Data-driven analysis of cold-formed pin structure characteristics in the context of versatile joining processes

Due to increasingly strict emission targets and regulatory requirements, especially for companies in the transport industry, the demand for multi-material-systems is continuously rising in order to lower energy consumption. In this context, mechanical joining processes offer an environmentally friendly and flexible alternative to established joining methods, especially in the field of lightweight design. For example, cold-formed cylindrical pin structures show high potentials in joining multi-material-systems without auxiliary elements. The pin structures are joined either by pressing them directly into the joining partner or by caulking with a pre-punched part. However, to evaluate the strength of the joint and to ensure the joining reliability for versatile processes, such as changing joining partners or batch variations, engineering designers currently have only limited design principles available compared to thermal joining processes. Consequently, the design of an optimal pin joint requires cost- and time-intensive experimental investigations and adjustments to design or process parameters. As a solution, data-driven methods offer procedures for structuring data and identifying dependencies between varying process parameters and resulting pin structure characteristics. Motivated by this, the paper presents an approach for the data-driven analysis of cold-formed pin structures and offers a deeper understanding of how versatile processes affect the pin characteristics. Therefore, the application of an intelligent design of experiment in combination with several machine learning methods enable the setup of a best-fitting meta-model. Resulting, the determination of a mathematical model provides the opportunity to accurately estimate the pin height considering only relevant geometrical and process parameters with a prediction quality of 95 %.