Data-Driven Optimization and Modelling of the Gap Bridgeability Performance of Multi-Pin Friction Stir Welded EN AW...

Area of Science:

• Materials Science and Engineering
• Manufacturing Processes
• Computational Materials Science

Background:

• Friction Stir Welding (FSW) of high-strength aluminum alloys presents challenges with weld line gaps, impacting joint integrity.
• Optimizing FSW process parameters is crucial for achieving defect-free welds and maintaining mechanical properties.

Purpose of the Study:

• To develop a predictive, data-driven framework for assessing and optimizing the gap bridgeability of FSW joints in EN AW 7020-T651 aluminum alloy.
• To identify key process parameters influencing weld quality under varying gap conditions (0-4 mm).

Main Methods:

• A structured experimental matrix systematically varied rotational speed, welding speed, axial force, and tool shoulder diameter.
• A multi-pin tool was utilized to ensure stable material flow and consistent weld quality.
• Weld quality was assessed using defect-free criteria, minimum ultimate tensile strength (230 MPa), and a novel weak area percentage (WAP) metric derived from micro-hardness mapping.

Main Results:

• The Random Forest machine learning model achieved 92.5% accuracy and an F1-score of 0.90 in classifying weld acceptability.
• Feature importance analysis revealed that "welding speed × gap size" and "rotational speed × gap size" interactions were the most significant predictors of weld quality.
• The developed framework successfully predicted and optimized FSW performance across a range of gap conditions.

Conclusions:

• A data-driven approach using machine learning effectively predicts weld quality in FSW of aluminum alloys with weld line gaps.
• The "welding speed × gap size" and "rotational speed × gap size" interactions are critical factors for controlling weld integrity.
• The novel WAP metric offers a more robust assessment of mechanical integrity in the heat-affected zone compared to traditional methods.