Scholarlink

<p>Machine learning techniques are increasingly applied in the field of pyrometallurgy. The integration of advanced machine learning technologies with electroslag remelting (ESR) remains inadequate, and there is a lack of indepth interpretation regarding the decision-making process of the machine learning model. The models for predicting titanium content of Incoloy 825 alloy refined by ESR based on machine learning algorithms and SHAP interpretation framework were established in the current study. An improved ranking-based decision method was employed to compare the prediction performance of these five machine learning models. The prediction accuracy of the XGBoost, regression tree, ridge regression and artificial neural network models surpasses the metallurgical mechanism models for forecasting the contents of Ti and Si in remelted ingots. The XGBoost model exhibits the highest accuracy with determination coefficient, mean absolute error and root mean square error values of 0.996, 0.006, and 0.010, respectively. The effects of slag composition, consumable electrode composition and ESR process parameters on the titanium content of remelted ingots were analyzed by the SHAP interpretation framework and metallurgical mechanism. The SHAP framework determined the key factors that influence the titanium content of remelted ingots.</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

Prediction of Titanium Content of Nickel–Iron-Base Alloy in Electroslag Remelting Based on Machine Learning and SHapley Additive ExPlanations

  • 期刊论文

  • 发表日期：

    2025年3月24日

查看全文
  • Huai Zhang,
  • Chengbin Shi,
  • Hui Wang,
  • Jiawei Liu,
  • Yifan Meng

摘要

Machine learning techniques are increasingly applied in the field of pyrometallurgy. The integration of advanced machine learning technologies with electroslag remelting (ESR) remains inadequate, and there is a lack of indepth interpretation regarding the decision-making process of the machine learning model. The models for predicting titanium content of Incoloy 825 alloy refined by ESR based on machine learning algorithms and SHAP interpretation framework were established in the current study. An improved ranking-based decision method was employed to compare the prediction performance of these five machine learning models. The prediction accuracy of the XGBoost, regression tree, ridge regression and artificial neural network models surpasses the metallurgical mechanism models for forecasting the contents of Ti and Si in remelted ingots. The XGBoost model exhibits the highest accuracy with determination coefficient, mean absolute error and root mean square error values of 0.996, 0.006, and 0.010, respectively. The effects of slag composition, consumable electrode composition and ESR process parameters on the titanium content of remelted ingots were analyzed by the SHAP interpretation framework and metallurgical mechanism. The SHAP framework determined the key factors that influence the titanium content of remelted ingots.