Scholarlink

<p>The study develops and validates a framework integrating machine learning into the Six Sigma DMAIC methodology to address quality challenges in complex manufacturing environments. While traditional DMAIC methodology offers a well-structured framework for process improvement, its applicability is limited when dealing with high-dimensional and complex datasets. The proposed framework was implemented in a foundry organisation producing large aluminium battery housings for car manufacturers using high-pressure die casting technology. Production data from 6,619 castings with 41 process parameters were analysed using an Ishikawa diagram and subsequently ML techniques (Random Forest, XGBoost, and LightGBM). The best performing model – Random Forest achieved moderate accuracy (77%). Its interpretation through feature importance, partial dependence plots and SHAP analysis revealed that die temperature, lubricant dosage, and casting sequence number are critical parameters contributing to casting defects. Implementation of ML-guided improvements, including core pin thermoregulation and spraying process optimisation, reduced nonconforming parts from 21% to 10% and lowered the costs of poor quality. The demonstrated case shows that even the model with moderate accuracy (given the complexity of the process) can support decision-making when it provides interpretable insights that enhance process understanding. The proposed approach provides a practical pathway for integrating Quality 4.0 principles into traditional Six Sigma initiatives.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integration of Six Sigma DMAIC methodology with machine learning in quality improvement: an application in the casting manufacturing process

  • 期刊论文

  • 发表日期：

    2026年6月16日

查看全文
  • Andrea Sütőová,
  • Radoslav Palacka,
  • Jiju Antony,
  • Gijo Edappattu Vijayan,
  • Elizabeth Cudney,
  • Cristina Ciliberto

摘要

The study develops and validates a framework integrating machine learning into the Six Sigma DMAIC methodology to address quality challenges in complex manufacturing environments. While traditional DMAIC methodology offers a well-structured framework for process improvement, its applicability is limited when dealing with high-dimensional and complex datasets. The proposed framework was implemented in a foundry organisation producing large aluminium battery housings for car manufacturers using high-pressure die casting technology. Production data from 6,619 castings with 41 process parameters were analysed using an Ishikawa diagram and subsequently ML techniques (Random Forest, XGBoost, and LightGBM). The best performing model – Random Forest achieved moderate accuracy (77%). Its interpretation through feature importance, partial dependence plots and SHAP analysis revealed that die temperature, lubricant dosage, and casting sequence number are critical parameters contributing to casting defects. Implementation of ML-guided improvements, including core pin thermoregulation and spraying process optimisation, reduced nonconforming parts from 21% to 10% and lowered the costs of poor quality. The demonstrated case shows that even the model with moderate accuracy (given the complexity of the process) can support decision-making when it provides interpretable insights that enhance process understanding. The proposed approach provides a practical pathway for integrating Quality 4.0 principles into traditional Six Sigma initiatives.