Scholarlink

<p>The current study provides a thorough experimental framework to compare the feasibility of micro-ED milling-based surface texturing on biodegradable AZ91 magnesium alloy with non-biodegradable Ti6Al46V alloy for medical implant applications. The study compared the surface characteristics and dimensional precision of micro-ED milled AZ91 and Ti6Al4V alloy. Spearman correlation coefficient heatmaps show the statistical correlation between micro-ED milled input features, i.e., voltage (V), pulse-on time (T<sub>on</sub>), and tool rotational speed (TRS), and output responses, i.e., surface roughness (Ra) and taper angle (α). It is evident that all the input factors positively affect Ra and α, with V showing the strongest correlation, whereas T<sub>on</sub> correlates moderately with Ra and α and TRS correlates the least. Additionally, optimized Gaussian process regression (OGPR)-based metamodels are developed to predict the responses. It is evident from the box plots that the OGPR predicted the Ra for AZ91 alloy and Ti6Al4V alloy with a prediction error ranging from − 0.01 to + 0.05 and predicted the α for AZ91 alloy and Ti6Al4V alloy with a prediction error ranging from − 0.01 to + 0.03, corroborating an adequate generalization prowess of OGPR models. Finally, SHapley Additive exPlanations (SHAP) is applied to interpret the predictive mechanism of OGPR models.</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

An Explainable Machine Learning-Assisted Comparative Assessment of Micro-ED Milling Performance of AZ91 Magnesium Alloy and Ti-6Al-4V Alloy

  • 期刊论文

  • 发表日期：

    2026年2月9日

查看全文
  • Subhankar Saha,
  • Rashed Mustafa Mazarbhuiya,
  • Saikat Ranjan Maity,
  • Ambika Swargiary,
  • Satish Chaurasia

摘要

The current study provides a thorough experimental framework to compare the feasibility of micro-ED milling-based surface texturing on biodegradable AZ91 magnesium alloy with non-biodegradable Ti6Al46V alloy for medical implant applications. The study compared the surface characteristics and dimensional precision of micro-ED milled AZ91 and Ti6Al4V alloy. Spearman correlation coefficient heatmaps show the statistical correlation between micro-ED milled input features, i.e., voltage (V), pulse-on time (Ton), and tool rotational speed (TRS), and output responses, i.e., surface roughness (Ra) and taper angle (α). It is evident that all the input factors positively affect Ra and α, with V showing the strongest correlation, whereas Ton correlates moderately with Ra and α and TRS correlates the least. Additionally, optimized Gaussian process regression (OGPR)-based metamodels are developed to predict the responses. It is evident from the box plots that the OGPR predicted the Ra for AZ91 alloy and Ti6Al4V alloy with a prediction error ranging from − 0.01 to + 0.05 and predicted the α for AZ91 alloy and Ti6Al4V alloy with a prediction error ranging from − 0.01 to + 0.03, corroborating an adequate generalization prowess of OGPR models. Finally, SHapley Additive exPlanations (SHAP) is applied to interpret the predictive mechanism of OGPR models.