Neural Networks: From Image Processing to Process Monitoring

Artificial intelligence (AI) is increasingly important in production technology. It offers the potential to train machines to optimize themselves and detect process errors. A networked machine park opens the possibility of optimizing entire production systems. However, there are still challenges, particularly in small and me-dium-sized manufacturing companies (SMEs), e.g., due to heterogeneous data from many produced compo-nents. Metadata, such as CAD files, are also not always available. In the "Cycle Optimization" demonstrator of the IIP Ecosphere project funded by the German Feder-al Ministry for Economic Affairs and Climate Action, an AI-based parameterization and monitoring for machin-ing cycles are therefore being developed that works for individual parts. Machining cycles control the motion sequences between the tool and workpiece during complex machining operations. Examples of cutting cycles are the turning of threads, the milling of form features, or the power skiving of gears. The power skiving process considered as an example enables particularly productive production of gears. In this process, a rotating tool at high speed "peels" a gear into a workpiece rotating in the same direction (Fig. 1). This process offers great potential, especially for SMEs, because the process can be carried out in con-ventional turning and milling centers, and no special machines are required.