AProKI – Analysis of process variables at friction stir welding using artificial intelligence
The aim of the project AProKI is the application of current findings in the field of artificial intelligence (AI) for the evaluation of process variables at friction stir welding. This should enable a non-destructive, purely data-based prediction of the weld quality. The developed algorithms will be provided to the small and medium-sized enterprises participating in the project in order to promote the transfer of AI applications to these companies.
Motivation
Friction stir welding (FSW) is particularly suitable for light metals such as aluminum alloys, which is why the technology is already used in various industrial sectors. Especially for e-mobility, FSW is considered an enabler for lightweight construction. The demand for high-quality products and cost-effective production increases the need for inline process monitoring. Since FSW is a highly automated process, the acquisition of process variables (e.g. forces, temperatures) is well realizable. Up to now, the monitoring of process variables has been conducted by simple methods, such as the definition of limit values that must not be exceeded or undercut. There is great potential for improvement through the use of artificial intelligence (AI). AI provides the tools to efficiently handle data from industrial processes and to interpret them for the benefit of companies and their customers. The transfer of new AI technologies to a broad spectrum of the German economy, which is characterized by small and medium-sized enterprises, has great potential for value creation in Germany and currently represents a challenge.
Research objective
Aims of the project are the transfer of current findings of AI research to the evaluation of process variables for the prediction of the weld quality and the transfer of the AI application to small and medium-sized enterprises.
Approach
The central stages of the project are the execution of experiments to generate an appropriate database, the inspection of the welds, the development and validation of the AI algorithms, and the development of a graphical user interface for an easier application of the algorithms.
Results and Use
A meaningful and valid evaluation of the process variables using AI could supplement or substitute other tests that follow welding. As expensive measuring equipment for examining the weld quality is often missing, especially in the case of small and medium-sized enterprises, these would benefit particularly from the results.