Scholarlink

<p>Burr formation remains a critical issue in the fabrication of micro-grooved structures, while existing studies on magnetic abrasive finishing (MAF) have primarily focused on simple geometries. Thus, this study proposes an electropolishing-assisted MAF process for effective burr removal on micro-grooved surfaces. Experimental investigations were conducted using a Taguchi orthogonal array to analyze the effects of process parameters on deburring performance. The optimal condition was identified at a magnetic flux density of 20 mT, a voltage of 10&#xa0;V, a working gap of 1.0&#xa0;mm, and a feed rate of 20&#xa0;mm/min. Under this condition, a maximum burr reduction of approximately 95% was achieved, reducing the burr height from 11.97&#xa0;µm to 0.59&#xa0;µm. To further model the nonlinear relationship between process parameters and deburring behavior, predictive models, namely multiple regression, support vector regression (SVR), and neural network (NN), were developed. The predictive performance was evaluated using the coefficient of determination (R<sup>2</sup>), root mean squared error (RMSE), and mean absolute error (MAE). Compared with the multiple regression model, the machine learning models improved prediction accuracy by approximately 5–10% in terms of R<sup>2</sup>, from 0.879 to 0.932–0.988, with lower RMSE and MAE. Additional validation experiments under three unseen conditions confirmed that both models can reliably predict deburring performance, while the NN model demonstrated slightly improved prediction stability. These findings highlight the potential of combining the hybrid finishing process with data-driven approaches for predicting and optimizing burr removal performance in precision micro-finishing applications.</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-aided prediction of burr removal in electropolishing-assisted magnetic abrasive finishing of micro-grooved surfaces

  • 期刊论文

  • 发表日期：

    2026年6月1日

查看全文
  • Jung-Hee Lee,
  • Jae-Seob Kwak

摘要

Burr formation remains a critical issue in the fabrication of micro-grooved structures, while existing studies on magnetic abrasive finishing (MAF) have primarily focused on simple geometries. Thus, this study proposes an electropolishing-assisted MAF process for effective burr removal on micro-grooved surfaces. Experimental investigations were conducted using a Taguchi orthogonal array to analyze the effects of process parameters on deburring performance. The optimal condition was identified at a magnetic flux density of 20 mT, a voltage of 10 V, a working gap of 1.0 mm, and a feed rate of 20 mm/min. Under this condition, a maximum burr reduction of approximately 95% was achieved, reducing the burr height from 11.97 µm to 0.59 µm. To further model the nonlinear relationship between process parameters and deburring behavior, predictive models, namely multiple regression, support vector regression (SVR), and neural network (NN), were developed. The predictive performance was evaluated using the coefficient of determination (R2), root mean squared error (RMSE), and mean absolute error (MAE). Compared with the multiple regression model, the machine learning models improved prediction accuracy by approximately 5–10% in terms of R2, from 0.879 to 0.932–0.988, with lower RMSE and MAE. Additional validation experiments under three unseen conditions confirmed that both models can reliably predict deburring performance, while the NN model demonstrated slightly improved prediction stability. These findings highlight the potential of combining the hybrid finishing process with data-driven approaches for predicting and optimizing burr removal performance in precision micro-finishing applications.