Integrated multiobjective optimization of RFSSW parameters for AA2024-T3 using ANOVA machine learning and NSGA II.

Area of Science:

• Materials Science and Engineering
• Manufacturing Technology
• Computational Science

Background:

• Intelligent manufacturing demands precise control over complex processes like Refill Friction Stir Spot Welding (RFSSW) to ensure product quality.
• Nonlinear interactions between process parameters significantly impact outcomes in solid-state joining.
• Optimizing RFSSW for materials like AA2024-T3 aluminum alloy is crucial for advanced applications.

Purpose of the Study:

• To develop a data-driven methodology for optimizing RFSSW parameters.
• To identify key process variables influencing joint load capacity.
• To achieve multi-objective optimization for maximum weld strength in AA2024-T3 aluminum alloy.

Main Methods:

• A 3³ full-factorial design of experiments was employed to collect data on rotational speed, plunge depth, and welding time.
• Six machine learning techniques were evaluated for predicting joint load capacity, with XGBoost showing superior performance.
• The NSGA-II evolutionary algorithm was utilized for multi-objective optimization, generating a Pareto frontier of optimal parameter sets.

Main Results:

• Plunge depth was identified as the most influential parameter affecting weld strength, confirmed by ANOVA and SHAP analysis.
• XGBoost model achieved high prediction accuracy (R² up to 0.89) for joint load capacity.
• A maximin strategy on the Pareto frontier yielded a robust compromise solution for optimal RFSSW parameters.

Conclusions:

• The integrated approach combining statistical analysis, machine learning, and evolutionary optimization effectively refines solid-state joining processes.
• This methodology provides a scalable template for intelligent manufacturing applications requiring multi-objective optimization.
• Accurate prediction and optimization of RFSSW parameters are vital for enhancing product quality and process efficiency.