Learning process adaptation for tool grinding

• Applicant: Professor Dr.-Ing. Berend Denkena
• Subject Area: Metal-Cutting and Abrasive Manufacturing Engineering
• Term: from 2020 to 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 445811009

The grinding process within the process chain for the production of solid carbide tools causes up to 60% of the manufacturing costs, defines the geometry and influences the substrate properties of the tool. For these reasons, the grinding process plays a central role in the process chain. The increasing number of variants and smaller batch sizes require an efficient process design and as few adjustment attempts as possible. Process planning has a considerable influence on the process result, both in terms of the resulting quality and the costs incurred. Mechanical and thermal loads act on the workpiece during the grinding process. These can lead to shape deviations and thermal damages. These effects can be compensated by selecting suitable parameters and adjusting the NC code. As a result of complicated process kinematics and the large number of process control variables, process planning is usually in the responsibility of the employee and in many cases running-in processes are required. According to the literature, automation systems and self-adapting grinding machines can be taken in order to counteract the increased costs in process planning. Approaches include the feedback of process information, the use of simulation systems and the application of machine learning methods to build process models. The main objective of the proposed project is the research of an automated process adaptation using a process-parallel material removal simulation and learning process models for grinding twist drills. A special focus lies on the consideration of the process force-induced displacement of the tools to be ground. A method for the process-parallel simulation of engagement conditions during tool grinding and the subsequent data fusion with measured process and process output variables is developed. This is followed by research into learning process models. With the achievement of this subgoal, fundamental knowledge about the possibilities and limits of methods of machine learning for modelling process force-induced displacement, roughness and edge zone damage during tool grinding is available. A method for optimizing the initial process planning will also be developed and investigated. The process control variables have to be adjusted and potential displacement processes have to be compensated by adapting the tool path. Finally, the method is researched and transferred to the grinding process of the tip and shell geometry for learning process adaptation during the grinding processes of an entire tool.

• DFG Programme: Research Grants