# Process parameter optimization of Friction Stir Welding on 6061AA using Supervised Machine Learning Regression-based Algorithms

**Authors:** Akshansh Mishra, Eyob Messele Sefene, Assefa Asmare Tsegaw  
**arXiv:** [2109.00570](https://arxiv.org/abs/2109.00570)  
**Date:** 2021-08-12  
**Summary page:** https://www.summarizepaper.com/en/paper/2109.00570v1/

## Key points

- Lightweight materials and their joining processes are gaining attention for weight reduction in various industries
- Friction Stir Welding (FSW) allows for joining materials without filler rods, below melting point
- FSW is useful for joining similar and dissimilar metals, especially lightweight non-ferrous materials like aluminum, copper, and magnesium alloys
- Researchers focused on optimizing FSW process parameters for enhanced mechanical properties in weld joint
- Experiment conducted using 5 mm thick sheet of 6061 aluminum alloy
- Process parameters considered: tool material, rotational speed, traverse speed, axial forces
- Tensile test performed to evaluate mechanical properties of weld joint
- Maximum joint strength efficiency reached 94.2%
- Supervised Machine Learning based Regression algorithms used to analyze and predict properties accurately
- Random Forest algorithm yielded highest coefficient of determination value (0.926) when predicting mechanical properties based on process parameters
- Tool Material did not significantly affect output feature (ultimate tensile strength)
- Future research could explore implementing Quantum Machine Learning based algorithms for further prediction accuracy enhancement in FSW optimization

## Summary

The field of lightweight materials and their joining processes has gained significant attention due to the increasing demand for heft reduction in various industries. Friction Stir Welding (FSW) is a recent advancement that allows for the joining of materials without the need for additional filler rods, and it can be performed below the melting point of the parent material. This process is particularly useful for joining similar and dissimilar metals, especially lightweight non-ferrous materials like aluminum, copper, and magnesium alloys. In this study, the researchers focused on optimizing the process parameters of Friction Stir Welding to achieve enhanced mechanical properties in the weld joint. The experiment was conducted using a 5 mm thick sheet of 6061 aluminum alloy. The process parameters considered were tool material, rotational speed, traverse speed, and axial forces. To evaluate the mechanical properties of the weld joint, a tensile test was performed. The results showed that the maximum joint strength efficiency reached 94.2%. In order to further analyze and predict these properties accurately, Supervised Machine Learning based Regression algorithms were employed. Specifically, Decision Trees, Random Forest, and Gradient Boosting Algorithm were used. Among these algorithms, it was found that the Random Forest algorithm yielded the highest coefficient of determination value of 0.926. This indicates that it provided the best fit compared to other algorithms when predicting mechanical properties based on process parameters. It is worth noting that during feature importance analysis it was observed that Tool Material did not significantly affect the output feature (ultimate tensile strength). Therefore, this input feature was dropped from further analysis. Based on these findings, future research could explore implementing Quantum Machine Learning based algorithms to further enhance prediction accuracy in Friction Stir Welding optimization. Overall, this study contributes to understanding how different process parameters impact mechanical properties in Friction Stir Welding joints.