The irruption of Artificial Intelligence (AI) in various areas of industry has led to an unprecedented transformation, and the field of welding is no exception. Contrary to conventional welding methods that involve melting the base materials, friction stir welding uses a rotating tool to generate frictional heat and plastically deform the material, resulting in a solid-state joint. This feature significantly reduces the risk of defects such as porosity and distortion. While this method leads to high-quality welds in aluminium alloys, in the case of steel, the process has limitations. The STWIN project, coordinated by CIDAUT, explores the potential of AI to guarantee the quality of FSWed joints of steel by ensuring that the process parameters, regardless of the type of material, are always appropriate. For that purpose, the project relies on non-destructive inspection methods such as acoustic emission monitoring, magnetic flux and thermography.
Machine intelligence
Reducing defects is essential to ensure productivity and sustainability in any industrial field: whenever a process is optimised, waste and energy consumption are reduced. In this regard, artificial intelligence, with its capability to process large amounts of data, learn from it and adapt its responses, has potential to address the above-described problem. By combining measurements from online inspection methods with process parameters, AI can be used to create inferences about events that cause defects. However, the real power of AI lies in its capability to predict which process parameters lead high-quality welds.
Tangible results
The STWIN project, and particularly CIDAUT, works on developing AI models for automatic prediction of process variables that optimise weld quality. The models transform raw data from online inspection methods into actionable insights, such as defect prediction and optimisation of process parameters. In this way, AI opens new horizons for friction stir welding in the industry in general.
The research leading to these results has received funding from Research Fund for Coal and Steel Programmes of the European Commission under Grant Agreement 101112504.