WEB Generation of meta-models based on varying input data for the data-driven design of clinch joints.Tuesday (22.09.2020) 14:45 - 15:00 Z: Special Symposia II Part of:
Due to the increasing requirements on lightweight constructions, the demands for efficient joining processes are constantly rising. In order to cope with these requirements, cold forming processes provide a faster and less expensive alternative to established thermal joining methods. In order to guarantee the joining reliability, not only the selection of a suitable process, but also the design and dimensioning of the joint is crucial. While development engineers can rely on a wide range of design and layout principles in the field of thermal processes, especially welded joints, the evaluation of mechanical joining processes is mainly based on expert knowledge and less experimental tests. Consequently, the design of an optimal mechanical joining connection often requires high cost- and time-intensive adjustments on design or process parameters. As a possible solution, data-driven methods offer procedures for the structuring of data. Especially, meta-models provide an efficient means to approximate and examine relationships between product/process features and key variables of a technical system including a sufficiently accurate reproduction of all related information. Based on the generated regression models, analytical design functions can be derived that enable the prediction of investigated joint properties. Due to the limitations of the available data, the main challenge is to generate design functions that contain only relevant product/process features without a critical loss of the prediction quality. For this purpose, sensitivity and robustness analysis provide efficient methods to identify how key variables are affected on the changes in product/process features. As a result, an analytical design function with a significantly reduced complexity can be derived. In order to show the applicability of the procedure, the analysis of a simulated clinching process is used as an example. The presentation will give an overview of how meta-models can be generated based on varying input data and analytical design function can be derived afterwards.