). The objective is to adjust the feed rate so that the force remains as close as possible to a given reference, but never exceeds it. In this way, the deflection of the cutter can be limited, thus guaranteeing that tolerances are met. To achieve this, the MPC solves a constrained optimization problem in each step by means of a short future horizon of the tool path. This is possible because the engagement conditions along the toolpath can be determined by an engagement calculation. With the help of a model of the machine behavior and of the process force, the immediate future can be predicted. The model of the process force must be identified at runtime to take into account the progression of tool wear. The system enables to safely operate the milling process at its technological limit.

In future projects the control system will be extended with regard to the five-axis milling process. It is also desirable to measure the force indirectly in order to identify the process model at runtime.