Scholarlink

<p>The primary focus of this work is to predict the corrosion behaviour of surface-engineered (micro-blasting and/or thermal oxidation) Ti6Al4V alloy using the machine learning approach. The potentiodynamic polarization (PDP) and electrochemical impedance spectroscopy (EIS) experiments were carried out for untreated and surface-engineered samples. Supervised machine learning (ML) algorithms such as multiple polynomial regression (MPR), support vector regression (SVR), decision tree (DT), and extreme gradient boosting (XGB) were used, and the experimental results were given as input data sets to predict the corrosion behaviour. The heat treatment temperature (600&#xa0;°C) and duration (24 and 48&#xa0;h), surface roughness, crystalline size and residual stresses were given as common independent variables to the ML models, and the feature importance analysis was carried out. A set of predetermined metrics was used to assess the performance of the models: R-squared (R2), mean absolute error (MAE), and root mean square error (RMSE). The XGB algorithm outperformed in predicting corrosion behaviour for PDP, and the corresponding R2 value is 0.88. Further, SVR performed best in predicting passive layer characteristics for EIS. This model showed the R2 value of 0.99 for both Nyquist and Bode Impedance plots, while 0.86 for the Bode phase angle plot. Finally, the feature importance analysis results showed that the surface roughness was the most influential process parameter on the corrosion behaviour of Ti6Al4V alloy.</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 Assisted Corrosion Behaviour Prediction of Dual-Engineered Ti6Al4V Alloy

  • 期刊论文

  • 发表日期：

    2025年5月29日

查看全文
  • P. Jeyalakshmi,
  • M. Naveen,
  • P. Ramkumar

摘要

The primary focus of this work is to predict the corrosion behaviour of surface-engineered (micro-blasting and/or thermal oxidation) Ti6Al4V alloy using the machine learning approach. The potentiodynamic polarization (PDP) and electrochemical impedance spectroscopy (EIS) experiments were carried out for untreated and surface-engineered samples. Supervised machine learning (ML) algorithms such as multiple polynomial regression (MPR), support vector regression (SVR), decision tree (DT), and extreme gradient boosting (XGB) were used, and the experimental results were given as input data sets to predict the corrosion behaviour. The heat treatment temperature (600 °C) and duration (24 and 48 h), surface roughness, crystalline size and residual stresses were given as common independent variables to the ML models, and the feature importance analysis was carried out. A set of predetermined metrics was used to assess the performance of the models: R-squared (R2), mean absolute error (MAE), and root mean square error (RMSE). The XGB algorithm outperformed in predicting corrosion behaviour for PDP, and the corresponding R2 value is 0.88. Further, SVR performed best in predicting passive layer characteristics for EIS. This model showed the R2 value of 0.99 for both Nyquist and Bode Impedance plots, while 0.86 for the Bode phase angle plot. Finally, the feature importance analysis results showed that the surface roughness was the most influential process parameter on the corrosion behaviour of Ti6Al4V alloy.