Machine Learning-Based Multi-Output Prediction of Weld Quality Using Ensemble Regression Models

This study presented a machine learning framework for multi-output prediction of welding quality characteristics using five supervised regression algorithms, namely Random Forest, XGBoost, LightGBM, CatBoost, and Artificial Neural Network. The developed framework integrated data preprocessing, feature engineering, model development, and performance evaluation into a unified workflow for intelligent welding quality prediction. The proposed methodology was evaluated using an experimental welding dataset, and identical training and testing conditions were maintained for all models to ensure a fair comparative assessment. Experimental results demonstrated that ensemble learning algorithms provide reliable prediction performance for welding quality estimation. The comparative evaluation indicated that the best-performing model achieved higher prediction accuracy with lower error values than the remaining algorithms. The incorporation of feature engineering further enhanced the predictive capability of the machine learning models by capturing nonlinear interactions among welding process parameters. Feature importance analysis also provided valuable engineering insights into the influence of individual process variables on weld quality characteristics. The proposed framework offers several practical advantages for intelligent manufacturing applications. By accurately predicting multiple weld quality characteristics, the developed model can reduce dependence on expensive experimental trials, improve process planning, and support data-driven decision-making during welding operations. The modular software architecture also enables future integration with industrial automation systems and smart manufacturing environments. Although the proposed framework demonstrated satisfactory predictive performance, several opportunities exist for further improvement. Future research may investigate larger industrial datasets, advanced deep learning architectures, Explainable Artificial Intelligence (XAI) techniques such as SHAP and LIME, and Bayesian Optimization for automatic welding parameter optimization. Integration with real-time sensor data and digital twin technologies may further improve the applicability of the framework for adaptive welding process control in Industry 4.0 and Industry 5.0 environments. Overall, the proposed machine learning framework demonstrates the effectiveness of ensemble regression techniques for intelligent welding quality prediction and provides a practical foundation for future research on AI-enabled welding process optimization.  

1.                 Akbari, M., Shamanian, M., & Salehi, M. (2022). Machine learning approaches for prediction and optimization of welding process parameters: A comprehensive review. Journal of Manufacturing Processes, 80, 351–370. 2.                 DebRoy, T., Mukherjee, T., Wei, H. L., Elmer, J. W., & Milewski, J. O. (2022). Artificial intelligence in welding and additive manufacturing. Annual Review of Materials Research, 52, 413–438. 3.                 Liu, Y., Zhang, H., Wang, X., & Chen, Z. (2023). Prediction of weld bead geometry using ensemble machine learning algorithms for gas metal arc welding. Journal of Materials Processing Technology, 320, 117960. 4.                 Patel, R., Kumar, S., & Singh, P. (2023). Intelligent prediction of welding quality using machine learning techniques: A comparative study. Engineering Applications of Artificial Intelligence, 124, 106531. 5.                 Sharma, V., Kumar, A., & Gupta, R. (2024). Explainable artificial intelligence for manufacturing quality prediction: A review. Journal of Intelligent Manufacturing, 35(2), 487–509. 6.                 Wang, J., Li, X., Zhang, Y., & Chen, L. (2024). Multi-output machine learning model for welding quality prediction using ensemble learning techniques. Journal of Manufacturing Systems, 73, 248–261. 7.                 Zhao, H., Sun, Y., & Li, Q. (2024). Intelligent welding parameter optimization using machine learning and data-driven manufacturing. Robotics and Computer-Integrated Manufacturing, 88, 102756. 8.                 Kim, S., Park, J., & Lee, D. (2025). Machine learning-enabled welding process monitoring and quality assessment: Recent advances and future directions. Journal of Manufacturing Processes, 98, 150–169. 9.                 Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5–32. https://doi.org/10.1023/A:1010933404324 10.              Chen, T., & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 785–794). ACM. https://doi.org/10.1145/2939672.2939785