Development of a grey-box model to understand and predict wear of coated cutting tools during turning

点击翻译按钮获取中文摘要

Today, the majority of cutting tools with a geometrically defined cutting edge are coated. Compared to uncoated tools, coated cutting tools enable longer tool lives, significant increases in performance and thus a more efficient chip removal. A reliable wear prognosis of coated tools, in particular of the transient failure behavior, is currently not possible. Therefore, in the research project, a deeper understanding of the stationary and transient system behavior of coated tools in high-performance machining is created. A reliable forecast of the onset of failure, the progress of wear and the remaining tool life should be made possible. For this purpose, experiments with coated cutting tools must be carried out, which characterize the initial condition of the cutting tools, record the wear behavior and thus create the database for the models. The deterministic model (white box) is then combined with a new data-driven model (black box) to a grey-box model. In this way, the purely deterministic non-describable temporal changes in the tool properties caused by wear and the end of the tool life can be described as a model. To create a database for understanding and predicting the tool wear, turning experiments are carried out, selected in situ measured variables are recorded and two- and three-dimensional wear parameters are continuously determined ex situ. In addition, a white box model (FEM machining simulation) is set up in order to determine the specific stress distribution in the cutting wedge, taking into account the respective state of wear. For a deeper understanding of the mechanical damage processes in the coating, suitable methods for coating characterization are applied and analogy experiments are carried out on the mechanical damage processes in the coating. Machine learning methods are used as part of a gray box model to predict the state of wear and possible coating failures. For this purpose, suitable measurement data features are processed using statistical data analysis, which are then implemented into the black box together with the data of the white box model. In the 2nd funding period, the grey-box model will be further refined by expanding the parameter space and the tribological elements. In this way, valid forecasts can be ensured. After the end of the second funding period, this enables knowledge-based qualification of coated tools for more efficient machining processes.