Analytical approach for parameter identification in machine tools based on identifiable CNC reference runs

As a result of the steadily growing importance of data-driven methods such as digital twins, approaches for automated parameter identification in production equipment are becoming increasingly important. Previous work has shown that AI-based approaches for classification are increasingly reaching their limits. As a result of new developments, CNC reference runs with a high information content that can be specifically identified via an ID can be generated. In this context, it was possible to achieve an oscillation state on a test machine that is particularly well suited for identification. In this paper, an analytical approach is presented which, in addition to classification, can assign the signal to the respective source and therefore establish interdependencies between signals. Here, on the test machine tool, with successfully excited oscillations, all signals could be classified and assigned via the ID with high accuracy. If the oscillation state cannot be reached, classification accuracies of over 90% could be achieved, depending on the motion generation.