Scholarlink

<p>Based on relevant domestic and international literature and laboratory sample data, this study developed machine learning (ML) models for predicting the high-temperature deformation flow stress of all titanium alloys. The employed ML models include BP-ANN, SVR, and RF. The correlation coefficient (<i>R</i>), root mean square error (RMSE), and average absolute relative error (AARE) were used to evaluate the model performance. The test results indicated that the correlation coefficients of all three models exceeded 0.98. The selected composition parameters were verified as important parameters through sensitivity analysis (SA). In the flow stress prediction for different types of alloys, this study took experimental measurements as the benchmark and compared the prediction performance of the ML models, constitutive models, and the BP-ANN model based on a single-alloy system. The results showed that among near-<i>β</i>, <i>α</i> + <i>β</i>, and <i>β</i>-type alloys, the prediction performance of the three ML models was significantly higher than that of the Arrhenius model and close to that of the BP-ANN model for single-alloy systems. In the analysis of the influence of Al, Fe, V, and H contents on flow stress, the BP-ANN and SVR models exhibit high predictive performance. When investigating the influence of deformation process parameters (deformation temperature and strain rate) on flow stress, the RF model achieved a prediction accuracy of <i>R</i> = 0.9854 and AARE = 23.47%.</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

Machine Learning-Based Model for Predicting Flow Stresses in High-Temperature Deformation of Titanium Alloys

  • 期刊论文

  • 发表日期：

    2026年5月3日

查看全文
  • Yong Niu,
  • Jiecheng Bai,
  • Yanchun Zhu,
  • Yaoqi Wang

摘要

Based on relevant domestic and international literature and laboratory sample data, this study developed machine learning (ML) models for predicting the high-temperature deformation flow stress of all titanium alloys. The employed ML models include BP-ANN, SVR, and RF. The correlation coefficient (R), root mean square error (RMSE), and average absolute relative error (AARE) were used to evaluate the model performance. The test results indicated that the correlation coefficients of all three models exceeded 0.98. The selected composition parameters were verified as important parameters through sensitivity analysis (SA). In the flow stress prediction for different types of alloys, this study took experimental measurements as the benchmark and compared the prediction performance of the ML models, constitutive models, and the BP-ANN model based on a single-alloy system. The results showed that among near-β, α + β, and β-type alloys, the prediction performance of the three ML models was significantly higher than that of the Arrhenius model and close to that of the BP-ANN model for single-alloy systems. In the analysis of the influence of Al, Fe, V, and H contents on flow stress, the BP-ANN and SVR models exhibit high predictive performance. When investigating the influence of deformation process parameters (deformation temperature and strain rate) on flow stress, the RF model achieved a prediction accuracy of R = 0.9854 and AARE = 23.47%.