In addition to the current requirements in the field of machining, the trend towards batch size one and the increasingly complex workpieces constitute further challenges. Highly automated machine tools are used for machining, which require a high degree of process understanding for cost-efficient use. Particularly with the emerging smaller batch sizes, production is carried out with conservatively selected process parameters far below the productivity optimum. Due to the highly varying customer-individualized products, the development of sustainable process understanding on the employees' side usually remains lacking, with the result that potentials remain unused despite many years of experience. Current systems for supporting employees at machine tools are geared towards avoiding downtimes and damage, but not towards increasing productivity by optimizing all process parameters in small batch and individual part production.
This is the starting point for the research project AICoM – Artificial Intelligence Controlled Milling. The aim is to develop a learning machine tool for metal-cutting production with the ability to autonomously adapt the process by using learned “knowledge” or learned “experience”. For manufacturing, a 3D model of the workpiece to be produced, including quality requirements, is transferred to the machine tool and the component is subsequently manufactured, taking into account the target value selected by the user, such as maximum productivity or maximum workpiece quality.
To be able to react to any changes of process and machine conditions, the core AI of AICoM is enabled to adapt not only the process parameters but also the previously calculated path points during the manufacturing process. The close-to real-time control loop thus targets an in-process adaptation.
The required information is provided to AICoM by processed machine-internal signals as well as external sensor data, which are used for the acquisition of further relevant signals. These enable the process status to be captured as precisely as possible and are preprocessed automatically in the Preprocessing & Data Acquisition module depending on the requirements for the control loops.
In addition to the processed sensor signals, the calculated variables serve as input variables for the artificial intelligence in the AI data processing module for automated optimization of the technology parameters depending on the requirements and target variables based on the knowledge learned by the AI.
The research project is sponosored by the German Federal Ministry of Education and Research (BMBF). We thank for the opportunity to work on this project.